Surrogate vs Clinical Endpoints in Drug Development: A Strategic Guide for Researchers and Regulators

Abstract

This article provides a comprehensive analysis of surrogate and clinical endpoints for researchers and drug development professionals. It explores the foundational definitions and regulatory context, details methodological frameworks for validation and application, addresses common challenges and optimization strategies in trial design, and offers a critical comparative evaluation of endpoint utility. With insights from recent FDA guidances, AACR workshops, and contemporary studies, this guide aims to equip stakeholders with the knowledge to navigate endpoint selection, accelerate drug development, and ensure new therapies deliver meaningful patient benefits.

Defining the Landscape: What Are Surrogate and Clinical Endpoints and Why Do They Matter?

In the landscape of clinical trials and drug development, the selection of appropriate endpoints stands as one of the most pivotal decisions, directly influencing a trial's duration, cost, interpretability, and ultimate regulatory success. This choice fundamentally centers on two distinct categories: surrogate endpoints and patient-relevant clinical outcomes. A surrogate endpoint is "an outcome measure used as a substitute for a clinically meaningful endpoint…changes induced by a therapy on a surrogate endpoint are expected to reflect changes in a clinically meaningful endpoint" [1]. In contrast, a clinical outcome is "a measurable change in symptoms, overall health, ability to function, quality of life, or survival outcomes that result from giving care to patients" [2]. These clinical outcomes directly measure how a patient feels, functions, or survives [3] [1]. The distinction is not merely semantic; it carries profound implications for accurately assessing the true therapeutic value of medical interventions and ensuring that healthcare resources are allocated to treatments that deliver meaningful patient benefits.

The use of surrogate endpoints has grown substantially over the past two decades, driven by the imperative to accelerate drug development. Between 2010 and 2012, the U.S. Food and Drug Administration (FDA) approved 45 percent of new drugs based on a surrogate endpoint, and recent analyses suggest this figure now exceeds 50% for both the FDA and the European Medicines Agency (EMA) [4] [5]. While this trend enables faster market access for promising therapies, it also introduces significant uncertainty about long-term clinical benefit, underscoring the critical need for researchers, regulators, and payers to thoroughly understand the strengths and limitations of each endpoint type.

Defining the Concepts: A Comparative Framework

Patient-Relevant Clinical Outcomes

Patient-relevant clinical outcomes, sometimes termed "clinical efficacy measures" or "true endpoints," represent the gold standard for evaluating therapeutic interventions in definitive clinical trials. These endpoints capture the direct, tangible effects of a treatment from the patient's perspective. The FDA defines them as measures that directly assess whether people in a trial "feel or function better, or live longer" [3]. They are the ultimate indicators of treatment success because they measure aspects of health that patients inherently value and experience directly.

Table 1: Characteristics of Patient-Relevant Clinical Outcomes

Attribute	Description	Examples
Definition	Direct measures of how a patient feels, functions, or survives [3] [1].	Overall survival, reduction in pain, improved ability to perform daily activities.
Primary Focus	Patient experience and tangible health benefit.	"How patients feel, function, or survive" [6].
Key Advantage	High interpretability and direct evidence of clinical benefit.	Clear, meaningful measure of therapeutic value.
Key Limitation	Often require longer, larger, and more costly trials to assess.	Measuring survival in chronic diseases can take many years.
Regulatory Acceptance	Gold standard for definitive evidence of efficacy.	Required for full traditional approval when feasible.

Scoping reviews of the literature identify a wide range of outcomes considered relevant to patients, with the most dominant categories being symptoms, adverse events/complications, survival/mortality, and pain [7]. These outcomes can be assessed through various tools, including Clinical Outcome Assessments (COAs), which are reports generated by clinicians, patients, or non-clinician observers about a patient's health status [5] [3].

Surrogate Markers and Endpoints

Surrogate markers (or surrogate endpoints) operate as substitutes for patient-relevant clinical outcomes. They are typically biological markers—such as physiological measurements, blood tests, radiographic findings, or other chemical analyses—that are objectively measured and are hypothesized to predict clinical benefit [3] [1]. A biomarker must undergo rigorous validation to be accepted as a surrogate endpoint. The underlying premise is that the surrogate lies on the causal pathway between the intervention and the ultimate clinical outcome of interest.

Table 2: Characteristics of Surrogate Endpoints

Attribute	Description	Examples
Definition	A substitute endpoint that predicts, but does not itself constitute, a clinical benefit [3] [1].	Blood pressure, LDL cholesterol, tumor shrinkage on a scan.
Primary Focus	Biological or physiological process correlated with a clinical outcome.	"A defined characteristic that is measured as an indicator of...pathologic processes, or responses to an intervention" [3].
Key Advantage	Can significantly reduce trial size, duration, and cost.	Enables faster regulatory approval and patient access.
Key Limitation	May not reliably predict true clinical benefit; risk of misleading conclusions.	A drug can improve the surrogate but fail to improve, or even harm, the clinical outcome.
Regulatory Acceptance	Accepted for traditional and accelerated approval, but require validation.	FDA maintains a table of over 200 surrogate endpoints used for approval [6].

A classic and successful example of a validated surrogate endpoint is the reduction of systolic blood pressure, which predicts a reduced risk of the clinical outcome of stroke [5] [3]. Similarly, reduction in low-density lipoprotein cholesterol (LDL-C) is an accepted surrogate for reduced cardiovascular morbidity and mortality in statin trials [8]. However, it is crucial to note that the predictive value of a surrogate is often context-dependent; for instance, LDL-C is a strong surrogate for statins but less predictive for other classes of lipid-lowering therapies like fibrates [4].

A Hierarchical View of Clinical Trial Endpoints

To better conceptualize the relative strength and reliability of different outcome measures, a hierarchical framework is useful. This model, adapted from Fleming (2012), classifies endpoints into four distinct levels based on the strength of evidence linking them to patient benefit [1].

Diagram 1: Endpoint Hierarchy

• Level 1: Clinical Efficacy Measures represent the highest standard, constituting direct measures of patient benefit, such as overall survival, reduction in pain, or prevention of symptomatic bone fractures [1].
• Level 2: Validated Surrogate Endpoints are biomarkers supported by strong evidence from multiple trials demonstrating that a treatment's effect on the surrogate reliably predicts its effect on the clinical outcome. Examples include HbA1c for microvascular complications in diabetes and blood pressure for cardiovascular events [1].
• Level 3: Reasonably Likely Surrogate Endpoints are biomarkers with a strong mechanistic or epidemiologic rationale but insufficient clinical data for full validation. These can support accelerated approval, contingent on post-marketing confirmation studies [3] [6].
• Level 4: Biomarkers (Correlates) are measures of biological activity that have not been established to predict clinical benefit. These are often used for secondary endpoints or patient selection but are inadequate as primary bases for approval. Examples include CD4 counts in HIV or PSA levels in prostate cancer prevention [1].

This hierarchy clarifies that not all biomarkers are surrogate endpoints, and not all surrogate endpoints are equally reliable. The ultimate goal of surrogate evaluation is to enable the use of Level 2 validated surrogates to make accurate inferences about a treatment's effect on Level 1 clinical outcomes in future trials.

Methodological Framework for Validating Surrogate Endpoints

The validation of a surrogate endpoint is a rigorous, multi-stage process that moves beyond biological plausibility to quantitative statistical assessment. A widely accepted framework, such as the "Ciani framework," proposes three levels of evidence for surrogate validation [4].

The Three Levels of Surrogate Validation

Table 3: Levels of Evidence for Surrogate Endpoint Validation

Level	Evidence Type	Description	Source of Evidence	Key Statistical Metrics
Level 3	Biological Plausibility	The biomarker lies on the known causal pathway of the disease and the clinical outcome.	Clinical data and understanding of disease biology.	Not applicable.
Level 2	Individual-Level Association	The surrogate endpoint and the target outcome are correlated at the level of the individual patient.	Epidemiological studies and/or clinical trials.	Correlation between the surrogate and the final outcome.
Level 1	Trial-Level Association	The treatment's effect on the surrogate endpoint predicts its effect on the target outcome across multiple trials.	Meta-analysis of multiple RCTs or a single large RCT.	Coefficient of determination (R²trial), Spearman’s correlation, Surrogate Threshold Effect (STE).

Level 1 evidence, also known as "trial-level surrogacy," is considered the most critical for health technology assessment (HTA) and reimbursement decisions [4]. It requires demonstrating that across a set of clinical trials, the magnitude of a treatment's effect on the surrogate consistently corresponds to the magnitude of its effect on the patient-relevant final outcome.

Statistical Approaches and Workflow

The statistical evaluation of surrogate endpoints employs several frameworks, including the meta-analytic approach, the proportion of treatment effect (PTE) explained, and principal stratification [9]. The most robust evaluations often rely on a meta-analysis of multiple randomized controlled trials (RCTs), preferably using individual participant data (IPD), to assess the association between treatment effects on the surrogate and the final outcome [4]. The strength of this association is quantified using metrics like the coefficient of determination (R²_trial), where a value close to 1 indicates that the treatment effect on the surrogate explains nearly all the variability in the treatment effect on the final outcome.

Diagram 2: Validation Workflow

An increasingly reported metric in this process is the Surrogate Threshold Effect (STE), defined as the minimum treatment effect on the surrogate needed to predict a statistically significant effect on the final clinical outcome [4]. This metric is particularly valuable for health technology assessment bodies, as it helps quantify the clinical implications of a treatment's effect on a surrogate in a future trial.

Case Study: GFR Slope as a Validated Surrogate in Chronic Kidney Disease

The validation of glomerular filtration rate (GFR) slope in chronic kidney disease (CKD) provides a contemporary example of a successfully validated "first in class" surrogate endpoint. CKD is a slowly progressive disease where the definitive target outcome—kidney failure requiring dialysis or transplantation—can take many years to observe [4].

Experimental Protocol and Evidence Base

The validation of GFR slope (the rate of decline in kidney function over time) followed a rigorous, multi-step process aligning with the Ciani framework:

• Biological Plausibility (Level 3): GFR is a direct measure of kidney function, and its progressive decline is a fundamental characteristic of CKD progression, logically leading to kidney failure.
• Individual-Level Association (Level 2): Epidemiological studies and clinical trial data established a strong correlation between an individual's GFR slope and their subsequent risk of reaching kidney failure.
• Trial-Level Association (Level 1): A meta-analysis of randomized controlled trials in CKD demonstrated a very strong association between the treatment effect on GFR slope and the treatment effect on the risk of kidney failure. The strength of this association was remarkably high, with a reported R²_trial of 97% [4].

This robust evidence base led the FDA and EMA to recently approve GFR slope as an acceptable primary endpoint for clinical trials of CKD therapies, significantly accelerating the development of new treatments for this condition [4].

The Scientist's Toolkit: Key Reagents and Methods

Table 4: Essential Research Toolkit for Surrogate Endpoint Evaluation

Tool / Method	Function in Validation	Application Example
Individual Participant Data (IPD) Meta-analysis	Enables standardized analysis of associations at both patient and trial levels; considered the optimal approach.	Combining IPD from multiple CKD trials to validate GFR slope.
Statistical Software (R/Python)	Implementation of surrogate evaluation frameworks (PTE, meta-analytic).	Fitting multivariate models to estimate the proportion of treatment effect explained.
Coefficient of Determination (R²trial)	Quantifies the proportion of variance in the clinical outcome effect explained by the surrogate effect.	Reporting an R²trial of 0.97 for GFR slope, indicating excellent predictive power.
Surrogate Threshold Effect (STE)	Defines the minimum treatment effect on the surrogate needed to predict a significant clinical benefit.	Used by HTA bodies to interpret the clinical meaning of a trial's results.

Regulatory and HTA Perspectives

Regulatory agencies and HTA bodies approach surrogate endpoints with different but overlapping priorities. The FDA, through its Accelerated Approval program and traditional approval pathways, may accept surrogate endpoints that are "reasonably likely to predict clinical benefit" or are fully "validated" [3] [6]. The FDA's public "Table of Surrogate Endpoints" lists over 200 such markers that have been or could be used for drug approval [6].

However, HTA agencies and payers, who make decisions about reimbursement based on longer-term comparative effectiveness and cost-effectiveness, have traditionally been more cautious [4]. They require a higher level of evidence, particularly strong Level 1 trial-level surrogacy, to reduce decision uncertainty. Overreliance on inadequately validated surrogates can lead to systematic overestimation of clinical benefit and cost-effectiveness, resulting in market access for treatments that may later be found to provide limited patient benefit [4]. This skepticism is well-founded in historical cases, such as anti-arrhythmia drugs, which successfully reduced arrhythmias (the surrogate) but were found to increase cardiac deaths (the clinical outcome), resulting in tens of thousands of preventable deaths [9].

The distinction between surrogate markers and patient-relevant clinical outcomes is fundamental to the integrity of clinical research and drug development. While surrogate endpoints offer a powerful tool to accelerate the delivery of new therapies, their value is entirely contingent on rigorous, multi-level validation demonstrating a reliable predictive relationship with meaningful clinical benefits. The hierarchical framework for endpoints and the structured validation process provide researchers with a clear roadmap for evaluating potential surrogates.

As the use of surrogate endpoints continues to grow, the imperative for transparency and ongoing evaluation intensifies. Stakeholders, including regulators, HTA bodies, clinicians, and patients, must critically assess the strength of evidence supporting each surrogate to ensure that the pursuit of efficiency in drug development does not come at the cost of certainty about genuine patient benefit. Future efforts should focus on strengthening the science of surrogate endpoint validation through collaborative evidence generation and robust post-marketing studies, ensuring that both innovation and patient interests are served.

In the landscape of drug development, endpoints serve as the definitive signposts that determine a therapy's regulatory journey and ultimate destination. These carefully selected measures form the foundation upon which drug sponsors and regulatory agencies assess whether a new medical product delivers a positive balance of benefit and risk [3]. Between 2010 and 2012, the U.S. Food and Drug Administration (FDA) approved 45 percent of new drugs based on a surrogate endpoint, highlighting the pivotal role these markers play in modern therapeutic development [3]. The choice between clinical outcomes that directly measure how patients feel, function, or survive, and surrogate endpoints that substitute for these direct measures, represents one of the most consequential decisions in clinical trial design—a decision that fundamentally shapes development timelines, resource allocation, and regulatory strategy.

This endpoint selection imperative exists within an evolving regulatory framework that increasingly recognizes the need for both scientific rigor and efficiency in bringing new treatments to patients. The 21st Century Cures Act codified this importance by mandating that the FDA publish and regularly update a list of surrogate endpoints that have formed the basis of drug approval or licensure [10]. As of 2025, this table contains over 200 surrogate markers that have been or would be accepted by the agency to support drug approval, providing a valuable roadmap for developers while underscoring the regulatory significance of endpoint selection [6]. Understanding the distinctions, applications, and evidence requirements for different endpoint categories is thus not merely an academic exercise but a practical necessity for navigating the complex drug approval pathway.

Endpoint Fundamentals: Definitions and Regulatory Classifications

Clinical Outcome Assessments (COAs) and Clinical Endpoints

Clinical endpoints directly measure how a patient feels, functions, or survives, providing unambiguous evidence of treatment benefit [3]. These measures include:

• Overall survival in oncology trials
• Reduction in stroke incidence in cardiovascular studies
• Improvement in symptoms such as pain or shortness of breath
• Preservation or improvement of physical function in neurodegenerative diseases

The FDA has created a Clinical Outcomes Assessment (COA) Compendium that summarizes how certain COAs have been used in clinical trials to measure the patient's experience and support labeling claims [5]. These assessments are measured through reports generated by clinicians, patients, non-clinician observers, or performance-based assessments, capturing the direct impact of a treatment on a patient's quality of life and functional status.

Surrogate Endpoints and Biomarkers

A surrogate endpoint is "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure, that is not itself a direct measurement of clinical benefit, but is known or reasonably likely to predict clinical benefit" [10]. Surrogate endpoints exist within a broader category of biomarkers, which the NIH Definitions Working Group defines as "a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention" [11].

Table 1: Categories of Surrogate Endpoints Based on Validation Status

Category	Definition	Level of Evidence	Typical Regulatory Use
Candidate Surrogate Endpoint	Under evaluation for ability to predict clinical benefit	Preliminary mechanistic or epidemiologic rationale	Early development, proof-of-concept studies
Reasonably Likely Surrogate Endpoint	Supported by strong mechanistic/epidemiologic rationale but limited clinical data	Strong biological plausibility but insufficient clinical validation	Accelerated Approval pathway
Validated Surrogate Endpoint	Supported by clear mechanistic rationale and clinical data demonstrating prediction of specific clinical benefit	Extensive evidence from epidemiological studies and clinical trials	Traditional approval

The FDA characterizes surrogate endpoints by their level of clinical validation, with "validated surrogate endpoints" representing those that have undergone extensive testing and are accepted as evidence of benefit for traditional approval [3]. These validated surrogates include well-established markers such as blood pressure for cardiovascular outcomes, HbA1c for diabetes complications, and tumor response rates in certain oncology settings.

The Regulatory Framework: Endpoint Selection and Approval Pathways

Traditional Approval Versus Accelerated Approval

The choice of endpoint directly determines which regulatory pathway a drug may pursue, with significant implications for development strategy and evidence requirements. The following diagram illustrates how endpoint selection dictates the available regulatory routes:

For traditional approval, drugs must demonstrate a direct effect on clinical outcomes or validated surrogate endpoints that are supported by extensive evidence predicting clinical benefit [3]. This pathway requires "substantial evidence of effectiveness" from adequate and well-controlled investigations, typically involving two or more pivotal trials [6].

The accelerated approval pathway provides patients with serious diseases more rapid access to promising therapies based on "reasonably likely" surrogate endpoints that are supported by strong mechanistic and/or epidemiologic rationale but lack sufficient clinical data to be considered validated [3]. This regulatory mechanism acknowledges that for serious conditions with unmet medical needs, the public health benefit of earlier availability may outweigh the uncertainty associated with less-validated endpoints. However, this approach requires sponsors to conduct post-marketing studies to verify the anticipated clinical benefit, and failure to demonstrate this benefit can result in withdrawal of the approval [6].

The FDA Surrogate Endpoint Table

Section 507 of the Federal Food, Drug, and Cosmetic Act, as amended by the 21st Century Cures Act, mandates that the FDA publish a list of "surrogate endpoints which were the basis of approval or licensure (as applicable) of a drug or biological product" [10]. This table, updated every six months, includes:

• Surrogate endpoints that sponsors have used as primary efficacy endpoints for approval of new drug applications (NDAs) or biologics license applications (BLAs)
• Surrogate endpoints that the Agency anticipates could be appropriate for use as a primary efficacy endpoint, though not yet used to support an approved NDA or BLA
• Separate sections for adult and pediatric endpoints

Table 2: Selected Examples from FDA's Surrogate Endpoint Table for Adult Non-Cancer Conditions

Disease or Use	Patient Population	Surrogate Endpoint	Type of Approval Appropriate For
Alzheimer's disease	Patients with mild cognitive impairment or mild dementia stage	Reduction in amyloid beta plaques	Accelerated Approval
Chronic kidney disease	Patients with chronic kidney disease secondary to multiple etiologies	Estimated glomerular filtration rate or serum creatinine	Traditional Approval
Duchenne muscular dystrophy (DMD)	Patients with DMD who have a confirmed mutation amenable to exon skipping	Skeletal muscle dystrophin	Accelerated Approval
Asthma/COPD	Patients with asthma or COPD	Forced expiratory volume in 1 second (FEV1)	Traditional Approval
Gout	Patients with gout	Serum uric acid	Traditional Approval

The table serves as a reference guide to inform discussions between sponsors and FDA review divisions, potentially speeding up drug and biologic development by providing greater clarity on potential endpoints [3]. However, the acceptability of any surrogate endpoint for a specific development program is determined on a case-by-case basis, considering factors such as the disease, studied patient population, therapeutic mechanism of action, and availability of current treatments [10].

Comparative Analysis: Clinical Endpoints vs. Surrogate Endpoints

Relative Advantages and Limitations

The selection between clinical and surrogate endpoints involves balancing multiple factors, including development timeline, cost, feasibility, and certainty about clinical benefit. The table below summarizes the key comparative characteristics:

Table 3: Comparative Analysis of Clinical Endpoints versus Surrogate Endpoints

Characteristic	Clinical Endpoints	Surrogate Endpoints
Directness of Benefit Measurement	Directly measure how patients feel, function, or survive [3]	Indirect measure; predicts rather than measures clinical benefit [3]
Trial Duration	Often lengthy, especially for chronic diseases with late outcomes [5]	Generally shorter, as surrogate markers can be measured earlier [5]
Trial Size	Typically requires larger sample sizes to detect clinically meaningful differences	Often feasible with smaller populations due to more frequent and measurable endpoints [5]
Development Costs	Higher due to longer duration and larger size [6]	Lower due to reduced timeline and smaller trials [6]
Regulatory Certainty	High certainty when benefit is demonstrated	Variable certainty depending on validation level; may require post-market confirmation [6]
Risk of Misleading Results	Low when properly measured and adjudicated	Higher risk if surrogate does not adequately predict clinical outcome [6]
Patient Relevance	High, as they measure outcomes that matter directly to patients	Variable, depending on how well patients understand the connection to their experience

Validation Challenges and Evidentiary Standards

A fundamental challenge in using surrogate endpoints lies in establishing their validity—demonstrating that effects on the surrogate reliably predict effects on clinically meaningful outcomes. The International Conference on Harmonisation (ICH) Guideline E9 outlines key criteria for establishing this relationship [11]:

• Biological plausibility: The surrogate should be on the causal pathway of the disease
• Statistical relationship in epidemiological studies: The surrogate should correlate with the clinical outcome in observational studies
• Evidence from clinical studies: Treatment effects on the surrogate should correspond to effects on clinical outcomes across multiple interventions

Despite these established criteria, reviews of validated surrogate markers used as primary endpoints in trials supporting FDA approvals suggest that many lack sufficient evidence of being associated with a clinical outcome [6]. In oncology, for instance, most validation studies of surrogate markers find low correlations with meaningful clinical outcomes such as overall survival or quality of life [6]. This validation gap represents a significant challenge in the increasing reliance on surrogate endpoints for regulatory decision-making.

Experimental Design and Methodological Considerations

Endpoint Adjudication in Clinical Trials

Clinical trial endpoint adjudication has emerged as a major component of clinical trials in recent years, driven by increasing complexity of trial design and growing requirements from Health Authorities [5]. Independent blinded review and adjudication of both efficacy and safety endpoints helps ensure objective, consistent endpoint assessment across study sites, particularly when using subjective clinical endpoints.

The use of surrogate endpoints can significantly impact the adjudication process. In some cases, well-validated, objectively measured surrogate endpoints may make adjudication unnecessary [5]. For example, while recognizing and defining whether a patient has suffered a stroke requires expert neurological assessment, measuring systolic blood pressure is a simple procedure that can be performed by any trained site personnel. However, in rare cases, the evaluation of surrogate endpoints may be more complex than that of the primary outcome or may need to be combined with other endpoints to adequately describe the patient's disease status [5].

Biomarker Qualification and Novel Endpoint Development

For novel surrogate endpoints not yet included in the FDA's table, sponsors can engage with the FDA through the Biomarker Qualification Program or scheduled meetings to discuss feasibility and evidence requirements [3]. The PDUFA VI Commitment Letter outlines a Type C meeting process specifically for sponsors who would like to employ a biomarker as a surrogate endpoint that has not been used previously as the primary basis for product approval in the proposed context of use [3].

These meetings typically occur when sponsors have preliminary clinical study results showing that the proposed biomarker responds to the candidate drug at generally tolerable doses. The meeting aims to discuss the feasibility of the surrogate as a primary efficacy endpoint, identify knowledge gaps, and discuss how those gaps could be addressed before the surrogate endpoint can serve as the primary basis for product approval [3].

The following diagram illustrates the workflow for developing and validating novel surrogate endpoints:

Global Perspectives: Endpoint Regulation Beyond the United States

While the United States has established clear frameworks for endpoint use in drug development, other regions approach endpoint regulation differently. A 2025 study examining surrogate endpoints in Japan for drugs approved from 1999 to 2022 found that of 2,307 pharmaceutical products approved during this period, 1,012 (43.9%) were indicated for diseases with surrogate endpoints specified in the FDA's Surrogate Endpoint Table [12].

The study revealed that 947 drugs (93.6%) were approved using the same surrogate endpoint as the FDA, while 65 (6.4%) were approved using a different endpoint [12]. Significant differences were observed across therapeutic categories:

• Metabolic drugs showed high consistency between Japan and the U.S., with 98.7% using the same surrogate endpoints
• Drugs against pathogenic organisms showed significantly lower consistency, with only 87.6% using the same endpoints as the FDA

Unlike the U.S., Japan lacks established rules or guidance regarding surrogate endpoint use, with discussions based primarily on past practices and consultations between regulatory authorities and sponsors for individual drugs [12]. This approach creates a situation that "lacks transparency, universality, and academic merit" according to researchers, highlighting the need for further consideration and guidance regarding SEPs in Japan [12].

Research Reagent Solutions for Endpoint Development

Table 4: Key Research Reagents and Resources for Endpoint Development and Validation

Resource Type	Specific Examples	Function in Endpoint Research
Biomarker Assays	High-sensitivity C-reactive protein (hs-CRP), Troponins, Creatine kinase MB band (CK-MB) [11]	Provide quantitative measures of biological processes for use as potential surrogate endpoints
Imaging Technologies	Quantitative coronary perfusion, Intravascular ultrasound, Magnetic resonance imaging (MRI), Nuclear imaging (99mTc-SPECT) [11]	Enable non-invasive visualization and quantification of pathological processes and treatment effects
Functional Assessment Tools	Endothelial function tests, Arterial stiffness measurements, Left ventricular systolic/diastolic volume assessment [11]	Measure physiological functions that may serve as surrogate markers for clinical outcomes
Genomic and Proteomic Platforms	Functional genomics, Proteomics, Modern analytical technologies [11]	Facilitate discovery of novel biomarkers through comprehensive molecular profiling
Preclinical Models	Ionic channel assays, hERG channel binding studies, Guinea pig myocytes, Rabbit or dog Purkinje fibers [11]	Provide initial assessment of biomarker response and safety signals before human trials
Data Analysis Tools	PK/PD modeling techniques, Computational methods/informatics [11]	Support quantitative assessment of relationship between biomarker response and clinical outcomes

Drug developers have access to several key regulatory resources when designing endpoints for clinical trials:

• FDA Surrogate Endpoint Table: Provides listed endpoints that have supported approvals or that FDA anticipates could be appropriate endpoints [10]
• FDA Clinical Outcomes Assessment (COA) Compendium: Summarizes how COAs have been used in clinical trials to measure patient experience [5]
• Biomarker Qualification Program: Allows biomarker developers to request regulatory qualification of a biomarker for a particular context of use in drug development [3]
• Type C Meetings for Novel Surrogate Endpoints: Enables sponsors to discuss biomarkers as surrogate endpoints that haven't been used previously as the primary basis for approval [3]

Future Directions and Emerging Trends in Endpoint Science

Advancing Endpoint Methodologies

The field of clinical trial endpoints continues to evolve, with several emerging trends shaping future approaches:

• Multistate Models: In critical care research, traditional endpoints like all-cause mortality are increasingly supplemented by more nuanced approaches. Multistate models conceptualize critical illness as a sequence of transitions among mutually exclusive clinical states (e.g., noninvasive ventilation, invasive ventilation, death), providing a dynamic alternative to cross-sectional assessments [13]. These models capture both transitions and states while intrinsically handling competing risks, offering more comprehensive assessment of treatment effects in complex critical illnesses [13].

• Longitudinal Frameworks: Interest is growing in longitudinal frameworks that represent patient trajectories, moving beyond traditional cross-sectional designs to better account for unequal follow-up, censoring, competing risks, and time-varying exposures [13]. These approaches align trial objectives, design, and analysis through the "estimands framework"—a structured approach that requires explicit specification of the treatment effect of interest and handling of intercurrent events [13].

• Patient-Reported Outcomes (PROs): Regulatory guidance and sponsor priorities are converging to incorporate PROs into early-phase trial designs, particularly in areas like oncology where they offer critical insights into symptomatic adverse events and patient tolerability [14]. For 2025, the inclusion of PROs in early-phase oncology trials is expected to become increasingly emphasized as part of comprehensive safety and tolerability profiles [14].

Addressing Current Challenges

Despite advances, significant challenges remain in endpoint science. Reviews suggest that most validation studies of surrogate markers find low correlations with meaningful clinical outcomes [6]. In a review of 15 surrogate validation studies conducted by the FDA for oncologic drugs, only one demonstrated a strong correlation between surrogate markers and overall survival [6]. This validation gap has prompted calls for:

• Enhanced Transparency: FDA should make more transparent the strength of evidence of surrogate markers included in its endpoint tables, including justifications and citations to relevant validation studies [6]
• Interagency Collaboration: The Department of Health and Human Services should establish an interagency working group including FDA, NIH, PCORI, ARPA-H, and CMS to collaboratively conduct or commission meta-analyses of existing clinical trials to determine whether there is sufficient evidence to establish surrogacy [6]
• Regular Reassessment: Congress should mandate that FDA and other federal health agencies re-evaluate listed surrogate endpoints annually, with authority to sunset those that fail to show association with meaningful clinical outcomes [6]

As endpoint science continues to evolve, the fundamental regulatory imperative remains: to balance the need for efficient drug development with the certainty that approved therapies provide meaningful clinical benefit to patients. The ongoing refinement of endpoint strategies will undoubtedly continue to shape drug approval pathways for the foreseeable future, requiring sponsors to maintain vigilance in their endpoint selection and validation approaches.

In the relentless pursuit of accelerating patient access to novel therapies, clinical trial design has undergone a fundamental transformation. The most clinically relevant endpoints, such as overall survival (OS) in oncology, often require extensive follow-up durations and larger sample sizes, creating significant logistical and financial challenges for drug developers [15]. In this context, surrogate endpoints have emerged as critical tools for streamlining clinical research. Defined as biomarkers or measures that are not direct assessments of clinical benefit but are expected to predict it, surrogate endpoints can substantially reduce trial duration and size while driving down research and development costs [4]. Regulatory agencies worldwide have increasingly accepted validated surrogate endpoints, particularly for serious conditions with unmet medical needs. This paradigm shift raises crucial questions for researchers and drug development professionals: How prevalent has this practice become? Which surrogate endpoints have proven most valid? And what methodological frameworks ensure their proper use? This guide provides a data-driven comparison of surrogate endpoint utilization, validation methodologies, and implementation across therapeutic areas, offering an objective analysis for professionals navigating this evolving landscape.

Quantitative Landscape: Prevalence Across Regions and Specialties

Global Adoption Patterns

The use of surrogate endpoints has become a mainstream strategy in drug development rather than an exception. Comprehensive research investigating drugs approved in Japan over a 24-year period (1999-2022) provides compelling quantitative evidence of this trend. Among 2,307 pharmaceutical products approved, 1,012 drugs (43.9%) were indicated for diseases where surrogate endpoints were specified in the FDA's Surrogate Endpoint Table [12]. This extensive analysis revealed that Japan's regulatory practices largely align with American standards, with 947 drugs (93.6% of those targeting indications with established surrogates) approved using the same surrogate endpoints as the FDA [12]. The consistency between these major regulatory systems underscores the global acceptance of surrogate endpoints in modern drug development.

Annual trends from this dataset demonstrate increasing standardization, with the use of different surrogate endpoints than the FDA (classified as EP-nSEP) decreasing to ≤5% in recent years [12]. However, significant specialty-specific variations persist. The proportion of drugs using the same SEPs as the FDA was significantly higher for metabolic drugs (98.7%) compared with agents against pathogenic organisms (87.6%), which more frequently employed Japan-specific surrogate endpoints (p < 0.001) [12]. This heterogeneity highlights how surrogate endpoint validation remains context-dependent, influenced by disease mechanism, patient population, and therapeutic mechanism of action.

Oncology: A Case Study in Surrogate Endpoint Utilization

Oncology represents a therapeutic area where surrogate endpoints have become particularly prevalent, driven by the urgent need to accelerate availability of life-extending therapies. The FDA's Accelerated Approval pathway has been instrumental in this transition, allowing drugs for serious conditions to be approved based on effects on a surrogate endpoint "reasonably likely" to predict clinical benefit [16]. This regulatory mechanism has played a major role in making innovative cancer treatments available more quickly, though it requires sponsors to conduct post-marketing confirmatory trials to verify anticipated benefits [16].

Table 1: Common Surrogate Endpoints in Oncology Drug Development

Surrogate Endpoint	Category	Definition	Predictive Strength for OS	Example FDA Use Case
Progression-Free Survival (PFS)	Reasonably Likely	Time from treatment start until disease progression or death	Varies by cancer type; R²=0.79 for ADCs [17]	Bevacizumab for recurrent glioblastoma [16]
Objective Response Rate (ORR)	Reasonably Likely	Proportion of patients with ≥30% tumor shrinkage per RECIST criteria	Moderate association; R²=0.47 for ADCs [17]	Pembrolizumab for MSI-H/dMMR solid tumors [16]
Pathologic Complete Response (pCR)	Validated	Absence of invasive cancer in breast and lymph nodes after neoadjuvant therapy	Strong correlation with EFS/OS in specific cancers [16]	Pertuzumab for neoadjuvant HER2+ breast cancer [16]
Major Molecular Response (MMR)	Validated	≥3-log reduction in BCR-ABL transcript levels in CML	Validated for chronic myeloid leukemia [16]	Imatinib for chronic myeloid leukemia [16]

Recent empirical research specifically evaluating antibody-drug conjugates (ADCs) for solid tumors provides crucial quantitative insights into the predictive strength of common oncology surrogates. A meta-analysis of 25 randomized clinical trials encompassing 26 treatment comparisons and 11,729 patients found that PFS demonstrated a strong trial-level association with OS (R² = 0.79; 95% CI = 0.66 to 0.92), while ORR showed only a moderate association (R² = 0.47; 95% CI = 0.11 to 0.83) [17]. This evidence supports PFS as a robust surrogate endpoint for OS in ADC trials, offering greater reliability than ORR for supporting accelerated approval decisions [17].

Methodological Frameworks: Validation and Evaluation

The Ciani Framework for Surrogate Endpoint Validation

For a surrogate endpoint to be considered valid, it must undergo rigorous evaluation across multiple evidence dimensions. The Ciani framework has gained widespread acceptance by the international health technology assessment community, proposing three hierarchical levels of evidence for surrogate endpoint validation [4]:

• Level 3 (Biological Plausibility): Evidence that the surrogate endpoint lies on the disease pathway with the final patient-relevant outcome, based on clinical data and understanding of disease mechanisms.
• Level 2 (Individual-Level Association): Epidemiological studies and/or clinical trials demonstrating the relationship between the surrogate endpoint and target patient-relevant outcome at the individual level.
• Level 1 (Trial-Level Surrogacy): The highest level of evidence, requiring randomized controlled trial data demonstrating an association between the treatment effect on the surrogate and the treatment effect on the target outcome [4].

This structured approach ensures that surrogate endpoints are not only statistically correlated with clinical outcomes but also biologically plausible and demonstrably responsive to therapeutic interventions in a manner that predicts ultimate clinical benefit.

Advanced Statistical Methodologies

Traditional methods for surrogate endpoint validation have relied heavily on the hazard ratio as a measure of treatment effect, which assumes proportional hazards that may not hold true in practice. Departures from proportional hazards are frequent in cancer RCTs, limiting the reliability of these conventional approaches [15]. Innovative statistical methodologies are emerging to address these limitations:

• Restricted Mean Survival Time (RMST) Differences: A novel two-stage meta-analytic model uses RMST differences to quantify treatment effects without requiring the proportional hazards assumption. This approach captures the strength of surrogacy at multiple time points and can evaluate surrogacy with a time lag between surrogate and true endpoints [15].
• Trial-Level Coefficient of Determination (R²): This statistic quantifies the variation in the true endpoint explained by variation in the surrogate endpoint. In the ADC meta-analysis, this method revealed the substantially stronger association between PFS and OS compared to ORR and OS [17].
• Surrogate Threshold Effect (STE): An increasingly reported metric, the STE represents the magnitude of treatment effect on the surrogate that would predict a significant treatment effect on the target outcome, providing a valuable benchmark for clinical decision-making [4].

Table 2: Experimental Framework for Surrogate Endpoint Validation

Validation Component	Methodology	Data Requirements	Key Output Metrics
Trial-Level Surrogacy	Meta-analysis of multiple RCTs assessing both surrogate and true outcomes	Aggregate trial-level data or individual patient data	Coefficient of determination (R²), Spearman's correlation, STE [4]
Individual-Level Association	Correlation analyses between surrogate and final outcome at patient level	Individual patient data from clinical trials	Correlation coefficients, hazard ratios [4]
Temporal Validation	RMST-based models evaluating surrogacy at multiple timepoints	Individual patient data with varying follow-up durations	Time-varying surrogacy strength, lag effects [15]
Biological Plausibility Assessment	Pathophysiological research on disease mechanisms	Basic science studies, biomarker research	Mechanistic evidence supporting causal pathway [4]

Experimental Protocols and Research Workflows

Protocol for Meta-Analytic Surrogacy Validation

The gold standard approach for validating surrogate endpoints involves meta-analyzing data from multiple randomized controlled trials. The following protocol outlines the key methodological steps:

• Trial Identification and Selection: Systematically identify RCTs testing the interventions of interest in the target patient population, which must report data on both the surrogate endpoint and the reference clinical outcome (e.g., OS) [17].
• Data Extraction and Harmonization: Extract treatment effect estimates for both surrogate and true endpoints from each trial. For time-to-event outcomes, this typically involves hazard ratios with confidence intervals [17]. When individual patient data are available, more advanced analyses using RMST differences are possible [15].
• Statistical Analysis of Trial-Level Associations: Employ linear regression models weighted by trial size or precision to assess the relationship between treatment effects on the surrogate and true endpoints. The coefficient of determination (R²) from this model quantifies the trial-level surrogacy [17] [4].
• Cross-Validation and Sensitivity Analyses: Validate findings through leave-one-out cross-validation or bootstrap procedures. Conduct subgroup analyses based on tumor type, line of therapy, or drug characteristics to assess consistency of the surrogate relationship across clinical scenarios [17].
• Assessment of Surrogate Threshold Effect: Determine the minimum treatment effect on the surrogate endpoint necessary to predict a statistically significant effect on the final clinical outcome [4].

This protocol adheres to the recently developed 'Reporting of Surrogate Endpoint Evaluation using Meta-Analyses' (ReSEEM) guidelines to ensure methodological rigor and transparent reporting [4].

Pathway for Surrogate Endpoint Evaluation

The following diagram illustrates the logical workflow and decision points in the surrogate endpoint evaluation process, integrating the key concepts from the Ciani framework and statistical validation methods:

The Scientist's Toolkit: Essential Research Reagents and Solutions

The following table details key reagents, biomarkers, and methodological tools essential for conducting surrogate endpoint research across therapeutic areas:

Table 3: Research Reagent Solutions for Surrogate Endpoint Studies

Research Tool Category	Specific Examples	Function in Surrogate Endpoint Research
Tumor Response Biomarkers	RECIST criteria, circulating tumor DNA (ctDNA), pathologic complete response (pCR)	Objective assessment of treatment effect in oncology trials; ctDNA shows promise as a non-invasive biomarker for molecular response [16]
Kidney Function Biomarkers	Estimated glomerular filtration rate (eGFR), proteinuria, urinary protein-to-creatinine ratio (uPCR)	Quantify kidney function decline and protein leakage in nephrology trials; supported by strong evidence in IgAN [18]
Cardiovascular Surrogates	LDL cholesterol, blood pressure, hemoglobin A1c	Established validated surrogates for cardiovascular outcomes and diabetes control; accepted by regulatory agencies [8] [10]
Statistical Software Packages	R, SAS, Python with specialized meta-analysis packages	Implement advanced surrogacy validation methods including RMST differences and copula models [15]
Data Resources	Individual patient data meta-analyses, FDA Surrogate Endpoint Table, clinical trial registries	Provide foundational data for surrogacy validation and reference for acceptable endpoints [10]

The quantitative evidence presented in this guide demonstrates that surrogate endpoints have become firmly established in modern clinical trial design, with nearly half of drugs in certain jurisdictions being approved based on these measures. The growing prevalence reflects a strategic balance between the need for efficient drug development and the imperative to demonstrate meaningful clinical benefit. The data reveals a nuanced landscape: while validated surrogates like PFS in specific oncology settings show strong predictive value (R² = 0.79 for ADCs), the strength of association varies considerably across endpoints and therapeutic areas [17]. This underscores the critical importance of context-specific validation rather than blanket application of surrogate measures.

For researchers and drug development professionals, the evolving landscape demands rigorous adherence to established validation frameworks like the Ciani criteria and sophisticated statistical approaches that account for non-proportional hazards and temporal dynamics [15] [4]. The fundamental challenge remains navigating the trade-off between speed of drug development and certainty of clinical benefit—a balance that must be continually recalibrated based on accumulating evidence about the predictive performance of surrogate endpoints across diverse clinical contexts [19]. As methodological innovations continue to emerge and more data become available from post-marketing confirmation studies, the evidence base for surrogate endpoints will further mature, enabling more precise quantification of their utility and limitations across the therapeutic development spectrum.

In the field of clinical drug development, the choice of endpoints fundamentally shapes trial design, duration, cost, and ultimately, regulatory decisions. While clinical endpoints such as overall survival (OS) and quality of life (QOL) measure what is inherently meaningful to patients, the pharmaceutical industry increasingly relies on surrogate endpoints—intermediate measures that predict clinical benefit [20] [21]. These biomarkers or intermediate outcomes serve as substitutes for clinical outcomes of interest to expedite research and decision-making [21]. This shift is particularly pronounced in oncology, where surrogate endpoints like progression-free survival (PFS) and response rate (RR) are now commonly used in trials supporting marketing authorisation [20].

The drivers behind this transition are multifaceted, rooted in practical necessities but balanced by significant limitations. This article examines the key drivers, evaluates the performance of surrogate versus clinical endpoints, details experimental methodologies for validation, and outlines essential tools for researchers navigating this complex landscape.

Key Drivers for Adoption

Efficiency and Resource Constraints

• Reduced Development Time: Surrogate endpoints allow for significantly shorter trial durations because they measure effects that occur sooner than final clinical outcomes [22]. For example, assessing tumor shrinkage takes far less time than determining whether a drug improves cancer patient survival [23].
• Smaller Sample Sizes: Trials using surrogate endpoints typically require fewer patients [22] [24], making them less expensive and operationally more feasible, especially for serious conditions with unmet medical needs.
• Accelerated Patient Access: Regulatory pathways like the FDA's Accelerated Approval allow drugs with meaningful advantages over existing therapies to reach the market faster based on surrogate endpoints [23].

Regulatory and Commercial Factors

• Support for Marketing Authorization: Global regulatory agencies have recognized surrogate endpoints as valid primary efficacy indicators in support of drug or biologic approval [22]. The FDA maintains a list of over 100 surrogate endpoints that have been used in approved drug development programs [22].
• Investment in Innovation: The ability to demonstrate drug effects more quickly with surrogate endpoints encourages continued investment in pharmaceutical research and development, particularly in high-risk areas like oncology [20] [25].

Scientific and Technical Advancements

• Biomarker Discovery: Advances in biomarker research have identified numerous physiological indicators that correlate with disease progression or treatment response [25] [21].
• Individualized Therapy: In emerging fields like mRNA cancer vaccines, early biological markers are crucial for assessing immune activation signals before traditional efficacy measures like tumor shrinkage become apparent [25].

Table 1: Categories of Common Surrogate Endpoints in Oncology

Category	Type of Measurement	Examples	Typical Context
Tumor Shrinkage	Time point measurement	Response Rate (RR), Pathological Complete Response (pCR), Circulating Tumor DNA (ctDNA)	Solid tumors - local & advanced [20]
Haematological Measures	Time point measurement	Minimal Residual Disease (MRD), Complete Remission (CR), Major Molecular Response (MMR)	Liquid/Haematological tumors [20]
Time-to-Event Endpoints	Composite time-to-event	Progression-Free Survival (PFS), Disease-Free Survival (DFS), Event-Free Survival (EFS)	Both solid and liquid tumors [20]

Performance Comparison: Surrogate Endpoints vs. Clinical Endpoints

Quantitative Assessment of Clinical Utility

A comprehensive study presented at the 2025 American Society of Clinical Oncology (ASCO) Annual Meeting evaluated 791 randomized controlled trials (RCTs) published between 2002 and 2024, representing 555,580 patients [24]. The findings reveal significant disparities between surrogate endpoint performance and actual patient benefit.

Table 2: Outcomes of Oncology Trials Using Surrogate Endpoints (n=791 RCTs)

Outcome Measure	Success Rate	Findings
Alternative Endpoint Superiority	55%	More than half of trials met their primary surrogate endpoint [24]
Overall Survival (OS) Improvement	28%	Fewer than one-third of trials demonstrated actual survival benefit [24]
Quality of Life (QOL) Improvement	11%	Only one in nine trials showed improved patient-reported QOL [24]
Both OS and QOL Improvement	6%	A minimal proportion delivered both survival and life quality benefits [24]

Limitations and Clinical Concerns

The disconnect between surrogate endpoint performance and genuine clinical benefit presents several challenges:

• Therapeutic Misalignment: Surrogates can result in inappropriate stopping or switching of therapy at the bedside [20]. There is a risk of ushering in new treatment strategies that may ultimately erode patient outcomes [20].
• Magnitude Discrepancies: Large improvements in surrogate endpoints may translate to only minimal clinical benefits. For example, in the NeoSphere trial, a 17-point improvement in pCR was linked to a less than 1% improvement in 3-year invasive disease-free survival [20].
• Validation Gaps: Analyses indicate that nearly 60% of surrogate endpoints used in FDA approvals for chronic non-oncologic diseases lack high-strength evidence from randomized trial meta-analyses supporting their relationship with target outcomes [22].

Methodological Framework and Experimental Protocols

Statistical Validation of Surrogate Endpoints

Robust validation requires demonstrating that treatment effects on the surrogate endpoint reliably predict effects on the true clinical outcome. A novel two-stage meta-analytic approach using Restricted Mean Survival Time (RMST) differences addresses limitations of traditional methods that rely on hazard ratios and assume proportional hazards [15].

Diagram 1: Two-stage surrogate validation model using RMST

Experimental Protocol: RMST-Based Surrogacy Validation

Objective: To evaluate trial-level surrogacy between a surrogate endpoint (e.g., Disease-Free Survival) and a true clinical endpoint (e.g., Overall Survival) using individual patient data from multiple randomized controlled trials.

Stage 1: RMST and Pseudo-Observation Calculation

• Data Preparation: Collect individual patient data from multiple RCTs, including event times for surrogate and true endpoints, censoring indicators, and treatment assignments [15].
• Milestone Selection: Define clinically relevant time points (τ₁, τ₂, ..., τₖ) for evaluation, considering varying follow-up durations across trials [15].
• RMST Estimation: For each trial i and endpoint p (surrogate or true), calculate the RMST as μ̂ᵢᵖ(τ) = ∫₀ᵗ Ŝᵢᵖ(r)dr, where Ŝᵢᵖ(r) is the Kaplan-Meier survival estimator [15].
• Pseudo-Observation Generation: Replace censored outcome data with pseudo-observations using the formula θ̂ᵢⱼᵖ(τ) = nᵢμ̂ᵢᵖ(τ) - (nᵢ - 1)μ̂ᵢ⁽⁻ʲ⁾ᵖ(τ), where μ̂ᵢ⁽⁻ʲ⁾ᵖ(τ) is the RMST estimate computed after eliminating subject j [15].

Stage 2: Two-Stage Generalized Linear Mixed Model

• Model Specification: Fit a generalized linear mixed model (GLMM) to the pseudo-observations, accounting for correlations between endpoints and time points through random intercepts for individuals and endpoints [15].
• Treatment Effect Estimation: Extract RMST differences between treatment groups for both surrogate and true endpoints at each timepoint.
• Surrogacy Quantification: Compute the between-study covariance matrix of RMST differences and calculate the coefficient of determination (R²) to assess how well treatment effects on the surrogate endpoint explain effects on the true endpoint [15].

Key Advantages: This protocol does not require proportional hazards, captures surrogacy strength at multiple time points, and can evaluate surrogacy with a time lag between endpoints [15].

Research Reagent Solutions and Essential Materials

Table 3: Essential Research Materials for Surrogate Endpoint Studies

Research Tool	Function/Application	Considerations
Individual Participant Data (IPD)	Meta-analysis of multiple RCTs for surrogacy validation [15]	Gold standard for surrogate validation; requires data sharing agreements
RMST Analysis Software	Statistical computation of restricted mean survival time	R packages (survRM2, pseudo) enable RMST and pseudo-observation calculation [15]
Tumor Assessment Tools	Standardized measurement of tumor response (RECIST criteria)	Essential for solid tumor surrogate endpoints like PFS and RR [20]
Biomarker Assays	Detection and quantification of molecular surrogates (e.g., ctDNA, MRD)	Circulating tumor DNA enables minimal residual disease detection [20]
Quality of Life Instruments	Patient-reported outcome measures (e.g., EORTC QLQ-C30)	Critical for validating that surrogate benefits translate to patient-experienced benefits [24]

The pharmaceutical industry's reliance on surrogate endpoints is driven by compelling needs for efficiency, accelerated development, and regulatory pragmatism. However, recent evidence indicates that only a minority of trials based on surrogate endpoints ultimately demonstrate meaningful improvements in survival or quality of life. The validation of surrogate endpoints requires sophisticated statistical methodologies, such as RMST-based models that can evaluate surrogacy patterns over time without relying on proportional hazards assumptions. As drug development evolves, particularly in innovative fields like mRNA cancer vaccines, the disciplined use of rigorously validated surrogate endpoints, balanced with ongoing assessment of clinical benefit, will be essential for delivering therapies that genuinely improve patient outcomes.

For chronic conditions where assessing the definitive patient-relevant outcome, such as death or organ failure, can take many years, the use of surrogate endpoints is critical for accelerating clinical research and drug development. A surrogate endpoint is "a biomarker... that replaces a clinical endpoint" and is used to predict clinical benefit based on scientific evidence [26]. This guide objectively compares two prominent examples: GFR slope in chronic kidney disease (CKD) and various surrogate endpoints used in oncology, such as progression-free survival (PFS). The analysis is framed around the levels of validation required for a surrogate to be considered reliable and the distinct challenges faced in these two therapeutic areas, providing a practical comparison for researchers and drug development professionals.

Surrogate Endpoint Validation: A Hierarchical Framework

The acceptance of a surrogate endpoint by regulators and health technology assessment (HTA) bodies relies on a multi-level validation framework. The "Ciani framework" outlines three levels of evidence needed to establish a surrogate endpoint's validity [4].

Diagram: The Three-Level Validation Framework for Surrogate Endpoints

Key Validation Metrics

• Level 1 (Trial-Level Association): This is considered the most important for HTA decision-making. It is typically quantified using the coefficient of determination (R² trial), which measures how much of the variability in the treatment effect on the true clinical endpoint is explained by the treatment effect on the surrogate. An R² value close to 1.0 indicates a strong predictive relationship [4].
• Surrogate Threshold Effect (STE): This metric defines the minimum treatment effect on the surrogate endpoint needed to predict a statistically significant effect on the final clinical outcome. It is crucial for designing trials and interpreting their results [4].

GFR Slope as a Surrogate Endpoint in Chronic Kidney Disease

Endpoint Definition and Measurement

The glomerular filtration rate (GFR) slope measures the rate of change in kidney function over time, typically expressed in mL/min/1.73 m² per year. In CKD, a steeper negative slope indicates faster progression toward kidney failure. The estimated GFR (eGFR) is calculated from serum creatinine and other factors using validated equations [27].

Table 1: Key Methodological Approaches for eGFR Slope Calculation in Clinical Trials

Methodological Aspect	Common Approaches in CKD Trials	Rationale
Slope Type	Total Slope: Uses all data from randomization.Chronic Slope: Calculated from month 3 or 4 onwards to exclude acute effects.	"Total slope" demonstrated superior performance in a major meta-analysis (R²=0.97 vs 0.55 for chronic slope) [28].
Evaluation Period	2-3 years is common, but 1 year may be feasible in advanced CKD.	Shorter periods allow for faster trials but may require larger sample sizes. In CKD stages 4-5, a 1-year slope showed a strong association with kidney failure [27].
Statistical Model	Linear mixed-effects models with random intercepts and random slopes.	Accounts for both within-individual and between-individual variability in eGFR measurements over time [27].

Validation Evidence and Performance

Recent large-scale meta-analyses have provided robust Level 1 validation for GFR slope.

Table 2: Quantitative Validation Data for GFR Slope in CKD

Validation Metric	Reported Value	Interpretation and Context
Trial-Level R²	0.97 (for 3-year total slope) [28] [4]	Extremely high. Indicates that nearly all variation in treatment effects on clinical outcomes is explained by effects on GFR slope.
Treatment Effect Association	Each 0.75 mL/min/1.73 m²/year slower decline in GFR slope was associated with a 23.3% lower hazard for the clinical composite endpoint (KFRT, sustained GFR<15, or doubling of serum creatinine) [28].	Provides a quantifiable link between the surrogate and the clinical outcome.
Clinically Meaningful Difference	A deceleration of 0.5–1.0 mL/min/1.73 m²/year is considered a reliable treatment effect on long-term outcomes [27].	This range helps determine the target effect size for clinical trials.

Regulatory and HTA Status

Based on this strong validation, the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have officially approved GFR slope as an acceptable primary endpoint for clinical trials of CKD therapies [4] [29]. However, HTA agencies and payers remain more cautious, often requiring additional evidence for reimbursement decisions, highlighting a disconnect between regulatory approval and market access [4] [26].

Surrogate Endpoints in Oncology

Common Endpoints and Definitions

In oncology, surrogate endpoints are used to evaluate the efficacy of new cancer therapies more rapidly than waiting for overall survival (OS) data.

Table 3: Common Surrogate Endpoints in Oncology Clinical Trials

Endpoint	Definition	Clinical Context of Use
Progression-Free Survival (PFS)	Time from randomization until tumor progression or death from any cause [30].	Widely used across many cancer types for accelerated and regular approvals.
Time to Progression (TTP)	Time from randomization until tumor progression (excludes death) [30].	Less common than PFS, as it ignores the competing risk of death.
Disease-Free Survival (DFS)	Time from randomization until disease recurrence (used in adjuvant setting after definitive therapy) [30].	Common in trials for solid tumors after surgery (e.g., colon, breast cancer).
Objective Response Rate (ORR)	Proportion of patients with a predefined reduction in tumor size [31].	Often used in single-arm trials for accelerated approval.

Validation and Performance Challenges

The validation landscape for oncology surrogates is more mixed and context-dependent than for GFR slope in CKD.

• Variable Correlation with OS: A review of 15 analyses by the FDA found that only one showed a strong correlation between a surrogate (event-free survival) and OS [31]. A study of 153 cancer drug approvals showed that while the use of surrogate endpoints for approval rose to 85% in 2018, the percentage of drugs improving OS fell to a low of 7% in 2017 [31].
• Regulatory Reliance and Confirmatory Trials: The FDA's accelerated approval program allows drugs for serious conditions to be approved based on an effect on a surrogate "reasonably likely to predict clinical benefit." However, a study found that 57% of cancer drugs granted accelerated approval did not show a benefit in OS or quality of life within 5 years [31]. This has led to market withdrawals for some indications when confirmatory trials failed to verify clinical benefit [31].
• Arguments for Use: Despite limitations, surrogate endpoints remain necessary in oncology. They shorten trial durations, get treatments to patients faster, and in some specific contexts (e.g., multiple myeloma, prostate cancer), improvements in PFS have been associated with improvements in OS [31]. They are also less susceptible to being confounded by subsequent lines of therapy than OS [30].

Direct Comparison: GFR Slope vs. Oncology Surrogates

Diagram: Contrasting Validation and Use of Surrogates in CKD and Oncology

Table 4: Side-by-Side Comparison of Key Characteristics

Characteristic	GFR Slope (CKD)	Oncology Surrogates (e.g., PFS)
Underlying Concept	Continuous measure of organ function decline.	Time-to-event measure based on tumor growth or death.
Strength of Validation (Level 1)	Exceptionally strong (R² = 0.97) across multiple CKD etiologies [28] [4].	Variable and often weak; highly dependent on cancer type, treatment mechanism, and line of therapy [31].
Regulatory Acceptance	Accepted for full approval by FDA/EMA [4] [29].	Frequently used for accelerated approval; full approval may require confirmatory trials showing OS benefit [31].
Key Challenge	Bridging the acceptance gap between regulators and HTA bodies/payers [26].	High rate of failure in confirmatory trials and lack of demonstrated OS/QoL benefit post-approval [31].

The Scientist's Toolkit: Essential Reagents and Materials

Table 5: Key Research Reagent Solutions for Featured Endpoints

Item / Reagent	Function / Application	Specific Example / Context
Serum Creatinine Assay	Essential for calculating eGFR. Measured repeatedly over time to establish the GFR slope.	Used in all CKD clinical trials and routine clinical practice to monitor kidney function [27].
CKD-EPI or JSN-Specific eGFR Equation	Standardized formula to estimate GFR from serum creatinine, age, sex, and race, ensuring consistency across study sites.	The 2021 CKD-EPI equation is recommended. The Japanese cohort study used a equation tailored to the Japanese population [27].
RECIST (Response Evaluation Criteria in Solid Tumors) Guidelines	Standardized protocol for measuring tumor size on imaging (CT/MRI) to define "progression" or "response."	Critical for objectively determining PFS, TTP, and ORR in solid tumor oncology trials [30].
Linear Mixed-Effects Model Software	Statistical software packages capable of fitting complex models with random effects to calculate individual and group-level eGFR slopes.	Used with R, SAS, or Python to model eGFR trajectories in CKD trials, as described in the CKD-JAC study [27].

GFR slope in chronic kidney disease stands as a benchmark for a highly validated surrogate endpoint, with robust Level 1 evidence demonstrating it can reliably predict the clinical outcome of kidney failure across a wide range of patient populations. In contrast, surrogate endpoints in oncology, such as PFS, are indispensable for accelerating drug development but demonstrate variable and often weaker predictive validity, leading to greater uncertainty in their ability to reflect true patient benefit. This comparison underscores that the utility of a surrogate endpoint is not absolute but is contingent upon the strength of its hierarchical validation and the specific clinical and regulatory context in which it is applied.

Frameworks and Validation: The Science Behind Acceptable Surrogate Endpoints

In the drive toward faster patient access to new therapies, surrogate endpoints have become integral components of modern drug development and regulatory evaluation. Defined as biomarkers or intermediate outcomes that substitute for and predict final patient-relevant outcomes (such as mortality or health-related quality of life), surrogate endpoints enable shorter clinical trials with reduced costs and faster outcome accrual compared to trials measuring definitive clinical outcomes [4] [32]. This acceleration is particularly valuable in chronic diseases like chronic kidney disease (CKD), where definitive outcomes such as kidney failure may take many years to manifest [4]. However, reliance on unvalidated surrogate endpoints carries significant risks, including overestimation of clinical benefit, underestimation of harms, and ultimately inaccurate value assessment by health technology assessment (HTA) bodies [4] [32]. The Ciani framework has emerged as a widely accepted methodological approach for establishing the validity of surrogate endpoints, providing a structured process for moving from biological plausibility to demonstrated trial-level surrogacy [4].

The Three-Level Ciani Validation Framework

The Ciani framework proposes a hierarchical approach to surrogate endpoint validation, establishing three distinct levels of evidence that build upon one another to provide comprehensive demonstration of a surrogate's validity [4]. This framework has gained widespread acceptance within the international HTA community and provides a systematic methodology for assessing whether a surrogate endpoint can reliably predict clinical benefit [4] [33].

Table 1: The Three-Level Evidence Framework for Surrogate Endpoint Validation

Evidence Level	Definition	Source of Evidence	Statistical Metrics
Level 3: Biological Plausibility	Surrogate endpoint lies on the disease pathway with final patient-relevant outcome	Clinical data and understanding of disease mechanism	Not applicable
Level 2: Observational Association	Association between surrogate endpoint and target outcome at the individual level	Epidemiological studies and/or clinical trials	Correlation between surrogate endpoint and target outcome
Level 1: Trial-Level Surrogacy	Association between treatment effect on surrogate and treatment effect on target outcome	RCTs demonstrating association between treatment change in surrogate and final outcome	Trial-level R², Spearman's correlation, Surrogate Threshold Effect (STE)

The framework emphasizes that Level 1 evidence (trial-level surrogacy) is considered most crucial for HTA decision-making, as it demonstrates that treatments affecting the surrogate endpoint consistently produce corresponding effects on the final clinical outcome [4]. This hierarchical approach ensures that surrogate endpoints are evaluated through progressively rigorous evidence standards, with each level providing additional validation of the surrogate's reliability.

Level 3: Establishing Biological Plausibility

The foundation of surrogate endpoint validation begins with establishing biological plausibility - the demonstration that the putative surrogate endpoint lies on the causal pathway between the intervention and the final patient-relevant outcome [4]. This level requires a thorough understanding of the disease mechanism and the intervention's mechanism of action, providing the theoretical basis for why the surrogate should predict clinical benefit.

The validation at this level is primarily qualitative, drawing on clinical data and pathophysiological understanding of the disease process [4]. For example, in chronic kidney disease, glomerular filtration rate (GFR) slope possesses strong biological plausibility as a surrogate because it directly measures kidney function decline, which progressively leads to kidney failure requiring replacement therapy [4]. Similarly, in cardiovascular disease, reduction in LDL-cholesterol has biological plausibility for predicting cardiovascular mortality due to its established role in atherosclerosis progression [8]. While this level does not involve statistical validation, it provides the essential scientific rationale for proceeding to higher levels of validation.

Level 2: Demonstrating Individual-Level Association

The second validation level requires demonstrating an observational association between the surrogate endpoint and the target clinical outcome at the individual patient level [4]. This evidence typically comes from epidemiological studies or clinical trial data that show a correlation between the values of the surrogate and the ultimate clinical outcome of interest.

Statistical evaluation at this level focuses on measuring the strength of association between the surrogate and final outcome within individuals [4]. The specific metrics used depend on the nature of the endpoints but may include correlation coefficients, hazard ratios, or other measures of association. This level provides important evidence that the surrogate and final outcome are related in the expected direction across a population. However, it is critical to note that a strong individual-level association, while necessary, is not sufficient to establish a surrogate as valid for predicting treatment effects [4]. The framework emphasizes that many biomarkers have shown strong individual-level associations with clinical outcomes but failed to reliably predict treatment effects in randomized trials.

Level 1: Establishing Trial-Level Surrogacy

The highest level of validation in the Ciani framework is trial-level surrogacy, which requires demonstrating that the treatment effect on the surrogate endpoint predicts the treatment effect on the final patient-relevant outcome [4]. This level is considered most important for HTA decision-making because it directly addresses whether changes in the surrogate caused by an intervention reliably translate to changes in clinical benefit [4].

Evidence for trial-level surrogacy typically comes from meta-analyses of multiple randomized controlled trials that have measured both the surrogate and final outcomes [4]. The strength of association is quantified using metrics such as the coefficient of determination (R² trial), Spearman's correlation coefficient (ρ), or Kendall's tau [4]. An R² value of 1 would indicate perfect prediction of the treatment effect on the final outcome based on the effect on the surrogate, while values closer to 0 indicate poor predictive ability. For example, GFR slope in chronic kidney disease has demonstrated exceptionally strong trial-level surrogacy with an R² trial of 97% for predicting kidney failure outcomes [4].

Table 2: Statistical Metrics for Trial-Level Surrogacy Validation

Metric	Interpretation	Strength Assessment	Application in Decision-Making
Trial-level R²	Proportion of variance in treatment effect on final outcome explained by treatment effect on surrogate	0-0.25: Weak; 0.25-0.65: Moderate; >0.65: Strong	Higher values reduce decision uncertainty for HTA agencies
Spearman's Correlation (ρ)	Monotonic relationship between treatment effects on surrogate and final outcome	-1 to +1, with values closer to ±1 indicating stronger relationship	Non-parametric measure less sensitive to outliers
Surrogate Threshold Effect (STE)	Minimum treatment effect on surrogate needed to predict significant effect on final outcome	Smaller STE indicates more sensitive surrogate	Used to establish whether observed treatment effect is sufficient to predict clinical benefit

The Surrogate Threshold Effect (STE) has emerged as a particularly valuable metric for health technology assessment, as it quantifies the minimum treatment effect on the surrogate that would predict a statistically significant treatment effect on the final outcome [4]. This metric helps HTA agencies and payers determine whether the observed effect on a surrogate endpoint is sufficient to infer clinical benefit.

Experimental Protocols for Validating Surrogate Endpoints

Meta-Analytic Validation of Trial-Level Surrogacy

Objective: To quantitatively assess the relationship between treatment effects on a surrogate endpoint and treatment effects on a final clinical outcome across multiple randomized controlled trials.

Methodology:

• Identify Relevant Trials: Systematically identify all RCTs investigating interventions for the specific disease condition that report both the surrogate endpoint and the final clinical outcome of interest [4].
• Extract Treatment Effects: For each trial, extract the estimated treatment effects (e.g., hazard ratios, mean differences) and their measures of precision (confidence intervals, standard errors) for both the surrogate and final outcomes [4].
• Meta-Analytic Synthesis: Conduct a multivariate meta-analysis to model the relationship between the treatment effects on the surrogate and final outcomes across trials [4]. The preferred approach uses individual participant data (IPD) meta-analysis when available, as it allows for standardized statistical methods across datasets and robust analysis at both patient and trial levels [4].
• Evaluate Surrogacy Strength: Calculate trial-level surrogacy metrics including R², Spearman's correlation, and the surrogate threshold effect (STE) [4].
• Assess Consistency: Evaluate whether the surrogate relationship remains consistent across different patient subgroups, intervention types, and trial characteristics [4].

Interpretation: The validation is considered strong when the R² trial value exceeds 0.65-0.70, indicating that the treatment effect on the surrogate explains most of the variance in the treatment effect on the final outcome [4]. The framework emphasizes that surrogate validation should be based on RCTs with appropriate populations, interventions, comparators, and outcomes reflective of the specific HTA decision problem [4].

Biomarker Analytical Validation Protocol

Objective: To establish the analytical reliability and reproducibility of a biomarker used as a surrogate endpoint.

Methodology:

• Pre-analytical Factors: Standardize specimen collection, processing, and storage conditions to minimize pre-analytical variability [8].
• Analytical Performance: Establish assay precision (repeatability and reproducibility), accuracy (deviation from true value), sensitivity (limit of detection), and specificity (ability to measure analyte exclusively) [8].
• Reference Standards: Implement appropriate calibration and reference standards to ensure consistent measurement across different laboratories and over time [8].
• Quality Control: Establish internal quality control procedures and participation in external quality assessment schemes [8].

This analytical validation is a prerequisite before a biomarker can undergo clinical validation for use as a surrogate endpoint [8].

Research Reagent Solutions for Surrogate Endpoint Studies

Table 3: Essential Research Reagents and Materials for Surrogate Endpoint Validation

Reagent/Material	Function in Validation	Application Examples
Validated Assay Kits	Quantitative measurement of biomarker levels	LDL-cholesterol kits for cardiovascular surrogates; HbA1c kits for diabetes surrogates
Reference Standards	Calibration and standardization across laboratories	Certified reference materials for analytical validation
DNA/RNA Extraction Kits	Isolation of genetic material for biomarker analysis	Molecular surrogate studies in oncology and genetic disorders
Cell Culture Reagents	In vitro modeling of disease pathways and drug effects	Functional assays for biological plausibility studies
Statistical Software Packages	Implementation of multivariate meta-analysis methods	R, SAS, or Stata with specialized surrogacy analysis modules
Clinical Data Management Systems	Secure storage and processing of individual participant data	IPD meta-analysis platforms for trial-level surrogacy evaluation

Comparative Performance of Validated Versus Non-Validated Surrogates

The consequences of using inadequately validated surrogate endpoints can be significant, leading to misleading conclusions about treatment efficacy and potentially harmful coverage decisions. A review of NICE technology appraisals in oncology between 2022-2023 found that of 18 appraisals utilizing surrogate endpoints, the evidence supporting the validity of the surrogate relationship varied considerably [34]. Only 11 provided RCT evidence, 7 provided evidence from observational studies, 12 relied on clinical opinion, and 7 provided no evidence for the use of the surrogate endpoints [34]. This variability in validation rigor creates substantial uncertainty for HTA decision-makers.

Well-validated surrogate endpoints like GFR slope in chronic kidney disease (with R² trial of 97%) provide high confidence for both regulatory and HTA decisions [4]. In contrast, historical examples such as CD4+ counts in HIV/AIDS and tumor response in oncology have demonstrated that weakly validated surrogates can lead to approval of treatments with questionable effects on overall survival [4]. The Ciani framework addresses these limitations by providing a standardized, evidence-based approach to surrogate validation that minimizes decision uncertainty.

The Ciani validation framework provides a systematic, hierarchical approach to establishing the validity of surrogate endpoints, moving from biological plausibility to demonstrated trial-level surrogacy. This framework has become increasingly important as HTA agencies and payers worldwide face growing pressure to make coverage decisions based on surrogate endpoint evidence [4] [33]. The rigorous application of this framework enables more informed decision-making while facilitating faster patient access to genuinely beneficial therapies. As drug development continues to accelerate, the appropriate validation and use of surrogate endpoints will remain critical for balancing innovation with evidence-based healthcare resource allocation.

In the rigorous world of drug development and clinical research, validating that a treatment provides genuine patient benefit is paramount. Clinical outcomes directly measure how patients feel, function, or survive, serving as the most reliable indicators of treatment efficacy [3]. However, measuring these ultimate benefits often requires large, lengthy, and expensive trials. To accelerate the development of promising therapies, researchers increasingly rely on surrogate endpoints – biomarkers or other measures that are intended to predict clinical benefit [3]. The validation of these surrogate endpoints depends critically on robust statistical metrics, primarily the coefficient of determination (R²) and the correlation coefficient (r), which form the foundation for advanced methodologies like the Surrogate Threshold Effect (STE).

This guide provides a comprehensive comparison of these essential statistical tools, framing them within the critical context of surrogate versus clinical endpoint evaluation. For researchers and drug development professionals, understanding the strengths, limitations, and proper application of these metrics is crucial for designing efficient yet reliable clinical trials and accurately interpreting their results.

Core Statistical Metrics Explained

R-squared (R²): The Coefficient of Determination

R-squared is a goodness-of-fit measure for linear regression models that indicates the percentage of the variance in the dependent variable that the independent variables explain collectively [35]. Statistically, R² is defined as:

R² = 1 - (SS₍res₎ / SS₍tot₎)

Where SS₍res₎ is the sum of squares of residuals and SS₍tot₎ is the total sum of squares proportional to the variance of the data [36]. In practical terms, R² measures the strength of the relationship between your model and the dependent variable on a convenient 0-100% scale [35].

• Interpretation: An R² of 0% represents a model that does not explain any of the variation in the response variable around its mean, while 100% represents a model that explains all the variation [35].
• Key Limitation: A fundamental caveat is that R² does not indicate whether a regression model provides an adequate fit to your data. A good model can have a low R² value, while a biased model can have a high R² value [35]. This makes residual plots essential for complementary assessment.

Correlation Coefficient (r)

The Pearson correlation coefficient (r) measures the strength and direction of a linear relationship between two variables [37]. Unlike R², r is a unitless measure that always ranges between -1 and 1 [37].

• Interpretation: An r value of -1 indicates a perfect negative linear relationship, +1 indicates a perfect positive linear relationship, and 0 indicates no linear relationship [37].
• Relationship to R²: In simple linear regression, r is simply the square root of R², with the sign indicating the direction of the relationship [37]. The correlation coefficient r is directly related to the coefficient of determination R² in the obvious way. If R² is represented in decimal form, then all we have to do to obtain r is to take the square root of R² [37].

The Surrogate Threshold Effect (STE)

The Surrogate Threshold Effect (STE) is an advanced meta-analytic concept used specifically in surrogate endpoint validation. It determines the minimum treatment effect required on a surrogate endpoint to predict a significant effect on the true clinical outcome [38] [39].

In practical application, the STE represents the maximum value of the hazard ratio for a surrogate endpoint (e.g., HRPFS for progression-free survival) that needs to be observed in a trial to ensure the possibility of concluding a significant effect on the final clinical outcome (e.g., overall survival) [39]. This metric becomes particularly valuable when the correlation between surrogate and final endpoint is in the medium range, where surrogate validity is unclear [39].

Table 1: Key Differences Between R², r, and STE

Metric	Statistical Question	Range	Application Context	Interpretation Caveats
R²	What proportion of variance in the outcome is explained by the model?	0% to 100%	General regression model evaluation	Can be artificially inflated; high value doesn't guarantee a good or unbiased model [35].
r	How strong is the linear relationship between two variables?	-1 to +1	Assessing bivariate relationships	Only captures linear relationships; sensitive to outliers [37].
STE	What threshold on the surrogate is needed to predict a significant clinical effect?	Context-dependent (e.g., HR < 1)	Surrogate endpoint validation (meta-analysis)	Dependent on the quality and heterogeneity of included studies [39].

Experimental Protocols for Validation

Protocol for Validating a Surrogate Endpoint Using STE

The following workflow outlines the key steps for conducting a meta-analysis to validate a surrogate endpoint and calculate the STE, based on established methodologies [39].

Diagram 1: Workflow for Surrogate Endpoint Validation and STE Calculation

The specific methodology can be detailed as follows [39]:

• Systematic Literature Search & Study Selection: Conduct a comprehensive search of databases (e.g., MEDLINE, EMBASE) following PRISMA guidelines. Define precise inclusion criteria for randomized controlled trials (RCTs), specifying the patient population, interventions, and mandatory reporting of hazard ratios (HRs) for both the surrogate and final clinical endpoint, along with confidence intervals or standard errors.

• Data Extraction: From each included study, extract the hazard ratios for both the surrogate endpoint (e.g., HRPFS) and the true clinical outcome (e.g., HROS). The standard error (SE) for HROS must be calculated or extracted, often recalculated from the 95% confidence interval if not reported.

• Correlation Analysis: Calculate the Pearson correlation coefficient (r) between the effect estimates (e.g., hazard ratios) of the surrogate and the final endpoint across all trials. Test this correlation for statistical significance (H₀: r = 0 vs. H₁: r ≠ 0). The strength of the correlation dictates the next step [39]:

  • High Correlation (Lower Confidence Limit, LCL > 0.85): The surrogate endpoint is considered validated.
  • Low Correlation (Upper Confidence Limit, UCL < 0.7): No validation statement is possible.
  • Medium Correlation: Proceed with STE analysis.

• Meta-Regression for STE: For a medium correlation, fit a random effects mixed-model with the surrogate endpoint's HR (e.g., HRPFS) as the moderator and the true outcome's HR (e.g., HROS) as the outcome variable. The model should be weighted by the standard error of the true outcome's HR.

• Prediction Band and STE Calculation: Based on the meta-regression fit, calculate a prediction band for the HROS at a specified significance level (e.g., α = 0.05). The STE is the value of the surrogate endpoint's HR (HRPFS) at which the upper limit of this prediction band intersects the line of no effect (HROS = 1).

Case Study: PFS as a Surrogate for OS in Metastatic Breast Cancer

A 2019 meta-analysis applied this exact protocol to validate Progression-Free Survival (PFS) as a surrogate for Overall Survival (OS) in hormone receptor-positive, HER2-negative metastatic breast cancer [39].

• Included Studies: 16 randomized controlled trials with 5,324 patients total.
• Correlation Result: The correlation between hazard ratios of PFS and OS was r = 0.72 (95% CI: 0.35–0.90), which was statistically significant (p = 0.0016) but classified as only medium strength according to the IQWiG criteria [39].
• STE Result: The meta-regression model revealed a Surrogate Threshold Effect (STE) for HRPFS of 0.60. Sensitivity analyses confirmed the robustness of this result [39].
• Conclusion and Application: Based on this derived STE, a hypothetical trial demonstrating an upper confidence limit for HRPFS below 0.60 allows for a conclusion of a significant effect on OS. However, the authors note that only final OS results can confirm if a clinically relevant difference in survival time is actually achieved [39].

Table 2: Key Results from the Metastatic Breast Cancer STE Analysis

Metric	Result	Interpretation
Pearson Correlation (r)	0.72	A positive, statistically significant linear relationship between the treatment effects on PFS and OS.
95% Confidence Interval for r	0.35 to 0.90	The correlation is medium-strength, as the LCL is not >0.85 and the UCL is >0.7.
Surrogate Threshold Effect (STE)	HRPFS = 0.60	A trial must show an HRPFS with an upper confidence limit < 0.60 to predict a significant OS benefit.
Residual Heterogeneity (τ²)	0.009	Low residual heterogeneity among studies, increasing confidence in the model.
I² Statistic	25%	Low to moderate heterogeneity (25% of total variability due to between-study differences).

For researchers embarking on surrogate endpoint validation, having the right "research reagents" – in this context, data sources, software, and statistical tools – is essential for a successful analysis.

Table 3: Essential Research Reagents for Surrogate Endpoint Validation

Tool / Resource	Function / Purpose	Example / Note
Bibliographic Databases	Identifying relevant randomized controlled trials for the meta-analysis.	MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL) [39].
Statistical Software with Meta-analysis Packages	Performing correlation analysis, meta-regression, and calculating prediction intervals.	R with the metafor package was used in the case study [39].
PRISMA Guidelines	Ensuring a rigorous and reproducible systematic literature review process.	Provides a standardized flowchart for reporting study selection [39].
FDA Surrogate Endpoint Table	Referencing surrogate endpoints previously accepted in drug approvals.	Aids in context and justification for studying a particular surrogate [3].
Validated Endpoint Definitions	Ensuring consistent endpoint evaluation across all included studies.	For oncology, RECIST (Response Evaluation Criteria in Solid Tumors) is critical for PFS [39].
Hazard Ratio (HR) & Confidence Interval (CI) Data	The primary quantitative input for the validation analysis.	Must be extractable from published trials or obtained from study authors.

The journey from a promising surrogate endpoint to a validated predictor of clinical benefit is paved with rigorous statistical evaluation. R-squared provides the initial measure of how well a model capturing the surrogate-final endpoint relationship fits the observed data, while the correlation coefficient quantifies the strength of their linear association. When this correlation is meaningful but imperfect, the Surrogate Threshold Effect emerges as a powerful, practical tool for determining the specific threshold of treatment effect on the surrogate that is required to predict a tangible benefit for patients.

The case study in metastatic breast cancer demonstrates that while a significant correlation (r=0.72) exists between PFS and OS, it is not strong enough to validate PFS outright. Instead, the derived STE (HRPFS < 0.60) provides a clear, quantitative benchmark for researchers and regulators to use when evaluating new therapies in this space. By mastering these statistical metrics and the protocols for their application, drug development professionals can make more informed decisions, potentially accelerating the delivery of effective treatments to patients while maintaining the rigorous standards of evidence that underpin clinical benefit.

For drug developers, achieving regulatory approval from agencies like the U.S. Food and Drug Administration (FDA) or the European Medicines Agency (EMA) is a critical milestone. However, in most advanced healthcare systems, a second, equally crucial hurdle exists: securing positive health technology assessment (HTA) and reimbursement from payers [4]. While regulators may accept surrogate endpoints as evidence of efficacy, HTA bodies and payers are often more skeptical, requiring a deeper and more robust demonstration of a therapy's value [4] [40]. This guide compares the distinct evidence requirements for these two gatekeepers, providing a framework for developers to successfully navigate the journey from regulatory approval to market access.

Surrogate vs. Clinical Endpoints: A Primer for Evidence Generation

Clinical trial endpoints measure the outcomes used to evaluate a therapy's efficacy [3].

• Clinical Outcomes: Directly measure how a patient feels, functions, or survives (e.g., overall survival, reduction in pain, improved mobility). These are the ultimate indicators of treatment benefit [3].
• Surrogate Endpoints: Act as substitutes for clinical outcomes. These are typically laboratory measures or physical signs (e.g., blood pressure, tumor shrinkage, progression-free survival, glomerular filtration rate) that are expected to predict clinical benefit [3]. Their primary advantage is efficiency, allowing for shorter, smaller, and less costly trials [4] [40].

The tension arises because over 50% of FDA and EMA drug approvals are now based on surrogate endpoints [4], yet HTA bodies emphasize that a statistical correlation does not necessarily guarantee true surrogacy for patient-relevant outcomes [40].

Comparative Analysis: Regulatory vs. HTA & Payer Requirements

The following table summarizes the key differences in focus and evidence requirements between regulatory agencies and HTA bodies.

Table 1: Key Differences Between Regulatory and HTA/Payer Perspectives

Aspect	Regulatory Agencies (e.g., FDA, EMA)	HTA Bodies & Payers (e.g., NICE, IQWiG, HAS)
Primary Focus	Efficacy, safety, and risk-benefit balance [3]	Comparative clinical effectiveness, cost-effectiveness, and overall value for the healthcare system [4]
Endpoint Preference	Accepts validated and "reasonably likely" surrogate endpoints [3] [6]	Prefer patient-relevant final outcomes (e.g., OS, QoL); cautious acceptance of surrogates [4] [40]
Key Requirement for Surrogates	"Reasonably likely to predict clinical benefit" (for accelerated approval); validated for traditional approval [3]	Strong, context-specific validation demonstrating a quantitative link to final outcomes [4] [41]
Economic Evidence	Generally not considered	Central to decision-making; requires cost-effectiveness analysis (e.g., cost per QALY) [4]
Post-Market Evidence	Often required for accelerated approval (confirmatory trials) [6]	Increasingly required via managed entry agreements and real-world evidence (RWE) collection [40]

The HTA Evidentiary Framework: Validating Surrogate Endpoints

HTA agencies employ structured frameworks to evaluate the validity of a surrogate endpoint. The widely accepted "Ciani framework" outlines three levels of evidence required [4]:

Table 2: The Ciani Framework for Surrogate Endpoint Validation

Evidence Level	Definition	Source of Evidence	Key Statistical Metrics
Level 3: Biological Plausibility	The surrogate lies on the causal pathway of the disease and the clinical outcome.	Clinical data and understanding of disease biology.	Not applicable
Level 2: Individual-Level Association	An association exists between the surrogate and the target outcome at the individual patient level.	Epidemiological studies and/or clinical trials.	Correlation coefficients
Level 1: Trial-Level Association	The treatment effect on the surrogate is consistently associated with the treatment effect on the final outcome across multiple trials.	Meta-analysis of multiple RCTs assessing both surrogate and final outcome.	Coefficient of determination (R² trial), Spearman's correlation, Surrogate Threshold Effect (STE) [4]

Level 1 evidence, particularly from individual participant data (IPD) meta-analysis, is considered the most important for HTA decision-making [4]. The strength of the association is often quantified by the R² value, where a value close to 1 indicates a strong predictive relationship. For example, the glomerular filtration rate (GFR) slope in chronic kidney disease has been validated as a surrogate with an R² trial of 97% [4].

Experimental & Statistical Protocols for Validation

To meet HTA standards, developers must undertake rigorous surrogate validation studies. The following workflow outlines the key methodological steps.

Detailed Methodologies:

• Data Collection (PICO Framework): The validation must be based on RCTs with a range of Populations, Interventions, Comparators, and Outcomes that reflect the specific HTA decision problem. Extrapolating validation from different contexts (e.g., different drug classes) is often not accepted [4].

• Statistical Modeling for Trial-Level Association (Level 1): Multiple statistical models can be used, and comparing their predictions is considered best practice [42].

  • Weighted Linear Regression (WLR): A common approach where the treatment effect on the final outcome is regressed on the treatment effect on the surrogate, weighted by trial size. It provides a useful reference, but weights must account for follow-up time [42].
  • Bayesian Bivariate Random-Effects Meta-Analysis (BRMA): This model accounts for within-trial and between-trial variability and is often more robust, especially with smaller datasets, though it may require informative priors [42].
  • Model Comparison & Cross-Validation: A recent comparison of 6 models showed predictions can vary significantly, particularly with moderate associations. Using a leave-one-out cross-validation procedure to assess extrapolation error is critical for understanding prediction uncertainty [43] [42].

• Calculating the Surrogate Threshold Effect (STE): The STE is the minimum treatment effect on the surrogate needed to predict a statistically significant effect on the final outcome. It is a crucial metric for HTA, as it helps quantify the uncertainty when translating surrogate effects into long-term health benefits [4].

Case Study: Olaparib and the Payer Evidence Challenge

The oncology drug Olaparib (Lynparza) illustrates the market access challenges posed by heavy reliance on surrogate endpoints. Its pivotal trials used progression-free survival (PFS) and invasive disease-free survival (iDFS), which were sufficient for regulatory approval [40]. However, HTA bodies like France's HAS and Germany's G-BA were reluctant to grant broad reimbursement, emphasizing the uncertainty about whether these gains would translate into improved overall survival (OS) or quality of life (QoL) [40]. Payers demanded additional real-world evidence to confirm the long-term value, constraining the recognized added benefit and, consequently, the achievable price in key European markets [40]. This case underscores that a purely surrogate-based value proposition is rarely sufficient for a positive HTA outcome.

The Scientist's Toolkit: Key Reagents for Surrogate Endpoint Research

Table 3: Essential Research Reagents and Resources for Endpoint Validation

Item / Resource	Function & Application in Research
Individual Participant Data (IPD)	The optimal data source for surrogate validation, allowing for standardized statistical methods and robust analysis at both patient and trial levels [4].
FDA Surrogate Endpoint Table	A public list of over 200 surrogate endpoints that have been used or could be used for drug approval. Serves as a reference, but does not detail strength of evidence [3] [6].
CONSORT-Surrogate Guidelines	A reporting checklist for trials using surrogate endpoints as primary outcomes. Improves transparency, interpretation, and usefulness of trial findings [44].
ReSEEM Guidelines	Guidelines for the "Reporting of Surrogate Endpoint Evaluation using Meta-Analyses," ensuring rigorous and transparent methodology [4].
Type C Meeting (FDA)	A dedicated meeting type for sponsors to discuss novel surrogate endpoints with the FDA early in development, identifying evidence gaps [3].
Real-World Evidence (RWE)	Data collected outside of traditional RCTs (e.g., from electronic health records). Used post-approval to validate surrogates and reduce payer uncertainty [40].

Strategic Recommendations for Navigating HTA Hurdles

To successfully transition from regulatory approval to favorable HTA and payer decisions, developers should adopt the following strategies:

• Design Holistic Trials: Where possible, include final outcomes like OS or QoL as co-primary or key secondary endpoints, even if the primary endpoint for regulatory approval is a surrogate [40].
• Validate Context-Specifically: Do not assume a surrogate validated for one drug class or patient population will be accepted for another. Generate Level 1 evidence within the specific PICO context of the new therapy [4].
• Quantify the Link: Move beyond qualitative associations. Use meta-analytic techniques to establish a quantitative relationship between the surrogate and final outcome, and calculate the STE to inform cost-effectiveness models [4] [41].
• Plan for Post-Market Evidence: Develop robust plans for collecting RWE and conducting post-approval studies to confirm clinical benefit and address payer uncertainties, potentially through managed entry agreements [40] [6].
• Engage Early and Often: Proactively engage with both regulators and HTA bodies (e.g., via FDA Type C meetings or EUnetHTA Joint Scientific Consultations) to align on evidence requirements and avoid costly missteps [3] [40].

By integrating these requirements into the core of drug development planning, researchers and developers can build a compelling evidence dossier that demonstrates value not just to regulators, but to the payers who ultimately control patient access.

Chronic Kidney Disease (CKD) presents a substantial and growing global health burden, affecting approximately 10-15% of the global adult population and projected to become the fifth leading cause of death worldwide by 2040 [4]. The traditional clinical endpoints in CKD trials—kidney failure requiring replacement therapy (dialysis or transplantation), doubling of serum creatinine, or death—present significant practical challenges for drug development. These definitive outcomes typically require extensive follow-up periods spanning many years, large sample sizes, and consequently, substantial financial investment [4] [45]. This creates a pressing need for validated surrogate endpoints that can accurately predict clinical benefit while accelerating therapeutic development.

The glomerular filtration rate (GFR) slope, which measures the rate of kidney function decline over time, has emerged as a leading candidate surrogate endpoint. Its adoption represents a paradigm shift in the evaluation of CKD therapies. This case study provides a comprehensive examination of the rigorous multi-level validation process that has established GFR slope as a robust surrogate endpoint, enabling more efficient clinical trials and faster access to effective treatments for patients with CKD [4] [46].

Validation Frameworks and Levels of Evidence for Surrogate Endpoints

The Ciani Framework for Surrogate Endpoint Validation

For a biomarker to be accepted as a valid surrogate endpoint, it must undergo rigorous evaluation against established scientific frameworks. The "Ciani framework," widely accepted by the international health technology assessment (HTA) community, proposes three hierarchical levels of evidence required for surrogate endpoint validation [4]:

• Level 3 (Biological Plausibility): Evidence that the surrogate endpoint lies on the causal pathway of the disease and the clinical outcome.
• Level 2 (Observational Association): Epidemiological evidence demonstrating a consistent relationship between the surrogate endpoint and the target clinical outcome at the individual level.
• Level 1 (Interventional/Treatment Effect Association): Evidence from randomized controlled trials (RCTs) demonstrating that treatment effects on the surrogate endpoint reliably predict treatment effects on the final clinical outcome.

This framework emphasizes that trial-level evidence (Level 1) is the most crucial for validation, as it directly tests whether modifying the surrogate endpoint through intervention translates to meaningful clinical benefit [4].

GFR Slope Through the Validation Framework

The validation of GFR slope has systematically addressed all three levels of this framework. The biological plausibility is well-established: GFR directly measures the kidney's filtering capacity, and its progressive decline is the central pathophysiological process leading to kidney failure [45]. Observational studies have consistently shown that steeper GFR decline strongly correlates with higher risks of end-stage renal disease (ESRD) [45]. Most importantly, recent large-scale meta-analyses of RCTs have now provided robust Level 1 evidence, demonstrating that treatment effects on GFR slope reliably predict effects on hard clinical outcomes [47] [4].

Table 1: Validation of GFR Slope Against the Ciani Framework

Validation Level	Type of Evidence Required	Evidence for GFR Slope
Level 1: Trial-Level Association	RCT data showing treatment effects on surrogate predict effects on clinical outcome	Meta-analyses of 66 trials showing R² of 0.95 for predicting clinical outcome [47]
Level 2: Individual-Level Association	Observational association between surrogate and clinical outcome	Steeper GFR decline associated with 5.4-32.1x higher ESRD risk depending on decline threshold [45]
Level 3: Biological Plausibility	Surrogate lies on causal pathway to clinical outcome	GFR directly measures kidney filtration function; its decline is central to CKD progression [45]

Quantitative Evidence: Meta-Analyses validating GFR Slope

Landmark Meta-Analysis of 66 Clinical Trials

A pivotal 2025 meta-analysis by Greene et al. provided compelling Level 1 evidence for GFR slope as a surrogate endpoint. This comprehensive analysis included 66 randomized treatment comparisons from previous CKD clinical trials and employed a novel Bayesian meta-regression framework to examine the relationship between treatment effects on GFR slope and established clinical endpoints [47].

The key findings demonstrated that treatment effects on both acute (before 3 months) and chronic (after 3 months) GFR slopes independently predicted the treatment effect on the established clinical endpoint with a remarkably high median R² of 0.95 (95% credible interval: 0.79 to 1.00) [47]. This indicates that changes in GFR slope explain approximately 95% of the variability in treatment effects on clinical outcomes across trials. The analysis further revealed that for a fixed treatment effect on the chronic slope, each 1 ml/min/1.73 m² greater acute GFR decline for the treatment versus control increased the hazard ratio for the established clinical endpoint by 11.4% (7.9%-15.0%), against the treatment [47].

Supporting Evidence from Broader CKD Populations

Additional large-scale meta-analyses have reinforced these findings across diverse CKD populations. A 2023 meta-analysis by Inker et al. analyzed data from 66 trials involving 186,312 participants across various disease groups including diabetes, glomerular diseases, CKD, and cardiovascular diseases [46]. This study found a strong association between treatment effects on the total GFR slope and clinical endpoints, confirming that GFR slope could reliably predict treatment effects on kidney failure [46].

Another meta-analysis encompassing over 1.7 million patients with kidney disease worldwide examined the relationship between percentage changes in eGFR over a two-year period and the risks of ESRD [45]. The results demonstrated that a 30% decline in eGFR was associated with an adjusted hazard ratio for ESRD of 5.4 (95% CI 4.5-6.4), while a 57% decline (equivalent to doubling of serum creatinine) was associated with a hazard ratio of 32.1 (95% CI 22.3-46.3) [45].

Table 2: Key Meta-Analyses Validating GFR Slope as a Surrogate Endpoint

Study	Trials/Participants	Statistical Strength	Key Findings
Greene et al. (2025) [47]	66 randomized treatment comparisons	R² = 0.95 (0.79-1.00)	Both acute and chronic GFR slope effects independently predict clinical outcomes
Inker et al. (2023) [46]	66 trials (N=186,312)	Strong association demonstrated	GFR slope predicts treatment effects on kidney failure across diverse diseases
Coresh et al. [45]	1.7 million patients worldwide	HR=5.4 for 30% eGFR decline	Established threshold declines in eGFR as predictors of ESRD risk

Methodological Protocols for GFR Slope Analysis

Experimental Workflow for GFR Slope Validation Studies

The robust validation of GFR slope has relied on sophisticated methodological approaches applied to large datasets. The following diagram illustrates the key steps in the experimental workflow for validating GFR slope as a surrogate endpoint:

Statistical Approaches for Surrogacy Validation

Recent methodological advances have enhanced the rigor of surrogate endpoint validation. A novel two-stage meta-analytic model has been developed that employs restricted mean survival time (RMST) differences to quantify treatment effects at the first stage [15]. At the second stage, the model assesses surrogacy through coefficients of determination at multiple timepoints using the between-study covariance matrix of RMSTs and differences in RMST [15]. This approach offers significant advantages: it does not require the proportional hazard assumption, captures the strength of surrogacy at multiple time points, and can evaluate surrogacy with a time lag between surrogate and true endpoints [15].

For GFR slope specifically, the Bayesian meta-regression framework used in the Greene et al. analysis enabled the separation of acute and chronic treatment effects on GFR slope, providing crucial insights into their independent contributions to clinical outcomes [47]. This methodological sophistication has been essential for understanding the complex relationship between short-term changes in kidney function and long-term clinical benefit.

GFR Slope in Clinical Practice and Trial Design

Regulatory Acceptance and Clinical Applications

The robust validation evidence for GFR slope has led to its formal acceptance by regulatory agencies. Both the United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA) now include estimated glomerular filtration rate or serum creatinine as accepted surrogate endpoints for drug approval in chronic kidney disease [10]. This regulatory acceptance has transformed clinical trial design for CKD therapies, substantially reducing the duration and cost of drug development programs [4].

The optimal implementation of GFR slope in clinical trials has been refined through the validation studies. The evidence supports the 3-year total slope—defined as the average slope extending from baseline to 3 years—as the primary slope-based outcome in randomized trials [47]. This timeframe adequately captures both acute and chronic treatment effects while remaining practically feasible for clinical trial implementation.

Distinct Roles in Clinical Research Versus Practice

It is important to distinguish between the use of GFR slope as a validated surrogate endpoint in clinical trials versus its application as a risk stratification tool in clinical practice, as these represent "two missions" for the same metric [46].

Table 3: GFR Slope in Clinical Trials versus Clinical Practice

Aspect	Surrogate Endpoint in Trials	Risk Stratification in Practice
Primary Purpose	Measure treatment efficacy	Identify high-risk patients
Users	Researchers, regulators	Clinicians, guideline developers
Measurement	Standardized (e.g., 3-year total slope)	Flexible (e.g., >5 mL/min/1.73m²/year decline)
Validation	Scientifically proven to predict kidney failure	Useful for guiding real-world care decisions
Impact	Accelerates drug development	Enables personalized treatment plans

In clinical practice, eGFR slope has demonstrated value beyond kidney outcomes alone. A recent prospective cohort study of 5,362 older adults found that among participants with preclinical cardiac abnormalities (Stage B heart failure), a steeper annual decline in eGFR significantly increased the risk of developing clinical heart failure, particularly heart failure with preserved ejection fraction (HFpEF) [46]. Individuals with the steepest eGFR decline (< -1.87 mL/min/1.73 m² per year) had a 58% higher risk of incident heart failure compared to those with moderate declines [46].

The Scientist's Toolkit: Essential Research Reagents and Methodologies

The validation of GFR slope has relied on specific methodological approaches and analytical tools that constitute the essential "toolkit" for researchers in this field.

Table 4: Essential Methodological Components for GFR Slope Research

Component	Function	Application in GFR Slope Validation
Linear Mixed-Effects Models	Model longitudinal eGFR measurements	Calculate individual patient eGFR slopes accounting for within-patient correlation
Bayesian Meta-regression	Quantify relationship between treatment effects on surrogate and clinical outcomes	Estimate independent contributions of acute and chronic slope effects [47]
Restricted Mean Survival Time (RMST)	Measure treatment effect without proportional hazards assumption	Evaluate surrogacy at multiple timepoints in time-to-event settings [15]
Individual Patient Data Meta-analysis	Pool raw data from multiple trials	Gold standard approach for surrogacy evaluation across diverse populations [4]
Coefficient of Determination (R²)	Quantify strength of surrogacy	Measure proportion of variance in clinical outcome explained by surrogate [47]

Comparative Performance: GFR Slope Versus Other Surrogate Endpoints

The validation strength of GFR slope stands in contrast to the performance of surrogate endpoints in some other therapeutic areas. A recent evaluation of 791 randomized controlled trials in oncology found that while surrogate endpoints like progression-free survival (PFS) were commonly used (in 63% of trials), only 28% of trials deemed "positive" based on these surrogates ultimately demonstrated improved overall survival, and merely 11% showed improved quality of life [24]. This highlights the exceptional validation status achieved by GFR slope in the CKD domain.

The relationship between GFR slope, acute/chronic effects, and clinical outcomes can be visualized as follows:

The validation of GFR slope as a surrogate endpoint for CKD progression represents a notable advance in clinical trial methodology. Through large-scale meta-analyses of randomized controlled trials, GFR slope has demonstrated exceptional predictive performance for clinical outcomes, with treatment effects on GFR slope explaining approximately 95% of the variation in treatment effects on kidney failure [47]. This evidence has established GFR slope as a robust surrogate endpoint that meets the highest levels of validation criteria.

The implications for drug development are substantial. By utilizing GFR slope as a primary endpoint, clinical trials for CKD therapies can be significantly shortened in duration and reduced in size, accelerating the availability of new treatments while maintaining confidence in their clinical benefit [4]. This is particularly important given the growing global burden of CKD and the urgent need for more effective therapies.

Future directions in this field include ongoing refinement of the optimal implementation of GFR slope in trial design, further exploration of combination endpoints incorporating both GFR slope and proteinuria reduction [48], and investigation of novel biomarkers that may complement or enhance the predictive value of GFR slope. The successful validation of GFR slope serves as a model for the rigorous evaluation of surrogate endpoints across therapeutic areas and underscores the importance of robust statistical approaches in bridging the gap between surrogate markers and meaningful clinical outcomes.

The U.S. Food and Drug Administration (FDA) has developed a sophisticated toolkit to advance drug development, with two components serving complementary roles: the Surrogate Endpoint Table and the Patient-Focused Drug Development (PFDD) Guidance series. The Surrogate Endpoint Table provides a curated list of biomarkers and intermediate outcomes that can substitute for direct measurements of clinical benefit, potentially accelerating drug approval pathways [10] [3]. In parallel, the PFDD Guidance series establishes a systematic methodology for incorporating patient experience data into medical product development and regulatory decision-making [49] [50]. These frameworks represent distinct approaches to endpoint evaluation—one focusing on biological and physiological measures that predict clinical outcomes, and the other prioritizing direct patient input on meaningful treatment benefits. Together, they reflect the FDA's evolving approach to balancing efficiency in drug development with comprehensive assessment of patient-relevant outcomes.

Comparative Analysis: Core Characteristics and Applications

The following table summarizes the key characteristics of these two regulatory tools, highlighting their distinct purposes, regulatory foundations, and applications in drug development.

Table 1: Comparison of FDA's Surrogate Endpoint Table and PFDD Guidance

Characteristic	Surrogate Endpoint Table	PFDD Guidance Series
Primary Purpose	Accelerate drug development using validated biomarkers as substitutes for clinical outcomes [10] [3]	Systematically incorporate patient experience data into medical product development [49] [50]
Regulatory Foundation	Section 507 of FD&C Act (21st Century Cures Act) [10]	21st Century Cures Act & FDARA 2017 [50]
Endpoint Focus	Biomarkers, laboratory measurements, radiographic images [10]	Clinical Outcome Assessments (COAs): PROs, ObsROs, PerfOs, ClinROs [3]
Key Applications	- Traditional approval- Accelerated approval- Context-dependent endpoint selection [10]	- Identifying patient-important outcomes- Developing fit-for-purpose COAs- Clinical trial endpoint selection [49] [50]
Validation Requirements	Epidemiologic, therapeutic, pathophysiologic evidence of prediction [3]	Qualitative and quantitative evidence of measuring concepts important to patients [49]
Stakeholder Engagement	Sponsors discuss novel endpoints with FDA via Type C meetings [3]	Patients, caregivers, researchers, medical product developers [50]

The Surrogate Endpoint Table: Framework and Implementation

Regulatory Framework and Endpoint Classification

The Surrogate Endpoint Table fulfills Section 507 of the FD&C Act, mandating publication of surrogate endpoints used as the basis for drug approval under both accelerated and traditional pathways [10]. The table categorizes endpoints according to their validation status and appropriate regulatory pathway. Validated surrogate endpoints are supported by strong mechanistic rationale and clinical data demonstrating that an effect on the surrogate predicts a specific clinical benefit [3]. These endpoints can support traditional approval. In contrast, reasonably likely surrogate endpoints have strong mechanistic or epidemiologic rationale but insufficient clinical data for full validation, making them appropriate for the Accelerated Approval program [3].

The table is organized into four sections: adult non-cancer, adult cancer, pediatric non-cancer, and pediatric cancer endpoints [10]. This structure reflects the context-dependent nature of surrogate endpoints, where acceptability depends on disease, patient population, therapeutic mechanism, and available treatments [10].

Experimental Validation Methodologies

Table 2: Surrogate Endpoint Validation: Methodologies and Evidence Requirements

Validation Level	Epidemiologic Evidence	Clinical Trial Evidence	Pathophysiologic Evidence
Candidate Surrogate	Observational studies showing association between surrogate and clinical outcome	Preliminary clinical data suggesting response to intervention	Biological plausibility for causal pathway
Reasonably Likely	Consistent associations across multiple studies	Strong mechanistic rationale with some clinical correlation	Understanding of intervention's effect on disease pathway
Validated	Extensive evidence from multiple sources confirming predictive value	Multiple clinical trials demonstrating consistent prediction of clinical benefit	Well-understood causal pathway between surrogate and clinical outcome

The FDA recommends that sponsors seeking to use novel surrogate endpoints schedule PDUFA VI Type C meetings to discuss the feasibility of the surrogate as a primary efficacy endpoint and identify evidence gaps [3]. The background package for these meetings should include comprehensive data supporting the proposed surrogate endpoint relationship.

Experimental Implementation: Case Examples

Oncology Applications: In oncology, surrogate endpoints like progression-free survival (PFS) and overall response rate (ORR) are frequently used. Between 2016 and 2018, the percentage of cancer drugs approved based on surrogate endpoints increased from 57% to 85% [31]. However, validation of these surrogates remains challenging; one analysis found only 1 of 15 FDA-assessed surrogate endpoints showed strong correlation with overall survival [31].

Rare Disease Applications: For Duchenne muscular dystrophy (DMD), skeletal muscle dystrophin levels serve as a surrogate endpoint supporting accelerated approval [10]. This endpoint exemplifies the "reasonably likely" standard, where increased dystrophin expression is mechanistically linked to clinical benefit, with confirmatory trials required to verify actual clinical improvement.

Patient-Focused Drug Development Guidance: Framework and Implementation

The Four-Guidance Framework

The PFDD Guidance Series adopts a sequential approach to incorporating patient experience into drug development [50]:

• Guidance 1 focuses on collecting comprehensive and representative patient input, including sampling methods and defining target populations.
• Guidance 2 addresses methods to identify what is important to patients through qualitative and quantitative research.
• Guidance 3 (released October 2025) covers selecting, developing, or modifying fit-for-purpose Clinical Outcome Assessments.
• Guidance 4 will address methodologies for collecting, analyzing, and interpreting COA data.

This structured approach ensures that patient experience data are systematically integrated throughout the drug development process.

Clinical Outcome Assessment Development Workflow

The diagram below illustrates the methodological workflow for developing and implementing Clinical Outcome Assessments as outlined in the PFDD Guidance.

Research Reagent Solutions for Patient Experience Measurement

Table 3: Essential Methodological Tools for Patient-Focused Endpoint Development

Research Tool	Primary Function	Application Context
Structured Interview Guides	Elicit comprehensive patient input on disease experience	Qualitative research phase to identify concepts of importance
Cognitive Debriefing Protocols	Test patient understanding of COA items	COA development and modification to ensure content validity
Psychometric Validation Packages	Establish reliability, validity, responsiveness of COAs	Quantitative evaluation of measurement properties
Electronic Clinical Outcome Assessment (eCOA) Systems	Capture patient-reported data in clinical trials	Standardized data collection with time-stamped entries
Meaningful Change Threshold Kits	Establish clinically important differences	Interpretation of COA results in clinical trials

Comparative Endpoint Evaluation: Case Studies and Data

Oncology Endpoint Performance Analysis

Table 4: Surrogate vs. Patient-Focused Endpoints in Oncology Drug Approvals

Endpoint Category	Specific Endpoint	Correlation with Clinical Benefit	Time to Measurement	Regulatory Acceptance
Traditional Surrogate	Progression-Free Survival (PFS)	Variable; only 1/15 FDA analyses showed strong OS correlation [31]	Intermediate (months)	Accelerated & Traditional Approval
Traditional Surrogate	Overall Response Rate (ORR)	Predicts tumor shrinkage but not always survival or QOL [31]	Short (weeks-months)	Accelerated Approval
Patient-Reported Outcome	Quality of Life (QOL) measures	Direct measurement of patient benefit	Longitudinal (throughout trial)	Supportive evidence for approval
Patient-Reported Outcome	Symptom improvement	Direct measurement of patient experience	Short to intermediate	Can support primary endpoint
Clinical Outcome	Overall Survival (OS)	Gold standard for clinical benefit	Long (years)	Traditional Approval

Integration Case Study: Alzheimer's Disease

The recent approval of amyloid-beta targeting therapies for Alzheimer's disease demonstrates the potential integration of surrogate and patient-focused approaches. The reduction in amyloid beta plaques serves as a surrogate endpoint supporting accelerated approval [10]. However, this surrogate exists alongside patient-focused assessments of cognitive and functional decline. This dual approach acknowledges the biological mechanism while requiring continued evaluation of clinical meaningfulness through confirmatory trials.

The FDA's Surrogate Endpoint Table and PFDD Guidance represent complementary rather than competing approaches to drug evaluation. The Surrogate Endpoint Table offers efficiency in drug development through biologically-based endpoints with validated predictive value, particularly valuable for serious conditions with unmet needs. Meanwhile, the PFDD Guidance ensures that drug development remains grounded in outcomes that matter directly to patients, addressing the limitation that surrogate endpoints do not directly measure how patients feel or function [21].

The evolving regulatory landscape suggests increased emphasis on integrating these approaches. For researchers and drug developers, this means considering surrogate endpoints within the context of patient-important outcomes, and using PFDD methods to validate that surrogate changes correspond to meaningful patient benefits. As noted in recent analyses, the increasing use of surrogate endpoints makes ongoing validation and transparency about their limitations essential for maintaining the integrity of drug evaluation [31]. The most robust drug development strategies will leverage both tools—using surrogate endpoints for efficiency while grounding research in patient-important outcomes through PFDD methodologies.

Navigating Pitfalls and Controversies: Limitations and Ethical Considerations

In the relentless pursuit of accelerating drug development, surrogate endpoints have become fundamental components of modern clinical trials. These biomarkers or intermediate measurements are intended to substitute for direct assessments of how a patient feels, functions, or survives, offering a pathway to substantially reduce the size, duration, and cost of clinical research [4]. The underlying premise is compelling: by using endpoints that can be measured earlier, more frequently, or more conveniently than definitive clinical outcomes, promising therapies can reach patients faster. Regulatory agencies worldwide have embraced this approach, with the US Food and Drug Administration (FDA) maintaining a list of over 100 surrogate endpoints considered acceptable for drug approval [4]. In Japan, 43.9% of drugs approved between 1999 and 2022 were for indications with established surrogate endpoints [12].

However, a critical disconnect has emerged between success measured by these surrogate markers and meaningful patient outcomes. A comprehensive study presented at the 2025 American Society of Clinical Oncology (ASCO) Annual Meeting examining 791 randomized controlled trials revealed a disturbing trend: while 53% of trials met their primary surrogate endpoints, only 28% demonstrated actual improvement in overall survival, and a mere 11% showed enhanced quality of life from the patient's perspective [24]. Even more striking, only 6% of these "positive" trials improved both survival and quality of life simultaneously [24]. This stark reality underscores the "predictive gap"—the troubling divergence between statistical success on surrogate measures and tangible patient benefit that forms the critical focus of this comparison guide.

Quantitative Evidence: Documenting the Disconnect

The disparity between surrogate endpoint success and patient-centered outcomes is not merely theoretical but is substantiated by robust empirical evidence across therapeutic areas. The following analysis systematically compares surrogate endpoint performance against ultimate clinical benefit.

Table 1: Trial Success Rates: Surrogate Endpoints vs. Patient-Centered Outcomes

Metric of Success	Success Rate	Data Source	Sample Size
Trials meeting primary surrogate endpoint	53%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Superiority on alternative/surrogate endpoint	55%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Demonstration of overall survival benefit	28%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Improvement in patient-reported quality of life	11%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Improvement in both survival AND quality of life	6%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Trials collecting QOL data	61%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)
Trials publishing global QOL results	34%	ASCO 2025 Analysis [24]	791 RCTs (555,580 patients)

The data reveals systematic limitations in how surrogate success translates to patient benefit. Beyond these overall statistics, the validation strength of surrogate endpoints varies substantially, necessitating a structured framework for evaluation.

Table 2: Validation Framework for Surrogate Endpoints

Evidence Level	Definition	Source of Evidence	Statistical Metrics
Level 1: Trial-Level Surrogacy	Association between treatment effect on surrogate and target outcome	RCTs assessing both surrogate and final outcome	Trial-level R², Spearman's correlation, Surrogate Threshold Effect (STE)
Level 2: Individual-Level Association	Relationship between surrogate and target outcome at patient level	Epidemiological studies or clinical trials	Correlation between surrogate and final outcome
Level 3: Biological Plausibility	Pathophysiological rationale for surrogate relationship	Clinical data and disease understanding	Not applicable

This validation framework, often called the "Ciani framework" after its developers, highlights that Level 1 evidence—demonstrating that a treatment's effect on the surrogate reliably predicts its effect on the final outcome—is considered most important for health technology assessment (HTA) decision-making [4]. The strength of this association is quantified using metrics like the coefficient of determination (R² trial), where values closer to 1.0 indicate stronger predictive validity.

Methodological Protocols: Assessing Surrogate Endpoint Validity

Meta-Analytic Validation of Candidate Surrogates

Objective: To establish whether a candidate surrogate endpoint meets Level 1 evidence criteria for surrogacy by demonstrating that treatment effects on the surrogate endpoint reliably predict treatment effects on the target patient-relevant outcome across multiple clinical trials.

Experimental Workflow:

• Data Collection: Individual participant data (IPD) is the optimal approach, though meta-analyses of published aggregate trial-level data can be used when IPD is unavailable [4].
• Trial Selection: Include multiple randomized controlled trials that have assessed both the surrogate endpoint and the final target outcome. The analysis should incorporate trials with both positive and negative results to avoid publication bias [51].
• Statistical Analysis: Apply multivariate or Bayesian meta-analytic methods to evaluate the relationship between treatment effects on the surrogate and final outcomes across trials [4].
• Surrogacy Evaluation: Quantify the strength of association using:
  • Trial-level R²: The coefficient of determination representing the proportion of variance in the treatment effect on the final outcome explained by the treatment effect on the surrogate. Values >0.85 are often considered strong.
  • Surrogate Threshold Effect (STE): The minimum treatment effect on the surrogate needed to predict a statistically significant effect on the final outcome [4].
• Reporting: Follow the Reporting of Surrogate Endpoint Evaluation using Meta-Analyses (ReSEEM) guidelines to ensure comprehensive reporting [4].

Key Considerations: Validation is context-specific. A surrogate validated for one drug class (e.g., statins for LDL-cholesterol) may not be valid for another (e.g., fibrates) [4]. The populations, interventions, comparators, and outcomes in the validation studies should reflect the intended use case.

Quality of Life Assessment in Clinical Trials

Objective: To rigorously evaluate whether a treatment that shows benefit on surrogate endpoints also improves how patients feel and function in their daily lives.

Experimental Workflow:

• Instrument Selection: Choose validated, patient-reported outcome (PRO) measures specific to the disease and treatment context. Generic instruments like EQ-5D can be supplemented with disease-specific modules.
• Study Design: Integrate QOL assessment at predefined, protocol-specified timepoints throughout the trial, not merely as an exploratory endpoint.
• Baseline Adjustment: Account for baseline QOL scores in statistical analyses, as patients present with different symptom burdens that can influence results. Notably, 82% of trials fail to make this adjustment [24].
• Statistical Analysis: Pre-specify QOL analysis in the statistical analysis plan, including methods for handling missing data, which is common in progressive diseases.
• Reporting Transparency: Publish global QOL results regardless of outcome. While 61% of trials collect QOL data, only 34% actually publish these results [24].

Case Studies: When Surrogates Mislead

The BELLINI Trial: A Cautionary Tale in Multiple Myeloma

The BELLINI Phase III clinical trial provides a stark example of surrogate endpoint failure. This trial evaluated adding venetoclax to standard treatment for patients with advanced multiple myeloma. Based on conventional surrogate endpoints, the results appeared promising: patients receiving venetoclax showed significantly improved treatment response rates, higher minimal residual disease (MRD) negativity, and longer progression-free survival (PFS) compared to the placebo group [51].

However, the ultimate clinical outcome revealed a disturbing contradiction. At an interim analysis, investigators discovered significantly more deaths in the venetoclax arm than in the placebo arm [51]. This finding led the FDA to suspend further enrollment in the clinical trial. The case underscores a critical limitation—while MRD and PFS suggested clinical benefit, they failed to capture the complete risk-benefit profile, ultimately failing as predictors of overall survival in this context.

The GFR Slope in Chronic Kidney Disease: A Validation Success

In contrast to the BELLINI example, the glomerular filtration rate (GFR) slope in chronic kidney disease (CKD) represents a rare example of a rigorously validated surrogate endpoint. The GFR slope, which measures the rate of decline in kidney function over time, has demonstrated a remarkably strong trial-level association (R² trial of 97%) with patient-relevant outcomes including kidney failure requiring dialysis or transplantation [4].

This robust validation, encompassing biological plausibility, individual-level association, and trial-level surrogacy, led both the FDA and European Medicines Agency (EMA) to approve GFR slope as an acceptable primary endpoint for CKD therapy trials [4]. The case highlights that when properly validated with comprehensive evidence across all three levels of the Ciani framework, surrogate endpoints can reliably predict clinical benefit and accelerate drug development without compromising patient welfare.

The Scientist's Toolkit: Essential Research Reagents & Solutions

Table 3: Essential Resources for Endpoint Research & Validation

Tool/Resource	Function & Application	Specific Examples
FDA Surrogate Endpoint Table	Reference of endpoints acceptable for regulatory submissions; guides trial design	Adult Surrogate Endpoint Table (100+ endpoints) [4]
Meta-Analytic Software	Statistical analysis for surrogate validation (Level 1 evidence)	R packages, Bayesian methods, multivariate models [4]
Validated PRO Instruments	Measure patient-reported quality of life and functional status	EORTC QLQ-C30 (cancer), SF-36 (generic health) [24]
Clinical Trial Registries	Access to trial protocols and results for surrogate validation	ICTRP, ClinicalTrials.gov, EU Clinical Trials Register [52]
Reporting Guidelines	Ensure comprehensive reporting of surrogate evaluation methods	ReSEEM guidelines [4]

The evidence compiled in this guide reveals a landscape where surrogate endpoints, despite their widespread adoption and undeniable utility in accelerating drug development, frequently fail to predict the outcomes that matter most to patients: longer survival and better quality of life. The 6% success rate in achieving both these goals simultaneously represents a concerning predictive gap that demands methodological refinement and heightened regulatory scrutiny [24].

Future progress requires a multi-faceted approach. First, surrogate endpoint validation must be strengthened through rigorous application of the Ciani framework, with particular emphasis on establishing robust Level 1 evidence across appropriate therapeutic contexts [4]. Second, trial designs must prioritize patient-centered outcomes by systematically integrating methodologically rigorous quality of life assessments with appropriate baseline adjustments and transparent reporting [24]. Finally, as noted in the 2025 FDA-AACR Workshop, the field must embrace an iterative, collaborative process for developing novel endpoints that balance the need for speed with the imperative of ensuring genuine patient benefit [51].

The scientific community stands at a crossroads where it must reconcile efficiency with efficacy. By acknowledging the predictive gap and implementing more rigorous standards for surrogate endpoint validation and application, researchers can ensure that trial success translates into meaningful improvements in patient care rather than merely statistical victories.

In the pursuit of accelerating drug development, surrogate endpoints—such as progression-free survival (PFS) or overall response rate (ORR)—are increasingly used as substitutes for direct measurements of patient clinical benefit, namely overall survival (OS) and quality of life (QoL). However, a growing body of recent evidence from oncology reveals a concerning and frequent disconnect between success in these surrogate markers and meaningful improvements in how patients feel or how long they live. This analysis synthesizes the latest quantitative data, explores the mechanisms behind this discordance, and details the experimental methodologies driving this critical field of research.

Quantitative Analysis of the Correlation Between Surrogate and Clinical Endpoints

Recent studies and reviews consistently demonstrate that surrogate endpoints often fail to reliably predict improvements in OS or QoL.

Table 1: Correlation Between Surrogate Endpoints and Overall Survival in Recent Analyses

Analysis Focus	Surrogate Endpoint(s)	Correlation with OS	Key Findings	Source/Context
FDA Analysis of 15 Surrogate Endpoints	Various (e.g., PFS, EFS)	Low	Only 1 of 15 analyses showed a strong correlation with OS.	[31]
Cancer Drug Approvals (2018)	PFS, ORR	Very Low	85% of approvals were based on surrogate endpoints, but only 7% demonstrated an OS improvement in 2017.	[31]
Accelerated Approvals (5+ Year Follow-up)	Various Unvalidated Endpoints	Low	57% of drugs (26 of 46) did not show improved OS or QoL within 5 years.	[31]
Extensive-Stage Small-Cell Lung Cancer (1st Line)	Progression-Free Survival (PFS)	Strong (r=0.77)	PFS demonstrated strong clinical value as a surrogate for OS in this specific setting.	[53]
Extensive-Stage Small-Cell Lung Cancer (1st Line)	Overall Response Rate (ORR), Disease Control Rate (DCR)	Not Significant	ORR and DCR did not correlate with OS, underscoring they are unreliable predictors of long-term outcomes.	[53]
Confirmatory Trials for Accelerated Approval	PFS, ORR	Low	When converted to regular approval, only 32% of trials showed an OS benefit, and 12% showed a QoL benefit.	[31]

The data reveal a clinical trial landscape heavily reliant on surrogate endpoints. A study of 153 cancer drug approvals showed that the percentage of drugs approved based on surrogates rose from 57% in 2016 to 85% in 2018 [31]. Conversely, the percentage of approved drugs that actually improved OS decreased to a low of 7% in 2017 [31]. This trend creates a paradox where drugs are reaching the market based on biomarker or imaging changes, without confirmed evidence that they help patients live longer or better lives.

Mechanisms and Contexts for Endpoint Discordance

The failure of a surrogate endpoint to predict clinical benefit is not random; it occurs through specific, identifiable mechanisms. Understanding these contexts is crucial for interpreting trial data.

Table 2: Mechanisms Behind Discordance Between Surrogate Endpoints and Clinical Benefit

Mechanism of Discordance	Description	Illustrative Example
Effect of Subsequent Therapies	Effective later-line treatments can mask the true OS benefit of a first-line therapy, diluting the observed survival difference. This makes PFS, which is less influenced by subsequent lines, a more robust signal for the initial drug's activity [54].	In genitourinary cancers where OS often exceeds 36 months due to multiple effective treatments, PFS may be a more timely endpoint, though its value is debated when not associated with QoL [54].
Toxicity Outweighing Benefit	A drug may slow disease progression but cause significant adverse events that ultimately shorten survival or severely degrade QoL.	In the BELLINI trial for multiple myeloma, adding venetoclax to standard therapy improved PFS but led to worse OS, partly due to fatal infection-related toxicity in a patient subgroup [55].
Tumor Heterogeneity & Molecular Subgroups	A treatment may be highly effective in a molecularly defined patient subgroup but ineffective or harmful in others. Analysis of the full trial population can obscure this heterogeneity.	The BELLINI trial showed a PFS benefit with venetoclax in the overall population, but OS was worse. Subsequent analysis revealed that patients without the t(11;14) genetic abnormality drove the negative OS result [55].
Lack of Validation for Novel Mechanisms	Surrogate endpoints validated for one drug class (e.g., chemotherapy) may not hold for another (e.g., immunotherapy or targeted therapy), as their mechanisms of action differ [51].	Disease-free survival was once validated for chemotherapy in adjuvant colon cancer, but this relationship may not hold for modern molecularly targeted therapies and immunotherapies [31].
Ascertainment Bias	Differences in how and when a surrogate is measured between study arms can artificially create or mask an effect.	In prostate cancer trials, metastasis-free survival (MFS) can be influenced by the intensity of imaging; more frequent scanning in the control group can shorten the time to detection, introducing bias [54].

The relationship between surrogate and clinical endpoints is not static. As highlighted in a recent FDA-AACR workshop, a surrogate endpoint validated in one context (e.g., for a specific drug mechanism, patient population, and standard of care) may not be appropriate in another [51]. This necessitates continuous re-evaluation and validation of these markers.

Experimental Protocols for Validating Surrogate Endpoints

The gold standard methodology for establishing the validity of a surrogate endpoint is the meta-analysis of multiple, patient-level randomized controlled trials (RCTs). The protocol below, derived from a 2025 study on small-cell lung cancer, exemplifies this rigorous approach [53].

Protocol: Meta-Analysis of Surrogate Endpoints for OS

Objective: To quantitatively evaluate the strength of correlation between candidate surrogate endpoints (PFS, ORR, DCR) and overall survival (OS) within a specific cancer and treatment setting.

Primary Endpoints: Hazard ratios (HRs) for PFS and OS; odds ratios (ORs) for ORR and DCR.

Methodology:

• Literature Search & Study Selection (PRISMA Guidelines):

  • Data Sources: Systematic search of electronic databases (e.g., MEDLINE, Embase, Cochrane Central, ClinicalTrials.gov).
  • Time Frame: Pre-defined period (e.g., 2014-2024).
  • Inclusion Criteria: Phase III RCTs in the target disease (e.g., extensive-stage small-cell lung cancer) that report data on both the surrogate endpoints and OS.
  • Exclusion Criteria: Non-randomized studies, phase I/II trials, and studies without extractable data for the endpoints of interest.

• Data Extraction:

  • Data is manually extracted from each eligible trial into a standardized form.
  • Key Data Points:
    • Trial identifiers and publication year.
    • Patient population and line of therapy (e.g., first-line, second-line).
    • Treatment and control regimens.
    • Hazard Ratios (HRs) for PFS and OS, with confidence intervals.
    • Odds Ratios (ORs) for ORR and DCR.

• Statistical Analysis:

  • Correlation Analysis: Weighted linear regression and Pearson correlation analyses are performed using the logarithms of the effect measures (HRs and ORs).
  • Correlation Coefficient (r): The correlation coefficient r is computed to evaluate the strength of the relationship between the treatment effect on the surrogate endpoint and the treatment effect on OS.
  • Interpretation: A correlation coefficient r close to 1.0 indicates a strong surrogate relationship, while a value close to 0 suggests no correlation. Statistical significance is typically set at P < 0.05.

This methodology was successfully applied in a 2025 meta-analysis of 23 phase III trials in SCLC, which conclusively demonstrated a strong correlation between PFS and OS (r = 0.77, P < 0.001) in the first-line setting, but no significant correlation for ORR or DCR [53].

Experimental Workflow for Endpoint Validation

The following diagram illustrates the multi-stage pathway from initial biomarker identification to its acceptance as a validated surrogate endpoint, highlighting the iterative nature of this process and the role of regulatory science.

The Scientist's Toolkit: Research Reagent Solutions

Advancing the field of endpoint validation requires a suite of specialized tools and methodologies. The following table details key resources for researchers in this domain.

Table 3: Essential Research Tools for Endpoint and Biomarker Analysis

Tool / Resource	Function & Application	Key Features & Notes
FDA Surrogate Endpoint Table	A regulatory reference listing surrogate endpoints that have supported drug approvals, facilitating discussion between sponsors and the FDA [10].	Mandated by the 21st Century Cures Act. Endpoints are context-dependent and not a substitute for direct FDA consultation [10].
Circulating Tumor DNA (ctDNA)	A liquid biopsy biomarker used to detect minimal residual disease (MRD) and predict relapse much earlier than radiographic imaging [51].	Emerging as a potential surrogate endpoint; evidence is being gathered to validate its correlation with PFS and OS in various cancers [51].
Minimal Residual Disease (MRD)	A highly sensitive measure of the small number of cancer cells remaining after treatment, typically assessed via flow cytometry or DNA sequencing in hematologic malignancies [55].	Recently supported by the FDA as an accelerated approval endpoint in multiple myeloma, with confirmation of PFS/OS required [55].
Biomarker Qualification Program (BQP)	An FDA pathway for qualifying biomarkers for a specific "context of use," making them available for any drug developer without needing re-validation [56].	The program has been slow, qualifying only 8 biomarkers since its inception, highlighting the complexity of robust biomarker development [56].
Patient-Level Data Meta-Analysis	The definitive statistical method for validating a surrogate endpoint by correlating treatment effects on the surrogate with effects on OS across multiple RCTs [53] [51].	Requires access to data from numerous trials. Considered the highest level of evidence for establishing a surrogate relationship [53].

The compelling body of recent evidence underscores a critical challenge in modern drug development: the rapid approval of therapies based on surrogate endpoints does not guarantee that these treatments will deliver the ultimate goals of prolonged survival or improved quality of life. While surrogates like PFS are indispensable for accelerating drug development, their application requires rigorous, context-specific validation. The future of reliable and ethical drug development depends on a balanced approach that leverages the speed of surrogate endpoints while investing in the robust, long-term evidence generation needed to confirm their true clinical value for patients.

The Accelerated Approval Program, established by the U.S. Food and Drug Administration (FDA), serves as a critical regulatory mechanism for expediting the availability of drugs that treat serious conditions and address an unmet medical need [57]. The cornerstone of this pathway is the use of surrogate endpoints, which are markers—such as laboratory measurements, radiographic images, or physical signs—that are considered reasonably likely to predict clinical benefit, but are not themselves a direct measurement of that benefit [57]. This approach allows for drug approval based on earlier, more readily measurable outcomes, potentially shortening clinical trial duration and bringing promising therapies to patients years sooner [58]. However, this regulatory flexibility is granted with the explicit requirement that sponsors must conduct post-approval confirmatory trials to verify the anticipated clinical benefit [59]. The fundamental challenge lies in the fact that a significant proportion of these required confirmatory studies are delayed, or in some cases, fail to conclusively demonstrate the clinical benefit they were designed to confirm [58] [24]. This creates a tension between providing rapid access to potentially life-saving treatments and ensuring that patients are not exposed for prolonged periods to drugs with unverified efficacy or safety profiles.

The Current Landscape of Surrogate Endpoints and Confirmatory Trial Delays

Quantitative Evidence of the Confirmatory Trial Problem

The scale of the challenge is substantial. Historically, delays in completing confirmatory trials have been a significant issue. A 2021 analysis revealed that a staggering 38 percent of all accelerated drug approvals (104 out of 278) still had pending completion and review of their confirmatory trials [58]. Furthermore, among these outstanding trials, 34 percent extended past their originally planned completion date [58]. This has led to situations where drugs remain on the market for years with unconfirmed clinical benefit, often referred to as "dangling" approvals [58]. A specific analysis of non-oncology drug indications approved through the pathway between 1992 and 2018 found that approximately 20% of confirmatory trials failed to meet FDA requirements, leaving clinical efficacy unconfirmed in many cases [59].

The Clinical Benefit Verification Gap

The ultimate test of a surrogate endpoint's validity is whether it reliably predicts improvements in how patients feel or function, or whether they live longer. A 2025 study presented at the American Society of Clinical Oncology (ASCO) Annual Meeting offers a sobering perspective on this issue [24]. This research, which assessed 791 randomized controlled trials published between 2002 and 2024, found that while alternative (surrogate) endpoints were the most common primary endpoints (63%), and 55% of these trials were deemed "positive" based on these surrogates, the translation to tangible patient benefit was weak.

Table 1: Translation of "Positive" Surrogate Endpoint Trials to Patient Benefit

Outcome Measure	Percentage of RCTs Showing Improvement	Notes
Alternative (Surrogate) Endpoint Superiority	55%	Basis for initial "positive" trial result
Overall Survival (OS) Improvement	28%	Only about half of "positive" trials showed a survival benefit
Quality of Life (QOL) Improvement	11%	Informed by patient-reported outcomes
Both OS and QOL Improvement	6%	Minimal overlap in survival and quality of life benefits

This data underscores a critical disconnect: the majority of trials that are successful based on surrogate markers do not ultimately demonstrate that patients live longer or better lives [24]. The reasons for the increased use of surrogate endpoints are understandable—they require fewer patients and resources, and results are available faster, which is incredibly important in fields like oncology. However, as Dr. Alexander Dean Sherry, lead author of the ASCO study, notes, this raises a fundamental question: "What are the true benefits of these new treatments, which are often more expensive and potentially even more toxic?" [24].

Methodologies for Validating Surrogate Endpoints

To address these challenges, the scientific community has developed rigorous statistical methodologies for evaluating the strength of surrogate endpoints. The "gold standard" for surrogacy validation involves modeling individual patient data from multiple randomized controlled trials (RCTs) [15]. The most common approach is to estimate a trial-level coefficient of determination (R²), which quantifies how much of the variation in the treatment effect on the true clinical endpoint (e.g., overall survival) is explained by the variation in the treatment effect on the surrogate endpoint (e.g., progression-free survival) [15].

A Novel Meta-Analytic Model for Time-to-Event Endpoints

Many current validation methods rely on the hazard ratio under the assumption of proportional hazards, which can be problematic as treatment effects often vary over time. A novel two-stage meta-analytic model proposed in 2025 addresses several key limitations [15]:

• Use of Restricted Mean Survival Time (RMST): This model employs differences in RMST, rather than hazard ratios, to quantify treatment effects. The RMST represents the average event-free survival time up to a pre-specified time point and remains valid even when the proportional hazards assumption is violated, which is common in cancer trials [15].
• Assessment at Multiple Timepoints: The framework allows for the evaluation of the strength of surrogacy (the R²) at multiple time points, capturing how the relationship between the surrogate and the true endpoint may change over the course of the disease [15].
• Explicit Modeling of Time Lag: A significant innovation is the model's ability to account for a time lag between the surrogate and the final clinical outcome. This allows researchers to evaluate whether a treatment effect on a surrogate measured over a shorter time frame (e.g., 1-year progression-free survival) reliably predicts a later effect on the clinical endpoint (e.g., 3-year overall survival), directly informing the feasibility of reducing trial duration [15].

The following diagram illustrates the workflow of this advanced surrogate endpoint validation methodology.

Diagram 1: Workflow for advanced surrogate endpoint validation. This two-stage meta-analytic model uses Restricted Mean Survival Time (RMST) to validate surrogate endpoints (S) against true clinical endpoints (T) across multiple randomized controlled trials (RCTs), accounting for non-proportional hazards and time lag.

Comparing Surrogacy Models and the Scientist's Toolkit

Choosing the appropriate statistical model is crucial for robust validation. A 2025 comparative study evaluated the performance of six different surrogacy models, including weighted linear regression, meta-regression, and Bayesian Bivariate Random-Effects Meta-Analysis (BRMA) [42]. The study found that while weighted linear regression provides a useful reference, Bayesian BRMA generally provided the most robust predictions, especially for smaller data sets, though it required informative priors for the heterogeneity distribution [42]. The predictions from different models showed a high degree of variation when the surrogate association was only moderate, highlighting the importance of not relying on a single method.

Table 2: Essential Research Reagents and Tools for Surrogate Endpoint Validation

Tool / Reagent	Function in Validation	Key Consideration
Individual Patient Data (IPD)	Raw data from multiple RCTs required for gold-standard meta-analytic validation.	Must include time-to-event data, treatment assignment, and censoring indicators.
Statistical Software (e.g., R, SAS)	Platform for implementing complex two-stage models, copulas, and BRMA.	Requires specialized packages for survival analysis and advanced meta-analysis.
Pseudo-Observation Algorithm	Technique to handle censored data in the RMST-based validation model.	Replaces censored outcomes with contributions to the RMST estimate.
Bayesian Priors	Incorporation of existing knowledge about heterogeneity in BRMA.	Critical for obtaining stable results from smaller meta-analyses.
Pre-Specified Analysis Plan	Protocol defining timepoints, statistical models, and criteria for validity.	Essential to avoid data dredging and ensure reproducible research.

Recent Regulatory Reforms and Their Implications

In response to the issues with delayed confirmatory trials, Congress passed the Food and Drug Omnibus Reform Act (FDORA) in 2022, granting the FDA enhanced authority to enforce timely completion of these studies [58] [60] [61]. In late 2024 and early 2025, the FDA issued two draft guidance documents to implement these new authorities, significantly reshaping the regulatory landscape for drug developers [58] [59] [60].

The New Requirement: "Underway" Confirmatory Trials

A central change is the FDA's newfound authority to require that confirmatory trials be "underway" prior to granting an Accelerated Approval [60] [61]. The January 2025 draft guidance clarifies that FDA "generally intends to consider a confirmatory trial to be 'underway'" only if it meets three key criteria, detailed in the table below [61].

Table 3: FDA Criteria for a Confirmatory Trial to be "Underway" (per Jan 2025 Draft Guidance)

Criterion	Detailed FDA Expectations	Sponsor Considerations
1. Target Completion Date	Date must be consistent with diligent conduct, considering disease natural history, alternative treatments, and recruitment timeline.	Sponsors must provide a "clear and sound justification" for the proposed timeline [61].
2. Sponsor Progress & Plans	Plans must provide "sufficient assurance" of timely completion. FDA will review accrual rates and site activation pace.	Sponsors should propose measurable benchmarks (e.g., recruitment goals, site activity) [61].
3. Initiation of Enrollment	Enrollment must have begun. The FDA does not specify a minimum number, but this is a "minimum expectation."	In cases of anticipated enrollment challenges (e.g., after market availability), FDA may require completed enrollment prior to approval [60].

The guidance emphasizes that if a confirmatory trial is required to be underway and is not, the FDA "does not intend to grant Accelerated Approval until this deficiency is addressed" [60]. This represents a significant shift from prior practice and is intended to address the historical problem of patient recruitment plummeting once a drug becomes commercially available [58].

Expedited Withdrawal Procedures and Heightened Scrutiny

The December 2024 draft guidance, "Expedited Program for Serious Conditions — Accelerated Approval of Drugs and Biologics," clarifies the FDA's procedures for withdrawing approvals when confirmatory trials fail to verify clinical benefit or are not conducted with due diligence [58] [59]. While the process still involves formal notice and sponsor appeal rights, the FDA's intent to act more swiftly is clear. The guidance also highlights heightened oversight of promotional materials, requiring that claims be aligned with the verified benefit and that materials may be subject to FDA review before dissemination [59]. The following diagram summarizes the reformed Accelerated Approval pathway under these new guidances.

Diagram 2: The reformed Accelerated Approval pathway. Recent FDA guidance mandates confirmatory trials be "underway" at approval and outlines expedited procedures for withdrawal if clinical benefit is not verified.

The landscape of FDA's Accelerated Approval pathway is undergoing a significant transformation. The core challenge remains balancing the imperative for rapid therapeutic development against the ethical and scientific necessity of confirming real patient benefit. Recent evidence suggests that an over-reliance on surrogate endpoints, without rigorous and timely confirmation, risks populating the market with drugs that offer uncertain value to patients [24]. The methodological advances in surrogate endpoint validation, such as RMST-based models that account for time lags and non-proportional hazards, provide more robust tools for assessing the strength of these markers [15]. Concurrently, the new regulatory framework established by FDORA and the 2025 draft guidances creates a stricter environment, demanding greater upfront commitment from sponsors to ensure confirmatory trials are feasible, timely, and diligently executed [58] [61]. For researchers and drug developers, success in this new era will require early and strategic engagement with the FDA, the adoption of methodologically sound surrogate validation techniques, and a unwavering focus on designing confirmatory trials that can definitively answer the question of whether a drug ultimately helps patients live longer or better lives.

The analysis of drug pricing is a critical component of medical product development, intersecting significantly with the evaluation of clinical evidence. The choice between surrogate endpoints and clinical endpoints in research not only influences regulatory approval but also fundamentally shapes subsequent pricing and reimbursement decisions. This guide provides an objective comparison of drug pricing landscapes across major markets, examining how different evidentiary standards contribute to the economic uncertainties facing researchers, scientists, and drug development professionals. The complex interplay between demonstrated therapeutic value, research methodology, and market access creates a challenging environment where pricing models must reconcile innovation incentives with affordability constraints.

Understanding international prescription drug price differentials provides crucial insights for strategic planning in drug development. Recent analyses reveal that U.S. manufacturer gross prices for all prescription drugs were 278% of prices in 33 OECD comparison countries combined in 2022, meaning prices in other countries were approximately one-third of U.S. prices [62] [63]. This disparity stems from dramatically different pricing structures for various drug categories, with U.S. prices for brand-name originator drugs reaching 422% of prices in comparison countries, while U.S. unbranded generics were actually cheaper at 67% of international prices [62]. These differentials create distinct market dynamics that influence how research outcomes translate into commercial success across different regions.

Quantitative Comparison of International Drug Pricing

Comprehensive Price Analysis Across Drug Categories

Table 1: International Prescription Drug Price Comparisons (2022 Data)

Country/Region	All Drugs Price Relative to U.S.	Brand-Name Originator Drugs Price Relative to U.S.	Unbranded Generics Price Relative to U.S.	Data Source/Year
United States (Baseline)	100%	100%	100%	RAND Corporation 2024
OECD Average (33 countries)	36%	24%	149%	RAND Corporation 2024
Canada	44%	N/A	N/A	RAND Corporation 2024
Mexico	58%	N/A	N/A	RAND Corporation 2024
Turkey	10%	N/A	N/A	RAND Corporation 2024

Table 2: Market Share and Spending Patterns by Drug Type (2022)

Drug Category	U.S. Prescription Volume Share	U.S. Spending Share (Gross Manufacturer)	Comparison Countries Volume Share	Comparison Countries Spending Share
Brand-name originator drugs	7%	87%	29%	74%
Unbranded generics	90%	8%	41%	13%

The tabular data reveals several critical patterns in global drug pricing. The United States demonstrates exceptional market conditions where brand-name originator drugs command premium prices while generics are remarkably inexpensive compared to other markets [63]. This bifurcated market structure has significant implications for how different types of therapeutic products are valued internationally. The volume distribution further highlights this dichotomy, with unbranded generics accounting for 90% of U.S. prescription volume but only 8% of spending, while brand-name originator drugs represent just 7% of volume but 87% of spending [63]. These patterns underscore how different evidentiary standards for drug approval—whether based on surrogate versus clinical endpoints—can dramatically influence economic returns across various markets.

Experimental Protocol for International Price Comparisons

The methodological framework for conducting international drug price comparisons follows rigorous standards to ensure valid cross-market assessments. The RAND study utilized 2022 IQVIA MIDAS data to calculate price indexes comparing prescription drug prices in the United States with those in 33 OECD comparison countries [62]. The experimental protocol included several critical components:

• Data Collection: Presentation-level data (separate records for each combination of active ingredient, formulation, and dosage strength) for all prescription drugs in the IQVIA MIDAS dataset [62].

• Bilateral Comparison Framework: Individual comparisons between the U.S. and each OECD country using overlapping drug presentations between markets [62].

• Weighting Methodology: Application of U.S. volume weights (share of total volume accounted for by each presentation) to calculate price indexes, reflecting U.S. policy perspectives [62].

• Data Quality Controls: Exclusion of outlier presentations with very low volume/sales or extreme price ratios to prevent undue influence on overall results [62].

• Market Basket Definition: Analysis comparing U.S. prices with all comparison countries combined utilized data from 4,690 presentations and 1,646 active ingredients [62].

This methodological approach enables consistent comparison of manufacturer gross prices—the starting point before negotiated rebates and discounts—which is particularly relevant for understanding how initial pricing decisions based on clinical evidence translate across different markets.

Market Dynamics and Stakeholder Influence

Drug Pricing Ecosystem Stakeholder Relationships

The pharmaceutical pricing ecosystem involves multiple stakeholders with competing interests and influence. Manufacturers set initial list prices based on R&D costs, production expenses, and therapeutic value propositions [64]. Pharmacy Benefit Managers (PBMs) and insurers function as powerful intermediaries that negotiate rebates and discounts with manufacturers while managing formularies that determine patient access [64]. The three largest PBMs—Express Scripts, CVS Caremark, and OptumRx—processed approximately 79% of all prescription drugs in 2022, serving about 290 million Americans [64], demonstrating their substantial market power. Government agencies influence pricing through regulatory frameworks, reimbursement policies, and increasingly through direct negotiation, as evidenced by the Inflation Reduction Act of 2022 which empowers Medicare to negotiate prices for certain high-cost drugs [64].

The evidentiary standards required for favorable pricing and reimbursement decisions increasingly depend on the type of endpoints used in clinical research. Drugs demonstrating value through clinical endpoints that measure how patients feel, function, or survive typically command higher premium pricing than those relying solely on surrogate endpoints [64]. This creates strategic decisions for drug development professionals about research investment, as the choice between surrogate versus clinical endpoints involves trade-offs between development speed, regulatory approval probability, and ultimate market pricing potential.

Regulatory Frameworks and Historical Evolution

Comparative Analysis of International Regulatory Approaches

Table 3: International Drug Pricing Regulatory Mechanisms

Country/Region	Primary Pricing Mechanism	Key Features	Impact on Prices
United States	Market-based with negotiated discounts	Medicare Part D negotiation ban (2003-2022), Inflation Reduction Act authorizes limited negotiation	Highest prices among OECD countries
Germany	Reference pricing & AMNOG process	Benefit evaluation of new drugs, reference pricing for older drugs	Moderate prices, ~35-40% of U.S. levels
France	Direct price controls & therapeutic assessment	Prices set based on added therapeutic value, no post-launch increases	Low prices, among lowest in OECD
United Kingdom	Profit controls & expenditure caps	Voluntary/statutory schemes regulating profits, spending caps	Moderate to low prices
Japan	Price premium system & biennial cuts	Premiums for innovation, mandatory biennial price revisions	Low prices

International approaches to drug price regulation reflect different societal balances between innovation incentives and access imperatives. European countries typically employ more direct intervention mechanisms, including product price controls (France, Italy, Portugal, Spain), reference pricing (Germany, Netherlands), and profit controls (UK) [64]. Germany's AMNOG process, implemented in 2011, requires benefit evaluation of new drugs followed by price negotiations, with older drugs often subject to reference pricing [64]. France prohibits price increases after launch and systematically reduces prices on older drugs to fund newer ones [64]. These differential regulatory approaches create a complex global landscape where the evidentiary standards required for favorable pricing vary significantly, influencing how drug developers approach research design and endpoint selection across different target markets.

Research Reagent Solutions for Economic Analysis

Table 4: Essential Methodological Tools for Drug Pricing Research

Research Tool	Function	Application Context
IQVIA MIDAS Database	Provides sales and volume estimates projected from audits of standardized list prices and manufacturer invoices	International price comparison studies, market trend analysis
Price Index Methodology	Enables standardized comparison of drug prices across formulations, strengths, and markets	Bilateral price comparisons, tracking price changes over time
Gross-to-Net Adjustment Models	Estimate actual manufacturer revenues after rebates and discounts	Net price analysis, understanding realized manufacturer returns
Volume-Weighted Averaging	Accounts for differences in drug utilization patterns across markets	Cross-national price comparisons that reflect actual consumption
Therapeutic Assessment Frameworks	Evaluate added clinical value of new drugs compared to existing treatments	Price premium justification, value-based pricing decisions

The methodological tools for conducting drug pricing research require sophisticated approaches to handle complex market data. The IQVIA MIDAS Database provides fundamental data infrastructure for international comparisons, containing sales and volume estimates projected from comprehensive audits [62]. Price index methodologies enable standardized comparisons, with researchers using U.S. volume weights to calculate indexes that reflect differences from U.S. policy perspectives [62]. Gross-to-net adjustment models attempt to address the significant limitation that list prices don't reflect manufacturer realized prices, though these adjustments introduce measurement uncertainty as rebates vary substantially across therapeutic classes and are confidential [62]. When the RAND study applied an adjustment reducing U.S. brand-name retail prices by 37.2% to approximate net prices, U.S. prices remained 308% of prices in other countries [62], indicating that rebates alone don't explain international differences.

Economic Foundations and Research Implications

Endpoint Selection Impact on Development and Pricing

The fundamental economics of pharmaceutical pricing revolve around several distinctive market characteristics. For many drugs addressing life-threatening conditions, demand is profoundly inelastic, as patients facing severe illnesses have limited alternatives and will pay premiums regardless of price [64]. This inelasticity grants manufacturers significant pricing power, particularly for drugs demonstrating unique clinical benefits [64]. Simultaneously, the immense R&D costs and high failure rates create economic pressure for high launch prices, with estimates ranging from $879.3 million to $2.3 billion per new drug, and only about 11% of candidates entering clinical trials ultimately succeeding [64]. These economic realities intersect with research design decisions, as drugs demonstrating value through clinical endpoints typically justify higher pricing but require more extensive and expensive trials, creating strategic trade-offs for development professionals.

The choice between surrogate endpoints and clinical endpoints represents a critical strategic decision with significant economic implications. Surrogate endpoints (e.g., biomarker reduction, tumor shrinkage) typically allow faster development cycles and accelerated approval pathways but may result in greater pricing and reimbursement uncertainty [64]. Clinical endpoints (e.g., survival, quality of life) provide more direct evidence of patient benefit but require longer, larger, and more expensive trials [64]. This creates a complex optimization problem for drug developers balancing speed to market against evidence strength needed for favorable pricing and reimbursement decisions across different health systems. The increasing emphasis on value-based pricing models in many markets further intensifies the importance of these endpoint decisions, as payers increasingly demand demonstrated clinical benefit rather than surrogate marker improvements alone.

In the pursuit of accelerating patient access to novel therapies, the use of surrogate endpoints has become increasingly prevalent in drug development. According to the U.S. Food and Drug Administration (FDA), a surrogate endpoint is "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure, that is not itself a direct measurement of clinical benefit," but is either known or reasonably likely to predict clinical benefit [10]. While these endpoints can significantly shorten drug development timelines, they introduce inherent uncertainty that must be mitigated through rigorous confirmatory trials and robust post-market surveillance systems.

This framework balances the need for timely access to promising therapies with the fundamental obligation to verify clinical benefits and ensure patient safety. The strategic integration of these mitigation approaches forms a critical safeguard in modern regulatory science, protecting public health while fostering innovation [10] [12]. This article examines the key strategies within this framework, providing researchers and drug development professionals with practical methodologies and comparative analyses to navigate the complexities of endpoint evaluation.

Surrogate versus Clinical Endpoints: A Comparative Analysis

Definition and Regulatory Context

Surrogate endpoints serve as substitutes for direct measurements of how a patient feels, functions, or survives. They are utilized in clinical trials when direct measurement of clinical benefit is impractical due to time constraints, cost, or feasibility [12]. The FDA categorizes surrogate endpoints into two distinct regulatory pathways: those "known to predict clinical benefit" for traditional approval, and those "reasonably likely to predict clinical benefit" for accelerated approval [10].

In contrast, clinical endpoints directly measure the therapeutic effect from the patient's perspective, including overall survival, reduction in disease-specific symptoms, or improved quality of life. These endpoints represent unambiguous evidence of treatment value but often require larger, longer, and more costly trials to demonstrate statistically significant effects [12].

Comparative Evaluation of Endpoint Types

The table below summarizes the fundamental characteristics, advantages, and limitations of both endpoint categories:

Table 1: Comparison of Surrogate and Clinical Endpoints

Characteristic	Surrogate Endpoints	Clinical Endpoints
Definition	Indirect measures of disease activity or response (e.g., lab values, imaging) [10]	Direct measurements of how a patient feels, functions, or survives [12]
Regulatory Use	Supports both traditional and accelerated approval pathways [10]	Required for traditional approval; gold standard for evidence [12]
Trial Duration	Shorter	Longer
Trial Size	Smaller	Larger
Cost	Lower	Higher
Risk of Uncertainty	Higher - requires validation [10]	Lower - directly measures benefit
Examples	Tumor shrinkage (oncology), HbA1c reduction (diabetes), LDL cholesterol (cardiology) [10]	Overall survival, symptom reduction, improved physical function

Methodologies for Confirmatory Trial Design and Analysis

The Role of Confirmatory Trials

Confirmatory trials serve as the critical bridge between initial approval based on surrogate markers and verified clinical benefit. For products receiving accelerated approval based on surrogate endpoints that are "reasonably likely" to predict clinical benefit, the FDA mandates post-marketing confirmatory trials to "verify and describe the anticipated effect on irreversible morbidity or mortality or other clinical benefit" [10]. These trials are not merely procedural hurdles but essential scientific validations that either confirm the predictive value of the surrogate endpoint or lead to regulatory actions, including product withdrawal.

Statistical Methods for Comparing Drug Efficacy

In the absence of direct head-to-head clinical trials, researchers employ several statistical methodologies to compare drug efficacies. The table below outlines key approaches:

Table 2: Statistical Methods for Comparative Drug Efficacy Analysis

Method	Description	Application & Acceptability
Head-to-Head Trials	Direct comparison within a single randomized controlled trial (RCT)	Gold standard; eliminates confounding variables but often costly and time-consuming [65]
Adjusted Indirect Comparison	Compares two interventions via a common comparator (e.g., both vs. placebo); preserves original randomization [65]	Accepted by many health technology assessment agencies (e.g., NICE, PBAC); mentioned in FDA guidelines [65]
Mixed Treatment Comparison (MTC)	Bayesian models incorporating all available data, including data not relevant to the direct comparator [65]	Reduces statistical uncertainty but not yet widely accepted by regulatory bodies [65]
Naïve Direct Comparison	Direct comparison of results from separate trials without statistical adjustment	Not recommended for conclusive evidence; breaks randomization and subjects results to significant confounding and bias [65]

Experimental Protocol for Adjusted Indirect Comparison

For researchers conducting indirect comparisons, the following protocol provides a methodological framework:

Protocol Title: Adjusted Indirect Comparison of Drug Efficacy Through Common Comparator

Objective: To compare the efficacy of Drug A versus Drug B using a common comparator (C) when no direct head-to-head trial exists.

Step 1: Trial Identification

• Identify two randomized controlled trials: Trial 1 (Drug A vs. Comparator C) and Trial 2 (Drug B vs. Comparator C).
• Ensure trials have similar patient populations, designs, and outcome definitions to minimize heterogeneity [65].

Step 2: Data Extraction

• For continuous outcomes (e.g., change in blood glucose): Extract mean difference and variance for each drug versus C.
• For binary outcomes (e.g., proportion of patients reaching HbA1c <7.0%): Extract relative risk or odds ratio and confidence intervals for each drug versus C [65].

Step 3: Statistical Analysis

• For continuous outcomes: Calculate the indirect comparison as (A vs. C) - (B vs. C). The variance of the estimate is the sum of the variances of the two direct comparisons [65].
• For binary outcomes: Calculate the indirect relative risk as (RRAvsC / RRBvsC). The confidence intervals are derived using statistical software that accounts for the compounded uncertainty [65].

Step 4: Interpretation

• A point estimate of 0 for continuous outcomes (or 1 for relative risk) indicates no difference between A and B.
• Acknowledge that the uncertainty (confidence interval) will be wider than in either direct comparison due to the summing of variances [65].

Post-Market Surveillance: Techniques and Systems for Risk Mitigation

The Imperative for Post-Market Surveillance

Post-marketing surveillance (PMS) represents a critical risk mitigation strategy that addresses the inherent limitations of pre-market clinical trials. These trials are typically conducted in controlled environments with limited patient populations (often <5,000 patients) and may exclude vulnerable groups such as pregnant women, children, or the elderly [66]. PMS systems monitor drug safety in real-world settings across diverse populations, detecting rare, long-term, or unexpected adverse drug reactions (ADRs) that occur at rates too low (e.g., 1 in 10,000) to be detected in pre-approval studies [67] [66].

Core Post-Marketing Surveillance Techniques

Multiple complementary techniques form a comprehensive PMS strategy:

Table 3: Post-Marketing Surveillance Techniques and Applications

Technique	Methodology	Primary Application
Spontaneous Reporting	Voluntary reports from healthcare professionals/patients to regulatory authorities (e.g., FDA's FAERS, UK's Yellow Card) [67] [66]	Early signal detection for unknown ADRs; most common PMS method [66]
Active Surveillance	Proactive monitoring through electronic health records (EHR), registries, and claims databases (e.g., FDA's Sentinel Initiative) [67]	Systematic assessment of ADR incidence in large populations
Phase IV Clinical Trials	Controlled studies conducted after approval	Confirm long-term safety, optimal usage, and effectiveness versus other treatments [67]
Registry Programs	Databases tracking patients with specific diseases or exposures	Generate real-world evidence on prescription patterns, off-label use, and outcomes in broad populations [67]
Data Mining & Analysis	Statistical techniques (e.g., disproportionality analysis, machine learning) applied to large safety databases [67]	Identify potential safety signals from massive datasets

PMS System Implementation: Yellow Card and FAERS

The Yellow Card Scheme (UK)

• History and Evolution: Established in 1964 in response to the thalidomide tragedy, initially for doctors and dentists. Expanded over decades to include pharmacists, nurses, coroners, and eventually patients [66].
• Key Features: Includes both mandatory reporting for "black triangle drugs" (newly licensed medicines under intensive monitoring) and voluntary reporting for serious suspected ADRs to established products [66].
• Reporting Criteria: Encourages reporting even when causality is doubtful, particularly for serious reactions including fatal, life-threatening, disabling, or congenital abnormalities [66].

FDA Adverse Event Reporting System (FAERS)

• Function: A database that contains adverse event and medication error reports submitted voluntarily by healthcare professionals and consumers [67].
• Process: FDA clinical reviewers evaluate and analyze reports to make recommendations on product safety [67].
• MedWatch: FDA's safety information and adverse event reporting program, which issues safety alerts for regulated products [67].

The following diagram illustrates the integrated workflow of post-market surveillance systems:

The Scientist's Toolkit: Essential Research Reagent Solutions

The following reagents and methodologies are fundamental for conducting research in endpoint validation and pharmacovigilance:

Table 4: Essential Research Reagents and Resources for Endpoint and Safety Research

Resource/Solution	Function/Application	Relevance to Research
FDA Surrogate Endpoint Table	Comprehensive list of endpoints used as basis for drug approval [10]	Reference for designing clinical trials; understanding accepted surrogate endpoints across therapeutic areas
Pharmacovigilance Databases (e.g., FAERS, VigiBase)	Databases of spontaneous adverse event reports [67]	Data source for identifying potential safety signals; understanding ADR profiles
Real-World Data Sources (e.g., EHRs, claims data, registries)	Longitudinal patient data from routine clinical care [68]	Generating real-world evidence on drug effectiveness and safety in diverse populations
Statistical Software for Indirect Comparisons	Specialized programs for network meta-analysis and indirect treatment comparisons [65]	Conducting adjusted indirect comparisons when head-to-head trials are unavailable
Data Mining Tools (e.g., machine learning algorithms)	Advanced analytics for large datasets [67]	Identifying patterns and signals in complex pharmacovigilance data

The strategic use of surrogate endpoints, when coupled with rigorous confirmatory trials and comprehensive post-market surveillance, creates a robust framework for balancing innovation with patient safety in drug development. Confirmatory trials provide the scientific validation necessary to verify that early signals of efficacy translate to meaningful clinical benefits, while post-market surveillance systems offer continuous monitoring in real-world settings where pre-market trials cannot reach. For researchers and drug development professionals, understanding the methodologies for indirect comparison, the operational mechanisms of surveillance systems, and the regulatory expectations for evidence generation is paramount. This integrated approach ensures that the pursuit of accelerated development pathways does not compromise the fundamental commitment to demonstrating genuine patient benefit and maintaining ongoing safety vigilance throughout a product's lifecycle.

A Critical Weighing: Comparing the Utility of Endpoints Across Therapeutic Areas

The assessment of new cancer therapies relies heavily on the use of clinical endpoints to determine treatment efficacy. Among these, Overall Survival (OS) has long been regarded as the gold standard endpoint in oncology clinical trials [30]. OS is defined as the time from randomization until death from any cause, providing a patient-centered, objective, and clinically meaningful measure of treatment benefit [30]. However, the practical challenges associated with OS measurement—including the need for large patient populations, extended follow-up periods, and substantial financial resources—have driven the exploration and adoption of surrogate endpoints [30].

The recent FDA draft guidance from August 2025, "Approaches to the Assessment of Overall Survival in Oncology Clinical Trials," marks a significant evolution in regulatory thinking, underscoring that "Overall survival is both an efficacy and a safety endpoint; it can be favorably impacted by the therapeutic benefits of a specific drug and negatively impacted by the drug's toxicity" [69]. This reaffirmation of OS's central role comes amidst growing sophistication in the use of surrogate endpoints that now support both accelerated and traditional approval pathways across numerous disease areas [10].

Defining the Endpoints: Characteristics and Applications

Overall Survival (OS) represents the most unambiguous and clinically relevant endpoint in oncology trials. Its strength lies in its objective nature—it measures survival time from randomization to death from any cause, with patients still alive at the time of analysis being censored [30]. This endpoint is definitive, easily measured, and not subject to interpretation bias, making it the preferred benchmark against which all other endpoints are evaluated [30].

However, OS has significant limitations that become particularly challenging in modern oncology drug development. The requirement for long-term follow-up means that trials take years to complete, especially in diseases with prolonged survival, and may require larger patient populations [30]. Additionally, OS can be confounded by subsequent therapies and crossover treatment effects, making it difficult to attribute survival benefits specifically to the investigational therapy being studied [30].

Common Surrogate Endpoints in Oncology

Surrogate endpoints are biomarkers or intermediate endpoints intended to substitute for a clinical endpoint, predicting clinical benefit based on epidemiological, therapeutic, pathophysiological, or other scientific evidence [10]. The FDA recognizes that these endpoints can "expedite the completion of clinical trials, resulting in earlier approval and enabling patients earlier access to new drugs" [12]. The table below summarizes key surrogate endpoints used in oncology and their relationship to OS.

Table 1: Key Surrogate and Clinical Endpoints in Oncology Trials

Endpoint	Definition	Relationship to OS	Advantages	Limitations
Overall Survival (OS)	Time from randomization to death from any cause [30]	Gold standard; direct measure of clinical benefit	Objective, unambiguous, clinically meaningful	Requires large sample size; long follow-up; affected by subsequent therapies
Progression-Free Survival (PFS)	Time from randomization until first evidence of disease progression or death [30]	Surrogate endpoint; does not always correlate with OS	Not influenced by subsequent therapies; earlier readout; smaller sample size	Requires rigorous and frequent radiologic assessment; may not translate to OS benefit
Time to Progression (TTP)	Time from randomization until first evidence of disease progression (excludes deaths) [30]	Weaker surrogate for OS than PFS	Focuses specifically on drug effect on tumor growth	Does not capture survival impact; may miss toxicity-related outcomes
Disease-Free Survival (DFS)	Time from randomization until evidence of disease recurrence (used in adjuvant setting) [30]	Validated surrogate for OS in some cancers (e.g., stage III colon cancer) [30]	Shorter follow-up than OS; relevant for curative settings	Definition of recurrence can be ambiguous; may include non-clinical recurrences
Event-Free Survival (EFS)	Time from randomization to any event (progression, treatment discontinuation, or death) [30]	Surrogate endpoint used in neoadjuvant settings [30]	Comprehensive capture of negative events	Composite nature can make interpretation challenging
Objective Response Rate (ORR)	Proportion of patients with predefined tumor shrinkage [10]	Basis for accelerated approval in many cancers [10]	Direct measure of drug activity; early readout	May not correlate with survival; does not capture duration of response

Regulatory Framework for Surrogate Endpoints

The FDA's Surrogate Endpoint Table, mandated by the 21st Century Cures Act, provides a comprehensive list of endpoints that have formed the basis for drug approval or licensure [10]. According to section 507(e)(9) of the FD&C Act, a surrogate endpoint is "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure, that is not itself a direct measurement of clinical benefit" but is either known to predict clinical benefit (for traditional approval) or reasonably likely to predict clinical benefit (for accelerated approval) [10].

The FDA emphasizes that the acceptability of any surrogate endpoint must be determined on a case-by-case basis, considering the disease context, patient population, therapeutic mechanism, and available treatments [10]. This contextual approach acknowledges that a surrogate endpoint appropriate for one development program may not be suitable for another, even in the same clinical setting.

Comparative Analysis: OS Versus Surrogate Endpoints

Methodological Comparison of Endpoint Assessment

The measurement and analysis of different endpoints require distinct methodological approaches, statistical considerations, and trial designs. The following diagram illustrates the typical evidence generation pathway for surrogate endpoints and their relationship to OS validation.

Diagram 1: Surrogate Endpoint Validation Pathway to OS Confirmation

Quantitative Comparison of Endpoint Characteristics

The utility of different endpoints varies significantly based on trial context, disease stage, and therapeutic mechanism. The table below provides a structured comparison of key operational and methodological characteristics.

Table 2: Operational Characteristics of Primary Endpoints in Oncology Trials

Characteristic	Overall Survival	Progression-Free Survival	Objective Response Rate	Disease-Free Survival
Measurement Objectivity	High (death is unambiguous) [30]	Moderate (requires blinded radiological review)	Moderate (requires standardized response criteria)	Moderate (requires standardized recurrence definition)
Required Sample Size	Large (hundreds to thousands) [30]	Moderate to Large	Smaller [30]	Moderate to Large
Typical Follow-up Period	Long (years) [30]	Intermediate (months to years)	Short to Intermediate (weeks to months)	Intermediate to Long
Susceptibility to Bias	Low [30]	Moderate to High (assessment bias)	Moderate to High (assessment bias)	Moderate (definition bias)
Influence of Subsequent Therapies	High [30]	Low [30]	Low	Moderate
Clinical Direct Relevance	High (direct patient benefit) [30]	Intermediate (clinical relevance debated)	Variable (disease and context dependent)	High in adjuvant setting
Regulatory Acceptance as Primary Endpoint	Gold standard for traditional approval [69]	Accepted for both traditional and accelerated approval [10]	Primarily for accelerated approval [10]	Accepted in adjuvant settings [30]

Contextual Factors Influencing Endpoint Selection

The appropriateness of different endpoints varies significantly based on disease context, treatment setting, and mechanism of action:

• Metastatic vs. Adjuvant Settings: In metastatic disease, PFS and OS are more commonly used, while DFS and EFS are preferred in adjuvant and neoadjuvant settings where the goal is prevention of recurrence [30].
• Tumor Type and Natural History: Slowly progressing cancers (like some prostate cancers or indolent lymphomas) make OS impractical, while aggressive malignancies with limited treatment options may still utilize OS as a feasible endpoint.
• Therapeutic Mechanism: Immunotherapies with delayed effects or pseudoprogression patterns may not be adequately captured by traditional response-based endpoints, requiring more sophisticated surrogate development [30].
• Treatment Landscape: In diseases with multiple effective subsequent therapies, OS differences may be diluted, making PFS a more sensitive measure of drug activity.

Experimental Design and Methodological Considerations

Statistical Analysis Approaches for OS and Surrogate Endpoints

Robust statistical methodologies are essential for proper endpoint evaluation. The recent FDA guidance emphasizes several key considerations for OS analysis, even when it is not the primary endpoint:

• Pre-specification: OS analyses must be pre-specified in protocols and statistical analysis plans, even when OS is not a primary or secondary endpoint [69].
• Handling of Immature Data: Sponsors should use simulations and calculations to model harm based on hypothetical future data when OS data is immature [69].
• Sensitivity Analyses: Robust sensitivity analyses should assess uncertainty, particularly in settings with non-proportional hazards [69].
• Subgroup Analyses: Pre-specified subgroup analyses must be biologically plausible if intended for labeling or regulatory decision-making [69].

For surrogate endpoints, the correlation with OS must be quantitatively established. The web-based survival analysis tool referenced in the search results provides a valuable methodology for this validation, enabling "univariate and multivariate Cox proportional hazards survival analysis" with appropriate correction for multiple testing [70].

Table 3: Essential Research Reagents and Tools for Endpoint Analysis

Tool/Reagent Category	Specific Examples	Primary Function in Endpoint Research
Statistical Analysis Software	R Survival package [70], SAS, Python lifelines	Perform survival analyses, generate Kaplan-Meier curves, compute hazard ratios
Web-Based Analysis Platforms	Custom survival analysis tools [70]	Enable survival analysis without specialized software; facilitate collaborative validation
Tumor Assessment Guidelines	RECIST 1.1, iRECIST, Lugano criteria	Standardize response and progression definitions for solid tumors and lymphomas
Biomarker Assay Platforms	IHC, PCR, NGS, flow cytometry	Quantify biomarker levels used as surrogate endpoints (e.g., serum proteins, genetic markers)
Clinical Data Management Systems	Electronic data capture (EDC) systems, clinical trial management systems (CTMS)	Collect, manage, and quality-control endpoint data across multiple sites
Independent Review Committees	Blinded independent central review (BICR)	Minimize bias in endpoint assessment, particularly for imaging-based surrogates

Methodological Workflow for Endpoint Validation

The validation of surrogate endpoints follows a rigorous methodological pathway, as illustrated in the following experimental workflow diagram.

Diagram 2: Surrogate Endpoint Validation Workflow

Regulatory Evolution and Global Perspectives

Changing Regulatory Landscape for OS Assessment

The FDA's evolving position on OS represents a significant shift in oncology drug development strategy. Key aspects of this evolution include:

• OS as Both Efficacy and Safety Endpoint: The recent guidance emphasizes that OS functions dually—it can demonstrate therapeutic benefit but also reveal unexpected harms, making its assessment critical even for drugs approved via surrogate endpoints [69].
• Assessment in All Randomized Studies: Sponsors must now assess OS in all randomized oncology studies used to support marketing approval, even when it is not the primary endpoint [69].
• Harm Assessment Focus: The guidance specifically recommends interim analyses for harm to limit patient exposure to potentially detrimental therapies [69].
• Long-term Follow-up Requirements: FDA may require post-marketing commitments to collect long-term OS data when such data is immature at the time of approval [69].

International Perspectives on Endpoint Acceptance

Global regulatory alignment on endpoint acceptance remains variable, as illustrated by research comparing US and Japanese practices. A comprehensive study found that among 1,012 drugs approved in Japan for diseases with FDA-recognized surrogate endpoints, 93.6% used the same surrogate endpoint as the FDA, while 6.4% used different endpoints [12]. Significant variation was observed across therapeutic categories—metabolic drugs showed high concordance (98.7%), while drugs targeting pathogenic organisms demonstrated greater divergence (87.6% concordance) [12].

This international perspective highlights that while surrogate endpoint acceptance is increasingly harmonized, local regulatory considerations, disease prevalence patterns, and historical practices continue to influence endpoint selection in global drug development programs.

The reaffirmation of OS as the gold standard endpoint in oncology occurs within an increasingly sophisticated landscape of surrogate endpoint development and validation. While surrogate endpoints remain essential tools for accelerating drug development—particularly in areas of high unmet medical need—the evolving regulatory emphasis on OS as both an efficacy and safety measure underscores its irreplaceable role in comprehensive benefit-risk assessment.

The future of oncology endpoint strategy lies in context-appropriate selection, rigorous validation of surrogate relationships, and thoughtful integration of OS assessment throughout the drug development lifecycle. As novel therapeutic modalities continue to emerge, the endpoint ecosystem will likewise evolve, but the fundamental importance of demonstrating survival benefit will remain central to oncology drug development.

Successful navigation of this complex landscape requires deep expertise in clinical trial design, statistical methodology, and regulatory strategy—ensuring that innovative therapies can reach patients efficiently while maintaining the evidentiary standards that protect patient safety and demonstrate meaningful clinical benefit.

In oncology clinical trials, the unequivocal gold standard for demonstrating clinical benefit is Overall Survival (OS), defined as the time from randomisation or treatment initiation until death from any cause. Its principal strength lies in its objectivity and direct reflection of a treatment's ultimate value to patients. However, the requirement for large sample sizes and extended follow-up periods to reach statistical maturity can significantly delay drug development and patient access to novel therapies. Consequently, the field has increasingly turned to surrogate endpoints, which are measures that can be evaluated earlier and more frequently than OS and are expected to predict clinical benefit.

For decades, Progression-Free Survival (PFS) and Overall Response Rate (ORR) have served as the workhorse surrogate endpoints in oncology. PFS measures the time from treatment initiation until disease progression or death, while ORR quantifies the proportion of patients with a predefined reduction in tumor burden. However, with the advent of immunotherapies and targeted agents, the limitations of these traditional metrics have become more apparent, particularly their reliance on RECIST (Response Evaluation Criteria In Solid Tumors) criteria, which focus primarily on tumor size. This has catalyzed the development and validation of novel, more sensitive endpoints. Among the most promising is Minimal Residual Disease (MRD), a molecular biomarker capable of detecting residual cancer cells after treatment at levels far below the resolution of conventional imaging.

This guide provides an objective comparison of the performance, methodologies, and utility of PFS, ORR, and MRD within the context of modern oncology drug development, framing the discussion around the critical balance between surrogate and clinical endpoint evaluation.

Comparative Analysis of Key Oncology Endpoints

The following table provides a structured comparison of the traditional and novel endpoints based on current evidence and regulatory precedent.

Table 1: Comparative Analysis of Key Oncology Endpoints

Endpoint	Definition & Measurement	Validation Status & Regulatory Context	Key Advantages	Key Limitations & Challenges
Overall Survival (OS)	Time from treatment start to death from any cause. [71]	Gold standard for clinical benefit; required for traditional approval.	Unambiguous; directly measures patient benefit; not subject to assessment bias.	Requires large sample size & long follow-up; can be confounded by subsequent therapies. [71]
Progression-Free Survival (PFS)	Time from treatment start to disease progression or death. [71]	Accepted surrogate for OS in specific cancers (e.g., SCLC); supports traditional & accelerated approval. [10] [71]	Not confounded by subsequent therapies; shorter trial duration than OS.	Correlation with OS is variable across tumor types; radiological assessments can be subjective. [71]
Overall Response Rate (ORR)	Proportion of patients with tumor shrinkage ≥30% (PR) or complete disappearance (CR) per RECIST. [72]	Common early development endpoint; supports accelerated approval. [10]	Direct measure of drug activity; rapid assessment.	Does not capture clinical benefit of stable disease; low correlation with OS in 91% of trials. [72]
Minimal Residual Disease (MRD)	Detection of residual tumor cells at molecular level post-treatment via ctDNA/BM analysis. [73] [74]	Emerging endpoint; FDA advisory committee supported for accelerated approval in multiple myeloma. [75]	Ultrasensitive (detection in parts per million); predicts relapse months before imaging; quantifiable. [76]	Lack of standardized assays; clinical utility for treatment guidance under investigation. [73]

The Rise of MRD and Circulating Tumor DNA (ctDNA) Analysis

MRD assessment, particularly via circulating tumor DNA (ctDNA), represents a paradigm shift in residual disease monitoring. In non-small cell lung cancer (NSCLC), for instance, ctDNA-based MRD detection can identify patients at high risk of recurrence with a median lead time of 10 months before clinical or radiographic relapse. [76] The analytical workflow for ctDNA-MRD involves two primary technological approaches, each with distinct methodologies and considerations.

Table 2: Comparison of Primary ctDNA-MRD Detection Approaches

Feature	Tumor-Informed Approach	Tumor-Naïve (Agnostic) Approach
Principle	Patient-specific mutations identified via tumor tissue sequencing (WES/WGS) are tracked in plasma. [74]	Plasma is screened for a predefined panel of recurrent cancer-associated alterations. [74]
Key Platforms	Signatera (Natera), RaDaR (Inivata), MRDetect (Veracyte) [74]	Guardant Reveal (Guardant Health), InVisionFirst-Lung (Inivata) [74]
Sensitivity	Very high (LoD as low as 0.0001% tumor fraction) [74]	Moderate (LoD typically 0.07–0.33% mutant allele frequency) [74]
Tissue Requirement	Requires high-quality tumor sample. [74]	No prior tumor tissue needed. [74]
Turnaround Time	Longer (weeks) for custom assay design. [74]	Shorter (days). [74]
Ideal Use Case	High-sensitivity applications in curative settings (adjuvant, neoadjuvant). [74]	Broad screening when tissue is unavailable or for rapid results. [74]

Experimental Protocols for Endpoint Evaluation

Protocol 1: Assessing PFS as a Surrogate for OS

A recent meta-analysis provides a robust methodological framework for evaluating PFS as a surrogate endpoint for OS at the trial level, specifically in small cell lung cancer (SCLC). [71]

• Study Design: A systematic review of randomized controlled trials (RCTs).
• Data Extraction: Hazard ratios (HRs) for PFS and OS, along with PFS rates (6-month, 1-year) and OS rates (1-, 1.5-, 2-year), were extracted from 43 eligible RCTs. [71]
• Statistical Analysis: A weighted linear regression of log HRs for PFS and OS was performed, with the Pearson correlation coefficient (R) used to quantify the relationship. Per the ReSEEM guidelines, R ≥ 0.7 indicates a strong correlation. [71]
• Key Findings: In treatment-naïve SCLC, PFS showed a moderate correlation with OS (R between 0.5 and 0.7), supporting its use as a surrogate in this context. The 1-year PFS rate was strongly correlated with longer-term OS (1.5-year and 2-year). [71]

Protocol 2: Ultrasensitive MRD Detection in Early-Stage Lung Cancer

The clinical performance of MRD is demonstrated by studies using assays like the Foresight CLARITY platform, which employs a phased-variant enrichment method to achieve ultra-high sensitivity. [76] [74]

• Sample Collection: Peripheral blood samples are collected at predefined timepoints (e.g., pre-operative, post-operative, during follow-up).
• ctDNA Extraction and Library Preparation: Cell-free DNA (cfDNA) is extracted from plasma and converted into sequencing libraries.
• MRD Detection (PhasED-Seq Method): This tumor-informed method does not rely on a single mutation. Instead, it identifies multiple mutations that are physically located on the same DNA molecule (in phase). Tracking these phased variants significantly boosts the signal-to-noise ratio, enabling detection of ctDNA fragments present at frequencies below one part per million. [74]
• Data Analysis: Bioinformatic pipelines analyze sequencing data to detect the presence of tumor-derived DNA. The primary readout is MRD-positivity or negativity.
• Outcome Correlation: MRD status is correlated with Recurrence-Free Survival (RFS). In a stage I lung cancer cohort, post-operative MRD detection was significantly associated with worse RFS (HR = 3.14 at the post-operative landmark). [76]

The following diagram illustrates the logical relationship and decision-making flow in clinical trials incorporating MRD assessment.

The Scientist's Toolkit: Essential Reagents and Technologies

Successful implementation of endpoint evaluation, particularly novel biomarkers like MRD, relies on a suite of specialized research reagents and platforms.

Table 3: Key Research Reagent Solutions for Endpoint Analysis

Research Tool	Primary Function	Example Application
RECIST v1.1 Guidelines	Standardized criteria for measuring tumor burden via CT/MRI to define PD, SD, PR, CR. [77]	Primary methodology for determining PFS and ORR in solid tumor trials. [72] [77]
PD-L1 IHC Assays (22C3, 28-8)	Immunohistochemistry kits to quantify PD-L1 expression via Tumor Proportion Score (TPS) or Combined Positive Score (CPS). [77]	Predictive biomarker for stratifying patients and analyzing outcomes in immunotherapy trials. [77]
NGS Panels for ctDNA	Targeted sequencing panels (e.g., OncoPanel, TSO-500) to profile tumor mutations from tissue or blood. [77]	Enables tumor-informed MRD assay design; assesses Tumor Mutational Burden (TMB) and MSI status. [77] [74]
Ultra-Sensitive MRD Assays	Platforms like Signatera or Foresight CLARITY to track patient-specific mutations in plasma cfDNA. [76] [74]	Detection of molecular relapse in adjuvant settings; correlating MRD negativity with improved survival. [75] [76]
Unique Molecular Identifiers (UMIs)	DNA barcodes ligated to individual DNA molecules before PCR amplification to correct for sequencing errors. [74]	Critical for achieving the high specificity required by tumor-naïve and hybrid-capture MRD assays. [74]

The following workflow diagram outlines the key steps in the two main approaches for ctDNA-based MRD detection.

The evolution of endpoints in oncology reflects the field's increasing sophistication. While PFS and ORR remain valuable, their limitations in the era of immuno-oncology and targeted therapies are clear. The emergence of MRD, powered by ultrasensitive ctDNA detection, offers a transformative opportunity to assess therapeutic efficacy with unprecedented speed and sensitivity. Its ability to predict clinical outcomes months before radiographic progression and its recent validation for regulatory use in hematologic malignancies underscore its potential.

For researchers and drug developers, the critical task is to continue the rigorous validation of these novel surrogates, particularly in solid tumors, and to standardize the complex technological platforms that underpin them. The future of oncology development lies in a multi-faceted endpoint strategy, where traditional surrogates are complemented by dynamic molecular biomarkers like MRD, accelerating the delivery of effective treatments to patients.

In the pursuit of accelerating therapeutic development, surrogate endpoints—biomarkers used as substitutes for direct measures of clinical benefit—have become fundamental tools in clinical trials. Regulatory agencies like the U.S. Food and Drug Administration (FDA) recognize that these endpoints can reduce trial duration, size, and cost when compared to trials requiring clinical outcomes such as improved survival or quality of life [3] [78]. The FDA maintains a public "Table of Surrogate Endpoints That Were the Basis of Drug Approval or Licensure," which catalogs over 200 surrogate markers accepted for regulatory decisions [10] [6]. However, a critical principle governs their use: a surrogate endpoint validated in one specific context cannot be automatically applied to another. This context-dependency represents a fundamental challenge in drug development, where misunderstanding these boundaries can lead to inaccurate assessments of a therapy's true clinical value.

The validation of a surrogate endpoint is not a universal endorsement but rather an acceptance for a specific context of use, which depends on factors including the disease, patient population, therapeutic mechanism of action, and available treatments [10] [4]. This article explores the scientific and methodological rationale behind this context-dependency, providing a structured comparison of evidence and experimental approaches essential for researchers, scientists, and drug development professionals.

Theoretical Framework: Levels of Evidence and Validation

The Validation Hierarchy for Surrogate Endpoints

The "Ciani framework" for surrogate endpoints, widely accepted by the international health technology assessment (HTA) community, proposes three levels of evidence required for surrogate validation [4]:

• Level 3: Biological Plausibility. The surrogate endpoint lies on the causal pathway of the disease and has a clear mechanistic rationale linking it to the final patient-relevant outcome.
• Level 2: Observational Association. Epidemiological studies or clinical trials demonstrate a correlation between the surrogate endpoint and the target patient-relevant outcome at the individual level.
• Level 1: Interventional/Treatment Effect Association. This highest level of evidence requires data from randomized controlled trials (RCTs) showing an association between the treatment effect on the surrogate endpoint and the treatment effect on the target outcome. This is often quantified using metrics like the trial-level coefficient of determination (R²) [4].

The following diagram illustrates the logical relationships in this validation framework and the core principle of context-dependency.

Diagram 1: Pathway to Surrogate Endpoint Validation and Its Context-Dependency.

The Regulatory Perspective on Context

The FDA explicitly emphasizes this context-dependent nature. Its Surrogate Endpoint Table is intended as a reference guide, but "the acceptability of these surrogate endpoints for use in a particular drug or biologic development program will be determined on a case-by-case basis" [10]. The agency cautions that a surrogate endpoint appropriate for one program should not be assumed appropriate for another in a different clinical setting. This stance is rooted in historical examples where plausible surrogates failed to predict clinical benefit when applied in new contexts, a risk that is particularly acute when using "reasonably likely" surrogate endpoints under the Accelerated Approval pathway [3] [6].

Comparative Evidence: Case Studies of Context-Dependency

Case Study 1: Cardiovascular Outcomes — LDL-Cholesterol

Low-density lipoprotein (LDL) cholesterol reduction is a validated surrogate endpoint for the reduction of cardiovascular events and forms the basis for the approval of statins [79]. However, its predictive value is not consistent across all drug classes. As noted in one review, "LDL-cholesterol has been shown to be a valid surrogate endpoint for cardiovascular related mortality for statins it appears to be less predictive for other classes of lipid lowering therapies such as fibrates" [4]. This illustrates that even with a strong overall validation record, the drug's mechanism of action is a critical element of the context.

Case Study 2: Oncology — Progression-Free Survival in Multiple Myeloma

Progression-Free Survival (PFS) is an accepted surrogate for Overall Survival (OS) in multiple myeloma and has supported many drug approvals [55]. However, its validity can be disrupted by molecular heterogeneity within the patient population. The BELLINI trial serves as a critical example:

• Overall Result: The trial demonstrated a significant PFS benefit with venetoclax (a BCL-2 inhibitor) but a significantly worse OS.
• Subgroup Analysis: Subsequent analysis revealed stark heterogeneity. Patients with the t(11;14) translocation or high BCL2 expression showed a trend toward OS benefit (HR 0.82), whereas those without these biomarkers experienced worse OS (HR 1.34) [55].
• Context Conclusion: The discordance between PFS and OS was attributed to a lack of efficacy and increased toxicity from venetoclax in the non-t(11;14) subgroup. This demonstrates that a surrogate endpoint (PFS) valid for the broad myeloma population was not valid for a specific molecular subgroup receiving a targeted therapy, highlighting the context-dependency introduced by molecular biomarkers.

Case Study 3: The Historical Counterexample — Cardiac Arrhythmia Suppression

A seminal case of surrogate endpoint failure is the Cardiac Arrhythmia Suppression Trials (CAST). Ventricular premature beats (VPB) were a strong predictor of increased sudden death risk after acute myocardial infarction. The drugs encainide, flecainide, and ethmozine effectively suppressed VPBs (the surrogate). However, the trials were halted when they revealed these drugs significantly increased mortality compared to placebo [79]. This powerful example shows that even a biomarker with strong epidemiological correlation (Level 2 evidence) can fail when used as a trial-level surrogate (Level 1 evidence), often due to off-target drug effects that are not captured by the surrogate.

Table 1: Comparative Analysis of Surrogate Endpoint Performance Across Contexts

Surrogate Endpoint	Validated Context	New Context Where Validation Failed	Key Contextual Difference	Outcome of Failure
LDL-Cholesterol Reduction	Statin drug class	Fibrate drug class	Drug Mechanism of Action	Weakened prediction of cardiovascular mortality benefit [4]
Progression-Free Survival (PFS)	Multiple myeloma (broad population)	Multiple myeloma (non-t(11;14) subgroup) with venetoclax	Molecular patient subgroup / Drug MoA	Significant PFS benefit did not translate to OS benefit; worse OS observed [55]
Ventricular Premature Beat (VPB) Suppression	Epidemiological predictor	Anti-arrhythmic drugs (encainide, flecainide)	Drug Intervention / Off-target effects	Effective VPB suppression led to increased mortality [79]
Hemoglobin A1c (HbA1c) Reduction	Diabetes treatments for microvascular complications	Specific drug mechanisms may be questioned	Specific Drug MoA / Long-term outcomes	While a validated surrogate, ongoing scrutiny ensures context-specific clinical benefit [79]

Methodological Approaches for Validation and Assessment

Robust validation of a surrogate endpoint across contexts requires specific methodological approaches, primarily based on meta-analysis of multiple randomized controlled trials (RCTs).

Meta-Analytic Validation Using Individual Patient Data

The gold standard for establishing trial-level surrogacy (Level 1 evidence) is an individual patient data (IPD) meta-analysis [15] [4]. This involves pooling raw data from multiple RCTs to assess the relationship between the treatment effects on the surrogate and the true clinical outcome across different trial settings. A novel two-stage meta-analytic model has been developed to address limitations of earlier methods. This model uses the difference in restricted mean survival time (RMST) as the treatment effect measure, which is valid even when the proportional hazards assumption is violated and allows for the evaluation of surrogacy strength at multiple timepoints [15].

Table 2: Key Reagents and Analytical Solutions for Surrogate Endpoint Research

Research Reagent / Solution	Function in Validation Research
Individual Patient Data (IPD)	Raw data from multiple RCTs, enabling standardized analysis and robust assessment of both patient-level and trial-level surrogacy [4].
Restricted Mean Survival Time (RMST)	A measure of treatment effect that does not rely on the proportional hazards assumption, allowing for more flexible and valid surrogacy evaluation over time [15].
Coefficient of Determination (R² trial)	A key statistical metric that quantifies the proportion of the treatment effect on the true outcome that is explained by the treatment effect on the surrogate endpoint at the trial level [15] [4].
Surrogate Threshold Effect (STE)	The minimum treatment effect on the surrogate endpoint needed to predict a statistically significant treatment effect on the final clinical outcome; useful for HTA and trial design [4].
Clayton Survival Copula Model	A widely used reference statistical model for surrogate endpoint validation with time-to-event outcomes, against which new methods are compared [15].

The following diagram outlines the experimental workflow for a two-stage IPD meta-analysis to validate a surrogate endpoint.

Diagram 2: Workflow for Meta-Analytic Validation of Surrogate Endpoints.

Statistical and Conceptual Pitfalls

Current practices are evolving to address several pitfalls:

• Reliance on Hazard Ratios (HR): Many standard methods rely on HRs, which can vary over time if the proportional hazards assumption is violated. The use of RMST differences is a more robust alternative [15].
• Ignoring Time Lags: The surrogate and true endpoints often occur on different timelines. Newer models can explicitly account for this time lag, which is crucial for accurately modeling their relationship [15].
• Extrapolation of Evidence: Validation evidence is context-specific. The FDA notes that "surrogate validation should therefore be based on RCTs with the appropriate range of populations, interventions, comparators and outcomes... reflective of the specific HTA decision problem" [4]. Using a surrogate validated in one disease setting for a different disease, or even for a different drug class within the same disease, requires new evidence.

The use of surrogate endpoints is indispensable for efficient drug development, but their context-dependent nature demands rigorous scientific validation for each proposed new use. As summarized in the evidence and case studies, a surrogate's validity is inextricably linked to the specific disease, patient population, and drug mechanism of action. Failure to respect these contextual boundaries has led to historic regulatory missteps and, more importantly, patient exposure to therapies without proven clinical benefit.

Future progress hinges on greater transparency and continued methodological refinement. Recommendations from the scientific community include having the FDA and other regulatory bodies provide more detailed justifications for the evidence supporting each listed surrogate endpoint [6]. Furthermore, the establishment of inter-agency working groups to conduct or commission public meta-analyses for surrogate validation could strengthen the evidence base independent of industry sponsorship [6]. For researchers and drug developers, the imperative is clear: early engagement with regulators and a commitment to robust, context-specific validation methodologies are essential to ensure that surrogate endpoints serve as reliable guides in the development of truly beneficial therapies.

In the rigorous world of drug development, the selection of endpoints is a pivotal decision that determines what constitutes success for a new therapy. This process has traditionally navigated between two principal paradigms: clinical endpoints, which directly measure how a patient feels, functions, or survives, and surrogate endpoints, which are biomarkers or other measures used as substitutes for direct clinical benefits [80]. The evolving landscape of healthcare development now demands a greater focus on the patient voice, emphasizing the integration of Clinical Outcome Assessments (COAs) and quality of life measures into endpoint selection. This shift is central to a patient-centered drug development approach, ensuring that the outcomes measured in clinical trials truly reflect aspects of health that are meaningful to those living with the condition [81].

A treatment benefit is formally defined as "a favorable effect on a meaningful aspect of how a patient feels or functions in their life or on their survival" [81]. The phrases "meaningful aspect" and "in their life" are crucial; an effect that does not impact a patient's usual life or is not meaningful to them cannot be considered a true benefit. This foundational concept underpins the growing imperative to integrate the patient perspective directly into endpoint selection through COAs.

Understanding Endpoint Types: From Surrogates to Clinical Outcomes

Endpoints in clinical trials are broadly categorized into surrogate endpoints and clinical endpoints. Understanding their distinct roles, strengths, and limitations is essential for designing trials that can effectively demonstrate treatment value.

Surrogate Endpoints are "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure, that is not itself a direct measurement of clinical benefit" [10]. They are intended to predict clinical benefit and can support both traditional and accelerated approval pathways. The U.S. Food and Drug Administration (FDA) maintains a table of surrogate endpoints that have been used as primary efficacy endpoints for drug approval [10].

Clinical Endpoints, in contrast, directly measure a patient's clinical benefit. The most definitive clinical endpoint is overall survival (OS). Other examples include measures of how a patient feels or functions, which are often captured using COAs.

Table 1: Comparison of Endpoint Types in Clinical Trials

Feature	Surrogate Endpoint	Clinical Endpoint (via COA)
Definition	A biomarker or measure that predicts clinical benefit [10]	A direct measurement of how a patient feels, functions, or survives [80]
Measurement	Objective lab values, imaging reads (e.g., tumor size, cholesterol) [80]	Can be patient-reported, clinician-observed, or based on patient performance [81]
Time to Result	Often shorter, potentially speeding up drug development [15]	Often longer, as it may require observing long-term patient status [15]
Interpretability	Can be complex; requires validation to ensure it predicts clinical benefit [80]	Directly interpretable as a treatment benefit if it measures a meaningful aspect of health [81]
Regulatory Context	Can support Accelerated Approval if "reasonably likely" to predict benefit [10]	Typically required for traditional approval and confirmation of benefit after Accelerated Approval [80]

The relationship between surrogate and clinical endpoints is a critical area of research. For a surrogate to be considered valid, substantial evidence must establish that it reliably predicts the clinical outcome of interest. However, this relationship can be tenuous, as a treatment may alter a biomarker without affecting the clinical endpoint [80]. Modern statistical methods, such as meta-analytic models using Restricted Mean Survival Time (RMST), are being developed to better evaluate trial-level surrogacy over time and account for potential time lags between the surrogate and true clinical outcome [15].

Clinical Outcome Assessments (COAs): Capturing the Patient Voice

Clinical Outcome Assessments (COAs) are measurements that come directly from the patient or through a clinician/observer report, and are based on a human assessment. They are distinguished from biomarkers because they are influenced by human choices, judgment, or motivation [81] [80]. COAs are the primary tools for quantifying the patient experience in clinical trials.

COAs are categorized based on who provides the assessment. The four main types are detailed in the table below.

Table 2: Categories of Clinical Outcome Assessments (COAs)

COA Type	Reporter	Description	Examples
Patient-Reported Outcome (PRO)	Patient	A report on their health condition coming directly from the patient, without interpretation by anyone else [80].	Symptoms, functional status, health-related quality of life [82] [80].
Clinician-Reported Outcome (ClinRO)	Clinician	An assessment based on a clinician's observation, reporting, and/or interpretation of a patient's health condition [81] [80].	Psoriasis Area and Severity Index (PASI), interpretation of radiographic images, global impressions of change [80].
Observer-Reported Outcome (ObsRO)	Non-Clinician Caregiver	An assessment of a patient's health by someone other than the patient or a healthcare professional, based on observable signs and behaviors [80].	A parent reporting the frequency of a child's vomiting; a caregiver noting observable behaviors [80].
Performance Outcome (PerfO)	Patient	A measurement based on a patient's performance of a defined task in a standardized manner [81] [80].	Distance walked in 6 minutes (6MWT), number of symbols correctly matched in a cognitive test [80].

The selection of an appropriate COA is a critical step in trial design. The measurement must be well-defined and possess adequate measurement properties—such as reliability, validity, and the ability to detect change—to demonstrate a treatment's benefit effectively [81]. The concept measured by the COA, known as the Concept of Interest (COI), must have a clear relationship to how a patient feels or functions [81].

Current Trends and Data on COA Use in Clinical Trials

The integration of COAs into clinical research has been increasing over time, though their adoption varies across therapeutic areas and trial phases. A comprehensive computational survey of the ClinicalTrials.gov registry provides insight into these trends.

An analysis of 35,415 oncology trials initiated between 1985 and 2020 found that only 18% reported using one or more COA instruments [82]. Among these COA-using trials, Patient-Reported Outcomes (PROs) were the most prevalent, utilized in 84% of cases [82]. The use of COAs is more likely in later-phase trials; they are more frequently incorporated in Phase 3 trials compared to Phase 1 or 2 [82]. Furthermore, trials focused on supportive care were nearly three times more likely to use COAs than those focused on direct treatment (Odds Ratio = 2.94) [82].

This trend is not unique to oncology. Across all non-oncology clinical trials (N = 244,440), the rate of COA use was higher, at 26% [82]. This data indicates that while progress has been made, there remains a significant need to further promote and integrate COA use, particularly in early-phase and treatment-focused trials in fields like oncology.

Experimental Protocols for COA Validation and Surrogate Endpoint Evaluation

Protocol for Establishing the Measurement Properties of a New COA

The development and validation of a new Clinical Outcome Assessment is a methodical process. The following protocol outlines the key stages, based on emerging good practices [81].

• Define the Treatment Benefit and Concept of Interest (COI): Start by clearly identifying the meaningful aspect of how a patient feels or functions that the treatment is intended to improve. This is the Concept of Interest (e.g., "reduction in the severity of daily pain") [81].
• Develop the Measurement Procedure: Create the instrument (e.g., a questionnaire, a structured interview guide) that will generate a score to represent the COI. This involves item generation, cognitive debriefing with patients, and ensuring the content comprehensively reflects the COI [81].
• Assess Measurement Properties: Conduct quantitative studies to evaluate key measurement properties [81]:
  • Reliability: The consistency of the scores (e.g., test-retest reliability).
  • Validity: The extent to which the instrument measures the intended COI (e.g., construct validity).
  • Ability to Detect Change: The instrument's sensitivity to changes in the patient's condition over time.
• Define the Endpoint and Analysis Plan: Specify how the COA will be used in the trial endpoint (e.g., "change from baseline to Week 12 in pain severity score") and pre-specify the statistical analysis method [81].

Protocol for a Meta-Analytic Validation of a Surrogate Endpoint

Validating a surrogate endpoint for a time-to-event clinical outcome (e.g., Progression-Free Survival for Overall Survival) requires evidence from multiple randomized controlled trials (RCTs). A modern two-stage meta-analytic approach using Individual Patient Data (IPD) is described below [15].

• Collect Individual Patient Data: Obtain IPD from multiple RCTs that have measured both the candidate surrogate endpoint (S) and the true clinical endpoint (T).
• Calculate Treatment Effects using RMST: For each trial, at pre-specified milestone timepoints (e.g., 3, 5, 7 years), calculate the treatment effect as the difference in Restricted Mean Survival Time (RMST) between the intervention and control groups for both S and T. RMST is the area under the survival curve up to a specific timepoint and is valid without assuming proportional hazards [15].
• Model the Relationship between Treatment Effects: In the second stage, model the relationship between the treatment effects on S and the treatment effects on T across the collection of trials. This is done using a generalized linear mixed model (GLMM) that accounts for correlations between endpoints and timepoints [15].
• Estimate Trial-Level Surrogacy: The strength of the surrogate is quantified by a coefficient of determination (R²) obtained from the model. An R² close to 1 indicates that the treatment effect on the surrogate explains most of the variation in the treatment effect on the true endpoint, suggesting a strong surrogate [15].

The following diagram illustrates this statistical workflow for surrogate endpoint validation.

The Scientist's Toolkit: Key Reagents and Materials

The following table lists essential "research reagents" and methodological components for conducting research in COA and endpoint development.

Table 3: Essential Reagents and Resources for Endpoint Research

Item / Resource	Function / Description
PROQOLID Database	A publicly available database of COA instruments, providing details on their names, acronyms, types, and therapeutic indications, which is essential for instrument selection [82].
FDA's Table of Surrogate Endpoints	A regulatory resource listing surrogate endpoints that have been used in drug approvals, serving as a reference for developers on potential endpoints for discussion with the agency [10].
Individual Patient Data (IPD) Meta-Analysis	The "gold standard" dataset for surrogate endpoint validation, comprising raw data from multiple randomized controlled trials, allowing for a powerful assessment of the surrogate-true endpoint relationship [15].
Restricted Mean Survival Time (RMST)	A statistical measure of treatment effect for time-to-event data, defined as the area under the survival curve to a specific time point. It is increasingly used in validation studies as it does not require the proportional hazards assumption [15].
ClinicalTrials.gov Registry	A comprehensive database of clinical studies worldwide. It is used for surveying trends in endpoint and COA usage across different diseases and trial phases [82].

The integration of Clinical Outcome Assessments into endpoint selection represents a fundamental shift toward a more patient-centered paradigm in drug development. While surrogate endpoints remain valuable tools for accelerating the development process, especially when validated using robust modern methodologies, the ultimate demonstration of a treatment's value lies in its direct, meaningful impact on a patient's life. The growing use of PROs and other COAs in clinical trials is a positive step, yet the relatively low adoption rate in certain areas highlights the need for continued advocacy and methodological refinement. As the field moves forward, the choice of endpoints will continue to balance scientific rigor, regulatory feasibility, and—increasingly—the imperative to capture the authentic patient voice, ensuring that new therapies deliver benefits that are truly meaningful to those they are designed to help.

In modern drug development, the choice of endpoints is pivotal in determining the success and efficiency of clinical trials. Clinical endpoints, which directly measure how a patient feels, functions, or survives, have long been the gold standard for evaluating therapeutic benefit [8]. However, the increasing complexity and cost of drug development have accelerated the adoption of surrogate endpoints – biomarkers expected to predict clinical benefit [8] [10]. The U.S. Food and Drug Administration (FDA) defines a surrogate endpoint as "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure, that is not itself a direct measurement of clinical benefit" but is known or reasonably likely to predict clinical benefit [10].

Against this backdrop, artificial intelligence (AI) and digital biomarkers are emerging as transformative technologies with the potential to redefine endpoint validation and utilization. AI approaches, particularly machine learning models and digital twins, are now being applied to enhance the predictive power of existing biomarkers, generate novel digital endpoints from patient-generated health data, and optimize clinical trial efficiency through prognostic covariate adjustment [83]. These technologies offer promising solutions to longstanding challenges in clinical development, including high variability in disease progression assessments, lengthy trial durations, and substantial sample size requirements [83] [84].

Traditional vs. AI-Enhanced Endpoints: A Comparative Analysis

The following table summarizes key characteristics of traditional clinical endpoints, conventional surrogate endpoints, and emerging AI-enhanced digital biomarkers:

Table 1: Comparison of Endpoint Types in Clinical Drug Development

Characteristic	Clinical Endpoints	Traditional Surrogate Endpoints	AI-Enhanced Digital Biomarkers
Definition	Direct measurement of how patients feel, function, or survive [8]	Laboratory measurements, radiographic images, or physical signs used to predict clinical benefit [10]	AI-derived measures from digital data streams (sensors, wearables) that indicate health status [84]
Validation Requirements	Demonstrated reliability and clinical meaningfulness [8]	Analytical validation, clinical validation, and evidence linking to clinical outcomes [8]	Algorithm validation, analytical verification, and clinical validation for context of use [83]
Typical Timeline for Assessment	Often long-term (months to years) [8]	Short to intermediate-term (weeks to months) [8]	Continuous or frequent sampling (real-time to days) [84]
Regulatory Acceptance	Gold standard for traditional approval [8] [10]	Accepted for both traditional and accelerated approval pathways [10]	Emerging regulatory frameworks; case-by-case assessment [83] [84]
Key Advantages	Direct measurement of patient benefit [8]	Faster assessment, smaller trial sizes, lower costs [8] [6]	Continuous monitoring, objective measurement, potential for early signal detection [83] [84]
Key Limitations	Lengthy, expensive trials; large sample sizes [8]	May not always predict clinical benefit; validation challenges [85] [6]	Immature regulatory pathways; technical validation requirements; privacy concerns [83] [56]

The Validation Gap in Traditional Surrogate Endpoints

Despite their widespread use, significant concerns persist about the validation of traditional surrogate endpoints. An analysis of FDA validation studies for oncologic surrogate endpoints from 2005-2022 revealed that only one of 15 studies demonstrated a strong correlation between surrogate markers and overall survival [85]. This validation gap is particularly concerning given that the number of drugs approved based on surrogate endpoints continues to rise [85] [6].

The FDA's Biomarker Qualification Program (BQP), formalized under the 21st Century Cures Act of 2016, was established to address these challenges by providing a transparent pathway for biomarker validation [56]. However, an analysis of this program shows limited impact, with only eight biomarkers fully qualified since the program's inception, and none of these being surrogate endpoints [56] [86]. This highlights the considerable challenges in qualifying novel biomarkers through formal regulatory pathways.

AI-Generated Digital Twins as Prognostic Covariates: A Case Study in Alzheimer's Disease

Experimental Protocol and Methodology

A recent study demonstrated the application of AI-generated digital twins to improve clinical trial efficiency in Alzheimer's disease (AD) [83]. The methodology followed these key steps:

• Data Harmonization and Model Training: Researchers trained a conditional restricted Boltzmann machine (CRBM) – an unsupervised machine learning model with probabilistic neural networks – on a harmonized dataset of 6,736 unique subjects from historical clinical trials and observational studies [83]. The integrated dataset combined data from the C-Path Online Data Repository for Alzheimer's Disease (CPAD) and the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, encompassing 66 variables including demographics, genetics, clinical severity scores, and laboratory measurements [83].

• Digital Twin Generation: The trained model used baseline data from participants in the AWARE trial (a Phase 2 study of tilavonemab in early AD) to generate individualized predictions of each participant's clinical outcomes if they had received placebo [83]. These digital twins served as prognostic covariates in the statistical analysis.

• Validation Approach: The methodology was validated using data from three independent clinical trials to ensure robustness and generalizability [83]. Positive partial correlation coefficients between the digital twins and actual change scores from baseline in key cognitive assessments demonstrated the predictive validity of the approach.

Table 2: Key Parameters and Outcomes from the Alzheimer's Disease Digital Twin Study

Parameter	Result	Context/Implication
Training Dataset Size	6,736 unique subjects	Combined data from 29 clinical trials and 4 observational studies [83]
Partial Correlation Coefficients	0.30 to 0.39 (Week 96, AWARE trial)	Consistent with validation results from three independent trials (0.30 to 0.46) [83]
Residual Variance Reduction	~9% to 15%	Indicates improved precision in measuring treatment effects [83]
Potential Sample Size Reduction	9% to 15% (total); 17% to 26% (control arm)	While maintaining statistical power [83]
Regulatory Status	Accepted by FDA and EMA for clinical trial applications	Part of PROCOVA-Mixed-Effects Model for Repeated Measures (MMRM) framework [83]

Signaling Pathways and Workflow

The following diagram illustrates the conceptual workflow and logical relationships in creating and applying AI-generated digital twins in clinical trials:

AI-Digital Twin Workflow in Clinical Trials

The Scientist's Toolkit: Essential Research Reagents and Technologies

Table 3: Key Research Reagents and Technologies for AI and Digital Biomarker Development

Tool/Category	Specific Examples	Function/Application
AI/ML Modeling Frameworks	Conditional Restricted Boltzmann Machine (CRBM) [83]	Generates digital twins by predicting individual clinical outcomes under placebo conditions
Data Harmonization Platforms	C-Path Online Data Repository for Alzheimer's Disease (CODR-AD) [83]	Provides standardized, pooled clinical trial data for model training
Digital Endpoint Sources	Wearable sensors, smartphone applications, smartwatches [84]	Capture continuous, real-world data on patient activity, physiology, and behavior
Validation Methodologies	PROCOVA-Mixed-Effects Model for Repeated Measures (PROCOVA-MMRM) [83]	Statistical framework for incorporating digital twins as prognostic covariates
Regulatory Pathway Tools	FDA Biomarker Qualification Program (BQP) [56]	Structured process for qualifying biomarkers for specific contexts of use in drug development

Regulatory Considerations and Implementation Challenges

The integration of AI and digital biomarkers into regulatory decision-making faces several significant challenges. The FDA's Biomarker Qualification Program has demonstrated limitations in advancing novel biomarkers, particularly surrogate endpoints [56] [86]. An analysis of this program revealed that only five of 61 accepted biomarker programs focused on surrogate endpoints, and these faced significantly longer development timelines – nearly four years compared to 31 months for other biomarker types [56]. This suggests that the current regulatory framework may not be well-suited for the rapid evolution of AI-driven biomarkers.

Furthermore, evidence standards for validating AI-based endpoints remain evolving. As one expert noted, "I spend half my time still repeating to my scientists: Don't trust what AI tells you, go verify" [87]. This highlights the critical need for rigorous validation of AI-generated insights while leveraging their pattern recognition capabilities. The PROCOVA framework (prognostic covariate adjustment) represents one approach that has received regulatory acceptance, having received a positive qualification opinion from the European Medicines Agency in September 2022 [83].

From a technical perspective, the implementation of digital biomarkers requires careful attention to data quality, algorithm transparency, and analytical validation. As noted in research on digital endpoints, these technologies must demonstrate robustness across diverse patient populations and clinical settings to gain regulatory acceptance [84]. The growing availability of multimodal data from sources such as genomics, proteomics, wearable sensors, and electronic health records creates both opportunities and challenges for AI-based endpoint development [88] [84].

AI and digital biomarkers represent a paradigm shift in how we conceptualize and validate endpoints for clinical drug development. While traditional surrogate endpoints have faced challenges in validation and correlation with meaningful clinical outcomes [85] [6], AI-enhanced approaches offer the potential for more personalized, continuous, and predictive measures of treatment response.

The case study of digital twins in Alzheimer's disease trials demonstrates that these technologies can already deliver measurable improvements in trial efficiency, including significant reductions in required sample sizes while maintaining statistical power [83]. However, realizing the full potential of these approaches will require addressing ongoing challenges in regulatory alignment, technical validation, and standardization.

As the field evolves, successful integration of AI and digital biomarkers will likely depend on collaborative efforts among researchers, regulatory agencies, and technology developers to establish robust frameworks for validation and qualification. Such efforts have the potential to accelerate the development of innovative therapies while maintaining rigorous standards for demonstrating patient benefit.

Conclusion

The choice between surrogate and clinical endpoints is not a binary one but a strategic balance. While validated surrogate endpoints are indispensable for accelerating drug development, especially for serious conditions with unmet needs, they carry inherent uncertainty and must not eclipse the ultimate goals of extending survival and improving quality of life. The future of endpoint evaluation lies in robust, context-specific validation using established frameworks, greater incorporation of the patient perspective through COAs, and transparent post-approval verification. Researchers and regulators must collaborate to ensure that the pursuit of efficiency does not compromise the delivery of therapies with proven, meaningful clinical benefit for patients.

References

• 1. Biomarkers and Surrogate Endpoints In Clinical Trials - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3551627/]
• 2. Clinical outcome - Toolkit [https://toolkit.ncats.nih.gov/glossary/clinical-outcome/]
• 3. Surrogate Endpoint Resources for Drug and Biologic ... [https://www.fda.gov/drugs/development-resources/surrogate-endpoint-resources-drug-and-biologic-development]
• 4. Use of surrogate endpoints in health technology assessment ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12418889/]
• 5. Clinical Outcomes versus Surrogate Endpoints [https://www.endpointadjudication.com/blog/surrogate-endpoints]
• 6. Validating Surrogate Endpoints to Support FDA Drug ... [https://fas.org/publication/validating-surrogate-endpoints/]
• 7. Patient-relevant outcomes: what are we talking about? A ... [https://bmchealthservres.biomedcentral.com/articles/10.1186/s12913-020-05442-9]
• 8. Biomarkers as Surrogate Endpoints in Drug Development [https://www.sciencedirect.com/science/article/abs/pii/S0149291825002693]
• 9. Surrogate Marker Evaluation: A Tutorial Using R - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC12403976/]
• 10. Table of Surrogate Endpoints That Were the Basis of Drug ... [https://www.fda.gov/drugs/development-resources/table-surrogate-endpoints-were-basis-drug-approval-or-licensure]
• 11. Biomarkers and Surrogate Endpoints in Clinical Drug Development | Applied Clinical Trials Online [https://www.appliedclinicaltrialsonline.com/view/biomarkers-and-surrogate-endpoints-clinical-drug-development]
• 12. Surrogate Endpoints in Pivotal Clinical Trials for Drug ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12605959/]
• 13. A New Generation of Clinical Endpoints for Critical Care ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12369439/]
• 14. The Future of Clinical Trials: 2025 Outlook [https://signanthealth.com/resources/blog/the-future-of-clinical-trials-2025-trends]
• 15. Meta-analytic evaluation of surrogate endpoints at multiple ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12614294/]
• 16. Surrogate Endpoints in Oncology [https://www.rxteach.com/surrogate-endpoints-in-oncology/]
• 17. Surrogate endpoints for overall survival in randomized ... [https://pubmed.ncbi.nlm.nih.gov/40036179/]
• 18. Association Between Surrogate Endpoints and Clinical ... [https://link.springer.com/article/10.1007/s12325-025-03331-3]
• 19. Surrogate end points in oncology: the speed–uncertainty ... [https://www.nature.com/articles/s41571-025-01007-z]
• 20. Bedside implications of the use of surrogate endpoints in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11557723/]
• 21. Surrogate endpoints: a key concept in clinical epidemiology [https://www.sciencedirect.com/science/article/pii/S0895435623003402]
• 22. JAMA：采用了替代终点的研究，在论文中如何规范报告？ [https://www.mediecogroup.com/news/2540/]
• 23. Benefits and Criticisms of Surrogate Endpoints [https://www.allucent.com/resources/blog/surrogate-endpoints-neither-end-nor-point]
• 24. “Positive” Cancer Trials With Surrogate Endpoints Rarely ... [https://www.oncologynewscentral.com/oncology/positive-cancer-trials-with-surrogate-endpoints-rarely-improve-survival-quality-of-life]
• 25. 2025 CBI核酸论坛：mRNA肿瘤疫苗临床开发全路径解读 [https://www.unionclin.com/News/457.html?page=1]
• 26. The REnal Surrogacy accEpTance in Chronic Kidney ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12579683/]
• 27. Clinically meaningful eGFR slope as a surrogate endpoint ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11822292/]
• 28. eGFR slope as a primary endpoint for clinical trials of CKD ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10805343/]
• 29. Use of surrogate endpoints in health technology ... [https://www.sciencedirect.com/science/article/pii/S2589537025003979]
• 30. Clinical endpoints in oncology - a primer - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC8085844/]
• 31. Surrogate Endpoints in Oncology & Use for Cancer Drug ... [https://www.cancertherapyadvisor.com/features/surrogate-endpoints-in-oncology-approvals/]
• 32. Development of a framework and decision tool for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9546394/]
• 33. Surrogate Endpoints in Health Technology Assessment [https://link.springer.com/article/10.1007/s40273-020-00935-1]
• 34. USE OF SURROGATE ENDPOINTS IN HEALTH ... [https://arxiv.org/html/2412.02380v2]
• 35. How To Interpret R-squared in Regression Analysis [https://statisticsbyjim.com/regression/interpret-r-squared-regression/]
• 36. Coefficient of determination [https://en.wikipedia.org/wiki/Coefficient_of_determination]
• 37. 2.6 - (Pearson) Correlation Coefficient r | STAT 462 [https://online.stat.psu.edu/stat462/node/96/]
• 38. Surrogate threshold effect based on a meta-analysis for the ... [https://www.springermedizin.de/surrogate-threshold-effect-based-on-a-meta-analysis-for-the-pred/16706096]
• 39. Surrogate threshold effect based on a meta-analysis for the ... [https://link.springer.com/article/10.1007/s10549-019-05262-4]
• 40. Surrogate Endpoints in HTA: Linking Evidence to Pricing & ... [https://remapconsulting.com/hta/surrogate-endpoints-hta-pricing-access/]
• 41. International collaboration provides new guidance on the ... [https://www.nice.org.uk/news/articles/international-collaboration-provides-new-guidance-on-the-use-of-surrogate-endpoints-in-cost-effectiveness-analysis]
• 42. A Comparison of the Performance of 6 Surrogacy Models, ... [https://www.sciencedirect.com/science/article/pii/S1098301525000233]
• 43. Evaluating surrogate endpoints, prognostic markers, and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4451440/]
• 44. Reporting of surrogate endpoints in randomised controlled ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11231881/]
• 45. Surrogate endpoints in diabetic kidney disease [https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2025.1557813/full]
• 46. One Metric, Two Missions: How eGFR Slope Guides Risk ... [https://ajkdblog.org/2025/08/20/one-metric-two-missions-how-egfr-slope-guides-risk-and-research/]
• 47. Clinical Implications of Slope of GFR in Clinical Trials of CKD ... [https://pubmed.ncbi.nlm.nih.gov/40237639/]
• 48. Validation of Estimated Glomerular Filtration Slope and ... [https://www.medrxiv.org/content/10.1101/2025.06.15.25329660v2.full-text]
• 49. Patient-Focused Drug Development: Fit-for Purpose ... [https://www.fda.gov/regulatory-information/search-fda-guidance-documents/patient-focused-drug-development-selecting-developing-or-modifying-fit-purpose-clinical-outcome]
• 50. Patient-Focused Drug Development Guidance Series [https://www.fda.gov/drugs/development-approval-process-drugs/fda-patient-focused-drug-development-guidance-series-enhancing-incorporation-patients-voice-medical]
• 51. FDA-AACR Workshop Explored Novel Oncology Endpoints [https://www.aacr.org/blog/2025/09/23/looking-beyond-overall-survival-fda-aacr-workshop-explored-novel-oncology-endpoints/]
• 52. Use of data from registered clinical trials to identify gaps ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3777147/]
• 53. Meta-analysis of surrogate endpoints for overall survival in ... [https://pubmed.ncbi.nlm.nih.gov/41101157/]
• 54. In Defence of Context: Surrogates, Sequencing, and Real- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12615315/]
• 55. Progression-free survival as a surrogate endpoint in ... [https://www.nature.com/articles/s41408-024-01109-4]
• 56. Sluggish biomarker program offers FDA opportunity for ... [https://www.agencyiq.com/blog/sluggish-biomarker-program-offers-fda-opportunity-for-reform/]
• 57. Accelerated Approval Program [https://www.fda.gov/drugs/nda-and-bla-approvals/accelerated-approval-program]
• 58. Accelerated Approval confirmatory trials requirement may ... [https://www.whitecase.com/insight-alert/accelerated-approval-confirmatory-trials-requirement-may-further-complicate-life]
• 59. Rethinking FDA's Accelerated Approval Pathway [https://www.mcguirewoods.com/client-resources/alerts/2025/1/rethinking-fdas-accelerated-approval-pathway-new-draft-guidances-and-implications-for-drug-companies/]
• 60. FDA Issues New Draft Guidance on Getting Confirmatory ... [https://www.arnoldporter.com/en/perspectives/advisories/2025/01/fda-guidance-confirmatory-trials-for-accelerated-approval]
• 61. FDA Issues Draft Guidance Describing When a ... [https://www.ropesgray.com/en/insights/alerts/2025/01/fda-issues-draft-guidance-describing-when-a-confirmatory-trial-of-a-drug-seeking]
• 62. International Prescription Drug Price Comparisons [https://pmc.ncbi.nlm.nih.gov/articles/PMC11147645/]
• 63. International Prescription Drug Price Comparisons [https://www.rand.org/pubs/research_reports/RRA788-3.html]
• 64. Decoding Drug Pricing Models: A Strategic Guide to Market ... [https://www.drugpatentwatch.com/blog/decoding-drug-pricing-models-a-strategic-guide-to-market-domination/?srsltid=AfmBOoogQcvsc5rg_esK3jiXQfzpYJuNP_c1LDuVJ17qETUSIG_TGXZC]
• 65. Overview of methods for comparing the efficacies of drugs ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3895352/]
• 66. Postmarket surveillance: a review on key aspects and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6661791/]
• 67. Post-Marketing Surveillance Strategies [https://prorelixresearch.com/post-marketing-surveillance-strategies/]
• 68. A Guide to Real-World Evidence in Clinical Trials [https://www.quanticate.com/blog/real-world-evidence-in-clinical-drug-development]
• 69. FDA Draft Guidance on Overall Survival in Oncology Trials [https://www.precisionformedicine.com/blog/fda-draft-guidance-on-overall-survival-in-oncology-trials-what-sponsors-need-to-know]
• 70. Web-Based Survival Analysis Tool Tailored for Medical ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8367126/]
• 71. Evaluation of progression-free survival as a surrogate ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12557674/]
• 72. Overall Response Rate and solid tumors – the magic ... [https://www.immunocore.com/bringing-science-to-life/detail/22769/overall-response-rate-and-solid-tumors-the-magic-number-to-date]
• 73. Measurable residual disease in hematologic malignancies [https://pmc.ncbi.nlm.nih.gov/articles/PMC12257026/]
• 74. A comprehensive overview of minimal residual disease in ... [https://www.nature.com/articles/s41698-025-00984-9]
• 75. Cancer Research in the Wake of a Key FDA Decision on ... [https://news.med.miami.edu/cancer-research-in-the-wake-of-a-key-fda-decision-on-minimal-residual-disease/]
• 76. Foresight Diagnostics to Present Clinical Performance ... [https://foresight-dx.com/press_releases/foresight-diagnostics-to-present-clinical-performance-data-on-foresight-clarity-mrd-at-iaslc-2025-world-conference-on-lung-cancer/]
• 77. Real-World Outcomes and Biomarker Analysis Based ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12524331/]
• 78. Establishing a Public Resource for Acceptable Surrogate ... [https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2022.820990/full]
• 79. Validated Surrogate Endpoint - NCBI - NIH [https://www.ncbi.nlm.nih.gov/books/NBK453484/]
• 80. Clinical Outcome Assessment, endpoints | AI [https://www.mapi-trust.org/news-events/news/focus-on-the-patient-the]
• 81. Clinical Outcome Assessments: Conceptual Foundation ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4610138/]
• 82. Clinical outcome assessment trends in clinical trials ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10501237/]
• 83. Using AI‐generated digital twins to boost clinical trial ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12639399/]
• 84. The next generation of evidence-based medicine [https://www.nature.com/articles/s41591-022-02160-z]
• 85. FDA validation of surrogate endpoints in oncology: 2005- ... [https://pubmed.ncbi.nlm.nih.gov/36155118/]
• 86. New Study Shows Future Biomarker Development Needs ... [https://friendsofcancerresearch.org/news/new-study-shows-future-biomarker-development-needs-enhancements/]
• 87. How AI & Biomarkers Are Reshaping the Next Frontier in ... [https://www.precisionformedicine.com/blog/how-ai-biomarkers-are-reshaping-the-next-frontier-in-drug-development]
• 88. Biomarker Utilization for Regulatory Decision Making [https://pmc.ncbi.nlm.nih.gov/articles/PMC12619682/]