Abstract
Background
Surrogate endpoints are increasingly used as the basis of cancer drug approvals. A strong association between treatment effects on the surrogate and true endpoint is essential for their validation as reliable endpoints. We assess the current evidence of trial-level surrogacy for colorectal cancer (CRC) endpoints through a novel framework.
Methods
We developed a framework to grade the quality of oncology surrogate endpoints in 7 domains of the surrogate relationship: data source quality (A-B); disease setting homogeneity (0-3); uniformness of surrogate and true endpoint definitions (0-3); number of trials evaluated (0-4); and individual- (0-2) and trial-level associations (0-8). We screened PubMed using keywords and medical subject headings terms for meta-analyses published before July 6, 2024, validating surrogate endpoint associations across CRC trials. Two reviewers blindly assessed each article before reaching a consensus score. The main outcomes were framework-derived evidence scores and reported trial-level associations (R2) with overall survival.
Results
A total of 18 articles were identified, containing 39 evaluations of 12 unique surrogate endpoints for overall survival. Evidence scores ranged from B6 to A23. Only disease-free survival consistently demonstrated high-quality evidence of trial-level surrogacy (≥A18, median R2 = 0.92). Progression-free survival and objective response rate were low-quality surrogates in metastatic CRC (progression-free survival: B6-A15, median R2 = 0.55; objective response rate: B6-A13, median R2 = 0.33). Other evaluated endpoints were demonstrated to be unreliable surrogates.
Conclusion
Disease-free survival shows high-quality surrogacy in stage II-III colon cancer. Endpoints identified in the metastatic setting demonstrate inconsistent surrogacy evidence. Reassessment and identification of new endpoints are warranted, particularly with immunotherapy.
Introduction
Surrogate endpoints are increasingly selected as primary endpoints in oncology trials to expedite drug development, including regulatory approvals. Within the US Food and Drug Administration (FDA) Accelerated Approval program, oncology indications made up 83% of all approvals from 2012 to 2021—a sharp increase from the first 10 years of the program, where they comprised less than 35% of approvals.1 In oncology, surrogate endpoints are substitute measures of efficacy for true endpoints, such as overall survival; however, a strong association between their treatment effects is essential for validation.2,3
In colorectal cancer (CRC), the FDA only considers 3-year disease-free survival (DFS) as a validated surrogate endpoint for 5-year overall survival in the adjuvant setting, while progression-free survival (PFS) and objective response rate are frequently accepted as reasonably likely surrogate endpoints in metastatic disease.3 The FDA evaluates validity of surrogates with a broad evidence base; however, the strongest evidence is collected in meta-analyses of randomized clinical trials measuring the association of treatment effects between a surrogate and clinical endpoint (termed surrogacy validation studies herein).2 The highest level of this evidence class is derived from a 2-stage approach using pooled individual patient data (IPD) to model the correlation between treatment effects at the trial level.2,4-6 PFS previously demonstrated strong surrogacy for overall survival in metastatic CRC when treatment options were limited beyond 5-fluorouracil monotherapy.7,8 However, newer treatments offer multiple therapies after the first-line, potentially attenuating this relationship.8,9 Moreover, immunotherapies and biomarker-selected treatments have introduced new treatment classes with mechanisms of action distinct from traditional cytotoxic chemotherapies, which may impact former assessments of surrogacy.10
Both clinical and methodological challenges of statistical validation have complicated efforts to identify strong surrogate endpoints of overall survival based on trial-level correlations. With a growing number of validation studies of varying reliability, not only does the strength of surrogacy need to be considered but also the quality of validation methods for a complete assessment of surrogate endpoints. In this study, we propose a rank-based framework to assess surrogate quality in endpoint validation studies of oncology trials. We identified example studies in CRC modeling the trial-level association between putative surrogate endpoints and overall survival and applied evidence scores using this framework. We also summarized the current evidence of trial-level surrogacy for CRC endpoints. DFS, PFS, and objective response rate were of primary interest as frequently FDA-accepted surrogate endpoints. Our aim is to provide a framework to guide discussions between regulators and trial sponsors in the selection of appropriate endpoints and to establish guidelines to inform future validation efforts of current and new surrogate endpoints.
Methods
Evidence domains and framework development
We developed an evaluation framework based on evidence criteria in 7 statistically and clinically relevant domains of the trial-level surrogate relationship: data sources (domain 1), disease population (domain 2), surrogate and true endpoint definitions (domains 3-4), number of studies included (domain 5), and individual- and trial-level associations (domains 6-7). Each domain contains evidence criteria associated with a rank, which are presented in Table 1. With the exception of the first domain (data sources), each domain is ordered numerically with higher ranks representing stronger evidence quality. Domain-specific ranks are then summed to compute a global evidence score ranging from B0 (lowest) to A23 (highest) reflecting the overall quality of evidence and strength of surrogacy. The evidence domains and rationale underlying the proposed ranking system are described extensively in the Supplementary Methods, Supplement 1.
Table 1.
List of evidence criteria to grade the quality and strength of individual- and trial-level association analysis of surrogate endpoints in oncology.
| Surrogate endpoint evaluation domains | Evidence criteria |
|---|---|
| Data sources |
|
| Disease population |
|
| Surrogate endpoint definition and derivation |
|
| True endpoint definition and derivation |
|
| Studies included in the meta-analysis |
|
| Individual patient level surrogacy analysis and results | 0 = Was not evaluated. 1 = Strong conventional prognostic association reported in observational studies or single arm designed interventional trials or moderate correlations based on copula models based on randomized control trials. 2 = Strong correlation, adjusting for treatment comparisons, measured by bivariate outcomes model (ie, copula parameter or its variations, eg, global odds ratio or Rho-Copula in bivariate copula models). |
| Trial level surrogacy analysis and results | 0 = Analysis is not feasible. 0-2 = The summary statistics (median time, success rate, etc) of the surrogate and true endpoints were extracted from published papers or calculated based on individual patient data within each treatment group (arm) nested in trials. The correlation between summary statistics of surrogate endpoint and summary statistics of true endpoints were measured by Spearman or Pearson correlation coefficient. |
| • 0 = correlation coefficient < 0.6 • 1 = correlation coefficient: 0.6 to < 0.9 • 2 = correlation coefficient: ≥0.9 0-2 = Treatment effects were measured as difference in median time or response rate between 2 concurrently randomized treatment arms. The correlation between summary statistics of surrogate endpoint and summary statistics of true endpoints were measured by Spearman or Pearson correlation coefficient. • 0 = correlation coefficient: <0.6 • 1 = correlation coefficient: 0.6 to <0.9 • 2 = correlation coefficient: ≥0.9 3-8 = Treatment effects (hazard ratio or odds ratio) on the surrogate and true endpoints were estimated based on 2 concurrently randomized treatment arms nested in trials. The correlation between treatment effects on surrogate endpoint and treatment effects on true endpoints were measured by R2-Copula and/or R2-WLS (Weighted Least Squares). • 0 = both R2: <0.5 • 3 = either R2: 0.5-0.8 and does not achieve strong level defined in score 8 • 8 = either R2: ≥0.8 with 95% confidence interval lower bound >0.6, and neither <0.7 |
This framework was developed by a working group including 4 gastrointestinal medical oncologists, who are also clinical trialists, and a biostatistician. The process began with a literature review of trial-level evidence for surrogate endpoints in CRC. Background information and summary results of the review were presented to the group during which key considerations of surrogacy analyses were highlighted as priority subjects for framework development to align with widely accepted statistical methodology. The literature search also served as the identification process for studies to be included for evidence grading. A preliminary version of the framework was developed by the biostatistician who is an expert in surrogate endpoint evaluation (Q.S.) and piloted; a subset of studies was graded with the proposed evidence criteria by another author (B.C.). A series of meetings followed in which the framework was iteratively tested and refined. Examples of framework revisions included replacing the scoring schema for domain 1 (data sources) from numeric to categorical (1-2 to A-B), creating an additional domain (disease population), and adjusting grading scales in the trial-level analysis domain (0-3 to 0-8).
To demonstrate herein the utility of the final framework, the example studies identified in the literature search were subject to evidence scoring. Each study was independently scored by 2 reviewers (B.C. and Q.S.) who were blinded to each other’s assessments during adjudication. Upon completion, reviewers’ scores were unblinded, and disagreements were resolved by consensus decision among the investigators.
Study selection and data extraction
We searched PubMed on July 5, 2024, using keywords and medical subject headings terminology, including “colorectal,” “meta-analysis,” “endpoint,” “surrogate,” and “coefficient of determination” combined with Boolean operators “AND” and “OR.” The full search strategy is presented in Table S1. Articles were included if they were meta-analyses or pooled analyses evaluating the trial-level association of any endpoint for overall survival across CRC trials. Articles were excluded if they did not attempt to statistically validate an endpoint, did not quantify trial-level surrogacy with R2 (or a statistical derivative), collected data from a single institution only, or were a methodology study. Abstracts and conference proceedings were excluded.
Titles and abstracts were screened by the first investigator (B.C.). Duplicates and articles without full-text articles available were removed. Additional articles either known to the investigators or identified during full-text examination were also reviewed. All articles were reviewed by 2 investigators (B.C. and Q.S.), and data listed in Table S2 in Supplement 1 were independently extracted.
Study outcomes
The main outcomes were the framework-based evidence scores and trial-level associations reported in the included validation studies. Trial-level association quantifies a surrogate’s strength based on the association of treatment effects with a clinical endpoint. If treatment effects were calculated with hazard ratios or odds ratios and hazard ratios, then the coefficient of determination (R2) was collected; otherwise correlation coefficients were preferred for static summary measures. If multiple analyses were performed for the same endpoint, the best analysis was selected (eg, stage III only [unified population] compared with stage II-III [mixed population]). Individual patient-level association was collected as global odds ratios and rank-correlation coefficients for binary and time-to-event surrogate endpoints, respectively. In a subanalysis of DFS, PFS, and objective response rate, surrogacy associations were descriptively analyzed using median and IQR.
Results
The study selection process from the literature search is illustrated in Figure 1. A total of 161 articles were identified in PubMed. After screening by title and abstract, 43 articles were retrieved for full review. An additional 7 articles were identified during full-text examination of included reports or were already known to the investigators. Of the 50 articles for full review, 18 studies containing 39 analyses of 12 unique endpoints were included in the grading stage. Included studies are listed in Appendix S1.
Figure 1.
Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram illustrating the literature search process and selection strategy. aFull-text was sought for an additional 7 articles by manual search based on investigator knowledge of eligible studies and in-text examination of included articles.
Surrogate endpoints
All endpoints assessed for surrogacy of overall survival were time to event or response related. Time-to-event endpoints included PFS (n = 11),7,9,11-19 DFS (n = 5),20-23 time to progression (n = 5),14,15,18,19,24 time to treatment failure (n = 1),14 time to failure of strategy (n = 1),17 duration of response (n = 1),14 duration of disease control (n = 1),17 and time to nadir (n = 1).25 Response-related endpoints included objective response rate (n = 9),11-16,19,24,26 disease control rate (n = 2),13,14 early tumor shrinkage (n = 1),13 and depth of nadir (n = 1).25
Study characteristics
Full characteristics of the 18 studies are provided in Table S3. Of the 39 surrogate endpoint analyses, 5 were limited to colon cancers and 34 included rectal cancers. Surrogacy was tested in 29 analyses of trials containing experimental biologic agents; 27 investigated targeted therapies and only 2 included immunotherapies for PFS and objective response rate. Chemotherapy-only interventions were tested in 10 analyses, 4 of which were 5-fluorouracil monotherapy only. Surrogate testing was conducted in stage IV disease 34 times and stage II-III disease 5 times. Stage II (n = 1), III (n = 3), and pooled II-III (n = 1) patients were evaluated separately. IPD was available in 13 cases, whereas summary trial data was used in 26 analyses. A measure of individual-level association was reported in 11 analyses (Figure S1). For trial-level association, a correlation coefficient was reported as the primary measure in 9 analyses and the coefficient of determination (R2) was the primary statistic in 30 cases. Treatment effects were characterized by paired hazard ratios (or odds ratio and hazard ratio for response-related surrogates) in 17 analyses, absolute difference or ratio of median effects in 17 cases, and both in 5 analyses.
Evidence scores for surrogate endpoint quality
The cumulative evidence scores of surrogate endpoint quality and reported trial-level associations are presented for each analysis in Figure 2. The best R2 outcome from either surrogacy model is presented if multiple were available. The highest evidence score assessed was A23, and the lowest score was B6. Of the IPD analyses (level A), the highest and lowest scores given were 23 and 12, respectively. The highest and lowest scores for summary data analyses (level B) were 12 and 6, respectively. Domain-specific evidence scores for each analysis are presented in Table 2.
Figure 2.
Forest plot of trial-level associations measured in each surrogate endpoint analysis. Best R2 value is reported for each surrogacy evaluation. The cumulative evidence score for each analysis is depicted. R2 estimates reported without confidence limits were assumed maximum uncertainty on the confidence interval. R2 measures were derived from rho-based estimates when R2 was not reported. Abbreviation: CI = confidence interval.
Table 2.
Evidence scores for each framework domain evaluating the quality of surrogate endpoints.
| Endpoint | Study | Data source | Disease population | Surrogate endpoint definition | Clinical endpoint definition | Number of trials | Individual-level surrogacy | Trial-level surrogacy | Total |
|---|---|---|---|---|---|---|---|---|---|
| Progression-free survival | Shahnam et al.11 | B | 2 | 0 | 3 | 4 | 0 | 0 | B6 |
| Elia et al.12 | B | 2 | 2 | 1 | 4 | 0 | 3 | B12 | |
| Colloca et al.13 | B | 3 | 2 | 0 | 2 | 0 | 1 | B8 | |
| Cremolini et al.16 | B | 3 | 0 | 0 | 4 | 0 | 3 | B10 | |
| Colloca et al.14 | B | 3 | 1 | 0 | 2 | 0 | 1 | B7 | |
| Ciana et al.15 | B | 2 | 2 | 2 | 4 | 0 | 0 | B10 | |
| Chirila et al.18 | B | 2 | 2 | 0 | 4 | 0 | 1 | B9 | |
| Chibaudel et al.17 | A | 3 | 3 | 3 | 2 | 1 | 3 | A15 | |
| Tang et al.19 | B | 3 | 0 | 0 | 4 | 0 | 1 | B8 | |
| Shi et al.9 | A | 3 | 3 | 3 | 4 | 1 | 3 | A17 | |
| Buyse et al.7 | A | 3 | 3 | 3 | 3 | 2 | 8 | A22 | |
| Disease-free survival | Sargent et al.22 | A | 3 | 3 | 3 | 4 | 2 | 8 | A23 |
| Sargent et al.23 (stage II only) | A | 3 | 3 | 3 | 4 | 2 | 3 | A18 | |
| Sargent et al.23 (stage III only) | A | 3 | 3 | 3 | 4 | 2 | 8 | A23 | |
| Sargent et al.20 | A | 3 | 3 | 3 | 2 | 0 | 8 | A19 | |
| Yin et al.21 | A | 3 | 3 | 3 | 3 | 2 | 8 | A22 | |
| Overall response rate | Shahnam et al.11 | B | 2 | 0 | 0 | 4 | 0 | 0 | B6 |
| Elia et al.12 | B | 2 | 1 | 1 | 4 | 0 | 0 | B8 | |
| Colloca et al.13 | B | 3 | 1 | 0 | 2 | 0 | 0 | B6 | |
| Cremolini et al.16 | B | 3 | 0 | 0 | 4 | 0 | 0 | B7 | |
| Colloca et al.14 | B | 3 | 1 | 0 | 2 | 0 | 1 | B7 | |
| Ciana et al.15 | B | 2 | 1 | 2 | 4 | 0 | 0 | B9 | |
| Johnson et al.24 | B | 3 | 0 | 0 | 4 | 0 | 0 | B7 | |
| Buyse et al.26 | A | 3 | 3 | 3 | 4 | 0 | 0 | A13 | |
| Tang et al.19 | B | 3 | 0 | 0 | 4 | 0 | 0 | B7 | |
| Disease control rate | Colloca et al.13 | B | 3 | 1 | 0 | 2 | 0 | 1 | B7 |
| Colloca et al.14 | B | 3 | 1 | 0 | 2 | 0 | 2 | B8 | |
| Early tumor shrinkage | Colloca et al.13 | B | 3 | 2 | 0 | 2 | 0 | 0 | B7 |
| Depth of nadir | Burzykowski et al.25 | A | 3 | 3 | 3 | 2 | 1 | 0 | A12 |
| Time to nadir | Burzykowski et al.25 | A | 3 | 3 | 3 | 3 | 1 | 3 | A16 |
| Time to progression | Ciana et al.15 | B | 2 | 2 | 2 | 2 | 0 | 3 | B11 |
| Chirila et al.18 | B | 2 | 1 | 0 | 4 | 0 | 0 | B7 | |
| Johnson et al.24 | B | 3 | 0 | 0 | 4 | 0 | 0 | B7 | |
| Tang et al.19 | B | 3 | 0 | 0 | 4 | 0 | 0 | B7 | |
| Colloca et al.14 | B | 3 | 2 | 0 | 2 | 0 | 1 | B8 | |
| Time to failure of treatment strategy | Chibaudel et al.17 | A | 3 | 3 | 3 | 2 | 1 | 3 | A15 |
| Time to treatment failure | Colloca et al.14 | B | 3 | 1 | 0 | 2 | 0 | 0 | B6 |
| Duration of response | Colloca et al.14 | B | 3 | 1 | 0 | 2 | 0 | 1 | B7 |
| Duration of disease control | Chibaudel et al.17 | A | 3 | 3 | 3 | 2 | 1 | 3 | A15 |
PFS, DFS, and objective response rate
IPD was available for evaluating surrogacy of DFS in all 5 analyses of stage II-III colon cancer patients. Evidence scores ranged from A18 to A23. The median R2 and IQR for DFS were 0.92 and 0.02, respectively (Figure 3). Targeted therapies were included in 1 DFS analysis in which surrogacy remained high (Figure 4). In the metastatic setting, 11 analyses were conducted for PFS, 3 of which had access to IPD. Evidence scores ranged from B6 to A15. The median R2 value for PFS was 0.55 with an IQR of 0.14 (Figure 3). Biologics, including targeted agents and immunotherapies, were tested in 9 analyses. Trial-level surrogacy was higher prior to the introduction of biologics (Figure 4). The strength of surrogacy was low (R2 = 0.38, 95% confidence interval [CI] = 0.18 to 0.59) in the only analysis containing immunotherapies (Table S3). Of the 9 objective response rate surrogacy evaluations, only 1 analysis was performed using IPD. Scores ranged from B6 to A13, and the median R2 was 0.33 (IQR = 0.28) (Figure 3). Biologics were included 8 times. Trial-level associations for objective response rates were low before and after the introduction of biologics (Figure 4). Surrogacy was poor in the analysis including immunotherapies (R2 = 0.35, 95% CI = 0.12 to 0.59) (Table S3). PFS and objective response rate surrogacy was stronger in analyses that limited trial inclusion to the first-line setting (Figure 5). Associations for individual-level surrogacy of DFS and PFS are listed in Figure S1. No studies reported individual-level surrogacy outcomes for objective response rates.
Figure 3.
Trial-level associations for US Food and Drug Administration–accepted surrogate endpoints of overall survival based on data source used in each validation study. Each endpoint is represented by a boxplot (with and without IPD). The number of analyses for each endpoint, median R2, and IQR are calculated. Summary statistics are presented for IPD and non-IPD analyses together for each endpoint. Boxplots represented by vertical dashes indicate less than 2 observations. All DFS evaluations used IPD. Abbreviations: DFS = disease-free survival; ORR = overall response rate; PFS = progression-free survival; IPD = individual-patient data.
Figure 4.
Individual- and trial-level associations of US Food and Drug Administration–accepted surrogate endpoints for overall survival across treatment settings. The datapoints represent the corresponding individual- and trial-level associations reported by each meta-analysis evaluating DFS, PFS, and ORR. Studies that did not report individual-level correlations were plotted at 0 on the horizontal axis. Analyses are weighted by their evidence score; larger points indicate higher global evidence scores. Treatment setting is defined by the class of experimental therapies evaluated across treatment arms of clinical trials within each validation study. Biologic agents include immunotherapies and targeted therapies. Abbreviations: DFS = disease-free survival; ORR = overall response rate; PFS = progression-free survival.
Figure 5.
Reported trial-level associations (R2) for overall response rate and progression-free survival by lines of therapy included in each meta-analysis. Meta-analyses that included either (1) second-line trials only, (2) first and later-line trials, or (3) did not specify treatment setting were characterized as analyses of endpoint surrogacy in the first line and beyond. Crossbars represent median R2 values. If multiple trial-level associations were reported, the highest R2 was selected.
Discussion
Herein, we propose a framework to assess the quality of surrogate endpoints at the trial level from meta-analytic validation studies of oncology trials. A rank-based system that captures the reliability of validation methods and strength of surrogacy allows for the stratification of evidence quality and evaluation of trial-level associations across multiple meta-analytic sources. In this study, we used this framework to summarize and evaluate the current state of trial-level surrogacy for overall survival in CRC, in which DFS remained the only strong surrogate (in stage II-III solid tumor colon cancers) across treatment classes. Although the focus of this evaluation was in CRC, this framework can be generalized to validation studies of any oncology indication.
Alternative criteria for evaluating surrogate endpoints have been previously proposed. One of the earliest frameworks was published in the Journal of the American Medical Association Users’ Guide monograph series,27 which was designed as a set of guiding questions to support clinicians’ appraisal of surrogate use in trials for managing individual patients’ care (Are the results valid? What were the results? Will the results help me in caring for my patients?). Although this framework raises important questions, it does not sufficiently address considerations of current surrogate validation paradigms. Alternatively, the Institute of Quality and Efficiency in Health Care proposed recommendations for accepting surrogates as validated endpoints based on 2 criteria: the correlation between surrogate and clinical endpoints (termed validity) and the reliability of those correlations.28 In principle, the reliability criterion evaluates the methods of validation studies similar to our framework (eg, restriction of disease population, clearly defined endpoints). However, like the Journal of the American Medical Association Users’ Guide, these criteria lack operationalized definitions, which limit objective assessment and hierarchical ranking of evidence.
The Biomarker-Surrogacy Evaluation Schema (BSES3) overcomes these limitations by using graded criteria to assess validity of a surrogate from levels A to F.29 Although the BSES3 is also a hierarchical classification system, it is conceptually and functionally different from our framework. For instance, the analytic unit of BSES3 is randomized data, whereas this framework evaluates meta-analyses of randomized trials. Because trial-level associations are conducted at the meta-analytic stage, coincident statistical modeling is also required by BSES3 users. Further, BSES3 aims to validate surrogates across multiple evidence types, including pharmacologic and epidemiology data—much like at the regulatory level. Consequently, assessment of trial-level validation principles is far less nuanced. Our proposed framework is a novel method of evaluating the surrogates in the context of meta-analytic statistical validation evidence.
The use of surrogate endpoints to support regulatory decisions in the FDA’s Accelerated Approval program has been criticized for failing to predict durable treatment effects on the intended clinical endpoint in postmarket confirmatory trials.30,31 The status of accelerated approvals including those yet to meet postmarketing requirements and withdrawn indications are made publicly available by the FDA.32 With the increase of drug approvals based on surrogate outcomes, the frequency of withdrawn applications after being granted initial approval has also increased,1 suggesting merit to criticisms of expedited approvals. At the same time, accountability for failed approval conversions in postmarket surveillance has improved markedly with the median time between date of initial approval and application withdrawal decreasing from 10.4 years during 1992-2001 to 3.5 years in the 2012-2021 period.1 Nevertheless, continued dependence on ostensibly reliable but nonvalidated endpoints further reinforces challenges of accelerated regulatory frameworks. For example, of all 83 oncology indications granted approvals by the FDA between 2009 and 2014, 66% were decided based on response rate or PFS.30 Within this analysis, objective response rate and PFS were low-quality surrogates owing to poor methodological reliability and strength of association, particularly in the modern biologics era. Despite this, objective response rate and PFS remain attractive endpoints because of their familiarity and the advantage of shorter trial durations compared with overall survival. Accordingly, the use of such endpoints to support regulatory decisions must be weighed in the context of these competing interests.
Our work highlights 2 important calls to action. First, greater accessibility of IPD is needed to support validation efforts. IPD data-sharing initiatives, such as the ACCENT (Adjuvant Colon Cancer End Points)33 and ARCAD (Aide et Recherche en Cancérologie Digestive)34 databases, have previously played key roles in identifying high-quality surrogates. Second, the development of new surrogate endpoints is warranted, specifically where traditional endpoints like PFS and objective response rate fall short in contemporary treatment settings. Multiple newer surrogate markers have already been adopted, such as clinical complete response in rectal organ preservation watch-and-wait trials and molecular response rate in biomarker-selected trials—neither of which are validated. It also remains unclear whether circulating tumor DNA (ctDNA) is an appropriate surrogate endpoint. Implementation of this framework will support robust evaluations of forthcoming trial-level evidence; however, much work is still needed before such evaluations can occur. First, multilateral efforts are needed to define additional criteria of evidence acceptability (eg, assay characteristics, molecular detection thresholds)—similar to the process of establishing minimal residual disease negativity as an accepted surrogate for accelerated approvals in multiple myeloma.35 Second, clinical trials should incorporate the collection and reporting of ctDNA kinetics. A number of ongoing trials are evaluating ctDNA clearance as a study endpoint, such as the COBRA trial (NCT04068103)—albeit, failing to meet its primary endpoint of ctDNA clearance—SU2C ACT3 trial (NCT03803553), and NCT05710406. Correlative overall survival data from these trials are crucial for future surrogate analyses. Additionally, few collecting ctDNA kinetics data, including SU2C ACT3 and the BREAKWATER trial (NCT0460742), are expected to make IPD available, enabling high-quality trial-level surrogacy investigations.
Key limitations of our study must be considered. First, this framework only addresses surrogate endpoint validity in-principle based on statistical association at the trial level. Assessments of validity for the basis of drug approvals require evaluation of additional evidence sources, such as biologic plausibility and epidemiologic data. Although this framework will support surrogate assessments, it yields necessary but not sufficient evidence for regulatory-level validation. Second, the application of this framework to support such decisions is limited by the availability of trial-level evaluations. As this process requires data on both endpoints, the inherent time lag between surrogate use for conditional approvals and validatory evidence remains a challenge. Third, our evidence scores should be interpreted with caution as this framework is not yet validated. Future validation of the proposed domains and criteria may improve certainty of scores for regulatory use. Lastly, surrogate endpoint evaluation in the immunotherapy setting was limited with only 2 analyses including immune checkpoint inhibitors and neither derived from IPD. As such, relationships between surrogate endpoints and overall survival in trials investigating immune checkpoint inhibitors are unclear. Future meta-analyses should evaluate immunotherapy-containing trials separately to better understand contemporary surrogacy relationships in CRC.
This evidence framework offers a standardized evaluation of surrogate endpoint quality in meta-analytic validation studies of oncology trials. DFS was identified as a high-quality surrogate of overall survival for stage II-III colon cancers in the adjuvant setting. PFS and objective response rate were identified as low-quality surrogates for stage IV CRC and should be continually evaluated given their acceptance as reasonably likely surrogates for the basis of accelerated approvals. This rank-based approach supports identification of high-quality endpoints to help direct discussions between health regulators and trial sponsors in the selection of trial endpoints for the basis of regulatory approvals of new cancer drugs. Additionally, our evidence criteria serve as recommendations to guide best practices in statistical validation of novel surrogate endpoints.
Supplementary Material
Acknowledgments
J.M.L.’s work is supported by the BC Cancer Foundation philanthropic donations. His role as a Canadian Cancer Trials Group senior investigator is supported by the Canadian Cancer Society. B.J.C. is a CIHR Canada Graduate Research Scholarship—Master’s and BC Cancer Rising Stars recipient, which help support his research. The funder did not play a role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.
Contributor Information
Brendan J Chia, Medical Oncology, BC Cancer—Vancouver Cancer Centre, Vancouver, BC, Canada.
Carmen Allegra, Hematology and Oncology, University of Florida College of Medicine, Gainesville, FL, United States.
Smitha Krishnamurthi, Medical Oncology, Cleveland Clinic, Cleveland, OH, United States.
Christopher Lieu, Medical Oncology, University of Colorado Cancer Center, Aurora, CO, United States.
Jonathan M Loree, Medical Oncology, BC Cancer—Vancouver Cancer Centre, Vancouver, BC, Canada.
Qian Shi, Quantitative Health Sciences, Mayo Clinic, Rochester, MN, United States.
Author contributions
Brendan J. Chia (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing—original draft, Writing—review & editing), Carmen Allegra (Conceptualization, Methodology, Validation, Writing—review & editing), Smitha Krishnamurthi (Conceptualization, Methodology, Validation, Writing—review & editing), Christopher Lieu (Conceptualization, Methodology, Validation, Writing—review & editing), Jonathan M. Loree (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing—original draft, Writing—review & editing), and Qian Shi (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Supervision, Validation, Writing—review & editing)
Supplementary material
Supplementary material is available at JNCI: Journal of the National Cancer Institute online.
Funding
This work was supported by the BC Cancer Foundation (#F25-02938).
Conflicts of interest
B.J.C. declares no conflicts of interest. J.M.L. has received consulting fees from Taiho, Bayer, Novartis, Ipsen, Sanofi, GSK, and Merck and nonfinancial support from Personalis, Agenus, Guardant Health, Quest Diagnostics, and Bayer for research outside this work.
Data availability
All data underlying this article are available in the article and in its online supplementary material.
References
- 1. Beakes-Read G, Neisser M, Frey P, Guarducci M. Analysis of FDA’s accelerated approval program performance December 1992–December 2021. Ther Innov Regul Sci. 2022;56:698-703. 10.1007/s43441-022-00430-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. U.S. Food and Drug Administration. Guidance for industry: expedited programs for serious conditions–drugs and biologics. fda.gov. Published May 2014. Updated November 28, 2023. Accessed June 18, 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/expedited-programs-serious-conditions-drugs-and-biologics
- 3. U.S. Food and Drug Administration. Table of surrogate endpoints that were the basis of drug approval or licensure. fda.gov. Updated August 29, 2025. Accessed August 12, 2025. https://www.fda.gov/drugs/development-resources/table-surrogate-endpoints-were-basis-drug-approval-or-licensure
- 4. Burzykowski T, Molenberghs G, Buyse M, Geys H, Renard D. Validation of surrogate end points in multiple randomized clinical trials with failure time end points. J R Stat Soc Ser C Appl Stat. 2001;50:405-422. http://www.jstor.org/stable/2680805 [Google Scholar]
- 5. Buyse M, Molenberghs G, Burzykowski T, Renard D, Geys H. The validation of surrogate endpoints in meta-analyses of randomized experiments. Biostatistics. 2000;1:49-67. 10.1093/biostatistics/1.1.49 [DOI] [PubMed] [Google Scholar]
- 6. Elston J, Taylor RS. Use of surrogate outcomes in cost-effectiveness models: a review of United Kingdom health technology assessment reports. Int J Technol Assess Health Care. 2009;25:6-13. 10.1017/S0266462309090023 [DOI] [PubMed] [Google Scholar]
- 7. Buyse M, Burzykowski T, Carroll K, et al. Progression-free survival is a surrogate for survival in advanced colorectal cancer. J Clin Oncol. 2007;25:5218-5224. 10.1200/JCO.2007.11.8836 [DOI] [PubMed] [Google Scholar]
- 8. Saad ED, Katz A, Hoff PM, Buyse M. Progression-free survival as surrogate and as true end point: insights from the breast and colorectal cancer literature. Ann Oncol. 2010;21:7-12. 10.1093/annonc/mdp523 [DOI] [PubMed] [Google Scholar]
- 9. Shi Q, de Gramont A, Grothey A, et al. Individual patient data analysis of progression-free survival versus overall survival as a first-line end point for metastatic colorectal cancer in modern randomized trials: findings from the analysis and research in cancers of the digestive system database. J Clin Oncol. 2015;33:22-28. 10.1200/JCO.2014.56.5887 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Baker SG. Five criteria for using a surrogate endpoint to predict treatment effect based on data from multiple previous trials. Stat Med. 2018;37:507-518. 10.1002/sim.7561 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Shahnam A, Hitchen N, Nindra U, et al. Objective response rate and progression-free survival as surrogates for overall survival treatment effect: a meta-analysis across diverse tumour groups and contemporary therapies. Eur J Cancer. 2024;198:113503. 10.1016/j.ejca.2023.113503 [DOI] [PubMed] [Google Scholar]
- 12. Elia EG, Städler N, Ciani O, Taylor RS, Bujkiewicz S. Combining tumour response and progression free survival as surrogate endpoints for overall survival in advanced colorectal cancer. Cancer Epidemiol. 2020;64:101665. 10.1016/j.canep.2019.101665 [DOI] [PubMed] [Google Scholar]
- 13. Colloca GA, Venturino A, Guarneri D. Analysis of response-related endpoints in trials of first-line medical treatment of metastatic colorectal cancer. Int J Clin Oncol. 2019;24:1406-1411. 10.1007/s10147-019-01504-z [DOI] [PubMed] [Google Scholar]
- 14. Colloca G, Venturino A, Guarneri D. Analysis of clinical end points of randomised trials including bevacizumab and chemotherapy versus chemotherapy as first-line treatment of metastatic colorectal cancer. Clin Oncol (R Coll Radiol). 2016;28:e155-e164. 10.1016/j.clon.2016.05.001 [DOI] [PubMed] [Google Scholar]
- 15. Ciani O, Buyse M, Garside R, et al. Meta-analyses of randomized controlled trials show suboptimal validity of surrogate outcomes for overall survival in advanced colorectal cancer. J Clin Epidemiol. 2015;68:833-842. 10.1016/j.jclinepi.2015.02.016 [DOI] [PubMed] [Google Scholar]
- 16. Cremolini C, Antoniotti C, Pietrantonio F, et al. Surrogate endpoints in second-line trials of targeted agents in metastatic colorectal cancer: a literature-based systematic review and meta-analysis. Cancer Res Treat. 2017;49:834-845. 10.4143/crt.2016.249 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Chibaudel B, Bonnetain F, Shi Q, et al. Alternative end points to evaluate a therapeutic strategy in advanced colorectal cancer: evaluation of progression-free survival, duration of disease control, and time to failure of strategy—an Aide et Recherche en Cancérologie Digestive Group study. J Clin Oncol. 2011;29:4199-4204. 10.1200/JCO.2011.35.5867 [DOI] [PubMed] [Google Scholar]
- 18. Chirila C, Odom D, Devercelli G, et al. Meta-analysis of the association between progression-free survival and overall survival in metastatic colorectal cancer. Int J Colorectal Dis. 2012;27:623-634. 10.1007/s00384-011-1349-7 [DOI] [PubMed] [Google Scholar]
- 19. Tang PA, Bentzen SM, Chen EX, Siu LL. Surrogate end points for median overall survival in metastatic colorectal cancer: literature-based analysis from 39 randomized controlled trials of first-line chemotherapy. J Clin Oncol. 2007;25:4562-4568. 10.1200/JCO.2006.08.1935 [DOI] [PubMed] [Google Scholar]
- 20. Sargent D, Shi Q, Yothers G, et al. ; Adjuvant Colon Cancer End-points (ACCENT) Group. Two or three year disease-free survival (DFS) as a primary end-point in stage III adjuvant colon cancer trials with fluoropyrimidines with or without oxaliplatin or irinotecan: data from 12,676 patients from MOSAIC, X-ACT, PETACC-3, C-06, C-07 and C89803. Eur J Cancer. 2011;47:990-996. 10.1016/j.ejca.2010.12.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Yin J, Salem ME, Dixon JG, et al. Reevaluating disease-free survival as an endpoint vs overall survival in stage III adjuvant colon cancer trials. J Natl Cancer Inst. 2022;114:60-67. 10.1093/jnci/djab187 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Sargent DJ, Wieand HS, Haller DG, et al. Disease-free survival versus overall survival as a primary end point for adjuvant colon cancer studies: individual patient data from 20,898 patients on 18 randomized trials. J Clin Oncol. 2005;23:8664-8670. 10.1200/JCO.2005.01.6071 [DOI] [PubMed] [Google Scholar]
- 23. Sargent DJ, Patiyil S, Yothers G, et al. ; ACCENT Group. End points for colon cancer adjuvant trials: observations and recommendations based on individual patient data from 20,898 patients enrolled onto 18 randomized trials from the ACCENT group. J Clin Oncol. 2007;25:4569-4574. 10.1200/JCO.2006.10.4323 [DOI] [PubMed] [Google Scholar]
- 24. Johnson KR, Ringland C, Stokes BJ, et al. Response rate or time to progression as predictors of survival in trials of metastatic colorectal cancer or non-small-cell lung cancer: a meta-analysis. Lancet Oncol. 2006;7:741-746. 10.1016/S1470-2045(06)70800-2 [DOI] [PubMed] [Google Scholar]
- 25. Burzykowski T, Coart E, Saad ED, et al. ; Aide et Recherche en Cancerologie Digestive Group. Evaluation of continuous tumor-size–based end points as surrogates for overall survival in randomized clinical trials in metastatic colorectal cancer. JAMA Netw Open. 2019;2:e1911750. 10.1001/jamanetworkopen.2019.11750 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Buyse M, Thirion P, Carlson RW, Burzykowski T, Molenberghs G, Piedbois P. Relation between tumour response to first-line chemotherapy and survival in advanced colorectal cancer: a meta-analysis. Lancet. 2000;356:373-378. 10.1016/S0140-6736(00)02528-9 [DOI] [PubMed] [Google Scholar]
- 27. Bucher HC, Guyatt GH, Cook DJ, Holbrook A, McAlister FA for the Evidence-Based Medicine Working Group. Users’ guides to the medical literature: XIX. Applying clinical trial results A. How to use an article measuring the effect of an intervention on surrogate end points. JAMA. 1999;282:771-778. 10.1001/jama.282.8.771 [DOI] [PubMed] [Google Scholar]
- 28. Institute for Quality and Efficiency in Health Care (IQWiG). Validity of surrogate endpoints in oncology: executive summary of rapid report A10-05, version 1.1. Published November. 2011;21. Accessed June 2, 2025. https://www.ncbi.nlm.nih.gov/books/NBK198799/ [PubMed]
- 29. Lassere MN, Johnson KR, Schiff M, Rees D. Is blood pressure reduction a valid surrogate endpoint for stroke prevention? An analysis incorporating a systematic review of randomised controlled trials, a by-trial weighted errors-in-variables regression, the surrogate threshold effect (STE) and the biomarker-surrogacy (BioSurrogate) evaluation schema (BSES). BMC Med Res Methodol. 2012;12:27. 10.1186/1471-2288-12-27 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Kemp R, Prasad V. Surrogate endpoints in oncology: when are they acceptable for regulatory and clinical decisions, and are they currently overused? BMC Med. 2017;15:134. 10.1186/s12916-017-0902-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Mittal A, Kim MS, Dunn S, Wright K, Gyawali B. Frequently asked questions on surrogate endpoints in oncology—opportunities, pitfalls, and the way forward. EClinicalMedicine. 2024;76:102824. 10.1016/j.eclinm.2024.102824 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. U.S. Food and Drug Administration. Resources for information | approved drugs. fda.gov. Updated November 16, 2022. Accessed October 25, 2025. https://www.fda.gov/drugs/drug-approvals-and-databases/resources-information-approved-drugs
- 33. Renfro LA, Shi Q, Sargent DJ. Mining the ACCENT database: a review and update. Chin Clin Oncol. 2013;2:18. 10.3978/j.issn.2304-3865.2013.03.05 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. de Gramont A, Haller DG, Sargent DJ, Tabernero J, Matheson A, Schilsky RL. Toward efficient trials in colorectal cancer: the ARCAD clinical trials program. J Clin Oncol. 2010;28:527-530. 10.1200/JCO.2009.25.2544 [DOI] [PubMed] [Google Scholar]
- 35. Landgren O, Devlin SM. Minimal residual disease as an early endpoint for accelerated drug approval in myeloma: a roadmap. Blood Cancer Discov. 2025;6:13-22. 10.1158/2643-3230.BCD-24-0292 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data underlying this article are available in the article and in its online supplementary material.