Cancer Treatment: Surrogate Endpoints and Patient Outcomes

Unlocking the Truth Behind Cancer Drug Approvals: A Deep Dive into Surrogate Endpoints

The landscape of cancer drug development and approval is increasingly complex, with a growing reliance on surrogate endpoints to bring new therapies to patients faster. However, this expedited process often comes with a critical caveat: do these drugs truly extend or improve patients’ lives, or do they merely affect intermediate measures that may not reliably predict actual clinical benefit? The illuminating review, “Frequently asked questions on surrogate endpoints in oncology-opportunities, pitfalls, and the way forward,” published in eClinicalMedicine, directly confronts these pressing questions. Authored by Abhenil Mittal, Myung Sun Kim, Shenna Dunn, Kristin Wright, and Bishal Gyawali, this article is an essential resource for anyone involved in the intricate world of oncology trials and patient care.

The Patient’s Perspective vs. Current Realities Patients battling cancer harbor fundamental expectations from their treatments: prolonged life (overall survival, OS) or a better quality of life (QOL). These are the clinical endpoints that directly measure a longer and better life, representing what truly matters to both patients and physicians. Yet, a striking majority of new cancer drugs are now approved not based on these pivotal outcomes, but on surrogate endpoints such as tumor shrinkage or delayed tumor progression. While these surrogate measures aim to provide quicker indications of efficacy, they often exhibit a weak correlation with the outcomes that truly matter to patients. This creates a significant tension between the desire for early access to potentially life-saving drugs and the imperative to confirm that these treatments improve patient outcomes, rather than just scan results.

A Comprehensive Review in Q&A Format This review offers a comprehensive examination of the surrogate endpoints currently employed in oncology trials across both curative and advanced disease settings. It meticulously clarifies their definitions, outlines the methodologies required for their validation, and presents existing evidence regarding their surrogacy. Structured in an accessible and engaging Q&A format, the article addresses the most commonly asked questions, providing clarity on a complex subject.

Key questions explored include:

• How are surrogate endpoints defined and validated?
• What is the distinction between patient-level and trial-level surrogacy?
• What are the inherent benefits and significant pitfalls of using surrogate endpoints?
• What level of surrogacy is required for regular versus accelerated approval, and are we currently overusing these endpoints?
• What is the surrogacy of Progression-Free Survival (PFS) for OS and QOL, and why does PFS frequently fail to correlate with these crucial clinical outcomes?
• Why do regulatory bodies continue to rely on surrogate endpoints despite their limitations?

Defining Surrogate Endpoints: A Spectrum of Measures The article classifies surrogate endpoints based on their disease setting and type (e.g., tumor response-based or time-to-event). In advanced settings, Overall Response Rates (ORR), also known as objective response rates, and Progression-Free Survival (PFS) are the most common. Their definitions are typically derived from the RECIST criteria, which standardize how tumor responses (complete response, partial response, stable disease, progressive disease) are measured radiologically. ORR represents the percentage of patients achieving a complete or partial response. PFS is defined as the time from randomization until disease progression or death, whichever occurs first.

In curative settings (neo-adjuvant and adjuvant), endpoints like pathological complete response (pCR), Disease-Free Survival (DFS), Invasive Disease-Free Survival (iDFS), and Recurrence-Free Survival (RFS) are used. The article notes that while definitions for advanced cancer surrogates are less controversial, debates persist in curative settings, particularly regarding what constitutes an “event” in measures like DFS. For example, Metastasis-Free Survival (MFS) has been identified as a strong surrogate for OS specifically in prostate cancer, but requires further validation in other contexts.

However, the review cautions that many tumor response endpoints, while reflecting biological activity and useful for early-phase “go/no-go” decisions, have not been validated as surrogates for OS or QOL at a trial level and should therefore not typically be used for drug approval or clinical decisions without confirmatory evidence of clinical benefit.

The Nuances of Validation: Patient-Level vs. Trial-Level Surrogacy A critical distinction highlighted is that between patient-level surrogacy (prognostic marker) and trial-level surrogacy (predictive marker). A valid surrogate endpoint for clinical trials must demonstrate utility in both capacities. For instance, pCR is a strong prognostic marker for individual patient outcomes in certain breast cancers, meaning achieving pCR predicts better long-term outcomes regardless of the specific neoadjuvant treatment. However, pCR fails as a trial-level surrogate because treatments that improve pCR rates in trials do not necessarily translate to improved event-free survival (EFS) or OS. This means while pCR is good for an individual patient, a drug that boosts pCR rates may not ultimately improve survival.

Validation of a surrogate endpoint at the trial level typically involves determining the correlation coefficient between treatment effects on the surrogate and the clinical endpoint across multiple randomized controlled trials (RCTs) through a meta-analysis. The German Institute for Quality and Efficiency in Health Care (IQWIG) provides guidelines, suggesting a lower limit of the 95% confidence interval for the correlation coefficient of ≥0.85, or a coefficient of determination (R2) of at least 0.7, for proven surrogacy. Worryingly, a recent systematic review found that only 23% of surrogate and survival pairs achieved a strong correlation with OS, highlighting a significant “surrogate-survival” gap in oncology.

Perceived Benefits and Real Pitfalls of Surrogate Use The primary appeal of surrogate endpoints like PFS and ORR lies in their ability to measure treatment efficacy earlier in the disease course than OS, potentially expediting drug approvals. This can lead to time-savings, averaging 11–19 months when ORR or PFS are used instead of OS as primary endpoints. Surrogates can also make trials more efficient, potentially requiring smaller sample sizes and reducing costs. They are considered reasonable for single-arm phase I/II trials to make quick go/no-go decisions, and may be particularly useful in rare diseases. Another argument is that PFS is not “diluted” by post-protocol treatments, which can make demonstrating OS benefits challenging in diseases with long survival post progression (SPP).

However, the authors emphasize that careful consideration of surrogate validation and confirmatory studies is crucial to prevent misleading results and the approval of ineffective drugs.

Regulatory Over-Reliance and the Vanishing Distinction The review critically examines the US FDA’s approach, noting that surrogate endpoints are accepted for both accelerated and regular (traditional) approval. Theoretically, traditional approval requires robust validation for clinical endpoints, while accelerated approval, being conditional, only requires surrogates “reasonably likely” to predict clinical outcomes, with a mandatory confirmatory trial.

However, studies reveal a troubling trend: the FDA has increasingly granted regular or traditional approvals based on surrogate endpoints, rather than accelerated approvals, between 2009 and 2018. This suggests the distinction between surrogacy strength required for traditional versus accelerated approvals has effectively vanished. Often, these regular approvals are granted based on response rates from single-arm trials, a practice more common in oncology than in other medical specialties. This is problematic because response rates are generally poor predictors of clinical benefit, and can be inflated in single-arm studies compared to randomized trials. The FDA’s own table of surrogate endpoints lacks clear differentiation for regular vs. accelerated approval, except for pCR.

The increasing use of surrogates for regulatory approvals is stark: analyses show a significant rise from 2000-2010 to 2011-2019 for both regular (31.5% vs 18.6%) and accelerated (29.5% vs 13.3%) approvals. In some years, such as 2018, 93% of approvals for adult solid tumors were based on surrogate endpoints. Disturbingly, for 28% of approved drugs, no OS or QOL data had been reported, and among those with no or unknown correlation to OS, 53% received regular approval. Studies also indicate that 86% of drugs approved based on surrogates have not demonstrated improved overall survival several years post follow-up.

Premature Regular Approval is particularly concerning, as it lacks the “safety net” of mandatory confirmatory trials found with accelerated approval. This makes withdrawing ineffective drugs incredibly difficult. Even when confirmatory trials are conducted, a study found that in 20% of cases, the confirmatory trial used the same surrogate endpoint as the original approval trial, further highlighting the lack of clarity.

Special Considerations for the Adjuvant Setting Adjuvant cancer trials present unique challenges, as treatment is administered in the absence of visible tumors, making response assessment impossible. The goal of adjuvant therapy is unequivocally to improve survival, as a majority of patients would have been cured without it, and those who relapse still suffer harms from therapy. Demonstrating OS benefit in this setting takes much longer than in metastatic disease, leading many recent trials to use surrogate measures like DFS. While DFS is justified in certain cancers with existing validation data, the authors argue for a deeper discussion incorporating evidence, ethics, and economics.

Furthermore, the validity of surrogates is heterogeneous; for example, DFS is a validated surrogate for OS in HER2-positive breast cancer but not others, and even this can change over time with new treatments. Its surrogacy for OS has not been validated in Triple Negative Breast Cancer (TNBC) or hormone receptor-positive breast cancer, yet it continues to be used as a primary endpoint for approvals without OS benefit.

Progression-Free Survival (PFS): The Most Used, Yet Most Problematic Surrogate PFS is the most commonly used surrogate endpoint in trials of advanced cancer, but its validity as a surrogate for OS is highly variable and mostly weak, with only rare exceptions. While correlation might be better in very poor prognosis settings, OS events occur quickly in such cases, diminishing the utility of surrogates.

In the era of targeted therapy and immunotherapy, the relationship between PFS and OS becomes even more complex. Immunotherapy, for instance, can lead to prolonged disease control despite early progression (pseudo-progression), which complicates the linear relationship expected between PFS and OS. In fact, PFS may underestimate OS gains in immunotherapy trials, in contrast to cytotoxic agents where it often overestimates them. Studies have consistently found weak to moderate correlations of ORR and PFS with OS in immune checkpoint inhibitor trials, underscoring the need for routine reporting of mature OS data.

Why PFS Often Fails to Correlate with OS: Multiple Explanations The review meticulously outlines several reasons for the frequent discordance between PFS gains and OS gains:

1. Arbitrary Definition: RECIST criteria arbitrarily define progression, lacking a biological explanation for why a 20% tumor growth delay should consistently correlate with OS.
2. Differential Survival Post Progression (SPP): Gains from an earlier line of treatment (PFS) can be negated by varied SPP. If SPP is long, and multiple treatment options are available after progression, the initial PFS benefit may not translate to OS benefit.
3. Crossover Effects: Patients in the control arm crossing over to the experimental arm upon progression can dilute the apparent OS benefit of the experimental treatment, although this information could show that the drug is better reserved for later lines.
4. Informative or Unequal Censoring: This occurs when the experimental treatment is toxic, leading to imbalanced discontinuation rates or patient dropouts. If sicker patients drop out of the experimental arm and healthier patients in the control arm seek experimental drugs outside the trial, the randomization can become void, leading to a false positive PFS gain. The BOLERO-2 study is cited as a classic example, where a significant PFS benefit for everolimus plus exemestane did not translate to OS benefit, likely due to much higher discontinuation rates in the everolimus arm.
5. Evaluation-Time Bias: If progression is evaluated more frequently in one arm, it can skew PFS estimates. This can happen if investigators believe an experimental treatment is superior and monitor control arm patients more closely to declare progression, moving them to experimental or subsequent therapy.
6. Reader-Evaluation Bias: Determining progression on scans is often subjective, especially in unblinded studies, leading to potential bias in PFS assessment. OS, being a definitive event, is less susceptible to these biases.

PFS and Quality of Life (QOL): Another Disconnect Compounding the issues with OS, PFS has also not been shown to correlate reliably with QOL, the other crucial patient-centric outcome. Studies reveal a weak correlation between PFS gains and improved QOL. In fact, among trials showing improved PFS, 15% actually worsened QOL, whereas drugs that improved OS never worsened QOL.

Patient and Public Misunderstanding of PFS There is significant misunderstanding among patients and the public about the true meaning of PFS. Experiments show that participants often confuse ORR, PFS, and OS, believing that improved PFS automatically means longer life, even with disclaimers. To address this confusion, the authors propose renaming PFS to Progression-Free Interval (PFI), removing “survival” to prevent misinterpretation. Studies on patient preferences further highlight that when fully informed, over 50% of patients do not prefer drugs that only improve PFS without also improving QOL or OS.

Why Regulators Persist: A Misguided Approach? The article suggests that regulators sometimes view PFS as a clinical endpoint in itself, leading to “premature regular approval” even when OS results are negative or could be available quickly. Examples like neratinib, everolimus, or margetuximab in breast cancer illustrate this, where regular approval was granted based on PFS despite available or imminent negative OS data. Given the lack of surrogacy evidence, patient preferences, and timely availability of OS data, considering PFS a clinical endpoint is deemed “deeply flawed and misguided”.

The Way Forward: Optimizing Trial Design and Ensuring True Benefit The review concludes with concrete proposals for a path forward:

• Mandatory Validation: Surrogate measures must be appropriately validated before being used as endpoints for randomized trials or as the basis for drug approval.
• Transparency: The strength of surrogacy required for accelerated versus traditional approval needs to be transparently defined. A prognostic marker (patient-level surrogate) should never be mistaken for a trial-level surrogate endpoint.
• “One-Trial Approach”: The FDA’s new guidance suggests a “one-trial approach” where a single trial is powered for both a surrogate (like ORR or PFS) and a clinical benefit endpoint (like OS) using adaptive designs. This approach could simultaneously provide OS data and validate/disprove surrogacy, potentially making the system more efficient.
• OS as Default Endpoint: Overall Survival (OS) should be the default endpoint, especially in later-line settings or where median OS is less than a year. This is logistically efficient in last-line settings due to earlier achievement of survival endpoints, avoiding issues like crossover and long SPP.
• Direct QOL Measurement: Quality of Life (QOL) outcomes should be directly measured in trials, not inferred from surrogates.
• Symptomatic PFS: Given the arbitrary nature of RECIST-defined PFS, symptomatic PFS (incorporating patient symptoms) may be more meaningful. Validation studies are needed to assess its performance as a surrogate.
• Mandatory Blinding: Any trial using PFS as an endpoint must be mandatorily blinded to mitigate assessment biases.
• Re-evaluating ORR: ORR should seldom be relied upon as a surrogate endpoint, though complete response rates (CR) could potentially be used for accelerated approval, as CR often reflects true beneficial drug effects.
• Accelerated Approval for Surrogates: Even when surrogates are used, they should ideally lead only to accelerated approval, maintaining the crucial safety net of confirmatory trials to withdraw ineffective drugs.
• Beyond Surrogacy: Critical appraisal of trials must also consider other factors, including the appropriateness of the control arm, presence of informative censoring, proportion of patients receiving active drug at relapse, and consistency with other trials. Overpowering trials to detect clinically insignificant differences exposes many patients to toxic treatments without real benefit.

In conclusion, while surrogate endpoints offer the promise of earlier trial completion, their potential to mislead is substantial. The article powerfully advocates for RCTs powered for OS to definitively establish true clinical benefit, or the “one-trial approach” for a balanced path between speed and efficacy. The authors’ call to rename PFS to PFI and to directly measure QOL outcomes reflects a commitment to clearer communication and patient-centered research. This review is an indispensable guide to navigating the opportunities and pitfalls of surrogate endpoints in oncology, urging for greater rigor and transparency to ensure that cancer treatments truly deliver longer and better lives.

APA Reference:

Mittal, A., Kim, M. S., Dunn, S., Wright, K., & Gyawali, B. (2024). Frequently asked questions on surrogate endpoints in oncology-opportunities, pitfalls, and the way forward. eClinicalMedicine, 76, 102824. https://doi.org/10.1016/j.eclinm.2024.102824