Improving the clinical meaning of surrogate endpoints: an empirical assessment of clinical progression in phase III oncology trials

Abstract

Disease progression in clinical trials is commonly defined by radiologic measures. However, clinical progression may be more meaningful to patients, may occur even when radiologic criteria for progression are not met, and often requires a change in therapy in clinical practice. The objective of this study was to determine the utilization of clinical progression within progression-based trial endpoints among phase III trials testing systemic therapies for metastatic solid tumors. The primary manuscripts and protocols of phase III trials were reviewed for whether clinical events, such as refractory pain, tumor bleeding, or neurologic compromise, could constitute a progression event. Univariable logistic regression computed odds ratios (OR) and 95% CI for associations between trial-level covariates and clinical progression. A total of 216 trials enrolling 148,190 patients were included, with publication dates from 2006 through 2020. A major change in clinical status was included in the progression criteria in 13% of trials (n = 27), most commonly as a secondary endpoint (n = 22). Only 59% of trials (n = 16) reported distinct clinical progression outcomes that constituted the composite surrogate endpoint. Compared with other disease sites, genitourinary trials were more likely to include clinical progression definitions (16/33 [48%] vs 11/183 [6%]; OR, 14.72; 95% CI, 5.99 to 37.84; P < 0.0001). While major tumor-related clinical events were seldom considered as disease progression events, increased attention to clinical progression in may improve the meaningfulness and clinical applicability of surrogate endpoints to patients with metastatic solid tumors.

Graphical Abstract

INTRODUCTION

Surrogate endpoints, such as progression-free survival (PFS), are commonly utilized in late-phase clinical trials for patients with metastatic solid tumors.¹ Surrogate endpoints are often composed of multiple different measures, including radiologic disease progression, biochemical or serologic progression, and death.² However, defining progression by radiologic or biochemical changes in the absence of clinical symptoms may hold less relevance to patients, particularly in scenarios in which progression-free survival poorly predicts overall survival.³ Conversely, major clinical progression related to worsening tumor-related symptoms, such as progressive pain, tumor bleeding, or spinal cord compression, is common among patients with metastatic solid tumors, yet may not meet formal radiologic criteria for progression.^4–8 For instance, tumor bleeding may be more closely related to tumor erosion and invasion into vessels rather than threshold tumor size changes. An increase in metastatic disease involving the spinal canal and resulting in symptomatic cord compression may be associated with only small changes in measurable tumor size.

Thus, although clinical symptomatic progression is common for patients with metastatic solid tumors and may occur without meeting radiologic progression criteria, it is unclear how often phase III trials define such tumor-related clinical events as disease progression. Here, we characterize the scope and reporting of clinical progression among phase III trials of systemic therapy for patients with metastatic solid tumors.

METHODS

This cross-sectional analysis identified published phase III trials from ClinicalTrials.gov, with search of publications through 2020 as previously reported.¹ Trials included in the study had a superiority, 2-arm design testing systemic therapies for metastatic solid tumors with matured results. Definitions of disease progression were considered to encompass clinical progression if they included clinical events. Examples of clinical events included, but were not limited to, pain, spinal cord compression, and tumor bleeding. To be considered, the clinical event was required to be definable as progression independent of any radiologic or serologic changes. Trials with multiple surrogate survival endpoints were considered to encompass clinical events if any progression endpoint had clinical criteria.

Univariable binary logistic regression examined associations between trial-level factors and outcomes to calculate odds ratios (ORs) with 95% CI. Multivariable analyses were not performed due to small event numbers. A probabilistic sensitivity analysis was performed by modeling the main outcomes as a beta distribution using a Bayesian approach. Random samples were drawn from the beta distribution following 10,000 simulations, and 95% credible intervals were computed. Analyses were completed using SAS v9.4 (Cary, NC) and R v.4.3.2 (Vienna, Austria), and code for all analyses is included in the Supplement. The graphical abstract was made using GraphPad Prism v10 (La Jolla, CA) and Biorender.com.

RESULTS

After excluding 569 trials (not superiority, n=138; supportive care, n=100; lack of mature overall survival data, n = 180; not 2-arm, n = 33; and non-metastatic, n = 118), a total of 216 trials enrolling 148,190 patients were included (see Supplement for a complete list of trials). Trial were published from 2006 through 2020 (Table 1). 94% of trials were industry sponsored (204 of 216), and 49% of trials (106 of 216) used a double-blind design.

Table 1.

Characteristics of 216 eligible trials included for analysis.

Characteristic	Value
Disease site, N (%)
Breast	30 (14)
Gastrointestinal	51 (24)
Genitourinary	33 (15)
Thoracic	62 (29)
Othera	40 (19)
Cooperative group ledb, N (%)	28 (13)
Industry sponsored, N (%)	204 (94)
Number of enrolled patients, median (IQR)	610 (413–893)
Publication yearc, median (IQR)	2014 (2012–2016)
Double-blind, N (%)	106 (49)
Overall survival primary or co-primary endpoint, N (%)	122 (56)

^aOther disease sites: central nervous system, endocrine, gynecologic, head and neck, sarcoma, and skin.

^bCooperative group and industry sponsorship were not considered mutually exclusive.

^cThe publication year range was from 2006 to 2020.

Abbreviation: IQR, interquartile range.

A major change in clinical status (e.g., progressive pain, tumor bleeding, or spinal cord compression) was considered a progression event in at least 1 surrogate endpoint by 13% of the trials (n = 27) (Figure 1) (see Supplement for the list of trials). A Bayesian probabilistic sensitivity analysis suggested an estimated mean inclusion rate of 12.8% (95% credible interval: 8.6% to 17.6%) (Figure S1). The most common method of including clinical events in progression criteria was through the composite terms progression-free survival or time-to-progression (59%, 16 of 27). Additional terms used were skeletal-related event (19%, n = 5), time to treatment failure (11%, n = 3), clinical progression (n = 1), time to symptomatic deterioration (n = 1), and time to opiate use (n = 1). The surrogate endpoints that included clinical progression criteria were most commonly secondary endpoints (81%, n = 22); in 3 trials, surrogate endpoints with clinical progression criteria represented the primary endpoint (11%, n = 3). In the remaining two instances, the surrogate endpoint was defined post hoc (exploratory). However, of the trials evaluating clinical progression, only 59% (n = 16) reported clinical progression outcomes separated from the overall composite surrogate endpoint findings. A Bayesian probabilistic sensitivity analysis suggested an estimated mean reporting rate of 58.8% (95% credible interval: 41% to 76%) (Figure S2).

Figure 1.

A forest plot is shown with univariable logistic regressions for trial-level covariates and the odds of using clinical criteria for progression. The 95% confidence interval is shown, and the red line represents an OR of 1. Abbreviation: OR, odds ratio.

The use of clinical progression criteria has not significantly changed over time (slope = −0.26; 95% CI, −1.53 to 1.01; P = 0.66). Compared with all other disease sites, genitourinary trials were more likely to include clinical progression definitions (16/33 [48%] vs 11/183 [6%]; OR, 14.72; 95% CI, 5.99 to 37.84; P < 0.0001) (Figure 1). Larger trials also had higher odds of using clinical progression definitions (per every additional 100 patients; OR, 1.14; 95% CI, 1.04 to 1.25; P = 0.006) (Figure 1). Conversely, open-label trials were less likely to include clinical progression criteria than double-blind trials (7/110 [6%] vs 20/106 [19%]; OR, 0.29; 95% CI, 0.11 to 0.69; P = 0.008) (Figure 1). Lastly, compared to trials using overall survival as a primary endpoint, trials with a surrogate primary endpoint were less likely to include clinical criteria in progression definitions (6/94 [6%] vs 21/122 [17%]; OR, 0.32; 95% CI, 0.12 to 0.80; P = 0.02) (Figure 1).

DISCUSSION

In this large-scale analysis of phase III randomized controlled trials assessing systemic therapy regimens for patients with metastatic solid tumors, most trials did not consider clinically symptomatic progression as disease progression. These findings argue for standardized approaches to defining, managing, analyzing, and reporting clinical progression events among patients enrolled in phase III systemic therapy trials. Further, by focusing on patient-centric events, phase III trials adding clinical progression to their surrogate endpoints have an opportunity to improve the meaningfulness and applicability of surrogate survival endpoints to patients with metastatic solid tumors.

Although the present study is the first large-scale assessment of the inclusion of symptomatic progression criteria in published phase III oncology trials, a multitude of other studies have argued for the importance of patient-centric endpoints in oncology clinical trials.^9–11 A review by Secord and colleagues details published trials that incorporated patient-reported outcomes into composite endpoint definitions, which is a laudable goal complementary to the goal of inclusion of clinical progression criteria.¹² In addition, for patients enrolled with symptomatic disease, cancer symptom response has been proposed as a trial endpoint.¹³ In our study, genitourinary trials were more likely to use clinical progression definitions than were trials of other histologies. This may be due to the efforts of the Prostate Cancer Working Group in recommending trial designs and/or to patients with metastatic prostate cancer being more likely to develop symptoms related to bone metastases than patients with other histologies (perhaps due to radiographic disease burden and/or long natural history).^14–16 As the lifespan of patients with metastatic cancer continues to increase with advances in systemic therapies, the impetus for including clinical progression in endpoint criteria will continue to increase.

While defining disease progression by the development of worsening tumor-related symptoms that interfere with function and quality of life may enhance the clinical relevance of progression criteria, blinded independent radiologic review is often considered to be a more objective approach for defining disease progression.^17–20 Therefore, criteria for clinical progression should be clearly defined in the protocol. The use of double-blind designs may reduce the bias associated with determining clinical progression; this may in part explain why, in our study, fewer open-label trials used clinical progression criteria compared with double-blinded studies. It may also explain the apparent reluctance of trials with surrogate primary endpoints to incorporate clinical progression criteria. In either of these scenarios, corroboration of clinical progression by a blinded independent central review committee may further help to reduce bias.²¹ If an endpoint is composed of clinical progression, radiologic progression, and death, it may be unclear which type of event is driving the observed differences between randomization arms.² Sensitivity analyses of composite surrogate endpoints, evaluating for the contribution of clinical progression to the overall findings, may additionally enhance interpretation of the findings.

There are several important limitations to this study. Due to the low numbers of trials with clinical progression criteria, multivariable adjustment was not performed for this analysis, and associations must be interpreted cautiously. Also, the trials may not be representative of global studies because they were obtained from the US ClinicalTrials.gov registry.

Together, this empirical analysis of phase III metastatic solid tumor trials demonstrates that most trials did not consider major changes related to focal tumor-related clinical symptoms as progression. Thoughtful consideration of clinical events at the time of study design may improve the meaningfulness and clinical applicability of surrogate endpoints to patients with metastatic disease.

Novelty & Impact Statement: IJC-24-1530.R1.

Among cancer patients in clinical trials, disease progression generally is assessed by radiologic criteria. However, changes in clinical status are highly relevant to quality of life, not least because they typically warrant a shift in therapeutic approach. Here, the authors evaluated more than 200 phase III trials to better understand the scope of clinical progression assessment. Despite the importance of tumor-related symptoms to patients, only 13 percent of trials included clinical progression criteria. The findings indicate that clinical progression is underevaluated and understudied in phase III trials, highlighting a need for greater inclusion of patient-centric endpoints in clinical trials.

Novelty and impact statement:

Clinical progression is highly meaningful to patients with a direct impact on clinical management and quality of life. Here, we comprehensively evaluated 216 phase III trials to characterize the scope of clinical progression assessment and outcomes. Despite the importance of tumor-related symptoms to patients, only 13% of trials included clinical progression criteria. This work is the first to comprehensively characterize the landscape of clinical progression is under-evaluated and under-studied in phase III trials.

Abbreviations:

OR

odds ratio

PFS

progression-free survival

Data availability:

All source code is publicly available on GitHub (https://github.com/adsherry2024/Analysis-code/tree/main). The other data that support the findings of this study are available from the corresponding author upon reasonable request.