Disease-free survival as a surrogate for overall survival in early-stage pancreatic cancer trials: a correlation meta-analysis

Abstract

Objective: Disease-free survival (DFS) is frequently used as the primary endpoint in trials of adjuvant and neoadjuvant therapies for early-stage pancreatic cancer (PC), but its validity as a surrogate for overall survival (OS) remains uncertain. This study evaluates the strength and consistency of DFS as a surrogate for OS in early-stage PC trials.

Methods and analysis: This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Trials of early-stage PC involving drug therapy were identified through PubMed, Embase and Cochrane CENTRAL databases, and data on DFS and OS with HR and CIs were extracted. Trial-level surrogacy of DFS for OS was assessed using weighted linear regression, calculating the coefficient of determination (R²) and surrogate threshold effect (STE).

Results: 29 trials involving 6777 patients were included. In adjuvant trials, HR of DFS strongly correlated with HR of OS (R²=0.70) at trial-level, with the STE of 0.94 indicating the maximum DFS HR beyond which OS benefit was unlikely. The correlation strength differed between phase III (R²=0.71) versus phase II (R²=0.67) trials. This correlation was stronger in trials including radiation therapy (R²=0.81) and trials in the neoadjuvant setting (R²=0.90). No trial in our study was a registration trial of a new molecule and all involved chemotherapy.

Conclusion: Treatment effects on DFS had a strong correlation with treatment effects on OS, making DFS a potential surrogate endpoint for OS in early-stage PC trials of cytotoxic chemotherapies, but its use in registration trials requires careful consideration due to variability across treatment settings and trial designs.

Prospero registration number: CRD42024595441.

Keywords: Adjuvant therapy; Pancreatic cancer.

Figure 1. Correlation between disease-free survival (DFS) and overall survival (OS) across clinical trials evaluating adjuvant treatments for early-stage pancreatic cancer. The scatter plot displays the HRs for DFS against OS, with each red circle representing a clinical trial. The size of the circles is proportional to the population size of the trials. The black line represents the linear regression fit, and the shaded grey area indicates the 95% CI for the predicted HRs. The dotted vertical line marks the surrogate threshold effect (STE=0.947), the point at which the upper bound of the CI crosses the null effect line (HR=1 for OS).

Figure 2. Correlation between disease-free survival (DFS) and overall survival (OS) in phase III or phase II adjuvant trials for pancreatic cancer. (A) Phase III adjuvant trials show a strong correlation with a surrogate threshold effect (STE) of 0.91 (B) Phase II adjuvant trials display a lower correlation with an STE of 0.83. Bubble sizes represent the population sizes of individual trials.

Figure 3. Leave-one-out cross-validation analysis of observed versus predicted HRs for overall survival (OS). Each trial was left out once, and a surrogate model was built using the remaining trials to predict the HR for OS based on the HR for disease-free survival. Green squares represent the observed HR for OS in each trial, while blue circles represent the predicted HR from the surrogate model. Red error bars indicate the 95% prediction intervals for the predicted HRs. A close agreement between observed and predicted HRs, with the observed HR falling within the prediction interval, indicates a reliable surrogate model.