Surrogate endpoints for overall survival in chemotherapy and radiotherapy trials in operable and locally advanced lung cancer: a re-analysis of meta-analyses of individual patients' data

Abstract

Background: The gold standard endpoint in clinical trials of chemotherapy and radiotherapy for lung cancer is overall survival. Although reliable and simple to measure, this endpoint takes years to observe. Surrogate endpoints that would enable earlier assessments of treatment effects would be useful. We assessed the correlations between potential surrogate endpoints and overall survival at individual and trial levels.

Methods: We analysed individual patients' data from 15,071 patients involved in 60 randomised clinical trials that were assessed in six meta-analyses. Two meta-analyses were of adjuvant chemotherapy in non-small-cell lung cancer, three were of sequential or concurrent chemotherapy, and one was of modified radiotherapy in locally advanced lung cancer. We investigated disease-free survival (DFS) or progression-free survival (PFS), defined as the time from randomisation to local or distant relapse or death, and locoregional control, defined as the time to the first local event, as potential surrogate endpoints. At the individual level we calculated the squared correlations between distributions of these three endpoints and overall survival, and at the trial level we calculated the squared correlation between treatment effects for endpoints.

Findings: In trials of adjuvant chemotherapy, correlations between DFS and overall survival were very good at the individual level (ρ(2)=0.83, 95% CI 0.83-0.83 in trials without radiotherapy, and 0.87, 0.87-0.87 in trials with radiotherapy) and excellent at trial level (R(2)=0.92, 95% CI 0.88-0.95 in trials without radiotherapy and 0.99, 0.98-1.00 in trials with radiotherapy). In studies of locally advanced disease, correlations between PFS and overall survival were very good at the individual level (ρ(2) range 0.77-0.85, dependent on the regimen being assessed) and trial level (R(2) range 0.89-0.97). In studies with data on locoregional control, individual-level correlations were good (ρ(2)=0.71, 95% CI 0.71-0.71 for concurrent chemotherapy and ρ(2)=0.61, 0.61-0.61 for modified vs standard radiotherapy) and trial-level correlations very good (R(2)=0.85, 95% CI 0.77-0.92 for concurrent chemotherapy and R(2)=0.95, 0.91-0.98 for modified vs standard radiotherapy).

Interpretation: We found a high level of evidence that DFS is a valid surrogate endpoint for overall survival in studies of adjuvant chemotherapy involving patients with non-small-cell lung cancers, and PFS in those of chemotherapy and radiotherapy for patients with locally advanced lung cancers. Extrapolation to targeted agents, however, is not automatically warranted.

Funding: Programme Hospitalier de Recherche Clinique, Ligue Nationale Contre le Cancer, British Medical Research Council, Sanofi-Aventis.

Figure 1

Kaplan-Meier curves of DFS and OS in assessment of adjuvant chemotherapy for non-small-cell lung cancers (A) Chemotherapy vs no chemotherapy. (B) Radiotherapy+chemotherapy vs radiotherapy alone. OS=overall survival. CT=chemotherapy. DFS=disease-free survival. RT=radiotherapy.

Figure 2

Correlation between treatment effects on disease-free and overall survival in the assessment of adjuvant treatment for non-small-cell lung cancers (A) Chemotherapy compared with no chemotherapy. (B) Radiotherapy plus chemotherapy compared with radiotherapy alone. Each trial is represented by a circle, with a size proportional to the number of patients. A logarithmic scale is used on axes. Correlation values are excellent (R=0·92 and R=0·99).

Figure 3

Kaplan-Meier curves of PFS and OS in the treatment of locally advanced disease (A) Radiotherapy plus concurrent chemoradiotherapy compared with radiotherapy alone in non-small-cell lung cancer. (B) Modified radiotherapy compared with standard radiotherapy in non-small-cell and small-cell lung cancers. OS=overall survival. RT=radiotherapy. CT=chemotherapy. PFS=progression-free survival.

Figure 4

Correlation between treatment effects on progression-free and overall survival in locally advanced disease (A) Radiotherapy plus concurrent chemotherapy compared with radiotherapy alone in non-small-cell lung cancer. (B) Modified radiotherapy compared with standard radiotherapy in non-small-cell and small-cell lung cancers. Correlation values are excellent (R=0·97 and R=0·96).

Figure 5

Internal validation of the prediction of overall survival by treatment effects on surrogate endpoints (A) Treatment effects on disease-free survival for adjuvant chemotherapy compared with no chemotherapy in non-small-cell lung cancer. (B) Treatment effects on progression-free survival effects for radiotherapy plus concurrent chemotherapy compared with radiotherapy alone in non-small-cell lung cancer. (C) Treatment effects on disease-free survival for radiotherapy plus adjuvant chemotherapy compared with radiotherapy alone in non-small-cell lung cancer. (D) Treatment effects on progression-free survival for modified radiotherapy compared with standard radiotherapy in non-small-cell and small-cell lung cancers. Predicted HRs for overall survival are calculated from the observed HR on disease-free or progression-free survival of that particular trial and the surrogate model built on all the other trials. Observed HRs are shown for overall survival. All values are shown with 95% prediction intervals. HR=hazard ratio.