Recurrence-free survival versus overall survival as a primary endpoint for studies of resected colorectal liver metastasis: a retrospective study and meta-analysis

Recurrence-free survival has been used as a surrogate endpoint for overall survival in trials involving patients with resected colorectal liver metastases. We aimed to assess the correlation between recurrence-free survival and overall survival after resection of colorectal liver metastases to determine the adequacy of this surrogate endpoint. In this retrospective study and meta-analysis, we compiled an institutional cohort of consecutive patients who had complete resection of colorectal liver metastases from the Memorial Sloan Kettering Cancer Center (New York, NY, USA) prospective database. Patients were eligible for inclusion if they were aged 18 years or older, and underwent hepatectomy, with or without operative ablation, between Jan 1, 1991, and April 30, 2019. We estimated overall survival and recurrence-free survival probabilities at various timepoints using the Kaplan-Meier method, and we assessed pairwise associations between these endpoints using Spearman's rank correlation. We also did a meta-analysis of adjuvant phase 3 clinical trials for colorectal liver metastases to assess the correlation between hazard ratios (HRs) for recurrence-free survival and overall survival. We searched MEDLINE for articles of phase 3 randomised controlled trials analysing adjuvant treatment strategies for resected colorectal metastases from database inception to Jan 1, 2022. The titles and abstracts of identified studies were screened before full-text screening and summary data were either recalculated or extracted manually from the published Kaplan-Meier curves (depending on data availability). Data were available for 3299 patients in the institutional database, of whom 2983 were eligible for inclusion in our cohort. Median follow-up was 8·4 years (95% CI 7·9–9·1) , during which time there were 1995 (67%) disease recurrences and 1684 (56%) deaths. Median recurrence-free survival was 1·3 years (95% CI 1·3–1·4) and median overall survival was 5·2 years (95% CI 5·0–5·5). 1428 (85%) of 1684 deaths were preceded by recurrence, and median time from recurrence to death was 2·0 years (IQR 1·0–3·4). Pairwise correlations between recurrence-free survival and overall survival were low to moderate, with a correlation estimate ranging from 0·30 (SD 0·17) to 0·56 (0·13). In the meta-analysis of adjuvant clinical trials, the Spearman's correlation coefficient between recurrence-free survival HR and overall survival HR was r=0·20 (p=0·71). We found a minimal correlation between