Event-free survival underpowered as surrogate for overall survival in osteosarcoma
Intermediate endpoints for overall survival (OS), such as event-free survival (EFS), appear to have limited value in randomized controlled trials (RCTs) involving osteosarcoma, a recent study has found.
A systematic database search yielded 20 eligible RCTs, corresponding to 5,620 patients. A cumulative of 45 treatment arms were reported, though all comprised combination chemotherapy regimens involving two to eight drugs. EFS was the primary endpoint in nine studies, while three used response rate (RR). Only two trials used OS as the main outcome of interest.
Neither EFS nor RR were sufficiently correlated with OS. Weighted least-square regression analysis found only a moderate association between the hazard ratios (HRs) of EFS and OS (R2, 0.456, 95 percent confidence interval [CI], 0.112–0.799).
Five-year EFS, on the other hand, showed good correlation with OS HR (R2, 0.530, 95 percent CI, 0.210–0.850). This was defined as having a coefficient of determination >0.5 in value. RR, in comparison, had a correlation with OS HR that was classified as poor (R2, 0.242, 95 percent CI, 0.00–0.709).
In a sensitivity analysis that excluded five studies with patients with metastatic disease, the correlation between EFS and OS worsened (R2, 0.156, 95 percent CI, 0.00–0.511). A better coefficient of determination was obtained when focusing on localized osteosarcoma alone (R2, 0.598, 95 percent CI, 0.075–1.00), but the correlation was inverse.
Similarly, excluding studies that employed radiation and immunotherapy did not meaningfully improve the correlation between EFS and OS.
“[O]ur first-of-its-kind trial-level investigation of the suitability of intermediate endpoints as surrogates for OS in RCTs of osteosarcoma did not confirm their suitability, even though the rareness of the disease and the difficulty inherent in conducting large-scale RCTs were taken into account,” said researchers.