CheckMate 026

KEYNOTE-024 and CheckMate 026 asked the same question at nearly the same time and got opposite answers. KEYNOTE-024 used pembrolizumab with a PD-L1 tumour proportion score threshold of 50 percent; CheckMate 026 used nivolumab with a threshold of 1 percent for entry and 5 percent for the primary analysis.

Among the 423 patients with PD-L1 expression of 5 percent or more, median progression-free survival was 4.2 months with nivolumab against 5.9 months with chemotherapy (hazard ratio 1.15, 95 percent confidence interval 0.91 to 1.45, p=0.25), and median overall survival was 14.4 against 13.2 months (hazard ratio 1.02, 0.80 to 1.30). Sixty percent of the chemotherapy arm crossed over to nivolumab. Treatment-related adverse events of any grade occurred in 71 percent on nivolumab and 92 percent on chemotherapy, and grade 3 or 4 events in 18 against 51 percent.

The failure is usually explained by the threshold and by an imbalance that put more patients with poor prognostic features in the nivolumab arm. Whatever the cause, CheckMate 026 is why single-agent PD-1 blockade in the first line is restricted to high PD-L1 expression, and why nivolumab's first-line indication came only later, with ipilimumab in CheckMate 227 and with chemotherapy in CheckMate 9LA.