Comment

1286 www. thelancet. com/oncology Vol 16 October 2015 previously treated with ipilimumab, the proportion of patients who achieved an objective response to nivolumab was 44% in patients positive for PD-L1 expression, but 20% of patients identified as PD-L1 negative also responded. Similar findings were reported in the Checkmate 066 trial, 2 which compared nivolumab with dacarbazine in patients with BRAF wild-type melanoma. The proportion of patients with an objective response to nivolumab was 52· 7% in those positive for PD-L1 and 33· 1% in patients negative for the protein. More importantly, patients negative for PD-L1 receiving nivolumab still had better overall survival than those receiving dacarbazine (1-year survival 67· 8% vs37· 4%). In the pivotal KEYNOTE 006 study, pembrolizumab resulted in a higher proportion of patients with melanoma who achieved an objective response and improved survival compared with ipilimumab in all subgroups, with the exception of 18% of patients with PD-L1 negative tumours. 3 The KEYNOTE 001 study4 assessed pembrolizumab in 655 patients with advanced melanoma. A strong correlation was reported between PD-L1 expression and the number of patients with an objective response, up to a maximum level of 65% PD-L1 expression, after which the proportion of responses decreased. Tumour type and histological subtype need to be taken into consideration. In non-small-cell lung cancer, PD-L1 expression was predictive of response for patients with non-squamous non-small-cell lung cancer, 5 whereas for squamous non-small-cell lung cancer, the survival benefit and improvement in objective responses with nivolumab compared with docetaxel were independent of PD-L1 status. 6

Which clinical endpoint should PD-L1 expression be correlated with? In the Checkmate 067 trial7 comparing the combination of nivolumab plus ipilimumab with nivolumab and ipilimumab monotherapies in untreated patients with advanced melanoma, subgroup analysis showed a higher proportion of responses for patients treated with the combination therapy compared with the nivolumab monotherapy, both for those who were PD-L1 positive (72· 1% vs 57· 5%) and those negative for PD-L1 (54· 8% vs 41· 3%). However, the median progression-free survival for patients who were PD-L1 positive was 14 months for combination therapy and single agent nivolumab, but side-effects with the combination was substantially greater. 7 Overall survival data are awaited.