A year of the immunotherapy nivolumab (Opdivo) given after lung cancer surgery did not lengthen the time patients lived free of recurrence, according to the federally sponsored ANVIL trial published in JAMA. The negative result stands apart from a string of positive adjuvant immunotherapy studies and tempers expectations for checkpoint blockade in earlier-stage disease.

ANVIL (NCT02595944), sponsored by the National Cancer Institute, was a randomized phase 3 study run at 378 sites in the US National Clinical Trials Network. Between May 2016 and September 2019 it enrolled patients with resected stage IB (at least 4 cm), II, or IIIA non-small cell lung cancer without sensitizing EGFR or ALK alterations, all of whom had completed planned adjuvant chemotherapy and/or radiotherapy. In total, 935 patients were randomized: 466 to nivolumab (480 mg intravenously every 4 weeks for up to one year) and 469 to observation.

No benefit on the co-primary endpoint

Disease-free survival, a co-primary endpoint alongside overall survival, showed no advantage. Median disease-free survival was 71.3 months with nivolumab versus 68.8 months with observation (hazard ratio for progression or death, 0.97; 97% CI, 0.79-1.20; 1-sided P = .39). In the prespecified subgroup with PD-L1 expression of 50% or higher, the hazard ratio was 0.86 (98% CI, 0.55-1.34; P = .22), also not significant.

“Adjuvant nivolumab was not associated with improved disease-free survival,” the authors concluded.

Because overall survival was to be tested only if disease-free survival was positive, it was not formally analyzed; an exploratory look showed a hazard ratio of 1.02 (95% CI, 0.82-1.26) in the overall population after a median follow-up of 72.6 months. Grade 3 to 5 treatment-related toxicities occurred in 116 nivolumab patients (25%), including two fatal respiratory events.

The result contrasts with adjuvant immunotherapy regimens that have won approval, and the authors note that nivolumab here was given after, rather than concurrent with, chemotherapy. The wide confidence interval crossing 1.0 leaves little room for a hidden benefit. Association is not causation, and these data do not constitute medical advice.