Background: Anti-PD-1 therapy (hereafter referred to as anti-PD-1) induces long-term disease control in approximately 30% of patients with metastatic melanoma; however, two-thirds of patients are resistant and will require further treatment. We aimed to determine the efficacy and safety of ipilimumab plus anti-PD-1 (pembrolizumab or nivolumab) compared with ipilimumab monotherapy in patients who are resistant to anti-PD-(L)1 therapy (hereafter referred to as anti-PD-[L]1).

Methods: This multicentre, retrospective, cohort study, was done at 15 melanoma centres in Australia, Europe, and the USA. We included adult patients (aged ≥18 years) with metastatic melanoma (unresectable stage III and IV), who were resistant to anti-PD-(L)1 (innate or acquired resistance) and who then received either ipilimumab monotherapy or ipilimumab plus anti-PD-1 (pembrolizumab or nivolumab), based on availability of therapies or clinical factors determined by the physician, or both. Tumour response was assessed as per standard of care (CT or PET-CT scans every 3 months). The study endpoints were objective response rate, progression-free survival, overall survival, and safety of ipilimumab compared with ipilimumab plus anti-PD-1.

Findings: We included 355 patients with metastatic melanoma, resistant to anti-PD-(L)1 (nivolumab, pembrolizumab, or atezolizumab), who had been treated with ipilimumab monotherapy (n=162 [46%]) or ipilimumab plus anti-PD-1 (n=193 [54%]) between Feb 1, 2011, and Feb 6, 2020. At a median follow-up of 22·1 months (IQR 9·5-30·9), the objective response rate was higher with ipilimumab plus anti-PD-1 (60 [31%] of 193 patients) than with ipilimumab monotherapy (21 [13%] of 162 patients; p<0·0001). Overall survival was longer in the ipilimumab plus anti-PD-1 group (median overall survival 20·4 months [95% CI 12·7-34·8]) than with ipilimumab monotherapy (8·8 months [6·1-11·3]; hazard ratio [HR] 0·50, 95% CI 0·38-0·66; p<0·0001). Progression-free survival was also longer with ipilimumab plus anti-PD-1 (median 3·0 months [95% CI 2·6-3·6]) than with ipilimumab (2·6 months [2·4-2·9]; HR 0·69, 95% CI 0·55-0·87; p=0·0019). Similar proportions of patients reported grade 3-5 adverse events in both groups (59 [31%] of 193 patients in the ipilimumab plus anti-PD-1 group vs 54 [33%] of 162 patients in the ipilimumab group). The most common grade 3-5 adverse events were diarrhoea or colitis (23 [12%] of 193 patients in the ipilimumab plus anti-PD-1 group vs 33 [20%] of 162 patients in the ipilimumab group) and increased alanine aminotransferase or aspartate aminotransferase (24 [12%] vs 15 [9%]). One death occurred with ipilimumab 26 days after the last treatment: a colon perforation due to immune-related pancolitis.

Interpretation: In patients who are resistant to anti-PD-(L)1, ipilimumab plus anti-PD-1 seemed to yield higher efficacy than ipilimumab with a higher objective response rate, longer progression-free, and longer overall survival, with a similar rate of grade 3-5 toxicity. Ipilimumab plus anti-PD-1 should be favoured over ipilimumab alone as a second-line immunotherapy for these patients with advanced melanoma.