Long-Term Outcomes and Responses to Retreatment in Patients With Melanoma Treated With PD-1 Blockade

Abstract

Purpose: To analyze long-term outcomes after treatment discontinuation of anti-programmed death-1 (anti-PD-1) therapy in a cohort of patients with melanoma with the longest follow-up yet available to our knowledge, including a majority of patients treated outside of a clinical trial. We also assessed efficacy of retreatment with anti-PD-1 therapy with or without ipilimumab in relapsing patients.

Methods: We retrospectively analyzed all patients with nonuveal, unresectable stage III/IV melanoma treated with single-agent anti-PD-1 therapy at Memorial Sloan Kettering from 2009-2018 who had discontinued treatment and had at least 3 months of follow-up after discontinuation (n = 396). Overall survival for patients with complete response (CR) was calculated from time of CR. Time to treatment failure for patients with CR was time from CR to the next melanoma treatment or death.

Results: CRs were seen in 102 of 396 patients (25.8%). The median number of months of treatment after CR was zero (range, stopped before CR to 26 months after CR). With a median follow-up of 21.1 months from time of CR in patients who did not relapse, the probability of being alive and not needing additional melanoma therapy at 3 years was 72.1%. There was no significant association between treatment duration and relapse risk. In multivariable analysis, CR was associated with M1b disease and cutaneous versus mucosal or acral primaries. Among the 78 patients (of 396) retreated after disease progression, response was seen in 5 of 34 retreated patients with single-agent anti-PD-1 therapy and 11 of 44 patients escalated to anti-PD-1 plus ipilimumab.

Conclusion: In our cohort, most patients discontinued treatment at the time of CR. Most CRs were durable but the probability of treatment failure was 27% at 3 years. Responses to retreatment were infrequent. The optimal duration of treatment after CR is not yet established.

FIG 1.

Time to treatment failure (TTF) and overall survival (OS) data for the entire cohort and by best overall response (BOR). (A) The median overall TTF was 7.9 months; (B) the median OS was 39 months. (C) TTF from the time of BOR showed a 3-year TTF of 72% for patients with complete response (CR), 26.9% for responders with less than CR, and 3.8% for stable disease (SD). (D) Overall survival from time of BOR showed a 3-year OS of 82.7%, 57.3%, 39%, and 18.9%, respectively, for patients with CR, response less than CR, SD, and progressive disease. Tick marks indicate censored patients. Blue shaded areas in (A) and (B) represent 95% CIs.

FIG 2.

Time to treatment failure and overall survival of patients with complete response (CR). (A) Time to treatment failure and (B) overall survival were calculated from the time of CR. Patients were determined to have a CR by RECIST confirmed by 1 of 2 reference radiologists (RECIST cohort), a radiologic CR that could not be confirmed by the reference radiologists (UC RECIST), or major response with a negative biopsy (pathologic). The “other” group includes patients who had a CR by clinical evaluation but did not have radiologically assessable disease or patients who had a radiologic CR as assessed by an outside radiologist but for whom the outside scans were not available for review at our institution.

FIG 3.

Event-history plot showing outcomes of patients who were retreated with (A) single-agent anti–programmed death-1 (anti–PD-1) therapy or (B) combination anti–PD-1 with anti-CTLA4. Time zero is the start of retreatment. Patients are grouped on the basis of the best overall response to initial anti–PD-1 therapy (complete response [CR], partial response [PR], stable disease [SD], progressive disease [PD]), which is noted on the left side of the panel. The duration of the first course of anti–PD-1 is indicated by light blue, progression of disease; and the reason for discontinuation is indicated by the color (red, suspected CR; teal, toxicity; orange, the course of planned therapy; purple, toxicity; yellow, progression of disease). Red dots indicate time of progression of disease. Open bars indicate nonimmunotherapy systemic treatments. In addition, some patients received radiation therapy (R) or surgery (S). After time zero, colored bars indicate the duration of retreatment. Purple bars show patients who responded; dark blue bars are patients whose best response to retreatment was disease progression. The aqua bars indicate the duration of survival. Patients have either died (red X) or were censored (arrowhead). BOR, best overall response; TFF, time to treatment failure.

FIG A1.

Histogram of (A) months on treatment and (B) No. of anti—programmed death-1 (PD-1) doses given to patients in this analysis.

FIG A2.

Distribution of (A) best overall responses and (B) basis of complete responses (CRs). Of the 76 radiologic CRs, 58 were confirmed by reference radiologists to meet formal RECIST for CR.