Anti-PD-1 therapy in unresectable desmoplastic melanoma: the phase 2 SWOG S1512 trial

Abstract

Desmoplastic melanoma is a distinct subtype of melanoma known to have preexisting immune infiltrates and high ultraviolet light damage, resulting in a high tumor mutational burden. We hypothesized that this may result in high response rates with single-agent anti-programmed death protein 1 (PD-1) therapy. SWOG S1512 was a two-cohort clinical trial testing the activity of pembrolizumab in patients with surgically resectable (cohort A) and unresectable (cohort B) desmoplastic melanoma. Here we report on the cohort B single-arm clinical trial, which enrolled 27 patients with unresectable desmoplastic melanoma receiving pembrolizumab 200 mg intravenously every 3 weeks for up to 2 years, with the primary endpoint of complete response rate. The complete response rate was 37% (95% confidence interval: 19-58%), and the post hoc endpoint of objective response rate was 89% (95% confidence interval: 71-98%). The estimated secondary endpoints of 3-year melanoma-specific progression-free survival and overall survival were 84% and 96%, respectively, with only one patient having died from melanoma progression. Ten patients (37%) experienced grade 3 or 4 adverse events, and nine patients (33%) discontinued treatment because of adverse events. Patients with advanced desmoplastic melanoma have a high response rate to single-agent PD-1 blockade therapy, supporting single-agent anti-PD-1 as the treatment of choice, but are limited by a frequency of toxicities that is numerically higher than in other patient populations. ClinicalTrials.gov identifier: NCT02775851.

Extended Data Fig. 1 |. CONSORT diagram.

Patient eligibility and disposition. PFS: Progression-free survival.

Extended Data Fig. 2 |. Examples of central pathology review of desmoplastic melanoma diagnosis.

Each biopsy had 2 sequential sections stained and imaged for a) S100 or b) H&E, and images were reviewed simultaneously by pathologists to confirm the presence of desmoplastic melanoma by viable tumor cells, shape, and collagenous stroma. In situations with multiple biopsies, pathologists reviewed all sections in parallel to determine consensus annotation.

Extended Data Fig. 3 |. Examples of central pathology review of response to therapy in baseline and on-study biopsies.

Each biopsy had 2 sequential sections stained for S100 (left) or H&E (right), and images were reviewed simultaneously by pathologists to determine presence of a) viable tumor regions with desmoplastic melanoma histology or b) features indicative of tumor regression and immune-mediated pathological response.

Extended Data Fig. 4 |. Tumor mutational burden and recurrently mutated genes.

Whole exome sequencing data was available from biopsies from 17 patients. Tumor mutational burden (TMB, in Mutations [Mut] per Megabase [Mb], top panel) and genes with mutations in at least 25% of cases are shown (bottom panels). Mutations are annotated by their putative impact (fill color) and whether they are present in chromosomal regions with loss-of-heterozygosity (LOH). Patients are further annotated based upon their clinical response by imaging at 9 weeks (complete response [CR], partial response [PR], stable disease [SD], or progressive disease [PD]).

Fig. 1 |. Depth of response and duration of response.

a, Waterfall plot demonstrating the best response, as percentage change in RECIST target lesions, to pembrolizumab. The vertical bars represent the largest percentage decrease or, if no decrease was observed, the smallest percentage increase in the size of target lesions against baseline. The lines indicate the threshold for objective response (≥30% decrease) or disease progression (>20% increase). One patient did not have follow-up disease assessment data and is denoted on the far left with a gray bar at 50% over baseline. b, Swimmer plot of patient outcomes, including timing of PRs and CRs and PFS events. The horizontal bars represent time from registration. Shaded regions of the bars represent duration of protocol treatment. Patients with arrows are alive and progression free. The patient represented by the gray bar at the bottom withdrew consent after one cycle of treatment; no follow-up outcome data were obtained. nivo-ipi, nivolumab plus ipilimumab; pembro, pembrolizumab.

Fig. 2 |. Examples of responses to therapy.

a, Photographic images of the locally advanced scalp desmoplastic melanoma from a patient, at baseline (left) and on two sequential clinic visits, showing response in cutaneous disease. The baseline lesion had bled, which apparently makes it look pigmented, but the primary lesion was not pigmented. b, PET scan images of the same patient at baseline (left) and at 42 days on-therapy (right) showing response in the cutaneous locally advanced primary (imaged photographically in a) and lymph node metastases. c, Images of CT of the chest of another patient at baseline (left) and at 3 monthsʼ follow-up (right) showing response in lung metastasis. CT, computed tomography.

Fig. 3 |. Analysis of PFS and OS.

a, Kaplan–Meier plot of PFS. The 3-year PFS estimate and 95% confidence interval (CI) are reported. One patient underwent resection and has been excluded from the PFS analysis. b, Kaplan–Meier plot of OS. The 3-year OS estimate and 95% CI are reported. c, Melanoma-specific PFS (MSPFS). The 3-year MSPFS estimate and 95% CI are reported. One patient underwent resection and has been excluded from the analysis. d, Melanoma-specific survival (MSS). The 3-year MSS estimate and 95% CI are reported, with the single event being a patient who died from melanoma brain metastases. YR, year.