9502 Background: Desmoplastic melanoma (DM) is a rare cancer defined by a dense fibrous collagen matrix. It is associated with high UV exposure leading to a high mutational load. When locally advanced, the standard of care is wide excision and radiation therapy due to its propensity for local relapses. Metastatic DM has a high response rate (RR) to PD-1 blockade therapy (Eroglu et al. Nature 2018). We hypothesized that neoadjuvant treatment with anti-PD-1 monotherapy may induce pathologically confirmed regressions in a high percentage of cases, potentially allowing for less extensive local treatment. Methods: Patients > 18 years old with histologically confirmed resectable (primary, recurrent, or regional lymph node metastasis) DM with clinical evidence of residual disease received pembrolizumab 200 mg q3 weeks (wk) 3 followed by excision. No adjuvant therapy was administered. Primary endpoint: pathological complete response (pCR), with the assumption that pCR of 25% would be considered a positive result worthy of future study. To test this hypothesis, a single arm trial with 25 eligible patients would have a 3.4% probability of a positive result with a true pCR of 5%, and a power of 90% of a positive result if the true pCR is 25%. Secondary endpoints: clinical RR by imaging and clinical exam, median overall survival (OS), and evaluation of safety/tolerability of neoadjuvant pembrolizumab. Adverse events were assessed q3 wk. Disease assessments occurred at baseline and q9 wk. NCT02775851. Results: We enrolled-29 eligible patients with resectable DM. One patient refused treatment and was omitted from further analysis. Median age was 75, 79% were male, primary sites of disease were 72% H&N, 10% torso, 14% extremities, 3% unknown. No patients received prior systemic therapy. Mean time from C1D1 treatment to surgery was 84.2 (range: 52-135) days, mean number of cycles received 3.3 (range: 2-4). 26/27 (93%) of patients underwent wide excision of the resectable disease, of which 14 (54%) underwent sentinel lymph node biopsy. One patient underwent resection of a nodal recurrence thus did not require wide excision. pCR was noted in 15/27 (56%) of patients (95% CI: 35%-75%). One patient without a pCR had a major pathologic response with 0.2 mm residual melanoma. In addition, one patient with a clinical CR did not undergo resection by choice. None became inoperable. Clinical RR was 52% (95% CI: 32%-71%). Median OS has not been reached, with two nontreatment related deaths (acute hypoxic respiratory failure; unknown). No > grade 2 related adverse events were observed. Conclusions: Neoadjuvant pembrolizumab in resectable DM results in a high pCR rate with excellent tolerance, which supports consideration of PD-1 blockade therapy prior to surgery. Funding: U10CA180888 and U10CA180819; and in part by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA. Clinical trial information: NCT02775851.