This is an open-label, non-randomized, multicenter, international, Phase 2 study of patients with distant or nodal mCSCC (Groups 1 and 3) (see online supplementary file 1, S1 for study sites and principal investigators). Enrollment for Group 3 opened after full enrollment of Group 1. The time point for the primary analysis of data from patients in Group 3 was reached.

Eligible patients were aged ≥18 years with histologically confirmed diagnosis of invasive CSCC, an Eastern Cooperative Oncology Group performance status score of 0 or 1, adequate organ function, and at least one measurable lesion per Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1).21

Key exclusion criteria included ongoing or recent (within 5 years) autoimmune disease requiring systemic immunosuppression, prior treatment with an agent blocking the PD-1/PD ligand-1 (PD-L1) pathway, a history of solid organ transplantation, concurrent cancer (unless indolent or non-life-threatening), or hematologic cancer.

Patients were administered cemiplimab 350 mg intravenously over 30 min Q3W for up to 54 weeks, with the option to extend treatment to 96 weeks (Group 3) or cemiplimab 3 mg/kg intravenously over 30 min Q2W for up to 96 weeks (Group 1). The primary endpoint was ORR per ICR independently in each group. Tumor assessments were performed at the end of each treatment cycle (every 9 weeks for Group 3 and every 8 weeks for Group 1) (see online supplementary file 1, S2 for further details).

Secondary endpoints included ORR per investigator review (INV), duration of response (DOR) by ICR and INV, progression-free survival (PFS) by ICR and INV, overall survival (OS), complete response (CR) rate per ICR, adverse events (AEs), and quality of life. Durable disease control rate, defined as the proportion of patients with response or stable disease for at least 105 days was also examined. Safety assessments included treatment-emergent AEs (TEAEs), laboratory tests, vital signs, and physical examinations. The severity of TEAEs was graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (version 4.03).

Archived tumor samples from prior CSCC biopsies or surgeries were provided during the screening period. TMB was estimated in DNA samples extracted from formalin-fixed paraffin-embedded tumor biopsies. TMB was calculated as the total number of somatic single nucleotide variants and indels in the coding regions of targeted genes per megabase of analyzed genomic sequence (see online supplmentary file 1, S2 for further details).

The study also includes Group 4 and pilot Group 5, which explore alternative doses and/or schedules of cemiplimab, and Group 6 which is designed to confirm the results with 350 mg intravenously Q3W observed in Group 3. Groups 4 to 6 have not reached primary analysis and are not included in this report.

The primary endpoint analyses for each group were statistically independent. Fifty patients in each group were needed to provide at least 85% power to reject a null hypothesis of an ORR of 15% at a two-sided significance level of no more than 5%, if the true ORR was 34%. The primary efficacy analyses were undertaken using 95% binomial exact confidence intervals (CIs), which were generated using the Clopper-Pearson method.22 The secondary efficacy analyses of DOR, PFS, and OS are summarized by their medians and 95% CIs, which were generated by the Kaplan–Meier method. CR rates are summarized descriptively. The full analysis set, which included all patients who passed screening and were eligible for study participation, was used for the analysis of all efficacy endpoints. The safety analysis set included all enrolled patients who received any study drug. The primary efficacy analysis of both groups was performed 6 months after the first dose of cemiplimab had been administered to the last patient enrolled; the results from the primary efficacy analysis of Group 1 (cut-off date of October 27, 2017) have been previously published.18 In the present analysis, data are reported separately for Group 1 and Group 3, as well as for both groups combined. The cut-off date for the primary analysis of Group 3 and follow-up analysis of Group 1 was September 20, 2018.

Results are presented for Group 3 primary analysis, and for Group 1 with approximately 11 months additional follow-up after the primary analysis. In total, 56 patients were enrolled and treated with cemiplimab 350 mg Q3W (Group 3) from July 2017 through March 2018, and 59 patients were enrolled and treated with cemiplimab 3 mg/kg Q2W (Group 1) from March 2016 through January 2017. Baseline characteristics were similar in regard to median age, gender, and extent of disease (nodal vs distant) (table 1). The median duration of follow-up was 8.1 (range, 0.6 to 14.1) months for Group 3 and 16.5 (range, 1.1 to 26.6) months for Group 1. Patients in Group 3 received a median of 11.5 (range, 1 to 20) doses of cemiplimab and were exposed to treatment for a median of 34.3 (range, 2.6 to 60.4) weeks. Patients in Group 1 received a median of 31.0 (range, 1 to 48) doses of cemiplimab and were exposed to treatment for a median of 65.0 (range, 2.0 to 96.1) weeks. Patient disposition is summarized in online supplementary file 2, figure S1.

For the Group 3 primary analysis, ORR per ICR was 41.1% (95% CI, 28.1% to 55.0%). For the Group 1 update, ORR per ICR was 49.2% (95% CI, 35.9% to 62.5%). The combined ORR for both groups was 45.2% (95% CI, 35.9% to 54.8%), including 39 partial responses (PRs) and 13 CRs per ICR (table 2). The ORR per INV was 51.8% (95% CI, 38.0% to 65.3%) in Group 3 and 49.2% (95% CI, 35.9% to 62.5%) in Group 1 (see online supplementary file 3, table S1).

Per ICR, most evaluable patients in both Group 3 and Group 1 had a decrease from baseline in the target lesion diameters (figure 1), and most responses were evident at the first response assessment in both groups (see online supplementary file 2, figure S2). Durable responses are most evident in Group 1 due to longer follow-up and are emerging for Group 3 (see online supplementary file 2, figure S2). The median DOR had not been reached in either group at data cut-off. Per ICR, the Kaplan–Meier estimate for DOR at 8 months was 95.0% (95% CI, 69.5% to 99.3%) in responding patients in Group 3, and at 12 months was 88.9% (95% CI, 69.3% to 96.3%) in responding patients in Group 1 (table 2).

The median time to response per ICR was 2.1 (range, 2.0 to 8.3) months for Group 3, 1.9 (range, 1.7 to 9.1) months for Group 1, and 2.1 (range, 1.7 to 9.1) months for both groups combined. The disease control rate per ICR was 64.3% (95% CI, 50.4% to 76.6%) in Group 3, 71.2% (95% CI, 57.9% to 82.2%) in Group 1, and 67.8% (95% CI, 58.5% to 76.2%) in both groups combined. The durable disease control rate per ICR was 57.1% (95% CI, 43.2% to 70.3%) in Group 3, 61.0% (95% CI, 47.4% to 73.5%) in Group 1, and 59.1% (95% CI, 49.6% to 68.2%) in both groups combined (table 2).

Median PFS had not been reached at the time of data cut-off. The median Kaplan–Meier estimated PFS per ICR was 10.4 (95% CI, 3.6 to not evaluable (NE)) months based on 44.6% event rate for Group 3, 18.4 (95% CI, 6.8 to NE) months based on 47.5% event rate for Group 1, and 18.4 (95% CI, 7.3 to NE) months based on 46.1% event rate for both groups combined. The Kaplan–Meier estimation of PFS at 12 months per ICR was 47.4% (95% CI, 29.6% to 63.3%) for Group 3, 52.9% (95% CI, 39.0% to 65.0%) for Group 1, and 51.2% (95% CI, 41.0% to 60.6%) for both groups combined (see online supplementary file 2, figure S3A).

Median OS had not been reached in either group at data cut-off. The Kaplan–Meier estimation of OS at 12 months was 76.1% (95% CI, 56.9% to 87.6%) for Group 3, 81.3% (95% CI, 68.7% to 89.2%) for Group 1, and 80.7% (95% CI, 71.9% to 87.1%) for both groups combined (see online supplementary file 2, figure S3B).

For Group 3, ORR per ICR was explored for patients with high body weight (>120 kg). Among four patients who weighed >120 kg at baseline, two experienced PR per ICR (see online supplementary file 3, table S2).

In both groups combined, 113 (98.3%) patients experienced at least one TEAE of any grade regardless of attribution, including 96.4% for Group 3% and 100.0% for Group 1 (table 3). The most common TEAEs in Group 3, Group 1, and both groups combined, were fatigue (28.6%, 25.4%, and 27.0%, respectively), diarrhea (17.9%, 28.8%, and 23.5%, respectively), and nausea (17.9%, 23.7%, and 20.9%, respectively). Grade ≥3 TEAEs regardless of attribution were reported in 45.2% of patients in both groups combined, with the most common in Group 3, Group 1, and both groups combined, being anemia (8.9%, 3.4%, and 6.1%, respectively), fatigue (5.4%, 1.7%, and 3.5%, respectively) and pneumonitis (0.0%, 5.1%, and 2.6%) (table 3). Treatment-related AEs (TRAEs) were reported in 71.3% (82/115) of patients, most commonly fatigue (13.0% (15/115)) (table 3; online supplementary file 3, table S2). TRAEs reported in ≥5% of patients in either treatment group are shown in online supplementary file 3, table S3. Immune-related AEs per INV are presented in online supplementary file 3, table S4.

Three (5.4%) patients in Group 3 discontinued treatment due to an AE (Grade 3 soft tissue necrosis, n=1; Grade 2 lethargy, n=1; and Grade 3 psoriasis, n=1). In Group 1, six (10.2%) patients discontinued treatment due to an AE; four of these were previously reported,18 and two occurred after the data cut for the Group 1 primary analysis. All AEs leading to treatment discontinuations were considered treatment-related, except for the patient in Group 3 with soft tissue necrosis on the head. One (1.8%) patient in Group 3 died due to arterial hemorrhage from their right lower extremity tumor which measured 12.5 cm in longest diameter at baseline. This death was not considered related to study treatment. No new AEs resulting in death were reported in the updated analysis of Group 1.