•OS was numerically longer with pembrolizumab in all subgroups.•PFS was similar between treatment arms in all subgroups.•More patients receiving pembrolizumab versus SOC had an objective response.•Responses with pembrolizumab were durable.•Results suggest benefit from pembrolizumab regardless of recurrence pattern. In the phase 3 KEYNOTE-040 study, pembrolizumab prolonged OS versus chemotherapy in previously treated recurrent or metastatic (R/M) HNSCC. We present a post hoc subgroup analysis by disease recurrence pattern: recurrent-only, recurrent and metastatic (recurrent-metastatic), and metastatic-only HNSCC. Patients had HNSCC that progressed during or after platinum-containing treatment for R/M disease or had recurrence or progression within 3–6 months of previous platinum-containing definitive therapy for locally advanced disease. Patients were randomly assigned (1:1) to pembrolizumab 200 mg Q3W or investigator’s choice of standards of care (SOC): methotrexate, docetaxel, or cetuximab. Outcomes included OS, PFS, ORR, and DOR. The data cutoff was May 15, 2017. There were 125 patients (pembrolizumab, 53; SOC, 72) in the recurrent-only subgroup, 204 in the recurrent-metastatic subgroup (pembrolizumab, 108; SOC, 96), and 166 in the metastatic-only subgroup (pembrolizumab, 86; SOC, 80). The hazard ratio (95% CI) for death for pembrolizumab versus SOC was 0.83 (0.55–1.25) in the recurrent-only, 0.78 (0.58–1.06) in the recurrent-metastatic, and 0.74 (0.52–1.05) in the metastatic-only subgroups. PFS was similar between treatment arms in all subgroups. ORR was 22.6% for pembrolizumab versus 16.7% for SOC in the recurrent-only, 10.2% versus 6.3% in the recurrent-metastatic, and 15.1% versus 8.8% in the metastatic-only subgroups. DOR was numerically longer with pembrolizumab in all subgroups. Pembrolizumab provided numerically longer OS and durable responses in all subgroups compared with SOC, suggesting that patients with previously treated R/M HNSCC benefit from pembrolizumab regardless of recurrence pattern.