Background: The development of biomarkers that reliably predict response to anti-PD-1 mAb therapy (IO) is a critical need. Small (30-150 nm) extracellular vesicles (sEV), aka exosomes, are “molecular mimics” of their parent cells. In cancer, plasma sEV are mixtures of tumor-derived exosomes (TEX) and vesicles produced by non-malignant cells. We conducted a retrospective study of R/M HNSCC patients treated with IO to evaluate sEV as predictive biomarkers. Methods: We evaluated sEV in plasma of R/M HNSCC patients (n = 24) obtained just prior to initiation of IO. Precleared ultrafiltered plasma was used to isolate total plasma sEV by size exclusion chromatography. Total plasma sEV were separated into two subsets, T cell-derived CD3(+) sEV and TEX-enriched CD3(-) sEV by immunocapture with anti-CD3 mAb. On-bead flow cytometry was used to estimate relative levels of proteins carried on the sEV surface. Specifically, the CD3(-) fraction was evaluated for stimulatory (stim; CD40, CD40L, OX40, OX40L, CD80) and suppressive (supp; TGFb, CTLA4, FasL, PD-L1, PD-1) proteins. Differences between responders (CR/PR/SD) and non-responders (PD) were assessed using Wilcoxon-Mann-Whitney test. Multivariate Cox proportional hazards regression was used to evaluate the relationship between the pre-treatment sEV characteristics and outcome. Results: Mean age was 62, 46% had oropharynx (82% HPV+), 17% oral cavity and 29% had larynx/hypopharynx primary. Treatment indication for IO was 46% platinum failure, remainder first line. Best response: 17% PR, 29% SD, and 54% PD. Median PFS and OS were 5.3 and 16.2 months, respectively. Total sEV protein level was not associated with efficacy. Total CD3(+) sEV level was low overall and did not correlate with lymphocyte count or neutrophil/lymphocyte ratio. High CD3(+) sEV level was associated with better OS [HR: 0.16 (95% CI 0.04-0.59), P = 0.007] and PFS [HR: 0.16 (95% CI 0.04-0.58), P = 0.005] but not response. While total TEX and levels of individual stim and supp proteins carried by TEX were not predictive of response, an increased supp/stim ratio was (P = 0.02). Importantly, high TEX supp score and stim score were both associated with better OS [HR: 0.19 (95% CI 0.05-0.77), P = 0.02, and HR: 0.26 (95% CI 0.07-0.95), P = 0.04, resp.]. High PD-L1, but not PD-1 on TEX, was associated with better OS [HR: 0.15 (95% CI 0.04-0.56), P = 0.005]; and PFS: HR: 0.33 (95% CI 0.11-0.94), P = 0.04]. Conclusions: Evaluation of two sEV subsets, T cell-derived-CD3(+) sEV and TEX-enriched CD3(-) sEV, indicated their potential utility as predictive biomarkers with IO. High T cell-derived sEV, TEX supp/stim ratio, and TEX PD-L1 expression levels were independently associated with significantly increased efficacy with IO. Our study uniquely analyzed the predictive value of plasma sEv subsets in IO treated R/M HNSCC patients, and evaluation in a larger cohort is warranted.