Background Breast and prostate tumors are known to be less responsive to immune checkpoint inhibitors (ICIs). Tissue-based tumor mutation burden (tTMB) has emerged as a predictive biomarker of response to ICIs, including in these "cold tumors". In clinical practice, when tTMB is not available, blood-based TMB score (bTMB) can be used to consider treatment with ICIs. Methods This retrospective, real-world study included a final cohort of metastatic breast and prostate cancer patients treated with an ICI following a liquid biopsy test. Multiple bTMB-High cut-offs were assessed. Clinical, genomic, and outcomes data were collected. We hypothesized that a cut-off of bTMB ≥10 mut/Mb is not a strong predictor of response to ICIs in this setting. The Guardant Health genomic database (GHGD) was then queried (N = 13,992) for associations of bTMB with genomic alterations. Results In the clinical cohort (N = 48), ICI treatment is offered after a median of 3 (1-9) lines of treatment. The median bTMB is 16.4 (10-186) mut/Mb. The median time on ICI and PFS is 2.1 (0-1.7) and 3.1 months (95%CI, 1.6-4.6) respectively; no difference by MSI/MMR status (p = 0.152). Response rate among eligible patients (n = 36) is 16.7%; only N = 1/6 in bTMB <16 mut/Mb. High bMSI is associated with higher bTMB (correlation test, r = 0.66, p = 0.000). In the GHGD, patients with bTMB high have significantly more alterations than bTMB low and TP53, PIK3CA, ATM, ESR1, NF1, BRCA2, ARID1A, and APC were the most frequently altered genes. Conclusions In this study, the practice of offering an ICIs based on bTMB was uncommon and did not independently predict ICI benefits in patients with refractory, advanced breast and prostate cancers.