Prognostic impact of radiological tumor burden in patients with metastatic urothelial carcinoma treated with pembrolizumab

•Clinical prognosticators are investigated in metastatic urothelial carcinoma receiving pembrolizumab.•Radiological tumor burden was represented by baseline tumor size (BTS) and baseline tumor number.•Large BTS and high baseline tumor number were independently associated with worse overall survival and cancer-specific survival.•A linear relationship was observed between BTS and survival outcomes.•BTS was not predictive of response to pembrolizumab. Radiological tumor burden has been reported to be prognostic in many malignancies in the immunotherapy era, yet whether it is prognostic in patients with metastatic urothelial carcinoma (mUC) treated with pembrolizumab remains uninvestigated. We sought to assess the predictive and prognostic value of radiological tumor burden in patients with mUC. We performed a retrospective analysis of 308 patients with mUC treated with pembrolizumab. Radiological tumor burden was represented by baseline tumor size (BTS) and baseline tumor number (BTN). Optimal cut-off value of BTS was determined as 50 mm using the Youden index (small BTS: n = 194, large BTS: n = 114). Overall (OS), cancer-specific (CSS), progression-free survival (PFS), and objective response rate (ORR) were compared. Non-linear associations between BTS and OS and CSS were evaluated using restricted cubic splines. Patients with large BTS were less likely to have undergone the surgical resection of the primary tumor (P = 0.01), and more likely to have liver metastasis (P < 0.001) and more metastatic lesions (P < 0.001). On multivariable analyses controlling for the effects of confounders (resection of primary tumor, metastatic site, number of metastases and lactate dehydrogenase level), large BTS and high BTN were independently associated with worse OS (HR 1.52; P = 0.015, and HR 1.69; P = 0.018, respectively) and CSS (HR 1.59; P = 0.01, and HR 1.66; P = 0.031, respectively), but not PFS. Restricted cubic splines revealed BTS was correlated with OS and CSS in linear relationships. Additionally, large BTS was significantly predictive of lower ORR and complete response rate on univariable analyses (P = 0.041 and P = 0.032, respectively), but its association disappeared on multivariable analyses. Radiological tumor burden has independent prognostic value with a linear relationship in pembrolizumab-treated patients with mUC and might help drive the earlier introduction of second-line pembrolizumab and/or switching to subsequent therapies.