Response and outcomes with immune checkpoint inhibitor (ICI) in patients (pts) with urothelial carcinoma (UC) and subtype histology (SH).

4580 Background: ICIs are used for UC in different settings. Most cases have pure urothelial carcinoma (PUC) but about a third have UC mixed with SH. We examined pts with UC treated with ICI and hypothesized that outcomes would not differ between PUC and mixed with SH. Methods: We retrospectively included pts with PUC and any SH component treated with ICI as adjuvant, 1 st line 1L upfront, maintenance avelumab mAV, ≥2 nd line 2+L therapy; pts were divided by SH type. Pts with multiple SH were included as separate group. We excluded pts with pure non-UC histology and pts treated with neoadjuvant ICI. We calculated overall response rate (ORR), median overall, progression-free and disease-free survival mOS, mPFS, mDFS using KM method from ICI start. Analysis was stratified by therapy line; multivariable models were adjusted by Khaki factors for 1L/upfront, mAV and Bellmunt factors for 2+L; reference group was PUC. We used descriptive statistics for small subsets. Results: We included 1525 pts from 26 sites; n=752 1L, n=609 2+L; 76% men, 80% White, 69% smoking history, 80% lower urinary tract primary, 28% SH. Median f/u time from 1L ICI start 24 months (mo): mPFS was 3 (95%CI 2.8-3.2) mo for PUC (n=340), 3 (95%CI 2-4) mo for squamous (n=60) (HR=1, 95%CI 0.7-1.4), 2 (95%CI 1.4-3.2) mo for micropapillary (n=36) (HR=1.3 95% 0.9-2) and 4 (3.5-4.5) mo for ≥2 SH (n=20) (HR=0.9 95%CI 0.6-1.5). Median f/u time from 2+L ICI start 26 mo: mPFS was 3 (95%CI 2.4-3.2) mo for PUC (n=264), 4 (95%CI 2-5) mo for squamous (n=33) (HR=0.8, 95%CI 0.6-1.2), 3 (95%CI 1.6-3.7) mo for micropapillary (n=17) (HR=0.9 95% 0.6-1.7) and 3 (2.5-3.1) mo for ≥2 SH (n=10) (HR=1.2 95%CI 0.6-2.2); Table shows ORR/OS data. Data on adjuvant and mAV settings will also be presented. Conclusions: We found no significant survival difference between PUC and SH. ORR was numerically higher with sarcomatoid SH and ≥2 SH in the 1L upfront setting. Limitations: retrospective design, small sample size in a few subsets, lack of randomization and central scan/path (% SH) review, selection bias, unmeasured confounders. Table: see text