Genomic determinants of metastatic burden and survival in metastatic urothelial carcinoma.

Abstract

4572 Background: Metastatic urothelial carcinoma (mUC) has variable outcomes. Metastatic site count (MSC) correlates with prognosis; the genomic landscape underlying metastatic burden and survival is poorly characterized. We investigated associations among somatic alterations, metastatic patterns, and survival in a large clinical-genomic cohort. Methods: We analyzed 1,157 patients with metastatic bladder cancer by integrating genomic data from the AACR Project GENIE Cohort v18.0-public with clinical annotations and metastatic site information from the MSK-MET database. Associations between 428 genomic alterations and MSC were assessed using Mann-Whitney tests. Survival analyses (n=1,153; 364 events) used Kaplan-Meier estimation, log-rank tests, and Cox regression. Models were adjusted for age, MSC (stratified), and location (lymph, liver, lung). FDR correction was applied, and proportional hazards assumptions verified. Results: Median age was 68.7 years; median OS was 15.8 months. MSC stratification revealed survival differences: 0-1 sites (n=515, median OS not reached), 2-4 sites (n=415, median OS 62.8 months), and >=5 sites (n=223, median OS 13.1 months; log-rank p < 2.5 x 10 -19 ). Each increase in MSC category was associated with worse OS (FDR p < 0.001). MSC predicted 2-year survival with an AUC of 0.78 (95% CI: 0.75–0.82). Six alterations were significantly associated with lower MSC (FDR < 0.05): ERCC2 (n=122; d=0.50, FDR p=5.6 x 10 -6 ), STAG2 (n=128; d=0.40, FDR p=1.9 x 10 -4 ), FBXW7 (n=96; d=0.36, FDR p=7.1 x 10 -3 ), FGFR3 (n=291; d=0.17, FDR p=8.1 x 10 -3 ), TP53BP1 (n=35; d=0.59, FDR p=4.3 x 10 -2 ), and ARID2 (n=70; d=0.41, FDR p=4.6 x 10 -2 ). ERCC2 mutation frequency was 15.7% in 0-1 sites and 4.0% in >=5 sites. Conversely, CDKN2A/B deletions were enriched in patients with >=5 sites (28.3%/27.9%) compared with 0-1 sites (16.5%/15.9%; FDR p < 0.01). Five alterations were independently associated with OS in univariable analysis (FDR < 0.05): CDKN2B deletion (HR 1.81, 95% CI: 1.43–2.28, FDR p=1.7 x 10 -4 ), CDKN2A deletion (HR 1.79, 95% CI: 1.42–2.26, FDR p=1.7 x 10 -4 ), ERCC2 mutation (HR 0.39, 95% CI: 0.24–0.62, FDR p=1.1 x 10 -2 ), CREBBP mutation (HR 0.53, 95% CI: 0.37–0.75, FDR p=3.1 x 10 -2 ), and KDM6A mutation (HR 0.66, 95% CI: 0.52–0.84, FDR p=4.4 x 10 -2 . In multivariable analysis adjusting for age, MSC, and location, no genomic alterations retained independent significance, suggesting MSC mediates these prognostic effects. Penalized regression confirmed MSC as the dominant prognostic variable (HR 1.39 per site increase). Conclusions: DNA damage repair and chromatin modifier mutations are associated with oligometastatic phenotypes and favorable survival, whereas CDKN2A/B deletions correlate with polymetastatic disease. The attenuation of genomic effects after MSC adjustment suggests that metastatic burden is a key mediator. These findings support using genomic profiling to guide mUC prognosis.