Background: Circulating tumor DNA (ctDNA)-based MRD assays can detect clinically occult disease with high sensitivity and specificity in patients with solid tumors. Previous trials that evaluated MRD testing in bladder cancer, such as ABACUS and IMVigor010, enrolled patients with a predominant urothelial carcinoma component. Given the known histological and molecular heterogeneity in urothelial carcinoma, we examined the prognostic value of a tumor-informed MRD assay in bladder cancer patients with mixed or pure variant histologies (M/VH). Methods: This retrospective study analyzed the results of post-cystectomy longitudinal ctDNA testing from a multi-institutional database of patients with urothelial carcinoma (n=205). A personalized, tumor-informed assay (Signatera, Natera, Inc.) was used for the detection and quantification of ctDNA. Patients with M/VH and at least 3 months of postoperative follow-up data on relevant demographic, clinical, pathologic, and NGS variables were eligible for inclusion. Time-dependent Cox regression analyses were performed for disease-free survival (DFS) based on ctDNA during the MRD window (<12 weeks from surgery, N=29) and surveillance window (>12 weeks post-surgery or post-AT, N=29). Multivariable regression was performed to adjust hazards for clinicopathologic factors predicting DFS. Results: Of the 205 patients, 46 (74% male, median age: 69 years) met the inclusion criteria for analysis. Whole exome sequencing of tumor tissue revealed that TP53 (45.7%), ARID1A (32.6%), and KDM6A (30.4%) were the top three genes with somatic mutations. Twenty-seven (59%) patients received neoadjuvant therapy, 37% of whom (10/27) also received adjuvant therapy. ctDNA-positivity during the MRD window and during surveillance was strongly associated with poor DFS (MRD window: HR=4.93, 95%CI: 1.77-20.77, p=0.03; surveillance: HR=28.3, 95% CI: 3.52-3660, p<0.001). None of the serially ctDNA-negative patients (N=25) recurred, whereas 58.9% (10/17) of patients who were ctDNA-positive anytime post-surgery recurred at a median follow-up of 9 months from the date of surgery, range: 3.1-36.4 months). When adjusted for all other risk factors, ctDNA-positivity anytime post-surgery was the only significant predictor of recurrence (HR=55.26, 95%CI: 5.3-7648, p<0.00001). Conclusions: Personalized, tumor-informed ctDNA testing can identify patients with mixed or variant histology bladder cancer who are at high risk of recurrence. Further studies are warranted to evaluate the potential benefit of ctDNA-guided adjuvant therapy in this patient population.