Neoadjuvant immunotherapies hold promise in muscle-invasive bladder cancer (MIBC). To report on 2-yr disease-free (DFS) and overall (OS) survival including novel tissue-based biomarkers and circulating tumor DNA (ctDNA) in the ABACUS trial. ABACUS was a multicenter, single-arm, neoadjuvant, phase 2 trial, including patients with MIBC (T2-4aN0M0) who were ineligible for or refused neoadjuvant cisplatin-based chemotherapy. Two cycles of atezolizumab were given prior to radical cystectomy. Serial tissue and blood samples were collected. The primary endpoints of pathological complete response (pCR) rate and dynamic changes to T-cell biomarkers were published previously. Secondary outcomes were 2-yr DFS and OS. A biomarker analysis correlated with relapse-free survival (RFS) was performed, which includes FOXP3, major histocompatibility complex class I, CD8/CD39, and sequential ctDNA measurements. The median follow-up time was 25 mo (95% confidence interval CI 25-26). Ninety-five patients received at least one cycle of atezolizumab. Eight patients did not undergo cystectomy (only one due to disease progression). The pCR rate was 31% (27/88; 95% CI 21-41). Two-year DFS and OS were 68% (95% CI 58-76) and 77% (95% CI 68-85), respectively. Two-year DFS in patients achieving a pCR was 85% (95% CI 65-94). Baseline PD-L1 and tumor mutational burden did not correlate with RFS (hazard ratio HR 0.60 95% CI 0.24-1.5, p = 0.26, and 0.72 95% CI 0.31-1.7, p = 0.46, respectively). RFS correlated with high baseline stromal CD8+ (HR 0.25 95% CI 0.09-0.68, p = 0.007) and high post-treatment fibroblast activation protein (HR 4.1 95% CI 1.3-13, p = 0.01). Circulating tumor DNA positivity values at baseline, after neoadjuvant therapy, and after surgery were 63% (25/40), 47% (14/30), and 14% (five/36), respectively. The ctDNA status was highly prognostic at all time points. No relapses were observed in ctDNA-negative patients at baseline and after neoadjuvant therapy. The lack of randomization and exploratory nature of the biomarker analysis are limitations of this work. Neoadjuvant atezolizumab in MIBC is associated with clinical responses and high DFS. CD8+ expression and serial ctDNA levels correlated with outcomes, and may contribute to personalized therapy in the future. We showed that bladder cancer patients receiving immunotherapy followed by cystectomy have good long-term outcomes. Furthermore, we found that certain biological features can predict patients who might have particular benefit from this therapy.