436 Background: A subset of mUC exhibits a DNA repair deficiency (DRD) phenotype predicting benefit from platinum based chemotherapy (PBC). We hypothesised that switch maintenance therapy with the PARP inhibitor rucaparib, in patients who have derived clinical benefit from PBC, would improve outcomes for patients with mUC harbouring a DRD biomarker. Methods: ATLANTIS is an adaptive, multi-comparison, phase II trial platform. It tests multiple biomarker selected maintenance therapies for mUC after 4 to 8 PBC cycles without disease progression. Biomarker allocation to the rucaparib comparison was based on ≥10% genomic loss of heterozygosity (%LOH) and/or somatic alteration in defined DRD associated genes ( ATM, BARD1, BRCA1, BRCA2, BRIP1, CDK12, CHEK2, FANCA, NBN, PALB2, RAD51, RAD51B, RAD51C, RAD51D, RAD54L) and/or germline BRCA1 or BRCA2 alteration. Biomarker positive patients were randomised (1:1) to maintenance rucaparib 600 mg BID PO, or matched placebo, within 10 weeks of completing PBC, until disease progression. The primary endpoint was progression free survival (PFS). Statistical analysis (data cut 17/Nov/2021) was pre-planned to target a hazard ratio of 0.5. We selected a 20% 1-sided alpha for this signal seeking phase II trial with 85.4% power (requiring 30 PFS events in 40 patients). PFS (RECIST 1.1) was compared between trial arms, by intention to treat, within a Cox model incorporating baseline minimisation factors. Adverse events (AE) were assessed by CTCAE v4.03. Results: 74 of 279 (26.5%) screened patients were biomarker positive. 40 were randomised within the rucaparib comparison (Dec 2017-Dec 2020). Biomarker positive status was by high %LOH in 22 (55%), DRD gene alteration in 11 (27.5%) and both in 7 (17.5%). Patient characteristics (median age 70.5; 82.5% male; 87.5% bladder primary; 52.5% ECOG PS 0; 62.5% prior cisplatin; 45% visceral metastases) were balanced by treatment arm. 12 (60%) and 20 (100%) PFS events have occurred in the rucaparib and placebo arms respectively (median duration follow up 94.6 weeks in those still alive). Median PFS was 35.3 weeks (80% confidence interval (CI) 11.7-35.6) with rucaparib and 15.1 weeks (80% CI 11.9-22.6) with placebo (hazard ratio 0.53, 80% CI 0.30-0.92, 1 sided p = 0.07). In the safety population (n = 39) treatment related adverse events were mostly low grade. Rucaparib was tolerable with a median duration of 10 rucaparib or 6 placebo cycles on treatment. The most frequent treatment related AEs (all grades) of fatigue (63.2% vs 30.0%, p = 0.03), nausea (36.9% vs 5.0%, p = 0.03) and rash (21.1% vs 0%, p = 0.04) were more common with rucaparib respectively. Conclusions: Maintenance rucaparib, following PBC, extended PFS in DRD biomarker selected patients with mUC and is tolerable. Further investigation of PARP inhibition for mUC is warranted. Clinical trial information: 25859465.