Background: Leveraging adaptive immunity to control mesothelioma is now a standard approach, however the factors that underpin clinical response are poorly understood. Here we report the final analysis of the CONFIRM trial (NCT03063450), a double-blind phase III randomized study of the PD-1 inhibitor nivolumab (N) versus placebo (P) in patients (pts) with unresectable mesothelioma. Genomic, transcriptomic and multiplex spatial phenotypic correlates were explored in mesotheliomas exhibiting either partial response or progressive disease as their best outcome. Methods: Pts with relapsed pleural or peritoneal mesothelioma with ECOG performance status 0–1 were randomized 2:1 to N (3 mg/kg) or P once every 2 weeks until progression, or a maximum of 12 months. Pts were stratified by epithelioid (E) vs non-epithelioid (NE) histology. Co-primary endpoints were progression-free survival (PFS) and overall survival (OS); key secondary endpoints included best overall response, safety and tolerability. PD-L1 tumour proportion score (TPS, Dako22C3) was evaluated. To decipher correlates response to N, blinded multi-omic analysis was conducted in a subset of responders (R, n=16) versus progressors (NR, n=13). Whole exomes (WES) and transcriptomes were sequenced, with immune landscapes profiled by 19x depth 4-panel multiplex immunofluorescence. Acquired resistance was explored by WES in a biopsied responder at the time of progression. Results: Between April 2017-March 2020, 332 pts were randomised to N (n=221) or P (n=111). Median follow up was 37.2m. Baseline characteristics were balanced between arms. Histology: E 88.3%, 3^rd line or greater 69.9%. PFS was longer for N vs P (events=324; median, 2.9 vs 1.6 months; HR 0.65; 95% CI, 0.51 – 0.82; P<0.001). Crossover from P was 18.0% (due to widespread availability of N during the covid19 pandemic). OS curves crossed at 32m; 9.5 vs 6.8 months; HR, 0.81; 95% CI, 0.64 – 1.04; P=0.096). Grade 3-4 treatment-related toxicity occurred in 20.4% vs 7.2% of pts; discontinuation occurred in 13.6% (N) versus 9.9%. PD-L1 TPS was not predictive of PFS or OS. The R-subgroup had longer PFS versus NR (319 versus 30 days, Hazard ratio, HR 0.02, p<0.001) and OS (670 vs 122 days HR 0.19, p<0.001). The R-group were enriched for BAP1 inactivation, CD8+ T-cell infiltration, T cell co-stimulation, cytolytic activity and mature tertiary lymphoid structures (TLSs). Conversely, the NR group had significantly more aneuploidy (notably uniparental disomy, UPD), and enrichment of epithelial mesenchymal transition, TGF beta signalling, and E2F transcription. Acquired resistance was associated with increased aneuploidy (UPD) and subclonal evolution including a new DNMT3A c2204A>G mutation. Conclusions: N met its co-primary endpoint of PFS with responding mesotheliomas harbouring a TLS enriched, inflamed tumour microenvironment and stable genome compared to non-responders. Clinical trial information: NCT03063450.