Background: Curative treatment for unresectable LA NSCLC comprises chemoradiotherapy (CRT) and immunotherapy (IO). However, this uniform approach does not take into account the intrinsic heterogeneity of NSCLC and nearly half of all pts relapse within the first year. This highlights the need to define distinct molecular and phenotypic subgroups of LA NSCLC so that we can interrogate the divergent responses to CRT-IO and identify biomarkers that can guide pt specific strategies. Methods: Consecutive pts treated with CRT +/- IO from May 2010 and recruited to an observational study were retrospectively analysed (N = 336). All pts with adenocarcinoma (ADC) were subject to reflex molecular profiling including NGS and FISH testing for ALK, ROS1, MET and RET. Of note, tracking of bespoke circulating tumour DNA (ctDNA) levels was performed for 25 pts. Plasma samples were collected at longitudinal timepoints pre, during and post CRT (N = 118; median 5 samples/ pt). WES of biopsies and matched normal DNA was used to design tumour informed ctDNA assays to track 16 single nucleotide variants in plasma samples. ctDNA levels were then quantified as mean tumour molecule per mL (MTM/mL). Primary endpoint was progression free survival (PFS). Results: Median age was 61 (IQR 48, 74). Of 336 pts, 74% were male; 32% never smokers; 54% were ADC, 33% squamous cell carcinoma (SCC); 18% had EGFR mutations (EGFRm), 37%EGFR wild type (EGFRw), 40% unknown EGFR status (EGFRu), 5% ALK rearranged, 1% ROS1 rearranged. At median follow-up of 44 months (m), median PFS for the entire cohort was 16.2m. Pts who received consolidation IO had longer PFS [IO, not reached vs no IO, 14.9m; P = 0.05]. Median PFS was shorter for pts with EGFRm [EGFRm, 13m vs EGFRw, 22.5m vs EGFRu, 15.8m]. Patterns of failure were analysed according to local in-field, regional out-field and distant. EGFRm pts were less likely to fail locally in-field [EGFRm, 12% vs EGFRw,18% vs EGFRu, 28%] and significantly more likely to develop distant metastases as first site of failure post CRT [EGFRm, 51% vs EGFRw, 28% vs EGFRu, 19%; P = 0.01]. For the 25 pts with ctDNA analysed, 24 pts had evaluable ctDNA post CRT. Significant heterogeneity was noted at baseline pre CRT [EGFRm, 6.45 vs EGFRw, 7.49 vs SCC, 37.67 MTM/mL; P = 0.05]. No difference was noted in PFS between pts with high vs low baseline ctDNA. Of note, 14 of 24 (58%) pts achieved undetectable ctDNA post CRT and had significantly longer PFS [undetectable, 24.1m vs detectable, 3.5m; P = 0.02]. Conclusions: Here, we demonstrate distinct phenotypic responses to CRT based on molecular subtypes of NSCLC. Crucially, we demonstrate clinical feasibility of deploying tumour-informed ctDNA assays in LA NSCLC with detection of ctDNA post CRT serving as a potential actionable biomarker to guide intensification strategies, providing a window into developing patient-specific combinatorial approaches.