Reproducible metabolomic fingerprinting strengthens postmortem evaluation of insulin intoxication

BackgroundFatal insulin intoxication remains difficult to diagnose because insulin undergoes rapid degradation after death, limiting the reliability of direct biochemical measurements. This creates diagnostic uncertainty when objective molecular confirmation of insulin excess are required. We hypothesised that insulin excess induces systemic metabolic alterations that persist beyond insulin degradation and can be captured using postmortem metabolomics in a forensic setting. MethodsHigh-resolution mass spectrometry (HRMS)-based metabolomics was applied to a national cohort comprising 51 fatal insulin intoxications. Orthogonal partial least squares-discriminant analysis (OPLS-DA) models were trained on cases collected between 2017-2022 to identify insulin-associated metabolite features using a shared-and-unique-structures approach. Performance was evaluated using two temporally distinct test sets (2023-2024): a matched validation cohort and a heterogeneous forensic cohort reflecting biological variability. ResultsHere we show that an insulin-associated metabolomic fingerprint comprising 91 features demonstrated reproducible discrimination across independent cohorts. In the matched cohort (n=59, including 14 insulin cases), insulin intoxication classification achieved 100% sensitivity and 73% specificity within the applicability domain. In the heterogeneous cohort (n=154, including 14 insulin cases), 100% sensitivity was maintained with a 72% specificity despite increased biological variability. Univariate analyses demonstrated significant alterations across multiple metabolite classes, including acylcarnitines, fatty acids/lipids, and purine/nucleoside metabolites, with moderate effect sizes, consistent with systemic effects of insulin-induced hypoglycaemia. ConclusionsFatal insulin intoxication is associated with a reproducible metabolomic fingerprint detectable after death. These findings demonstrate that postmortem metabolomics may serve as a complementary decision-support tool when conventional biomarkers are unreliable.