Insights from large real-world cohorts towards inflammatory versus metabolic phenotypes of schizophrenia spectrum disorders

Background: Schizophrenia spectrum disorders (SSDs) are clinically coherent yet biologically heterogeneous. Converging evidence suggests that systemic inflammation may delineate an “inflammatory biotype”, but the behavioural and clinically relevant correlates of inflammation in large real-world cohorts remain poorly characterized. Methods: This retrospective cohort study utilized the TriNetX Analytics network. Data extraction and analysis were conducted on February 3, 2026. Adults with SSD diagnoses and documented antipsychotic exposure were included. Two cohorts were defined a priori: Higher Severity Schizophrenia Spectrum Disorders (HS-SSD) and Lower Severity Schizophrenia Spectrum Disorders (LS-SSD). The HS-SSD cohort was identified using electronic health record proxy indicators of recent clinical instability, including hospitalization in the last 6 months, intentional self-harm, hospital admission, hallucinations, and unspecified stupor, restlessness and agitation; the LS-SSD cohort met the same diagnostic and treatment criteria in the absence of these severity signs/events. The index event was the first recorded simultaneous presence of diagnostic criteria, antipsychotic treatment, and severity signs/events. Primary outcomes were all-cause mortality, hospitalization rates, and inflammatory state, defined as C-reactive protein (CRP) ≥ 3.00 mg/L and/or lactate dehydrogenase (LDH) ≥ 220 U/L (most recent available measurements). Secondary outcomes included diastolic blood pressure, QTc interval, HbA1c, and LDL cholesterol. Propensity score matching (1:1 nearest-neighbour) was performed on age, sex, ethnicity, sleep disorders, body mass index, and systolic blood pressure, yielding two balanced cohorts (N = 140,862 each). Results: All-cause mortality was 12.38% in HS-SSD and 11.58% in LS-SSD (RR = 1.068; 95% CI = 1.047–1.090; p < 0.001; HR = 1.096; 95% CI = 1.073–1.119). Hospitalizations were 0.56% vs 0.02% (RR = 24.594; 95% CI = 17.272–35.018; p < 0.001). Positive inflammatory markers were 12.55% vs 11.30% (RR = 1.110; 95% CI = 1.086–1.135; p < 0.001). Secondary outcomes differed significantly, with slightly lower mean HbA1c, LDL, blood pressure, and QTc in HS-SSD. Conclusions: The results are consistent with the hypothesis of an inflammatory–behavioural phenotype within the psychosis spectrum, characterized by higher inflammatory burden alongside increased all-cause mortality and a persistence of an instability trajectory.