A comparative atudy of immune and neuroendocrine protein biomarkers with latent profiles in all-cause hospitalisation: a time-to-event analysis

Early identification of individuals in the general population at risk for hospitalisation is important to reducing public health burden. Immune and neuroendocrine proteins are robust indicators of disease, but their efficacy in epidemiological studies has not consistently translated to clinical trials. Subgroup identification among the population, while capturing complex patterns of biomarker expression through latent profile analysis, may improve predictive accuracy for clinical outcomes compared to individual biomarkers alone. Immune-neuroendocrine biomarkers (C-reactive protein [CRP]; fibrinogen [Fb]; white blood cell counts [WBCC]; and insulin growth-factor-1 [IGF-1]) were collected in 2008 in the English Longitudinal Study of Ageing (ELSA). Administration data on hospitalisations, linked to ELSA, derived from the Hospital Episode Statistics (HES). Hospitalisation for 12 disease classes of varying sample sizes (n=3163–4276) was monitored from 2008–18 (Mage=66·3 ± 9·4 at baseline). Analyses were adjusted for genetic, demographic, socioeconomic, lifestyle, and clinical variables selected a priori. A three-class latent profile solution offered the most parsimonious fit to the data (low-risk [52·43%]; moderate-risk [35·89%]; high-risk [11·68%] inflammatory status). Profiles offered greater specificity than individual biomarkers in risk for all-cause hospitalisation, with effect sizes from the high-risk profiles being larger for risk of hospitalisation for sleep, blood, infection, and musculoskeletal conditions. A risk gradient was seen in sleep, circulatory, endocrine, and genitourinary disorders, with the magnitude of effects being notably higher in the high-risk group. In one instance, high-risk profile membership, but not moderate-risk, was associated with a 38% greater hazard of hospitalisation for infectious disorders 10 years later (HR 1·38, 95% CI 1·08–1·78, p=0·011). There were no associations between individual biomarkers or profiles on hospitalisation for digestive, nervous, and skin disorders. For the first time, we show that immune-neuroendocrine profiles better characterise disease risk to the point of hospitalisation, beyond individual biomarkers. A precision medicine approach may enable more precise risk-stratification, and subgroup-specific analyses, with risk estimates less susceptible to dilution or inflation. By ensuring that resources are directed towards those most susceptible to disease, these strategies have the potential to make hea