Identifying modifiable factors and their joint associations on late-onset schizophrenia risk in the UK Biobank: a prospective exposure-wide association study

Abstract

Background One in four cases of schizophrenia begins in late life, resulting in high unemployment and reduced life expectancy. However, knowledge of the modifiable risk factors for late-onset schizophrenia and their combined effects is limited. Aims To identify modifiable risk factors for late-onset schizophrenia and estimate their joint disease risk effects. Methods This prospective cohort study using UK Biobank data included 482708 participants without late-onset schizophrenia at baseline, followed up for a mean of 14.36 years. We conducted an exposure-wide association study of 232 potentially modifiable factors linked to late-onset schizophrenia risk. Late-onset schizophrenia is diagnosed using ICD-10 (International Classification of Diseases, 10th Revision) criteria. Cox proportional hazard models identified significant factors across six domains: lifestyle, environment, medical history, physical measures, mental health and socioeconomic status (SES). Domain-specific weighted scores were calculated from Cox model coefficients and stratified into tertiles (favourable, intermediate, unfavourable) for risk assessment. Population attributable fractions (PAFs) quantified prevention potential. Results During follow-up, 1276 participants developed late-onset schizophrenia. We identified 109 significant potentially modifiable factors, with intellectual disability (HR 35.15, 95% CI 11.23 to 110.09), manic episode (HR 33.14, 95% CI 21.16 to 51.90) and bipolar affective disorder (HR 32.91, 95% CI 27.07 to 40.01) showing the strongest risks, while higher household income (>£100000: HR 0.14, 95% CI 0.09 to 0.22), regular friends/family visits (HR 0.23, 95% CI 0.18 to 0.28) and higher hand grip strength (HR 0.35, 95% CI 0.29 to 0.44) showed the strongest protection. PAF estimations indicated that shifting individuals from unfavourable to intermediate/favourable risk profiles could prevent 71.3% (95% CI 71.2% to 71.4%) of late-onset schizophrenia cases, mainly from mental health (25.1%, 95%CI 25.0% to 25.2%), medical history (13.6%, 95%CI 13.5% to 13.7%) and SES domain (11.2%, 95%CI 11.1% to 11.3%); shifting individuals from intermediate/unfavourable risk profiles to favourable could prevent 89.2% of cases. Conclusions A substantial proportion of late-onset schizophrenia risk appears modifiable, with mental health and medical history as key contributors. Physical health and natural environment exposure provided protective benefits. Findings supported integrating clinical interventions and structural changes addressing socioeconomic and environmental factors to reduce late-onset schizophrenia burden.