Mortality following cancer diagnosis among people with non-affective psychoses: mediation by stage at diagnosis and time to treatment initiation

Background: People with psychotic disorders have greater mortality rates following cancer diagnosis, compared to people without psychotic disorders. Prior examinations of the effect of psychotic disorders on survival following cancer diagnosis mediated by stage at diagnosis and treatment disparities did not accommodate for multiple mediators and post-exposure confounding. Methods: We identified cases of cancer diagnosed between 1995 and 2019 within a group of people with non-affective psychotic disorders (NAPD) and people without NAPD, constructed using Ontario health administrative data held by Institute for Clinical Evaluative Sciences (ICES). Death from any cause was identified using register data. Inverse probability of treatment weighted Cox models were used to estimate the effect of NAPD on mortality following cancer diagnosis mediated by stage at diagnosis and time to treatment initiation, adjusting for relevant confounders. Results: The cohort included 3,643 people with NAPD and 15,174 people without NAPD who developed cancer. People with NAPD had a 66% greater adjusted hazard of all-cause mortality than people without NAPD (95%CI for the adjusted hazard-ratio: 1.55,1.78). The HR estimate for the indirect effect mediated through stage at diagnosis was 1.09 (95%CI=1.05,1.13) and HR estimate for the indirect effect mediated through time to treatment initiation was 1.00 (95%CI =.96,1.04). Conclusions: Our findings suggest that a relatively small proportion of the effect of NAPD on mortality is mediated by stage at diagnosis, while time to treatment initiation does not mediate that effect. This excess risk is potentially mediated by other patient, provider, and system-related factors in cancer care.