Medication Exposure and Mortality in Patients With Schizophrenia

Abstract

Importance: The use of antipsychotics, antidepressants, and benzodiazepines may influence the risk of mortality in people with schizophrenia. However, many observational studies have not accounted for immortal time bias (ITB), which occurs when there is a period during which patients in the exposed group are necessarily alive and misclassified as exposed (the period between start of follow-up and initiation of drug). Ignoring ITB may lead to misinterpretation of the association between these drugs and mortality.

Objectives: To examine whether the cumulative dose of antipsychotics, antidepressants, and benzodiazepines is associated with mortality risk in patients with schizophrenia and discuss the potential impacts of ignoring ITB.

Design, setting, and participants: This cohort study used administrative data from Québec, Canada, including patients aged 17 to 64 years diagnosed with schizophrenia between January 1, 2002, and December 31, 2012. Data analysis was performed from June 22, 2022, to September 30, 2024.

Main outcomes and measures: The primary outcome was all-cause mortality, with follow-up from January 1, 2013, to December 31, 2017, or until death. Mortality risk was assessed for low, moderate, and high exposure to antipsychotics, antidepressants, and benzodiazepines. Cox proportional hazards regression models with time-fixed exposure (not controlling for ITB) and time-dependent exposure (controlling for ITB) were performed.

Results: The cohort included 32 240 patients (mean [SD] age, 46.1 [11.6] years; 19 776 [61.3%] men), of whom 1941 (6.0%) died during follow-up. No dose-response association was found for antipsychotics with mortality using the time-fixed method. However, high-dose antipsychotic use was associated with increased mortality after correcting for ITB (adjusted hazard ratio [AHR], 1.28; 95% CI, 1.07-1.55; P = .008). Antidepressants showed a reduced mortality risk using the time-fixed method, but only at high doses when correcting for ITB (AHR, 0.86; 95% CI, 0.74-1.00; P = .047). Benzodiazepines were associated with increased mortality risk regardless of the method.

Conclusions and relevance: The findings of this study do not dispute the known efficacy of antipsychotics in schizophrenia, but they call into question the magnitude of long-term mortality benefits.

Figure 1.. Illustration of the Immortal Time Bias (ITB) in the Context of Drug Exposure and Mortality

Period during which the outcome of the exposure cannot occur (A), and time-dependent method of analysis (B). The ITB is a type of bias occurring when, among the exposed group, there is a follow-up period between study entry and the determination of the exposure variable in which the outcome cannot occur, resulting in an artificially lower incidence of the event in the exposed group compared with the unexposed group. Another way of seeing the ITB is the following: ITB occurs when, during follow-up, cohort members develop the outcome of interest before they had the chance to be exposed to the drug because the follow-up period was too short. Consequently, patients who survive longer are more likely to be defined as exposed, leading to an artificial protection of exposed patients. DDD indicates defined daily dose.

Figure 2.. Comparative Results of the Risk Estimates for All-Cause Mortality Between Studies and Methods

Cox proportional hazards regression model with time-fixed exposure not corrected for immortal time bias (ITB) and with time-dependent exposure corrected for ITB for antipsychotics, antidepressants, and benzodiazepines (N = 32 240). NS indicates nonsignificant; S, significant (P < .05).

Figure 3.. Adjusted Hazard Ratios (AHRs) for All-Cause Mortality for Cox Proportional Hazards Regression Models

Time-fixed (not corrected for immortal time bias [ITB]) and time-dependent (corrected for ITB) exposure for antipsychotics, antidepressants, and benzodiazepines (N = 32 240). Error bars indicate 95% CIs.

Figure 4.. Comparison of Adjusted Hazard Ratios (AHRs) for All-Cause Mortality

Findings shown for studies of Tiihonen et al (N = 21 492, not corrected for immortal time bias [ITB]), Lin et al (N = 102 964, not corrected for ITB), Li et al (N = 6433, not corrected for ITB), and the current study (N = 32 240, time-fixed method, not corrected for ITB). To compare similar methods, time-fixed exposure to antidepressants and benzodiazepines was used to report the results. NS indicates nonsignificant (P ≥ .05). Error bars indicate 95% CI.