Psychotic Disorders and the Risk of Type 2 Diabetes Mellitus, Atherosclerotic Cardiovascular Diseases, and All-Cause Mortality: A Population-Based Matched Cohort Study

Data sources

For this population-based matched cohort study, the Korean National Health Insurance Service (KNHIS) database from January 2002 to December 2018 was analyzed. The Korean government operates the single nationwide insurer, KNHIS, and all Korean residents are covered by this system [8,9]. The KNHIS constructed an anonymized public database, the KNHIS database, which contains information on demographics, date of death for the deceased, household income state, disability, primary and all secondary diagnoses summarized according to the International Classification of Diseases-10th Revision (ICD-10), prescriptions, medical or surgical procedures, and dates of hospitalization or hospital visits for the entire population in Korea [8].

The Institutional Review Board of Samsung Medical Center approved this study (SMC 2019-09-030) and granted informed consent exemption because only anonymous data were provided to the researchers from the KNHIS.

Patient selection and 1:1 exact matching

A representative sample cohort of patients with psychotic disorders was constructed using the KNHIS database. From all individuals who had been first diagnosed with a psychotic disorder (schizophrenia, schizoaffective disorder, schizotypal disorder, and/or delusional disorder, ICD-10 codes F20–29) between 2003 and 2017, a representative sample cohort comprised of 20% of the total was randomly selected using systematic stratified random sampling with proportional allocation within each stratum. The strata were established based on sex, age, residential region, and household income level. In this sample population with claims for psychotic disorder, 112,491 individuals who were prescribed antipsychotics between 2003 and 2017 were selected. The incidence date of psychotic disorder was determined as the date when the antipsychotics were prescribed for the first time in individuals with psychotic disorder. For comparison, a representative 20% sample cohort was constructed of healthy controls (n=405,111) who had never been diagnosed for major psychiatric disorders (depression, bipolar disorder, and/or psychotic disorder) between 2002 and 2018 after systematic stratified random sampling based on sex, age, residential region, and household income level. The index date (baseline) was defined as the incidence date of psychotic disorder for individuals with psychotic disorders, and as January 1, 2003 for those without major psychiatric disorders. Among the sample population with psychotic disorders treated with antipsychotics and subjects without any major psychiatric disorder, individuals with missing data in at least one variable, those who died within 90 days after baseline, those who had, before or within 3 months after baseline, claims with codes for diabetes (E10–14), obesity (E66), dyslipidemia (E78), hypertension (I10–15), atherosclerosis (I70), aortic aneurysm and dissection (I71–72), arterial embolism and thrombosis (I74), heart failure (I50), ischemic heart disease (I20–25), or cerebrovascular disease (I61–69), subjects who underwent coronary artery bypass graft operation, percutaneous transluminal coronary angioplasty, percutaneous coronary intervention, or thrombolytic treatment before or within 3 months after baseline, those who had claims for transient ischemic attack (G45) during admission before or within 3 months after baseline, and individuals aged <18 years at baseline were excluded (Supplementary Fig. 1). Finally, 357,382 adults without any metabolic or cardiovascular disease were selected; 48,250 were individuals with psychotic disorders treated with antipsychotics and the remaining 309,132 had never been diagnosed with major psychiatric disorders. Exact 1:1 matching based on age and sex was applied for individuals with psychotic disorders and those without major psychiatric disorders. Finally, 48,162 individuals with psychotic disorders and 48,162 matched control participants without major psychiatric disorders were selected for analysis (Supplementary Fig. 1).

Included antipsychoticsg

To define individuals with a psychotic disorder treated with antipsychotics, prescription data for antipsychotics available in Korea were examined. The antipsychotics included amisulpride, aripiprazole, blonanserin, chlorpromazine, clozapine, haloperidol, olanzapine, paliperidone, pimozide, quetiapine, risperidone, sulpiride, and ziprasidone. Among them, chlorpromazine, clozapine, and olanzapine were classified as antipsychotics with substantial weight gain liability [10]. Quetiapine, risperidone, and paliperidone (an active metabolite drug of risperidone) in addition to aforementioned three agents with substantial weight gain liability were considered as antipsychotics with any weight gain liability [10].

Outcomes and follow-up

The outcomes of interest were incident T2DM, AMI, ischemic stroke, composite of all cardiometabolic diseases, and all-cause death during follow-up. All-cause mortality included deaths from physical diseases as well as from suicide or accident. T2DM was defined as one or more claims per year for the prescription of anti-diabetes medication with ICD-10 codes E11–14 [11-13] and AMI was defined as at least one claim with ICD-10 codes I21–22 during hospitalization or two or more claims with these codes [13-15]. Ischemic stroke was defined as recorded ICD-10 code I63 during hospitalization with claims for brain magnetic resonance imaging or computed tomography [14,15]. All cardiometabolic diseases included T2DM, obesity, dyslipidemia, hypertension, ischemic heart disease, cerebrovascular disease, atherosclerosis, aortic aneurysm and dissection, arterial embolism and thrombosis, and hospitalization for heart failure. Definitions of each component of cardiometabolic diseases are summarized in Supplementary Table 1. The study population was followed up from baseline until the date of death, development of outcomes, or December 31, 2018, whichever came first.

Measurements and definitions

The Charlson comorbidity index (CCI) was calculated according to the established method [16] using the diagnostic codes summarized in a previous report [17]. In Korea, according to the Korean Disability Act, persons with physical, mental, or intellectual disabilities diagnosed by the respective medical specialists are registered in a National Disability Registry System [18]. Presence of disability at baseline was verified using anonymized records in this registry system. Among individuals with psychotic disorders, use of other psychiatric medications (mood stabilizers, benzodiazepines, stimulants, antidepressant, and zolpidem) within 3 months after baseline and history of admission to psychiatry from baseline to the end of the follow-up were examined.

Statistical analysis

SAS Enterprise Guide version 7.1 (SAS Institute, Cary, NC, USA) was used for statistical analyses. Two-sided P-values <0.05 were considered statistically significant. Continuous variables are presented as mean±standard deviation and categorical variables as numbers (percentages). The incidence rates of outcomes were derived as number of incident cases divided by total follow-up duration (person-years). The cumulative incidence rates of outcomes were compared between individuals with psychotic disorders and age- and sex-matched controls without major psychiatric disorders using Kaplan–Meier curves; the difference between groups was evaluated using a log-rank test. The hazard ratios (HRs) and 95% confidence intervals (CIs) for outcomes were compared using multivariable Cox regression analysis for individuals with psychotic disorders and age and sex-matched controls. Model 1 was unadjusted and model 2 adjusted for CCI. Model 3 was adjusted for CCI, household income, and disability.

The HRs (95% CIs) for outcome incidence were compared between subjects with psychotic disorders and age- and sex-matched controls without major psychiatric disorders (reference) in subgroups. The subgroups were categorized based on age (18–39, 40–64, or ≥65 years), sex, household income quartile (lowest quartile [Q1] vs. Q2–Q4), CCI (<1 vs. ≥1), and presence of disability. The P value for interaction was derived after evaluating the potential effect modification by factors determining the subgroups.

Sensitivity analyses

To account for competing risk by all-cause death, a cumulative incidence function for T2DM, AMI, ischemic stroke, and composite of all cardiometabolic diseases was plotted, and the subdistribution HRs based on the Fine and Gray method [19] for the four outcomes were estimated after adjusting for the same potential confounders used in the main analyses. Furthermore, we divided individuals with psychotic disorders according to whether they were constantly treated with antipsychotics or not, and estimated the HRs for outcome incidence in these subpopulations with psychotic disorders compared with controls without major psychiatric disorders. Constant antipsychotics therapy was defined as a total of ≥three prescriptions of antipsychotics and at least one prescription of antipsychotics between 365 and 730 days after the first prescription of antipsychotics. We also conducted sensitivity analyses excluding individuals who had ever treated with antipsychotics with substantial weight gain liability or any weight gain liability. During the latter two sensitivity analyses, models were adjusted for age and sex in addition to the potential confounders used in the main analyses.

Baseline characteristics of the study population

After 1:1 exact matching based on age and sex, a total of 48,162 individuals with psychotic disorders (cases) and 48,162 matched controls without major psychiatric disorders were included (Supplementary Fig. 1).

The baseline characteristics of the two groups are presented in Table 1. The mean age was 38.24 years and 49.06% of individuals were males in both groups. Subjects with psychotic disorders were more likely to be in the lowest household income quartile and have a disability. Among individuals with psychotic disorder treated with antipsychotics, use of mood stabilizers, benzodiazepines, stimulants, antidepressant, and zolpidem were identified in 14.05%, 72.80%, 3.68%, 30.12%, and 17.79%, respectively. At least one hospitalization to the psychiatry department was found in 22.67% of individuals with psychotic disorder treated with antipsychotics during the entire follow-up period.

Table 1.

Baseline characteristics of the study population

Psychotic disorders and hazard of outcomesn

In the study cohort, 7,686 cases of T2DM were diagnosed during a median 13.26 years (1,102,508.3 person-years) and 899 AMI cases occurred after a median 14.14 years (1,140,531.5 person-years) (Table 2). During a median 14.14 years (1,140,161.0 person-years), 1,216 ischemic stroke cases occurred. After a median 7.38 years (779,755.6 person-years), 53,683 composite cases of all cardiometabolic diseases were detected. A total of 7,092 deaths occurred during a median 14.23 years (1,145,061.4 person-years). The cumulative incidence of all five outcomes was higher in individuals with psychotic disorders than in matched controls without major psychiatric disorders (Fig. 1) The hazards of all outcomes were significantly higher in subjects with psychotic disorders treated with antipsychotics compared with matched controls without major psychiatric disorders (adjusted HR [95% CI]: 1.522 [1.446 to 1.602] for T2DM; 1.455 [1.251 to 1.693] for AMI; 1.568 [1.373 to 1.790] for ischemic stroke; 1.595 [1.565 to 1.626] for composite of all cardiometabolic diseases; and 2.747 [2.599 to 2.904] for all-cause mortality during follow-up) (Table 2).

Table 2.

HRs and 95% CIs for outcomes according to the presence of psychotic disorders treated with antipsychotics

Fig. 1.

Kaplan–Meier plots of the cumulative incidence of (A) composite of all cardiometabolic diseases, (B) acute myocardial infarction, (C) ischemic stroke, (D) type 2 diabetes mellitus, and (E) all-cause mortality during follow-up in patients with psychotic disorders treated with antipsychotics versus matched controls without major psychiatric disorders.

Subgroup analyses

The HRs (95% CIs) of outcomes in individuals with psychotic disorders treated with antipsychotics compared with matched controls were estimated in subgroups categorized based on age, sex, household income quartile, CCI, and presence of disability (Fig. 2). Subjects with psychotic disorders treated with antipsychotics exhibited higher hazard of outcomes compared with their matched controls in all subgroups except individuals ≥65 years of age and subjects with CCI ≥1 (for outcomes of T2DM, AMI, and ischemic stroke), as well as individuals with disability (for outcomes of T2DM, AMI, ischemic stroke, and all-cause death).

Fig. 2.

Subgroup analyses for the hazard of (A) composite of all cardiometabolic diseases, (B) acute myocardial infarction, (C) ischemic stroke, (D) type 2 diabetes mellitus, and (E) all-cause mortality during follow-up in patients with psychotic disorders treated with antipsychotics versus matched controls without major psychiatric disorders. HR, hazard ratio; CI, confidence interval; CCI, Charlson comorbidity index.

The HRs of T2DM, AMI, and ischemic stroke in the psychotic disorder group showed a non-significant increase compared with the control group also in subjects with CCI ≥1. The association between psychotic disorders and hazards of all five outcomes was more prominent in younger individuals and attenuated in the elderly (P for interaction <0.0001). Furthermore, association between psychotic disorders and hazards of other four outcomes except for AMI were more pronounced in subjects without disability and attenuated in disabled individuals (P for interaction ≤0.0002). Although statistical significance was not observed, with respect to the outcome of AMI, the HR tended to be numerically higher in individuals without disability than in subjects with disability (P for interaction= 0.0588). Regarding outcomes of T2DM, AMI, and composite of all cardiometabolic diseases, the association between psychotic disorders and outcomes was more pronounced in subjects with the lowest household income quartile. Furthermore, association between psychotic disorders and the outcomes of T2DM and all-cause death was more prominent in subjects with CCI <1.

Sensitivity analyses

When all-cause deaths were considered as competing events, the corresponding cumulative incidence function and subdistribution HRs for T2DM, AMI, ischemic stroke, and composite of all cardiometabolic diseases did not differ from the main findings (Supplementary Table 2, Supplementary Fig. 2).

Compared to controls without major psychiatric disorders, those with psychotic disorders constantly treated with antipsychotics exhibited higher hazards of all five outcomes while those with psychotic disorders without constant antipsychotics therapy showed significantly higher hazards of ischemic stroke, composite of all cardiometabolic diseases, and all-cause death during follow-up (Supplementary Table 3). In those with psychotic disorders without constant antipsychotics therapy, although hazards of T2DM and AMI were not significantly higher than in controls in model 3, the HR (95% CI) of allcause death during follow-up compared with controls was 5.019 (4.694 to 5.366) whereas that in individuals with psychotic disorders constantly treated with antipsychotics was 2.070 (1.947 to 2.201).

Additional sensitivity analyses excluding patients who had ever treated with antipsychotics with substantial weight gain liability demonstrated findings consistent to the main analyses (Supplementary Table 4). When patients who had ever treated with antipsychotics with any weight gain liability were excluded, attenuations in HRs were observed throughout all five outcomes (Supplementary Table 5). Nevertheless, people with psychotic disorders exhibited significantly higher hazards of composite of all cardiometabolic diseases and all-cause death during follow-up even in this sensitivity analysis.