Association of Low-Dose Quetiapine and Diabetes

Key Points Question Is the use of quetiapine in low doses associated with increased risk of diabetes? Findings In this nationwide cohort study that included 57 701 new users of quetiapine in low doses and without severe mental illness, the incidence of diabetes was approximately 9 cases per 1000 person-years, similar to that of a reference population treated with selective serotonin reuptake inhibitors for other psychiatric disorders. Meaning Quetiapine used in low doses was not associated with an increased risk of diabetes among individuals with nonsevere mental illness in comparison with use of selective serotonin reuptake inhibitors.


Case-control analysis
The case-control analysis aimed to investigate the association between cumulative doses of quetiapine, used as low-dose treatment, and type 2 diabetes mellitus (T2DM). The case-control analysis was nested among low-dose quetiapine users and, thus, did not include a comparator group. Each T2DM case was matched on age and sex with ten controls, using risk set sampling. The observation period for each low-dose quetiapine user was similar to the follow-up period used in ITT-analyses, i.e., follow-up was confined to use of low-dose quetiapine alone and did not include time with use of SSRIs, other antipsychotics, or higher strengths of quetiapine tablets than equivalents of 25 mg or 50 mg per day.
We calculated odds ratios (ORs) for the association between cumulative quetiapine dose and T2DM, using conditional logistic regression in two ways: 1) Using cumulative quetiapine doses transformed by the binary logarithm as the independent variable. In this analysis, the OR represents the increase in risk for each doubling of the cumulative dose (preplanned analysis). 2) Using predefined strata of cumulative dose (6.26-12.5/12.51-25/25.01-50/100.01-200/>200 DDD) as the independent variable. Use of ≤6.25 DDD was used as reference in both approaches, as this amount is equivalent to 100 tablets of 25mg quetiapinethe smallest package marketed in Denmark.

Sensitivity analyses
To test the impact of analytical choices on the observed association between use of low-dose quetiapine and T2DM, we conducted the following six sensitivity analyses (reported in eFigures 3-4 and eTables 8-9): 1) The grace period was varied from 90 to 60 and 120 days in as-treated analyses to test the impact of grace period definition on incidence rate estimates. 2) The washout window was extended from one to five years (applying to prior use of quetiapine, SSRIs, or other antipsychotics), to assess the impact of previous use of these drugs on T2DM risk. 3) The maximum follow-up period was extended from five to 10 years, to assess if a potential long-term risk existed beyond our initial follow-up window. 4) Exclusion of individuals with history of recurrent depression from cohort entry, to assess the impact of long-term depression on T2DM risk. 5) Using inverse probability of censoring weights (IPCW) in the main analysis, to explore the magnitude of selection bias due to potential informative censoring (e.g., using higher doses quetiapine or being diagnosed with a severe mental illness). To construct ICPWs, time was partitioned into 90-day periods, and the probability of being uncensored at the beginning of each period was estimated using logistic regression analysis. The regression model included treatment status, calendar time, and all covariates from the hdPS-model (as measured at baseline). Stabilized ICPWs were then calculated as the cumulative probability of remaining uncensored, conditional on treatment status and calendar time, divided by the cumulative probability of remaining uncensored on treatment status, calendar time and baseline covariates (in the hdPS-model). We then obtained IPCWweighted hazard ratios using pooled logistic regression analyses. 6) Using standardized mortality ratio weights (SMRW) in a Cox proportional hazards regression model as an alternative to propensity score (PS) matching. Users of low-dose quetiapine were given a weight of 1, and SSRI-users were weighted according to their PS with a weight equal to PS/(1-PS).

Supplementary analyses
To test the impact of different exposures or comparator, and to test assay sensitivity, we conducted the following four supplementary analyses: 1) Extending the exposure definition to include 100 mg quetiapine tablets. This alternative exposure definition was tested in both AT-and ITT-analyses as well as the case-control analysis of the association with cumulative dose (eFigure 3 and eTable 7). 2) Extending the exposure definition to include all strengths of quetiapine tablets. This alternative exposure definition was tested in both AT-and ITT-analyses as well as the case-control analysis of the association with cumulative dose (eFigure 3 and eTable 7). 3) Using Z-drugs as active comparator to assess the appropriateness of choosing SSRIs as the comparator in the main analyses (eFigure 5). 4) Using olanzapine as alternative exposure to test assay sensitivity of the main analyses (eFigure 6).

eFigure 1: Graphical representation of baseline assessment and follow-up
Notes: a) History of diabetes was assessed as prescriptions for antidiabetic medications, hospital diagnoses of diabetes, or blood levels of glycosylated hemoglobin above reference values, b) History of severe mental illness was defined hospital diagnoses of schizophrenia, schizoaffective disorder, or bipolar affective disorder, c) High-dose quetiapine was defined as filling of prescriptions with tablet strengths >50mg, d) Baseline covariates for the propensity score model included prescriptions and hospital diagnoses from the prior 365 days, e) Follow-up ended at the earliest of diabetes, death, emigration, end of study, five years of follow-up, or censoring for meeting baseline exclusion criteria (i.e. diagnosis of SMI, use of high-dose QUE, use of other study drug, use of other AP, or new diagnosis of type 1 diabetes mellitus

eTable 2: Covariates included in high-dimensional propensity score model
The 50 most common prescriptions and hospital diagnoses among the included low-dose quetiapine or SSRI-users within 12 months before cohort entry. Use of drugs or occurrence of hospital diagnoses are divided into three groups: At least one occurrence, sporadic use (i.e., more than the median number of prescriptions/hospital contacts in the cohort), and frequent use (i.e., more than the 75 th percentile of prescriptions/hospital contacts in the cohort).