This cohort study examines the association between initiation of antipsychotic medication and annual prevalence of pharmacologically treated type 2 diabetes among youths in the Netherlands.
Key Points
Question
Does the annual prevalence of pharmacologically treated type 2 diabetes increase after antipsychotic initiation in youth?
Findings
In this cohort study of 89 991 Dutch youths aged 0 to 19 years, a design controlling for stable, person-specific confounders showed that the prevalence of pharmacologically treated type 2 diabetes increased significantly after antipsychotic initiation, reaching approximately 9 additional cases per 10 000 antipsychotic users by year 5. This elevated prevalence persisted among those with brief antipsychotic exposure.
Meaning
Antipsychotic initiation is associated with a rapid, sustained increase in pharmacologically treated type 2 diabetes prevalence independent of underlying time-invariant factors, emphasizing the need for routine metabolic monitoring.
Abstract
Importance
Antipsychotic medications are widely prescribed to children and adolescents worldwide, raising concerns about treatment-emergent type 2 diabetes (T2D). Previous studies have reported elevated T2D risks but have relied on between-person comparisons susceptible to unmeasured confounding and have not characterized risk trajectories around treatment initiation.
Objective
To determine the annual prevalence of pharmacologically treated T2D before and after antipsychotic initiation in Dutch youths.
Design, Setting, and Participants
This cohort study used data from Dutch population-based registers to track yearly pharmacologically treated T2D prevalence up to 5 years before and after antipsychotic initiation among youths aged 0 to 19 years initiating such medication. Using fixed-effects models, the event study design controlled for all stable, person-specific confounders that limit standard between-person comparisons. Data were collected from January 1, 2006, to December 31, 2022, and analyzed beginning in February 2025.
Exposure
Antipsychotic initiation (first recorded dispensation).
Main Outcome and Measures
The primary outcome was annual prevalence of pharmacologically treated T2D, defined as at least 1 dispensation of a noninsulin glucose-lowering medication within a given year. Annual risk differences (RDs) and relative risk changes were estimated using the year preceding antipsychotic initiation as the reference.
Results
A total of 89 991 youths who were dispensed at least 1 antipsychotic prescription between 2006 and 2022 (median [IQR] age at antipsychotic initiation, 13.6 [9.8-16.7] years; 55 794 [62.0%] males) were included in the sample. In this cohort, pharmacologically treated T2D prevalence was stable before treatment and increased significantly after antipsychotic initiation. The RD increased from 2.47 (95% CI, 1.09-3.86) per 10 000 antipsychotic users in the initiation year to 9.02 (95% CI, 5.99-12.06) per 10 000 users by year 5. Females had greater absolute excess risk of T2D than males by year 5 (RD, 16.48 [95% CI, 8.95-24.00] per 10 000 users vs 5.50 [95% CI, 2.57-8.43] per 10 000 antipsychotic users). Adolescents aged 13 to 19 years had a larger long-term increase in risk than children aged 7 to 12 years (year-5 RD, 14.09 [95% CI, 8.19-19.98] per 10 000 users vs 5.47 [95% CI, 2.76-8.18] per 10 000 users). While recurrent users experienced greater, progressively increasing risk (RD, 10.17 [95% CI, 6.47-13.87] per 10 000 antipsychotic users by year 5), elevated risk of pharmacologically treated T2D persisted among youths using antipsychotics in the initiation year only.
Conclusions and Relevance
In this within-individual cohort study, antipsychotic initiation in youths was associated with a rapid, sustained increase in treated T2D risk independent of underlying time-invariant factors. These findings indicate that even brief antipsychotic exposure is associated with a sustained metabolic vulnerability, underscoring the need for routine metabolic monitoring and preventive strategies from antipsychotic treatment initiation.
Introduction
Antipsychotic medications are widely prescribed to children and adolescents worldwide.1,2 In the Netherlands, antipsychotic prescribing rates are high compared with those in other European countries, with annual prescription rates reaching 8 to 10 per 1000 children.3 Most pediatric antipsychotic use has become off label, used primarily for autism spectrum disorder–related irritability, attention-deficit/hyperactivity disorder, anxiety, and sleep disturbances rather than psychotic disorders.4 Although guidelines recommend routine metabolic monitoring in young people treated with antipsychotics, adherence in practice is low.5 This is concerning because early-onset type 2 diabetes (T2D) often follows an aggressive clinical course, with early nephropathy and neuropathy.6,7
Observational studies and meta-analyses consistently associated antipsychotic use during youth and increased T2D risk, with a 2- to 3-fold higher incidence in patients treated with an antipsychotic than in unexposed peers.4,8,9,10 Excess risk appears within the first treatment year, particularly for second-generation agents, such as olanzapine and clozapine.11,12 Evidence on how risk differs by sex, age, or socioeconomic status remains limited, despite well-established socioeconomic gradients in cardiometabolic risk.13,14
However, several clinical questions remain unanswered, such as when after initiation does excess risk first emerge, does it persist after brief exposure, and is the postinitiation increase in risk associated with trends already under way before treatment. First, most pediatric studies have relied on between-person comparisons, typically contrasting antipsychotic-treated youths with unexposed psychiatric patients or initiators of other psychotropic drugs, using covariate adjustment or propensity score methods.8,9,10 These designs may not fully account for unmeasured, time-invariant factors such as genetic liability, baseline metabolic risk, and family and neighborhood conditions. Large population-based cohorts and umbrella reviews show that T2D incidence and prevalence vary substantially across psychiatric disorders, even after adjustment for sociodemographic and lifestyle factors, suggesting that residual confounding by underlying disorder profile and social disadvantage is likely.15,16 Second, most studies report 1 or a few hazard ratios or odds ratios over fixed follow-up, sometimes split only by crude exposure categories. These summary estimates show that excess risk appears early, but they say little about how risk changes year by year around initiation, the longer-term shape of the risk, or the degree to which long-term excess risk is associated with brief vs sustained antipsychotic use. Third, previous studies have not formally examined pretreatment trends in pharmacologically treated T2D, making it difficult to distinguish treatment-emergent risk from metabolic deterioration that was already under way before treatment.
To address these gaps, we conducted a nationwide pre-post event study using Dutch population–based registry data. This study examined how the yearly prevalence of pharmacologically treated T2D changed before and after antipsychotic initiation in youths, whether it showed preinitiation trends, and how this trajectory varied by sex, age at initiation, neighborhood household income, and exposure duration. By clarifying when excess risk emerges, how it evolves, and how it varies across subgroups, this study aimed to inform metabolic monitoring schedules, identification of high-risk groups, and prescribing and follow-up policies in young people.
Methods
This registry-based cohort study used pseudonymized data from Statistics Netherlands (CBS); therefore, ethics review and informed consent were not required under the applicable Dutch legal and institutional frameworks. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Study Design and Data Sources
Beginning in February 2025, we conducted a register-based pre-post study of antipsychotic initiation and pharmacologically treated T2D prevalence in Dutch youths. We used a within-person event study design, meaning that we compared each individual’s annual prevalence of pharmacologically treated T2D in the years before and after they started antipsychotic medication, with the year immediately preceding initiation as the reference. This approach produced year-by-year estimates of how the pharmacologically treated T2D prevalence changed around treatment initiation while controlling for all time-invariant characteristics. Data were obtained from CBS, which contains near-complete information on prescription dispensation, demographics, socioeconomic status, migration, and health care use. All residents of the Netherlands are registered in the national population registry and those younger than 18 years have basic health insurance without premiums or deductibles.
Study Population
The study population included youths aged 0 to 19 years who initiated an antipsychotic medication (Anatomical Therapeutic Chemical [ATC] code N05A) between January 1, 2006, and December 31, 2022. Individuals whose first antidiabetic medication was insulin (ATC A10A) prior to any noninsulin antidiabetic (ATC A10B) use were excluded to minimize misclassification. Follow-up ended at death or age 25 years.
We applied a within-individual design, aligning each youth’s outcomes relative to their own antipsychotic initiation date and tracking the yearly prevalence of pharmacologically treated T2D over an 11-year period, spanning 5 years before and 5 years after treatment initiation. The fixed-effects models allowed us to control for stable person-specific confounders and calendar time trends. We quantified year-specific absolute and relative changes in the prevalence of pharmacologically treated T2D, assessed preinitiation trends, and examined differences by sex, age at initiation, neighborhood household income, and brief vs sustained exposure.
Antipsychotic exposure was defined as the first recorded dispensation of any medication under ATC code N05A. This date marked year 0 for each individual and served as the anchor for their observation period. Each individual contributed only 1 antipsychotic initiation date and 1 observation window. For each initiator, we constructed an unbalanced 11-year panel spanning 5 years before to 5 years after the first antipsychotic dispensation, conditional on being observed in that calendar year. This window captured short- and longer-term risk patterns. The primary outcome was pharmacologically treated T2D, defined as receiving at least 1 dispensation of a noninsulin glucose-lowering drug (ATC A10B) within a given calendar year.
Subgroup Analysis
Subgroup analyses examined sex, age at antipsychotic initiation, socioeconomic status, and exposure duration. Age subgroups were restricted to children aged 7 to 12 years and adolescents aged 13 to 19 years; children aged 0 to 6 years were excluded because their short preexposure window did not allow reliable estimation of pretreatment trends. Socioeconomic status was included given prior evidence of socioeconomic disparities in diabetes risk and care among youths.17 Socioeconomic status was defined using neighborhood-level household income from CBS (median population, 2780 individuals). Individuals were classified as living in lower- or higher-income neighborhoods based on whether the area’s average disposable income fell below or above the cohort median. Finally, we compared individuals whose antipsychotic exposure was confined to the initiation year (<1 year) with those who were dispensed antipsychotic prescription in at least 1 subsequent year (≥1 year) to assess differences between brief and recurrent exposure.
We also performed several sensitivity and supplemental analyses. To assess the robustness of the primary findings, we repeated the main analysis in the complete case population and in models additionally adjusted for antidepressant, systemic corticosteroid, and anxiolytic use. To evaluate the outcome definition, we conducted a supplementary diagnosis-based analysis of T2D for the period in which diagnosis data were available. To examine whether risk varied with exposure intensity, we estimated pharmacologically treated T2D risk according to cumulative antipsychotic exposure, measured using defined daily doses. Finally, to address potential time-varying confounding, we repeated the main analysis with additional adjustment for parental mental health and health care utilization, using health care costs as a proxy.
Statistical Analysis
Fixed-effects linear probability models were used to estimate annual risk differences relative to 1 year prior to initiation (year −1) (eMethods in Supplement 1). Relative risk (RR) changes were calculated by dividing each risk difference (RD) by the crude prevalence of pharmacologically treated T2D in year –1, providing the proportional increase or decrease in risk relative to baseline. Parallel preinitiation trends were assessed using a joint Wald test of all preinitiation coefficients. A nonsignificant result supported the validity of the design. Analyses were conducted in R, version 4.3.1 (R Project for Statistical Computing) with individual-level clustered SEs. Statistical significance was set at 2-sided P < .05.
Results
The study included a total of 89 991 youths who were dispensed at least 1 antipsychotic prescription between 2006 and 2022. The median (IQR) age at antipsychotic initiation was 13.6 [9.8-16.7] years, and 55 794 males (62.0%) were included (Table 1).
Table 1. Baseline Characteristics of the Study Population at Antipsychotic Initiation (Year 0).
| Characteristic | No. (%) |
|---|---|
| Total No. of participants | 89 991 |
| Sex | |
| Male | 55 794 (62.0) |
| Female | 34 197 (38.0) |
| Median (IQR) age at initiation, y | 13.6 (9.8-16.7) |
| Median (IQR) neighborhood household income, €a | 33 400 (30 000-36 300) |
| Age group at initiation, y | |
| 0-6 | 6669 (7.4) |
| 7-12 | 34 934 (38.8) |
| 13-19 | 48 388 (53.8) |
To convert to US dollars, multiply by 1.05.
Figure 1 and Table 2 show the annual prevalence of pharmacologically treated T2D per 10 000 youths who initiated antipsychotic treatment. During the preinitiation period, the prevalence remained low and relatively stable, ranging from 0.76 (95% CI, 0.09-1.42) per 10 000 antipsychotic users in year –5 to 2.78 (95% CI, 1.68-3.88) per 10 000 users in year –1. A marked increase was observed beginning in year 0 (6.22 [95% CI, 4.60-7.85] per 10 000 users), followed by a continued upward pattern in the years thereafter, reaching 14.58 (95% CI, 11.56-17.60) per 10 000 antipsychotic users by year 5.
Figure 1. Dot Graph of Yearly Prevalence of Pharmacologically Treated Type 2 Diabetes (T2D) by Year Relative to Antipsychotic Initiation.
Years –5 through –1 represent the 5-year period preceding the first recorded antipsychotic prescription dispensed, with year –5 indicating 5 years prior and year –1 representing the year immediately preceding initiation. Years 0 through 5 reflect the year of initiation and the subsequent 5 years of follow-up.
Table 2. Estimated Prevalence of Pharmacologically Treated T2D by Time Relative to Antipsychotic Initiation.
| Time relative to antipsychotic initiation, y | Point estimate (95% CI) | P valueb | ||
|---|---|---|---|---|
| T2D prevalence per 10 000 users | RDa | RR changeb | ||
| −5 | 0.76 (0.09 to 1.42) | −0.41 (−1.74 to 0.92) | −0.15 (−0.63 to 0.33) | .55 |
| −4 | 1.39 (0.53 to 2.25) | −0.24 (−1.69 to 1.21) | −0.09 (−0.61 to 0.44) | .75 |
| −3 | 1.54 (0.66 to 2.42) | −0.35 (−1.60 to 0.90) | −0.13 (−0.58 to 0.32) | .58 |
| −2 | 1.55 (0.71 to 2.40) | −0.67 (−1.72 to 0.39) | −0.24 (−0.62 to 0.14) | .21 |
| −1 | 2.78 (1.68 to 3.88) | Reference | Reference | NA |
| 0 | 6.22 (4.60 to 7.85) | 2.47 (1.09 to 3.86) | 0.89 (0.39 to 1.39) | <.001 |
| 1 | 9.14 (7.10 to 11.18) | 5.11 (3.14 to 7.08) | 1.84 (1.13 to 2.55) | <.001 |
| 2 | 9.33 (7.19 to 11.47) | 5.05 (2.99 to 7.11) | 1.82 (1.08 to 2.56) | <.001 |
| 3 | 9.21 (7.01 to 11.40) | 4.55 (2.33 to 6.77) | 1.64 (0.84 to 2.44) | <.001 |
| 4 | 13.46 (10.67 to 16.24) | 8.29 (5.52 to 11.07) | 2.98 (1.99 to 3.98) | <.001 |
| 5 | 14.58 (11.56 to 17.60) | 9.02 (5.99 to 12.06) | 3.25 (2.15 to 4.34) | <.001 |
Abbreviations: NA, not applicable; RD, risk difference; RR, relative risk; T2D, type 2 diabetes.
RDs and RR changes are presented relative to year –1.
P values correspond to comparisons with the reference year.
Event study estimates controlling for individual and calendar-year effects showed no evidence of differential preinitiation trends relative to year –1. The preinitiation coefficients were not jointly different from 0 (P = .63 by joint Wald test). A statistically significant increase appeared in year 0 (RD, 2.47 [95% CI, 1.09-3.86] per 10 000 users), corresponding to 2 to 3 additional pharmacologically treated T2D cases per 10 000 users compared with year –1. The RD increased to 5.11 (95% CI, 3.14-7.08) per 10 000 users in year 1, increasing to 9.02 (95% CI, 5.99-12.06) additional pharmacologically treated T2D cases per 10 000 users in year 5, corresponding to a cumulative excess of 0.34% over the 5 years after initiation (Figure 2). The corresponding RR changes were 0.89 (95% CI, 0.39-1.39) in year 0, increasing to 1.84 (1.13-2.55) in year 1 and reaching 3.25 (2.15-4.34) in year 5 (eFigure 1 in Supplement 1).
Figure 2. Dot Graph of Model-Estimated Annual Risk Differences in Prevalence of Pharmacologically Treated Type 2 Diabetes (T2D) by Year Relative to Antipsychotic Initiation.
Years −5 to −1 represent the preinitiation period, year 0 is the year of initiation, and years 1 through 5 indicate follow-up years. Estimates are derived from a fixed-effects linear probability model, adjusted for calendar year. Year –1 (the year immediately preceding antipsychotic initiation) serves as the reference category.
aP < .001.
Subgroup Analyses
Sex Differences
Both sexes showed increases in pharmacologically treated T2D after initiation, but with differing magnitudes. In year 0, increases were modest, with an RD of 2.75 (95% CI, 0.22-5.29) per 10 000 antipsychotic users in females and 2.32 (95% CI, 0.70-3.94) per 10 000 users in males. By year 1, risk increased further (females: RD, 6.21 [95% CI, 2.26-10.17] per 10 000 users; males: RD, 4.49 [95% CI, 2.37-6.62] per 10 000 users). From year 2 onward, estimates for females consistently exceeded those for males. By year 5, RDs were 16.48 (95% CI, 8.95-24.00) per 10 000 users for females and 5.50 (95% CI, 2.57-8.43) per 10 000 users for males (between-sex P = .008).
In terms of RR, both sexes showed sustained postinitiation increases. By year 5, RR changes were comparable at 3.32 (95% CI, 1.80-4.83) for females and 3.83 (95% CI, 1.79-5.88) for males, indicating similar proportional increases but markedly different absolute burdens (eFigure 2 and eTable 1 in Supplement 1).
Age at Antipsychotic Initiation
Differences by age were evident both in timing and magnitude of increased risk. In year 0, adolescents (aged 13-19 years) showed a significant increase in RD at 4.52 (95% CI, 1.97-7.06) per 10 000 users, whereas younger children (aged 7-12 years) did not yet show a significant increase, with an RD of 0.09 (95% CI, –0.59 to 0.78) per 10 000 users. By year 1, both groups demonstrated statistically significant increases, with adolescents again experiencing larger increases (6.67 [95% CI, 3.29-10.05] per 10 000 users vs 3.98 [95% CI, 1.83-6.13] per 10 000 users).
Differences widened progressively. From year 3 onward, adolescents accumulated greater excess risk, and by year 5, the RD among adolescents reached 14.09 (95% CI, 8.19-19.98) per 10 000 users compared with 5.47 (95% CI, 2.76-8.18) per 10 000 users among younger children. The between-group difference became evident by year 4, when the RD was 12.26 (95% CI, 7.01-17.50) per 10 000 users among adolescents vs 5.49 (95% CI, 2.89-8.08) per 10 000 users among younger children (P = .02).
RR changes could not be calculated for younger children (aged 7-12 years) because no cases were observed in year −1. Among adolescents, proportional increases were evident from year 0 (RR change, 0.87 [95% CI, 0.38-1.37]) and rose steadily, reaching 2.73 (95% CI, 1.58-3.87) by year 5 (eFigure 3 and eTable 2 in Supplement 1).
Neighborhood Household Income
We assessed differences by neighborhood household income at antipsychotic initiation, using the cohort median of €33 400 (IQR, €30 000-€36 300; $35 070 [IQR, $31 500-$38 115]) as the cutoff. After initiation of antipsychotic medication, RDs increased in both income groups. In year 0, the RD was 3.88 (95% CI, 1.74-6.02) per 10 000 users among individuals from below–median income neighborhoods and 1.14 (95% CI, –0.67 to 2.96) per 10 000 users among those from above–median income neighborhoods, with only the former reaching statistical significance. From year 1 onward, both groups showed significant and sustained increases. By year 5, the RD reached 7.62 (95% CI, 2.94-12.29) per 10 000 users in the below-median group and 9.89 (95% CI, 5.68-14.09) per 10 000 users in the above-median group, indicating comparable absolute increases.
Both income groups experienced persistent increases in RR change in pharmacologically treated T2D after antipsychotic initiation. In year 0, the RR change was 1.07 (95% CI, 0.48-1.66) for the below-median group and 0.58 (95% CI, –0.34 to 1.49) for the above-median group. Relative increases became statistically significant from year 1 onward in both groups. By year 5, the RR change reached 2.10 (95% CI, 0.81-3.39) among below-median households and 4.98 (95% CI, 2.86-7.10) among above-median households. Overlapping CIs indicate no clear socioeconomic gradient in treatment-emergent T2D (eFigure 4 and eTable 3 in Supplement 1).
One-Year vs Recurrent Antipsychotic Exposure
Both 1 year or less and more than 1 year exposure groups experienced elevated pharmacologically treated T2D risk after initiation of an antipsychotic medication. Among individuals with 1 year of exposure or less, the RD was 3.38 (95% CI, 0.48-6.27) per 10 000 users in year 0, increasing to 6.31 (95% CI, 2.41-10.22) in year 1. Risk remained elevated through year 5 (RD, 6.62 [95% CI, 1.02-12.22] per 10 000 users). In contrast, youths with recurrent exposure showed consistently larger and statistically significant increases across all postinitiation years. The RD increased from 1.95 (95% CI, 0.59-3.30) per 10 000 users in year 0 to 4.36 (95% CI, 2.23-6.49) per 10 000 users in year 1 and reached 10.17 (95% CI, 6.47-13.87) per 10 000 users in year 5.
On the relative scale, youths with 1 year of exposure or less exhibited moderate increases, with an RR change of 0.62 (95% CI, 0.09-1.15) in year 0 and 1.16 (95% CI, 0.44-1.87) in year 1, with RR changes remaining positive through year 5 (1.21 [95% CI, 0.19-2.24]). In contrast, the more than 1 year exposure group showed a markedly steeper rise, with an RR change of 2.08 (95% CI, 0.63-3.52) in year 0, 4.65 (95% CI, 2.37-6.93) in year 1, and 10.85 (95% CI, 6.91-14.80) by year 5 (eFigure 5 and eTable 4 in Supplement 1).
Sensitivity and Supplemental Analyses
The findings of the sensitivity and supplemental analyses were consistent with the primary findings. In the complete case analysis, the postinitiation increase in pharmacologically treated T2D persisted, with an RD of 4.00 (95% CI, 1.08-6.92) per 10 000 antipsychotic users in year 1 and 11.45 (95% CI, 6.43-16.47) per 10 000 users in year 5 (eFigures 6 and 7 and eTable 5 in Supplement 1). A separate model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medication use showed a similar postinitiation pattern (eFigures 8 and 9 and eTable 6 in Supplement 1). A supplementary diagnosis-based analysis restricted to 2016 to 2022 showed a comparable postinitiation pattern, although interpretation was limited by the shorter observation window (eFigure 10 and eTable 7 in Supplement 1). In a descriptive analysis of cumulative antipsychotic exposure, the highest defined daily dose quartile had the highest observed prevalence of pharmacologically treated T2D (eTable 8 in Supplement 1). Additional adjustment of the model for parental mental health and health care utilization, using health care costs as a proxy, did not materially change the observed postinitiation pattern (eTable 9 in Supplement 1).
Discussion
In this cohort study of 89 991 Dutch youths, antipsychotic initiation was followed by a rapid and sustained increase in pharmacologically treated T2D risk. Excess risk appeared in the initiation year, nearly doubled in the first year, and continued thereafter. Both sexes showed sustained relative increases, but females accumulated a larger absolute RD by year 5. Adolescents had a significantly greater increase than younger children from year 4 onward. Youths in both lower- and higher-income neighborhoods showed increased risk, with no strong evidence of socioeconomic differences. Risk patterns differed markedly by exposure duration: those with only initiation-year use showed modest but persistent elevation in risk, whereas recurrent users experienced greater and progressively increasing risk.
The pattern of flat preinitiation trends followed by a sharp and sustained postinitiation increase supports a treatment-related mechanism rather than a mechanism associated solely with underlying illness trajectories. This interpretation is biologically plausible given the known metabolic effects of second-generation antipsychotic medications, which promote weight gain and insulin resistance through combined histaminergic, serotonergic, and dopaminergic receptor blockade.18 Prior between-person studies have established that youths treated with antipsychotic medications have approximately 2- to 3-fold higher T2D risk than unexposed peers.8,9,10 Our within-individual event study design, which compared individuals over time while controlling for all stable person-specific confounders, added 3 quantitative refinements to this picture. First, the year-by-year resolution showed that excess risk emerged in the year of initiation, nearly doubled by year 1, and continued to accumulate through year 5, a trajectory that aggregate hazard ratios over a fixed follow-up cannot reveal. Second, the formal pretrend test (joint Wald P = .63) was consistent with flat preinitiation trends, providing empirical support for the interpretation that the postinitiation increases reflected exposure rather than metabolic deterioration already under way. Third, the absolute RDs, a cumulative excess of 0.34% over 5 years, were of the same order of magnitude as those reported in the Danish registry study despite different designs, lending convergent support to the magnitude while identifying a previously undescribed temporal pattern of progressive accumulation after treatment initiation.9
The greater cumulative risk among females is consistent with previous reports of higher rates of antipsychotic-induced weight gain and metabolic complications.9 Earlier pubertal onset, with associated increases in fat mass and shifts in body composition, may contribute to greater insulin resistance compared with resistance in boys of similar age, and lower average physical activity levels in adolescent girls may further exacerbate adverse metabolic effects.19 The higher risk in adolescents likely reflects heightened metabolic vulnerability during puberty. Pubertal hormonal changes induce transient insulin resistance, which, when combined with antipsychotic-induced weight gain and dysmetabolism, may make T2D more likely to develop in susceptible adolescents.20 The divergence in risk after approximately 4 years suggests cumulative treatment effects within this developmental window. Notably, recent global and Dutch data indicate that obesity prevalence is not higher in adolescents than in younger children, arguing against baseline obesity as the main mechanism of the observed age differences and pointing instead toward physiological and exposure-related mechanisms.21,22
Although both neighborhood household income groups experienced increased pharmacologically treated T2D risk after antipsychotic initiation, the increases were broadly similar. This may partly reflect universal access to basic health care in the Netherlands. However, population-based evidence indicates that socioeconomic disparities persist in Dutch populations with diabetes. Children and adults from lower socioeconomic backgrounds are more likely to be hospitalized and less likely to receive advanced technologies, such as continuous glucose monitoring.23 Our findings therefore do not exclude socioeconomic differences in long-term metabolic outcomes; rather, they suggest that, once exposed to antipsychotics, youths across the socioeconomic spectrum may face similar short- to medium-term increases in pharmacologically treated T2D risk, while downstream consequences may still reflect social patterns.
The findings for youths with brief antipsychotic exposure are noteworthy. Even when antipsychotic treatment was limited to the initiation year, pharmacologically treated T2D risk remained significantly elevated throughout follow-up. This finding suggests that early treatment-related metabolic changes may not fully resolve after discontinuation in susceptible youths. In contrast, recurrent antipsychotic users experienced greater and steadily increasing risk, indicating that continued treatment is associated with greater cumulative T2D burden. These findings suggest that both short and long antipsychotic courses warrant metabolic monitoring, with especially close surveillance during sustained exposure.
Despite guideline recommendations, cardiometabolic monitoring in youths who receive antipsychotic medication remains poor in routine practice. Dutch outpatient record reviews showed low rates of glucose and lipid monitoring after initiation, with more than half of children receiving no laboratory monitoring during follow-up.24,25,26 Our results highlight the clinical importance of closing this gap. Structured monitoring should be implemented from treatment initiation, with routine assessments at defined intervals, and should be accompanied by early lifestyle interventions, such as dietetic referral, physical activity promotion, and family-based education embedded in standard care. Quality improvement programs that integrate electronic health record alerts and engage both clinicians and families have been shown to increase glucose and lipid monitoring rates in pediatric antipsychotic users, and similar multifaceted approaches may be needed to achieve sustained adherence in routine practice.27,28 For selected youths at high risk, pharmacologic prevention strategies, such as off-label metformin may be considered, in consultation with pediatric specialists.29
Youth-onset T2D progresses rapidly and leads to early multimorbidity; fewer than 60% of adolescents remain free of complications 15 years after diagnosis, and nearly one-third develop multiple complications, such as nephropathy and hypertension.7 Each decade of earlier T2D onset is estimated to reduce life expectancy by 3 to 4 years, underscoring the long-term consequences of disease onset shifting into adolescence.30 Beyond individual burden, these trajectories translate into substantial societal costs. Dutch estimates indicate that T2D reduces employment probability by approximately 2.4% and income by 3.6%.31 Direct health care expenditures average roughly €2400 ($2520) per patient per year, with complications accounting for most of these costs.32 By advancing the age at onset, antipsychotic-related T2D in youths therefore amplifies both clinical and economic burdens across the life course.
Strengths and Limitations
Key strengths of this study include nationwide coverage, a large sample size, long follow-up, and a within-individual event study design that controlled for stable person-specific characteristics and mapped pharmacologically treated T2D risk around antipsychotic initiation. We also examined clinically relevant subgroups by sex, age at initiation, neighborhood household income, and exposure duration. In addition, sensitivity analyses showed a similar postinitiation pattern in the complete case analysis and in models additionally adjusted for antidepressant, systemic corticosteroid, and anxiolytic use (eAnalysis, eFigures 6 to 9, and eTables 5 and 6 in Supplement 1).
Study limitations should also be acknowledged. Our outcome captured only pharmacologically treated T2D and likely underestimated total T2D occurrence by missing undiagnosed or lifestyle-managed cases; limited laboratory monitoring may also have contributed to underascertainment, although increased health care contact after initiation may have increased detection. Because noninsulin glucose-lowering medications, particularly metformin, may also be prescribed for conditions other than T2D, including prediabetes and polycystic ovarian syndrome, as well as for management of antipsychotic-associated weight gain, outcome misclassification is possible. However, a supplementary diagnosis-based analysis from 2016 to 2022 suggested a similar postinitiation pattern, although the restricted time window limits interpretation . We could not distinguish between antipsychotic classes or individual agents, and although higher antipsychotic exposure, measured using defined daily doses, showed greater pharmacologically treated T2D risk, these findings do not replace agent-specific analyses. The ATC N05A category is heterogeneous, and prior pediatric research suggests variation in T2D risk across antipsychotic classes and individual agents.8 Therefore, our aggregate exposure definition may have masked important drug-specific differences. We also lacked information on clinical indication and intermediate metabolic measures, and residual confounding by time-varying disease severity, lifestyle factors, comedication, diagnostic workup, and treatment complexity remains possible. To partly address time-varying confounding, we performed a supplementary analysis additionally adjusting for parental mental health and health care utilization, using health care costs as a proxy; this analysis showed a similar postinitiation pattern (eTable 9 in Supplement 1).
Conclusions
In this within-individual cohort study of Dutch youths initiating antipsychotics, antipsychotic initiation was associated with a substantial and sustained increase in pharmacologically treated T2D risk. The risk accumulated steadily, particularly among females and adolescents, with comparable increases across income groups. Elevated pharmacologically treated T2D prevalence also persisted among youths whose antipsychotic use was limited to the initiation year, suggesting that even brief exposure was associated with sustained metabolic vulnerability. These findings underscore the need for cautious prescribing, structured metabolic monitoring, and early preventive strategies to limit the long-term health and economic burdens of treatment-emergent T2D in young people.
eMethods. Modeling approach
eFigure 1. Relative risk for the pre-post analysis
eFigure 2. Subgroup analysis by sex
eTable 1. Subgroup analysis by sex
eFigure 3. Subgroup analysis by age group
eTable 2. Subgroup analysis by age group
eFigure 4. Subgroup analysis by neighborhood household income
eTable 3. Subgroup analysis by neighborhood household income
eFigure 5. Subgroup analysis: 1-year versus recurrent antipsychotic exposure
eTable 4. Subgroup analysis: 1-year versus recurrent antipsychotic exposure
eAppendix. Complete case analysis
eFigure 6. Prevalence of complete case analysis
eFigure 7. Risk difference of complete case analysis
eTable 5. Complete case analysis
eFigure 8. Model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eFigure 9. Relative risk of model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eTable 6. Model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eTable 7. Diabetes diagnosis data over the period 2016-2022
eFigure 10. Risk difference of model built using diabetes diagnosis data over the period 2016-2022
eTable 8. Defined daily dose (DDD) of antipsychotic users (2020-2022)
eTable 9. Model adjusted for parent mental health and health care utilization (2009-2022)
Data Sharing Statement
References
- 1.Hálfdánarson Ó, Zoëga H, Aagaard L, et al. International trends in antipsychotic use: a study in 16 countries, 2005-2014. Eur Neuropsychopharmacol. 2017;27(10):1064-1076. doi: 10.1016/j.euroneuro.2017.07.001 [DOI] [PubMed] [Google Scholar]
- 2.Bushnell G, Lloyd J, Olfson M, Cook S, Das H, Crystal S. Antipsychotic medication use in Medicaid-insured children decreased substantially between 2008 and 2016. Health Aff (Millwood). 2023;42(7):973-980. doi: 10.1377/hlthaff.2022.01625 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Gangapersad RN, Zhou G, Garcia-Gomez P, et al. Comparison of antipsychotic drug use in children and adolescents in the Netherlands before and during the COVID-19 pandemic. Eur Child Adolesc Psychiatry. 2024;33(8):2695-2703. doi: 10.1007/s00787-023-02340-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Galling B, Roldán A, Nielsen RE, et al. Type 2 diabetes mellitus in youth exposed to antipsychotics: a systematic review and meta-analysis. JAMA Psychiatry. 2016;73(3):247-259. doi: 10.1001/jamapsychiatry.2015.2923 [DOI] [PubMed] [Google Scholar]
- 5.Sanyal S, Rowan PJ, Ochoa-Perez M, et al. Metabolic monitoring for children and adolescents prescribed second-generation antipsychotics: a qualitative study with child psychiatrists. J Child Adolesc Psychopharmacol. 2024;34(8):359-365. doi: 10.1089/cap.2024.0026 [DOI] [PubMed] [Google Scholar]
- 6.Zheng L, Chen X, Luo T, et al. Early-onset type 2 diabetes as a risk factor for end-stage renal disease in patients with diabetic kidney disease. Prev Chronic Dis. 2020;17:E50. doi: 10.5888/pcd17.200076 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bjornstad P, Drews KL, Caprio S, et al. ; TODAY Study Group. Long-term complications in youth-onset type 2 diabetes. N Engl J Med. 2021;385(5):416-426. doi: 10.1056/NEJMoa2100165 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bobo WV, Cooper WO, Stein CM, et al. Antipsychotics and the risk of type 2 diabetes mellitus in children and youth. JAMA Psychiatry. 2013;70(10):1067-1075. doi: 10.1001/jamapsychiatry.2013.2053 [DOI] [PubMed] [Google Scholar]
- 9.Nielsen RE, Laursen MF, Vernal DL, et al. Risk of diabetes in children and adolescents exposed to antipsychotics: a nationwide 12-year case-control study. J Am Acad Child Adolesc Psychiatry. 2014;53(9):971-979.e6. doi: 10.1016/j.jaac.2014.04.023 [DOI] [PubMed] [Google Scholar]
- 10.Lee H, Song DH, Kwon JW, Han E, Chang MJ, Kang HY. Assessing the risk of type 2 diabetes mellitus among children and adolescents with psychiatric disorders treated with atypical antipsychotics: a population-based nested case-control study. Eur Child Adolesc Psychiatry. 2018;27(10):1321-1334. doi: 10.1007/s00787-018-1123-2 [DOI] [PubMed] [Google Scholar]
- 11.Correll CU, Manu P, Olshanskiy V, Napolitano B, Kane JM, Malhotra AK. Cardiometabolic risk of second-generation antipsychotic medications during first-time use in children and adolescents. JAMA. 2009;302(16):1765-1773. doi: 10.1001/jama.2009.1549 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Newcomer JW. Second-generation (atypical) antipsychotics and metabolic effects: a comprehensive literature review. CNS Drugs. 2005;19(1)(suppl 1):1-93. doi: 10.2165/00023210-200519001-00001 [DOI] [PubMed] [Google Scholar]
- 13.Derks IPM, Koster A, Schram MT, et al. The association of early life socioeconomic conditions with prediabetes and type 2 diabetes: results from the Maastricht study. Int J Equity Health. 2017;16(1):61. doi: 10.1186/s12939-017-0553-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Abreu TC, Beulens JW, Heuvelman F, Schoonmade LJ, Mackenbach JD. Associations between dimensions of the social environment and cardiometabolic health outcomes: a systematic review and meta-analysis. BMJ Open. 2024;14(8):e079987. doi: 10.1136/bmjopen-2023-079987 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lindekilde N, Scheuer SH, Rutters F, et al. Prevalence of type 2 diabetes in psychiatric disorders: an umbrella review with meta-analysis of 245 observational studies from 32 systematic reviews. Diabetologia. 2022;65(3):440-456. doi: 10.1007/s00125-021-05609-x [DOI] [PubMed] [Google Scholar]
- 16.Lee MK, Lee SY, Sohn SY, Ahn J, Han K, Lee JH. Type 2 diabetes and its association with psychiatric disorders in young adults in South Korea. JAMA Netw Open. 2023;6(6):e2319132. doi: 10.1001/jamanetworkopen.2023.19132 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Yen FS, Wei JCC, Liu JS, Hwu CM, Hsu CC. Parental income level and risk of developing type 2 diabetes in youth. JAMA Netw Open. 2023;6(11):e2345812. doi: 10.1001/jamanetworkopen.2023.45812 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Miedlich SU, Lamberti JS. Connecting the dots: understanding and addressing the metabolic impact of antipsychotic and antidepressant medications. Ann N Y Acad Sci. 2025;1546(1):35-57. doi: 10.1111/nyas.15301 [DOI] [PubMed] [Google Scholar]
- 19.Candler TP, Mahmoud O, Lynn RM, Majbar AA, Barrett TG, Shield JPH. Continuing rise of type 2 diabetes incidence in children and young people in the UK. Diabet Med. 2018;35(6):737-744. doi: 10.1111/dme.13609 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Hannon TS, Arslanian SA. The changing face of diabetes in youth: lessons learned from studies of type 2 diabetes. Ann N Y Acad Sci. 2015;1353(1):113-137. doi: 10.1111/nyas.12939 [DOI] [PubMed] [Google Scholar]
- 21.Zhang X, Liu J, Ni Y, et al. Global prevalence of overweight and obesity in children and adolescents: a systematic review and meta-analysis. JAMA Pediatr. 2024;178(8):800-813. doi: 10.1001/jamapediatrics.2024.1576 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Plasmans M, Van Diest B, Zantinge E. Overgewicht Jongeren Volksgezondheid En Zorg. August 21, 2025. Accessed October 22, 2025. https://www.vzinfo.nl/overgewicht/jongeren
- 23.de Vries SAG, Sas TCJ, Bak JCG, et al. Socio-economic disparities in hospital care among Dutch patients with diabetes mellitus. Diabetes Obes Metab. 2024;26(4):1386-1394. doi: 10.1111/dom.15440 [DOI] [PubMed] [Google Scholar]
- 24.Biemond KMC, Breur JMPJ, Kloosterboer SM. Antipsychotica: monitoring van bijwerkingen bij kinderen en adolescenten. Psyfar. Published online June 1, 2024. Accessed October 22, 2025. https://www.psyfar.nl/tijdschrift/editie/artikel/t/antipsychotica-monitoring-van-bijwerkingen-bij-kinderen-en-adolescenten
- 25.Minjon L, Brozina I, Egberts TCG, Heerdink ER, van den Ban E. Monitoring of adverse drug reaction-related parameters in children and adolescents treated with antipsychotic drugs in psychiatric outpatient clinics. Front Psychiatry. 2021;12:640377. doi: 10.3389/fpsyt.2021.640377 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Dinnissen M, Dietrich A, van der Molen JH, et al. Guideline adherence of monitoring antipsychotic use for nonpsychotic indications in children and adolescents: a patient record review. J Clin Psychopharmacol. 2021;41(1):13-18. doi: 10.1097/JCP.0000000000001322 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Soda T, Richards J, Gaynes BN, et al. Systematic quality improvement and metabolic monitoring for individuals taking antipsychotic drugs. Psychiatr Serv. 2021;72(6):647-653. doi: 10.1176/appi.ps.202000155 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Shin S, Moon S, Wang J, Choi YJ. Impact of institutional quality improvement initiatives on metabolic monitoring in mental disorder in patients treated with antipsychotics: a meta-analysis of intervention studies. J Glob Health. 2024;14:04074. doi: 10.7189/jogh.14.04074 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Mansuri Z, Makani R, Trivedi C, et al. The role of metformin in treatment of weight gain associated with atypical antipsychotic treatment in children and adolescents: a systematic review and meta-analysis of randomized controlled trials. Front Psychiatry. 2022;13:933570. doi: 10.3389/fpsyt.2022.933570 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Kaptoge S, Seshasai SRK, Sun L, et al. ; Emerging Risk Factors Collaboration . Life expectancy associated with different ages at diagnosis of type 2 diabetes in high-income countries: 23 million person-years of observation. Lancet Diabetes Endocrinol. 2023;11(10):731-742. doi: 10.1016/S2213-8587(23)00223-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Vermeulen W, Pomp M, ter Weel B, William L, Andras H. Diabetes en werk. 2025. Accessed October 23, 2025. https://www.seo.nl/publicaties/diabetes-en-werk/
- 32.Peters ML, Huisman EL, Schoonen M, Wolffenbuttel BHR. The current total economic burden of diabetes mellitus in the Netherlands. Neth J Med. 2017;75(7):281-297. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Modeling approach
eFigure 1. Relative risk for the pre-post analysis
eFigure 2. Subgroup analysis by sex
eTable 1. Subgroup analysis by sex
eFigure 3. Subgroup analysis by age group
eTable 2. Subgroup analysis by age group
eFigure 4. Subgroup analysis by neighborhood household income
eTable 3. Subgroup analysis by neighborhood household income
eFigure 5. Subgroup analysis: 1-year versus recurrent antipsychotic exposure
eTable 4. Subgroup analysis: 1-year versus recurrent antipsychotic exposure
eAppendix. Complete case analysis
eFigure 6. Prevalence of complete case analysis
eFigure 7. Risk difference of complete case analysis
eTable 5. Complete case analysis
eFigure 8. Model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eFigure 9. Relative risk of model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eTable 6. Model adjusted for antidepressant, systemic corticosteroid, and anxiolytic medications
eTable 7. Diabetes diagnosis data over the period 2016-2022
eFigure 10. Risk difference of model built using diabetes diagnosis data over the period 2016-2022
eTable 8. Defined daily dose (DDD) of antipsychotic users (2020-2022)
eTable 9. Model adjusted for parent mental health and health care utilization (2009-2022)
Data Sharing Statement
