Psychotropic Classes Concomitant with Stimulants and the Association with the Risk of Diabetes in Medicaid-Insured Youth

Abstract

Importance:

Psychotropic classes concomitant with stimulants (PCCWS) in children and adolescents have shown an inconsistent impact on the risk of diabetes mellitus. PCCWS with 2–5 classes is common but the risk of diabetes subsequent to stimulant initiation is unknown.

Objective:

To assess the risk of diabetes in youth with PCCWS regimens with 2–5 additional psychotropic classes.

Design:

A retrospective cohort study was conducted using Medicaid claims data from 2007–2014. Youth aged 2–17 years with 1–7.5 years of continuous enrollment who were new stimulant users with clinician-reported psychiatric diagnosis were followed.

Setting:

Outpatient clinic and inpatient records for statewide Medicaid-insured youth in a mid-Atlantic state.

Participants:

The study cohort comprised 30,112 youth with an average follow-up of 6.4 years.

Exposures:

Among stimulant users, five groups were defined according to concomitant use.

Main Outcomes and Measures:

The major outcome is diabetes risk and was assessed using discrete-time failure models, after adjustment for disease risk score which was categorized using more than 120 baseline and time-dependent covariates.

Results:

Among 30,112 new stimulant users, 43 were new diabetes patients and 30,069 were nondiabetes patients. The absolute risk of diabetes in youth had an eight-fold increase from 3 to 5 class PCCWS regimens that included antipsychotics, antidepressants, or anticonvulsant-mood stabilizers (0.5; 1.13; 4.20 per 10,000 person-months, respectively). Compared with stimulant monotherapy, PCCWS with any of these 3 classes were significantly more likely to develop diabetes (adjusted relative risk [ARR], 2.58, (95% CI 1.05–6.82); 5.81, (2.29–14.75); 18.91, (6.07–58.90) for 3, 4 and 5 class PCCWS, respectively). Similarly, there was a significantly greater risk of diabetes for those with 120–779 days of cumulative duration than for shorter exposures, and in 4 and 5 class combined PCCWS including antipsychotics, antidepressants, or anticonvulsant-mood stabilizers [ARR, 3.78, (1.16–12.40)].

Conclusions:

In a large Medicaid-insured, long-enrolled youth cohort, the use of PCCWS, particularly concomitant use with antipsychotic, antidepressant, and/or antipsychotic-mood stabilizers, and with long duration of these combinations were associated with an increased risk of diabetes.

Relevance:

The findings support a call for corroboration in a large national cohort of continuously enrolled Medicaid-insured pediatric youth with long follow-up.

Keywords

Medicaid youth stimulants polypharmacy long-term use risk of diabetes

Introduction

Stimulants for the psychiatric treatment of children and adolescents are mainly prescribed to treat attention-deficit/hyperactivity disorder (ADHD) and other disruptive behavior disorders (Greenhill et al., 2002). Polypharmacy in youth population is on the rise with as many as 300,000 youth receiving three or more psychotropic classes concomitantly (Zito et al., 2021) and these regimens frequently reflect psychotropic classes concomitant with stimulants (PCCWS).

Stimulants are commonly used with atypical antipsychotics with or without antidepressants (Burcu et al., 2016; Kreider et al., 2014). Even though antipsychotics (Bobo et al., 2013; Burcu et al., 2017b; Rubin et al., 2015) and antidepressants (Burcu et al., 2017a) have been shown to be associated with increased risk of diabetes, concomitant use with stimulants has shown inconsistent results. For example, subgroup analyses showed that concomitant use of stimulants and antipsychotics was not significantly associated with diabetes when compared with antipsychotic monotherapy in 5–20-year-olds (Burcu et al., 2017b) and 10–20-year-olds (Rubin et al., 2015). However, another study of concomitant use of antipsychotics and stimulants showed a significant association with diabetes risk in 6–24-year-olds (Bobo et al., 2013).

To our knowledge, no study has fully investigated the association between 3–5 psychotropic classes in PCCWS and diabetes subsequent to stimulant initiation among long continuously enrolled youth. Our study design is closely allied with clinical practice for ADHD: start with stimulant and if further improvement is needed, add other psychotropic classes. We analyzed a longitudinal cohort of young people who were new stimulant users to assess the association between 3–5 class PCCWS and risk of diabetes.

Methods

Data source

We used 2007–2014 Medicaid administrative claims data from a mid-Atlantic state for the analysis, which included enrollment files, outpatient and physician claims files, and dispensed prescription drug files. The files in the data source were linked together via an encrypted identification number.

The enrollees’ sociodemographic characteristics, monthly Medicaid enrollment status, and Medicaid eligibility category are included in the enrollment files. The clinician-reported psychiatric diagnoses are based on the ICD-9-CM codes. The drug dispensing information such as national drug codes, dispensing date, days supplied, and quantity supplied were in the dispensed prescription drug file.

This study was reviewed and approved by the Institutional Review Board of the University of Maryland, Baltimore.

Study design and study population

This is a retrospective cohort study. The cohort is comprised of new stimulant users. A new-user design was applied by restricting the study cohort to those who initiated stimulant treatment, defined as not having stimulant dispensing in the prior 180 days. Furthermore, the study population of 2–17-year-olds was enrolled in Medicaid between 2007 and 2014 for continuous (1–7.5 years) enrollment and had a psychiatric diagnosis. The stimulant initiation date serves as the index date for cohort entry. If the patient stopped using a stimulant, s/he was censored. The Medicaid enrollment data were generated based on each patient’s monthly Medicaid enrollment status. To be eligible for a full year enrollment, the patient had to be enrolled for at least 10 months per year (Zito et al., 2013).

The flowchart in Figure 1 illustrates the inclusion and exclusion criteria that were applied to define the study cohort. There are 30,112 new stimulant users in the final study population: 43 developed diabetes and 30,069 formed a comparison group.

FIG. 1.

Flow chart for the study cohort of Medicaid-insured youth (2–17-year-olds) who were new users of stimulant, 2007–2014.

Newly diagnosed diabetes

In this study, the dependent variable is newly diagnosed diabetes, which is quantitatively expressed as the number of diabetes cases subsequent to stimulant initiation per 10,000 person-months of follow-up. Diabetes is ascertained adapting a computerized database algorithm that was previously validated in a cohort of Medicaid-insured youth with a positive predictive values of 83.9% (Bobo et al., 2013) and later applied by Burcu and colleagues(Burcu et al., 2017a; Burcu et al., 2017b). New cases of diabetes were identified using diabetes-related medical care encounters (Supplementary Table S1 and Supplementary Fig. S1):

An inpatient with a primary diagnosis of diabetes, by itself, met the case definition of diabetes. Otherwise, a combination of at least 2 out of 3 different diabetes-related medical care encounters within a period of 120 days met the case definition of diabetes (i.e., an outpatient visit with a primary diagnosis of diabetes, an inpatient secondary diagnosis of diabetes, or a dispensing of anti-diabetic medication). A single outpatient diabetes diagnosis or oral hypoglycemic agent dispensing in the absence of a diabetes diagnosis may indicate another disease than diabetes, such as polycystic ovarian syndrome (Legro et al., 2013), or prophylactic treatment to attenuate weight gain associated with antipsychotics (Barnard et al., 2013; Brufani et al., 2013). We excluded four patients diagnosed with polycystic ovary.

For youth who met the case definition of diabetes, the first diagnosis date for diabetes is assigned as the date of the initial diabetes-related medical care encounter. Youth were followed up from the first stimulant use to first diabetes diagnosis for the diabetic population while nondiabetic patients were followed up until the end of continuous enrollment, or the end of study (December 31, 2014), whichever came first. Patients who had diabetes before the start of stimulant use or before the start of PCCWS were excluded from the study.

Psychotropic medication and PCCWS

Psychotropic medications are grouped into seven classes (Supplementary Table S2a). In this study, PCCWS is defined as concomitant use of stimulants with any other psychotropic class(es) for at least 60 consecutive days (allowing for 15-day gaps to get prescriptions refilled). Based on each patient’s maximum number of drug classes taken concomitantly with stimulants, we defined five mutually exclusive groups: stimulant monotherapy and 2–5 class PCCWS.

The length of psychotropic medication use was calculated using time-dependent duration of exposure measures during the follow-up period. The patient’s daily stimulant usage was calculated based on the stimulant dispensing dates, days supplied, and allowable 15-day gaps for prescription refills. Based on daily stimulant usage in the database, we built a daily stimulant usage string up to 7.5 years for each patient in the study within their continuous enrollment period. Following the same methodology, the daily usage of the other six psychotropic classes was calculated as well.

Patient characteristics as risk factors

Multiple studies have indicated numerous risk factors for diabetes, which include clinical and other health care utilization factors (Mikkelsen et al., 2015; Nishi, 2018; Petrie et al., 2018; Pini et al., 2016; Rayner et al., 2019; Semenkovich et al., 2015). In this study, we included more than 120 baseline and time-dependent covariates, which include psychiatric disorders, metabolic disorders, cardiovascular diseases, etc. Supplementary Table S2 c–i lists subgroups of risk factors which were used in prior published studies assessing diabetes risk with psychotropic class in youth (Bobo et al., 2013; Burcu et al., 2017a; Burcu et al., 2017b). The disease risk score method to control for these factors is discussed in the statistical methods section.

Outcome measures

Sociodemographic and Diagnostic Factors.

Patient characteristics were grouped into categories: age at initial start of stimulants, gender, race/ethnicity, Medicaid eligibility category, and length of continuous Medicaid enrollment. Clinician-reported psychiatric diagnoses were grouped into 11 categories via ICD-9 codes (Supplementary Table S2b).

Absolute Risk of Diabetes According to the Number of PCCWS Classes and Cumulative Duration.

The absolute risk of diabetes was calculated based on the number of diabetes cases following stimulant initiation and the person-time to diabetes. The absolute risks of diabetes in duration subgroups were calculated.

Relative Risk of Diabetes.

The relative risk was calculated using discrete time failure models.

Stimulant Monotherapy and PCCWS Regimens and Psychiatric Diagnoses.

The stimulant monotherapy and PCCWS regimens for diabetic patients, and the clinician-reported psychiatric diagnoses for diabetic and nondiabetic patients are presented in number and percentage and compared using Chi-square analysis.

Statistical analysis

Chi square analysis was used to assess the demographic, administrative and clinician-reported diagnostic patterns of the study cohort. All analyses were done using SAS software (version 14.0). A two-sided p value <0.05 was considered statistically significant.

The absolute risk per 10,000 person-months was calculated by dividing the number of diabetes cases that developed subsequent to stimulant use by the number of person-months of observation. For relative risk, we applied a discrete time failure model in a longitudinal framework to estimate the effect of PCCWS on the risk of diabetes with person months as the unit of analysis. Discrete time analysis has been previously used (Magder and Petri, 2012; Rubin et al., 2015; Winterstein et al., 2012) and produced estimates approximately identical to those from Cox proportional hazard models while having computational efficiency and practical advantages, especially when several time-varying exposure groups exist (Alison, 2010; D’Agostino et al., 1990). In discrete time analyses, to adjust for confounding, we utilized the disease risk score methodology using the Miettinen full-cohort approach (Arbogast and Ray, 2009; Arbogast and Ray, 2011; Glynn et al., 2012).

The disease risk score was constructed by fitting a logistic regression model with the study outcome, the exposure of interest and the covariates for the entire study population: it is computed as the fitted value from that regression model for each study subject. Then the subjects were grouped according to the quartiles of disease risk score (<25%, 25%–50%, 51%–75%, >75%). The final discrete failure time model was fitted using another logistic regression, which was adjusted for disease risk score, time of follow up, age, gender, and eligibility.

Results

Characteristics of the study cohort

The study cohort comprised 30,112 youth aged 2–17 years with new stimulant use from January 01, 2007 through December 31, 2014 (Table 1). The study population had an average follow-up time of 6.4 years. There were 43 patients who developed diabetes subsequent to initiation with a stimulant. There were 30,069 non-diabetes patients for comparison.

Table 1.

Baseline Characteristics of Diabetic and Nondiabetic New Stimulant Users

	Diabetic patients N = 43		Nondiabetic patients N = 30,069
	N	%	N	%	p value
Age group (years)^a
2–5	4	9.3	11,035	36.7	<0.001
6–11	22	51.2	13,411	44.6
12–17	17	39.5	5,623	18.7
Gender
Male	19	44.2	19,906	66.2	<0.001
Female	24	55.8	10,163	33.8	<0.001
Race/ethnicity
African American	26	60.5	13,351	44.4	<0.001
White	12	27.9	12,328	41.0
Hispanic/Nat Am/Asian/Pac Islander/Unknown	5	11.6	4,390	14.6
Eligibility category
Foster care	8	18.6	2,646	8.8	<0.001
SCHIP	3	7.0	6,314	21.0
SSI	15	34.9	5,353	17.8
TANF	17	39.5	15,756	52.4
Continuous enrollment (years)
1–2	0	0	3,097	10.3	<0.001
3–4	2	4.7	4,059	13.5
5–6	9	20.9	5,443	18.1
7–8	32	74.4	17,470	58.1
Clinician-reported psychiatric diagnosis
Externalizing disorders only^b	12	27.9	17,771	59.1	<0.001
Internalizing disorders only^b	0	0	1,293	4.3
Both^b	30	69.8	10,133	33.7
Other psychiatric diagnoses	1	2.3	872	2.9
No. PCCWS classes
Stimulant only	15	34.9	22,732	75.6	<0.001
2 class PCCWS	10	23.3	4,961	16.5
3 class PCCWS	7	16.3	1,624	5.4
4 class PCCWS	7	16.3	662	2.2
5 class PCCWS	4	9.2	90	0.3
Diabetes disease risk score
<25%	6	14.0	15,305	50.9	<0.001
25%–50%	8	18.6	5,954	19.8
51%–75%	12	27.9	5,412	18.0
>75%	17	39.5	3,398	11.3

Age is the age of first stimulant use.

Externalizing disorders include attention-deficit/hyperactivity disorder, disruptive behavioral disorders; internalizing disorders include depression, anxiety disorders; other psychiatric diagnoses include adjustment disorder, autism spectrum disorder, learning disorder, schizophrenia, intellectual disability, tic disorder, bipolar disorder, and other mental health diagnoses.

SCHIP, State Children’s Health Insurance Program; SSI, Supplemental Security Income; TANF, Temporary Assistance for Needy Families; PCCWS, psychotropic classes concomitant with stimulants.

Among the diabetic patients, there were more 6–11-year-olds (51.2%), females (55.8%), African Americans (60.5%), youth eligible for disability (SSI) (34.9%) or family poverty level (TANF) (39.5%), and continuously enrolled for 7–8 years (74.4%). Patients had both externalizing and internalizing disorders (69.8%) or only externalizing disorders (27.9%). There were more diabetic patients with 2–5 PCCWS treatment (65.1%) than stimulant monotherapy. Most diabetic patients had more than 25% disease risk scores (86.0%) (Table 1).

PCCWS was more frequent among 12–17-year-olds, males, African Americans, those with internalizing disorders, and those who had disease risk scores greater than 75% (Supplementary Table S3).

In the nondiabetic population, patients were younger, more often male, white, enrolled in TANF, and slightly fewer had lengthy enrollments. Nondiabetic patients were more frequent in the stimulant monotherapy group (75.6%) and 51% had lowest disease risk scores (Table 1).

Absolute risk of diabetes according to stimulant monotherapy and PCCWS

There were 15 new cases of diabetes following stimulant use during 1,095,724 person-months of follow-up (absolute risk [AR], 0.14 per 10, 000 person-months) in the stimulant monotherapy group. Among 2–5 class PCCWS with antipsychotics, antidepressants, and anticonvulsant-mood stabilizers, there is a linear trend for increased absolute risk with increasing class. Consequently, AR grew from 0.50 to 1.13 to 4.20 per 10,000 person-months for 3, 4, and 5 class regimen users, respectively (Table 2).

Table 2.

Absolute and Relative Risk of Diabetes According to Exposure to Stimulant Monotherapy and PCCWS

Exposure status	Person-months	DM case No.	Absolute risk per10,000 Person-Months	Relative risk (95% CI)^a
Stimulant monotherapy	1,095,724	15	0.14	1 [Reference]
2 class PCCWS	476,702	10	0.21	1.45 (0.62–3.39)
With (ATP or ATD or ATC-MS)	243,674	7	0.29	1.70 (0.65–4.50)
Without (ATP or ATD or ATC-MS)	233,028	3	0.13	1.06 (0.30–3.67)
3 class PCCWS	153,228	7	0.46	2.34 (0.90–6.19)
With (ATP or ATD or ATC-MS)	140,924	7	0.50	2.58 (1.05–6.82)
Without (ATP or ATD or ATC-MS)	12,576	0	—	—
4 class PCCWS	63,236	7	1.11	5.71 (2.25–14.46)
With (ATP or ATD or ATC-MS)	62,101	7	1.13	5.81 (2.29–14.75)
Without (ATP or ATD or ATC-MS)	1,135	0	—	—
5 class PCCWS	9,805	4	4.08	18.61 (6.0–57.77)
With (ATP or ATD or ATC-MS)	9,513	4	4.20	18.91 (6.07–58.90)
Without (ATP or ATD or ATC-MS)	292	0	—	—

Adjusted for disease risk score, time of follow-up, age, gender, and eligibility.

AA, alpha agonist; ATP, antipsychotic; ATD, antidepressant; ATC-MS, anticonvulsant-mood stabilizer; AX, anxiolytic; PCCWS, psychotropic classes concomitant with stimulants.

Absolute risk of diabetes according to cumulative duration

Duration of PCCWS regimens that include antipsychotics, antidepressants, and anticonvulsant-mood stabilizers showed 1.84-fold to 3.78-fold increase of AR with longer PCCWS duration. Cumulative duration for 4 and 5 class PCCWS were combined due to the small numbers of diabetes patients in 5 class PCCWS. In the 3-class PCCWS group, AR was 0.80 per 10,000 person-months in the longer duration group (120–1055 days). In the combined 4–5 class PCCWS group; AR was 3.91 per 10,000 person-months in the longer duration group (120–779 days) (Table 3).

Table 3.

Absolute and Relative Risk of Diabetes According to Cumulative Duration of PCCWS Including ATP or ATD or ATC-MS*

Exposure status	Person-months	Number of case	Absolute risk per 10,000 person-months	Relative risk^a
2 class PCCWS with ATP or ATD or ATC-MS
Cumulative duration (days)
60–119	200,112	5	0.25	1 [Reference]
120–1,414	43,562	2	0.46	1.84 (0.36–9.47)
Total	243,674	7
3 class PCCWS with ATP or ATD or ATC-MS
Cumulative duration (days)
60–119	103,236	4	0.39	1 [Reference]
120–1,055	37,688	3	0.80	2.05 (0.46–9.18)
Total	140,924	7
4 and 5 class PCCWS with ATP or ATD or ATC-MS
Cumulative duration (days)
60–119	58,817	6	1.04	1 [Reference]
120–779	12,797	5	3.91	3.78 (1.16–12.40)
Total	71,614	11

Adjusted for disease risk score, time of follow-up, age, gender, and eligibility.

ATP, antipsychotic; ATD, antidepressant; ATC-MS, anticonvulsant-mood stabilizer; PCCWS, psychotropic classes concomitant with stimulants.

Absolute risk of diabetes according to disease risk score

Diabetes increased with increasing disease risk scores. The risk of diabetes was 0.08 per 10,000 person-months in those with <25% disease risk score and increased to 0.27, 0.45, and 1.05 per 10,000 person-months in 25%–50%, 51%–75% and ≥75% disease risk score groups (Supplementary Fig. S2).

Relative risk of diabetes according to stimulant monotherapy and PCCWS

The stimulant plus 3 class PCCWS with antipsychotics, antidepressants, and/or anticonvulsant-mood stabilizers use was significantly associated with a 2.58-fold (95% CI 1.05–6.82) increased risk of diabetes. The combined risk of 4 class PCCWS and 5 class PCCWS with these 3 classes was significantly associated with diabetes (RR = 5.81, 95% CI 2.29–14.75; RR = 18.91, 95% CI 6.07–58.90), respectively (Table 2).

Relative risk of diabetes according to cumulative duration

Similar to the absolute risk findings, there was a significant association of longer cumulative duration in the combined 4 and 5 class PCCWS group. The patients with cumulative duration of 120–779 days were more likely to have diabetes diagnosed following stimulant initiation (RR = 3.78, 95% CI 1.16–12.40) (Table 3).

Relative risk of diabetes according to disease risk score

Greater disease risk score is significantly associated with higher risk of diabetes: patients with >75% disease risk score was more likely to have incident diabetes (RR 7.74, 95% CI 2.73–21.89) followed by 51%–75% disease risk score group (RR 3.72, 95% CI 1.25–11.01) (Supplementary Table S4).

Distribution of PCCWS regimens and psychiatric diagnoses

Atypical antipsychotics, antidepressants, and anticonvulsant-mood stabilizers were the most frequent concomitant psychotropic medications in 3 to 5 class PCCWS among diabetes patients (Supplementary Table S5and Supplementary Table S6).

With respect to clinician-reported diagnosis, ADHD is the leading diagnosis in nearly all diabetic patients (95.4%) and nondiabetic patients (90.1%). More than half the diabetic patients had clinician-reported diagnosis of disruptive disorders (53.5%), exceeding the pattern of diagnosis in nondiabetic patients. ADHD and disruptive disorders were present in nearly all diabetic patients (95.4% and 53.5%) and nondiabetic patients (90.1% and 32.1%), respectively (Supplementary Table S7).

Discussion

To our knowledge, this is the first article using the disease risk score methodology and discrete failure time model for pediatric PCCWS among patients with long continuous enrollment to show that a higher number of concomitant psychotropic classes in PCCWS is associated with a higher risk of diabetes diagnosed subsequent to stimulant initiation. The outcome is most pronounced among 4 and 5 class PCCWS users in a real-world population across 7.5 years.

The major findings of this study: 1)

In a pediatric cohort followed up for 6.4 years on average, the relative risk of diabetes was significantly associated with 3 to 5 class PCCWS, longer cumulative duration, and with substantial disease risk score.

The absolute risk of diabetes showed an 8-fold significant linear increase from 3 to 5 class PCCWS regimens that included antipsychotics, antidepressants, and/or anticonvulsant-mood stabilizers and increased with longer cumulative duration.

Compared with nondiabetic patients, diabetic patients were more frequently 6–11-years old, female, African American, with disability (SSI) or poverty family income (TANF) coverage, and with 5 or more years of continuous enrollment. Diabetic patients received more complex PCCWS regimens. Atypical antipsychotics, antidepressants, and anticonvulsant-mood stabilizers were nearly twice as common as in nondiabetic patients.

A minor but concerning finding is the association of diabetes among very young stimulant monotherapy users.

Strengths of the methodology

We used a previously validated algorithm to identify new diabetes cases in this youth cohort (Arbogast and Ray, 2009; Arbogast and Ray, 2011; Glynn et al., 2012). In the algorithm, there is a wide range of risk factors related to the development of diabetes including medication (type, duration), underlying disease, and heath care access. Instead of adding these risk factors to the model, we utilized the disease risk score methodology from the Miettinen full-cohort approach (Burcu et al., 2017b; Glynn et al., 2012). Similar to the propensity score, the disease risk score helps build a parsimonious model and improves the performance of the model. It has advantages over the propensity score in the presence of multiple time-dependent exposure groups (Arbogast and Ray, 2009).

Comparison of current findings with published studies

There are three published studies that investigated the association between 2 class concomitant psychotropic medications (stimulant and antipsychotic or antidepressant) and diabetes in the pediatric Medicaid-enrolled population. The studies vary in age of the youth cohorts, follow-up time, and outcome (Table 4). By contrast, the 2–17-year-old population in our study covered children who started stimulants at a very young age, while earlier studies featured 5–20 year olds (Burcu et al., 2017b), 6–24 year olds (Bobo et al., 2013), and 10–18 year olds (Rubin et al., 2015). In our study, we had four diabetic patients in the 2–5-year-old age group: one patient had stimulant monotherapy, and three patients had 2 class PCCWS (stimulant plus antidepressant in two patients and stimulant plus alpha agonist in one patient). These patients had high diabetes disease risk score (≥75%) and a median of 83 months (6.9 years) from the start of treatment of stimulant to diabetes diagnosis.

Table 4.

Comparison of Current Findings with Published Studies

	Population	Average follow-up time	Findings
Current study	2–17-year-olds, Medicaid Mid-Atlantic state	6.4 years	3 class poly with (ATP/ATD/ATC-MS) RR 2.584 class poly with (ATP/ATD/ATC-MS) RR 5.815 class poly with (ATP/ATD/ATC-MS) RR 18.91
Burcu, 2017b	5–20-year-olds, Medicaid Analytic eXtract (4 states)	24.8 months	In concomitant users of atypical antipsychotics and stimulants, there was no significant association: neither duration nor cumulative dose of stimulants was associated with an increased risk of diabetes
Bobo et al., 2013	6–24-year-olds, Tennessee Medicaid	4 years	Atypical antipsychotics + StimulantsRR = 5.71 (95%CI 1.35–24.07)
Rubin et al., 2015	10–18-year-olds, more than 19 million across U.S. Medicaid population	20 months	Atypical antipsychotics + StimulantsRR = 0.84 (0.64–1.11)

ATP, antipsychotic; ATD, antidepressant; ATC-MS, anticonvulsant-mood stabilizer; RR, relative risk; PCCWS, psychotropic classes concomitant with stimulants.

Compared with the follow-up time in prior studies [4 years (Bobo et al., 2013), 24.8 months (Burcu et al., 2017b), and 17.2 months (Rubin et al., 2015)], the long follow-up time in our study is a strength: the majority of patients had an average of 6.4 years of follow-up (median follow-up 6.9 years). Bobo and colleagues showed in 4+ years of follow-up that concomitant stimulant and atypical antipsychotic was significantly associated with increased diabetes risk in 6–24 year olds (Bobo et al., 2013). In the published studies with ≤2 years of follow-up, the combination was not significant for stimulant plus atypical antipsychotic in 5–20 year olds (Burcu et al., 2017b) and in 10–18 year olds (Rubin et al., 2015). Our study results follow those of Bobo (Bobo et al., 2013) in showing that the significance of medication use with diabetes risk in youth largely depends on long follow-up of PCCWS with antipsychotics, antidepressants, and/or anticonvulsant-mood stabilizers (Table 4).

The findings of our study show that PCCWS is associated with increased diabetes risk which may be explained by the long duration of stimulant with other psychotropic class exposure and younger aged stimulant users. The greater duration of exposure is particularly important for the study of relatively rare stimulant adverse events, such as pediatric diabetes.

A recent compelling finding from the influential MTA study revealed that 16 years of consistent stimulant monotherapy was found to be associated with increased weight gain (Greenhill et al., 2020) warranting further research. Their finding is consistent with our finding of absolute risk of 0.14 per 10,000 person-months among stimulant monotherapy users associated with subsequent diabetes in our study (15/43, 34.9%).

Future research directions

In the current study, all the foster care enrollees had PCCWS and all developed diabetes. In view of continuing national concern for PCCWS among foster care youth, this is a research priority (GAO, 2011).

Future research on 3 or more PCCWS is critical not only because these combinations are typically off-label and commonly used by physicians, but more importantly, the safety signal based on multiclass PCCWS regimens needs to be further investigated to assure that benefits exceed risks (Baker et al., 2021). In a 2015 international consensus of distinguished child psychiatric researchers, studies on effectiveness and safety of commonly used off-label regimens is a leading priority (Persico et al., 2015). Research in large simple trials (Stroup, 2011) are feasible and cost-efficient (Horton et al., 2021) and are facilitated by the growing availability of electronic health record systems in large academic settings. Perhaps, the way forward includes deprescribing studies (Barnett et al., 2020) to learn whether individuals with complex regimens can be managed successfully on fewer classes of medication.

Limitations

Several limitations should be noted. First, causality cannot be inferred from observational studies; thus, the study findings support corroboration from a large cohort of continuously enrolled youth with long follow-up. Second, we focus on a single state Medicaid dataset and the number of diabetes cases subsequent to stimulant initiation is limited. Third, diabetes cases could have been misclassified if diagnosis information was missing. Differential misclassification of diabetes is theoretically possible. Fourth, the psychotropic medication dispensings may not be accurate because patients may not actually have consumed the medication. However, the long refill patterns diminish the likelihood of nonadherence. Fifth, the statistical power was calculated based on the published diabetes risk in 2 class polypharmacy [antipsychotic (ATP) and stimulants] (Burcu et al., 2017b). Due to limited data on diabetes risk, the statistical power could not be conducted for 3 or more class PCCWS. Sixth, residual confounding caused by conditions such as severity of disease, could not be adjusted in the study due to limited variables in the dataset. Seventh, the diabetes risk and the possible benefit of PCCWS is still unclear and needs further analysis, perhaps in large electronic health records. Eighth, due to the administrative payment record dataset, data on medication reconciliation were not available. Finally, there is possible surveillance bias because of metabolic monitoring. However, in a previously published article (Morrato et al., 2010), the children who had antipsychotics did not receive adequate glucose and lipid screening. Thus, surveillance bias is likely to be modest.

Conclusions

In a large cohort of Medicaid-insured youth with long continuous enrollment, the increased relative risk of diabetes and longer cumulative concomitant duration were significant for 3 to 5 class PCCWS regimens with atypical antipsychotics, antidepressants, and/or anticonvulsant mood-stabilizers. In view of the continuing growth of PCCWS in the United States, this study supports further investigation of effectiveness and safety of PCCWS regimens, particularly in large, vulnerable community-treated youth populations e.g., foster care.

Footnotes

Author Disclosure Statement

The authors declare that they have no relevant or material financial interests that relate to the research described in this article.

Supplemental Material

References

Alison

. Analysis of tied or discrete data with proc logistic. Survival Analysis Using SAS: A Practical Guide;2ed Cary(2010:235-256). 2010.

Arbogast

, Ray

. Performance of disease risk scores, propensity scores, and traditional multivariable outcome regression in the presence of multiple confounders. Am J Epidemiol, 2011; 174(5):613–620; doi: 10.1093/aje/kwr143

Arbogast

, Ray

. Use of disease risk scores in pharmacoepidemiologic studies. Stat Methods Med Res, 2009; 18(1):67–80; doi: 10.1177/0962280208092347

Baker

, Huefner

, Bellonci

, et al. Polypharmacy in the management of attention-deficit/hyperactivity disorder in children and adolescents: A review and update. J Child Adolesc Psychopharmacol, 2021; 31(3):148–163; doi: 10.1089/cap.2020.0162

Barnard

, Peveler

, Holt

. Antidepressant medication as a risk factor for type 2 diabetes and impaired glucose regulation: Systematic review. Diabetes Care, 2013; 36(10):3337–3345; doi: 10.2337/dc13-0560

Barnett

, Trepman

, Fuson

, et al. Deprescribing psychotropic medications in children: Results of a national qualitative study. BMJ Qual Saf, 2020; 29(8):655–663; doi: 10.1136/bmjqs-2019-010033

Bobo

, Cooper

, Stein

, 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

Brufani

, Crinò

, Fintini

, et al. Systematic review of metformin use in obese nondiabetic children and adolescents. Horm Res Paediatr, 2013; 80(2):78–85; doi: 10.1159/000353760

Burcu

, Safer

, Zito

. Antipsychotic prescribing for behavioral disorders in us youth: Physician specialty, insurance coverage, and complex regimens. Pharmacoepidemiol Drug Saf, 2016; 25(1):26–34; doi: 10.1002/pds.3897

10.

Burcu

, Zito

, Safer

, et al. Association of antidepressant medications with incident type 2 diabetes among medicaid-insured youths. JAMA Pediatr, 2017a;171(12):1200–1207; doi: 10.1001/jamapediatrics.2017.2896

11.

Burcu

, Zito

, Safer

, et al. Concomitant use of atypical antipsychotics with other psychotropic medication classes and the risk of type 2 diabetes mellitus. J Am Acad Child Adolesc Psychiatry, 2017b;56(8):642–651; doi: 10.1016/j.jaac.2017.04.004

12.

D’Agostino

, Lee

, Belanger

, et al. Relation of pooled logistic regression to time dependent cox regression analysis: The framingham heart study. Stat Med, 1990; 9(12):1501–1515; doi: 10.1002/sim.4780091214

13.

GAO. United States Government Accountability Office (GAO). Foster Children: Hhs Guidance Could Help States Improve Oversight of Psychotropic Prescriptions. GAO; 2011.

14.

Glynn

, Gagne

, Schneeweiss

. Role of disease risk scores in comparative effectiveness research with emerging therapies. Pharmacoepidemiol Drug Saf, 2012; 21(Suppl 2):138–147; doi: 10.1002/pds.3231

15.

Greenhill

, Pliszka

, Dulcan

, et al.; American Academy of Child and Adolescent Psychiatry. Practice parameter for the use of stimulant medications in the treatment of children, adolescents, and adults. J Am Acad Child Adolesc Psychiatry, 2002; 41(2 Suppl):26s–49s; doi: 10.1097/00004583-200202001-00003

16.

Greenhill

, Swanson

, Hechtman

et al. Trajectories of Growth Associated With Long-Term Stimulant Medication in the Multimodal Treatment Study of Attention-Deficit/Hyperactivity Disorder. J Am Acad Child Adolesc Psychiatry, 2020; 59(8):978–989; doi: 10.1016/j.jaac.2019.06.019

17.

Horton

, Blum

, Burcu

. Real-world evidence for assessing treatment effectiveness and safety in pediatric populations. J Pediatr, 2021; 238:312–316; doi: 10.1016/j.jpeds.2021.06.062

18.

Kreider

, Matone

, Bellonci

, et al. Growth in the concurrent use of antipsychotics with other psychotropic medications in medicaid-enrolled children. J Am Acad Child Adolesc Psychiatry, 2014; 53(9):960–970.e962; doi: 10.1016/j.jaac.2014.05.010

19.

Legro

, Arslanian

, Ehrmann

, et al.; Endocrine Society. Diagnosis and treatment of polycystic ovary syndrome: An endocrine society clinical practice guideline. J Clin Endocrinol Metab, 2013; 98(12):4565–4592; doi: 10.1210/jc.2013-2350

20.

Magder

, Petri

. Incidence of and risk factors for adverse cardiovascular events among patients with systemic lupus erythematosus. Am J Epidemiol, 2012; 176(8):708–719; doi: 10.1093/aje/kws130

21.

Mikkelsen

, Knop

, Frost

, et al. Use of antibiotics and risk of type 2 diabetes: A population-based case-control study. J Clin Endocrinol Metab, 2015; 100(10):3633–3640; doi: 10.1210/jc.2015-2696

22.

Morrato

, Nicol

, Maahs

, et al. Metabolic screening in children receiving antipsychotic drug treatment. Arch Pediatr Adolesc Med, 2010; 164(4):344–351; doi: 10.1001/archpediatrics.2010.48

23.

Nishi

. Diabetes mellitus and thyroid diseases. Diabetol Int, 2018; 9(2):108–112; doi: 10.1007/s13340-018-0352-4

24.

Persico

, Arango

, Buitelaar

, et al.; European Child and Adolescent Clinical Psychopharmacology Network. Unmet needs in paediatric psychopharmacology: Present scenario and future perspectives. Eur Neuropsychopharmacol, 2015; 25(10):1513–1531; doi: 10.1016/j.euroneuro.2015.06.009

25.

Petrie

, Guzik

, Touyz

. Diabetes, hypertension, and cardiovascular disease: Clinical insights and vascular mechanisms. Can J Cardiol, 2018; 34(5):575–584; doi: 10.1016/j.cjca.2017.12.005

26.

Pini

, Obara

, Battell

, et al. Histamine in diabetes: Is it time to reconsider? Pharmacol Res, 2016; 111:316–324; doi: 10.1016/j.phrs.2016.06.021

27.

Rayner

, McGovern

, Creagh-Brown

, et al. Type 2 diabetes and asthma: Systematic review of the bidirectional relationship. Curr Diabetes Rev, 2019; 15(2):118–126; doi: 10.2174/1573399814666180711114859

28.

Rubin

, Kreider

, Matone

, et al. Risk for incident diabetes mellitus following initiation of second-generation antipsychotics among medicaid-enrolled youths. JAMA Pediatr, 2015; 169(4):e150285; doi: 10.1001/jamapediatrics.2015.0285

29.

Semenkovich

, Brown

, Svrakic

, et al. Depression in type 2 diabetes mellitus: Prevalence, impact, and treatment. Drugs, 2015; 75(6):577–587; doi: 10.1007/s40265-015-0347-4

30.

Stroup

. What can large simple trials do for psychiatry? Am J Psychiatry, 2011; 168(2):117–119; doi: 10.1176/appi.ajp.2010.10111652

31.

Winterstein

, Gerhard

, Kubilis

, et al. Cardiovascular safety of central nervous system stimulants in children and adolescents: Population based cohort study. BMJ, 2012; 345:e4627; doi: 10.1136/bmj.e4627

32.

Zito

, Burcu

, Ibe

, et al. Antipsychotic use by medicaid-insured youths: Impact of eligibility and psychiatric diagnosis across a decade. Psychiatr Serv, 2013; 64(3):223–229; doi: 10.1176/appi.ps.201200081

33.

Zito

, Zhu

, Safer

. Psychotropic polypharmacy in the us pediatric population: A methodologic critique and commentary. Front Psychiatry, 2021; 12:644741; doi: 10.3389/fpsyt.2021.644741

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.10 MB