Antipsychotic Drug Prescribing in Children Previously Treated With Stimulants for ADHD: A Population-Based Longitudinal Study: La prescription d'antipsychotiques chez les enfants précédemment traités avec des stimulants pour le TDAH : une étude longitudinale basée sur la population

Abstract

Objective

Stimulant drug treatment in preschool-age children for attention-deficit hyperactivity disorder (ADHD) as well as the concomitant use of antipsychotic drugs is largely unstudied in terms of longitudinal outcomes. We characterized longitudinal patterns of stimulant drug use in children diagnosed for ADHD and analyzed the mental health disorders leading to add-on therapy with antipsychotics.

Method

The study population comprised of children and adolescents (age: 0–19 years) in the province of British Columbia (BC), Canada, with at least one dispensing for any psychotropic drug between 1997 and 2017 (N = 144,825). BC health administrative databases were used to identify children with diagnosis for ADHD and dispensings for stimulant and antipsychotic drugs. Longitudinal patterns of drug use and diagnostic codes proximal to the add-on of antipsychotics were assessed.

Results

We found that residence in rural regions and lack of child psychiatrists are significantly associated with higher rates of stimulant drug prescription in preschool and early school-age children. Residence in rural regions was also associated with a higher rate for the concomitant use of antipsychotics over the course of stimulant treatment. When comparing children starting stimulant therapy before the age of 6 with those starting therapy after 6 years, we found an 82% increase in the likelihood of antipsychotic add-on in those starting stimulants at younger ages (HR: 1.82, 95% CI [1.63–2.04]). Moreover, children starting stimulant therapy before the age of 6 years had 3.57-fold higher rates of diagnostic codes for specific delays in development (ICD-9 315) in close proximity to the antipsychotic add-on.

Conclusions

The question remains whether the add-on of antipsychotics is a consequence of insufficient action of the stimulant in ADHD, or required to ameliorate the adverse effects of the stimulant drug. Our result suggests that care need to be taken in the diagnosis for ADHD in children at the age when entering elementary school.

Plain Language Title

Antipsychotic use in children previously diagnosed for ADHD and treated with stimulant medication.

Plain Language Summary

Why was the study done ?

Over the last 20 years, a rapid increase in the prescribing of stimulant medication for treatment of ADHD (attention-deficit hyperactivity disorder) in children is observed in the western hemisphere. More recent studies find that up to 20% of children on continuous stimulant medication suddenly require antipsychotic drugs years after. Antipsychotic drug use in children is a major concern due to metabolic adverse effects, such as weight gain, obesity and diabetes.

What did the research do ?

We investigated prescriptions of stimulant and antipsychotic drugs, and analyzed the medical diagnoses associated with those prescriptions in all the children and adolescents in the province of British Columbia, Canada, over a period of 21 years. Our main objective was the search for factors responsible for antipsychotic prescription, such as child´s age at start of stimulant medication, duration of stimulant medication, child´s sex, the prescribers specialty and region in British Columbia.

What did the research find ?

We were able to identify the age at start of stimulant therapy and rural regions as factors associated with antipsychotic add-on. We found that children who start stimulant medication at the time when entering elementary school or even before, had an 82% higher risk for antipsychotic use later in life, as compared to those children who start with stimulant medication later. In addition, rural regions with a lack of child psychiatrists show more stimulant prescriptions in pre-school children with ADHD and a higher add-on of antipsychotics when compared to urban regions.

What do the findings mean ?

In order to reduce the need for antipsychotic drugs, we suggest that doctors should carefully diagnose ADHD and monitor stimulant use in young children under the age of 6 years. Rural regions need more psychiatry subspecialty services for children available for referral.

Introduction

In the past 20 years, a surge in the prescribing of stimulant drugs such as methylphenidate and amphetamines for treatment of attention-deficit hyperactivity disorder (ADHD) has been observed in higher-income countries, especially in North America.^1,2 The sudden increase in the prescription of drugs against a highly heritable disease raises intriguing questions, since ADHD exhibits the highest level of heritability among all neuro-psychiatric disorders, estimated at 74%. ³ Consequently, factors besides genetics, such as alterations in the diagnostic- and treatment guidelines for ADHD may serve as explanation for the rapidly expanding use of stimulant drugs. Indeed, new guidelines on the management of ADHD recommend a more liberal approach to prescribing in the United Kingdom, such as methylphendiate as first-line pharmacological treatment for children over 5 years of age. ⁴ Moreover, diagnosis of ADHD may be biased by societal pressure to perform in school^5,6 and increases the risk that children, especially boys at the time when entering school, are erroneously diagnosed with ADHD.^7,8 As a result, a certain percentage of children around the age when entering elementary school may receive stimulant drugs due to a misdiagnosis of ADHD.

While restoring the levels of synaptic neurotransmission in ADHD, it remains unknown whether stimulant drugs are able to trigger dopamine neurotransmission beyond the physiological level, especially with long-term use and in the case of patients without ADHD due to a diagnostic error. It is unknown whether adverse effects such as appetite suppression, weight loss, growth retardation and the potential for raising aggression is higher in children without ADHD. A recently published long-term, prospective, controlled study suggests that treatment with methylphenidate for 2 years is safe. ⁹ This study found no evidence for adverse effects on physical growth and neurological and psychiatric outcome in children diagnosed with ADHD. However, the study has excluded children under the age of 6 years and did not account for patients with concomitant psychotropic medication besides stimulants, such as antipsychotics.

Recently, the high prevalence of off-label psychotropic comedication in stimulant users has attracted considerable attention, ¹⁰ especially the add-on of antipsychotics during the course of stimulant therapy.^11,12 Several second-generation antipsychotics, such as risperidone, olanzapine and aripiprazole exhibit antiaggressive and sedative characteristics, which ameliorate irritability and aggression behaviour in children and adolescents with disruptive behaviour disorders, a frequent comorbidity in ADHD. ¹³ However, the metabolic adverse effects in those children, such as weight gain, hyperlipidemia and hyperglycemia are reported with the use of risperidone even with short-term exposure.^14,15 A recent meta-analysis has not found sufficient evidence for the efficacy of antipsychotics in ADHD, ¹⁶ and a comparison between stimulants alone and concomitant treatment with risperidone did not reveal any improvement in scores of ADHD or aggression in a 12-month follow up. ¹⁷ Although risperidone and aripiprazole have previously received FDA approval for the treatment of irritability, aggression, deliberate self-injury and quickly changing moods, the relevant studies included children with preexisting intellectual deficits only. ¹⁸ Since school functioning was never included in the study outcome, the effects of risperidone and aripiprazole on cognitive and intellectual development remains unknown. Overall, there is a lack of long-term data on metabolic- and neurologic parameters, including the cognitive and intellectual development in children on continuous antipsychotic treatment in the absence of preexisting intellectual deficits.

The aim of the current study was to characterize the longitudinal patterns of stimulant drug use in children diagnosed with ADHD, and to assess the reason and pattern for a subsequent add-on therapy with an antipsychotic drug.

Materials and Methods

Study Design and Data Sources

A population-based retrospective observational study was conducted using health administration databases from the province of British Columbia (BC), Canada. BC administration data includes all the general practitioner and specialists visits and prescriptions for all BC residents covered under the medical service plan (MSP). Indigenous patients living on reserve are covered federally through the noninsured health benefits for First Nations and Inuit and are therefore not included in the database.

Data extracted includes all children under the age of 19 years who had dispensings of psychotropic drugs between fiscal years 1997 and 2017. Dispensings were picked up from the pharmacy by patient or caregiver. Research Ethics approval was provided by the University of British Columbia, Children's and Women's Hospital's Research Ethics Board (H18-01247).

The following health resource utilization data were obtained: MSP Payment Information File, ¹⁹ Discharge Abstract Database (Hospital Separations), ²⁰ PharmaNet Data and Consolidation File (MSP Registration & Premium Billing). ²¹ These data files provide patients demographics, diagnosis codes and prescription dispensing records. Prescriptions distributed for hospitalized patients are not in the PharmaNet database which captures the prescriptions from ambulatory care pharmacies.

This study was focused on stimulant- and antipsychotic drug use in children and adolescents up to 19 years of age. Stimulant drugs encompass methylphenidate and amphetamines, including lisdexamfetamine. Supplemental Table S1 lists the antipsychotics.

Our study cohort includes children having used any stimulant drug continuously (at least 180 days in yearly rolling window) for at least 1 year, and with a diagnosis of ADHD up to 3 months before the index date (Supplemental Figure S1). Index date is defined as start of stimulant therapy. Patients using any antipsychotic drug before the index date were excluded. The primary outcome was time to event when patients had antipsychotic medication added to their stimulant therapy. End of treatment was defined as first antipsychotic add-on after index date. For those patients who did not experience antipsychotic add-on, their follow-up time was censored at either their last psychotropic drug use dates, or the dates when turning 19 years of age, or at the study cutoff date (December 31, 2017), whatever comes first.

Statistical Analysis

The rates of stimulant users were characterized and graphically displayed across age, sex and health regions in British Columbia. The Kaplan–Meier method was used to estimate the cumulative incidence rate of antipsychotic add-on with stratification between age group, sex, prescriber specialty and geographical regions to demonstrate rate variation. Cox regression analyses were conducted to calculate the hazard ratios and their 95% confidence intervals to identify factors associated with the add-on of antipsychotics. Statistical analyses were performed using SAS (version 9.4, SAS Inc, Cary NC).

Assessment of Mental Health Diagnoses

Diagnostic codes related to mental health disorders were compared between patients with index date ≤6 and >6 years of age. Stimulant users with a subsequent add-on of antipsychotics at ≥7 years of age were evaluated. Patients were matched between the 2 groups of index date ≤6 and >6 years of age at a ratio 1:3. The matching was applied to correct for the differences in age and sex at the time when antipsychotic was firstly dispensed and diagnoses were assessed. All the ICD-9 codes up to 3 months before the first prescription of the antipsychotic drug were tracked and evaluated. Three months was chosen as the appropriate time window to assess a broader range of diagnostic codes potentially related to adverse events encompassing more than one physician visit. Diagnostic codes of mental health disorders were assigned to the categories of mental health disease as noted in Supplemental Table S2. ICD-9 codes related to laboratory diagnostics (01L) or radiology diagnostics (01X) are shown separately. Diagnostic codes related to mental health disorders, but not mentioned in Supplemental Table S2, are shown and expressed as “remaining mental health codes.” Diagnostic codes not related to mental health disease were evaluated and displayed as “all other nonmental health codes.”

Results

Stimulant Drugs Prescription Patterns

The study identified a total of 12,550 patients who meet the inclusion criteria; 77.9% males and 22.1% females. The mean age of starting stimulant therapy was 9.4 years (sd = 2.9). Broken down in health authority regions the mean age of stimulant initiation for Vancouver Coastal is 9.9 (sd = 3.2); Interior Region 9.3 (sd = 2.9); Vancouver Island 9.4 (sd = 2.9); Fraser Valley 9.2 (sd = 2.8); and Northern region 8.9 (sd = 2.8).

Figure 1 shows the age- and sex distribution of children between birth and 19 years of age who received their first stimulant medication. Although patients up to the age of 6 years represent a minority (8.5%), the highest increase among all new users of stimulants were at the ages between 4–5 (3.3-fold) and ≥5–6 (2.8-fold) years of age. Moreover, patients up to the age of 6 years show the highest number of males in the sex ratio.

Figure 1.

Age and sex distribution at first prescription of stimulant drug between 1997 and 2017 in British Columbia, Canada.

Table 1 shows detailed characteristics of stimulant users across BC health authority regions. While Vancouver Coastal and Fraser Valley represent the urban areas with available access to specialized health care, the Northern Region represents the most remote rural area. In rural areas, the prescribers of stimulant drugs are most likely family practitioners (45.8% at Northern Region vs. 31.1% at Vancouver Coastal). Consequently, only 3.7% of stimulant drug prescribers are licensed psychiatrists in Northern Region, versus 25.3% at Vancouver Coastal. Importantly, the highest percentage of stimulant users under the age of 6 years are found in the rural areas, such as 12.8% and 9.1% in the Northern- and Interior Regions respectively, versus 5.1% at Vancouver Coastal. Methylphenidate is the most frequently prescribed stimulant drug independent of Health Authority Region, with an average prescription rate of 69.6% versus 30.4% for amphetamines.

Table 1.

Characteristics of Stimulant Users and Prescriber Specialty Across BC Health Authorities.

	British Columbia health authority regions:
	Northern region (n = 836, 6.7%)	Interior region (n = 2,770, 22.1%)	Vancouver island (n = 2,110, 16.8%)	Fraser valley (n = 4,641, 37.0%)	Vancouver coastal (n = 2,179, 17.4%)
Sex
Female	165 (19.7%)	628 (22.7%)	500 (23.7%)	988 (21.3%)	497 (22.8%)
Male	671 (80.3%)	2,142 (77.3%)	1,610 (76.3%)	3,653 (78.7%)	1,682 (77.2%)
Age levels
0-5	107 (12.8%)	252 (9.1%)	180 (8.5%)	411 (8.9%)	111 (5.1%)
6-11	550 (65.8%)	1,840 (66.4%)	1,361 (64.5%)	3,165 (68.2%)	1,407 (64.6%)
12-18	179 (21.4%)	678 (24.5%)	569 (27.0%)	1,065 (30.0%)	661 (30.3%)
Prescriber specialty
Psychiatrist	31 (3.7%)	297 (10.7%)	278 (13.2%)	602 (13.0%)	551 (25.3%)
Pediatrician	422 (50.5%)	1,373 (49.6%)	1,098 (52.0%)	2,282 (49.2%)	971 (44.6%)
Physician	383 (45.8%)	1,100 (39.7%)	734 (34.7%)	1,757 (37.8%)	657 (31.1%)
Stimulants type
Methylphenidate	604 (72.3%)	1,884 (68.0%)	1,505 (71.3%)	3,190 (68.7%)	1,560 (71.6%)
Dextroamphetamine	215 (25.7%)	724 (26.1%)	500 (23.7%)	1,119 (24.1%)	399 (18.3%)
Lisdexamfetamine	17 (2.0%)	162 (5.9%)	105 (5.0%)	332 (7.2%)	220 (10.1%)

Note. Table 1 shows the regional distribution of continuous stimulant users at the time of first stimulant prescription (index date) between the Health Authority Regions in BC. Patients were characterized according to sex, age, prescriber speciality and type of stimulant drug. Distribution is expressed as number of patients (n) and their proportion (in brackets).

Add-on Therapy With Antipsychotics

There were 2,659 (21.2%) patients who required antipsychotic add-on into their stimulant therapy; 78.6% males and 21.4% females. The median follow-up time from stimulant initiation up to antipsychotic add-on is 2.1 years (IQR: 0.5-5.5). Risperidone was the most common add-on antipsychotic (58.6%) and subsequent antipsychotic treatment lasts for several years (median 2.4 years, IQR: 0.5-5.9). Moreover, in half of those patients (50.9%), the antipsychotic was further prescribed even beyond the end of stimulant treatment.

Considering the higher prevalence of young stimulant users in rural areas, a Cox regression analysis was conducted with all health authority regions in British Columbia, and comparing children starting stimulant therapy ≤6 years with those >6 years of age. Our analysis revealed that children from rural regions (e.g., Interior Region) were 32% more likely to have antipsychotic add-on (HR: 1.32, 95% CI [1.16-1.50]) (Table 2). Children starting stimulant therapy under the age of 6 years are 82% more likely for antipsychotic add-on than those started over 6 years of age (HR: 1.82, 95% CI [1.63–2.04]) (Table 2 and Figure 2).

Table 2.

Hazard Ratios for Antipsychotic Add-on and Their 95% Confidence Intervals From Unadjusted and Adjusted Cox Regression Models.

Parameters	Unadjusted Models		Adjusted Model
	Hazard Ratios (95% CI)	P-values	Hazard Ratios (95% CI)	P-values
Age at starting stimulants
>6 years	—	—	—	—
≤6 years	1.97 (1.76-2.19)	<0.001	1.82 (1.63-2.04)	<0.001
Sex
Female	—	—	—	—
Male	1.13 (1.03-1.24)	0.012	1.10 (1.01-1.21)	0.043
Number prescriptions of stimulants
<8 prescriptions	—	—	—	—
≥8 prescriptions	1.61 (1.48-1.74)	<0.001	1.55 (1.43-1.68)	<0.001
Health regions
Vancouver Costal	—	—	—	—
Fraser Valley	1.18 (1.05-1.33)	0.007	1.14 (1.01-1.28)	0.031
Vancouver Island	1.21 (1.06-1.38)	0.006	1.17 (1.02-1.34)	0.023
Interior Region	1.38 (1.21-1.56)	<0.001	1.32 (1.16-1.50)	<0.001
Northern Region	1.37 (1.15-1.63)	0.001	1.26 (1.06-1.50)	0.009

Note. Hazard ratios and their confidence intervals, expressing the risk for an add-on of antipsychotic drugs in continuous stimulant users. Risk factors such as index date ≤6 years of age, male sex, ≥ 8 stimulant prescriptions, and residence in Health Authority Regions outside Vancouver Coastal, were compared with their respective counterparts.

Figure 2.

Kaplan–Meier cumulative incidence rates of antipsychotic add-on by age group.

Diagnoses Associated With the Add-on of Antipsychotics

The total number of diagnostic codes given in the period of 3 months before the add-on was used to determine the diagnoses that triggered antipsychotic drug prescribing. We found the rates of diagnoses for most disease categories, such as ADHD, psychosis, pervasive developmental disorders and drug addiction almost equally distributed between the index date ≤6 and >6 years of age groups (Table 3a). The rates for disturbance of conduct, depression and anxiety are slightly differing between children with index date ≤6 years and >6 years (4.6% vs. 6.5% for disturbance of conduct; 12.1% vs. 10.2% for depression and anxiety, respectively). However, we found major disparities for the diagnosis codes related to specific delays in development, radiology diagnostics and mental health disease codes outside our defined categories. Specific delay in development was 2.5% versus 0.7% and radiology codes 2.7% versus 0.8%, both in favour of the group with index date ≤6 years of age. When analyzing the remaining mental health codes in detail, we found unspecified eating disorder (ICD-9 code 307.5) at 1.73% in patients with index date ≤6 years, while this code does not appear in the >6 years of age cohort. Unspecified eating disorder and specific delay in development remain the main cause for the higher rate in mental health codes in the cohort of ≤6 years of age.

Table 3a.

All Diagnostic Codes up to 3 Months Before First-Time Add-on of Antipsychotic Drug. ICD-9 Codes Were Assigned to Disease Categories and Their Rate Expressed in Percentage.

Mental Disease Categories	Start of Stimulant Drug Therapy:
	≤6 years of age N = 197 1,681 codes	>6 years of age N = 591 5,484 codes
ADHD	29.1%	29.6%
Disturbance of conduct	4.6%	6.5%
Depression and anxiety	12.1%	10.2%
Disturbance of emotions and adjustment disorder	5.7%	4.7%
Episodic mood disorders and psychosis	3.5%	3.8%
Specific delays in development	2.5%	0.7%
Pervasive developmental disorders	0.5%	0.6%
Drug abuse and drug addiction	0.3%	0.3%
Remaining mental health codes	3.6%	1.9%
Nonmental health diagnostic codes
Laboratory diagnostics	13.3%	17.6%
Radiology diagnostics	2.7%	0.8%
All other nonmental health codes	22.1%	23.3%

Table 3b.

Diagnostic Codes for Episodic Mood Disorder and Psychosis in Detail. Rates of ICD-9 Codes Are Expressed in Percentage.

	Start of Stimulant Drug Therapy
	ICD-9 codes	≤6 years of age 1,681 codes	>6 years of age 5,484 codes
Drug-induced mental disorders	292	0%	0.034%
Episodic mood disorders	296	2.80%	2.56%
Psychosis-related codes	293; 295; 297; 298	0.70%	1.47%

Note. ADHD: attention-deficit hyperactivity disorder.

(a) 1,681 diagnostic codes related to 197 patients with index date ≤6 years of age, and 5,484 diagnostic codes related to 591 patients with index date >6 years of age were assigned to disease categories. Mental health disorders and their associated ICD-9 codes are shown in Supplemental Table S2. (b) Diagnostic codes related to drug-induced mental disorders (ICD-9 292), episodic mood disorders (ICD-9 296) and psychosis (ICD-9 293; 295; 297 and 298) were analyzed separately.

Drug-induced mental disorders was used for coding in a low percentage (0.034%) of visits in patients with index date >6 years of age (Table 3b). Diagnostic codes associated with psychosis, such as schizophrenic disorders (ICD-9 295) represent 0.7% and 1.47% in patients with index date ≤6 and >6 years of age, respectively (Table 3b).

In order to elucidate the discrepancy in the rate of diagnostic codes related to radiology diagnostics (01X) between the 2 groups, we analyzed all ICD-9 codes unrelated to mental health (all other nonmental health codes). Thereby we found diagnostic codes for fractures of metacarpal bones (ICD-9 815) with a rate of 0.17%, and codes for open wound of head (ICD-9 873) with a rate of 0.3%, exclusively in the index date ≤6 years of age group, which may explain the relative higher rate of radiology-associated diagnostic codes observed.

Discussion

The ongoing brain development in children under the age of 6 years ²² and the 3.3-fold rise in stimulant prescription in the year of school admission (Figure 1) prompted us to investigate the patterns of stimulant use specifically in this young population.

We found the highest percentage of young stimulant users under the age of 6 years in the rural areas of British Columbia (12.8% in rural vs. 5.1% in urban areas) (Table 1). At the same time the density of licensed psychiatrists is the lowest in rural areas (3.7% in rural vs. 25.3% in urban area), suggesting that family practitioners more likely prescribe stimulant drugs to preschool- and early school-age children, given the lack of psychiatry subspecialists available for referral. Rural prescribing, therefore, may be due to the lack of child behavioural and psychiatric services in these communities with the risk for overdiagnosis of ADHD in young children.

The probability for a therapeutic add-on of an antipsychotic drug to Medicaid-insured users of stimulants were reported in 2 previous studies as 19.2% and 22%.^11,23 These results precisely reflect our current data in British Columbia, since 21.2% of users in our continuous stimulant user cohort experience an add-on of an antipsychotic drug during their course of therapy. In contrast, a population-based longitudinal study using German health insurance data, found a probability for an add-on of any antipsychotic drug to methylphenidate users with a previous diagnosis for ADHD as low as 6% over a period of 9 years. ¹² Importantly, this study shares otherwise close similarities with our results, such as the mean age at start of stimulant therapy (10 years of age), the preferred type of antipsychotic drug (risperidone), and an overlapping study period (2004-2013). The major difference compared to our longitudinal study is the proportion of prescriber specialties. In Scholle et al., 42.5% of prescribers were child and adolescent psychiatrists, versus only 16.3% overall in British Columbia. This suggests, that the lack of psychiatric specialty services may not only lead to overprescription of stimulant drugs, but also increase the risk for a subsequent add-on of antipsychotics.

In order to distinguish whether the rural area or rather the young age at start of stimulant therapy is the major responsible factor for an increased risk of antipsychotic add-on, the analysis was extended to encompass all Health Authority regions in British Columbia, comparing children starting stimulant therapy ≤6 years with those >6 years of age. We found an 8.2-fold higher rate for antipsychotic add-on in children starting stimulant therapy ≤6 years (HR: 1.82, 95% CI [1.63–2.04]) (Table 2 and Figure 2). These results suggest that age at start of stimulant therapy is a key factor for predicting subsequent antipsychotic add-on.

We found risperidone as most frequent antipsychotic (58%) for the add-on, and about half of patients (50.9%) continued the antipsychotic beyond their stimulant therapy. The median treatment duration for the antipsychotic was 2.4 years (IQR: 0.5-5.9; mean 3.4 years). Concerns are the high risk for metabolic adverse effects of risperidone, such as weight gain, obesity, insulin resistance and elevation of serum lipids, ²⁴ which are not neutralized or reduced by the concomitant treatment with stimulant drugs. ²⁵

A prospective pediatric study suggests that patients require antipsychotic comedication due to higher disease severity and comorbid conditions in ADHD. ²⁶ However, data from prospective studies or clinical trials may not reflect real world evidence. Several adverse effects of stimulant drugs, especially sleep disturbances, nervousness and irritability can trigger aggressive behaviour, or may be confused with comorbidities associated with ADHD. In our study we found the antipsychotic add-on in average 5.6 years after the index date (median 2.1 years, IQR: 0.5-5.5), which raises the question whether the disease treated with the stimulant drug was different from ADHD, or whether the stimulant itself has triggered an underlying comorbidity. Consequently, the question remains whether the addition of an antipsychotic drug is the consequence of unresolved symptoms of ADHD, or rather an adverse effect inflicted by the stimulant drug.

Analysis of the diagnostic codes in proximity to the antipsychotic add-on revealed that disorders related to psychosis represent 1.3% of all codes up to 3 months before the add-on (Table 3b). This result suggests that psychosis is not the main cause for the observed antipsychotic add-on in the majority of continuous stimulant users. Instead, we found a high prevalence of codes related to disorders associated with irritability, such as disturbance of conduct (ICD-9 312), disturbance of emotions (ICD-9 309) and adjustment disorders (ICD-9 313), which represent altogether a total rate of 11.1% of all codes in close proximity to the add-on of the antipsychotic drug. We conclude that the occurrence of irritability was a major factor for the add-on of antipsychotics in our cohort of continuous stimulant users.

A comparison of continuous stimulant users between index date ≤6 and >6 years of age revealed a 3.6-fold higher rate for diagnostic codes related to specific delays in development and problems with learning (ICD-9 315 and V40.0) in those children who started with stimulant therapy ≤6 years of age (Table 3a). Conclusions on causality remains unclear, since both ADHD and the stimulant drug may impair developmental milestones and learning ability. Moreover, diagnostic codes for fractures of metacarpal bones (ICD-9 815) with a rate of 0.17%, and codes for open wound of head (ICD-9 873) with a rate of 0.3% were exclusively found in the index date ≤6 years of age group. Although, a recent publication reports on a reduced bone mineral density in children and adolescents using stimulant drugs for ≥3 months, ²⁷ conclusions on causality of the observed metacarpal fractures remain unclear, since the hyperactivity and impulsivity related to ADHD are associated with a 2.55-fold increase in the prevalence of fractures in children and adolescents. ²⁸

Interestingly, 1.73% of diagnostic codes were related to unspecified eating disorder (ICD-9 307.59) in those with index date ≤6 years, versus 0% in the index date >6 years of age group. This unspecified eating disorder diagnosis most likely reflects appetite and weight loss causally related to stimulant drug use. It is well known that stimulant drugs decrease appetite and frequently lead to weight loss and growth retardation, ¹⁶ however, an association of this adverse effect with the age at start of stimulant therapy is not reported yet.

One limitation of this study is the lack of information on patients with a diagnosis for ADHD but without any prescription of a psychotropic drug. Patients with less severe symptoms may not receive a stimulant drug. This limitation may cause a bias toward more severe cases of ADHD in our investigated cohort. As discussed in the previous section, another potential confounding relates to severity of ADHD symptoms, such as children treated with stimulants at younger age may have more severe symptoms which could increase the likelihood of antipsychotic add-on years later. Another limitation of this study is the uncertainty of medication adherence. Our information on drug usage is limited to drugs dispensed by the pharmacy and picked up by the patient or caregiver. The use of the drug itself cannot be assured. Stimulant users have the tendency to omit drug intake on weekends and school holidays and stimulants are often prescribed with those instructions. ²⁹ Therefore, we have defined continuous stimulant users as having received dispensings for only 180 days in a rolling 1-year window. However, we cannot estimate the impact of potential drug holidays on antipsychotic add-on.

The strength of this study are its longitudinal nature in drug dispensings combined with medical diagnoses, encompassing the entire spectrum of diagnostic codes in hospital- and ambulatory care settings for every patient between birth and the age of 19 years. Second, the length of the study period and a public health insurance system providing a sufficient number of patients to improve statistical power, even when subgroups are analyzed.

In summary, our study found significant associations between the requirement for antipsychotics and the age of onset of stimulant therapy and the Health Authority Region. Starting stimulant therapy under the age of 6 years, or in rural regions, significantly increased the likelihood for antipsychotic drug-requirement later in life. In order to exclude comorbidities and adjust for severity of ADHD, controlled studies are necessary to determine causality between age of stimulant user and antipsychotic add-on.

Although the current short-term studies up to few months provide strong evidence that stimulant drugs are superior in terms of efficacy compared to placebo, behavioural therapy or nonstimulant medications,^16,30 long-term studies on efficacy and safety are lacking. ³⁰

We suggest that clinicians should carefully diagnose ADHD and monitor stimulant use in male children at the age when entering elementary school. The risk of antipsychotic add-on therapy, in particular of risperidone, is higher in stimulant users. Therefore clinicians should ensure the diagnostic necessity for the add-on of risperidone several years after stimulant therapy.