Registers
The cohort was established by linking health care registers including the National Patient Register (NPR), with psychiatric diagnoses and dates with codes in accordance with International classification of diseases (ICD) from 1973 (hospitalizations) and 2001 (specialized out-patient care); the Total Population Register with information on year of birth on all residents in Sweden; and the Swedish Prescribed Drug Register (PDR), with information on all filled prescriptions from pharmacies in Sweden from July 1, 2005 (drug class according to the Anatomical Therapeutic Chemical [ATC] Classification System, number of items [tablets/capsules], doses, and dispensation dates, for details see Supplement). The filled prescriptions in the PDR provide information on type and amount of medication taken out from the pharmacy. Written instructions from individual doctors regarding indication or recommended doses for the prescription or information on prescriber’s name or clinic was not available. Ethical approval was granted by the Regional Ethics Review Board in Stockholm (Dnr: 2013/5:8).
Study cohort
From NPR we identified all individuals (children and adults) with two or more diagnoses of ADHD with the first diagnosis of ADHD (ICD-10, F90) appearing from July 1, 2005 and onwards, which matches the launch of the Swedish PDR, allowing us to capture new treatment periods. For flow-chart of the study inclusion, see Supplement figure 1. For distribution of when the year of the first ADHD diagnosis appears in the NPR, see Supplemental table 1. To be able to follow individuals’ new treatment periods during three years until the end of register data in December 31, 2013, we excluded individuals with their first ADHD diagnosis after January 1, 2011. Similarly, to follow filled prescriptions from the PDR during three years, individuals were excluded, if their first filled prescription for ADHD medication occurred after January 1, 2011. Individuals with a first ADHD or ASD-diagnosis registered before age three were excluded. Despite a wide age range for the first diagnosis of ASD and/or ADHD (see Table 1), we chose not to exclude individuals who received an ADHD and/or ASD diagnosis at an advanced age.
From the cohort of individuals with ADHD, we identified those with two or more diagnoses of ASD appearing after age three, according to ICD-9 (299) and ICD-10 (F84.1, F84.5, F84.8, F84.9), but the first ASD diagnosis had to appear before January 1, 2011. ADHD and ASD are both neurodevelopmental disorders, and to be correctly diagnosed symptom onset is required from childhood. Therefore, the first ASD date that appears in the register is not an indicator of onset, but instead indicate when the diagnosis first was decided on in conjunction with the neuropsychiatric assessment. Subsequently, we considered individuals as ASD cases regardless if the first ASD date appeared before or after the first ADHD date (Supplement).
Prescriptions of ADHD medication in Sweden
In Sweden, ADHD medications are usually prescribed in specialized psychiatric units, but may also be prescribed by medical doctors specialized in e.g. pediatric neurology. Individuals must have a documented neuropsychiatric assessment that resulted in an ADHD diagnosis. Between the years of 2005 and 2013, the substances methylphenidate and atomoxetine were approved drugs for ADHD treatment in Sweden by the Swedish Medical Products Agency and recommended for first-line treatment (Swedish Medical Agency). Prescription of dexamphetamine, and amphetamine substances (e.g. Adderall) required a special application to the Swedish Medical Products Agency during the study period. Modafinil was not approved for ADHD treatment, but was prescribed off-label for ADHD. Short-acting and extended release formulations were dispensed for methylphenidate and amphetamines, while for dexamphetamine only short acting formulations were dispensed.
ATC codes for ADHD medication
From the PDR we identified filled prescriptions for the following substances according to the ATC drug class: amphetamine (N06BA01), dexamphetamine (N06BA02), methylphenidate (N06BA04), modafinil (N06BA07), and atomoxetine (N06BA09). There were no dispensations recorded for lisdexamfetamine (N06BA12), guanfacine (C02AC02), dextromethamphetamine (N06BA03) and dexmethylphenidate (N06BA11) during the study period and these were not included in the analysis.
Definitions for treatment with ADHD medication
We defined treatment periods for ADHD medication overall as “no treatment” (no filled prescription from any ADHD medication), “single dispense” (one filled prescription from any ADHD medication) and “continuous treatment” (a sequence of two filled prescriptions from one ADHD medication at separate dates where the gap did not exceed 90 days).
Treatment periods were defined for each separate ADHD medication (i.e. methylphenidate, atomoxetine, dexamphetamine, modafinil, and amphetamine) as “no treatment” or “continuous treatment” (See Supplement for details). Each individual was followed during three years, starting from the first period of “continuous treatment”, and it was possible for one single individual to start a “continuous treatment” with several types of ADHD-medications, either sequentially or simultaneously, and were therefore included in more than one of the separate ADHD medications analyses.
Definitions for methylphenidate daily doses
We studied an approximation of doses per day for methylphenidate based on filled prescriptions in individuals who started a continuous treatment with methylphenidate between July 1, 2005 and January 1, 2011. Individuals were followed during a three-year period and information on dose and number of items per filled prescription was used to calculate the total dose per six months (Supplement figure 2). The dose was divided with number of days to receive an approximation of milligram per day (mg/day). The total dose was set to missing if there were no dispensations during a six-month interval.
Covariates
We adjusted for sex, birth year categories, year of first ADHD diagnosis (Supplement table 1), and psychiatric comorbidities. The comorbid diagnosis had to appear before the first ADHD diagnosis. The following comorbid diagnoses were adjusted for: substance use disorder (alcohol or narcotics), depression, anxiety disorder, obsessive compulsive disorder (OCD), any suicidal attempt, stress related disorders and post-traumatic stress disorder (PTSD), bipolar disorder, psychotic disorder, schizophrenia, eating disorder, and intellectual disability, for ICD-codes see supplement methods.
Statistical analysis
Descriptive statistics were presented as frequencies, percentages, or medians with minimum and maximum values. Group differences were calculated with Pearson χ2-tests for categorical data or Wilcoxon rank sum test for continuous data.
First, we examined whether individuals with ADHD and coexisting ASD are less likely to initiate treatment with ADHD medications overall in all individuals. We then separated into age-categories based on age when the first ADHD diagnosis appeared in the register (adults, age ≥ 18 years, adolescents, age 13-17 years, and children, age ≤ 12 years). We fitted logistic regression models (SAS PROC LOGISTIC) and analyzed the odds for continuous treatment versus no treatment, which included those who had a single dispense. In a secondary analysis, we analyzed the odds for continuous treatment versus no treatment and continuous treatment versus one single dispense. Odds ratios (ORs) were calculated with the associated two-sided 95% confidence intervals (CIs), for the likelihood of being prescribed ADHD medication, comparing individuals with ADHD without coexisting ASD to individuals with ADHD with coexisting ASD. First, we fitted a model without any other covariates, except for sex and birth categories (Birth year<1961, 1961-1970, 1971-1980, 1981-1990, 1991-2000, >2000). Next, we included covariates adjusting for year of first ADHD diagnosis as a continuous linear covariate and the presence of psychiatric comorbidities (yes/no), which could be suspected to confound our OR estimates.
Second, in those who started a continuous treatment we examined whether individuals with ADHD and coexisting ASD are more likely to be prescribed second line treatment options (i.e. dexamphetamine, amphetamine or modafinil). We fitted a logistic regression model for each type of ADHD medication in all individuals and separated into age-categories defined as in the first analysis.
Third, we examined whether individuals with ADHD and coexisting ASD were dispensed different average doses of ADHD medication. Due to non-normal distribution of the data, we applied the Wilcoxon rank-sum test (SAS PROC NPAR1WAY). Methylphenidate doses per day (mg/day) were calculated over six-month intervals, during a total time-period of three years. We also stratified into age-categories defined as in the first analysis. The Wilcoxon rank sum test is a non-parametric test that does not require the data to follow a particular data distribution, e.g. normal distribution, and is robust against single gross outliers [21]. Two-sided p-values of <0.05 corresponding to two-sided 95% confidence intervals not covering the value one were considered as statistically significant. All statistical analyses were performed using SAS 9.4.
Sensitivity analyses were performed, by excluding individuals with their first ADHD diagnosis at age 40 or later, to ascertain that the results were not biased by possible differences regarding coexisting ADHD and ASD in young adults as compared to elderly.