Stimulant treatment profiles predicting co-occurring substance use disorders in individuals with attention-deficit/hyperactivity disorder

Individuals with attention-deficit/hyperactivity disorder (ADHD) are at increased risk of developing substance use disorders (SUDs) and of starting smoking [1]. Stimulant treatment is the first choice pharmacological treatment of ADHD [2] because it has been proven efficacious in reducing core symptoms of the disorder [3, 4]. In recent years, concerns that stimulant treatment might increase SUDs and smoking in ADHD have been invalidated [5, 6]. One meta-analysis [6] found that stimulant treatment did not affect the development of SUDs or nicotine dependence (ND), whereas the other meta-analysis [5] found a protective effect of stimulants on tobacco use. Possibly, stimulant treatment may have a protective effect in earlier phases of smoking, but not in later stages (i.e., ND). These inconclusive results may be explained by differences in outcome measure severity (smoking vs ND), or indicate unidentified moderators on the development of SUDs and smoking.

Studies have reported earlier initiation of stimulant treatment [7‐9] and longer duration of stimulant use [10] as possibly enhancing the protective effect on the development of SUDs; however, other studies did not replicate these findings [11, 12]. Preclinical studies suggest that the brain may be more sensitive to the effect of stimulants during adolescence (i.e., critical or sensitive age periods) [13]. A recent study predicted substance-related behavior from both age of treatment onset and duration of treatment, and found that short, late-onset stimulant treatment increased the risk of SUDs. Unfortunately, this study did not account for ADHD-severity, a factor related to both stimulant treatment and the risk of SUDs [14], and looked at both factors separately. In contrast, earlier onset of use [7‐9], longer duration [10], and higher treatment continuity [5] could have positive long-term effects on SUDs and smoking. Previous studies have looked at these factors individually, but to the best of our knowledge, no studies have investigated these factors in concert to assess their joint predictive power on SUDs and smoking.

The objective of this study is to investigate how stimulant use profiles are associated with the risk of SUDs and ND. Here, we used a novel technique of community detection to identify distinct subgroups of patients with ADHD based on multiple indicators of stimulant treatment history (i.e., stimulant use profiles). This technique has previously been used in a partly overlapping, but smaller sample where we, successfully predicted increased brain activation during reward receipt in those treated early and intensely [15], in a brain area important in the development of SUDs and smoking. With this study, we build on the previous results comparing stimulant-treated subjects with stimulant-naïve and controls [16]. Start age, treatment duration, total dose, maximum dose, variability, and stop age were derived from highly detailed individual pharmacy transcripts. We hypothesized that adolescents with ADHD who started treatment at younger age, and were treated longer and at a more stable dose, would have a lower risk of SUDs and ND compared to adolescents with a history of later, lower dose, and variable treatment.

We aimed to examine the association between stimulant medication and the development of SUDs and smoking in individuals with ADHD using community detection to construct stimulant use profiles from highly detailed pharmacy records. This allowed us to look at stimulant treatment history in a new manner. We hypothesized that adolescents with ADHD who started treatment at younger age and were treated longer at a more stable dose, would have a lower risk of SUDs and smoking compared to adolescents with a history of later, lower dose and variable treatment. We confirmed our hypothesis and found that those individuals with ADHD who received treatment at a young age and with a high dose, were at a lower risk of developing SUDs and ND, but those who received treatment at a later age with a lower dose were not. This shows that, when looking at stimulant treatment effects, multiple indicators of stimulant medication use should be taken into account.

The current study significantly advances prior studies by being the first to look at multivariate profiles of medication use using highly detailed pharmacy records, as opposed to prior studies looking at a global measure of medication use (yes/no) using self-report scales. Our findings are in line with prior studies looking at indicators of medication use reporting that treatment duration [14] and age of treatment onset [7, 9] affect the development of SUDs and ND. In addition to these treatment characteristics, our data suggest that cumulative dose, maximum dose, and dose variability also play a role in the development of addictive disorders in ADHD. More specifically, confirming our hypothesis, we found that multivariate profiles characterized by a young start age of medication, a high maximum and cumulative dose are associated with lower risks of SUDs and ND in individuals with ADHD.

Our findings regarding smoking showed a distinction between daily smoking and ND. Whereas the early-and-intense group (characterized by a young onset of treatment age, a variable trajectory of medication use with a long duration, high total dosage, high maximum dosage and early age at treatment offset) and early-and-moderate group (characterized by a young onset age, a long duration of use, and a late offset of treatment age) were at lowest risk of developing ND, the risk of daily smoking was unaffected by stimulant use profile. Furthermore, these findings show the need to take severity of nicotine use into account in future studies. A possibility could be that early in the trajectory of ND, stimulant use does not have an effect, but in the later phases (i.e., ND) it does. However, this seems unlikely, as previous studies have suggested a delay in onset of substance-related disorders as a consequence of stimulant use, with much larger effects of stimulant use on the development in adolescence [11, 25, 26] than in adulthood [27, 28]. However, to make conclusive inferences a later follow-up of the current sample is necessary.

In studies of long-term medication effects such as ours, that are inevitably observational, one should be wary of potential confounding by unmeasured variables (i.e., endogeneity). As an example, unmeasured parental characteristics rather than stimulant treatment may account for some of the differences regarding SUD and ND between treatment groups. Similarly, we cannot rule out the possibility that the early-and-intense use and the early-and-moderate use subgroups differed from the late-and-moderate use group with regard to treatment response or factors associated with MPH response (e.g., genetic predispositions), which in turn may drive the association with SUDs and ND. Treatment response was not assessed in the current study. One may argue, however, that treatment response is most likely associated with treatment duration rather than with age of treatment onset. We recommend future studies of long-term stimulant outcomes to take treatment response into account.

The current study has several strengths. First, we introduce a novel approach of looking at stimulant treatment history that allows integrated analysis of multiple related treatment parameters. This data-driven approach resulted in distinct and ecologically valid subgroups, that had predictive validity with regard to important long-term outcomes. Second, we had access to extensive and highly detailed pharmacy records for the majority of our patients with ADHD. Third, unlike many previous studies of tobacco use, we distinguished between smoking and ND and found that indeed the effects of stimulant treatment on these two outcomes are not the same. Some limitations should be noted as well. Long-term outcomes of stimulant medication can only be studied using naturalistic longitudinal studies, that are inevitably at risk for endogeneity, making inferences about causality impossible. Second, we did not have the opportunity to distinguish between different SUDs, or SUDs of different severity, in a similar fashion as we did for smoking and ND, as the numbers of drug use disorders were low in our sample. While the exactness of pharmacy records were very high detailed and gave us the possibility to use community detection, we cannot assure that medications picked up from the pharmacy were actually taken by the individual. Furthermore, our sample is a clinical sample, and we can only draw conclusion on those subjects in clinical practice. As is common with clinical samples, our ADHD sample contains more males than our controls. While we statistically corrected for this, this could have clouded our results. We feel that our results are meaningful since our ADHD sample is representative of those seeking help for their problems. Additionally, comorbidities are common in those seeking help, and while disruptive behavioral problems are most common, other comorbidities are frequent, and should be taken into account in future studies. Especially since treatment effects can be different in different comorbid subgroups (e.g., [29]). Of note, factors associated with early-and intense-treatment, such as higher ADHD-severity and higher levels of CD, are both also associated with a higher risk of SUDs. Interestingly, this group was associated with a lower risk of SUDs and ND. Furthermore, we did not have further information on treatment response and tolerability of treatment, and for this reason we recommend future studies to take these into account. Finally, no data were available regarding psychosocial interventions; if the reduced risk of SUDs reported in this study is associated with a reduction in symptoms of ADHD (caused by the use of stimulants), one would expect to find other treatments with the potential to lower ADHDs core symptoms, such as effective psychosocial interventions, to have a protective effect as well.

In conclusion, we add to current literature by showing that, in stimulant-treated patients with ADHD, there are distinct trajectories of medication use that are differentially related to the risk for SUDs and ND. There is evidence to support the idea that untreated ADHD is related to worse outcomes than treated ADHD [30‐32]. Here, we corroborate this evidence, and expand on this by showing that a medication profile characterized by a late start age, low dose, and low duration of stimulant treatment also has worse outcomes compared to medication profiles with an early start age, high or moderate dose, and long duration of stimulant treatment. We want to emphasize the importance of optimal titration and proper monitoring of stimulant medication in the treatment of ADHD; stimulant treatment should be at adequate dosages to reduce the risk of SUDs and ND as negative long-term outcomes associated with ADHD. Furthermore, our results show an association between starting stimulant treatment at an early age and a reduced risk of developing negative long-term outcomes. Importantly, long-term outcomes of stimulant-treated adolescents with ADHD are associated with treatment characteristics, something that is often ignored when treated individuals are compared to untreated individuals.