Increased Risk for Substance Use-Related Problems in Autism Spectrum Disorders: A Population-Based Cohort Study

Up to now, many clinicians have assumed that substance use-related problems are rare among patients with ASD and, if present, primarily due to comorbid ADHD (Palmqvist et al. 2014). This notion has also been supported by clinical studies. A retrospective chart review of 97 youths with ASD found lower rates of substance use compared to psychiatrically treated controls (3.1 vs. 16.7 %); the three boys with ASD and substance use also had comorbid ADHD (Santosh and Mijovic 2006). Similarly, when 70 adults with ASD were compared to 70 subjects with ADHD, substance use rates were lower among those with ASD than ADHD patients (30 vs. 58 %). (Sizoo et al. 2010) However, both ADHD (Groenman et al. 2013) and psychiatrically treated patients (Mangerud et al. 2014) are at increased risk of substance use-related problems, which makes them less suitable as control subjects. To our knowledge, although prior research argued that substance use-related problems are not an issue among ASD individuals (Ramos et al. 2013), no prior clinical study with ASD patients has compared them to non-ASD population controls. However, more recent twin studies provide a different perspective. Lundström et al. were the first to report that autistic-like traits actually increase the risk of substance abuse (OR 7.4; 95 % CI 3.5–15.7) (Lundstrom et al. 2011), findings which were recently confirmed in Australia (De Alwis et al. 2014). These twin studies focused on autistic-like traits as a behavior pattern within the normal spectrum of social interest and competence and similar to that found in ASD, but without investigating the formal diagnostic criteria for persistence, distress or functional impairment required for a diagnosis. Thus, these studies did not address whether an ASD diagnosis was related to substance use-related problems.

Increased risk of substance use-related problems seems to contradict global negative attitudes towards psychoactive substances observed among ASD patients (Ramos et al. 2013). Individuals with ASD may find them helpful to reduce tension and enhance social skills more often than non-ASD controls do (Cludius et al. 2013).

So, why have the idea that ASD patients are somehow protected from substance use-related problems been quite persistent? One possibility is that substance use-related problems in individuals with ASD were indeed less common in the past, but that some factor(s) caused an increase over time. In fact, substantial time trends in substance use have been described for the general population (Kraus et al. 2015b, a). A cohort effect is one of several factors that may explain such temporal changes in substance use-related problems. The broadening of diagnostic criteria has previously been blamed for increase in ASD prevalence (Lundstrom et al. 2015). Hence, while ASD patients diagnosed after 1996 with ICD-10 appeared to have increased risk of substance use-related problems relative to control subjects, prior more narrow diagnostic practice may have excluded ASD patients with substance use-related issues or assigned other diagnoses to them. Thus, the remaining, narrowly defined ASD patient group will be perceived as being “protected” from substance use-related problems.

The risk of substance use-related problems may still differ across patient subgroups depending on ADHD and ID comorbidity. For example, increased rates of substance use-related problems in ASD have been attributed to comorbid ADHD (Palmqvist et al. 2014). Santosh et al. argued that comorbid ID may protect ASD patients from substance use-related problems (Santosh and Mijovic 2006). This was supported by studies suggesting that substance use-related problems were increased among patients with ASD, but without intellectual disability comorbidities (Hofvander et al. 2009; Sizoo et al. 2010).In this study, the increased risk of substance use-related problems suggested among ASD patients was unlikely to result entirely from comorbid conditions, since probands diagnosed solely with ASD had an almost doubled risk of substance use-related problems compared to non-ASD controls. However, comorbid ADHD and ID seemed to modify the overall risk. For example, co-occurring ADHD was associated with further increased risk of substance use-related problems, whereas ID was associated with a lowered risk. It is possible that the highly increased risk among patients with comorbid ADHD is due to diagnostic bias related to interpretation of ICD by clinicians. Taken literally, ICD-10 does not allow a comorbid ADHD diagnosis in the presence of several other diagnoses; ASD, anxiety-, and mood disorders. Simultaneous diagnoses of ASD and ID is allowed, provided that autistic-like symptoms cannot be explained by ID. Thus, clinicians may be reluctant to assign ASD and ID diagnoses to patients already diagnosed with ASD. In contrast, among patients with ASD and later substance use disorder, clinicians may be more likely to exchange ASD with an ADHD diagnosis. To test this possibility, we performed sensitivity analyses with those ASD patients who assigned with ASD, ADHD and ID before a substance use disorder diagnosis. As a result, it turned out that patients with ASD only and ASD with ADHD are actually on comparable risks of substance use-related problems (OR 1.6 vs. 1.9) and previously described extremely high risk in patents with ASD and ADHD seems to be due diagnostic biases.

To further investigate a possible shared familial background to the association between ASD and substance use-related problems, we analyzed the risk among non-ASD relatives of ASD probands. Consistent with prior reports of high rates of alcohol abuse among relatives of ASD patients (Miles et al. 2003) and higher risk of ASD among offspring of parents with alcohol abuse (Sundquist et al. 2014), our results suggested increased risk of substance use-related problems among 1st degree relatives and half-siblings without an ASD diagnosis. This supports a shared familial liability which may in turn reflect one or more possible explanations; shared genetic or environmental factors, or epigenetic mechanisms. First, ASD and substance use-related problems may share genetic risk variants (Zuo et al. 2013). Second, parental substance use disorder may also increase de novo mutation rates, found to be involved in ASD (Sanders et al. 2012). Exposure to psychoactive substances may be related to epigenetic modifications in germ cells (Govorko et al. 2012) and lead to high risk of ASD in offspring. Third, associations between ASD and substance use-related problems may be due to shared environmental factors. For example, exposure to alcohol during pregnancy may lead to fetal alcohol spectrum disorder and autistic-like symptoms within the course of this condition (Stevens et al. 2013). In addition, severely neglected children of parents addicted to psychoactive substance may present symptoms of reactive attachment disorder which, particularly when accompanied with autistic-like symptoms, may increase the probability of receiving an ASD diagnosis (McCullough et al. 2013).

Interestingly, full siblings and parents of ASD probands also had substantially increased risks of substance-related death. This association among parents may be explained by older age at the time of study (median age 47.1 years; IQR 41.4–53.9) enabling sufficient number of outcomes to occur. However, siblings were not older than probands at the time of inclusion (median age 16.9 years; IQR 10.8–22.7). We can only speculate that the same familial factors may be causal in substance use-related problems among ASD probands may lead also to higher risks of substance-related death among their non-ASD relatives. For example, as previously mentioned, a rigid norm-abiding interpersonal style characteristic for ASD may protect from life-threatening activities under the influence of a psychoactive substance.

Strengths include the large scale population-based design, prospectively collected data from nationwide registries, stratification by comorbid disorders, statistical control for socio-demographic confounders, and analysis of familial aggregation data from relatives. Nevertheless, some limitations deserve comments.

First, information bias should be considered. For example, substance use-related problems may be more likely to be detected among individuals with ASD who do have regular contact with habilitation and mental health services. However, similar results were also obtained from other resources; significantly increased risk for substance-related crime from the National Convictions Register and for alcohol-related somatic disease, diagnosed by other medical specialists. An information bias may also act in an opposite direction. Individuals with ASD had not statistically significant, but slightly higher prevalence of substance use disorder prior to ASD diagnosis then healthy controls.

Second, we could stratify our analysis on comorbid ADHD and ID only when those disorders were diagnosed, but it impossible to rule out that patients with seemingly “pure” ASD have undiagnosed ID, ADHD, or subthreshold ADHD-symptoms involving substance use-related problems. Third, the register-based design needs consideration while comparing absolute estimate rates from our study to those from research using population survey designs. For example, our absolute figures on alcohol use disorder are comparable to 1.5 % rates of alcohol abuse among men and 0.7 % among females estimated in a register-based cohort of Swedish citizens born 1973–1984 (Gauffin et al. 2013). Yet, they evidently differ from the 8.9 % 12-month prevalence of alcohol use disorder found in a Swedish population survey of those aged 15 years and older (WHO 2014). In other words, register-based studies are likely to underestimate the absolute prevalence of substance use disorder since users not in contact with the treatment system are not taken into account (EMCDDA 2004).

Fourth, it was not possible to disentangle if the shared familial background between ASD and substance use-related problems were best explained by genetic or shared environmental factors. Finally, we did not differentiate between Autistic disorder, Asperger´s disorder or Pervasive developmental disorder not otherwise specified which might have yielded different associations. Multi-site studies, although, do question the validity of the demarcation between these three disorders (Lord et al. 2012) and the DSM-5 (APA, 2013), groups them all under the one single category of autism spectrum disorders.