Articles

Understanding the substance use of autistic adolescents and

adults: a mixed-methods approach
Elizabeth Weir, Carrie Allison, Simon Baron-Cohen

Summary
Background Autistic individuals might be more likely to misuse substances than non-autistic individuals. Better Lancet Psychiatry 2021;
understanding of these patterns can help clinicians identify strategies to reduce substance use, protecting physical 8: 673–85

and mental health. The aim of this study was to compare the experiences of substance use between autistic and non- Published Online
July 1, 2021
autistic adolescents and adults.
https://doi.org/10.1016/
S2215-0366(21)00160-7
Methods This study is a mixed-methods study, including both quantitative (closed-ended questions) and qualitative See Comment page 641
(one open-ended question) online assessments. Data were collected as part of a larger study, the Autism and Physical Autism Research Centre,
Health Survey, in which we administered an anonymised, online questionnaire to autistic and non-autistic individuals Department of Psychiatry,
aged 16–90 years. In the present study, we investigated data on substance use or misuse, using two overlapping but University of Cambridge,
separate samples from the survey (one sample with complete quantitative responses and one sample with complete Cambridge, UK (E Weir BA,
C Allison PhD,
qualitative responses). Binary measures of substance use were investigated using unadjusted and adjusted binomial Prof S Baron-Cohen PhD)
logistic regression models. Content analysis was used to compare experiences of autistic and non-autistic adolescents Correspondence to:
and adults. We used Fisher’s exact tests to assess differences in frequency of reporting particular qualitative themes Elizabeth Weir, Autism Research
and subthemes. Centre, University of Cambridge,
Cambridge CB2 8AH, UK
[email protected]
Findings Survey recruitment was done between Feb 7, 2018, and Aug 26, 2019. At the end of the recruitment,
3657 individuals had accessed the survey. After excluding duplicates as well as participants with missing or incomplete
responses, we had data from 2386 participants (1183 autistic and 1203 non-autistic participants; 1571 female and
815 male participants) for the quantitative analyses and data from 919 participants (429 autistic and 490 non-autistic
participants; 569 female and 350 male participants) in the qualitative analyses. The samples for the quantitative and
qualitative analyses were predominantly composed of female individuals, White individuals, UK residents, and those
without intellectual disability. Autistic individuals were less likely than non-autistic individuals to report consuming
alcohol regularly (16·0% of autistic individuals vs 22·2% of non-autistic individuals; adjusted model: odds ratio
[OR] 0·69, 95% CI 0·55–0·86; p=0·0022) or binge-drinking (3·8% vs 8·2%; adjusted model: OR 0·38, 0·26–0·56;
p<0·0001). Autistic male participants were less likely than non-autistic male participants to report ever having smoked
(50·8% of autistic male participants vs 64·6% of non-autistic male participants; adjusted OR 0·50; 0·32–0·76;
p=0·0022) or ever using drugs (35·4% vs 52·7%; adjusted OR 0·53; 0·35–0·80; p=0·0022). Regarding our qualitative
analyses, among participants who reported a specific motivation for drug use, compared with non-autistic individuals,
autistic individuals were nearly nine times more likely to report using recreational substances to manage behaviour
(OR 8·89, 2·05–81·12; p=0·0017) and more likely to report using recreational substances to manage mental health
symptoms (OR 3·08, 1·18–9·08; p=0·032). Autistic individuals were also more likely to report vulnerability associated
with substance use (OR 4·16, 1·90–10·05; p=0·00027), including childhood use of drugs and being forced or tricked
into using drugs.

Interpretation Autistic individuals might be less likely than non-autistic individuals to report engaging in substance
misuse. They also report using drugs to self-medicate. Clinicians should be aware of vulnerability linked to substance
use among autistic patients and should work cooperatively with patients to effectively manage autistic and comorbid
symptoms.

Funding Autism Research Trust, Rosetrees Trust, Cambridge and Peterborough NHS Foundation Trust.

Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.

Introduction ability.1 1–2% of the general population are diagnosed as

Autism spectrum conditions (henceforth autism) are a autistic2 and male individuals are diagnosed three to
group of lifelong, neurodevelopmental conditions denoted four times more frequently than female individuals.2,3
by social and communication difficulties, repetitive There is a growing body of evidence to indicate that
behaviours, restricted interests, and variances in cognitive autistic female individuals might be underdiagnosed
profile, including atypical sensory experience and owing to a different presentation of autism symptoms,
information processing, motor abilities, and intellectual higher rates of internalising difficulties, and higher rates

www.thelancet.com/psychiatry Vol 8 August 2021 673

 Articles

Research in context
Evidence before this study key information about sex differences, highlighting that
We searched PubMed and Google Scholar using various autistic male individuals were less likely to ever have smoked or
combinations of the search terms “autis*”, “substance use”, engaged in recreational drug use than non-autistic male
“substance misuse”, “dependence”, “addiction”, “quantitative”, individuals, whereas there were no significant differences
“qualitative”, and “adult” with no language restrictions for all between autistic and non-autistic female groups. Qualitatively,
studies from database inception until Nov 1, 2017, before our results suggest that autistic individuals were more likely to
beginning the study, and again from database inception until report using substances to manage behavioural symptoms
Feb 11, 2021, after study completion. Existing studies vary (including autism symptoms) and using them to manage
greatly in size and scope. Multiple large population-based mental health symptoms than were non-autistic individuals.
studies suggest that autistic individuals have an increased risk The findings also provide evidence for the reduced likelihood of
of substance misuse or misuse. Several qualitative studies with reporting social motivations for drug use among autistic
small sample sizes have been collectively described in reviews individuals compared with non-autistic individuals. New areas
and meta-analyses. They have identified motivations, of self-reported vulnerability have been identified by this study,
protective factors, and risk factors for substance use or misuse including childhood use of drugs and being forced or tricked
among autistic individuals; however, none of these studies into using drugs.
have attempted to quantify the relative likelihood of autistic
Implications of all the available evidence
and non-autistic adolescents and adults reporting these
Health-care providers should work with autistic people to
behaviours or experiences.
identify and effectively manage the autistic symptoms as well
Added value of this study as the comorbid behavioural, mental, and physical health
This mixed-methods study provides new evidence of symptoms that require additional support, to prevent
differences in smoking and alcohol use, as well as differences in self-medication and possible substance misuse. Clinicians
motivations for substance use, among autistic and non-autistic should be aware of increased risk of adverse life events for
adolescents and adults. This analysis includes large samples of autistic individuals, some of which might be connected to
autistic female individuals and older autistic adults, which are substance use. This study reaffirms the importance of early
groups that remain neglected in research. Our findings show autism diagnosis and supportive health care across the lifespan.

of camouflaging.4–7 Camouflaging can be defined as (including 7528 autistic individuals), autistic individuals
altering one’s behaviour or personality traits to align with were twice as likely to have substance use problems than
social norms; and compensation (a related but distinct non-autistic indi­ viduals; and even their non-autistic
concept) can be understood as conscious or subconscious siblings and parents were at increased risk of substance
processes that allow individuals with neurodevelopmental use problems compared with controls, suggesting that
conditions (eg, autism) to reduce the presentation of their genetic or environmental factors might contribute to
symptoms to others, despite ongoing difficulties. Autistic risks.17,19 Additional diagnoses of ADHD and intellectual
individuals might attempt to use strategies of cam­ disability seem to moderate the risk of substance misuse,
ouflaging or compensation to minimise or obscure their with an ADHD diagnosis increasing the risk and a
autism symptoms to align with societal expectations of diagnosis of intellectual disability decreasing the risk.17–19
behaviour;5,8 however, this behaviour can come at huge Existing studies indicate that participants with ADHD,
cost, worsening mental health problems and even ADHD and autism, as well as those with other
increasing risk of suicidality.4,5,8,9 Please note that we use developmental disorders were all at greater risk of
identity-first language (eg, autistic individual) throughout substance use problems than the autistic participants,
the manuscript, as this terminology is preferred by the making it difficult to quantify the risk of substance use
majority of the autistic community in the UK.10 or misuse that is specific to autism—even in large,
Several studies taken from both clinical and general population-based samples.17,19,20
population samples (with sample sizes ranging from Several population-based studies of neurotypical adults
89 to 4123 autistic participants) suggest that autistic suggest that substance use or misuse (including alcohol,
individuals are less likely to smoke, use tobacco, use tobacco, prescription and recreational drugs) is associated
nicotine,11–13 or misuse substances (including alcohol) with many physical health risks, including respiratory
than non-autistic individuals.12–16 By contrast, results from problems, cancer, heart disease, hypertension, heart attack,
larger studies and one systematic review indicate that stroke, reproductive morbidity, diabetes, liver damage or
autistic individuals might have an increased likelihood disease, and sleep conditions.21 In a previous study that we
of developing substance use-related problems.17–20 did with the same sample as the current study, we found
Specifically, in large population-based studies in Sweden that substance use was not associated with increased
(including 26 986 autistic individuals) and Norway health risks among the autistic indi­viduals;22 however, this

674 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

study only used current alcohol consumption and greatest website Qualtrics to collect data regarding demographics,
smoking frequency as covariates, which does not capture autistic traits (we used a short version of the Autism
all aspects of substance use. In addition to physical health Spectrum Quotient, AQ-10, administered to non-autistic
risks, substance misuse might negatively affect the quality participants only), lifestyle-related factors (including both
of life of autistic individuals and exacerbate existing quantitative and qualitative substance use information),
difficulties with functional outcomes, such as main­tenance personal medical history, and family medical history. We
of employment and education.18 used a convenience sampling framework to recruit
Six studies have attempted to establish motivations, participants via the Cambridge Autism Research Database,
protective factors, and risk factors for substance use or Autistica’s Discover Network, autism support groups and
misuse using qualitative methods;18,23 however, these charities, as well as social media (especially Twitter and
studies included small samples of autistic individuals Facebook). Because of our recruitment strategies, our
(n<50) and even fewer autistic females (n≤16). Autistic control sample might have been biased towards
individuals were more likely to use substances to individuals with an interest in autism or those with
compensate for comorbid mental health conditions undiagnosed autism. In an attempt to recruit a general
(as well as psychological distress) and perceived social population sample, we advertised the study via Facebook
difficulties; weak executive functioning, maladaptive and did not target specific autism groups or forums. All
coping style, late autism diagnosis, few social resources, advertisements encouraged participation from both
lack of structure in daily life or leisure activities, family autistic and non-autistic adolescents and adults.
history of substance misuse, early smoking onset, and All questions related to substance use were developed
adverse childhood experiences were additional risk by our research team by consulting publicly available
factors for substance use or misuse.18,23 information and questions from surveys from the UK
Only two studies considered sex or gender differences National Health Service, UK National Institute for Health
in substance misuse.14,19 In the general population, male and Care Excellence, US National Institutes of Health,
individuals are far more likely to use and misuse and WHO. All substance use data were collected as part
substances than female individuals;24 yet, this pattern of APHS, with the dual purposes of describing the
appears more complex among autistic individuals, with substance use or misuse of autistic individuals directly,
smaller differences between male and female while also con­ sidering the relationships between
individuals.14,19 Crucially, none of these studies consider substance use and risk of physical health conditions
sex or gender differences in substance use patterns or in among autistic individuals.22 The present analyses used
qualitative studies when considering motivations, binary measures of smoking and alcohol use (the
protective factors, or risk factors for substance use or binarisation was done before looking at the results).
misuse. These analyses also incorporated measures of
Research into the substance use of autistic individuals recreational drug use and second-hand smoke exposure
is limited in sample size and scope; however, it is clear (which were originally measured as binary [yes or no]
that differences in substance use might leave autistic responses). Analysing the data in this binarised structure
individuals susceptible to wide-ranging negative con­ allowed us to establish cutoff points for possible
sequences regarding daily functioning and physical substance misuse (eg, ≥5 alcoholic drinks per average
health. The present study attempts to explore whether session, consumption of alcohol on ≥3 days per week)
there are quantitative or qualitative differences in rather than describing substance use more generally, as
substance use between autistic and non-autistic the binarised data might have greater clinical relevance.
adolescents and adults. As questions relating to substance use might be
considered sensitive to some participants, we made all
Methods questions optional, and participants were informed of
Study design and participants this at the beginning of this section of the survey. We had
This study is a mixed-methods study, including both high response rates (>98·99%) to all quantitative
quantitative (closed-ended questions) and qualitative questions. The specific phrasing of all relevant questions
(one open-ended question) online assessments. Data were have been provided in the appendix (pp 5–8). See Online for appendix
collected as part of a larger study, the Autism and Physical This study obtained ethics approval from the University
Health Survey (APHS). Any consenting individual of at of Cambridge Human Biology Research Ethics com­mittee
least 16 years of age was eligible to participate in the (HBREC.2017.28). Written, online informed consent was
APHS. To include as many relevant participant records as obtained from all participants. The study protocol has
possible, the quantitative and qualitative analyses of our been included in the appendix (pp 12–21).
study include two different but overlapping samples—
one sample with complete quantitative responses and Procedures
one sample with complete qualitative responses—both The qualitative section of the survey relied on responses to
taken from the APHS dataset. We used an anonymous, a single question related to substance use. All participant
self-report, cross-sectional survey in English via the records without a response to this question were excluded

www.thelancet.com/psychiatry Vol 8 August 2021 675

 Articles

their own designation for autism diagnosis. Although

3657 responses this choice makes it possible that the qualitative section
of our study underestimates true group differences in
1102 excluded due to incomplete response the frequency of reporting particular codes, subthemes,
or themes between autistic and non-autistic individuals,
we felt it was important to explore the real-world
2555 responses experiences and opinions of self-diagnosed and
undiagnosed individuals.
The non-autistic group of the quantitative sample
112 duplicates excluded 1600 excluded due non-response included individuals who provided relevant responses
to qualitative question without an autism diagnosis and for whom an autism
diagnosis or duplicate response were not suspected. The
non-autistic group of the qualitative sample included all
2443 responses to quantitative question 955 responses to qualitative question
non-duplicate individuals who provided a relevant
response and self-identified as non-autistic (even if
56 excluded due to unconfirmed 36 duplicates excluded autism was suspected or the individual was awaiting
autism status autism diagnosis).
Sex assigned at birth was self-reported as female, male,
2387 responses or other; current gender identity was also recorded but is
not included as part of this analysis. Education level was
coded as a categorical variable and used as a proxy
1 excluded due to responding
“other” for biological sex
measure of socioeconomic status; it was defined as the
highest qualification held with the following options: no
formal qualifications, secondary school or high school
2386 responses included in final 919 responses included in final level qualifications, further vocational qualifications,
quantitative sample qualitative sample
university undergraduate level qualifications (BA,
BSc, etc), and university postgraduate level qualifications
Figure 1: Study profile
(MA, MSc, PhD, Certificate, etc). Because of low
response rates of individuals from non-White ethnic
from the qualitative sample. Then, all suspected duplicate backgrounds, we used a binary representation of
records were eliminated from the quantitative and ethnicity (White vs non-White) in all of our analyses.
qualitative samples. As we did not collect any personally Descriptive breakdowns of ethnicity for both the
identifiable data, we developed an algorithm to exclude quantitative and qualitative samples are provided in the
potential duplicate responses. Any responses that matched appendix (p 5). We derived a categorical variable of
a previous response across the following 11 criteria were country of residence based on frequency with the
excluded: autism diagnosis (yes/no), specific autism following options: UK, USA, Germany, Australia, and
diagnosis, type of diagnosing practitioner, year of autism other.
diagnosis, country of residence, sex assigned at birth, We included one open-ended, free-text question,
current gender identity, education level, age, maternal age namely: “Please list any recreational substances/drugs
at birth, and paternal age at birth. To provide the most you have used and how long you used them for. Please
conservative analysis, we did not exclude individuals with provide any information that you think may be
a high autism quotient score from the control group. relevant.” With this question, we intended to give
The autistic cohorts of each sample included all autistic and non-autistic individuals the opportunity to
individuals who self-reported an autism diagnosis made provide information related to their experiences of
by a medical practitioner. As the survey was anonymous, substance use organically and to explore various topics
participants did not provide diagnostic assessments; yet without explicit direction of themes from the
we required that they disclose additional information to questionnaire. We used the content analysis method,
verify their diagnosis: type of practitioner who diagnosed because it flexibly uses some principles of thematic
them, the year of their diagnosis, their specific analysis while also providing the opportunity to
diagnosis, and whether they have a syndromic form of quantify the relative frequency of reporting of specific
autism. Individuals whose autism status could not be code-level and thematic data.25–27 As the analysis was
confirmed were excluded from both the autistic and intended to be exploratory in nature, EW developed
control cohorts of the quantitative sample. By contrast, codes inductively to answer two research questions:
the qualitative sample included individuals who self- (1) what experiences or themes related to substance use
diagnosed as autistic, suspected autism, or were do autistic and non-autistic individuals choose to
awaiting autism assessment; they were included in discuss without explicit direction, and (2) are there
either the autistic or non-autistic cohorts on the basis of differences in the relative frequency of discussion of

676 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

these themes between autistic and non-autistic Results

individuals. EW organised the codes into an original Survey recruitment began on Feb 7, 2018, and ended on
schema with possible themes and subthemes identified Aug 26, 2019; with two pauses in survey recruitment (from
on the basis of the explicit, word-level information from May 13, 2018, to June 28, 2018, and from April 8, 2019, to
the participants (rather than at the latent, interpretive Aug 20, 2019) to consider different means of advertising
level); all authors then discussed and clarified these the survey. No changes were made to the survey after these
themes or categories and established the final structure pauses. We did a sensitivity analysis covarying for each of
together. the three time periods for adjusted regression analyses;
there were no significant differences, as determined by
Statistical analysis z-tests and full results are provided in the appendix (p 4).
We used R version 3.6.2 to do all quantitative analyses. At the end of the recruitment, 3657 individuals across
We employed both unadjusted and adjusted models 62 different countries accessed the survey; 1102 (30%) of
(binomial logistic regression using the “glm” function the original participant responses were excluded due to
from the “stats” package); the adjusted model controlled incomplete responses, meaning that they withdrew from
for self-reported sex assigned at birth, age, ethnicity, the survey before completing required questions related
education level, and country of residence. Descriptive to their demo­ graphics (figure 1). 783 (71%) of these
demographic statistics were done using the “CrossTable” individuals withdrew from the survey before answering
function from the “gmodels package” for Pearson’s χ²
tests (for categorical and binary covariates) and the
Autistic group Non-autistic group Effect size p values
“wilcox.test” function from the “stats” package for the
Mann–Whitney U tests (for continuous covariates). Quantitative sample demographics (1183 autistic individuals, 1203 non-autistic individuals)
Missingness for the covariates of age, education level, Age, years 41·04 (14·41) 41·86 (15·59) 0·019* 0·34
ethnicity, and country of residence was addressed by Age categories, years
creating five completed datasets (using predictive mean 16–29 303 (25·61%) 311 (25·85%) ·· ··
matching for five imputations) and pooling the results 30–39 250 (21·13%) 240 (19·95%) ·· ··
according to Rubin’s rules using the “mice” and “pool” 40–49 252 (21·30%) 252 (20·95%) ·· ··
functions of the “MICE” package.28 To minimise 50–59 214 (18·09%) 206 (17·12%) ·· ··
type 1 errors from multiple testing, we used the false 60–69 113 (9·55%) 127 (10·56%) ·· ··
discovery rate correction and used a p threshold of 0·05 ≥70 25 (2·11%) 52 (4·32%) ·· ··
across all analyses.29 Missing 26 (2·20%) 15 (1·25%) ·· ··
We ran a binomial logistic regression model controlling Biological sex ·· ·· 0·058* 0·0045
for all the covariates listed previously, as well as the Female 746 (63·06%) 825 (68·58%) ·· ··
interaction of sex and diagnosis for all outcomes in our Male 437 (36·94%) 378 (31·42%) ·· ··
main analyses. If there was a significant interaction, we Ethnicity ·· ·· 0·055* 0·0068
used the “glht” function of the “multcomp” package (for White 1045 (88·33%) 1020 (84·78%) ·· ··
the adjusted model) and sex-stratified Fisher’s exact tests Non-White 135 (11·42%) 183 (15·21%) ·· ··
(for the unadjusted model) to estimate sex-specific values Missing 3 (0·25%) 0 ·· ··
and reported the sex-specific results in place of the main Education ·· ·· 0·094† <0·0001
models. No formal qualifications 57 (4·82%) 14 (1·16%) ·· ··
Regarding the qualitative analysis, we used NVivo Further vocational 215 (18·17%) 138 (11·47%) ·· ··
version 12 (for the coding process) and R version 3.6.2. qualifications
Using the frequency data from the content analysis, we Secondary school or high 211 (17·84%) 171 (14·21%) ·· ··
did Fisher’s exact tests to quantify the relative frequency school
of self-reporting aspects of recreational drug use among University undergraduate 354 (29·92%) 354 (29·43%) ·· ··
autistic and non-autistic individuals. To reduce risk of University postgraduate 344 (29·08%) 523 (43·47%) ·· ··
type 1 errors, we ran tests of significance to identify the Missing 2 (0·17%) 3 (0·25%) ·· ··
relative frequency of discussing particular themes, rather Country of residence ·· ·· 0·053* <0·0001
than using individual codes to do so. We limited our UK 842 (71·17%) 759 (63·09%) ·· ··
analyses on subthemes to participants who reported USA 120 (10·14%) 174 (14·46%) ·· ··
information related to that theme, rather than on the Germany 31 (2·62%) 33 (2·74%) ·· ··
entire sample. We used the false discovery rate correction Australia 33 (2·79%) 20 (1·66%) ·· ··
and a p threshold of 0·05 across all analyses. Other 156 (13·19%) 214 (17·79%) ·· ··
Missing 1 (0·08%) 3 (0·25%) ·· ··
Role of the funding source Intellectual disability ·· ·· 0·071* 0·00054
The funders of the study had no role in study design, Self-identified 21 (1·78%) 4 (0·33%) ·· ··
data collection, data analysis, data interpretation, or (Table 1 continues on next page)
writing of the report.

www.thelancet.com/psychiatry Vol 8 August 2021 677

 Articles

individuals, White individuals, UK residents, and those

Autistic group Non-autistic group Effect size p values
without intellectual disability (table 1). There were
(Continued from previous page) significant group differences, and these differences were
Qualitative sample demographics (429 autistic individuals, 490 non-autistic individuals) expected based on the meth­ odology and recruitment
Age, years 40·95 (14·42%) 41·81 (15·60%) –0·047* 0·15 strategies used. There were no significant differences in
Age categories, years age between the autistic and control groups; the mean
16–29 76 (17·72%) 120 (24·49%) ·· ·· age was 41·04 years (SD 14·41) in the autistic group and
30–39 108 (25·17%) 114 (23·27%) ·· ·· 41·86 years (15·59) in the control group of the quantitative
40–49 128 (29·84%) 121 (24·69%) ·· ·· sample. Data presented in table 1 are demographic data
50–59 64 (14·92%) 77 (15·71%) ·· ·· before imputation. The results remain highly similar
60–69 35 (8·16%) 45 (9·18%) ·· ·· after imputation. The mean ages of the qualitative
≥70 10 (2·33%) 8 (1·63%) ·· ·· sample were similar and can also be found in table 1.
Missing 8 (1·86%) 5 (1·02%) ·· ·· Overall, we found some quantitative differences in
Biological sex ·· ·· 0·047* 0·13 alcohol use between autistic and non-autistic adolescents
Female 278 (64·80%) 291 (59·39%) ·· ·· and adults. Compared with non-autistic individuals,
Male 150 (34·97%) 199 (40·61%) ·· ·· autistic individuals were less likely to report regularly
Other 1 (0·23%) 0 ·· ·· consuming alcohol (≥3 days per week on average;
Ethnicity ·· ·· 0·057* 0·083 16·0% of autistic individuals vs 22·2% of non-autistic
White 383 (89·28%) 421 (85·92%) ·· ·· individuals; adjusted model: odds ratio [OR] 0·69,
Non-White 44 (10·26%) 69 (14·08%) ·· ·· 95% CI 0·55–0·86; p=0·0022) or engaging in binge-
Missing 2 (0·47%) 0 ·· ·· drinking (≥5 alcoholic beverages per average session;
Education ·· ·· 0·10† <0·0001 3·8% vs 8·2%; adjusted model: OR 0·38, 0·26–0·56;
No formal qualifications 15 (3·50%) 9 (1·84%) ·· ··
p<0·0001; table 2). The binomial logistic regression
Further vocational 84 (19·58%) 59 (12·04%) ·· ··
analyses that included an additional interaction term for
qualifications sex and diagnosis showed significant interactions for
Secondary school or high 70 (16·32%) 47 (9·59%) ·· ·· ever having smoked, smoking weekly or more, and ever
school having used recreational substances (table 2). Autistic
University undergraduate 142 (33·10%) 150 (30·61%) ·· ·· male individuals were less likely than non-autistic male
University postgraduate 118 (27·51%) 225 (45·92%) ·· ·· participants to report ever having smoked (50·8% of
Missing 0 0 ·· ·· autistic male participants vs 64·6% of non-autistic male
Country of residence ·· ·· 0·060* 0·0099 participants; adjusted OR 0·50; 95% CI 0·32–0·76;
UK 300 (69·93%) 289 (58·98%) ·· ·· p=0·0022) or ever having engaged in recreational drug
USA 49 (11·42%) 82 (16·73%) ·· ·· use (35·4% vs 52·7%; adjusted OR 0·53; 95% CI
Germany 13 (3·03%) 13 (2·65%) ·· ·· 0·35–0·80; p=0·0022). There were no significant
Australia 8 (1·86%) 10 (2·04%) ·· ·· differences between female groups (table 2). Bar graphs
Other 59 (13·76%) 96 (19·59%) ·· ·· showing the full distribution for responses regarding
Missing 0 1 (<1%) ·· ·· alcohol use and smoking are provided in the
appendix (pp 9–11).
Data are mean (SD) or n (%). p values and effect sizes were from Pearson’s χ² tests (for binary or categorical covariates)
Regarding our qualitative analyses, we identified
or from a Mann-Whitney U test (test of differences in means for continuous covariates). Effect sizes were based on
measures of Phi (φ) for binary covariates or Cramér’s V (V) for categorical covariates, or for measures of r (equal to 111 individual codes inductively that correspond to
Z statistic/√[sample size]) for continuous covariates. All records with missing outcome data were excluded from the various aspects of the experience of substance use for
relevant test. *Indicates a small effect. †Indicates a small-to-medium effect. 919 autistic and non-autistic individuals. Figure 2
Table 1: Participant demographics provides selected codes and the themes (and subthemes
for the “barriers” and “motivations for using” themes). A
version of the figure that incorporates all codes into the
any survey questions. An additional 56 (2%) individuals structure can be found in the appendix (p 1). The panel
were excluded from the quantitative sample due to provides selected quotes supporting each of the themes
unconfirmed autism diagnosis status (specifically, self- and subthemes identified.
diagnosed as autistic, suspected autism, or were awaiting We also identified the frequency of all codes for autistic
autism assessment). However, in the qualitative sample, and non-autistic individuals separately (appendix pp 2–4).
these individuals were not excluded and were instead Autistic individuals were less likely than non-autistic
incorporated into the autistic (n=6) or non-autistic (n=23) individuals to provide any qualitative information about
cohorts on the basis of their own designation. One intersex their substance use (OR 0·83, 95% CI 0·70–0·98;
participant was excluded from the quantitative sample, as p=0·045). Autistic individuals were more likely to report
our analysis strategy controlled for sex. an atypical response to drugs (regarding their experience
Both the samples for the quantitative and qualitative or tol­erance, or both; OR 6·56, 2·20–26·38; p=0·00027),
analyses were predominantly com­ posed of female as well as experiences of substance use that indicate

678 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

Autistic group Non-autistic group Unadjusted model Adjusted model

Yes, n* Total, n† % Yes, n* Total, n† % Odds ratio (95% CI) p values Odds ratio (95% CI) p values
Consumes alcohol ≥3 days per week 189 1182 15·99% 267 1202 22·21% 0·67 (0·54–0·82) 0·00043 0·69 (0·55–0·86) 0·0022
Consumes ≥5 alcoholic beverages per average 45 1182 3·81% 98 1202 8·15% 0·45 (0·31–0·64) <0·0001 0·38 (0·26–0·56) <0·0001
session
Ever smoked (female individuals) 367 746 49·20% 447 824 54·25% 0·82 (0·67–1·00)‡ 0·077 0·81 (0·60–1·09)§ 0·47
Ever smoked (male individuals) 222 437 50·80% 244 378 64·55% 0·57 (0·43–0·75)‡ 0·00036 0·50 (0·32–0·76)§ 0·0022
Smoking weekly or more (female individuals) 236 746 31·64% 271 824 32·89% 0·94 (0·76–1·17)‡ 0·67 0·93 (0·68–1·29)§ 1·00
Smoking weekly or more (male individuals) 143 437 32·72% 146 378 38·62% 0·77 (0·58–1·03)‡ 0·11 0·64 (0·41–1·00)§ 0·077
Second-hand smoke exposure 384 1178 32·60% 336 1184 28·38% 1·22 (1·02–1·46) 0·052 1·19 (0·99–1·42) 0·10
Ever used drugs (female individuals) 280 745 37·58% 294 822 35·77% 1·08 (0·88–1·33)‡ 0·55 1·11 (0·82–1·50)§ 1·00
Ever used drugs (male individuals) 154 435 35·40% 199 378 52·65% 0·49 (0·37–0·65)‡ <0·0001 0·53 (0·35–0·80)§ 0·0022
Unless otherwise specified, the unadjusted model refers to the outcome tested in the full population and the adjusted model refers to the binomial logistic regression analyses adjusted for age, biological sex,
ethnicity, education, and country of residence. *Number of respondents who designated each statement as true. †Number of respondents who answered each question. ‡Used a sex-stratified unadjusted model
to estimate sex-specific odds ratios, CIs, and p values. §Adjusted model refers to sex-specific outputs of binomial logistic regression analyses adjusted for age, biological sex, interaction of sex and diagnosis,
ethnicity, education, and country of residence.

Table 2: Substance use of autistic individuals compared with non-autistic individuals

vulnerability (OR 4·16, 1·90–10·05; p=0·00027), inclu­ between these patterns for autistic female individuals. To
ding being forced or tricked to use substances, childhood our knowledge, this finding is the first evidence that
use of substances, suicidality, trauma, and addiction or autistic male individuals are par­ ticularly unlikely to
dependence (table 3). engage in substance use compared with non-autistic
Among those who reported a specific motivation for male individuals; it might also suggest that the sex-
using drugs, autistic individuals were more likely to specific pattern of substance use in the general population
report using substances to manage mental health (eg, male individuals are more likely than female
symptoms and behaviour, but were less likely to report individuals to engage in substance use) might be different
social motivations for drug use (table 3). Among to that of autistic individuals,24 although two previous
participants who reported a specific motivation for drug studies showed small or non-significant sex differences.14,19
use, autistic individuals were nearly nine times more Cross-sectional convenience samples (as used in the
likely to report using recreational substances to manage present study) provide a unique opportunity for rec­
behaviour than non-autistic individuals (OR 8·89; ruitment of large samples of diagnosed autistic female
95% CI 2·05–81·12; p=0·0017) and over three times individuals,3–7 without which analyses on sex differences
more likely to reporting using recreational substances would not be possible. This work emphasises the
to manage mental health symptoms than were non- importance of recruiting autistic female individuals in
autistic individuals (OR 3·08; 95% CI 1·18–9·08; research, as they might have unique risk factors and are
p=0·032). Autistic individuals also reported using now widely reported to have worse outcomes regarding
substances as a form of self-medication for physical physical health, mental health, and mortality.14,22,30
health symptoms (including sleep, digestion or eating, Among those who provided any information regarding
and pain) but the frequency of this reporting was their motivations for using drugs, autistic individuals
not different from that of non-autistic individuals. were nearly nine times more likely than non-autistic
Five individuals specifically reported that receiving their individuals to report using them to manage behaviour
autism diagnosis was relevant to reducing or ending specifically. Although the terms masking, compensation,
their substance use. and camouflaging were not directly used by participants
and might not apply to all instances of descriptions of
Discussion managing behaviour, the descriptions provided fre­quently
Autistic individuals were less likely to report consuming correspond to existing literature and definitions of these
alcohol regularly or binge-drinking than non-autistic concepts;4–8 individuals described using drugs to
individuals; however, our mixed-methods approach eliminate, control, or reduce autism or symptoms of
revealed possible points of concern regarding the autism (eg, sensory overload, stimming behaviour,
substance use of autistic individuals, including wide- improving overall function, improving perception) and
ranging sex differences and qualitative differences in other comorbid symptoms (eg, ADHD). One autistic
motivations for drug use. Autistic male individuals were individual noted “I smoke pot to make my anxiety and
far less likely to report ever having smoked or having ever autism go away. It’s the only time I fell on the same wave
used recreational substances than non-autistic male length as everyone else”). Although causal links have not
individuals, whereas there were no significant dif­ferences yet been established, compensation and camouflaging

www.thelancet.com/psychiatry Vol 8 August 2021 679

 Articles

Autism diagnosis helped to reduce or

stop use
Forced, tricked, or accidentally took
drugs Vulnerability
Used to deal with trauma
Childhood use Positive experience
Dependence or addiction Not addicted or habit
No unusual experiences or bad trips Non-negative experience
No negative effects or long-term damage
Did not impede daily functioning

Did not get high or no effect

Atypical response
Abnormal effect

Stopped using temporarily

Adolescent use of substances
Regular user Usage Information
Long-term user
Heavy use
Test effect
Used when provided
No sense of risk Experimental interest
No impulse control
Liked to experiment
Manage social anxiety
Social drug use
Only time to feel part of something Social motivations
Enabled socialisation
Peer pressure
Used as pain relief
Self-medicating
Help sleep Manage physical symptoms or health Motivations for using
Help inflammation
Help digestion
To feel alert
Manage autism symptoms
Reduce sensory overload Manage behaviour
Functions better on drugs
Manage OCD Improved quality of life
Reduce nightmares
Manage suicidal thoughts Manage mental health
Manage depression
Manage anxiety
Illegality
Too expensive
Scared of addiction Other reasons
Did not like it or neutral
Need full control of faculties Motivations to stop use
Adverse side-effects
Messed up sleep patterns
Made sick Negative consequences
Caused mental health symptoms
Caused sensory problems

Figure 2: Selected codes, subthemes, and themes of substance use among autistic and non-autistic individuals
The blue boxes represent the themes. The green boxes represent the subthemes for the theme called motivations for using. The red boxes represent the subthemes
for the theme called motivations to stop use. The white boxes provide selected codes for each theme or subtheme; a list of all codes can be found in the appendix
(p 1). OCD=obsessive-compulsive disorder.

have been linked to high rates of mental health conditions Understanding the complex relationships among
and increased risk of suicidality.4,5,8,9 Future research substance use, physical health, and mental health is
should investigate the role of com­ pensatory and essential, as several studies now indicate that autistic
camouflaging strategies in motivating substance use of individuals are at increased risk of a wide variety of
autistic individuals more specifically. chronic physical and mental health conditions,12–14,16,20,22
Autistic individuals reported using substances as a and substance use can have deleterious effects on physical
form of self-medication for both mental and physical and mental health.21 A previous study from our research
health symptoms, although only mental health symptoms group at the Autism Research Centre, University of
showed a significant increase in reporting among autistic Cambridge, UK, has shown that alcohol use and smoking
compared with non-autistic individuals. Many individuals do not fully explain differences in prevalence of physical
did not view this self-medication as negative, instead health conditions between autistic and non-autistic
indicating that use of marijuana (or more rarely other individuals;22 yet, it is possible that substance use operates
substances) provided them with a higher quality of life, as in a positive feedback loop by worsening phy­sical and
proposed previously by the self-medication hypothesis.18,31 mental health conditions.

680 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

Panel: Quotations supporting each theme and subtheme

Usage information last up to 6 hour per day and the pain at least 4 hours. i lose
“MDMA and derivates, around three years, occasionally. all quality of life without amphetamine. i used to get it on
LSD, around three years, occasionally THC (marihuana), prescription The Cannabis i have smoked for 3/4 years and
three years, quite regularly, though not daily. It was my first only continue to do so because my Hospital told me i was
drug and a harmful one for me.” – Non-autistic male individual, curing my diseases” – Autistic female individual,
aged 40–49 years aged 40–49 years
“in the past (up til 10 yrs ago): alcohol, daily, up until 9 yrs ago “Cannabis - I still use, I consume one hash truffle per evening to
LSD, occasionally for a decade ether, a few weekends a year keep my pain levels under control, keep my mood level and help
ketamine, occasionally heroin, daily for a couple years (ending me sleep.” – Autistic non-binary individual, aged 40–49 years
10yrs ago) cocaine, few times a week, up until 10yrs ago lots of “Cannabis: to try and treat seizures, approx 4 year, age 17–21…”
pils, tons of pills...amphetamines, methadone, etc. – Non-autistic male individual, aged 30–39 years
(misc. pharms) recent: DXM daily up til last year. past and
current: cannabis, daily for the past 25+yrs sugar, lots until “Cannabis - Occaisional recreational aged 18–25. For past
recently. TONS. stuff i cant remember, of course.” – Autistic 18 months - more regularly to control pain & muscle spasm”–
male individual, aged 40–49 years Autistic female individual, aged 50–59 years

Non-negative experience Motivations for using: managing behavior

“Smoked cannabis everyday for just over a decade between the “Cannabis - ongoing for mental health, sensory overload & pain
ages of 19 to 33. Took small amounts of LSD (1 tab) every relief” – Autistic female individual, aged 30–39 years
month or so for around two years, when I was 22–24. “…I used amphetamines approx twenty times over same time
Occasionally took Ecstasy (maybe once every three months?) period. I found that this drug use alleviated my autistic
for around 5 years. Never experienced any ill effects.” – symptoms enabled me to socialise/go out and reduced
Non-autistic male individual, aged 40–49 years agoraphobia…” – Autistic female individual, aged 40–49 years
“I am 33. I first took ecstasy in Jan 1999. I took it, because I was “…Sativa seems to alter my behavior in a way I become more
suicidal and I thought it would kill me. [...] it didn’t kill me and I sociable, stim less and am less prone to meltdowns.” – Autistic
experienced happiness for the first time in my life…I consider individual, aged 20–29 years
myself to have been a functioning recreational drug addict “marijuana, speed, MDMA, cocaine, mephedrone - used over a
between the ages of 14–28. I saw it as a point of pride that I few years during adolescence/ early 20s mainly and mostly
partied harder than everyone else and still got up for school/ amphetamines used to feel alert and more ‘normal’ rather than
uni/work. I have never missed anything or messed up anything to get high…” – Autistic non-binary individual, aged 30–39 years
in life through taking drugs, I am too intelligent and have too
much self control to allow that to happen. The only reason I “…Out of all the drugs my favourite was ecstasy it got me doing
have friends is because I used to take drugs. I have no interest in things, thinking clearly and I could chat to people. I also
other people and don’t trust anyone, but ecstasy changes that. appreciated the effects mushrooms had on my synesthesia.” –
Going to Gatecrasher and being part of trance & hardhouse club Autistic transgender male individual, aged 30–39 years
culture was the best time of my life. I am glad that I got to have Atypical response
that experience as it was the only time I have ever been able to “…I no longer try any recreational drugs since I seem to have
feel part of something…” – Autistic non-binary individual, abnormal responses to them.” –Autistic female individual,
aged 30–39 years aged 30–39 years
Motivations for using: experimental interest “…No one suspects me of being “high”, I don’t get the
“marijuana, about 1 year since it became legal in my state” – munchies, or talk endlessly. Instead of watching every muscle
Autistic non-binary individual, aged 30–39 years move in people’s faces and frantically trying to process whether
“Good few timemes amphetamines or ecstasy .. twice lsd/.. I am in trouble or not, I can just work and be okay. I tried cocaine
smoked daily joints good few years I didn’t like the loss of short in the 1970s but it did nothing and I didn’t like being around
term memory, I stopped... I liked to experiment in 20s but I love the people who used it…” – Autistic female individual,
reality and to be myself without any use of substances.” – aged 60–69 years
Non-autistic female individual, aged 30–39 years Motivations for using: managing mental health
Motivations for using: managing physical symptoms or “…Weed/hash - Currently. Helping better to deal with anxiety/
health depression than all the medication I’ve been diagnose so far.
“…The amphetamine allows me to eat food as otherwise I I don’t get high, but gives me a sense of calm and control of
cannot eat anything without extreme pain, discomfort would myself…” – Non-autistic male individual, aged 30–39 years

(Continues on next page)

www.thelancet.com/psychiatry Vol 8 August 2021 681

 Articles

(Panel continued from previous page)

“…I’ve recently started to use small amounts of pure weed again, “…Used infrequently (couple of times per year), and only at
been using and significantly lower the dosage of the other parties or clubs since then (ongoing). Often in combination
antidepressant I used. It has drastically improved my life, with each other or alcohol…” – Autistic male individual,
as both antidepressants were not without side effects. (Weight aged 40–49 years
gain, extreme sweating, significantly lowered libido & almost “I haven’t used drugs for years. I used to use MDMA/ecstasy pills
unable to orgasm)”– Autistic male individual, aged 30–39 years about once or twice a month - I found it really helped me
“Cannabis it helped with the panic attacks and nightmares, i also socially. I used speed occasionally - this turned me into an
stopped smoking everything on 28 dec 2017 not had anything outgoing extrovert that could talk to anyone (the polar
since then” – Autistic male individual, aged 40–49 years opposite of what I’m usually like). I used cocaine and cannabis
“…I’ve been on mild opiates for various chronic pain conditions occasionally (I didn’t enjoy smoking cannabis and it didn’t have
for several years, but find they help my autistic depression and a pleasant effect on me, I only did it because of social pressures).
rage issues, sleep and PTSD as well. It’s an open question as to I tried ketamine and it made me very ill. I hardly ever drink
whether that could be called “recreational,” however I consider alcohol now because my body can’t tolerate it these days. I used
it legitimately medical, but it’s worth mentioning. Especially as to drink a lot when I was younger and it helped me socially at
I think opiates are far too demonised and under-explored, lot but it also used to make me feel very depressed as it was
considering how incredibly helpful they are for autism wearing off...” – Autistic female individual, aged 30–39 years
specifically--autism is like being born with PTSD because you’re Vulnerability: substance misuse
triggered by everything and constantly anxious, but opiates “Marihuana, several grams a Day for 2 maybe 3 years. Became
smooth that out and make your brain feel “normal”. I’ve never addicted. Had help from addiction care, both outpatient and
got similar help from any antidepressants/neuroleptics, and the clinical” – Autistic female individual, aged 40–49 years
side effects of the latter have been far more devastating for my
“I got drunk every day for 20 years. It helped me cope. I stopped
health than opiates ever have been, causing permanent
drinking in 2011 and couldn’t cope. I was sent to psychiatrist
physical damage. Opiates, on the other hand, have improved
and told I was autistic” – Autistic male individual,
my mental and physical health dramatically (leading me to
aged 50–59 years
suspect that whatever it was I was born with is to do with some
kind of dysregulation of the opiate receptors).” “I first took ecstasy in Jan 1999. I took it, because I was suicidal
Autistic female individual, aged 30–39 years and I thought it would kill me. Unfortunately, it didn’t kill me
and I experienced happiness for the first time in my life…
Motivations to stop use: negative consequences I consider myself to have been a functioning recreational drug
“…i started smoking when i was 18, and stopped smoking addict between the ages of 14–28. I saw it as a point of pride
regularly when i was 24. i stopped because it just stopped being that I partied harder than everyone else and still got up for
fun. i found that it began to make me quite depressed...” – school/uni/work…There were many times when I tried to stop
Autistic male individual, aged 40–49 years taking drugs, but I always ended up back in the same place
“Cannabis for 3 years. Several daysper week for 2 years. taking them. The only reason I stopped was getting diagnosed
I stopped because it gave paranoide symptomes including with Asperger’s and finally getting psychological help when I
delusions.” – Non-autistic female individual, aged 30–39 years was 28. If I had easy and safe access to opiates I would start
“Tried cannabis once when younger. Wasn’t for me. Gave me taking them. I no longer have interest in drugs which speed
severe anxiety.” – Autistic female individual, aged 30–39 years things up, I would like something which knocks me out and
makes me forget about how much I hate my life.”– Autistic
“Lsd a few times in high school and pot. Lsd made me non-binary individual, aged 30–39 years
absolutely pain free but made me yak, all pot gives me severe
migraines. No drugs since 1992.” – Autistic non-binary Motivations to stop use: other reasons
individual, aged 40–49 years “Cannabis, intermittently. Most years not at all. But if I could
get it on nhs would help me with my sleep pattern” – Autistic
Motivations for using: social motivations male individual, aged 20–29 years
“marijuana - probably used on and off for 3 years once or twice
a week in my late teens/ early 20’s (not enjoyable made me “…But I don’t do it much anymore as money is a problem so I
throw up) - did it due to peer pressure. Had one 24 hour just put up with feeling like crap.” – Autistic male individual,
psychosis when I ate a large amount instead of smoking it. age unknown
amphetamines - once (felt very unpleasant agitation)…” – “…opium - about 3 times (fantastic but was too scared of
Autistic female individual, aged 60–69 years addiction to continue, plus it wasn’t something that easy to get
hold of)” – Autistic female individual, aged 60–69 years
(Continues on next page)

682 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

(Panel continued from previous page)

Vulnerability major, I made serious attempts in the past. I seriously regret the
“lsd, mdma, mushrooms, ketamine found them to be integral pain I caused to my children. At present, and for a number of
in processing previous long term abuse and preparing years now, I have been smoking it daily and relay on it as if it
(unintentionally) for therapy, and giving me a reason to live in a were a trusted friend. It works at keeping me going, I say,
bad time in my life” – Autistic female individual, ‘to stay alive’. Since 1978 I have had psychiatric treatment with
aged 30–39 years all sorts of meds, CAT, CBT and DBT therapy , even 3 courses of
“It’s part of my shameful past, but desomorphine, ECT and different diagnoses that didn’t “fit” and all helped
amphetamine(s)[uncertain which] and I’ve sniffed glue... Over a marginally. I am still on medication but totally resigned that
period of about 2 years, between the ages of 8 and 10, depression is here to stay. After some 12 years of asking to be
on several occasions. I couldn’t say how often exactly. It was given an AS assessment only to be met with rejection and
sometimes forced upon me. As of then,never again. I never humiliation, I was even told I was choosing an easy way out to
wanted it, but I can’t deny that I did do it...” – Autistic male avoid the responsibility of changing my pattern of life, in 2015 I
individual, aged 16–19 years was finally diagnosed as an Asperger’s. It “fits” and has helped
me to understand most of my life and difficulties....
“…I tried speed once by mistake on holiday in Corfu when a Club unfortunately there is hardly any support out there for adults,
Rep offered me a vitamin unfortunately I believed him (age 20). very few understand, even medics, or make allowances for
I tried Cocaine once (age 33)…” – Autistic female individual, people of my age. I perceive this as saying, “You got here so far,
age unknown get on with it then. You must know what to do by now.’ Hence
“…Soon cannabis became my best screen to hide behind. I have cannabis as a trusted comforter at 67.” – Autistic female
lived here ever since. I married an Englishman in 1971, individual, aged 60–69 years
then children and teaching happily put a stop to my bohemian
Grammar or typographical errors were not corrected; specific names of prescribed
life. In the 90s I went back to cannabis, it felt calming and medications have been omitted to protect anonymity of participants.
scattered my suicidal thoughts. My depression can become

This study bolsters previous findings by supporting not be receiving appropriate management of behavioural,
greater likelihood of self-reported vulnerability associated physical, or mental health symptoms from medical
with substance use among autistic compared with non- providers. Third, marijuana and other substances
autistic individuals.32,33 Five autistic participants in this currently used for recreational drug use should be
study specifically noted that their autism diagnosis was investigated as possible medical interventions for
crucial to discontinuing their use or misuse of substances, managing physical and mental health symptoms fre­
emphasising the importance of timely diagnosis of quently comorbid to autism. Fourth, unwanted symp­
autism. In addition, we have identified new areas of risk toms of autism, mental health, and physical health
among autistic individuals, including forced or accidental conditions (identified by patients themselves) might
use of drugs and childhood use of drugs. This study serve as key targets for intervention for reducing
provides preliminary evidence that substance use substance use. Fifth, approaching sensitive topics
or misuse might have a complex association with vul­ (eg, substance use) is essential to ensure appropriate
nerability, with substances being used both to cope with safeguarding, particularly in light of evidence that
symptoms (eg, to deal with trauma, suicidality) and to substance use might be associated with risk of
serve as a vehicle to exacerbate other forms of vulnerability vulnerability among autistic individuals. As differences
(eg, eating disorders). with social communication are a core feature of autism,
Previous qualitative studies have suggested that autistic autistic individuals have previously endorsed that taking
adults might be motivated to use substances for the extra time within appointments and honouring alter­
reasons such as compensatory or camouflaging strat­ native forms of communication (such as written or
egies, mental health, physical health, and adverse life online communication) might improve patient–provider
events or vulnerability.18,23 However, our findings are, to communication;34 however, the efficacy of these strategies
the best of our knowledge, the first to clarify that autistic has not been tested directly.
adults were far more likely than non-autistic adults to Although the study includes a large sample of
report substance use for these reasons. These findings qualitative responses, several limitations should be
have clinical implications. First, adverse life events, noted. First, the study might be subject to the so-called
autism symptoms causing difficulty, mental health winner’s curse, meaning that it might include artificially
symptoms, and physical health symptoms might all inflated point estimates for group differences.
serve as possible risk factors for substance use among Second, our study is subject to sampling and recruitment
autistic individuals. Second, autistic individuals might biases, as advertisements were circulated via social

www.thelancet.com/psychiatry Vol 8 August 2021 683

 Articles

non-autistic individuals across the adult lifespan

Autistic group, Non-autistic Odds ratio (95% CI) p values
n (%) group, n (%) regarding their substance use. Finally, the effect sizes
and frequency data presented in relation to our
Frequency of reporting themes (429 autistic individuals, 490 non-autistic individuals)*
qualitative analyses should be interpreted not as
Atypical response 22 (5·13%) 4 (0·82%) 6·56 (2·20–26·38) 0·00027
providing information about the actual prevalence of
Motivations for using 91 (21·21%) 47 (9·59%) 2·54 (1·71–3·79) <0·0001
these experiences, but instead about the relative
Motivations to stop use 58 (13·52%) 27 (5·51%) 2·68 (1·63–4·49) 0·00023
likelihood of autistic and non-autistic individuals
Non-negative experience 11 (2·56%) 6 (1·22%) 2·12 (0·71–7·05) 0·21
reporting these themes in an unprompted fashion. A
Usage information 427 (99·53%) 488 (99·59%) 0·88 (0·06–12·12) 1·00
key limitation of this strategy is that autistic and non-
Vulnerability 31 (6·99%) 9 (1·84%) 4·16 (1·90–10·05) 0·00027
autistic individuals might be differentially likely to
Frequency of specific motivations for using (91 autistic individuals, 47 non-autistic individuals)† report information in an unprompted fashion, owing to
Experimental interest 4 (4·40%) 2 (4·26%) 1·03 (0·14–11·85) 1·00 differences in com­munication style. To mitigate this risk
Social motivations 36 (39·56%) 31 (65·96%) 0·34 (0·15–0·75) 0·0095 where possible, reported group differences in specific
Manage physical 39 (42·86%) 12 (25·53%) 2·18 (0·95–5·23) 0·097 subthemes were tested only among individuals who
symptoms or health
discussed the overarching theme (eg, we compared
Manage mental health 32 (35·16%) 7 (14·89%) 3·08 (1·18–9·08) 0·032
symptoms
frequency of reported drug use to manage behaviour
Manage behaviour 26 (28·57%) 2 (4·26%) 8·89 (2·05–81·12) 0·0017
only among participants who discussed any specific
Frequency of specific motivations to stop use (58 autistic individuals, 27 non-autistic individuals)‡
motivation for using drugs).
The themes and subthemes identified by autistic
Negative consequences 25 (43·10%) 12 (44·44%) 0·95 (0·34–2·65) 1·00
individuals have clear clinical relevance, particularly
Other reasons 43 (74·14%) 21 (77·78%) 0·82 (0·23–2·66) 1·00
with regards to the evidence of vulnerability and the
*These analyses provide results from Fisher’s exact tests in which the total sample includes all participants who provided evidence about unmet needs for managing behavioural,
any qualitative data. †These analyses provide results from Fisher’s exact tests in which the total sample includes only
mental health, and physical health symptoms. To the
participants who noted a specific motivation for using substances. ‡These analyses provide results from Fisher’s exact
tests in which the total sample includes only participants who noted a specific motivation to stop using substances. best of our knowledge, this study provides the largest
sample of qualitative data of its kind, and it newly
Table 3: Relatively increased likelihood of reporting specific themes and subthemes related to substance identified and supports several themes and subthemes
use by autistic adults compared with non-autistic adults
related to the experiences of substance use of autistic
individuals without intellectual disability across the
media, autism charities and support groups, and adolescent and adult lifespan (aged 16–90 years). On the
two networks of autistic individuals. As such, the non- basis of the results of the present study, clinicians
autistic sample might be biased towards individuals with should be aware of possible vulnerability related to
high autistic traits, an interest in autism, or undiagnosed substance use (including being forced or tricked to use
autism; therefore, our results might underestimate true substances, childhood use of substances, suicidality,
group differences between autistic and non-autistic trauma, and addiction or dependence) and should work
individuals. Third, our sample primarily includes cooperatively with patients to provide effective means of
autistic individuals who are White, who do not have managing autistic patients’ behavioural, mental health,
comorbid intellectual disability, and who have completed and physical health symptoms.
high school or higher education; therefore, the findings Contributors
are probably not representative of the experiences of EW did the literature search, data collection, and analysis. EW also
substance use of all autistic individuals. Fourth, we did contributed to the study design, inductive coding, original construction
of the qualitative schema, data interpretation, writing, and editing of the
not measure levels of autistic traits among formally original manuscript. SBC contributed to the study design,
diagnosed autistic individuals; although individuals data collection, revision of the qualitative schema, and editing of the
would need to meet a threshold of autistic traits to manuscript. CA contributed to the study design, data collection, revision
receive a formal diagnosis of autism, we cannot be of the qualitative schema, and editing of the manuscript. EW, SBC,
and CA have all accessed and verified all the data in the study.
certain that our sex-specific groups have similar levels of All authors had final responsibility for the decision to submit for
autistic traits or symptoms, which is a clear limitation of publication.
our analyses testing sex differences in substance use. Declaration of interests
Fifth, our quantitative analyses are based on binary We declare no competing interests.
measures of substance use (some of which have been Data sharing
simplified from categorical reports), to provide the most We can provide group-level data but not the underlying material itself,
clinically relevant information, which will have resulted as our participants did not consent to having their data shared publicly.
in a loss of some power and information. We have also Underlying anonymised data will be stored until 5 years after the study
ends and will only be made available to potential collaborators with
provided the unadjusted distributions of each original ethics approval, after they submit a research proposal to the Autism
categorical variable (appendix pp 9–11). Sixth, our Research Centre, University of Cambridge, UK.
qualitative analysis is based on responses to a single Acknowledgments
open-ended question, which might not fully reflect all Funding for this project was generously provided by the Autism
the attitudes, motivations, and experiences of autistic or Research Trust (grant RG72423), the Rosetrees Trust (grant G102199),

684 www.thelancet.com/psychiatry Vol 8 August 2021

 Articles

and the Cambridge and Peterborough NHS Foundation Trust 14 Croen LA, Zerbo O, Qian Y, et al. The health status of adults on the
(grant G102307). EW is supported by funding from the Corbin autism spectrum. Autism 2015; 19: 814–23.
Charitable Trust. SBC received funding from the Wellcome Trust 15 McLeod JD, Hawbaker A, Meanwell E. The health of college
(214322\Z\18\Z). SBC received funding from Innovative Medicines students on the autism spectrum as compared to their neurotypical
Initiative 2 Joint Undertaking under grant agreement number 777394. peers. Autism 2020; 25: 719–30.
The Joint Undertaking receives support from the European Union’s 16 Vohra R, Madhavan S, Sambamoorthi U. Comorbidity prevalence,
Horizon 2020 research and innovation programme as well as the healthcare utilization, and expenditures of Medicaid enrolled adults
European Federation of Pharmaceutical Industries and Associations, with autism spectrum disorders. Autism 2017; 21: 995–1009.
AUTISM SPEAKS, Autistica, and Simons Foundation Autism Research 17 Butwicka A, Långström N, Larsson H, et al. Increased risk for
Initiative. SBC and CA received funding from the Autism Research substance use-related problems in autism spectrum disorders:
Trust, Autistica, the Medical Research Council, and the UK National a population-based cohort study. J Autism Dev Disord 2017;
47: 80–89.
Institute for Health Research (NIHR) Cambridge Biomedical Research
Centre. The research was supported by the NIHR Collaboration for 18 Ressel M, Thompson B, Poulin M-H, et al. Systematic review of risk
and protective factors associated with substance use and abuse in
Leadership in Applied Health Research and Care East of England at
individuals with autism spectrum disorders. Autism 2020;
Cambridgeshire and Peterborough NHS Foundation Trust. The views 24: 899–918.
expressed are those of the authors and not necessarily those of the NHS,
19 Solberg BS, Zayats T, Posserud M-B, et al. Patterns of psychiatric
NIHR, or Department of Health and Social Care. We are grateful to comorbidity and genetic correlations provide new insights into
Paula Smith and Rosemary Holt for assistance with advertisement and differences between attention-deficit/hyperactivity disorder and
Varun Warrier for helpful input in developing our analysis plan. autism spectrum disorder. Biol Psychiatry 2019; 86: 587–98.
In particular, we wish to thank Simon R White for his statistical advice 20 Weiss JA, Isaacs B, Diepstra H, et al. Health concerns and health
and support. Thanks also to all our participants, as well as the service utilization in a population cohort of young adults with
Cambridge Autism Research Database, Autistica’s Discover Network, autism spectrum disorder. J Autism Dev Disord 2018; 48: 36–44.
and various autism support groups and charities for assisting our 21 Schulte MT, Hser Y-I. Substance use and associated health
recruitment. conditions throughout the lifespan. Public Health Rev 2014; 35.
References 22 Weir E, Allison C, Warrier V, Baron-Cohen S. Increased prevalence
1 American Psychiatric Association. Diagnostic and statistical manual of non-communicable physical health conditions among autistic
of mental disorders, 5th edn. Washington DC: American Psychiatric adults. Autism 2020; 25: 681–94.
Association, 2013. 23 Arnevik EA, Helverschou SB. Autism spectrum disorder and
2 Maenner MJ, Shaw KA, Baio J, et al. Prevalence of autism spectrum co-occurring substance use disorder—a systematic review.
disorder among children aged 8 years—autism and developmental Subst Abuse 2016; 10: 69–75.
disabilities monitoring network, 11 sites, United States, 2016. 24 Center for Behavioral Health Statistics and Quality. 2016 National
Surveillance Summaries 2018; 67: 1–23. Survey on drug use and health: detailed tables. https://www.
3 Loomes R, Hull L, Mandy WPL. What Is the male-to-female ratio in samhsa.gov/data/sites/default/files/NSDUH-DetTabs-2016/
autism spectrum disorder? A systematic review and meta-analysis. NSDUH-DetTabs-2016.pdf (accessed Feb 11, 2021).
J Am Acad Child Adolesc Psychiatry 2017; 56: 466–74. 25 Bengtsson M. How to plan and perform a qualitative study using
4 Bargiela S, Steward R, Mandy W. The experiences of late-diagnosed content analysis. NursingPlus Open 2016; 2: 8–14.
women with autism spectrum conditions: an investigation of the 26 Elo S, Kyngäs H. The qualitative content analysis process.
female autism phenotype. J Autism Dev Disord 2016; 46: 3281–94. J Adv Nurs 2008; 62: 107–15.
5 Hull L, Petrides KV, Allison C, et al. “Putting on my best normal”: 27 Erlingsson C, Brysiewicz P. A hands-on guide to doing content
social camouflaging in adults with autism spectrum conditions. analysis. Afr J Emerg Med 2017; 7: 93–99.
J Autism Dev Disord 2017; 47: 2519–34. 28 Azur MJ, Stuart EA, Frangakis C, Leaf PJ. Multiple imputation by
6 Hull L, Petrides KV, Mandy W. The female autism phenotype and chained equations: what is it and how does it work?
camouflaging: a narrative review. Rev J Autism Dev Disord 2020; Int J Methods Psychiatr Res 2011; 20: 40–49.
7: 306–17. 29 Benjamini Y, Hochberg Y. Controlling the false discovery rate:
7 Lai M-C, Szatmari P. Sex and gender impacts on the behavioural a practical and powerful approach to multiple testing. J R Stat Soc B
presentation and recognition of autism. Curr Opin Psychiatry 2020; 1995; 57: 289–300.
33: 117–23. 30 Hirvikoski T, Mittendorfer-Rutz E, Boman M, Larsson H,
8 Livingston LA, Shah P, Happé F. Compensatory strategies below Lichtenstein P, Bölte S. Premature mortality in autism spectrum
the behavioural surface in autism: a qualitative study. disorder. Br J Psychiatry 2016; 208: 232–38.
Lancet Psychiatry 2019; 6: 766–77. 31 Khantzian EJ. The self-medication hypothesis of substance use
9 Cassidy S, Bradley L, Shaw R, Baron-Cohen S. Risk markers for disorders: a reconsideration and recent applications.
suicidality in autistic adults. Mol Autism 2018; 9: 42. Harv Rev Psychiatry 1997; 4: 231–44.
10 Kenny L, Hattersley C, Molins B, Buckley C, Povey C, Pellicano E. 32 Griffiths S, Allison C, Kenny R, Holt R, Smith P, Baron-Cohen S.
Which terms should be used to describe autism? Perspectives from The vulnerability experiences quotient (VEQ): a study of
the UK autism community. Autism 2016; 20: 442–62. vulnerability, mental health and life satisfaction in autistic adults.
11 Bejerot S, Nylander L. Low prevalence of smoking in patients with Autism Research 2019; 12: 1516–28.
autism spectrum disorders. Psychiatry Res 2003; 119: 177–82. 33 Rigles B. The relationship between adverse childhood events,
12 Davignon MN, Qian Y, Massolo M, Croen LA. Psychiatric and resiliency and health among children with autism.
medical conditions in transition-aged individuals with ASD. J Autism Dev Disord 2017; 47: 187–202.
Pediatrics 2018; 141 (suppl 4): S335–45. 34 Nicolaidis C, Raymaker DM, Ashkenazy E, et al. “Respect the way I
13 Fortuna RJ, Robinson L, Smith TH, et al. Health conditions and need to communicate with you”: healthcare experiences of adults
functional status in adults with autism: a cross-sectional evaluation. on the autism spectrum. Autism 2015; 19: 824–31.
J Gen Intern Med 2016; 31: 77–84.

www.thelancet.com/psychiatry Vol 8 August 2021 685