Persistent Attention-Deficit/Hyperactivity Disorder predicts socially-oriented, but not physical/physiologically-oriented, alcohol problems in early adulthood

Abstract

Background:

Although individuals with histories of childhood attention-deficit/hyperactivity disorder (ADHD) report more alcohol-related problems in adulthood relative to those without, it remains unknown whether there are group differences in certain types of alcohol problems. We tested whether the nature of alcohol problems differed for individuals with and without childhood ADHD, as well as adulthood-persistent ADHD, to facilitate a personalized medicine approach for alcohol problems in this high-risk group.

Methods:

Data were drawn from a prospective, observational study. Children diagnosed with ADHD and demographically-similar individuals without childhood ADHD were followed prospectively through young adulthood (N=453; 87.6% male). ADHD symptom persistence was assessed using self- and parent-reports. Alcohol problems and heavy drinking were assessed repeatedly from 18–30 years old to construct lifetime measures.

Results:

Full sample confirmatory factor analyses identified five alcohol problem “types:” interpersonal problems/risky behaviors, occupational/academic impairment, impaired control/treatment seeking, tolerance/withdrawal, and drinking to blackout. Latent class analyses of items within each “type” yielded best fit for three-class solutions for all sets of items except for blackout drinking, for which two classes emerged. Children with ADHD were more likely than those without to belong to high-risk latent classes for interpersonal problems/risky behaviors, occupational/academic problems, and impaired control (whose high-risk class indexed treatment-seeking behavior). These effects were driven by individuals whose ADHD symptoms persisted into adulthood. Few group differences emerged for tolerance/withdrawal and blackout drinking, with the exception that individuals with only childhood ADHD (no persistence) were more likely to belong to the low-risk groups relative to those with adulthood-persistent ADHD and without ADHD.

Conclusions:

Individuals with ADHD histories whose symptoms persist into adulthood may be more likely to experience socially-oriented alcohol problems and impaired control/treatment seeking relative to individuals without an ADHD history and those with childhood ADHD only. Tailored alcohol prevention and treatment programs may be beneficial for this high-risk population.

Introduction

Attention-Deficit/Hyperactivity Disorder (ADHD) is a childhood disorder characterized by inattention, impulsivity, and hyperactivity or restlessness (American Psychiatric Association, 2013). Children diagnosed with ADHD report more alcohol-related problems and consequences in adulthood relative to those without childhood ADHD (Pedersen et al., 2016), However, it remains unknown whether those with and without ADHD differ with respect to the types of alcohol problems they are more likely to experience, such as alcohol-related interpersonal, physiologic, or academic/work problems. This is important as better understanding the symptom profiles of alcohol use disorder that are most common for those with ADHD could help inform tailored treatment and prevention strategies (Read et al., 2006). Understanding these associations could also elucidate unique etiological processes underlying alcohol use disorder, in line with the growing emphasis on understanding the heterogeneity of alcohol use disorder (Litten et al., 2015).

Individuals diagnosed with ADHD may show differing types of alcohol problems for several reasons. First, those with ADHD have been shown to become more disinhibited than those without ADHD when under the influence of alcohol (e.g., Weafer et al., 2009). Additionally, relative to those without, individuals with ADHD displayed increased attentional bias to alcohol-related stimuli following alcohol consumption, which in turn predicted later ad-libitum alcohol consumption (Roberts et al., 2012).

Heightened disinhibition and inattention while intoxicated could cause those with ADHD to exhibit greater alcohol-related relational difficulties (Tan et al., 2017), occupational/academic impairments (Yoneda et al., 2019), risk-taking behavior (Cyders, 2013), and impaired control over drinking (Leeman et al., 2014). For example, impulsivity (an indicator of disinhibition) and inattention have been shown to interfere with successful communication and conflict management (Tan et al., 2017), which could heighten drinking-related relational difficulties for those with ADHD. As impulsive decision-making while drinking could cause an individual to drink more excessively, stay out longer, or use additional substances, they may be more likely to experience hangover- or intoxication-related poor performance or absenteeism from work or school. Moreover, heightened impulsivity while intoxicated could result in increases in immediately-rewarding, risky behaviors (Cyders, 2013), like risky sex, for those with ADHD. Given that impulsivity makes it more difficult to control alcohol consumption (i.e., impaired control over drinking; Leeman et al., 2014), those with ADHD may also be more likely to exhibit features of impaired control, such as drinking more than intended or attempts to cut back consumption (e.g., seeking treatment).

Second, irrespective of alcohol consumption, individuals with ADHD have been shown to experience heightened interpersonal problems, impairments in work and school, and risk-taking behavior in childhood and adulthood (e.g., Barkley, 2002; Canu & Carlson, 2004). Additional problems in these functional domains that are incurred as a result of drinking could more easily “tip the scales” towards endorsing these alcohol problems for those with ADHD, such as losing a job or getting into trouble for risky behavior.

However, individuals with ADHD may not be more likely to experience other kinds of alcohol problems relative to those without ADHD, such as developing alcohol tolerance and withdrawal or blackout-drinking behaviors. Consistent with this possibility, two prior studies demonstrated that young adults with childhood ADHD reported similar or lesser levels of heavy drinking compared to those without childhood ADHD (Molina et al., 2018; Pedersen et al., 2016). Those with ADHD could be less or similarly prone to tolerance, withdrawal and blackout drinking relative to those without, as heavy drinking is strongly linked to these physiological alcohol consequences (Dawson, 2000; Jennison and Johnson, 1994).

In line with these hypotheses, McDowell et al. (2019) found that externalizing behaviors (i.e., symptoms of conduct disorder, antisocial personality disorder, and impulsivity) were more strongly related to alcohol-related social problems and role interference in adults. The authors attributed this pattern of results to shared irresponsibility characteristics among social problems, role interference, and externalizing problems and to the increased vulnerability of those with high levels of externalizing problems to the disinhibiting effects of alcohol. Moreover, they also found that externalizing problems were most weakly associated with alcohol tolerance, likely because this neuroadaptation is more directly linked to alcohol consumption than to personality characteristics. Notably, this study did not include ADHD in their conceptualization of externalizing behaviors. Doing so is particularly important due to the high prevalence and unique sequelae and symptomatology of this disorder.

Adulthood ADHD Symptom Persistence

Evidence suggests that the persistence of ADHD symptoms into adulthood accounts for worse functioning compared to those with only childhood ADHD or no ADHD histories, including more severe substance use outcomes (Hechtman et al., 2016). Thus, those with ADHD persistence may exhibit similar patterns of relations with the different types of alcohol problems predicted to be related to childhood ADHD, but with a higher level of severity. Alternatively, it is possible that predicted relationships with specific alcohol problem types could be solely driven by those with adulthood-persistent ADHD. In this study, we compared those with adulthood-persistent ADHD, childhood ADHD only, and those without ADHD on their likelihoods of these different types of alcohol problems.

Current Study

In this study, we utilized prospective data from the Pittsburgh ADHD Longitudinal Study (PALS) to assess whether children diagnosed with ADHD varied in the types of adulthood alcohol problems they were likely to experience relative to those without childhood ADHD. Additionally, we examined the role of ADHD symptom persistence in these associations. We first derived empirically validated alcohol problem “types” by testing a previously established, multi-dimensional model of alcohol problems (Read et al., 2006). We hypothesized that a six-factor structure capturing alcohol-related interpersonal consequences, occupational/academic impairment, risky behaviors, impaired control/treatment seeking, tolerance/withdrawal, and blackout drinking would best fit our data.

Next, in the full sample we conducted separate latent class analyses of the alcohol problems within each “type” to identify heterogeneity in severity that might discriminate adults with and without ADHD histories and symptom persistence. We then examined whether childhood ADHD or adulthood-persistent ADHD predicted membership in the latent classes found for each distinct alcohol problem type. We hypothesized that adults with ADHD histories would be more likely than adults without ADHD histories to belong to the highest risk latent classes for alcohol-related interpersonal problems, occupational/academic impairment, risk-taking behavior, and impaired control/treatment seeking in adulthood, but would be equally or less likely to belong to the latent classes for tolerance/withdrawal and blackout drinking. Moreover, we expected that those with only childhood ADHD would show levels of risk intermediate between those with ADHD persistence and those without ADHD.

Materials and Methods

Participants

PALS ADHD study participants were diagnosed with ADHD (DSM-III-R or DSM-IV criteria) as children (M_age=9.40 years, SD_age=2.27, range=5.0–16.92, 90% were ages 5–12) and participated in a summer treatment program (Pelham et al., 2005) between 1987–1996. ADHD diagnoses were made using standardized parent and teacher ADHD symptom rating scales (Pelham et al., 1992) and a standardized semi-structured diagnostic interview administered to parents by a PhD-level clinician. This semi-structured interview also assessed for childhood conduct disorder. Exclusion criteria included full-scale IQ<80, history of seizures or other neurological problems, and/or history of pervasive developmental disorder, schizophrenia, or other psychotic or similarly impairing severe disorders.

Of those eligible for longitudinal study (n=516), 70.5% participated (n=364; M_age=17.75, SD_age=3.39). Enrollment occurred an average of 8.35 (SD=2.79) years after childhood diagnosis of ADHD and summer treatment program participation. Enrolled PALS participants differed from non-enrolled individuals on 1 of 14 childhood demographic and diagnostic variables: conduct disorder symptoms were lower (t=3.09, p<.01) for participants (M=2.13; SD=1.93) than nonparticipants (M=2.74; SD=2.19; Cohen’s d=.30).

Demographically-similar PALS participants without ADHD were recruited at the time of initial follow-up for the ADHD sample and from the greater Pittsburgh area from several sources (e.g., pediatric practices, university hospital newsletter, local schools; Molina et al., 2007). In addition to the exclusion criteria above, individuals who met DSM-III-R criteria for ADHD (presence of 8 or more symptoms), currently or historically, were excluded from the non-ADHD sample. Via telephone screening, parents (and participants age 18+) provided information on sociodemographics, history of ADHD diagnosis/treatment, conduct disorder, and other behavior problems, presence of exclusion criteria, and ADHD symptoms. Initial enrollment occurred from 1999–2003. Annual interviews, begun in 1999, occurred until 2008; assessments have since been age-based (annually to 23, and then at approximately ages 25, 27, and 29). Informed consent was obtained from all individual participants. The University of Pittsburgh Institutional Review Board approved this study.

Current Participants

Participants were included in this study if they answered the lifetime alcohol problem questions (see below) at ages 28, 29, and/or 30 to ensure that we captured these experiences through the late 20s (n=453). Relative to those excluded (n=133), those included did not differ on any predictors or covariates (ADHD diagnosis: χ²(1)=1.33, p=0.25; race or ethnicity: χ²(1)=1.60, p=0.21; childhood conduct disorder: χ²(1)=0.02, p=0.88) except they were more likely to be female (χ²(1)=4.26, p=0.04). Those included also did not differ from those excluded on average lifetime heavy drinking (t(584)=0.72, p=0.47) or any of the lifetime alcohol problem items (χ²(1)=0.00–3.5, p=0.06–0.99) except they were more likely to have driven drunk (χ²(1)=7.12, p=0.01).

Measures

Lifetime alcohol problems.

During the age 18–30 assessments, participants self-reported whether they had experienced 28 alcohol problems in their lifetime using the Young Adult Alcohol Problems Screening Test (Hurlbut and Sher, 1992) that corresponded with the factors and items used in Read et al. (2006). To capture lifetime alcohol problems, each participant was coded as having experienced an alcohol problem (yes = 1) if they endorsed the alcohol problem when responding from the ages of 18–30, and, if not, coded as “no” (0).

Lifetime heavy drinking.

During the age 18–30 year old assessments, participants reported their frequency of past-year heavy drinking using the Substance Use Questionnaire (Molina & Pelham, 2003; “In the past 12 months, how often did you drink 5 or more drinks when you were drinking?”; 0=not at all; 1=1–3 times; 2=4–7 times; 3= 8–11 times; 4=Once/month; 5=2–3 times/month; 6=Once a week; 7=2–3 times a week; 8=4–6 times a week; 9=once/day; 10=Twice/day; 11=several times/day). To reflect heavy drinking across the adulthood age span, justified from our previous work (Wang et al., 2020), we computed an average from all available assessments from ages 18–30 (M=2.71, SD=2.22; Range=0–8.33).

Adulthood-persistent ADHD.

Participants and their parents reported on 18 ADHD symptoms in the past year during selected follow-up visits using the Adult ADHD Rating Scale (0 = never to 3 = very often; Barkley, 2011). Adulthood-persistent ADHD was defined as five or more ADHD symptoms (consistent with DSM-5) endorsed by participants or parents at ages 25, 27, or 29 and a diagnosis of childhood ADHD (18.2% of participants met these symptom count criteria for adulthood-persistent ADHD). Childhood ADHD only was defined as those with childhood ADHD but fewer than five or more ADHD symptoms.

Demographics.

Participants self-reported their gender (87.6% male, 12.4% female) and responded to a question about their race and/or ethnicity (83% White (not Hispanic), 11.3% Black or African American, 0.7% Hispanic or Spanish American [Mexican, Puerto Rican, Cuban, Latin American], 0.2% Asian American, 4.2% Mixed race, 0.7% “Other”). Those identifying as White (non-Hispanic) were coded “1” and all other racial and ethnic groups “0.”

Data Plan

Mplus v.7.2 was used for all analyses. Confirmatory factor analyses (CFA) were conducted using the WLSMV estimator (for dichotomous indicators). Based on Read et al., (2006), we tested a six-factor model comprised of alcohol-related interpersonal problems, occupational/academic impairment, risky behaviors, impaired control/treatment seeking, tolerance/withdrawal, and blackout drinking. We were unable to test two additional factors found by Read et al., (2006) due to an insufficient number of items in these domains (i.e., self-care and self-perception). Conventional selection criteria were used to determine model fit (i.e., CFI, RMSEA, chi-square).

We tested separate latent class analyses (LCAs) of dichotomous alcohol problem items according to the latent factors derived from the CFAs.¹ LCAs were conducted separately by alcohol problem type, rather than together in one LCA, as there is a growing literature highlighting the importance of examining type of alcohol problem in addition to number of problems (McDowell et al., 2019; Read et al., 2006) and one of our aims was to extend this literature to individuals with a history of ADHD.

No participants were missing on the alcohol problems variables. One to five class models were tested. The optimal model was selected using several indicators of model fit: Bayesian Information Criteria (BIC), adjusted BIC, and log likelihoods (Nylund et al., 2007). We also considered group proportion sizes (>5%) and theoretical relevance of identified groups (Muthén & Muthén, 2000). The R3STEP procedure in Mplus was used to examine childhood ADHD as a latent class predictor while controlling for sex, race and/or ethnicity, childhood conduct disorder, and lifetime heavy drinking. For adulthood-persistent ADHD analyses, this same procedure was used in two separate models to test all possible planned contrasts using dummy codes: (1) non-ADHD vs. persistent ADHD; (2) non-ADHD vs. childhood ADHD only; (3) childhood ADHD only vs. persistent ADHD.

Results

CFA.

In the six-factor model, the interpersonal and risk-taking factors were extremely highly correlated (r=0.98, p<0.001), so they were combined to form a single factor. The resultant five-factor model fit the data well (χ²=647.07, p<0.001, CFI=0.99, RMSEA=0.04, SRMR=0.08; r’s between factors=0.75–0.88). Factor loading ranges were 0.74–0.89 for the risky behavior/interpersonal problems factor, 0.89–0.96 for the occupational/academic problems factor, 0.85–0.95 for the impaired control/treatment seeking factor, 0.86–0.98 for the tolerance/withdrawal factor, and 0.89–0.97 for the blackout drinking factor.

LCAs.

See Table 1 for results and Figures 1–5 for the final LCAs. BIC, adjusted BIC, and loglikelihoods improved from the one-to-three class solutions for the interpersonal problems/risky behavior, occupational/academic, impaired control/treatment seeking, and tolerance/withdrawal LCAs, and from the one-to-two class solutions for the blackout drinking LCA. Thereafter, for all models but the blackout drinking LCA, BIC worsened, adjusted BIC worsened or showed only small improvements (<10), and loglikelihoods showed only small improvements (<20). For the blackout drinking LCA, three-to-five class models each yielded a non-positive-definite matrix, suggesting an untenable model. Thus, for subsequent analyses, we selected the three class solutions (hereafter referred to as high, intermediate, and low problems classes) for the interpersonal problems/risk-taking, occupational/academic impairments, impaired control/treatment seeking, and tolerance/withdrawal LCAs, and the two class solution (hereafter referred to as high and low classes) for blackout drinking (all group sizes >5%).

Table 1.

Fit Statistics for 1- to 5-class Solutions for Each Alcohol Problem Type

# Classes	AIC	BIC	Adjusted BIC	Log Likelihood	Entropy	LMR LRT	p	B LRT	p	% Classified to Each Class
										1	2	3	4	5
Interpersonal Problems/Risky Behaviors
1	4891.741	4928.784	4900.221	−2436.87						100%
2	3881.201	3959.403	3899.104	−1921.60	0.858	1013.96	<.001	−2436.87	<.001	48.6%	51.4%
3	3711.841	3831.202	3739.166	−1826.92	0.827	186.31	<.001	−1921.60	<.001	20.0%	43.1%	36.8%
4	3692.167	3852.687	3728.914	−1807.08	0.768	39.035	.14	−1826.92	<.001	10.5%	27.9%	30.7%	31.0%
5	3690.58	3892.259	3736.75	−1796.29	0.761	21.239	.18	−1807.08	.67	10.9%	9.6%	19.7%	29.5%	30.3%
Occupational/Academic Problems
1	2612.841	2633.421	2617.552	−1301.42						100%
2	1908.222	1953.497	1918.587	−943.111	0.904	97.61	<.001	−1301.40	<.001	42.9%	57.1%
3	1844.434	1914.404	1860.452	−905.217	0.858	73.78	<.001	−943.11	<.001	13.7%	54.9%	31.4%
4a	1846.833	1941.499	1868.505	−900.417	0.893	9.346	.038	−905.22	.09	14.1%	0.9%	30.9%	54.1%
5	1856.216	1975.577	1883.541	−899.108	0.918	2.548	.32	−900.42	1.00	9.5%	5.9%	29.4%	54.2%	0.8%
Impaired Control/Treatment Seeking
1	3618.626	3647.437	3625.222	−1802.31						100%
2	2771.572	2833.311	2785.706	−1370.79	0.903	845.77	<.001	−1802.31	<.001	29.3%	70.7%
3	2655.513	2750.179	2677.185	−1304.76	0.876	129.41	<.001	−1370.79	<.001	24.9%	16.6%	58.4%
4	2638.528	2766.12	2667.737	−1288.26	0.893	32.33	.03	−1301.76	<.001	24.8%	14.9%	2.2%	58.1%
5a	2610.148	2770.668	2646.895	−1266.07	0.875	21.44	.009	−1277.01	<.001	10.4%	11.3%	2.2%	35.0%	41.2%
Tolerance/Withdrawal
1	2010.586	2027.05	2014.355	−1001.29						100%
2	1493.944	1530.987	1502.424	−737.972	0.954	5509.97	<.001	−1001.29	<.001	49.4%	50.5%
3	1439.309	1496.931	1452.5	−705.654	0.929	62.59	<.001	−737.97	<.001	11.8%	50.4%	37.8%
4a	1446.999	1525.201	1464.902	−704.5	0.822	2.54	.02	−705.81	.15	34.9%	18.1%	13.2%	33.8%
5a	1453.83	1552.611	1476.443	−702.915	0.937	0.708	.53	−703.281	.38	0.6%	37.2%	16.0%	1.5%	45.0%
Blackout Drinking
1	1840.237	1856.701	1844.006	−916.119						100%
2	1347.408	1384.451	1355.888	−664.704	0.91	486.907	<.001	−916.119	<.001	71.3%	28.7%
3a	1306.352	1363.974	1319.543	−639.176	0.895	42.438	<.001	−661.09	<0.001	22.2%	14.4%	63.5%
4a	1316.217	1394.419	1334.12	−639.109	0.923	0.130	.15	−639.176	1.00	68.9%	19.3%	0.22%	11.5%
5a	1325.304	1424.086	1347.918	−638.652	0.923	0.884	.02	−639.109	0.67	7.5%	5.1%	60.0%	14.6%	12.8%

Note. AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; LMR LRT = Lo-Mendell-Rubin Adjusted Likelihood Ratio Test; B LRT = Bootstrap Likelilhood Ratio Test. Shaded rows indicate the selected solution for each alcohol problem type.

^aAnalyses yielded a non-positive definite error message, suggesting an untenable model.

Figure 1.

Final Three-Class Solution for the Interpersonal Problems/Risky Behaviors Latent Class Analysis.

Figure 5.

Final Two-Class Solution for the Blackout Drinking Latent Class Analysis.

To aid in interpretation of the latent classes, we conducted chi-square tests to determine on which items individuals in separate classes differed (see Table S1, Figures 1–5). Table S1 also includes the percentage endorsement for each item, as well as the number of alcohol problems of each corresponding type, within the high, intermediate, and low classes of each LCA. This provides a sense of how meaningfully different classes were from one another.

For the interpersonal problems/risky behaviors LCA, all items were significantly different among the three groups in accordance with severity levels.

For the occupational/academic impairment LCA, individuals in the high and intermediate classes were equally likely to miss work/school as a result of drinking (nearly all endorsed this) and individuals in the intermediate and low classes were equally likely to have been fired/suspended/expelled (nearly all never endorsed this); the remaining items were significantly different among the three groups in accordance with their severity levels.

For the impaired control/treatment seeking LCA, relative to those in the intermediate class, individuals in the high class were less likely to drink more than intended, equally likely to try controlling or cutting-down drinking, and more likely to continue drinking despite promises not to, to seek any or professional help to control one’s drinking, and to attend AA. The remaining items were significantly different among the three groups in accordance with their severity levels, although the intermediate and low classes could not be compared on seeking professional help as no individuals in these groups endorsed this. It may seem paradoxical that the high class was both less likely than the intermediate class to drink more than intended but more likely to continue drinking despite promises. Though statistically significant, the differences between the high and intermediate classes on these items were quite small, with 95.9% vs. 100% of high vs. intermediate endorsing “larger amounts” and 86.3% vs. 72.9% endorsing “drank despite promises,” respectively. In contrast, for the treatment seeking items on the same LCA, differences were large: 83.56–87.7% vs. 0–36.5% of those in high vs. intermediate endorsed them (see Table S1). This is visually apparent in Figure 3. Thus, the difference between the high vs. intermediate classes lies more in the large differences in treatment seeking items (see Table S1 and Figure 3).

Figure 3.

Final Three-Class Solution for the Impaired Control/Treatment Seeking Latent Class Analysis.

For the tolerance/withdrawal LCA, relative to the intermediate class, those in the high class were less likely to need more alcohol than previously to get the same effect, equally likely to no longer get drunk on the same amount as previously, and more likely to have gotten shakes after cutting down or needed a drink upon awakening. Those in the intermediate and low classes were equally likely to have gotten the shakes than those in the low class (nearly all never endorsed this). The remaining items were significantly different among the three groups in accordance with their severity levels. Note that those in the high class can be interpreted as being relatively more severe in terms of tolerance/withdrawal despite showing a lower likelihood of endorsing “need more alcohol” than those in the intermediate class. Indeed, group differences on this item were small, with 93.4% endorsement from those in the high class and 100% endorsement from those in the intermediate class. In contrast, there were large differences between the high and intermediate classes for the two withdrawal items. 100% of those in the high class vs. 0.63% in the intermediate class and 68.85% of those in the high class vs. 13.75% of those in the intermediate classes endorsed each withdrawal item, respectively (see Table S1 and Figure 4).

Figure 4.

Final Three-Class Solution for the Tolerance/Withdrawal Latent Class Analysis.

For the blackout drinking LCA, those in the high class were significantly more likely to endorse all items relative to those in the low class.

Differences between those with and without childhood ADHD.

See Table 2 for full results from the multinomial logistic regressions. Childhood ADHD predicted a greater likelihood of membership in the high relative to intermediate classes for interpersonal problems/risky behaviors, occupational/academic impairment, and impaired control/treatment seeking. Childhood ADHD did not predict membership in any of the tolerance/withdrawal or blackout drinking classes.

Table 2.

Results of Multinomial Logistic Regression Analyses Predicting Class Membership from Childhood ADHD Diagnosis and Covariates

		High vs. Intermediate				High vs. Low				Intermediate vs. Low
Predictor		B	SE B	p		B	SE B	p		B	SE B	p
Interpersonal Problems/Risky Behaviors
ADHD Diagnosisa		0.81	0.37	.029		0.199	0.47	.684		−0.62	0.36	.085
Sexb		−2.12	1.29	.101		−1.27	1.31	.334		0.85	0.44	.052
Racec		0.69	0.57	.223		−0.64	0.68	.346		−1.33	0.43	.002
Conduct Disorderd		0.27	0.44	.537		0.82	0.49	.096		0.55	0.40	.170
Heavy Drinkinge		0.26	0.10	.007		1.31	0.20	<.001		1.05	0.18	<.001
Occupational/Academic Problems
ADHD Diagnosis		1.20	0.47	.010		0.82	0.44	.062		−0.38	0.33	.254
Sex		−15.43	1.41	<.001		−15.03	1.43	<.001		0.40	0.43	.352
Race		1.09	0.78	.160		0.75	0.75	.316		−0.34	0.39	.374
Conduct Disorder		−0.52	0.51	.309		−0.20	0.44	.654		0.32	0.40	.421
Heavy Drinking		−0.04	0.11	.728		0.63	0.10	<.001		0.66	0.08	<.001
Impaired Control/Treatment Seeking
ADHD Diagnosis		0.95	0.34	.006		0.65	0.35	.060		−0.29	0.31	.352
Sex		−0.47	0.58	.419		−0.41	0.58	.475		0.05	0.43	.900
Race		0.16	0.51	.748		0.02	0.45	.967		−0.14	0.37	.700
Conduct Disorder		0.52	0.47	.263		−0.25	0.38	.505		−0.78	0.43	.072
Heavy Drinking		0.22	0.08	.003		0.63	0.08	<.001		0.41	0.08	<.001
Tolerance/Withdrawal
ADHD Diagnosis		0.51	0.46	.268		−0.07	0.47	.877		−0.58	0.30	.051
Sex		−0.21	0.78	.790		0.56	0.78	.472		0.77	0.41	.061
Race		1.54	1.76	.381		1.24	1.79	.487		−0.10	0.34	.383
Conduct Disorder		0.15	0.55	.781		0.43	0.55	.436		0.28	0.34	.412
Heavy Drinking		0.17	0.12	.169		0.83	0.13	<.001		0.67	0.08	<.001
Blackout Drinking
ADHD Diagnosis		−	−	−		−0.60	0.34	.076		−	−	−
Sex		−	−	−		0.04	0.36	.906		−	−	−
Race		−	−	−		−0.33	0.41	.421		−	−	−
Conduct Disorder		−	−	−		0.16	0.45	.723		−	−	−
Heavy Drinking		−	−	−		1.52	0.48	.001		−	−	−

Note. N = 452.

^aChildhood ADHD=1, non-ADHD=0

^bMale=1, Female=0

^cWhite=1, Other race=0

^dChildhood Conduct Disorder=1, no Conduct Disorder=0

^eLifetime adulthood average heavy drinking frequency. Bolded entries are statistically significant results.

Differences between those with adulthood-persistent ADHD, those with childhood ADHD only, and those without ADHD.

See Table 3 for full results. Relative to those without ADHD, those with adulthood-persistent ADHD showed a greater likelihood of being in the high than intermediate class for interpersonal/risky behaviors, as well as a greater likelihood of being in the high than intermediate and low classes for occupational/academic and impaired control/treatment seeking. Relative to those with childhood ADHD only, those with adulthood-persistent ADHD showed a greater likelihood of being in the: (1) high than low class for occupational/academic problems, (2) in the high than intermediate and low classes for impaired control/treatment seeking and, (3) in the high and intermediate than low classes for tolerance/withdrawal. Finally, those with childhood ADHD only were more likely to be in the low than intermediate class for tolerance/withdrawal relative to those without ADHD.

Table 3.

Results of Multinomial Logistic Regression Analyses Predicting Class Membership from ADHD Persistence Classification and Covariates

		High vs. Intermediate				High vs. Low				Intermediate vs. Low
Predictor		B	SE B	p		B	SE B	p		B	SE B	p
Interpersonal Problems/Risky Behaviors
ADHD Persistent v non-ADHD		1.31	.17	.005		.61	.55	.271		−.70	.49	.155
Child ADHD only v non-ADHD		.48	.42	.255		−.08	.52	.882		−.56	.39	.150
ADHD Persistent v Child ADHD only		.83	.46	.075		.68	.52	.190		−.14	.47	.761
Sexa		−1.95	1.08	.071		−1.11	1.11	.319		.84	.44	.056
Raceb		.82	.55	.137		−.56	.68	.411		−1.38	.44	.001
Conduct Disorderc		.43	.45	.336		.93	.51	.068		.50	.41	.220
Heavy Drinkingd		.25	.10	.012		1.30	.20	<.001		1.05	.18	<.001
Occupational/Academic Problems
ADHD Persistent v non-ADHD		1.76	.56	.002		1.48	.52	.004		−.28	.50	.575
Child ADHD only v non-ADHD		.71	.57	.216		.33	.52	.517		−.37	.34	.281
ADHD Persistent v Child ADHD only		1.05	.59	.074		1.14	.50	.022		.09	.48	.847
Sex		−3.04	3.39	.371		−2.70	3.31	.415		.34	.43	.429
Race		1.39	1.07	.192		.99	1.02	.330		−.40	.39	.304
Conduct Disorder		−.18	.59	.760		.04	.51	.940		.22	.40	.579
Heavy Drinking		−.04	.11	.687		.62	.10	<.001		.67	.08	<.001
Impaired Control/Treatment Seeking
ADHD Persistent v non-ADHD		1.73	.47	<.001		1.42	.41	.001		−.31	.44	.480
Child ADHD only v non-ADHD		.41	.40	.306		.15	.40	.714		−.26	.34	.450
ADHD Persistent v Child ADHD only		1.32	.48	.006		1.27	.39	.001		−.06	.44	.902
Sex		−.49	.58	.406		−.45	.57	.437		.04	.43	.926
Race		.34	.53	.528		.14	.49	.774		−.20	.37	.602
Conduct Disorder		.75	.48	.120		−.02	.40	.954		−.77	.44	.076
Heavy Drinking		.22	.08	.005		.63	.08	<.001		.41	.08	<.001
Tolerance/Withdrawal
ADHD Persistent v non-ADHD		.90	.49	.065		.87	.52	.092		−.03	.39	.938
Child ADHD only v non-ADHD		−.12	.84	.888		−.91	.82	.267		−.80	.33	.014
ADHD Persistent v Child ADHD only		1.02	.71	.151		1.78	.73	.015		.77	.37	.039
Sex		−.03	.88	.971		.71	.85	.407		.74	.42	.077
Race		2.71	6.13	.658		2.43	6.16	.694		−.28	.35	.415
Conduct Disorder		.52	.80	.510		.89	.77	.248		.37	.34	.285
Heavy Drinking		.20	.16	.207		.88	.16	<.001		.68	.08	<.001
Blackout Drinking
ADHD Persistent v Control		−	−	−		−.52	.47	.273		−	−	−
Child ADHD only v Control		−	−	−		−.61	.37	.101		−	−	−
ADHD Persistent v Child ADHD only		−	−	−		.09	.49	.847		−	−	−
Sex		−	−	−		.03	.36	.933		−	−	−
Race		−	−	−		−.35	.41	.391		−	−	−
Conduct Disorder		−	−	−		.16	.47	.736		−	−	−
Heavy Drinking		−	−	−		1.52	.49	.002		−	−	−

Note. N =450.

^aMale=1, Female=0

^bWhite=1, Other race=0

^cChildhood Conduct Disorder=1, no Conduct Disorder=0

^dLifetime adulthood average heavy drinking frequency. Bolded entries are statistically significant results. Two separate models with different dummy codes were conducted to compare all possible contrasts across ADHD persistence, childhood ADHD only, and non-ADHD, but all results are presented together in this table because the effects of covariates were identical and to save space.

Post-hoc analyses.

We compared the average lifetime heavy drinking by ADHD diagnosis and by ADHD persistence by running separate ANCOVAs (controlling for sex and race or ethnic minority) overall and within each latent class (Table 4) and correcting for multiple comparisons. Overall, those with childhood ADHD showed significantly less lifetime average heavy drinking than those without. This difference was driven by the fact that those with childhood ADHD only drank less heavily than those without ADHD. Across all the group comparisons by latent class, only one was statistically significant: for those in the low blackout drinking class, those with adulthood-persistent ADHD showed significantly greater heavy drinking than those with childhood only ADHD and no ADHD.

Table 4.

ADHD Group Differences in Lifetime Average of Past-Year Heavy Drinking Frequency by Alcohol Problems Class

Class	Mean (N)			F
	ADHD		Non-ADHD		corrected pa
	Persistent	Childhood only
Full Sample
	2.47		3.07 b	9.178	.045
	2.32	2.80 b		6.055	.030
Interpersonal Problems/Risky Behaviors
Low	.95		1.38	3.643	.145
	1.01	.81		1.981	.264
Intermediate	3.14		3.54	2.495	.193
	3.08	3.27		4.440	.326
High	4.26		4.65	.786	.436
	3.77	4.95		3.119	.123
Occupational/Academic Problems
Low	1.52		1.82	1.829	.266
	1.53	1.52		.883	.445
Intermediate	3.58		4.47	7.222	.060
	3.67	3.34		3.69	.105
High	4.00		4.15	.104	.801
	3.57	4.47		1.40	.326
Impaired Control/Treatment Seeking
Low	1.72		2.18	3.886	.145
	1.86	1.22		3.661	.105
Intermediate	3.60		4.42	4.719	.120
	3.33	4.10		3.306	.123
High	3.68		4.20	1.011	.398
	3.31	4.02		1.427	.326
Tolerance/Withdrawal
Low	1.53		1.50	.016	.901
	1.58	1.35		.260	.772
Intermediate	3.56		4.03	2.810	.193
	3.54	3.59		1.402	.326
High	4.09		4.87	2.629	.193
	3.59	4.58		2.730	.159
Blackout Drinking
Low	.94		.63	1.305	.349
	1.02 c	.77 c		5.556	.030
High	3.23		3.73	5.33	.110
	2.99	3.73		.943	.445

Note. Results of ANCOVAs, adjusting for race and sex. Bolded lines are statistically significant results per corrected p-value; significantly different means in pairwise comparisons are bolded. Response scale for lifetime average past-year heavy drinking: 0=not at all; 1=1–3 times; 2=4–7 times; 3= 8–11 times; 4=Once/month; 5=2–3 times/month; 6=Once a week; 7=2–3 times a week; 8=4–6 times a week; 9=once/day; 10=Twice/day; 11=several times/day.

^aCorrected p-values for multiple comparisons using the false discovery rate, which was applied to the 15 ANCOVAs comparing ADHD to non-ADHD group and separately to the 15 ANCOVAs comparing ADHD persistence, childhood ADHD only, and non-ADHD. Pairwise comparisons only interpreted for ANCOVAs yieldinf significant corrected p-values. Tukey’s multiple comparison tests used for pairwise comparisons, which adjust significance level for multiple testing.

^bNon-ADHD group significantly differed from childhood ADHD only group.

^cADHD persistent group significantly differed from childhood ADHD only and non-ADHD group, but childhood ADHD only group did not significantly differ from non-ADHD group.

Discussion

Although prior research showed that individuals with ADHD were at higher risk for alcohol problems and disorder relative to those without ADHD (Charach et al., 2011; Pedersen et al., 2016), this is the first study to our knowledge to demonstrate that those with ADHD exhibited different types of alcohol problems. Results have clinical implications for this high-risk population and highlight important issues in the comorbidity between and measurement of ADHD and alcohol use disorder.

Interpersonal Problems/Risky Behaviors and Occupational/Academic Impairment

We found that individuals with childhood ADHD were more likely to belong to the high relative to the intermediate risk classes for alcohol-related interpersonal problems/risky behaviors and occupational/academic impairment. However, findings were driven by those with adulthood-persistent ADHD. Those with adulthood-persistent ADHD were more likely to be in the high groups of these alcohol problem types when compared to both those with childhood ADHD only and without ADHD (see Table 3 for more details).

These results are highly consistent with a prior study, not focused on individuals with ADHD, that found stronger associations between externalizing problems (i.e., conduct disorder, antisocial personality disorder, and impulsivity) and alcohol-related social problems and role interference (McDowell et al., 2019). Notably, findings held after controlling for heavy drinking and, for those classified in the same latent classes, the heavy drinking frequencies of those with adulthood-persistent ADHD were generally similar to that of those without ADHD. Although those with adulthood-persistent ADHD may not be drinking more frequently than those without ADHD, when they drink they tend to experience heightened problems with people, risk taking, and school or work. Thus, it may be fruitful to tailor alcohol treatment and prevention for those experiencing concurrent ADHD symptoms. Clinicians may provide psychoeducation about the unique ways in which alcohol consumption interacts with ADHD-related problems, including how worsening impulsivity and inattentiveness can impair already-strained social functioning, relationships, and professional/academic functioning. Moreover, for those with adulthood-persistent ADHD, limiting the frequency of drinking sessions may not be as useful as limiting the amount consumed each time or engaging social supports or self-monitoring to regulate behavior while drinking. Our findings suggest that a history of childhood ADHD may not alone convey higher risk for socially-oriented alcohol problems. Moreover, results shed some light on the heterogeneity in alcohol problems that has been observed among those with ADHD in prior studies (Wang et al., 2020) and suggest that ADHD persistence is one individual difference contributing to this heterogeneity. Thus, ongoing treatment for those with ADHD symptom persistence may be key in preventing certain types of alcohol problems for this population.

Impaired control/treatment seeking

Although individuals with childhood ADHD were also more likely to belong to the high relative to the intermediate class for impaired control/treatment seeking, we again found this result to be driven by adulthood-persistent ADHD. Indeed, those with adulthood-persistent ADHD were more likely to belong to the high than to the intermediate and low classes for impaired control/treatment seeking relative to both those with childhood ADHD only and those without ADHD. Notably, the impaired control/treatment seeking latent classes exhibited a distinct pattern whereby those in the high class were lower or equal to those in the intermediate class on items assessing inability to control drinking in the moment (i.e., drinking more than intended, controlling or cutting-down on drinking), but higher on items assessing alcohol treatment seeking. Thus, those with adulthood-persistent ADHD were particularly elevated in the treatment seeking components of impaired control in addition to showing high rates of the other features of impaired control. Perhaps individuals with ADHD symptom persistence remain accustomed to accessing mental health care due to their early psychiatric diagnoses and continuing problems, increasing their likelihood of doing so for drinking problems in adulthood. Alternatively, those with ADHD persistence could be more likely to be mandated to treatment due to other problems arising from alcohol use (e.g., unlawful, risky behaviors). Increased contact with treatment providers represents an important opportunity for tailored alcohol interventions within this high-risk group.

Tolerance/Withdrawal and Blackout Drinking

Individuals with and without childhood ADHD did not show differences in class membership for alcohol-related tolerance/withdrawal or blackout drinking. After parsing adulthood-persistent ADHD, we found that those with childhood ADHD only were more likely to be in the low tolerance/withdrawal class relative to those with adulthood-persistent ADHD and to those without ADHD (see Table 3 for details). Overall, the absence of ADHD-associated risk for tolerance/withdrawal and blackout drinking classes is consistent with the weak associations found among externalizing problems and alcohol tolerance in McDowell et al. (2019). It is also in line with the similar or lower frequencies of heavy drinking of our ADHD participants (both persistent and childhood only) relative to those without ADHD (see Table 4). Indeed, tolerance/withdrawal and blackout drinking behaviors are directly linked with the amount of alcohol consumed (Dawson, 2000; Jennison and Johnson, 1994).

Comorbidity and Measurement Considerations

Results add to our understanding of the comorbidity between ADHD and alcohol use disorder. First, results suggest that alcohol use disorders characterized by interpersonal problems/risky behaviors, occupational/academic, and impaired control/treatment seeking may be more commonly comorbid with ADHD. Second, findings add to existing theories of the causes of ADHD and alcohol use disorder comorbidity (Molina & Pelham, 2014). Namely, perhaps pre-existing deficits caused by one’s ADHD symptoms (social skills deficits, interpersonal problems, occupational problems, risk taking) are further magnified by any missteps caused by drinking, rendering these alcohol consequences to be endorsed more readily by those with ADHD. More research is needed to test this theory.

However, such a finding also points to potential issues with using certain types of alcohol consequences in the diagnosis of alcohol use disorder. According to Martin et al. (2014), alcohol-related consequences like interpersonal problems, risky behaviors, and occupational problems may introduce systematic bias to and degrade the validity of alcohol use diagnoses, in part because these alcohol consequences are multiply determined and may not be reliably traced back to actual alcohol pathology. Our results are consistent with this idea. Because those with adulthood-persistent ADHD did not drink more heavily than those without ADHD (a critical component of pathological alcohol use), their endorsement of alcohol-related consequences such as interpersonal and occupational problems could more heavily reflect their ADHD than their alcohol symptomatology. Continued consideration of this issue is important and could suggest the need to weight these types of alcohol consequences less heavily in diagnoses of alcohol use disorder.

This study had several strengths and limitations. We capitalized on prospective data to comprehensively assess lifetime alcohol problems while mitigating retrospective recall bias. Rigorous, research-quality ADHD diagnoses were measured in childhood and assessment of adulthood symptom persistence was based on multiple informants. Limitations included the lack of generalizability to races and ethnicities beyond non-Hispanic White individuals and female participants given the demographic make-up of the sample. It will be critical to replicate these results with more diverse samples. As problematic alcohol use was only assessed via self-report, utilizing collateral reporters would also strengthen the conclusions drawn. Despite the strengths in our assessment of alcohol problems, one drawback was that lifetime reports may miss effects specific to certain developmental periods (e.g., early vs. late 20s). Unfortunately, we did not have the necessary data to conduct the analyses herein separately by these developmental periods. Moreover, we were unable to create separate heavy drinking variable for men (5+ drinks) and women (4+ drinks) due to limitations of the data.

Conclusion

Individuals with childhood ADHD who continue to exhibit ADHD symptoms into adulthood may show distinctive patterns of drinking-related consequences. As individuals with ADHD may be more prone to social-, relational-, and occupational-drinking consequences, this population may benefit from alcohol interventions that more heavily focus on skills to mitigate these distressing outcomes.

Figure 2.

Final Three-Class Solution for the Occupational/Academic Problems Latent Class Analysis.