Age of Onset and Alcohol and Cannabis Use Disorders among Mexican American Young Adults: Robust Substance-Specific Effects of Early Use as a Risk Factor

Abstract

We investigated the substance-specific and cross-substance risk associated with early onset (before age 15) of drunkenness and cannabis use in the subsequent development of alcohol (AUD) and cannabis use disorder (CUD) in Mexican American young adults. Survival analyses employed Cox proportional hazards models for AUD and CUD, separately. In cross-risk analyses, we modeled estimates for those participants reporting lifetime use of both substances. Early onset of drunkenness and early onset of cannabis use were associated with shorter time to AUD and CUD, respectively, even after accounting for psychiatric disorders. While there were no cross-risk associations, adjusting for psychiatric disorders and early onset cannabis use attenuated the association of early drunkenness with AUD.

Introduction

Hispanic/Latinx people born in the United States (US) have rates of anxiety, mood, and alcohol (AUD) and cannabis (CUD) use disorders that are two to three times higher than those of their immigrant parents (Borges et al., 2016; Caetano & Clark, 2000). Further, earlier ages of first alcohol intoxication and first cannabis use, along with psychiatric co-morbidity, may be associated with more rapid development of use disorders among Hispanic/Latinx people in the US (Alderete, Vega, Kolody, & Aguilar-Gaxiola, 2000; Vega et al., 1998). The current study assesses associations of early age of onset of first drunkenness and cannabis use (defined here as occurring prior to age 15), with rapidity of AUD and CUD development in Mexican American young adults. We also establish whether early onset is a robust predictor for substance use disorders (SUD) after accounting for psychiatric disorders (anxiety/affective and antisocial/conduct disorders) and other demographic risk factors in this group.

Early Age of Onset

Earlier age of first alcohol and cannabis use is related to increased risk for later AUD and CUD development (Feingold, Livne, Rehm, & Lev-Ran, 2020; Newton-Howes & Boden, 2016). Young people who use alcohol and cannabis before mid-adolescence show greater risks for and shorter time to AUD and CUD in adulthood than those who initiate use after mid-adolescence (Dawson, Goldstein, Chou, Ruan, & Grant, 2008; Ehlers et al., 2010b; Rioux et al., 2018). Data from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) show Hispanic/Latinx people have among the highest prevalence rates of transitioning to AUD after their first alcohol use and the highest rate of transitioning to CUD after their first cannabis use relative to both White and Black populations (Lopez-Quintero et al., 2011). Because some investigators have found that Hispanic/Latinx populations have elevated risk for SUDs, it is critical to investigate correlates of AUD and CUD onset among this group.

Focus on Hispanic/Latinx populations

In addition to higher transition rates from first alcohol and cannabis use onset to AUD and CUD development (Lopez-Quintero et al., 2011), US national data show Hispanic/Latinx young people ages 12 to 21 have greater likelihood of initiating cannabis use before alcohol or cigarette use compared to their White counterparts (Fairman, Furr-Holden, & Johnson, 2019). Among Hispanic/Latinx national groups in the US, Mexican Americans have the highest prevalence of alcohol dependence, consumption volume, and youngest age at first drink compared to Puerto Ricans, Cuban Americans, and Central/South Americans (Jetelina, Reingle Gonzalez, Vaeth, Mills, & Caetano, 2016). Furthermore, earlier age of alcohol use onset is related to greater drinking volume, greater probability of binge drinking, more alcohol-related problems, and increased likelihood of AUD in Mexican Americans (Caetano, Mills, Vaeth, & Reingle, 2014). It is important to focus on subpopulations given evidence that progression to SUDs manifest differently than results from analyses using large, general population samples (Ehlers, Wall, Betancourt, & Gilder, 2004). To our knowledge, there are no studies assessing the time between early drunkenness and early cannabis use onset and AUD and CUD onset, respectively, in Mexican Americans. Identifying risk factors for SUD progression among Mexican Americans, the largest Hispanic/Latinx ethnic subgroup in the US, will enable development of targeted intervention strategies to mitigate harms from SUDs.

Cross-Risk Associations of Age at First Drunkenness and Cannabis Use

In addition to substance-specific transitions to use disorder (e.g., alcohol and AUD), there is some evidence of a cross-risk association between substances and SUD (e.g., alcohol and CUD). In one study using the NESARC, Weinberger and colleagues (2016) found current cannabis use by US adults, for both those with and those without histories of AUD, was associated with increased incidence of and likelihood of persisting AUD over time (Weinberger, Platt, & Goodwin, 2016). The empirical evidence is less clear for early alcohol use in relation to development of CUD, however. Data from community samples in Germany and New Zealand showed age at first drink was not associated with CUD after adjusting for confounders (Behrendt et al., 2012; Newton-Howes & Boden, 2016). One longitudinal study from New Zealand differentiated age at first drunkenness from age at first drink and concluded age at first drunkenness is a stronger predictor of substance use problems in adulthood (ages 18 to 35) than age at first drink (Newton-Howes, Cook, Martin, Foulds, & Boden, 2019).

The cross-risk relationship between age at first drunkenness and cannabis use and CUD and AUD, respectively, within the US context currently is unknown, especially among Mexican Americans. In addition to filling a critical gap to ascertain whether early drunkenness and early cannabis use onset each are related to more rapid development of AUD and CUD, we explore whether early onset of drunkenness and cannabis use additionally are associated with greater cross-risk of later CUD and AUD among a largely US-born sample of young adult Mexican Americans. In particular, we investigate whether early onset elevates cross-risk of later AUD or CUD while also accounting for psychiatric comorbidities.

Psychiatric Risk Factors

Empirical evidence on psychiatric internalizing pathways (including depression and anxiety) to substance use most consistently indicate that depressive symptoms in childhood and/or early adolescence are associated with later substance use and dependence, and this remains true across different cultural groups (Hussong, Ennett, Cox, & Haroon, 2017; Rothenberg et al., 2020), including Hispanic/Latinx groups (Gilder, Lau, Gross, & Ehlers, 2007). Depressive symptoms in adolescence often predict frequency and amount of alcohol use (Colder, Lee, Frndak, Read, & Wieczorek, 2019; Hussong et al., 2017) while, conversely, associations between depressive symptoms and subsequent cannabis use are not as strong (Colder et al., 2019; Englund & Siebenbruner, 2012; Farmer et al., 2015). Findings on the relationship between anxiety symptoms and substance use are varied (Hussong et al., 2017), though some research shows that anxiety disorders are more strongly associated with earlier onset SUD for White and Black men than for Hispanic/Latinx men (Gil, Wagner, & Tubman, 2004).

By contrast, externalizing behaviors and psychopathology (including impulsiveness and conduct disorder) are predictive of both alcohol use (Chassin, Pitts, & Prost, 2002; Englund & Siebenbruner, 2012; Grant et al., 2015; Hussong et al., 2017; Pedersen, Thomsen, Pedersen, & Hesse, 2017; Thompson, Leadbeater, & Ames, 2015) and cannabis use (Englund & Siebenbruner, 2012; Farmer et al., 2015; Hawes, Trucco, Duperrouzel, Coxe, & Gonzalez, 2019; Hayatbakhsh et al., 2008; Oshri, Rogosch, Burnette, & Cicchetti, 2011; Pedersen et al., 2017). Childhood conduct disorder and attention deficit hyperactive disorder, a condition accompanied by impulsivity (Ortal et al., 2015) and frequently comorbid with conduct disorder (Abikoff & Klein, 1992), is associated with earlier onset AUD and other SUDs (Meyers & Dick, 2010; Sung, Erkanli, Angold, & Costello, 2004). Antisocial personality disorder also is associated with earlier onset AUD for men across all racial and ethnic groups (Gil et al., 2004). A sub-aim of the current study investigates whether psychiatric predictors, specifically anxiety/affective disorders and antisocial personality/conduct disorders, explain the association between early onset of drunkenness and early onset of cannabis use and development of AUD and CUD in a unique sample of Mexican American young adults.

Current Study

Building on prior literature to fill a critical gap for young adult Mexican Americans, we examine two key questions: First, is there evidence of substance-specific and cross-substance risk from early onset (prior to age 15) to AUD and CUD? Second, are associations of early onset with later use disorders robust after accounting for other known psychiatric risk factors (internalizing and externalizing disorders)? We hypothesize early onset will be associated with more rapid development of both AUD and CUD among Mexican Americans. We further expect stronger associations of early cannabis use on subsequent AUD than early alcohol intoxication on CUD. Finally, we also expect externalizing disorders to be strongly associated with onset of both AUD and CUD.

Methods

Participants and procedure

Participants were recruited as part of a larger study of risk and protective factors for mental and physical health problems in Mexican Americans (Ehlers, Gilder, Criado, & Caetano, 2009; Ehlers, Stouffer, & Gilder, 2014). The participants were recruited using a commercial mailing list that provided the addresses of individuals with Hispanic/Latinx surnames in 11 ZIP codes in San Diego County located within 25 miles of the research site, with each having at least 20% Hispanic/Latinx residents. As described previously (Ehlers et al., 2014), mailed letters invited potential participants between ages 18 and 30, who were documented US residents, and who could complete the survey instruments in English. The participant also must have self-reported Mexican American heritage of at least one grandparent. Based on the overarching aims of the main study, participants were excluded if they were pregnant or nursing, currently had a major neurological disorder or head injury, or otherwise had a current major medical disorder that precluded travel to the research site. Exclusion criteria were applied in an initial phone screening step. Participants were asked to refrain from alcohol or any other substance use for 24 hours prior to testing, and their breathalyzer blood alcohol levels had to be 0.00 g/dl to be enrolled in the study. Participants gave written informed consent, and the study protocol was approved by the Scripps IRB.

Each participant completed the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA), which is a fully structured, poly-diagnostic psychiatric interview that has undergone both reliability and validity testing in multiethnic populations (Bucholz et al., 1994; Hesselbrock, Easton, Bucholz, Schuckit, & Hesselbrock, 1999). Interviewers were trained by personnel from the Collaborative Study for the Genetics of Alcoholism (COGA). The SSAGA collected information on self-reported substance use and psychiatric disorder symptoms, as well as demographic details (including ethnicity, highest education completed, sex, income, employment status, and marital status) and medical history (Bucholz et al., 1994). In addition to information from the SSAGA, electrophysiological measures, blood for genetic testing, and other psychological and alcohol use data were obtained (Ehlers et al., 2009; Ehlers, Gilder, Criado, & Caetano, 2010a; Ehlers et al., 2016; Ehlers, Phillips, Criado, & Gilder, 2011; Ehlers et al., 2019; Ehlers et al., 2014; Ehlers, Wills, Karriker-Jaffe, & Gilder, 2022; Gilder et al., 2007).

Measures

Age of onset for first drunkenness and for first cannabis use were determined by asking the respondent in the SSAGA interview, “How old were you the first time you got drunk, that is, your speech was slurred or you were unsteady on your feet? / How old were you the first time you used marijuana?” Diagnoses of lifetime AUD and CUD used DSM-V criteria¹ to assess whether the respondent had ever experienced symptoms up until the time of interview and the age at which they first experienced symptoms. Moderate use disorder was reporting 4 to 5 symptoms, and severe use disorder was reporting 6 or more symptoms in a 12-month period (e.g., substance taken in greater amount than intended, persistent desire to cut down, cravings, reduced social or work-related roles because of use, tolerance) (American Psychiatric Association, 2013). The outcome variables included all cases of higher severity, such that models of moderate use disorder also included cases that met criteria for severe use disorder (thus, including all participants reporting at least 4 symptoms). The current study focused analyses on onset of at least moderate use disorders given potential overdiagnosis of mild use disorders using DSM-V criteria in young people (Kaminer & Winters, 2015; Lane & Sher, 2015; Malone & Hoffmann, 2016). Lifetime diagnoses of antisocial personality disorder/conduct disorder (ASPD/CD), any independent affective disorder (major depressive disorder, dysthymia, bipolar I disorder), and any independent anxiety disorder (social phobia, agoraphobia, panic disorder, obsessive compulsive disorder) were defined by DSM-IV-TR criteria (American Psychiatric Association, 2000). A research psychiatrist and addictions specialist (DAG) made all final diagnoses for this study.

Control variables included demographic characteristics associated with early onset, SUD, and mental health, including female gender (vs. male gender as referent), highest educational level completed, income of $20,000 or more (vs. less than $20,000), employment (vs. currently unemployed), and marital status (vs. never married). Given genetic vulnerability for AUD (Chartier, Thomas, & Kendler, 2017) especially among Hispanic/Latinx people in the US, family history of AUD was determined by the Family History Assessment Module (FAM) in the SSAGA, and it was defined by AUD in a first-degree family member. The FAM was developed by the COGA team.

Analysis Sample

Analyses were conducted on the subsamples of individuals who provided information on age of onset of drunkenness and who had not developed AUD before age 15 (N = 609; AUD sample) and individuals who provided information on age of onset of cannabis use and who had not developed CUD before age 15 (N = 500; CUD sample). A subsample of respondents (n = 480) had been drunk and also had initiated cannabis use; this group was used in cross-risk analyses.

Analysis Strategy

For each sample, descriptive statistics (means and standard deviations or proportions) were calculated for key demographics and focal variables. We employed survival analysis to assess risk of onset of AUD and CUD. The outcomes for the survival analyses were the time from the start at age 15 to the age of AUD onset (for the AUD analysis sample) and the time from the start at age 15 to the age of CUD onset (for the CUD analysis sample). These outcomes were censored at the age of interview for individuals without lifetime AUD or CUD diagnoses. Survival analyses for the effects of early onset of drunkenness and cannabis used Cox proportional hazards models for the AUD and CUD outcomes, separately. For each outcome, the reduced model included the control variables listed above, and the full models included anxiety/affective disorders and ASPD/CD as additional predictors.

For all analyses, we examined violations of the proportionality assumption by plotting Kaplan-Meier (KM) survival curves stratified by age of onset of drunkenness (before age 15 vs. after) and age of onset of cannabis use (before age 15 vs. after), respectively. Log rank tests were calculated to test the difference in KM curves stratified by age of onset of use. In the cross-risk analyses, we estimated our reduced and full models in the subset of individuals who reported both lifetime drunkenness and lifetime cannabis use (thus, they were at risk for both types of use disorders), including both indicators for early age of onset of use as simultaneous predictors of onset of moderate use disorders. We also performed several sets of sensitivity analyses: We modeled time to onset of mild AUD and mild CUD in a similar way, plotting the KM survival curves stratified by early onset of drunkenness and cannabis use. Finally, we replicated analyses for our research questions using respondents confirmed as US-born only.

Results

Descriptive statistics for the AUD and CUD analysis samples are presented in Table 1.

Table 1.

Sample characteristics¹

	AUD2 Sample (N=609)	CUD3 sample (N=500)	Cross-risk AUD sample (n=480)	Cross-risk CUD sample (n=473)
Age at interview, Mean (SD)	24.03 (3.71)	23.98 (3.74)	24.10 (3.71)	24.08 (3.73)
Female (%)	58.62	56.60	56.04	56.03
Foreign-Born4 (%)	10.67	9.60	8.96	9.09
Years schooling completed, Mean (SD)	13.55 (1.78)	13.41 (1.80)	13.43 (1.79)	13.45 (1.79)
Annual income (%)
More than $20,000	73.6	73.6	72.5	73.1
Less than $20,000	18.6	17.8	18.5	18.0
Not reported	7.9	8.6	9.0	8.9
Currently employed (%)	65.2	64.4	64.8	65.1
Married (%)	14.8	12.8	12.9	12.5
Family history of AUD (%)	38.9	40.0	40.4	40.0
Any anxiety/affective disorder (%)	48.8	51.2	50.8	50.3
Antisocial personality disorder/conduct disorder (%)	9.8	10.8	11.9	10.8
Lifetime AUD status (%)
Mild AUD	53.2	—	59.0
Moderate AUD	26.1	—	31.0
Severe AUD	12.8	—	15.8
Lifetime CUD status (%)
Mild CUD	—	36.6		37.8
Moderate CUD	—	20.2		20.7
Severe CUD	—	9.2		9.3

Note.

¹Analyses were limited to people who reported ever having been drunk (for the AUD sample), used cannabis (for the CUD sample), and those who reported both ever having been drunk and used cannabis (for the cross-risk samples).

²AUD = alcohol use disorder.

³CUD = cannabis use disorder.

⁴Confirmed nativity status; 8.05%, 8.20%, 8.13% and 8.03% of the AUD, CUD, cross-risk AUD and cross-risk CUD samples, respectively, had missing data on nativity status. Despite missing data, any respondent who would identify as foreign-born have a small likelihood of being included in the analyses as rates of recent drunkenness and cannabis use remain low in adolescence, i.e. to capture early onset (Caetano et al., 2014; Salas-Wright et al., 2019; Szaflarski, Cubbins, & Ying, 2011)

The KM survival curves of time to AUD and CUD were stratified by early onset of drunkenness (before age 15 vs. after) in Figure 1 and by early onset of cannabis use (before age 15 vs. after) in Figure 2.

Figure 1.

Survival plots for moderate AUD: Early vs. later onset of drunkenness

Figure 2.

Survival plots for moderate CUD: Early vs. later onset of marijuana use

In both figures, the early onset group consistently had lower survival rates than the later onset group (log rank tests p-value < .001 for both moderate AUD and moderate CUD). A few cases met criteria for mild use disorder prior to age 15 (n = 13 for moderate AUD models and n = 5 for moderate CUD models). In sensitivity analyses we confirmed that all key findings remained consistent when these early onset cases were excluded from the final analyses and also when the sample included participants who met criteria for mild use disorders (see Supplemental Table 1 for models assessing relationships between early onset and mild use disorders) or was limited to US-born respondents only.

The survival models (Table 2) showed a moderately strong association between early onset of drunkenness with early occurrence of moderate AUD, even after accounting for family history of AUD (Models 1 & 2) (HR = 2.18, p <.001) and psychiatric conditions (Model 2) (HR = 1.71, p <.01).

Table 2.

Results from survival models assessing relationships between early onset and moderate use disorders

	Moderate AUD onset (N=609)
	Model 1: Without psychiatric disorders				Model 2: With psychiatric disorders
	HR	95% CI		p-value	HR	95% CI		p-value
Drunkenness Early Onset (<15 vs. >=15)	2.183	1.529	3.117	<.001	1.707	1.171	2.489	0.005
Female gender	0.530	0.384	0.733	<.001	0.549	0.394	0.764	<.001
Highest grade completed	0.939	0.856	1.030	0.182	0.951	0.867	1.044	0.295
Income $20K or more	0.876	0.578	1.327	0.531	0.940	0.617	1.431	0.771
Missing income	0.824	0.390	1.741	0.612	0.759	0.357	1.613	0.474
Currently employed	0.980	0.700	1.372	0.908	0.999	0.710	1.405	0.994
Married	0.955	0.616	1.482	0.839	1.004	0.645	1.561	0.987
Family history of AUD	1.558	1.131	2.147	0.007	1.481	1.074	2.044	0.017
Any anxiety/affective disorder					1.615	1.156	2.256	0.005
Antisocial personality disorder or conduct disorder					2.191	1.435	3.345	<.001
	Moderate CUD onset (N=500)
	Model 1: Without psychiatric disorders				Model 2: With psychiatric disorders
	HR	95% CI		p-value	HR	95% CI		p-value
Cannabis Use Early Onset (<15 vs. >=15)	1.992	1.313	3.023	0.001	1.854	1.213	2.836	0.004
Female gender	0.525	0.349	0.790	0.002	0.521	0.342	0.793	0.002
Highest grade completed	0.933	0.830	1.049	0.245	0.933	0.831	1.049	0.248
Income $20K or more	0.732	0.441	1.216	0.228	0.795	0.475	1.330	0.382
Missing income	0.917	0.418	2.014	0.830	1.097	0.495	2.428	0.820
Currently employed	0.901	0.598	1.360	0.621	0.959	0.634	1.451	0.842
Married	0.929	0.516	1.671	0.806	0.870	0.481	1.574	0.645
Family history of AUD	1.276	0.854	1.909	0.235	1.229	0.825	1.831	0.310
Any anxiety/affective disorder					1.955	1.285	2.973	0.002
Antisocial personality disorder or conduct disorder					1.764	1.064	2.923	0.028

Other than gender (with men having higher risk than women), none of the other demographic variables predicted moderate AUD onset. A similar pattern was found for moderate CUD, where early onset of cannabis use was associated with early occurrence of moderate CUD (HR = 1.99, p <.01), even after accounting for psychiatric conditions (HR = 1.85, p <.01) (Models 1 and 2).

The cross-risk models (Table 3) showed early onset of cannabis use was not a significant predictor of moderate AUD, but early onset of drunkenness was a significant predictor of moderate AUD (Model 1) (HR = 1.76, p <.01), until the psychiatric conditions were added (Model 2) (Figure 3).

Table 3.

Results from survival models assessing relationships between early onset of both substances simultaneously and use disorders

	Moderate AUD onset (N=480)
	Model 1: Without psychiatric disorders				Model 2: With psychiatric disorders
	HR	95% CI		p-value	HR	95% CI		p-value
Early onset of drunkenness (<15 vs. >=15)	1.763	1.171	2.654	0.007	1.470	0.967	2.236	0.072
Early onset of cannabis use (<15 vs. >=15)	1.144	0.762	1.716	0.517	1.037	0.687	1.566	0.861
Female gender	0.526	0.377	0.734	0.000	0.543	0.386	0.765	<.001
Highest grade completed	0.976	0.885	1.076	0.623	0.979	0.888	1.078	0.663
Income $20K or more	0.898	0.583	1.382	0.625	0.940	0.609	1.452	0.781
Missing income	0.792	0.373	1.684	0.545	0.724	0.339	1.548	0.405
Currently employed	0.958	0.678	1.353	0.807	0.970	0.685	1.375	0.865
Married	1.037	0.657	1.636	0.875	1.074	0.678	1.702	0.759
Family history of AUD	1.472	1.057	2.050	0.022	1.422	1.021	1.981	0.037
Any anxiety/affective disorder					1.546	1.095	2.181	0.013
Antisocial personality or conduct disorder					2.119	1.377	3.258	0.001
	Moderate CUD onset (N=4731)
	Model 1: Without psychological disorders				Model 2: With psychological disorders
	HR	95% CI		pvalue	HR	95% CI		pvalue
Early onset of drunkenness (<15 vs. >=15)	0.828	0.484	1.416	0.490	0.677	0.393	1.165	0.159
Early onset of cannabis use (<15 vs. >=15)	2.144	1.347	3.411	0.001	2.165	1.355	3.460	0.001
Female gender	0.491	0.323	0.747	0.001	0.491	0.319	0.756	0.001
Highest grade completed	0.941	0.833	1.062	0.323	0.937	0.829	1.059	0.297
Income $20K or more	0.748	0.444	1.260	0.274	0.813	0.480	1.377	0.442
Missing income	0.908	0.409	2.016	0.813	1.053	0.472	2.349	0.899
Currently employed	0.872	0.572	1.329	0.524	0.925	0.605	1.415	0.719
Married	1.025	0.569	1.848	0.934	0.995	0.550	1.799	0.986
Family history of AUD	1.379	0.917	2.073	0.123	1.334	0.892	1.996	0.161
Any anxiety/affective disorder					2.022	1.318	3.103	0.001
Antisocial personality or conduct disorder					1.783	1.059	3.000	0.029

Note:

¹Seven people were missing data on CUD age of onset for CUD in the cross-risk model predicting CUD.

Figure 3.

Survival plots for moderate AUD based on cross-risk analysis sample in Table 3, risk related to early onset of drunkenness (left panel) and cannabis use (right panel)

Similarly, early onset of drunkenness was not a significant predictor of moderate CUD, but early cannabis use remained a significant predictor of moderate CUD (HR = 2.14, p <01), even when adjusting for the psychiatric conditions (Figure 4) (HR = 2.17, p <.01).

Figure 4.

Survival plots for moderate CUD based on cross-risk analysis sample in Table 3, risk related to early onset of drunkenness (left panel) and cannabis use (right panel)

In all models, anxiety/affective disorders and ASPD/CD were both significant predictors of AUD and CUD, in which anxiety/affective disorders were more strongly associated with CUD and ASPD/CD with AUD.

Discussion

The current study examined relationships between early onset of drunkenness and cannabis use with development of AUD and CUD in a sample of young adult Mexican Americans, accounting for key demographic predictors and psychiatric disorders. Our hypotheses were partially supported. First, we tested associations between early age of onset, i.e., before age 15, with AUD and CUD and found early onset of drunkenness and cannabis use were associated with later development of the same-substance use disorder. This substance-specific risk also was evident in the cross-risk models, as controlling for early onset of the other substance did not affect the association between age of onset and same-substance use disorder. As expected and in alignment with general population studies (Flórez-Salamanca et al., 2013), respondents with early onset of drunkenness and early onset of cannabis use experienced a shorter time to AUD and CUD, respectively, relative to those without early onset. To our knowledge, this is the first study assessing the time between early drunkenness and cannabis use onset and AUD and CUD development among Mexican Americans.

Second, the current study found no evidence of cross-risk associations between early cannabis use and AUD and early drunkenness and CUD. The null result on the early cannabis use and AUD risk relationship runs counter to a US national study showing increased risk of AUD for those who used cannabis (Weinberger et al., 2016). It also is possible we did not observe an association between early onset drunkenness and CUD because, relative to most extant studies, we utilized a more stringent indicator of alcohol use onset—first drunkenness—rather than first drink in general (Newton-Howes et al., 2019). One report found a significant association of early drunkenness with CUD and with faster transition to CUD, even after controlling for internalizing behaviors (Newton-Howes et al., 2019). That study, however, used a general population sample from New Zealand, so it may be that findings in the current study are unique to the histories and lived experiences of Mexican Americans in the US. Current findings fill a critical literature gap on the cross-substance associations of early onset with development of later SUD in a US context, especially relevant for Mexican American young adults. While Hispanic/Latinx individuals in the US have among the highest rates of transition from first alcohol and cannabis use onset to AUD and CUD (Lopez-Quintero et al., 2011), these results showed no cross-substance risk for later AUD and CUD development among this specific ethnic group. Thus, it is critical to continue to disaggregate the diverse US population in future studies of SUD as developmental pathways to SUD may vary by ethnicity.

Third, it is important to highlight that significant associations between early onset and use disorders remained after accounting for lifetime psychiatric comorbidity in this sample of young adult Mexican Americans. While we expected externalizing disorders (antisocial personality or conduct disorders) to be associated with both AUD and CUD (Meyers & Dick, 2010; Sung et al., 2004), internalizing disorders (anxiety or affective disorders) also were associated with both AUD and CUD. It is likely that these externalizing and internalizing disorders co-evolved with SUD symptoms and thus produced strong associations (Gilder & Ehlers, 2012). Of note was a finding in the cross-risk assessments, in which early onset of drunkenness was no longer significantly associated with AUD after accounting for psychiatric disorders and early onset of cannabis use. This lends some indication of other comorbidities at play and mechanisms underlying this relationship that would be worthwhile for further investigation.

Poor mental health, nevertheless, is increasingly prevalent in Hispanic/Latinx youth in the US (Platt et al., 2020), and it is critical for prevention efforts to identify psychiatric symptomology and substance use through school-based and other context-specific risk assessments. Psychiatric comorbidity may also be associated with transition from first use to dependence, likely via using substances to cope with stressors (Flores, Tschann, Dimas, Pasch, & de Groat, 2010). Stressors relevant to Hispanic/Latinx young people include coping with discrimination (Gilbert & Zemore, 2016) and negotiating with multiple cultures in their households (Perreira et al., 2019; Piña-Watson, Ojeda, Castellon, & Dornhecker, 2013), which may lead to elevated rates of acculturative stress (Ehlers et al., 2009). Furthermore, due to the current study’s location in Southern California, Mexican American young people are adjacent to the US-Mexico border where there is greater likelihood of experiencing or witnessing trauma, compared to those who live further from the border, due to violence in the area (Sabo et al., 2014) and the potential risk of family separation (Cobb et al., 2021). Future work should examine these factors as additional predictors influencing internalizing and externalizing pathways to SUDs. Because of gender differences in both SUDs and psychiatric conditions (Sher et al., 2015; Vasilenko, Evans-Polce, & Lanza, 2017), future research also should investigate these relationships for subgroups defined by gender identity.

There are several limitations to note for these analyses. First, temporality is not guaranteed. That is, a participant could have had an internalizing or externalizing disorder diagnosed after their reports of AUD or CUD symptomology. Conduct disorders as characterized by antisocial personality disorder, however, usually emerge during the childhood and adolescent development periods (Fairchild et al., 2019) and precede our study age range of 18 to 30 years. This age range may also underestimate AUD or CUD onset. Future studies could use a prospective design to assess for symptomology. Retrospective reporting of early onset and lifetime diagnoses in the SSAGA also are limitations given increased likelihood for participant recall bias. We also did not have data on the age of first intoxication by cannabis, so analyses are limited to age of first use. Some of the psychiatric disorders were relatively rare in this sample, necessitating aggregation into categories such as “any anxiety or affective disorder” which may obscure some differences in developmental trajectories that could be detected in a larger sample.

Finally, study results cannot be generalized to all Mexican Americans in the US. The current sample was comprised of those who responded to an outreach flier within a particular region. Also of note is that a small subset of our sample was foreign-born; second- and later- Hispanic/Latinx generations, or those born in the US, generally exhibit greater rates of behavioral and health risks than their foreign-born counterparts, and this includes SUDs (Otiniano Verissimo, Grella, Amaro, & Gee, 2014; Strunin, Edwards, Godette, & Heeren, 2007). Our sensitivity analyses, however, confirmed focal relationships were similar when the analysis sample was limited to US-born respondents. Because study eligibility was limited to respondents who could answer the survey instruments in English, our foreign-born participants may be more English-proficient than foreign-born respondents excluded from participation; linguistic acculturation is associated with recent substance use in Hispanic/Latinx adolescents (Parsai, Voisine, Marsiglia, Kulis, & Nieri, 2009) and SUDs in Hispanic/Latinx adults in the US (Blanco et al., 2013).

Despite the aforementioned limitations, these analyses provide data on the substance-specific and cross-risk associations between age of onset and two common SUDs in a high-risk Mexican American sample. Identifying correlates of SUD onset is critical for age-based interventions in a population subgroup that is less likely to access specialty alcohol or drug treatment relative to their non-Hispanic/Latinx peers (Guerrero, Marsh, Khachikian, Amaro, & Vega, 2013). Finally, this evidence that early cannabis use onset is related to rapid development of CUD, coupled with data suggesting Hispanic/Latinx young people are likely to initiate cannabis use first compared to other substances (Fairman et al., 2019), highlights that future studies may need to account for the changing legal landscape of cannabis and its increasing use in the US (Jones et al., 2020).

Conclusions

Young Mexican Americans who engage in early substance use are at greater risk for a faster transition to SUDs than those who do not use at such an early age, even after accounting for mental health diagnoses. The Hispanic/Latinx population is becoming increasingly US-born (Budiman, Tamir, Lauren, & Noe-Bustamante, 2020) and comprised of diverse sub-ethnicities that contribute to varying within-group prevalence rates of substance use and related problems. Hence, it is essential to continue to explore these differences. It remains critical to work towards targeted interventions for Hispanic/Latinx populations, and in particular for Mexican Americans, given evidence pointing to lower rates of access to care for both SUDs and mental health despite increasingly higher rates of diagnoses.