Elucidating the Links Between Mother and Father Alcohol Use Disorder and Adolescent Externalizing Psychopathology: A Test of Transmission Specificity Within Competing Factor Structures and Genetic and Environmental Liabilities

Abstract

Parental alcohol use disorder (AUD) is a substantiated risk factor for adolescent externalizing psychopathology; however, the level of specificity at which risk from parental AUD is transmitted to adolescent offspring should be interrogated further. The current study modeled competing factor structures of psychopathology in a sample of 502 adolescent twin pairs (M_age= 13.24 years) and tested associations with mother and father AUD. The bifactor model exhibited the best fit to the data when contrasted with correlated factors and general factor models. Paternal AUD predicted the externalizing and internalizing correlated factors, the adolescent P-factor but not the residual externalizing and internalizing factors, and the general factor. No significant associations with maternal AUD were noted. Lastly, the latent factors of adolescent psychopathology were all moderately heritable (h²= .44–.59) and influenced by the nonshared environment. Shared genetic factors primarily explained externalizing and internalizing covariance. Findings suggest that efforts to mitigate risk in offspring of fathers exhibiting AUD require broader approaches that address the full range of adolescent symptomology.

Adolescence is a developmental period marked by a complex convergence of developing competencies in socioemotional, intellectual, and behavioral domains as well as biological change. Risk for psychopathology emerges when development across these domains is incongruous and fails to coordinate (Steinberg, 2005). Rates of externalizing psychopathology increase significantly in adolescence (Moffitt, 2006), and these increases have been associated with myriad negative consequences, including peer victimization and poor academic functioning (Vaillancourt et al., 2013). Identifying risk factors that underlie externalizing problems in adolescence and precipitate negative developmental cascades is a necessary means of intervening to divert snowballing effects. A well-established risk factor for externalizing psychopathology is parental alcohol use problems (Park & Schepp, 2015). Children of parents with alcohol use problems are subject to elevated genetic and environmental risk for developing externalizing problems themselves (McGue, 1997). While this notion that psychopathology tends to cluster within families has garnered virtually axiomatic status, the level of specificity at which we should examine the familial aggregation of mental health symptomology has been challenged in response to high rates of comorbidity both within and across the externalizing and internalizing domains, calling for more generalist approaches (McLaughlin et al., 2012). Consequently, there has been a recent departure from traditional nosologies that tenuously differentiate disorders, failing to explain the heterogeneity within disorders and domains as well as the frequent co-occurrence across them.

Calls to interrogate pervasive, yet potentially flawed conceptualizations of adolescent symptom presentation and the conferral of intergenerational risk stand to illuminate critical processes at a developmental period ripe with opportunity to intervene and subvert persistent and detrimental cycles. The current study aims to do so by 1) elucidating externalizing symptom presentation in adolescence and contextualizing it within various factor structures of psychopathology, 2) examining whether parental alcohol use confers a specific risk for adolescent externalizing symptomology or a more general risk to psychopathology, and 3) determining the degree of genetic and environmental influence on adolescent externalizing problems, shared etiology with internalizing problems, and a hierarchical general factor of psychopathology capturing co-occurrence across externalizing and internalizing domains.

Parental Alcohol Use Problems and Adolescent Outcomes

Recent estimates indicate that approximately 7.5 million children in the United States live in households with at least one parent exhibiting an alcohol use disorder (Lipari & Van Horn, 2017). These children are susceptible to myriad adverse psychological outcomes, including early-onset alcohol use problems (Handley & Chassin, 2013; Johnson et al., 2019). They are also liable to display externalizing problems more broadly, such as hyperactivity, conduct problems, and delinquent behavior (Kendler et al., 2013). Moreover, they develop symptoms that transcend beyond the broadband domain of externalizing and into that of internalizing problems (Hussong et al., 2008). A consideration of the variable outcomes exhibited by children of parents with alcohol use problems prompts interrogation of the specificity at which intergenerational risk is conferred.

A number of familial factors have been implicated as primary mechanisms through which intergenerational risk from parental alcohol use problems is transmitted to children. For example, problematic parent-adolescent communication and a lack of family cohesion have been found to mediate associations between paternal problem drinking and adolescent externalizing problems (Finan et al., 2015). However, the robust area of literature examining the negative consequences incurred by children of parents with alcohol use problems has tended to neglect the key role of shared genetics in mediating parent-offspring associations. Behavioral genetic studies can address this oversight. For example, Knopik and colleagues (2006) examined attention deficit hyperactivity disorder (ADHD) in offspring of identical and fraternal twin parents concordant and discordant for alcohol use disorder using the children-of-twins design. They found that offspring of mothers with a history of alcohol use disorder and mothers without but who had an identical twin sibling with a history were more likely to exhibit ADHD. These findings imply the transmission of pleiotropic genetic effects, such that some of the genes passed on from parents to offspring are likely to influence both alcohol use disorder and ADHD. Pleiotropic effects suggest that genetic influences may explain common variance across externalizing psychopathologies rather than unique variance. Moreover, the influence may be even broader, with covariance across the externalizing and internalizing domains explained by common genetic proclivity.

Heritability of Externalizing Problems in Adolescence

Behavioral genetic studies can be used to elucidate the influence of genetic and environmental factors on the variance of externalizing problems in adolescence by leveraging the data of monozygotic (MZ) twin pairs who share 100% of their genes and dizygotic (DZ) twin pairs who share 50% of their segregating genes. Moreover, they can be used to examine the shared etiology underlying covariance across different forms of psychopathology. Adolescent twin studies examining genetic and environmental influences on externalizing behavior broadly and oppositional defiant disorder and conduct disorder specifically typically report moderate heritability estimates, though estimates can range substantially from 0% to 77% (Button et al., 2008; Dick et al., 2005; Ehringer et al., 2006; Gjone & Stevenson, 1997; Korhonen et al., 2012; Kuo et al., 2004; Scourfield et al., 2004; Silberg et al., 1994; Young et al., 2000 34,66). Studies consistently report a substantial influence of the nonshared environment, however there is an inconsistent detection of significant shared environmental effects. There is a relative paucity of research examining the etiology of attention deficit hyperactivity disorder in adolescence when compared to analogous research completed with child-aged samples, however these studies do report moderate to high estimates of heritability, modest to moderate influence of the nonshared environment, and no effect of the shared environment, which is comparable to estimates from child-aged samples (Dick et al., 2005; Ehringer et al., 2006; Hay et al., 2004; Silberg et al., 1996; Young et al., 2000).

The Factor Structure of Adolescent Psychopathology

The multi-final outcomes incurred by children of parents exhibiting alcohol use problems span both externalizing and internalizing domains and frequently co-occur. Studies examining associations between parental alcohol use problems and adolescent externalizing problems imply a specific conduit for parent-offspring transmission that does not reflect the complex reality of symptom presentation. There has been considerable debate surrounding the utility of parsing symptomology into broad dimensions versus specific disorders (Cuthbert, 2005), but the categorization of symptomologies into externalizing and internalizing dimensions has been largely embraced. However, an even broader, hierarchical approach may be needed. Modeling techniques, both phenotypic and behavioral genetic, must be revisited to reflect complex structures of symptom presentation in adolescence.

Genetically informed research stands to elucidate the underlying mechanisms which drive development of psychopathology at variable levels of specificity. Within the domain of externalizing symptomology, covariation among impulsivity, inattention, conduct problems, and oppositional defiant behavior has been found to be mediated by overlapping genetic factors in adolescence (Dick et al., 2005; Knopik et al., 2009; Knopik et al., 2014). These findings suggest a heritable, common liability to externalizing psychopathology. Furthermore, studies utilizing common factor models to extract behavioral disinhibition and externalizing latent factors further bolster evidence of moderately to highly heritable common liabilities to externalizing in adolescence (Cosgrove et al., 2011; Young et al., 2000).

The degree of heritability underlying these latent factors prompts the question: how do these broader externalizing phenotypes aggregate within families and can they potentially provide greater insight into the intergenerational transmission of psychopathology? Two studies leveraged externalizing symptom data in parents and adolescent twin offspring in order to address these questions. Latent factors indexing common liability to various externalizing disorders were formed in parent and offspring generations and correlated. Then, parent and offspring disorder-specific residual variances were covaried to determine if specific transmission effects characterized familial aggregation of externalizing psychopathology above and beyond general transmission. The best fitting models allowed for general transmission between parents and offspring only, suggesting that a general liability to externalizing sufficiently explained familial resemblance for psychopathology. These general vulnerabilities to externalizing were highly heritable (h²=.81 & .80), indicating that the covariation among adolescent externalizing disorders is largely mediated by overlapping genetic factors (Bornovalova et al., 2010; Hicks et al., 2004).

These findings suggest that parent-offspring transmission of externalizing psychopathology in adolescence is mediated by the transmission of a highly heritable general liability rather than disorder-specific effects. However, behavioral genetic research examining the shared etiology across both externalizing and internalizing domains in adolescence implicates a broader, heritable common liability underlying both (Cosgrove et al., 2011; Gjone & Stevenson, 1997; O’Connor et al., 1998; Subbarao et al., 2008). This research illuminates shared etiological mechanisms underlying the high phenotypic correlations observed among externalizing and internalizing disorders and conferring transdiagnostic risk, suggesting that adolescents with genetic proclivity toward externalizing psychopathology may also be predisposed to internalizing. This and recent phenotypic evidence indicating the presence of a hierarchical, general factor underlying all common adolescent psychopathology suggests that these approaches to elucidating the intergenerational transmission of externalizing psychopathology may, in fact, have been too specific.

Studies leveraging diagnostic data in adult samples have tested various factor structures in an attempt to elucidate the dimensional structure of psychopathology, including correlated factor and bifactor models extracting subordinate specific externalizing and internalizing factors and hierarchical general factors. In these models, the general P-factor accounted for all of the shared variance across disorders while the subordinate specific externalizing and internalizing factors accounted for the common, residual variance among the respective disorders. The bifactor models fit the data best, and tests of external validity demonstrated that the P-factor was more strongly associated with history of parental psychopathology than the specific dimensions, suggesting that intergenerational transmission effects may operate at a more general level than considered prior (Caspi et al., 2014; Lahey et al., 2012)..

These results lend further support for a need to rigorously test whether the transmission of specific or more general liabilities explains the familial aggregation of psychopathology. If the latter, elucidating the mechanisms which underlie these general transmission effects is an important next step. Measured gene research reports a single nucleotide polymorphism heritability of 38% for the P-factor in a sample of children (Neumann et al., 2016), and twin research has found the P-factor to be moderately heritable in childhood and adolescence (Allegrini et al., 2020; Waldman et al., 2016). If effectively pursued, the proposed research stands to expand upon these findings, providing insight into the specificity at which parental alcohol use disorder confers risk for adolescent psychopathology and the etiological mechanisms conferring common and specific risk for externalizing and internalizing psychopathology at multiple levels of analysis- insight, which if harnessed appropriately, could meaningfully inform efforts to ensure positive developmental trajectories for vulnerable youth.

Hypotheses

We aimed to first elucidate externalizing symptom presentation in adolescence and contextualize it within competing factor structures of psychopathology. We hypothesized that the bifactor model would represent the data best when contrasted with the correlated factors and general factor models, given high rates of co-occurrence across the externalizing and internalizing domains and also meaningful variance unique to each. Furthermore, we expected the correlated factors to be highly correlated (and thus lacking specificity) and the bifactor residual externalizing and internalizing factor scores to be negatively and modestly correlated, after accounting for shared variance. Second, we examined whether parental alcohol use confers a specific risk for adolescent externalizing symptomology or a more general risk to psychopathology. We hypothesized that mother and father alcohol use disorder would predict both externalizing and internalizing correlated factor model scores, P-factor scores, and general factor model scores, indicating nonspecific parent-offspring transmission. Despite similar labels, the dimensions derived from the three factor structures differ in their explanatory breadth and interpretation. As such, associations between parental AUD and the various dimensions were not assumed to be equivocal and in turn tested separately. We did not expect any differences in predictive utility between mother and father alcohol use disorder. Third, we determined the degree of genetic and environmental influence on adolescent externalizing problems, shared etiology with internalizing problems, and a broader, hierarchical dimension of psychopathology capturing co-occurrence across externalizing and internalizing domains. We hypothesized that externalizing, internalizing, and general factors of psychopathology would be influenced by additive genetic and nonshared environmental influences and that the genetic influence on externalizing and internalizing would be largely shared, supporting evidence of general genetic risk (Cosgrove et al., 2011; Waldman et al., 2016).

Methods

Participants

The sample comprised 502 twin pairs and their parents drawn from the Wisconsin Twin Project, a population-based study of child and adolescent emotion, temperament, and psychopathology (Schmidt et al., 2019). Twin births between 1989 and 2004 were identified through state birth records, and families were invited to participate when the twins were in toddlerhood, middle childhood, and adolescence (See Schmidt et al., 2019 for details regarding sample recruitment at each wave). The current study employs parent and twin offspring data from the early adolescence assessment (Range = 11–18 years, M = 13.24 years, SD = 1.52). The sample comprised 22% monozygotic (MZ) female, 15% MZ male, 16% dizygotic (DZ) female, 18% DZ male, and 29% opposite-sex DZ twin pairs. Approximately 83% of the sample was categorized as White (8% Black; 2% Native American; 3.8% multiracial; 2% other; less than 1% Filipino, Hmong, or Other Pacific Islander). Mothers and fathers had an average education of 15.28 (SD=2.35), and 14.64 (SD=2.52) years, respectively. The median income bracket ranged from $60,001 to $70,000 with approximately 19% reporting a family income of $40,000 or less and approximately 40% reporting a family income of $80,000 or more. The recruitment procedures for the Wisconsin Twin Project do not reflect those of a traditional longitudinal study in that not all families who participated at the initial assessment were invited to participate in subsequent assessments. For example, the sample was mildly enriched for psychopathology in middle childhood. As such, delineating the ethnic/racial and socioeconomic composition of the current sample relative to that of the initially recruited sample is a complex endeavor. The ethnic/racial composition of the current sample does, however, reflect that of the state of Wisconsin. Furthermore, the current sample comprises a broad range of socioeconomic backgrounds, with representation at both lower and higher income levels. The median household income and average education level in the current sample are also comparable to those of the state of Wisconsin (U.S. Census Bureau, 2019).

Procedure

Adolescent twins were interviewed separately during a home visit to acquire independent reports of psychiatric symptomology using structured clinical interviews administered on laptop computers by a trained staff member. Parents were interviewed separately over the telephone using similar methods. Participants were monetarily compensated for their participation. All protocols were approved by a University of Wisconsin–Madison IRB.

Measures

Zygosity Questionnaire for Young Twins

The Zygosity Questionnaire for Young Twins (Goldsmith, 1991) is a 32-item measure designed to assess the zygosity of twin pairs. Caregivers responded to questions regarding their pregnancy, the physical appearance of each twin, and the presence of observable differences between the twins. The agreement of the questionnaire with genotyping has been estimated at 96% (Forget-Dubois et al., 2003), rendering it a less burdensome and more cost effective alternative. The questionnaire was administered on multiple occasions, and photographs and birth records were also examined. Ambiguous cases were resolved with genotyping.

Composite International Diagnostic Interview

The Composite International Diagnostic Interview (CIDI; Robins et al., 1988) was administered to both biological mothers and fathers independently. The CIDI is a fully structured comprehensive lifetime interview designed to yield DSM-IV-based psychiatric diagnoses. CIDI diagnoses are significantly related to independent clinical diagnoses, and test-retest reliability is high (Kessler & Üstün, 2004). Alcohol dependence and abuse diagnoses were collapsed and recoded into an Alcohol Use Disorder category.

Diagnostic Interview Schedule for Children

The National Institute of Health’s computer-assisted Diagnostic Interview Schedule for Children, version IV (C-DISC-IV; Shaffer et al., 2000), was administered to each twin independently. The C-DISC-IV is a structured diagnostic instrument based on the DSM-IV designed to assess childhood and adolescent diagnoses and associated symptom counts occurring over the past 12 months and the past 4 weeks. Reliability and validity of the C-DISC are acceptable and represent the gold standard in the field (Shaffer et al., 2000). Symptom counts (rather than diagnoses given the young age of the adolescents) of major depressive disorder, obsessive-compulsive disorder, panic disorder, social and specific phobias, generalized anxiety disorder, separation anxiety, attention deficit hyperactivity disorder, conduct disorder, and oppositional defiant disorder were assessed and used in confirmatory factor analyses to test various factor structures of adolescent psychopathology.

Covariates

Symptom counts of adolescent psychopathology were residualized on age and sex, following standard practice for twin model fitting (McGue & Bouchard, 1984). The residualized scores were used in subsequent confirmatory factor analyses. Factor scores were extracted and used in phenotypic parent-offspring models and twin biometric models. Socioeconomic status (i.e., mean composite of standardized family gross income, years of mother education and father education) and ethnicity were included as additional covariates in phenotypic parent-offspring models.

Statistical Approach

Correlations, descriptive statistics, confirmatory factor analysis, and multivariate regression were conducted in MPlus version 7.4 (Muthén & Muthén, 2012). Confirmatory factor analysis tested three alternative models of the dimensional structure of adolescent psychopathology: 1) correlated factors model, 2) bifactor model, and 3) general factor model (Figure 1). Adolescent symptom counts were normally distributed and not zero-inflated, qualifying MLR estimation as appropriate. The MLR estimator produces maximum likelihood parameter estimates and standard errors which are robust to non-normality and non-independence of observations when used with the “type = complex” command. Model fit was verified using multiple indices, including the Tucker-Lewis Index (TLI: Tucker & Lewis, 1973), the Root Mean Square Error of Approximation (RMSEA; Browne & Cudeck, 1992), the Standardized Root Mean Square Residual (SRMR; Bentler, 1995), and the Bayesian Information Criteria (BIC; Raftery, 1995).

Figure 1.

Alternative models of the factor structure of adolescent psychopathology. Three models were tested using confirmatory factor analysis: 1) correlated factors model, 2) bifactor model, and 3) general factor model. Ovals represent latent (unobserved) factors and boxes represent observed symptoms counts. MDD = major depressive disorder; OCD = obsessive compulsive disorder; PD = panic disorder; Soc Phob = social phobia; Spec Phob = specific phobia; GAD = generalized anxiety disorder; Sep Anx = separation anxiety; ADHD = attention deficit hyperactivity disorder; CD = conduct disorder; ODD = oppositional defiant disorder.

Next, multivariate regression analyses testing associations between parental alcohol use disorder and adolescent factor structure scores were conducted in Mplus 7.4 using the “type = complex” command to account for twin interdependence and the MLR estimator to account for missing data using full information maximum likelihood with robust standard errors. Three separate models were tested in which outcomes were 1) EXT and INT factor scores derived from the correlated factors model, 2) P-, residual EXT, and residual INT factor scores derived from the bifactor model, and 3) factor scores derived from the general factor model.

Twin intraclass correlations on factor scores were computed in OpenMx, an R-based program that utilizes maximum likelihood estimation (Neale et al., 2016). We then fit univariate and bivariate ACE models estimating genetic and environmental influences on the latent, dimensional factors of adolescent psychopathology. The ACE model is a multigroup structural equation model that utilizes observed phenotypic variances and covariances of MZ and DZ twins to estimate latent factors A (or additive genetic influence, representing the sum of the average effects of individual genetic variants across the genome), C (or shared environmental influence, representing environmental exposures shared by both cotwins that contribute to cotwin similarity), and E (or non-shared environmental influence, representing environmental exposures uniquely experienced by one cotwin, contributing to intrapair differences, as well as measurement error). ADE models were tested when MZ twin correlations exceeded DZ twin correlations by more than double, estimating latent D (or nonadditive genetic influence, representing the interaction of alleles at the same or different loci) in addition to A and E factors.

MZ twins share 100% of their segregating genes, and DZ twins share, on average, 50%; as such, the A factor correlation is fixed to 1 for MZ twins and to .5 for DZ twins. C is fully shared by cotwins; as such, the C factor correlation is fixed to 1, regardless of zygosity. Lastly, the E factor correlation is fixed to zero for both MZ and DZ twins. In the case of ADE models, MZ twins share 100% of nonadditive genetic effects, whereas DZ twins inherit the same alleles at a locus 25% of the time; as such, the D factor correlation is fixed to 1 for MZ twins and to .25 for DZ twins. First, full models estimating latent A, C (or D), and E factors were fit. Next, parameters were systematically dropped and the fits of reduced, nested models were compared against the full model using the −2 log likelihood chi-square test of fit to determine the most parsimonious solution. A nonsignificant difference in fit suggests that the reduced model represents the observed data as well as the full model, whereas a significant change in fit indicates that the dropped parameter should be retained, as it is required to accurately represent the observed data. E is always retained because it contains measurement error.

Bivariate twin models were also fit to determine the genetic and environmental contributions to shared etiology across adolescent externalizing and internalizing factors derived from the correlated factors model. The bivariate Cholesky decomposition estimates unique additive genetic (A22), shared environmental (C22), and nonshared environmental influences (E22) on phenotype 2 while accounting for shared additive genetic (A21), shared environmental (C21), and nonshared environmental (E21) influences on phenotype 1. Estimates from the bivariate Cholesky decomposition can be used to calculate the proportion of shared variance explained by A, C, and E influences as well as the proportion of total variance in phenotype 2 that is explained by shared and independent A, C, and E influences. When MZ cross-twin cross-trait correlations are higher than those of DZ twins, additive genetic effects are implicated as underlying co-occurrence. When there are no differences between MZ and DZ cross-twin cross-trait correlations, shared environmental influences are implicated. Additionally, the bivariate twin model can estimate the degree to which genetic (rG), shared environmental (rC), and nonshared environmental factors (rE) for two traits are correlated.

Results

Preliminary Analyses

Correlations and descriptive statistics for twin symptom counts are presented in Table I. All but one of the variables were beneath the recommended cutoffs for skew (+/−2.00) and kurtosis (+/−7.00; Muthén & Kaplan, 1985); the adolescent conduct disorder symptom count was square root transformed to approximate normality. Symptom counts were normally distributed and not zero inflated making Pearson correlation and MLR estimation appropriate. Symptom counts across all of the adolescent diagnoses were significantly and positively correlated, ranging from r = 0.16 to r = 0.62, p < .01.

Table I.

Partial Correlations Controlling for Age and Sex and Descriptive Statistics for Twin Symptom Counts

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.
1. ADHD	-
2. CD	.49**	-
3. ODD	.56**	.56**	-
4. MDD	.62**	.44**	.53**	-
5. OCD	.47**	.30**	.30**	.54**	-
6. PD	.30**	.20**	.26**	.37**	.39**	-
7. Soc Phob	.27**	.16**	.24**	.35**	.34**	.22**	-
8. Spec Phob	.33**	.22**	.23**	.35**	.35**	.28**	.36**	-
9. GAD	.46**	.27**	.39**	.59**	.50**	.38**	.50**	.40**	-
10. Sep Anx	.40**	.26**	.26**	.49**	.50**	.38**	.44**	.51**	.54**	-
Mean	3.55	1.37	3.11	4.06	.82	.40	3.53	1.14	2.87	1.87
SD	3.99	2.38	2.81	4.03	1.14	.61	3.71	1.39	2.44	2.13
Minimum	0	0	0	0	0	0	0	0	0	0
Maximum	20	20	12	20	6	4	13	8	11	11
Skewness	1.46	2.89	.86	1.14	1.63	1.88	.61	1.49	.69	1.41
Kurtosis	1.93	11.20	−.01	.73	2.77	5.45	−1.02	2.28	−.35	1.67

Note. ADHD = attention deficit hyperactivity disorder; CD = conduct disorder; ODD = oppositional defiant disorder; MDD = major depressive disorder; OCD = obsessive compulsive disorder; PD = panic disorder; Soc Phob = social phobia; Spec Phob = specific phobia; GAD = generalized anxiety disorder; Sep Anx = separation anxiety; SD = standard deviation;

^**Correlation is significant at the 0.01 level (2-tailed).

Adolescent Factor Structure of Psychopathology

Confirmatory factor analysis addressed the first aim of the study, testing competing models of the factor structure of adolescent psychopathology. Fit statistics and standardized factor loadings for the correlated factors, bifactor, and general factor models are presented in Table II. The correlated factors model, arguably the most frequently modeled dimensional structure of psychopathology in the literature, demonstrated inconsistently acceptable fit: χ² (34) = 285.969, TLI = 0.878, RMSEA = 0.087, SRMR = 0.051, BIC = 37568.843. Loadings for the EXT and INT factors were all positive and high, ranging from 0.647 to 0.799 for the EXT factor and from 0.501 to .770 for the INT factor. The factors were highly correlated at r = 0.751, p < .001.

Table II.

Factor Structure of Adolescent Psychopathology: Model Fit Statistics, Standardized Factor Loadings, and Factor Correlations

	Correlated factors model				Bifactor model					General Factor model
Statistics, loadings, and correlations	Model fit	EXT	INT		Model fit	P	EXT	INT		Model fit	GEN
Statistic
Chi-square	285.969				88.947					461.011
df	34				25					35
TLI	.878				.958					.800
SRMR	.051				.023					.067
BIC	37568.843				37374.245					37785.233
RMSEA[90% CI]	.087				.051					.111
	[.077, .096]				[.040, .063]					[.102, .120]
Standardized Factor loading
Attention Deficit Hyperactivity		.799				.713	.266				.716
Conduct		.647				.492	.521				.521
Oppositional Defiant		.727				.588	.516				.599
Generalized Anxiety			.761			.657		.367			.731
Obsessive Compulsive			.684			.614		.277			.670
Panic			.501			.431		.242			.486
Separation Anxiety			.700			.538		.560			.658
Social Phobia			.544			.396		.444			.509
Specific Phobia			.545			.399		.464			.519
Major Depressive			.770			.878		.014			.801
Factor correlation		.751

Note. EXT = externalizing factor; INT = internalizing factor; P = P-factor; GEN = general factor; df= degrees of freedom; TLI = Tucker-Lewis index; SRMR = standardized root mean square residual; BIC = Bayesian information criterion; RMSEA = root-mean-square error of approximation; CI = confidence interval.

In contrast, the bifactor model consistently met criteria for good fit: χ² (25) = 88.947, TLI = 0.958, RMSEA = 0.051, SRMR = 0.023, BIC = 37374.245. Loadings for the P-factor were all significant and moderate to high, ranging from 0.396 to 0.878. Loadings for the residual EXT factor were all significant and moderate, ranging from 0.266 to 0.521. All but one loading for the residual INT factor were significant and moderate. The INT factor loading onto major depressive symptoms was near zero and nonsignificant, as most of the variance was subsumed by the P-factor. The residual externalizing and internalizing factors were modeled as orthogonal, but extracted factor scores were negatively correlated at r = −0.201, p < .01.

Finally, the general factor model was tested to determine whether the subordinate specific externalizing and internalizing factors were necessary when modeling adolescent psychopathology. This model demonstrated the worst fit of the three: χ ² (35) = 461.011, TLI = 0.800, RMSEA = 0.111, SRMR = 0.067, BIC = 37785.233.

Parent-Offspring Transmission Specificity Models

Maternal and paternal alcohol use disorders were positively and significantly correlated (r = 0.24, p < .01; Table III). The only significant correlation between maternal alcohol use disorder and the offspring latent factor scores of psychopathology was that with the bifactor model residual EXT factor score (r = 0.12, p < .05). Paternal alcohol use disorder was significantly correlated with the correlated factors model EXT and INT factor scores (r = 0.16, p < .05 and r = 0.14, p < .05, respectively), the p-factor score (r = 0.12, p < .05), and the general factor score (r = 0.15, p < .05).

Table III.

Parent-offspring Correlations, Twin Intraclass Correlations, and Descriptive Statistics for Latent Factor Scores of Psychopathology

	1.	2.	3.	4.	5.	6.	7.	8.
1. Maternal AUD	-
2. Paternal AUD	.24**	-
3. Offspring EXT	.04	.16*	-
4. Offspring INT	−.01	.14*	.85**	-
5. Offspring P-factor	.00	.12*	.90**	.94**	-
6. Offspring Residual EXT	.12*	.08	.54**	.10**	.16**	-
7. Offspring Residual INT	−.03	.08	.14**	.50**	.17**	−.20**	-
8. Offspring General Factor	.00	.15*	.91**	.99**	.97**	.21**	.39**	-
MZ ICC			.49	.47	.47	.54	.43	.48
DZ ICC			.34	.38	.34	.22	.16	.38
N Dx	133/410	160/294
Mean			.00	.00	.00	.00	.00	.00
Standard Deviation			1.81	2.85	3.23	1.02	.67	2.99
Minimum			−2.98	−4.20	−4.76	−3.24	−2.41	−4.33
Maximum			6.82	11.10	12.96	3.84	3.02	11.42
Skewness			.95	.87	.97	.58	.84	.86
Kurtosis			.54	.26	.43	.76	1.47	.23

Note. AUD = alcohol use disorder; EXT = externalizing factor; INT = internalizing factor; MZ ICC = monozygotic twin intraclass correlation; DZ ICC = dizygotic twin intraclass correlation; N Dx= number of parents diagnosed with Alcohol Use Disorder out of total sample; Offspring P, Residual INT, and Residual EXT factor scores extracted from bifactor model;

^**Correlation is significant at the 0.01 level (2-tailed);

^*Correlation is significant at the 0.05 level (2-tailed).

Next, multivariate regression analyses using the “type=complex” command to control for twin interdependence were used to test associations between parental alcohol use disorder and adolescent latent factor scores of psychopathology. Predictors included maternal and paternal alcohol use disorder, socioeconomic status, and ethnicity (offspring age and gender were controlled for prior to confirmatory factor analyses). Three separate models were tested in which outcomes were 1) EXT and INT factor scores derived from the correlated factors model, 2) P-, residual EXT, and residual INT factor scores derived from the bifactor model, and 3) factor scores derived from the general factor model. Paternal alcohol use disorder was associated with higher factor scores for both EXT and INT derived from the correlated factors model (ß = .141, SE = .049, p < .01 and ß =.138, SE = .049, p < .01, respectively). Higher socioeconomic status was associated with lower EXT factor scores (ß = −.115, SE = .053, p < .05). Paternal alcohol use was also associated with higher P-factor scores (ß = .123, SE = .050, p < .05), but not residual EXT and INT factor scores derived from the bifactor model (ß = .067, SE = .047, p = .156 and ß =.080, SE = .048, p = .092, respectively). Higher socioeconomic status was associated with lower residual INT factor scores (ß = −.085, SE = .043, p < .05). Finally, paternal alcohol use was associated with higher factor scores from the general factor model (ß = .139, SE = .049, p < .01). No significant associations occurred between maternal alcohol use and the adolescent latent factors scores of psychopathology.

Twin Biometric Models of Adolescent Psychopathology Factors

Twin intraclass correlations for the latent factor scores of psychopathology are presented in Table III. MZ twins were more similar than DZ twins on all latent factor scores, suggesting additive genetic influence and nonadditive genetic influence in the case of residual EXT and INT factors scores where MZ twin intraclass correlations were more than double those of DZ twins. In addition, MZ correlations were less than 1.00 across all latent factor scores, suggesting nonshared environmental influences.

Saturated models were fit to test for twin order, rater contrast, and assimilation effects. Fully saturated multigroup models freely estimating means and variances for MZ and DZ twins were tested against a series of models constraining means and variances to be equal across twin order and zygosity. Higher DZ vs. MZ twin variance suggests that reporters may be inflating DZ twin differences (contrast effects) or MZ twin similarities (assimilation effects), whereas lower DZ twin variance suggests possible sibling cooperation or imitation effects (Neale & Cardon, 1992). Means and variances could be equated across twin order and zygosity for all latent factor scores of psychopathology, indicating no evidence of twin order, rater contrast, or assimilation effects.

Biometric estimates were then obtained from structural equation model fitting. Standardized A, C (or D), and E squared parameter estimates for the latent factor scores of psychopathology and fit statistics for the full and best fitting reduced univariate models are given in Table IV. For all latent factor scores of psychopathology, the best fitting model was the reduced AE model. Heritability estimates were moderate in magnitude, ranging from .44 to .59.

Table IV.

Univariate ACE/ADE Model Fit and Parameter Estimates

Phenotype	Model	−2LL	df	Δ−2LL	Δdf	p	AIC	A	C/D	E
EXT Factor	ACE	3823.68	973	-	-	-	1877.68	.51 (.27–.82)	.06 (.00–.46)	.43 (.35–.53)
	AE	3823.97	974	.29	1	.59	1875.97	.57 (.46–.70)	-	.42 (.35–.51)
	CE	3835.57	974	11.90	1	< .001	1887.57
	E	3912.64	975	88.96	2	< .001	1962.64
INT Factor	ACE	4702.23	973	-	-	-	2756.23	.42 (.20–.74)	.13 (.01–.42)	.45 (.35–.55)
	AE	4703.78	974	1.54	1	.21	2755.78	.58 (.47–.70)	-	.42 (.35–.51)
	CE	4710.59	974	8.36	1	< .001	2762.59
	E	4795.23	975	93.00	2	< .001	2845.23
P-factor	ACE	4946.82	973	-	-	-	3000.82	.48 (.24–.79)	.07 (.00–.43)	.45 (.36–.55)
	AE	4947.30	974	.48	1	.49	2999.30	.57 (.45–.69)	-	.43 (.35–.52)
	CE	4957.13	974	10.31	1	< .001	3009.13
	E	5032.99	975	86.18	2	< .001	3082.99
Residual EXT Factor	ADE	2723.73	973	-	-	-	777.73	.28 (.01–.88)	.27 (.01–.92)	.45 (.36–.68)
	AE	2725.09	974	1.36	1	.24	777.09	.53 (.41–.65)	-	.47 (.39–.57)
	E	2797.86	975	74.14	2	< .001	847.86
Residual INT Factor	ADE	1921.01	973	-	-	-	−24.99	.13 (.05–91)	.35 (.04–.98)	.52 (.41–.63)
	AE	1923.17	974	2.16	1	.14	−24.83	.44 (.31–.58)	-	.56 (.46–.68)
	E	1965.73	975	44.72	2	< .001	15.73
General Factor	ACE	4791.53	973	-	-	-	2845.53	.43 (.21–.74)	.13 (.01–.41)	.44 (.35–.54)
	AE	4793.11	974	1.59	1	.21	2845.11	.59 (.48–.71)	-	.41 (.34–.50)
	CE	4800.40	974	8.88	1	< .001	2852.40
	E	4887.96	975	96.43	2	< .001	2937.96

Note. A, C, D, and E are standardized squared parameter estimates for additive genetic, shared environmental, nonadditive genetic, and nonshared environmental factors, respectively. Standardized confidence intervals are reported in parentheses. The most parsimonious final model is indicated in bold. EXT = externalizing; INT = internalizing; −2LL = − 2 log likelihood; df = degrees of freedom; Δ = change; AIC = Akaike’s information criterion.

Fit statistics for the full and best fitting reduced bivariate model of EXT and INT factor scores derived from the correlated factors model, as well as standardized A, C, and E squared parameter estimates are given in Table V. The best fitting model was an AE-AE model. 59% of the covariance between EXT and INT was explained by A and 41% by E. The A1 factor explained 57% of the variance in EXT and 44% of the variance in INT. An additional 14% of the variance in INT was explained by the specific, A2 factor. The E1 factor explained 43% of the variance in EXT and 27% of the variance in INT. An additional 15% of the variance in INT was explained by the specific, E2 factor. Genetic and environmental correlations showed that additive genetic influences on EXT and INT were correlated at 0.87, and nonshared environmental influences were correlated at 0.80.

Table V.

Bivariate Model Fit and Parameter Estimates for the Correlated Factors Model

	Model	−2LL	df	Δ−2LL	Δdf	p	AIC
EXT - INT	Bivariate Cholesky (full)	7353.151	1943	-	-	-	3467.15
	Bivariate Cholesky (final)	7354.85	1946	1.70	3	.64	3462.85
	Phenotype	A1	C1	E1	A2	C2	E2
Full Model	EXT	.44 (.24–.72)	.11 (.00–.36)	.45 (.36–.55)	-	-	-
	INT	.29 (.09–.50)	.14 (.01–.42)	.29 (.21–.39)	.13 (.08–.19)	.00 (.00–.03)	.15 (.12–.18)
Final Model	EXT	.57 (.46–.70)	-	.43 (.35–.51)	-	-	-
	INT	.44 (.35–.56)	-	.27 (.20–.35)	.14 (.10–.18)	-	.15 (.12–.18)

Note. Standardized squared parameter estimates for A, C, and E are reported. Standardized confidence intervals are reported in parentheses. The most parsimonious final model is indicated in bold. −2LL = − 2 log likelihood; df = degrees of freedom; Δ = change; AIC = Akaike’s information criterion.

Discussion

We aimed to elucidate externalizing symptom presentation in adolescence and contextualize it within competing factor structures of psychopathology, examine whether parental alcohol use disorder confers a specific risk for adolescent externalizing symptomology or a more general risk to psychopathology, and determine the degree of genetic and environmental influence on adolescent externalizing problems, shared etiology with internalizing problems, and a broader, hierarchical dimension of psychopathology capturing co-occurrence across externalizing and internalizing domains. In contrast to the correlated factors and general factor models, the bifactor structure exhibited the best fit to the adolescent symptom count data, evidencing important co-occurrence across externalizing and internalizing symptomology while substantiating a need to still capture the residual variance specific to the broadband domains. Next, paternal alcohol use disorder predicted both the externalizing and internalizing factor scores derived from the correlated factors model, the adolescent P-factor but not the residual externalizing and internalizing factor scores derived from the bifactor model, and the general factor scores derived from the general factor model. There were no significant associations with maternal alcohol use disorder. Lastly, the latent factors of adolescent psychopathology were all moderately heritable (h²= .44–.59) and subject to additional nonshared environmental influence. Associations between externalizing and internalizing factor scores derived from the correlated factors model were primarily explained by shared genetic influence. These findings indicate that paternal alcohol use disorder confers a general risk for adolescent psychopathology. Furthermore, variation in adolescent general psychopathology is primarily influenced by genetic factors.

These analyses are a distinct contribution from the Wisconsin Twin Project (Schmidt et al., 2019), which has produced related papers with samples that overlap ours. Among the related papers are those that investigate the genetics of psychopathology at an earlier, childhood age (Moore et al., in press; Vendlinski et al., 2014). In adolescence, previous twin analyses have examined the co-occurrence of symptoms from specific pairs of psychopathology domains, such as ADHD and anxiety (Brooker et al., 2020), and sensory over-responsivity and obsessive-compulsive disorder (Van Hulle et al., 2019). Here, we take a much more comprehensive approach to this issue of co-occurrence. Finally, two prior papers have examined parent-to-offspring transmission issues (Van Hulle et al., 2018; Oro et al., 2019), but this paper is the initial examination of parental substance use in relation to adolescent symptoms. We contextualize these findings within the broader literature and present strengths, limitations, and future directions.

Adolescent Factor Structure of Psychopathology

The bifactor structure exhibited the best fit to the adolescent symptom count data, corroborating our hypothesis and replicating previous findings from studies testing the bifactor model with subordinate specific externalizing and internalizing factors in similarly aged samples using questionnaire assessments (Castellanos-Ryan et al., 2016; Lahey et al., 2015; Snyder et al., 2017; Tackett et al., 2013; Waldman et al., 2016). Depression and ADHD exhibited the highest loadings on the P-factor which is generally consistent with the reviewed literature. The general factor, P-factor, and externalizing and internalizing factors from the correlated factors model were all very highly correlated (above .90). This high degree of overlap suggests that many features of externalizing psychopathology are, in fact, transdiagnostic and shared with internalizing. The bifactor model captures the important co-occurrence across externalizing and internalizing symptomology that the popular correlated factors model inadequately addresses. Nevertheless, important broadband domain-specific variance remains; in fact, the bifactor structure indicates significant residual variance unaccounted for by the P-factor which must be modeled. More research examining bifactor loadings at the item level is needed to ascertain which symptoms are indicators of general versus more specific dimensions. For example, bifactor analyses using item level symptom data indicate that somatization problems rather consistently load higher on the residual internalizing factor than the P-factor (McElroy et al., 2018). Broadly, these findings compel us to consider that which is in common across presumably distinct domains as clinically meaningful rather than nuisance. Our results bolster the rationale for examining intergenerational transmission from parental alcohol use disorder to adolescent psychopathology at a broader level.

The recent proliferation of bifactor modeling of psychopathology has garnered criticism (Bonifay et al., 2017). The major contention is interpretability; specifically, latent factor definitions. Critics appropriately warn against the premature misattribution of the P-factor as emerging from some speculative, unitary cause. Various hypotheses have been posited, including that the P-factor captures “disordered form and content of thought” (Caspi & Moffit, 2018), though more research is needed before such an interpretation can be responsibly extended for clinical use. However, this caution does not discount other valid applications of the bifactor model, including testing questions of specificity such as in the current study.

Despite these unknowns, genomics research has compared genetic correlations across major psychiatric disorders and found support for a genetic P-factor, indicating that the P-factor is not simply a statistical artifact; rather, meaningful, genetically influenced mechanisms underlie symptom co-occurrence (Selzam et al., 2018). Moreover, genome-wide association studies have compiled atlases of genetic associations to synthesize data for millions of genetic variants in a manner that effectively represents high degrees of polygenicity and pervasive pleiotropic effects. Findings indicated that tens of thousands of the genetic variants tested were associated with more than one trait (Canela-Xandri et al., 2018). This high degree of pleiotropy indicates that genetic influences on human complex traits, including psychopathology, are nonspecific. Recently, polygenic scores for a general factor of psychopathology in adolescence have been derived and examined for associations with symptoms trajectories, including that of depression (Li et al., 2020). These findings should be applied clinically in the implementation of a generalist genes/specialist environments framework that informs interventions aimed at mitigating risk for the development of psychopathology in adolescence.

An additional concern is that the bifactor model has a tendency to over-fit data, with critics warning against the blind selection of models as a function of best fit (Murray & Johnson, 2013). Rather, a theoretical approach should always inform model selection and, for the current study, the bifactor model provides the ideal structure to test questions of transmission specificity. Finally, concerns regarding the validity of the P-factor have arisen. This is a burgeoning area of research, and these concerns are likely to be quelled as new associations emerge in the literature. For example, the current study provides external validation of the adolescent bifactor structure by demonstrating associations with paternal alcohol use disorder.

Associations between Parental Alcohol Use Disorder and Adolescent Psychopathology

Phenotypic models testing associations between parental alcohol use disorder and offspring latent factor scores of psychopathology indicated significant paternal but not maternal effects operating at a nonspecific level, partially corroborating our hypotheses. Paternal alcohol use disorder predicted both the externalizing and internalizing factor scores derived from the correlated factors model, the adolescent P-factor but not the residual externalizing and internalizing factor scores derived from the bifactor model, and the general factor scores derived from the general factor model. Results align with epidemiological research with limited measurement indicating that parental psychopathology is a non-specific risk factor for offspring psychopathology (McLaughlin et al., 2012). These findings suggest that interventions designed to prevent the development of psychopathology in offspring of parents exhibiting alcohol use disorder necessitate broader approaches that address the full range of child symptomology.

Research examining differential magnitudes of effect across maternal and paternal alcoholism has been mixed. Some studies indicate that paternal alcoholism is more detrimental to children (Loukas et al., 2001; Ohennessian, 2013), while others indicate that maternal alcoholism is more detrimental (Rognmo et al., 2012). Moreover, some research indicates that child gender plays a role, with stronger effects of parental alcoholism shown for same-sex parent-child dyads (Ohennessian, 2012). More research is needed to determine whether differential effects of alcohol use disorder across parent gender exist and whether these effects are also impacted by child gender. Moreover, future studies should elucidate factors that potentially amplify or mitigate maternal and/or paternal effects of alcohol use disorder. For example, Homish and colleagues (2006) found that husbands’ alcohol problems were associated with wives’ depressive symptoms but found no associations between wives’ alcohol problems and husbands’ depressive symptomology. It may be that children of fathers exhibiting alcohol use disorder are susceptible to compounded negative effects as a result of their mother’s own psychological distress, whereas fathers whose wives exhibit alcohol use disorder provide care that buffers against the negative effects of maternal alcohol use problems. A family systems approach is needed to clarify these dynamics and better understand familial contexts that confer greatest risk to vulnerable children.

Genetic and Environmental Liabilities to Psychopathology in Adolescence

Twin biometric models of the latent factor scores of adolescent psychopathology indicated that all factors were moderately heritable, with estimates ranging from .44 to .59. Specifically, 57% of the variance in P-factor scores indexing common variance across externalizing and internalizing domains was explained by genetic factors. Extant twin research also reports the P-factor to be moderately heritable in adolescence (Allegrini et al., 2020; Waldman et al., 2016). Furthermore, we found that covariance between externalizing and internalizing factor scores derived from the correlated factors model was primarily explained by shared genetic influence as well as nonshared environmental influence. Importantly, a majority of the genetic influence on externalizing and internalizing psychopathology was shared across the two domains. Extant behavioral genetic research examining the shared etiology across externalizing and internalizing domains in adolescence also indicates that the domains are moderately heritable and influenced by common genetic and nonshared environmental influences (Cosgrove et al., 2011). These findings suggest that researchers in behavior genetics should moderate the language they employ when disseminating findings on the genetic influence of externalizing psychopathology and implications regarding specificity, as these influences underlie a host of other related constructs that may have more central or more proximal influences on outcomes. Future research should attempt to identify phenotypes that are more proximal to genetic risk and linked to both externalizing and internalizing psychopathology. Identifying transdiagnostic features that have more proximity to the genotype stands to further elucidate mechanisms of co-occurrence and garner meaningful clinical information that can be harnessed to effectively mitigate risk for psychopathology.

Importantly, a sociocultural lens is also needed to enhance our understanding of the etiology of adolescent psychopathology and the mechanisms through which risk is conferred from parents to offspring. Studies examining the effects of parental alcohol use problems on adolescent offspring psychopathology should intentionally consider the role of sociocultural influences, as parent-offspring associations may vary across racial/ethnic groups as a function of cultural practices; however, little research takes this approach. Slightly more is known regarding the genetic etiology of adolescent externalizing behavior and psychopathology in diverse groups, yet this research is scant and in a nascent state. For example, Elam and colleagues (2018) found that genetic risk for aggression and maternal substance use were associated with poorer family cohesion in adolescence for both Mexican-American and European-American families, yet family cohesion served as a greater buffer against later alcohol use for Mexican-American adolescents. These findings point to the role of an important Latino cultural value, familismo, in strengthening family ties and fostering cohesion that protects against negative outcomes. More studies like this are needed to bridge the divide between genetically- and culturally-informed research, to elucidate pathways of intergenerational transmission, and to identify etiological mechanisms underlying psychopathology in adolescence in just and equitable ways that advance science for all.

Strengths, Limitations, and Future Directions

We utilized rigorous methods to elucidate the factor structure of adolescent psychopathology, pathways of intergenerational transmission from parental alcohol use disorder at various levels of specificity, and the genetic architecture underlying latent dimensions of psychopathology in adolescence. Strengths of the study include a consideration of both maternal and paternal data within a kinship framework, as research has tended to neglect the role of fathers in the development of child psychopathology (Cassano et al., 2006). Additionally, interview-based diagnostic assessment of psychopathology is a formidable strength of this study, particularly within the genetic literature where rich phenotypic measurement is often not the standard. This method avoids the confound of common method variance often encountered in parent-offspring research when reports are acquired from a single informant.

Limitations of the study should also be acknowledged. Generalizability of the findings is a primary concern. First, the sample is predominantly Caucasian, limiting our understanding of whether these findings extend to other ethnic groups. We also examined the effect of parental alcohol use disorder on offspring psychopathology in a community sample when transmission effects may operate differentially as a function of symptom severity. However, this limitation does not negate the importance of examining these intergenerational associations at sub-clinical levels. Additionally, rates of parental AUD in the current sample (just over half of participating fathers and nearly one third of mothers) are higher than national prevalence rates where 7.6% and 4.1% of adult males and females endorse AUD, respectively (SAMSHA, 2019). These rates likely reflect the drinking culture in the state, where Wisconsinites consume alcohol and binge drink at higher rates than the U.S. median yet perceive less risk incurred from said drinking (CDC, 2019; Wisconsin Department of Health Services, 2019), potentially resulting in less biased disclosure of drinking behavior in the current sample. Lastly, the generalizability of twin study results of psychopathology to singleton populations has been questioned. Overall, disorder-specific rates appear to be generally similar for twins and singletons (Kendler et al., 1995), and twin vs. singleton differences in rates of co-occurrence in symptoms seem unlikely.

An additional limitation is the cross-sectional design of the study, precluding the examination of continuity and change in patterns of co-occurrence across externalizing and internalizing psychopathology during adolescence, in associations with parental alcohol use disorder over time, and in etiology. Phenotypic evidence points to strong homotypic continuity in P- and residual externalizing and internalizing factors across adolescence, suggesting that findings from the current study may hold over time (Snyder et al., 2017). Furthermore, twin studies of psychopathology point to an increase in variance explained by genetic factors across development (Plomin et al., 2016). Longitudinal twin studies are needed to examine the extent to which genetic and/or environmental influences mediate stability in co-occurring symptomology across adolescence and to determine whether heritability is amplified with time.

Overall, this study makes a significant contribution to the literature examining specificity of risk transmitted from parents exhibiting alcohol use disorder to their adolescent offspring and provides novel insight into the etiology underlying symptom co-occurrence across externalizing and internalizing domains in adolescence. Findings illuminate the breadth of influence paternal alcohol use disorder has on offspring psychopathology during a developmental period when susceptibility is amplified, informing future prevention and intervention efforts aimed at disrupting persistent and detrimental intergenerational cycles.

Data Availability of Data and Material

Data are available from the first author upon request.