Commentary: Sex Differences in the Pathways to Symptoms of Alcohol Use Disorder: A Study of Opposite‐Sex Twin Pairs

KENDLERANDCOLLEAGUES (2015) identify a latticework of risk and protective factors that underlie liability to alcohol problems (AP) in opposite-sex siblings. The authors examine the interplay between 18 risk and protective factors, broadly classified into 5 developmentally relevant domains relating to childhood (e.g., childhood abuse, familial risk), early adolescence (e.g., neuroticism), late adolescence (e.g., conduct disorder, lifetime traumas), adult (e.g., divorce), and last year (e.g., drinks/month). Their Figs 1 and 2, illustrating results from comprehensive and computationally intensive structural equation models, are more reminiscent of a subway grid or an anatomical model of human vasculature than the typical bar charts found in epidemiological studies. These figures outline several unique routes via which liability to AP develop in men and women. For instance, familial risk and nicotine dependence were found to have stronger links with AP in females, while conduct disorder and low self-esteem were found to exert a more pronounced influence among males. Importantly, this assortment of 18 parameters explained an impressive 71– 73% of the variance in liability to alcohol use problems (AUP). In particular, the study has 2 innovative features. First, the authors leverage the genetic and familial matching afforded by dizygotic opposite-sex sibling pairs. In traditional discordant twin analyses, such pairs are less useful because traditional exposures and outcomes studied in addiction (e.g., the association between early onset alcohol use and later alcohol dependence) are confounded with sex (Prescott and Kendler, 1999). However, Kendler and colleagues (2015) show the utility of these pairs for the study of the origins of sex differences in AP. Second, the authors allow the developmental contributors to interface with each other, estimating direct and indirect paths. For instance, familial risk is among the most robust predictors of AP, and arguably, more prognostic than any genetic variant identified to date (Yan et al., 2014). Yet, the current study shows that only about 80% of its relationship with AP is direct. The remainder, albeit modest, is attributable to its interface with lifetime trauma, neuroticism, nicotine dependence, anxiety, and conduct disorder. While the partial matching for genes identical by descent is a notable strength, it also raises the intriguing question of whether existing research is adequately addressing the putative role of sex differences in genetic analyses. The classical twin analysis directly utilizes data on opposite-sex twin pairs to examine qualitative sex differences, that is, whether the same or different genes influence heritable (and shared environmental) variation in men and women (Neale and Cardon, 1992). This is different from quantitative sex differences, or the test of whether the magnitude of heritable variation is sex invariant. For AUP, little support exists for quantitative sex differences (i.e., about 50 to 60% heritability in men and women). However, evidence of whether the genetic influences that comprise these heritability estimates overlap has been inconclusive. For instance, Prescott and colleagues (1999) found that unlike the genetic sharing of 0.5 that is expected for DZ twin pairs, the genetic correlation in opposite-sex pairs, when freely estimated, was 0.20 to 0.24. However, this qualitative sex difference was not identified by Heath and colleagues (1997). Genomic studies can also benefit from additional consideration of sex. For instance, while all genomewide association studies include the main effects of sex as a covariate, few have considered the joint analysis of the single nucleotide polymorphism (SNP) and SNP 9 Sex term. This 2 degree of freedom test has been powerful in uncovering gene–environment interactions for other environmental factors (Hancock et al., 2012). It was also employed in one of the earliest GWAS of nicotine dependence, the candidate gene component of which was one of the first studies to identify the role of rs16969968 in the etiology of smoking (Saccone et al., 2007). While the effects of rs16969968, the strongest and most replicable genetic signal for tobacco smoking to date, have been noted in men and women, other SNPs have been From the Department of Neuroscience (MK), Icahn School of Medicine at Mount Sinai, New York City, New York; and Department of Psychiatry (AA), Washington University in St. Louis, St. Louis, Missouri. Received for publicationMarch 17, 2015; acceptedMarch 23, 2015. Reprint requests: Manav Kapoor, PhD, Assistant Professor, Department of Neuroscience, 10-75B, 1425 Madison Avenue, New York, NY 10029; Tel.: 212-659-5967; Fax: 212-849-2611; E-mail: manav. kapoor@mssm.edu Copyright© 2015 by the Research Society on Alcoholism.