Childhood neglect is one of the most common forms maltreatment in the United States (Sedlak & Broadhurst, Reference Sedlak and Broadhurst1996). The Fourth National Incidence Study of Child Abuse and Neglect reported that 60% of all maltreatment cases in the United States included some form of neglect, with educational and physical neglect being the most prevalent type of neglect (Sedlak et al., Reference Sedlak, Mettenburg, Basena, Petta, McPherson and Greene2010). Research has demonstrated that childhood neglect may have serious negative consequences that can extend from childhood into adolescence and perhaps adulthood (Trickett & McBride-Chang, Reference Trickett and McBride-Chang1995). For instance, neglected youths are more likely to exhibit externalizing behaviors (i.e., defiant, disruptive) in childhood than nonneglected youths. The behavior of neglected youths is primarily characterized by social awkwardness, social isolation, poor impulse control, internalizing behaviors, and a negative view of oneself and others (Cicchetti & Toth, Reference Cicchetti and Toth1995; Hildyard & Wolfe, Reference Hildyard and Wolfe2002). Neglected youths often exhibit developmental delays in language and social skills in early and middle childhood, more so than nonneglected youths and youths who experience other forms of maltreatment. These developmental problems often continue into late childhood and early adolescence (Hildyard & Wolfe, Reference Hildyard and Wolfe2002).
Although there a large body of literature linking neglect to behavioral problems in childhood and early adolescence, there is less research that examines whether neglect is associated with antisocial and maladaptive behaviors in adulthood. Recently, researchers have begun to examine the long-term effects of neglect, and these studies show that individuals who were neglected during childhood are at risk for engaging in criminal behavior and substance abuse later in life (Dube et al., Reference Dube, Felitti, Dong, Chapman, Giles and Anda2003; Widom, Reference Widom1989b). For instance, Widom (Reference Widom1989a) found that neglected and abused individuals were more likely to be arrested during adolescence and adulthood than nonneglected/abused individuals. A later study by Widom and Ames (Reference Widom and Ames1994) revealed that the prevalence of adulthood arrests for property offenses, order offenses, and drug offenses were higher among abused and neglected youths relative to nonmaltreated youths. Other studies have revealed that neglected individuals have a higher risk of developing drug and alcohol problems during adulthood than nonneglected individuals (McCord, Reference McCord1983; Widom, Ireland, & Glynn, Reference Widom, Ireland and Glynn1995; Widom, White, Czaja, & Marmorstein, Reference Widom, White, Czaja and Marmorstein2007). These findings suggest that neglect may place individuals at risk for antisocial behavior beginning in adolescence, and this risk may extend into adulthood.
Although studies have shown that individuals subjected to neglect or abuse as children are more likely to engage in crime and substance abuse, studies indicate that a large percentage of these youth are resilient to the effects of such experiences (Cicchetti, Rogosch, Lynch, & Holt Reference Cicchetti, Rogosch, Lynch and Holt1993; McGloin & Widom, Reference McGloin and Widom2002). In an analysis of 2019 runaway youth, Kurtz, Kurtz, and Jarvis (Reference Kurtz, Kurtz and Jarvis1991) found that only 27.7% of runaway youths were abused during childhood. Similarly, Widom (Reference Widom1989a) found that abused children were arrested more frequently as juveniles and adults, but more than 70% had never been arrested. A follow-up study by Widom and colleagues found that approximately 50% of abused youths demonstrated resilience during adolescence, and approximately 30% of abused youths demonstrated resilience during adulthood (DuMont, Widom, & Czaja, Reference DuMont, Widom and Czaja2007). Resilience was broadly defined as graduating from high school, employment, participating in social activities, never being homeless, refraining from violent behavior, lacking a psychiatric diagnosis, and lacking a substance use or dependence diagnosis. These findings, along with results from other studies, suggest that a substantial percentage of neglected and abused individuals do not go on to engage in antisocial and maladaptive behaviors (Collishaw et al., Reference Collinshaw, Pickles, Messer, Rutter, Shearer and Maughan2007). Thus, there may be some characteristics that increase or decrease one's vulnerability to the effects of abuse and neglect.
There are a number of theoretical reasons to suspect that some individuals may be more sensitive to the effects of neglect than other individuals. From a psychological point of view, the diathesis stress model and specific vulnerability hypothesis argue that stressors are more likely to lead to maladaptive behaviors (i.e., depression, schizophrenia, bipolar disorder) among individuals who have a predisposition toward that maladaptive behavior (Hammen, Ellicott, Gitlin, & Jamison, Reference Hammen, Ellicott, Gitlin and Jamison1989; Monroe & Simons, Reference Monroe and Simons1991). A similar hypothesis is garnered from Agnew's general strain theory (Agnew, Reference Agnew1992). General strain theory posits that stressors, such as physical and emotional neglect, are more likely to lead to criminal behavior when individuals are impulsive or have an inclination toward antisocial behavior. Further, Agnew argues that the stability and continuity of criminal behavior will be greatest among individuals who are predisposed toward antisocial behavior and who experience stressful life events (i.e., neglect, abuse; Agnew, Reference Agnew and Thornberry1997). Similarly, Moffitt's (Reference Moffitt1993) taxonomy of criminal offenders also notes that biological/genetic risk factors and environmental risk factors interact to predict an early onset of antisocial behavior and continuity of antisocial behavior into adulthood. Together, these perspectives suggest that (a) individuals who have a biological or genetic propensity toward antisocial behavior are more sensitive to the criminogenic effects of neglect than individuals who do not have a propensity toward antisocial behavior (i.e., Genetic/Biological × Neglect interaction) and (b) these individuals may be at risk for offending across the life course.
Recent research has begun to examine genetic risk factors as a possible explanation for variation in criminal and substance abuse outcomes in abused youth. For example, a landmark study by Caspi et al. (Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002) found that the monoamine oxidase A (MAOA) polymorphism interacted with childhood maltreatment to predict severe antisocial behavior in adolescence and adulthood in a sample of male subjects (Caspi et al., Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002). In addition, Jaffee and colleagues (Reference Jaffee, Caspi, Moffitt, Dodge, Rutter and Taylor2005) found that genetic risk moderated the effects of physical maltreatment on conduct problems. Their results revealed that the probability of being diagnosed with conduct disorder being highest among participants who had the highest genetic risk and had experienced maltreatment. Although several studies suggest that genetic factors may moderate the effects of victimization on antisocial behavior (Brendgen et al., Reference Brendgen, Boivin, Vitaro, Girard, Dionne and Pérusse2008; Foley et al., Reference Foley, Eaves, Wormley, Silberg, Maes and Kuhn2004; Haberstick et al., Reference Haberstick, Lessem, Hopfer, Smolen, Ehringer and Timberlake2005), the available research examining these factors has generally focused on a polymorphism in the MAOA gene or a broad measure of genetic risk.
Serotonergic polymorphisms may also moderate the effects of neglect on antisocial behavior and substance abuse. A large body of literature has examined the interaction of the serotonin transporter linked polymorphic region (5-HTTLPR) polymorphism and stressful life events to predict depression and other maladaptive behaviors (Caspi et al., Reference Caspi, Sugden, Moffitt, Taylor, Craig and Harrington2003; Eley et al., Reference Eley, Sugden, Corsico, Gregory, Sham and McGuffin2004; Kaufman et al., Reference Kaufman, Yang, Douglas-Palumberi, Houshyar, Lipschitz and Krystal2004; Taylor et al., Reference Taylor, Way, Welch, Hilmert, Lehman and Eisenberger2006). For instance, Scheid et al. (1997) found that the rates of depression were highest among women who carried the short allele and who had experienced violent or sexual victimization. Cicchetti, Rogosch, and Struge-Apple (Reference Cicchetti, Rogosch and Sturge-Apple2007) compared 207 maltreated youth and 132 youth with no history of maltreatment and found that youth that had encountered sexual abuse and possessed a variant of 5-HTTLPR displayed greater anxiety, depression, and somatic symptoms. A recent study has revealed that maltreated youths who carried a copy of the short allele were more likely to abuse alcohol than maltreated youths who did not carry the short allele of 5-HTTLPR (Kaufman et al., Reference Kaufman, Yang, Douglas-Palumberi, Crouse-Artus, Lipschitz and Krystal2007). Data from nonhuman female primates have also shown that maternally neglected youth who carry a short allele exhibited a greater preference for alcohol, and they consumed more alcohol than neglected youths who did not carry a short allele (Barr, Newman, Lindell, et al., Reference Barr, Newman, Lindell, Shannon, Champoux and Lesch2004). These findings suggest that 5-HTTLPR interacts with maltreatment to influence substance use and other maladaptive behaviors.
Other studies, however, have not found that 5-HTTLPR moderates the effects of stressful life events on maladaptive behaviors (Gillespie, Whitfield, Williams, Heath, & Martin, Reference Gillespie, Whitfield, Williams, Heath and Martin2005; Surtees et al., Reference Surtees, Wainwright, Willis-Owen, Luben, Day and Flint2006). A recent meta-analysis of 14 studies did not find a statistically significant interaction between 5-HTTLPR and stressful life events on depression (Risch et al., Reference Risch, Herrell, Lehner, Liang, Eaves and Hoh2009). Similarly, Munafò, Durrant, Lewis, and Flint's (Reference Munafò, Durrant, Lewis and Flint2009) meta-analysis of five studies revealed that the interaction between 5-HTTLPR and stressful life events on depression was nonsignificant. The authors further noted that many of the gene–environment studies were underpowered, and that significant interaction effects between 5-HTTLPR and environmental risk factors may be due to a small percentage of cases that are exposed to environmental risk. In light of the meta-analysis results and the limited amount of research on 5-HTTLPR × Victimization/Abuse and antisocial behavior, it remains unclear whether 5-HTTLPR conditions the effect of neglect on criminal behavior and substance use.
The mechanisms underlying the interaction between 5-HTTLPR and maltreatment remain open to empirical evaluation. There is reason to believe, however, that the 5-HTTLPR × Maltreatment interaction influences functioning of the hypothalamic–pituitary–adrenal (HPA) axis and the serotonin system, which subsequently increase individual's risk for substance use and antisocial behavior. Studies of nonhuman primates have shown that abused youths who carried a short allele had a larger increase in adrenocorticotropic hormone (ACTH) and cortisol in response to stressors than abused youths who did not carry a short allele (Barr, Newman, Lindell, et al., Reference Barr, Newman, Lindell, Shannon, Champoux and Lesch2004; McCormack, Newman, Higley, Maestripieri, & Sanchez, Reference McCormack, Newman, Higley, Maestripieri and Sanchez2009). There is also evidence that maltreated subjects who carry the short allele may also display lower levels of serotonin activity than maltreated subjects who are homozygous for the long allele. Bennett et al.'s (Reference Bennett, Lesch, Heils, Long, Lorenz and Shoaf2002) analysis of 132 rhesus macaques revealed that subjects who carried a short allele and who experienced maternal separation had lower cerebrospinal fluid 5-hydroxyindoleacetic acid concentrations than separated subjects who did not carry a short allele. Lower serotonin activity resulting from the interaction of 5-HTTLPR and maltreatment is somewhat expected given the animal literature that (a) maltreatment is associated with lower levels of serotonin (Champagne & Curley, Reference Champagne and Curley2009; Ichise et al., Reference Ichise, Vines, Gura, Anderson, Suomi and Higley2006; Maestripieri, Reference Maestripieri, Pfaff, Kordon, Chanson and Christen2005; Matthews, Dalley, Matthews, Tsai, & Robbins, Reference Matthews, Dalley, Matthews, Tsai and Robbins2001) and (b) the short allele of 5-HTTLPR is associated with lower levels of overall serotonin activity (Williams et al., Reference Williams, Marchuk, Gadde, Barefoot, Grichnik and Helms2001), by way of the short allele leading to fewer serotonin transporters, less reuptake of serotonin from the synaptic cleft, and the subsequent downregulation of postsynaptic receptors (David et al., Reference David, Murthy, Rabiner, Munafo, Johnstone and Jacob2005; Lesch et al., Reference Lesch, Bengel, Heils, Sabol, Greenberg and Petri1996). The lack of inhibition that may arise from lower serotonin activity, coupled with a greater sensitivity to stress from a hyperactive HPA axis, may increase the likelihood that maltreated individuals who carry the short allele will engage in disruptive and maladaptive behaviors (i.e., substance use, depression; Andersen & Teicher, Reference Andersen and Teicher2009). Although this hypothesis seems plausible, there is some evidence that the interaction between 5-HTTLPR and maltreatment does not affect HPA axis activity and serotonin transporter activity in nonhuman primates (Kinnally, Lyons, Abel, Mendoza, & Capitanio, Reference Kinnally, Lyons, Abel, Mendoza and Capitanio2008; Kinnally et al., Reference Kinnally, Tarara, Mason, Mendoza, Abel and Lyons2010). Further, the majority of research on the neurobiological implications of 5-HTTLPR and neglect is conducted on nonhuman primates and rodents, rather than humans; thus, the neuropsychological consequences of the 5-HTTLPR short allele and neglect among humans remains an empirical question.
Despite questions regarding the mechanisms underlying the interaction between 5-HTTLPR and maltreatment, there is evidence that 5-HTTLPR moderates the effects of abuse and neglect on antisocial behavior (Kaufman et al., Reference Kaufman, Yang, Douglas-Palumberi, Crouse-Artus, Lipschitz and Krystal2007). In light of this evidence, the current study uses data from the National Longitudinal Study of Adolescent Health (Add Health) to examine the interaction between 5-HTTLPR and neglect on criminal behavior and substance abuse. More specifically, the current study examines whether 5-HTTLPR moderates the effects of neglect on alcohol use problems, frequency of marijuana use, and criminal behavior in adulthood. The current study focuses on neglect rather than maltreatment (i.e., a combination of neglect, physical abuse, and sexual abuse) because previous research has shown that neglect may have a stronger effect on criminal behavior than physical abuse and sexual abuse (Kingree, Phan, & Thompson, Reference Kingree, Phan and Thompson2003; Widom & Ames, Reference Widom and Ames1994); thus, combining neglect, physical abuse, and sexual abuse into a single composite measure of maltreatment may result in aggregation bias and may mask important differences in the effects of each type. Further, researchers have recently criticized the aggregation of stressful life events, even events that are similar in nature, into one scale because there may be important differences in the effects of each type of event (Moffitt, Caspi, & Rutter, Reference Moffitt, Caspi and Rutter2006; Monroe & Reid, Reference Monroe and Reid2008). As a result, the current study focuses on whether 5-HTTLPR moderates the effects of neglect on substance use and criminal behavior.
It is important to note that the current study examines the effects of 5-HTTLPR and neglect on substance use and criminal behavior during adulthood. The current study focuses on antisocial behaviors in adulthood for two reasons. First, there is a limited amount of research on the relationship between neglect and adulthood antisocial behavior, and the research that is available is predominantly from Widom and colleagues (Widom, Reference Widom1989a; Widom & Ames, Reference Widom and Ames1994). Although the effects of neglect on adulthood antisocial behavior are consistent across Widom's studies, it is important to examine whether these findings are replicated across samples. Second, previous research suggests that the effects of genetic polymorphisms on antisocial behavior may be more pronounced in adulthood than adolescence (Guo, Wilhelmsen, & Hamilton, Reference Guo, Wilhelmsen and Hamilton2007). Therefore, the current study focuses on adulthood antisocial behavior rather than antisocial behavior at earlier in the life course.
The current study also examines whether the effects of 5-HTTLPR and neglect on antisocial behaviors vary by gender. Previous research has found that gender conditions the effects of Gene × Environment interactions on antisocial behavior (Brendgen et al., Reference Brendgen, Boivin, Vitaro, Girard, Dionne and Pérusse2008). For instance, Beaver (Reference Beaver2008a) found that sexually abused males who have higher levels of genetic risk are more likely to engage in violent crime than males with lower genetic risk. However, sexually abused females with higher genetic risk are not more likely to engage in violent crime. Although recent research indicates that gender may play an important role in the effects of Gene × Environment interactions, this is the first study, to our knowledge, that examines whether the interaction between 5-HTTLPR and neglect on antisocial behavior varies by gender. The current study examines two questions related to the intersection of gender, 5-HTTLPR, and neglect on antisocial behavior: (a) Are 5-HTTLPR and neglect significantly related to alcohol use problems, marijuana use, and criminal behavior for both males and females? (b) Are the independent or interactive effects of 5-HTTLPR and neglect on these outcomes significantly stronger for one gender? Thus, the first question deals with whether an association exists for each gender (i.e., are the predictors the same), whereas the second question deals with the strength of the association across genders.
Methods
Sample
Data for this study came from the Add Health study. This nationally representative, prospective study currently contains three waves of data. Wave I data were collected between 1994 and 1995, when youths were between the ages of 11 and 19. Wave I also included a genetic supplemental sample that included 2,658 twins, 208 nontwin siblings of twins, 1,611 full siblings, 1,177 half siblings, and 491 nonrelated siblings (Harriset al., Reference Harris, Florey, Tabor, Bearman, Jones and Udry2003; Touranguea & Shin, Reference Tourangeau and Shin1999). In 1996, Wave II in-home interviews were conducted with youths that participated in the Wave I in-home interview and/or were in the genetic supplemental sample. Wave III surveys were administered between 2001 and 2002, when youths were between the ages of 18 and 26. Wave III also included a DNA subsample comprising 2,574 individuals from the Wave I genetic supplemental sample that were full siblings or twins. These participants provided a saliva sample, which was used to target six polymorphisms, including a serotonin transporter polymorphism (5-HTTLPR), for genetic analysis. The final analytical sample includes 2,403 subjects (1,246 females, 1,157 males), after randomly removing one individual from each monozygotic twin pair.
Measures
Dependent variables
The current study uses three dependent variables: alcohol use problems, frequency of marijuana use, and general criminal offending in Wave III. Alcohol use problems was measured by summing eight questions about the number of times the individual experienced negative consequences from consuming alcohol during the past 12 months (0 = never, 1 = once, 2 = twice, 3 = three or four times, 4 = five or more times). The eight negative consequences included problems at school or work because of drinking, problems with friend due to drinking, problems with dating partner due to drinking, being hung over, being sick to their stomach due to drinking, getting into a physical fight because they had been drinking, and were they drunk at school or work. These scores were summed to form a composite score in which higher values indicate higher levels of abuse (α = 0.77). Previous studies have used a similar scale to measure alcohol use problems (Russell, Driscoll, & Truong, Reference Russell, Driscoll and Truong2002; Vaske, Beaver, Wright, Boisvert, & Schnupp, Reference Vaske, Beaver, Wright, Boisvert and Schnupp2009).
Frequency of marijuana use was assessed by asking subjects during Wave III the number of times they had used marijuana in the past 30 days. This is a continuous variable ranging from 0 to 30, where 30 represented use of marijuana 30 or more times.
General criminal behavior for Wave III was assessed by asking participants whether they had engaged in 13 criminal behaviors (0 = no, 1 = yes). Participants were asked whether, during the past 12 months, they had deliberately written a bad check, used someone else's credit card without their knowledge, received stolen property, stolen something worth more than $50, stolen something worth less than $50, gone into a building or house to steal something, used or threatened to use a weapon to get something from someone, taken part in a group fight, used a weapon in a fight, taken part in a fight where they were physically injured, taken part in a fight where the other party was physically injured, pulled a knife or gun on someone, or shot or stabbed someone (α = 0.70). The 13 dichotomous items were summed to form a composite scale of general behavior. Previous studies that use the Add Health data have used similar scales to measure adulthood criminal behavior (Beaver, Reference Beaver2008b). The current study dichotomizes each item so that the less serious items do not overinflate the score for the general criminal behavior scale (i.e., less serious and more serious are equally weighted); thus, respondents’ scores represent the variety of one's criminal behavior (Moffitt, Caspi, Rutter, & Silva, Reference Moffitt, Caspi, Rutter and Silva2001).
Independent variables
Childhood neglect was measured retrospectively in Wave III to determine how often an individual's caretaker did not meet their basic physical needs (Bernstein, Ahluvalia, Pogge, & Handelsman, Reference Bernstein, Ahluvalia, Pogge and Handelsman1997; Bernstein, Fink, Handelsman, & Fotte, Reference Bernstein, Fink, Handelsman and Fotte1994). Respondents were asked, “How often had your parents or other adult caregivers not taken care of your basic needs, such as keeping you clean or providing food or clothing?” This question is similar to physical neglect questions in the Childhood Trauma Questionnaire (Bernstein et al., Reference Bernstein, Fink, Handelsman and Fotte1994). The original scale was recoded for this analysis as an ordinal level variable (0 = never, 1 = one to five times, 2 = six or more times).
The 5-HTTLPR insertion/deletion polymorphism was genotyped using a modified polymerase chain reaction assay that has been described elsewhere (Add Health Biomarkers Team). 5-HTTLPR is coded as codominant for the short allele. Research has shown that the short allele increases the risk of involvement in criminal and antisocial behavior (Comings et al., Reference Comings, Gade-Andavolu, Gonzalez, Wu, Muhleman and Blake2000; Haberstick, Smolen, & Hewitt, Reference Haberstick, Smolen and Hewitt2006; Lesch et al., Reference Lesch, Bengel, Heils, Sabol, Greenberg and Petri1996; Sen, Burmeister, & Ghosh, Reference Sen, Burmeister and Ghosh2004). Individuals were scored 0 = long/long genotype, 1 = one short allele and one long allele, or 2 = short/short genotype. Consistent with previous studies (Kaufman et al., Reference Kaufman, Yang, Douglas-Palumberi, Houshyar, Lipschitz and Krystal2004), African Americans are more likely to have the long/long genotype (54.3% males, 57.9% females) compared to Hispanics (20.1% males, 23.7% females) and Caucasians (31.7% males, 34.2% females), whereas Hispanics are more likely to have the short/short genotype (29.6% males, 24.9% females) compared to Caucasians (18.0% males, 18.3% females) and African Americans (11.7% males, 6.9% females).
The genotypes were in Hardy–Weinberg equilibrium for males (χ2 = 3.35, p = .07), but not for females (χ2 = 4.99, p = .02). Failure to meet Hardy–Weinberg equilibrium for females may reflect sampling error and/or measurement error. The genetic subsample is not a probability sample of the population, and so the sample may not be representative of the population, at least in terms of the distribution of 5-HTTLPR alleles. In terms of measurement error, Add Health researchers genotyped 5-HTTLPR as a biallelic polymorphism, but research has shown that 5-HTTLPR may be a multiallelic polymorphism (Nakamura, Ueno, Sano, & Tanabe, Reference Nakamura, Ueno, Sano and Tanabe2000). For instance, Hu et al. (Reference Hu, Lipsky, Zhu, Akhtar, Taubman and Greenberg2006) discovered a (A → G) single nucleotide polymorphism in the long allele, where the LG allele was functionally similar to the short allele. The authors noted that failure to genotype 5-HTTLPR as triallelic (LA, LG, and S) may mask relationships between 5-HTTLPR and phenotypes, but that this masking effect would be minimized for phenotypes that corresponding to lower levels of serotonin activity such as aggressive or antisocial behavior. Thus, failure to genotype 5-HTTLPR as multiallelic may reduce the strength of the relationships between 5-HTTLPR and antisocial behavior, but this masking effect is expected to be minimal in size.
Control variables
Age was included in the analysis as a continuous variable measured in years during Wave III. Race was controlled through two dummy variables: (a) African American (0 = Caucasian non-Hispanic [n = 1443] and all Hispanics [n = 357], 1 = African American non-Hispanic [n = 439]), and (b) all Hispanics (0 = Caucasian non-Hispanic and African American non-Hispanic, 1 = Hispanic). Caucasian is the reference category for both variables.
Statistical analyses
The analyses are divided into three stages. First, t test, chi-square, and correlational analyses are used to examine the relationships between the independent and dependent variables. More specifically, the t test and chi-square tests are used to examine whether there are any gender differences in the independent and dependent variables. Correlations are used to examine the zero order relationships between gender, 5-HTTLPR, neglect, alcohol use problems, frequency of marijuana use, and criminal behavior. Second, negative binomial regression models are used to estimate the independent and interactive effects of the 5-HTTLPR and childhood neglect on the three dependent variables: alcohol use problems, frequency of marijuana use, and criminal behavior. Negative binomial models are used instead of ordinary least squares due to the overdispersion in our dependent variables. Negative binomial models are typically appropriate for count variables. Although our measure of alcohol use problems is a sum of ordinally scaled items and the criminal behavior scale is a sum of dichotomous items, the distribution of the alcohol use problems and criminal behavior scales are positively skewed and overdispersed, which fits the distribution of data assumed under the negative binomial model. Also, the use of a negative binomial model, rather than a tobit model with a logged dependent variable, allows one to account for the nested nature of the genetic subsample data. Huber/White standard errors are estimated due to the clustered or nested nature of the genetic/sibling subsample data (i.e., siblings within a cluster; Rogers, Reference Rogers1993).
Separate models are estimated for males and females, in order to explore whether there are any gender differences in the effects of 5-HTTLPR and neglect on alcohol use problems, marijuana use, and offending. These separate models allow us to examine whether 5-HTTLPR and neglect are associated with substance use and criminal behavior for both males and females. Negative binomial coefficients are considered statistically significant if p ≤ .025 (Bonferroni corrected, two tailed). Finally, a difference of coefficients test is used to examine whether the independent or interactive effects of 5-HTTLPR and neglect on the antisocial outcomes are significantly stronger for males or females (p ≤ .05, two tailed, z ≥ 1.96; Clogg, Petkova, & Haritou, Reference Clogg, Petkova and Haritou1995). Previous research suggests that gender conditions the independent effects of 5-HTTLPR (Cadoret et al., Reference Cadoret, Langbehn, Caspers, Troughton, Yucuis and Sandhu2003; Gelernter, Kranzler, Coccaro, Siever, & New, Reference Gelernter, Kranzler, Coccaro, Siever and New1998) and neglect (English, Widom, & Brandford, Reference English, Widom and Brandford2001; Widom, Reference Widom1989a, Reference Widom1989b) on antisocial behaviors, but it is not known whether gender conditions the interaction of these factors.
Results
Bivariate statistics
Before proceeding to the analyses, it is important to describe the number of cases falling into the joint categories of the interaction (5-HTTLPR × Neglect) for females and males (Table 1). Chi-square analyses reveal that males (4.7%) are more likely to be neglected six or more times than females (3.2%, χ2 = 28.557, p < .001), but there are no gender differences in the frequency of 5-HTTLPR short alleles (χ2 = 3.443, p = .179; Table 2). Males, on average, report higher levels of offending (t = 8.843, p < .001), alcohol use problems (t = 7.665, p < .001), and frequency of marijuana use (t = 7.077, p < .001).
Number of cases in the interaction categories
Descriptive statistics for Add Health
Note: 5-HTTLPR, serotonin transporter linked polymorphic region gene.
Similar to previous studies, descriptive statistics suggest that respondents are somewhat resilient to the effects of neglect. Of individuals who experienced neglect six or more times during childhood, 71% of females and 47% of males did not report any problems related to alcohol use. A similar percentage of frequently neglected respondents (six + times) did not use marijuana during early adulthood (87% of females, 70% of males). Approximately 77% of females who reported six or more incidents of neglect stated that they did not engage in criminal behavior, and 60% of frequently neglected males did not offend in Wave III. Males seemed to be less resilient to the effects of neglect than females, especially in regard to alcohol use problems. Regardless, these results suggest that a substantial percentage (47–87%) of frequently neglected (six + times) females and males do not engage in criminal behavior or substance abuse during adulthood.
There are a number of correlations between the theoretically relevant independent and dependent variables (Table 3). For females, 5-HTTLPR is positively associated with childhood neglect (r = .074, df = 1213, p = .009) and alcohol use problems (r = .056, df = 1228, p = .049). Neglect is also associated with offending (r = .134, df = 1204, p < .001) for females. For males, childhood neglect is positively correlated with criminal behavior (r = .176, df = 1054, p < .001). 5-HTTLPR is not related to neglect for males (r = .013, df = 1073, p = .682).
Zero-order correlations for females and males
Note: 5-HTTLPR, serotonin transporter linked polymorphic region gene.
*p ≤ .05. **p ≤ .01.
Zero-order correlations suggest a significant gene–environment correlation between 5-HTTLPR and childhood neglect for females. To further investigate the validity and stability of this gene–environment correlation, we estimate separate multinomial regression model for females and males where childhood neglect is regressed onto 5-HTTLPR, controlling for race and age. “Never experienced neglect” is chosen as the reference category. These models also take into account the nested nature of the cases in the genetic subsample by using Huber/White standard errors. The results reveal that 5-HTTLPR is associated with six or more incidences of neglect for females (b = 0.660, relative risk ratio [RRR] = 1.936, p = .020), but 5-HTTLPR does not distinguish between females who never experienced neglect and females who experienced one to five incidents of neglect (b = 0.143, RRR = 1.154, p = .405). Approximately 7% of females who carried two copies of the short allele experienced six or more incidents of neglect (n = 15), compared to 2.2% of females with one short allele (n = 12), and 2.8% of females with zero short alleles (n = 12). For males, 5-HTTLPR is not associated with one to five incidents of neglect (b = −0.121, RRR = 0.885, p = .433) or six or more incidents of neglect (b = 0.377, RRR = 1.458, p = .078).
Multivariate statistics
Alcohol use problems
As shown in the first column of Table 4, 5-HTTLPR has a significant direct effect on alcohol use problems for females (b = 0.183, incident rate ratio [IRR] = 1.201, p = .020). Exponentiation of the 5-HTTLPR coefficient {(exp[b] − 1) × 100} reveals that the rate of alcohol use problems for females increases 20% with a one unit increase in the number of short 5-HTTLPR alleles {(exp[0.183] − 1) × 100 = 20.08%}. The main effect of neglect and the interaction coefficient are not significant for females. For males, the main effects and interaction effect of 5-HTTLPR and neglect are not significant.
Independent and interactive effects of 5-HTTLPR and childhood neglect on alcohol use problems, frequency of marijuana use, and criminal behavior for females and males
Note: 5-HTTLPR, Serotonin transporter linked polymorphic region gene.
*p ≤ .025.
The effects of race, however, are statistically significant for both females and males. The results show that Caucasian women and Caucasian men have more alcohol use problems than African American women (b = −0.962, IRR = 0.381, p < .001), Hispanic women (b = −0.448, IRR = 0.638, p = .004), African American men (b = −0.770, IRR = 0.462, p < .001), and Hispanic men (b = −0.423, IRR = 0.654, p = .002), respectively. More specifically, the predicted rates of alcohol use problems are approximately 2.5 times higher among Caucasian women compared to African American women, and 1.5 times higher for Caucasian women compared to Hispanic women. For men, Caucasian men are two times more likely to report alcohol use problems than African American men, and 1.5 times more likely to have alcohol use problems compared to Hispanic men.
Frequency of marijuana use
Table 4 also displays the negative binomial models for frequency of marijuana use for females and males. As shown in the table, 5-HTTLPR interacts with childhood neglect to increase the frequency of marijuana use for females (b = 1.325, IRR = 3.762, p < .001). The independent effects of 5-HTTLPR and childhood neglect are not significant. This finding suggests that childhood neglect only leads to significantly higher levels of marijuana use for females when they carry two copies of the 5-HTTLPR short allele. Figure 1 shows that the risk of frequent marijuana use is highest for frequently neglected females (six + times) who carry two copies of the short allele and moderately neglected females (one to five times) who are homozygous for the short allele. Neglected females who do not carry two copies of the 5-HTTLPR short allele and nonneglected females (regardless of genotype) are not at an increased risk of using marijuana on a frequent basis. The interaction effect is not significant for males, therefore suggesting the interaction effect may be gender specific. A difference of coefficients test reveals that the interaction effect of 5-HTTLPR and neglect on marijuana use is significantly stronger for females (z = 3.331, p < .001). This finding suggests that males and females who have the same genotype and who are exposed to the same environmental risk factor (i.e., childhood neglect) have different outcomes, at least in terms of marijuana use.
Serotonin Transporter Linked Polymorphic Region (5-HTTLPR) × Neglect interaction on frequency of marijuana use for females. [A color version of this figure can be viewed online at http://journals.cambridge.org/dpp]
Criminal behavior
Finally, the last two columns of Table 4 show the results of criminal behavior regressed on 5-HTTLPR, neglect, 5-HTTLPR × Neglect, race, and age for females and males. For females, criminal behavior is a function of childhood neglect (b = 0.617, IRR = 1.853, p = .001), age (b = −0.151, IRR = 0.859, p = .005), being African American (b = 0.594, IRR = 1.811, p = .008), and being Hispanic (b = 0.569, IRR = 1.766, p = .013). For males, the rate of offending is higher for neglected subjects (b = 0.613, IRR = 1.845, p < .001) and younger subjects (b = −0.172, IRR = 0.841, p < .001). The predicted rate of offending is approximately three times higher for males who experienced six or more incidents of neglect (expected rate = 1.68) compared to males who did not experience any neglect events (expected rate = 0.49). A difference of coefficients test reveals that the strength of the association between neglect and offending does not significantly vary by gender (z = 0.018, p = .992). There is not a significant interaction between 5-HTTLPR and childhood neglect for either females or males.
Discussion
The current study examined whether a serotonin transporter polymorphism (5-HTTLPR) moderated the effects of childhood neglect on alcohol use problems, marijuana use, and criminal behavior for young adult females and males. The results revealed that 5-HTTLPR interacted with childhood neglect to increase the rate of marijuana use for females, but not for males. More specifically, neglected females who were homozygous for the short allele had higher rates of marijuana use within the past 30 days than neglected females who were heterozygous or homozygous for the long allele (long/short and long/long genotypes). Childhood neglect did not have a significant, independent effect on marijuana use for females. These findings suggest that childhood neglect may only lead to significantly higher rates of marijuana use when females carry one or more copies of the short allele. The results also revealed that 5-HTTLPR did not moderate the effects of childhood neglect on alcohol use problems and criminal behavior for both females and males. It is important to note that marijuana use was measured within the past 30 days, whereas alcohol use problems and criminal behavior tapped into behavior over the past 12 months; thus, 5-HTTLPR moderated the effects of neglect on more recent antisocial behaviors more so than behaviors over a long period of time.
The significant Gene × Environment interaction is in line with previous research that has found that genetic factors may moderate the effects of childhood abuse and maltreatment on maladaptive behaviors (Brendgen et al., Reference Brendgen, Boivin, Vitaro, Girard, Dionne and Pérusse2008; Caspi et al., Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002). For instance, studies have found that dopaminergic and monoamine oxidase A polymorphisms may increase individuals’ sensitivity to the criminogenic effects of childhood maltreatment, abuse (Beaver, Reference Beaver2008a; Ducci et al., Reference Ducci, Enoch, Hodgkinson, Xu, Catena and Robin2008), and neglect (Widom & Brzustowicz, Reference Widom and Brzustowicz2006). 5-HTTLPR has been shown to moderate the effects of victimization on internalizing disorders and substance abuse (Scheid et al., Reference Scheid, Holzman, Jones, Friderici, Nummy and Symonds2007). Kaufman et al. (Reference Kaufman, Yang, Douglas-Palumberi, Crouse-Artus, Lipschitz and Krystal2007) recently found that maltreated youths who carried a copy of the short allele were more likely to abuse alcohol at an earlier age than maltreated youths who did not carry the short allele of 5-HTTLPR. The current study failed to replicate Kaufman et al.'s (Reference Kaufman, Yang, Douglas-Palumberi, Crouse-Artus, Lipschitz and Krystal2007) finding that 5-HTTLPR interacts with neglect to influence alcohol use. This failure to replicate Kaufman et al.'s findings may lie in our measure of neglect (rather than maltreatment in general), and our measurement of alcohol use problems in adulthood (rather than early onset of alcohol use). The current findings, however, do show that neglected females who carry one or more copies of the short allele are more likely to abuse marijuana than neglected females who do not carry a copy of the short allele. Our findings, along with the previous findings linking 5-HTTLPR × victimization to depression, lead us to hypothesize that 5-HTTLPR may interact with childhood neglect and abuse to influence both internalizing disorders and marijuana use. It may be beneficial for future studies to examine whether the interaction of 5-HTTLPR and neglect predicts the intersection of internalizing disorders and marijuana use.
Another important finding that emerged from the current analyses is that gender conditions the interaction between 5-HTTLPR and childhood neglect on substance abuse (Gender × 5-HTTLPR × Neglect). Analyses revealed that 5-HTTLPR moderated the effects of neglect on the frequency of marijuana use for females, but not for males. Further, the interaction coefficient on marijuana use was significantly stronger for females. These findings suggest that gender may be acting as a proxy for some unmeasured risk factor, which exacerbates the negative effects of 5-HTTLPR × Neglect on marijuana use.
In relation to the current study, it may be hypothesized that gender may be acting as a proxy for some biological risk factor, such as dysfunctions in the HPA axis and serotonergic system. Studies have shown that females, on average, exhibit higher levels of HPA axis activity (Young, Korszun, Figueiredo, Banks-Solomon, & Herman, Reference Young, Korszun, Figueiredo, Banks-Solomon, Herman, Becker, Berkley, Geary, Hampson, Herman and Young2007) and lower levels of serotonin activity than males (Biver et al., Reference Biver, Lotstra, Monclus, Wikler, Damhaut and Mendlewicz1996; Fischette, Biegon, & McEwen, Reference Fischette, Biegon and McEwen1983; Rubinow, Schmidt, & Roca, Reference Rubinow, Schmidt and Roca1998). Further, nonhuman primate research has revealed that maltreated subjects who carry a copy of the short allele exhibit lower levels of 5-hydroxyindoleacetic acid, exhibit higher levels of ACTH responses to stress, and exhibit more aggressive and anxious behaviors in response to stress than maltreated subjects who do not carry a copy of the short allele (Maestripieri, Reference Maestripieri, Pfaff, Kordon, Chanson and Christen2005; McCormack et al., Reference McCormack, Newman, Higley, Maestripieri and Sanchez2009). These negative effects have been shown to be more pronounced for females than males (Barr, Newman, Schwandt, et al., Reference Barr, Newman, Schwandt, Shannon, Dvoskin and Lindell2004). Thus, it may be hypothesized that the intersection of gender, the short allele of 5-HTTLPR, and neglect acts as a “triple whammy” on the HPA axis and serotonergic functioning. It may be expected that neglected females who carry the short allele will exhibit the lowest serotonin levels (both basal and tonic) and the greatest hypersensitivity to stress, relative to all other groups. As a result, these females may be at risk for developing substance abuse problems and a range of mood disorders (Andersen & Teicher, Reference Andersen and Teicher2009).
The current analyses also revealed a significant gene–environment correlation for females. The results revealed that females who carried one or more copies of the short allele were more likely to be neglected than females who did not carry a copy of the short allele. This finding is somewhat unexpected given the growing body of behavioral genetics literature that shows that genetic factors may explain a small amount of variance in childhood maltreatment (Jaffee, Caspi, Moffitt, & Taylor, Reference Jaffee, Caspi, Moffitt and Taylor2004; Kendler & Prescott, Reference Kendler and Prescott2006). A recent analysis of the Add Health data, however, reveals that genetic factors account for approximately 25% of the variance in childhood neglect (Schultz Heik et al., Reference Schulz-Heik, Rhee, Silvern, Lessem, Haberstick and Hopfer2009). Further, the study found that genetic effects accounted for approximately 33% of the variance in neglect for females, but only 3% of the variance for males; thus, magnitude of genetic effects for neglect were larger for females than for males. In light of the current findings and Schultz Heik et al.'s (Reference Schulz-Heik, Rhee, Silvern, Lessem, Haberstick and Hopfer2009) findings, it appears that genetic factors, especially 5-HTTLPR, may place females at risk for childhood neglect more so than males.
The mechanisms underlying this relationship remain open for investigation, but the mechanism may lie in females’ mental and emotional health. It may be hypothesized that 5-HTTLPR and low serotonin activity increases females’ risk of developing internalizing and mood disorders (Gonda, Juhasz, Laszik, Rihmer, & Bagdy, Reference Gonda, Juhasz, Laszik, Rihmer and Bagdy2005; Walderhaug et al., Reference Walderhaug, Magnusson, Neumeister, Lappalainen, Lunde and Refsum2007). Subsequently, females’ internalizing and mood disorders may strain the parent–child relationship and cause parents to emotionally and physically withdraw from their daughters. It is also likely that carriers of the short allele are at risk of negative emotionality (Gerra et al., Reference Gerra, Garofano, Santoro, Bosari, Pellegrini and Zaimovic2004), and thus they may be more likely to overemphasize and dwell on negative parent–child relations than individuals who do not carry the short allele. Thus, the 5-HTTLPR genotype may be related to measurement error in the retrospective self-reported neglect measure. It would be beneficial for future studies to further investigate the molecular and neurobiological mechanisms underlying this gene–environment correlation.
A final important finding from the current study is that neglect significantly influenced adult criminal behavior for females and males, but neglect did not have a significant impact on adult alcohol use problems or the frequency of marijuana use. This finding is in line with those from Widom and Ames (Reference Widom and Ames1994), which found that abuse and neglect were significantly related to arrests for property, order, and drug offenses (i.e., possession, trafficking, and use) during adulthood, but neglect was not associated with arrest for alcohol offenses. Other studies, however, have found that neglect is associated with greater risk of drug use and abuse (Dube et al., Reference Dube, Felitti, Dong, Chapman, Giles and Anda2003; Widom, Reference Widom1989b). For instance, a recent analysis of data from the Add Health study revealed that the odds of binge drinking among adolescents were approximately 1.2 times higher for neglected youths compared to nonneglected youths (Shin, Edwards, & Heeren, Reference Shin, Edwards and Heeren2009). Further, the finding that neglect is not significantly related to substance use for females is surprising given the literature that argues that females are more likely to cope with neglect through drug and alcohol abuse than criminal behavior (Broidy & Agnew, Reference Broidy and Agnew1997; Horwitz, Widom, McLaughlin, & White, Reference Horwitz, Widom, McLaughlin and White2001). The current study's findings suggest that neglect may lead to adulthood substance use for some females (i.e., those carrying two copies of the short allele), but that many neglected females (and males) are really at greater risk for criminal behavior during adulthood than substance abuse problems. The context and etiology of these problems may differ, which may suggest different intervention approaches for neglected females (and males) that show criminal behavior problems (i.e., cognitive behavioral therapy) versus neglected females that show marijuana use problems (i.e., cognitive behavioral therapy plus pharmacological intervention).
From a developmental point of view, these data contribute to the literature that shows that neglect may have long-term negative consequences on the individual. Yet, although these data illustrate the long-term effects of neglect on behavior, the findings also raise questions as to the mechanisms underlying these relationships. Previous research shows that neglected youths may be disruptive, but they are characterized by poor language ability, internalizing disorders, and social isolation (Hildyard & Wolfe, Reference Hildyard and Wolfe2002). It remains a question how these neglected children, with their internalizing disorders and poor social skills, grow up to be adults who engage in a variety of criminal behaviors. Further, given the nature of the criminal behavior scale in the current study, it is not known whether neglect is important for property crimes, violent crimes, or both. The answer to this question, as well as what other characteristics neglected children display in adulthood, may help researchers understand the development and behavior of neglected youths from childhood to adulthood.
The current study contributes to the growing body of literature showing that genetic factors condition the effects of childhood maltreatment on antisocial behaviors. More specifically, the current findings reveal that 5-HTTLPR moderates the effect of childhood neglect on substance abuse for females, but not for males. Although this is the first study to examine the interaction of 5-HTTLPR and childhood neglect on antisocial behaviors, there are at least six limitations that should be considered when interpreting the current findings. First, the current study used a single item, retrospective, self-report measure of neglect rather than an official reported and/or prospective measure. The use of a single indicator may underrepresent the construct of neglect, and it is not entirely clear how well the single item captures the concept of neglect (relative to capturing other concepts). Further, research has revealed that retrospective self-report measures of childhood maltreatment may underestimate the prevalence of maltreatment (relative to official reports) and introduce error into statistical models (Brown, Craig, Harris, Handley, & Harvey, Reference Brown, Craig, Harris, Handley and Harvey2007; Widom & Shepard, Reference Widom and Shepard1996). Although retrospective measures increase error in statistical models, Widom and Morris (Reference Widom and Morris1997) found that the error associated with retrospective measures of abuse was not large enough to severely attenuate the relationships between childhood abuse and antisocial behavior in adulthood.
Second, the 5-HTTLPR polymorphism deviated from Hardy Weinberg equilibrium. As previously noted, this may be attributed to sampling error or measurement error. Sampling error may attenuate the significance of the relationship by increasing the standard error and decreasing the z value. In terms of measurement error, Hu et al. (Reference Hu, Lipsky, Zhu, Akhtar, Taubman and Greenberg2006) argued that the strength of the relationship should be attenuated when 5-HTTLPR is genotyped as biallelic (vs. triallelic), because the LG is functionally similar to the short allele, but it is genotyped as the long allele; thus, the difference between the long allele and short allele should be less than if it was properly genotyped as triallelic. Thus, the presence of sampling error and measurement error may provide a more “conservative” estimate of the relationship between 5-HTTLPR and antisocial behavior because the statistical significance and strength of the effect are dampened by both types of error.
Third, the current study used a nonclinical measure of alcohol use problems; thus, it is possible that the significant interaction between 5-HTTLPR and childhood neglect for females may not generalize to clinical populations. Although this is a concern, the authors attempted to minimize this limitation by including items in the alcohol use problems measure that proxies the DSM-IV criteria for alcohol dependence.
Fourth, the current study did not control for some theoretically relevant variables, such as household poverty and child's behavioral problems. The current study did not include such theoretically relevant controls because this information was not available for subjects at the appropriate time period. The childhood neglect measure was a retrospective measure that asked about neglect prior to the sixth grade, but Add Health researchers did not ask about other circumstances or characteristics prior to the sixth grade; thus, information on household poverty and behavioral problems could not be taken into account in the models. The current study did however control for relevant demographics, which could proxy unmeasured, theoretically relevant control variables.
Fifth, the number of cases falling into the joint categories, while similar to other Gene × Environment interaction analyses, was relatively small. This may lead to instability in the parameter estimates for the current analyses.
Sixth, and finally, the current study did not include measures of other genetic polymorphisms (such as MAOA polymorphisms) that have been linked to serotonergic functioning and antisocial behavior (Murphy et al., Reference Murphy, Uhl, Holmes, Ren-Patterson, Hall and Sora2003). For instance, Passamonti et al. (Reference Passamonti, Cerasa, Gioia, Magariello, Muglia and Quattrone2008) found that an MAOA polymorphism interacted with the 5-HTTLPR polymorphism to influence activity in the anterior cingulate cortex, an area that is important for decision making, impulse control, and emotional regulation. Thus, it is important for future research to take into account MAOA polymorphisms.
It is important to note that the 5-HTTLPR short allele has been conceptualized as a risk factor, but it may also be conceptualized as merely an alternative way of responding to the environment. The view that genetic variation produces variation in behavioral responses, rather than producing cognitive or behavioral deficits, is an emerging theme in neuropsychological and molecular genetic research (Zhang, Parent, Weaver, & Meaney, Reference Zhang, Parent, Weaver and Meaney2004). This “alternative response” view may accurately reflect the complexity of genetic variation and expression, because many studies find conflicting results on which allele is considered the “risk allele” (Vaske et al., Reference Vaske, Beaver, Wright, Boisvert and Schnupp2009). Thus, the current results may be interpreted as the short allele increasing one's sensitivity to a particular environment (i.e., neglect), rather than the synergy of risk factors.
The current study revealed that 5-HTTLPR moderated the effects of childhood neglect on the frequency of marijuana use for females. Neglected females who carried two copies of the short allele had higher rates of marijuana use than neglected females who were not homozygous for the short allele. A significant interaction effect was not found for males. Further, the analyses revealed a significant gene–environment correlation between 5-HTTLPR and neglect for females. These results suggest an interesting connection between gender, 5-HTTLPR, neglect, and marijuana use. Future research should investigate the role of anxiety and depression in these connections, given the large body of research linking 5-HTTLPR to depression, and gender to depression and substance use (Belknap & Holsinger, Reference Belknap, Holsinger and Zaplin1998; Teplin, Abram, McClelland, Dulcan, & Mericle, Reference Teplin, Abram, McClelland, Dulcan and Mericle2002).
