- Email: [email protected]
Examining the association between MAOA genotype and incarceration, anger and hostility: The moderating influences of risk and protective factors
Journal of Research in Personality 45 (2011) 279–284
Contents lists available at ScienceDirect
Journal of Research in Personality journal homepage: www.elsevier.com/locate/jrp
Examining the association between MAOA genotype and incarceration, anger and hostility: The moderating influences of risk and protective factors Kevin M. Beaver ⇑, Joseph L. Nedelec, Meghan Wilde, Courtney Lippoff, Dylan Jackson College of Criminology and Criminal Justice, Florida State University, United States
a r t i c l e
i n f o
Article history: Available online 26 February 2011 Keywords: Add Health Anger Gene-environment interaction Hostility Incarceration MAOA
a b s t r a c t Findings from molecular genetic research have indicated that a polymorphism in the promoter region of the MAOA gene interacts with environmental liabilities to predict antisocial phenotypes. We use these findings as a springboard to examine whether a global protective-risk factor index moderates the effect of MAOA genotype on the probability of being incarcerated and on a measure of anger and hostility. Analysis of data from the National Longitudinal Study of Adolescent Health (Add Health) indicates that exposure to risk and protective factors in adolescence are able to moderate the effect of MAOA genotype on anger and hostility in adulthood for males. The results in relation to the probability of being incarcerated were consistently null. Ó 2011 Elsevier Inc. All rights reserved.
1. Introduction A long line of behavioral genetic research has examined the genetic and environmental underpinnings to virtually every measurable antisocial behavior. The results of these studies, which are based on thousands of kinship pairs, collected in different geographical regions, and at different time periods, have converged to reveal that approximately 50% of the variance in antisocial behaviors is attributable to genetic factors (Moffitt, 2005; Rhee & Waldman, 2002). As a result, there has been increasing interest in moving away from only decomposing phenotypic variance and instead focusing on identifying the specific genetic polymorphisms that are involved in explaining variance in antisocial behaviors. Extant research has indicated that the genes and gene systems that are most likely to contribute to antisocial behaviors are those that are involved in neurotransmission (Ferguson & Beaver, 2009). Of all the genes that have been studied in relation to antisocial phenotypes, the monoamine oxidase A (MAOA) gene has produced the most consistent results. The MAOA gene is located on the X chromosome (Xp11.23-11.4) and is responsible for encoding the MAOA enzyme which degrades neurotransmitters, such as serotonin, dopamine, and norepinephrine. The MAOA gene has a polymorphism (MAOA-uVNTR) that is the result of a 30-base-pair (bp) variable number of tandem repeats upstream in the 50 regulatory region of the gene. This polymorphism has been shown to ⇑ Corresponding author. Address: Florida State University, College of Criminology and Criminal Justice, 634 West Call Street, Tallahassee, FL 32306-1127, United States. Fax: +1 850 644 9614. E-mail address: [email protected] (K.M. Beaver). 0092-6566/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.jrp.2011.02.007
affect the functioning of the MAOA enzyme with some of the alleles encoding a low activity MAOA enzyme and others encoding a high activity MAOA enzyme. Genotyping MAOA via PCR typically produces the following five fragment sizes: 2 repeats (2R), 3 repeats (3R), 3.5 repeats (3.5R), 4 repeats (4R), and 5 repeats (5R). A general consensus has been reached in that the 2R and 3R alleles correspond to low MAOA activity, while the 3.5R and 4R alleles correspond to high MAOA activity. The 5R allele, however, has been shown to produce both low MAOA activity (Sabol, Hus, & Hamer, 1998) and high MAOA activity (Deckert et al., 1999). Human genetic research has examined the direct association between MAOA genotypes and antisocial behaviors, revealing that the alleles that encode the low activity MAOA enzyme confer an increased risk to antisocial phenotypes. For example, the low MAOA activity alleles have been linked to delinquent behavior in adolescents and young adults (Guo, Ou, Roettger, & Shih, 2008) as well as more serious types of violence, such as weapon use and gang membership (Beaver, DeLisi, Vaughn, & Barnes, 2010). While studies have documented the main effects that MAOA might have on violence and aggression, the most consistent evidence linking MAOA genotype to antisocial phenotypes comes from research examining gene-environment interactions. The logic underlying this line of inquiry suggests that MAOA genotype only maintains an association with antisocial phenotypes in the presence of an environmental pathogen. In the first study that tested this possibility, Caspi et al. (2002) examined the interrelationships among MAOA genotype, childhood maltreatment, and violence in a sample of males from New Zealand. Their analysis revealed that MAOA genotype was unrelated to violence for the entire sample. However, they found that MAOA genotype explained a significant amount of
280
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
variance in violence solely among males who had been maltreated as children; MAOA was unrelated to violence for males without a history of maltreatment. Importantly, a relatively recent metaanalysis of the studies examining the MAOA-maltreatment interaction found this interaction to be statistically significant across studies (Kim-Cohen et al., 2006). The link between MAOA genotype and antisocial behaviors and between the MAOA-maltreatment interaction and antisocial behaviors is highly complex and likely involves a long chain of intermediary processes and phenotypes. Known as endophenotypes, these intermediary processes/phenotypes are thought to fall somewhere between genotype and phenotype and are decidedly easier to identify in genetic association studies (Gottesman & Gould, 2003). Imaging genomics has provided evidence that certain neurobiological functions and structures may be endophenotypes that partially explain the association between MAOA genotype and antisocial behaviors. For example, carriers of the low MAOA activity alleles have been shown to have reduced limbic volume, amygdala hyperresponsivity, reduced prefrontal cortex reactivity, and structural changes to the orbitofrontal cortex (MeyerLindenberg et al., 2006). All of these neurobiological markers have been found to be, or have been posited to be, related to antisocial behaviors (Viding & Frith, 2006). Of particular importance is that genomic-imaging research has also drawn attention to the potential for certain personality traits to be endophenotypes in the MAOA-antisocial behaviors association (Alia-Klein et al., 2009). Buckholtz and Meyer-Lindenberg (2008), for instance, showed that some of the neurobiological endophenotypes mentioned previously are associated with higher scores on the personality trait anger and hostility. Even more applicable to the current research are the studies that have examined the association between MAOA genotype and antisocial personality traits. Williams et al. (2009) found, for example, that carriers of the low MAOA activity alleles, in comparison with carriers of the high MAOA activity alleles, scored significantly higher on measures of antisocial personality traits. Similar results were reported by Yang et al. (2007) in their analysis of Korean women. Importantly, however, not all studies have detected an association between MAOA genotype and antisocial personality traits (Koller, Bondy, Preuss, Bottlender, & Soyka, 2003). In general, studies investigating the nexus between MAOA and antisocial personality traits have failed to test for the role of moderating factors. As a result, heterogeneity in these study findings could be the result of differential exposure to risk and/or protective factors, a possibility that has not been fully explored to date. The current study builds off and extends previous research in three important ways. First, consistent with prior research, we examine whether MAOA interacts with certain factors to predict involvement in serious criminal behavior. Second, unlike existing studies, we do not focus on maltreatment as the moderating variable, but instead employ a protective-risk factor index. This index measures exposure to protective and risk factors as a continuum ranging from heavy exposure to protective factors to heavy exposure to risk factors. In this way, we are able to examine whether the presence of protective factors is able to blunt the effects of MAOA and whether the presence of risk factors is able to exacerbate the effects of MAOA (Belsky & Pluess, 2009). Moreover, this protective-risk factor index includes more than only social-environmental factors; instead, it is much more global and examines an array of environmental- and individual-level factors. As a result, we are able to explore the possibility that individual-level characteristics, such as verbal abilities, interact with genotype to affect phenotypic outcomes. Third, instead of only using a measure of antisocial behavior as the outcome of interest, we also employ a measure of anger and hostility. Because anger and hostility has previously been linked to antisocial behaviors (Gullone & Moore,
2000; Samuels et al., 2004), we propose that anger and hostility could be an intermediary phenotype that explains part of the mechanisms by which MAOA interacts with the environment to predict antisocial behaviors. We test these issues by analyzing genotypic and phenotypic data drawn from a longitudinal sample of American youths and adults. 2. Materials and methods 2.1. Participants Participants for this study were drawn from the National Longitudinal Study of Adolescent Health (Add Health; Udry, 2003). The Add Health is a longitudinal and nationally representative sample of American youths who were enrolled in middle or high school during the 1994–1995 school year. Four waves of data have been collected thus far. The first wave of data was comprised of two different components: the wave 1 in-school survey and the wave 1 inhome survey. The wave 1 in-school survey was administered to more than 90,000 youths while they were at school. To gain indepth information about some of the adolescents, a subsample of youths was selected to be re-interviewed at their home along with their primary caregiver. A total of 20,745 adolescents and 17,700 of their primary caregivers participated in the wave 1 in-home component of the Add Health study. About one to 2 years after the wave 1 data were collected, the second round of surveys was administered. Overall, 14,738 adolescents were included in the wave 2 of the Add Health data. Subsequently, between 2001 and 2002, the third wave of data was collected from 15,197 participants. The fourth and final round of surveys was distributed between 2007 and 2008 when most of the respondents were 24– 32 years old. A total of 15,701 respondents participated in the wave 4 component of the Add Health study. During wave 3 data collection, a subset of respondents was asked to submit samples of their buccal cells for genotyping. Respondents who had a sibling who was also participating in the Add Health study were eligible to participate. Overall, more than 2500 subjects submitted usable samples of their DNA, making the Add Health one of the largest samples in the world to include genotypic data (Harris, Halpern, Smolen, & Haberstick, 2006). After removing cases because of attrition and missing data via listwise deletion, the final analytical sample ranged between N = 420 and 493. 2.2. Genotyping Add Health participants were genotyped for the MAOA-uVNTR polymorphism using a variant of a previously developed assay (Sabol et al., 1998). DNA amplification was achieved by using the following primer sequences: forward, 50 ACAGCCTGACCG-TGGA GAAG-30 (fluorescently labeled), and reverse, 50 -GAACGTGACGCTC CATTCGGA-30 . This assay resulted in the PCR products of 291 (2-repeat allele), 321 (3-repeat allele), 336 (3.5-repeat allele), 351 (4-repeat allele), and 381 (5-repeat allele) base pairs. Each genotype was scored by two independent raters. Following previous researchers analyzing the Add Health data (Haberstick et al., 2005), alleles of the MAOA gene were pooled to form two groups: a low MAOA activity group and a high MAOA activity group. The low MAOA activity group was created by pooling together the 2-repeat allele and the 3-repeat allele. The high MAOA activity group was created by pooling together the 3.5-repeat allele, the 4-repeat allele, and the 5-repeat allele. With this coding strategy employed, 17.8% of females were homozygous for the low MAOA activity allele, 43.0% were heterozygous, and 39.1% were homozygous for the high MAOA activity allele. For males, who have only one MAOA allele, 40.2% possessed the low
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
MAOA activity allele, while 59.8% possessed the high MAOA activity allele. Descriptive statistics for the distribution of MAOA alleles as well as the other variables/scales used in the analyses are presented in Table 1. 2.3. Ever incarcerated Previous research has revealed that MAOA interacts with environmental liabilities to predict contact with the criminal justice system (Caspi et al., 2002). In order to examine whether MAOA would interact with the protective-risk factor index, a one-item ever incarcerated variable was employed. During wave 4 interviews, respondents were asked to indicate whether they had ever spent time in a jail, prison, juvenile detention center, or other correctional facility. Responses to this item were coded dichotomously, where 0 = no and 1 = yes. Importantly, alternative measures of contact with the criminal justice system were also used (e.g., ever arrested, ever sentenced to probation) and the substantive results were identical to the ones reported with the ever incarcerated variable. 2.4. Anger and hostility On wave 4 surveys, respondents were asked a range of questions designed to measure personality traits. Four of the available items, drawn from the neuroticism factor, measured individual variation in anger and hostility (Miller, Lynam, Widiger, & Leukefeld, 2001). Specifically, respondents were asked whether they rarely get irritated, whether they keep their cool, whether they lose their temper (reverse coded), and whether they get angry easily (reverse coded). Responses to these items were coded as follows: 1 = strongly agree, 2 = agree, 3 = neither agree or disagree, 4 = disagree, and 5 = strongly disagree. Factor analysis revealed that all of the items loaded on a unitary factor. As a result, responses to these four items were summed together to create the anger and hostility scale, where higher values indicated greater levels of anger and hostility (a = .77). 2.5. Protective-risk factor index To examine the potential role that protective and risk factors play in the moderation of MAOA on anger and hostility, a protective-risk factor index was created using data collected at wave 1. Unlike most gene-environment studies that focus on the negative-side of the environment, this index represents a continuum that ranges from protective/positive factors on the one end to risk/negative factors on the other end. By using such a measure, it is possible to determine whether protective factors can dampen Table 1 Descriptive statistics for Add Health study variables for females and males. Females Mean Ever incarcerated Yes No Anger and hostility Protective-risk factor index Race Caucasian African–American MAOA Low/low Low/high High/high
Males SD
%
N
5.9 94.1
29 464
Mean
SD
10.18
2.72
9.80
2.91
0.97
2.39
0.86
2.21
%
N
21.9 78.1
92 329
77.9 22.1
384 109
79.3 20.7
333 87
17.8 43.0 39.1
88 212 193
40.2
169
59.8
251
281
the effects of MAOA, and whether risk factors can exacerbate the effects of MAOA. Five protective factors and five risk factors were included in the index. The five protective factors included in the index were: verbal skills (as measured by the Peabody Picture Vocabulary Test), a 4-item grade point average (GPA) variable (a = .73), a 5-item attachment to school scale (a = .78), an 8-item social support scale (a = .77), and a 3-item religiosity scale (a = .73). These items, which were all scored such that higher values represent more positive outcomes, were then split at the mean. Respondents who scored at or above the mean for that variable received a value of ‘‘1’’; otherwise, they were assigned a value of ‘‘0.’’ The five risk factors included in the index were: a 3-item neighborhood disadvantage scale (a = .62), an 18-item depression scale (a = .86), a 22-item maternal negativity scale (a = .73), a 3-item delinquent peers scale (a = .75), and a dichotomous broken home variable (0 = raised in a non-broken home, 1 = raised in a broken home). All of these items were coded such that higher values reflect more negative outcomes. The items were then divided at the mean (except for the variable, broken home, which was coded yes/no) and respondents who scored at or above the mean for that variable were assigned a value of ‘‘+1’’; otherwise, they were assigned a value of ‘‘0.’’ All of the protective and risk factors were then summed together to create the protective-risk factor index that ranged between 5 and +5. Values below zero indicate more protective factors in comparison with risk factors, while values above zero indicate more risk factors in comparison with protective factors. Although continuous measures are often viewed as preferable to dichotomous variables, we followed the lead of previous researchers (Turner, Hartman, Exum, & Cullen, 2007) and dichotomized the risk and protective factors at the mean and then summed them together. We employed this measurement strategy for four main reasons. First, there is a significant body of research indicating that the most antisocial adolescents are saturated with multiple risk factors (Wachs, 2000). Focusing on just a single risk factor in isolation fails to delineate those who are highly antisocial versus those who are only marginally antisocial. Second, research has examined the effects that dichotomizing variables have in psychiatric and criminological studies (Farrington & Loeber, 2000). The results reveal that dichotomizing variables has very little effect on the substantive conclusions of the study. In fact, risk factors that are split at the mean tend to maintain a correlation of close to ‘‘1’’ with the original continuously measured variable and tend to have the same magnitude of effects on measures of antisocial phenotypes. Third, dichotomizing risk factors aids in the detection of interaction effects (Farrington & Loeber, 2000). From a statistical and methodological standpoint, if we had retained the original coding of the continuous variables, the data would not have been able to support an interaction analysis. Specifically, the cell frequencies of some genotype/protective-risk factor combinations would have been zero, which would reduce the statistical power needed to detect interactions and would essentially be estimating parameters where no cases exist. Fourth, we reestimated all of the models using different cut-points for the protective and risk factors (e.g., split at the median, split at one standard deviation above the mean) and the pattern of results was virtually identical to those reported when the risk and protective factors were split at the mean. We also reestimated the models by summing the risk and protective factors as continuous items (z-transformed). The substantive results were once again identical to the ones reported with the dichotomous items. As a result, the findings reported here are robust and consistent across different measurement strategies. 2.6. Statistical analysis The potential interaction between MAOA and the protectiverisk factor index was estimated by creating a multiplicative
282
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
interaction term between these two variables. Since the ever incarcerated variable is dichotomous, binary logistic regression models were estimated to test for interactions. Ordinary least squares (OLS) regression models were estimated to test for interactions between MAOA and the protective-risk factor index for the anger and hostility variable. Since MAOA is located on the X-chromosome, all statistical models were estimated separately for males and females. To help avoid population stratification effects due to race/ ethnicity, all statistical models included a dichotomous dummy variable for race (0 = Caucasian, 1 = African American). Finally, because some of the observations lacked independence (i.e., more than one sibling from the same household), all models were estimated using Huber/White standard errors. In no case were both twins from a monozygotic (MZ) twin pair included in the analyses (Haberstick et al., 2005).
3. Results To ensure that any interactions between MAOA and the protective-risk factor index were not the result of gene-environment correlation, we followed a two-step process. First, we tested for gene-environment correlation between MAOA and the protective-risk factor index by calculating difference in means tests (ANOVA for females and a t-test for males). No statistically significant differences in the average scores on the protective-risk factor index for females (F = .603, p > .05) or for males (t = 1.26, p > .05) were detected. Second, we examined whether the untransformed (i.e., not dichotomized) risk and protective factors were correlated with MAOA. As Table 2 shows, MAOA maintained a statistically significant association with two risk-protective factors for males (verbal skills and broken home) and only one for females (neighborhood disadvantage). We next proceeded to test for interactions between MAOA and the protective-risk factor index in predicting the probability of being incarcerated. Table 3 depicts the results of the logit models for females and males separately. For females, the interaction between MAOA and the protective-risk factor index, while nonsignificant, was marginally close to reaching the conventional significance level (p = .061). For males, the interaction between MAOA and the protective-risk factor index was not statistically significant. The results thus far provide little evidence that the protectiverisk factor index interacts with MAOA genotype to predict contact with the criminal justice system. The next models, however, examine whether MAOA interacts with the protective-risk factor index to predict variation in the anger and hostility scale. Table 4 displays the results of these models. The results for females can be seen in
the left-hand column and show that of all the predictor variables, only the protective-risk factor index was significantly related to scores on the anger and hostility scale. Specifically, higher scores on this scale corresponded to higher scores on anger and hostility. The right-hand column of Table 4 presents the results for males. Of particular interest was that the interaction between MAOA and the protective-risk factor index was a statistically significant predictor of scores on the anger and hostility scale. To probe this interaction more closely, we plotted scores on the anger and hostility scale (expressed as z-scores) as a function of MAOA genotype across different values on the protective-risk factor index. As can be seen in Fig. 1, for respondents in the low MAOA activity group, there was a steep increase in anger and hostility scores as scores on the protective-risk factor index increased (b = 0.12, p < .05). For respondents in the high MAOA activity group, the association between the protective-risk factor index and anger and hostility was not nearly as sharp (b = 0.03, p > .05). A post-hoc z-test (Paternoster, Brame, Mazerolle, & Piquero, 1998) confirmed that the slope for the protective-risk factor index was significantly greater for the low MAOA activity group in comparison with the high MAOA activity group (z = 1.91, p < .05, one-tailed test).
4. Discussion Contemporary molecular genetic research has underscored the close interplay between genetic polymorphisms and environmental factors in the production of human phenotypes. This is especially true for antisocial behaviors, where a growing body of research reveals that candidate genes for antisocial phenotypes tend to have their strongest effects when paired with certain environmental liabilities (Moffitt, 2005). The most studied gene-environment interaction in relation to human antisocial behavior involves the MAOA gene, which has been shown to interact with measures of childhood maltreatment to explain variance in measures of antisocial behaviors (Kim-Cohen et al., 2006). The current study expanded on this line of inquiry and examined whether MAOA would interact with a cumulative protective-risk factor index, measured in adolescence, to predict the lifetime probability of being incarcerated and to predict variation in the personality trait anger and hostility, measured in adulthood. Analysis of genotypic data drawn from the Add Health study revealed two key findings. First, for females there was not an association between MAOA genotype and incarceration or between MAOA genotype and anger and hostility. In addition, there was no evidence that MAOA genotype interacted with the protectiverisk factors index to predict the probability of being incarcerated or to predict scores on the anger and hostility scale. This finding
Table 2 Correlation matrix for selected Add Health study variables for females and males. X1 Anger and hostility Ever incarcerated MAOA Verbal skills GPA School attachment Social support Religiosity Neighborhood dis. Depression Maternal negativity Delinquent peers Broken home
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13
X2 .10*
.14* .04 .05 .13* .09* .12* .03 .04 .14* .12* .11* .02
.04 .13* .23* .17* .15* .08* .11* .08* .12* .24* .13*
X3
X4
X5
X6
X7
X8
X9
X10
X11
X12
X13
.02 .05
.05 .05 .04
.18* .11* .04 .37*
.13* .07* .02 .04 .27*
.18* .08* .01 .00 .27* .45*
.12* .07* .03 .02 .17* .22* .22*
.04 .07* .07* .33* .20* .04 .00 .08*
.20* .08* .01 .19* .35* .39* .51* .16* .08*
.11* .05 .05 .02 .21* .31* .61* .22* .00 .37*
.09* .15* .06 .04 .32* .23* .32* .27* .02 .32* .18*
.05 .08* .03 .18* .17* .11* .11* .12* .30* .13* .08* .09*
.07* .03 .01 .01 .05 .05 .01 .00 .00 .10*
.30* .05 .04 .11* .38* .29* .04 .04 .20*
.26* .20* .15* .21* .24* .17* .21* .24*
.45* .18* .07* .31* .24* .24* .12*
Note: Female estimates are above the diagonal and male estimates are below the diagonal. * Significant at the .05 level, two-tailed tests.
.23* .02 .34* .54* .29* .07*
.16* .06 .23* .24* .03
.15* .06 .02 .29*
.18* .20* .10*
.22* .02
.05
283
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284 Table 3 Logit models testing for gene environment interaction in the probability of being incarcerated for females and males. Females
MAOA Protective-risk factor index MAOA protective-risk factor index Race N
Males
B
SE
OR
Z
p
B
SE
OR
Z
.43 .10 .25 .14
.34 .10 .13 .48
1.54 1.10 1.28 .871 493
1.29 0.95 1.87 0.77
.198 .340 .061 .773
.12 .29 .07 .65
.26 .06 .11 .29
1.12 1.34 1.07 1.92 421
0.44 5.11 0.57 2.26
p .658 <.001 .571 .024
Note: Models were estimated using Huber/White standard errors. Race was coded dichotomously, where 0 = Caucasian and 1 = African American. For females, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles, 1 = the presence of one low MAOA activity allele, and 2 = the presence of two low MAOA activity alleles. For males, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles and 1 = the presence of one low MAOA activity allele.
Table 4 OLS regression models testing for gene environment interaction in the prediction of anger and hostility for females and males. Females
MAOA Protective-risk factor index MAOA protective-risk factor index Race N
Males
b
Beta
SE
t
.04 .26 .01 .13
.01 .23 .01 .02
.16 .05 .07 .30 493
0.26 5.12 0.16 0.43
p .792 <.001 .874 .664
b .50 .20 .33 .85
Beta .08 .15 .12 .12
SE
t
p
.29 .07 .14 .36 420
1.74 2.96 2.39 2.36
.083 .003 .017 .019
Note: Models were estimated using Huber/White standard errors. Race was coded dichotomously, where 0 = Caucasian and 1 = African American. For females, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles, 1 = the presence of one low MAOA activity allele, and 2 = the presence of two low MAOA activity alleles. For males, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles and 1 = the presence of one low MAOA activity allele.
Fig. 1. The interaction between MAOA and the protective-risk factor index in the prediction of anger and hostility for males.
is consistent with prior research suggesting that the effect of MAOA on antisocial phenotypes may be gender-specific and may not extend to females (Beaver et al., 2010), but is inconsistent with research showing that MAOA genotype is related to anger in a sample of Korean females (Yang et al., 2007). Second, and in line with the findings garnered for females, there was not a statistically significant interaction between MAOA genotype and the protectiverisk factor index in predicting the odds of being incarcerated for males. However, there was a statistically significant interaction between MAOA genotype and the protective-risk factor index in the prediction of anger and hostility. Post-hoc probing analysis indicated that males who carried the low MAOA activity alleles, in comparison with males who carried the high MAOA activity alleles, scored the lowest on anger and hostility when exposed to the most
protective factors and the highest on anger and hostility when exposed to the most risk factors. This latter finding is directly in line with the differential susceptibility hypothesis which suggests that certain alleles, such as the low MAOA activity alleles, may be more appropriately viewed as plasticity alleles (Belsky & Pluess, 2009). These plasticity alleles act as biological markers that influence the degree to which a person is susceptible to environmental pressures. In the face of adverse environments, carriers of plasticity alleles are more likely to be affected negatively, while in the presence of advantageous environments, carriers of plasticity alleles are more likely to be affected positively. Although the current study was not designed to test directly the differential susceptibility hypothesis, the results are nonetheless consistent with it. Additionally, the interaction between MAOA genotype and the protective-risk factor index is consistent with recent experimental research revealing that carriers of the low MAOA activity alleles employ more aggression in response to a provocation (McDermott, Tingley, Cowden, Frazzetto, & Johnson, 2009). The results of our study indicate that levels of anger and hostility are higher for low MAOA activity allele carriers in comparison with high MAOA activity allele carriers, but only if they were exposed to multiple risk factors. As a result, it is quite possible that anger and hostility serve as an adaptive response to high-risk situations for those with the genetic predisposition for antisocial phenotypes (as measured by the low MAOA activity alleles). Exactly why the interaction between MAOA and the protectiverisk factor index was statistically significant for the anger and hostility scale, but not for the odds of being incarcerated is not immediately apparent. However, we do offer one possible explanation. Being incarcerated is the end result of a complex chain of events that is affected by a host of different factors, including the seriousness of the offense being committed, the number of prior arrests, and being apprehended by law enforcement. In addition, socioeconomic status also plays a role in the ultimate outcome of a criminal case, with those of higher socioeconomic status being able to afford the most high-powered lawyers. Given all of the contributing factors to incarceration, it is possible that the effects of MAOA are
284
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
either too small to detect or are overshadowed by other factors. Future research should explore this possibility more closely by estimating the differential effects that MAOA has across a range of antisocial phenotypes. The results of the current study need to be interpreted with caution in light of a number of limitations. First, although the Add Health study was designed to be nationally representative of American youths, the genotypic sample was not. As a result, it is quite possible that the results reported here would not generalize to the larger population of Americans or to populations outside of the United States. Replication studies using different samples that include more cases are needed to determine the robustness of these findings. Second, the measure of anger and hostility was based on only four self-reported items. Ideally, the scale would have been comprised of many more items scored by multiple raters, but because the Add Health data only included four items measuring anger and hostility, we were forced to use a four-item scale. Psychometric analysis of the scale, however, revealed that it was reliable. Nonetheless, follow-up research is needed to ensure that the results generated in the current study would be applicable to other measures of anger and hostility. Lastly, the protective-risk factor index included both environmental and individual-level factors. In this way, we were examining a host of factors that might be able to moderate the association between MAOA and anger and hostility. Therefore, caution should be used when treating this study as a direct test of gene-environment interaction because the ‘‘environment’’ component of this interaction was inclusive of individual-level factors. The factors included in the protective-risk factor index, as a result, were capturing genetic effects. As a result, the interaction between MAOA and the protective-risk factor index may really be detecting gene–gene interactions as opposed to gene-environment interactions. Whether the results of the current study would remain once these shortcomings are addressed is an open-empirical question awaiting future investigation. Acknowledgments This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/ addhealth). No direct support was received from grant P01HD31921 for this analysis. References Alia-Klein, N., Goldstein, R. Z., Tomasi, D., Woicik, P. A., Moeller, S. J., Williams, B., et al. (2009). Neural mechanisms of anger regulation as a function of genetic risk for violence. Emotion, 9, 385–396. Beaver, K. M., DeLisi, M., Vaughn, M. G., & Barnes, J. C. (2010). MAOA genotype is associated with gang membership and weapon use. Comprehensive Psychiatry, 51, 130–134. Belsky, J., & Pluess, M. (2009). Beyond diathesis-stress: Differential susceptibility to environmental influence. Psychological Bulletin, 135, 885–908.
Buckholtz, J. W., & Meyer-Lindenberg, A. (2008). MAOA and the neurogenetic architecture of human aggression. Trends in Neurosciences, 31, 120–129. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., et al. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297, 851–854. Deckert, J., Catalano, M., Syagailo, Y. V., Bosi, M., Okladnova, O., Di Bella, D., et al. (1999). Excess of high activity monoamine oxidase A gene promoter alleles in female patients with panic disorder. Human Molecular Genetics, 8, 621–624. Farrington, D. P., & Loeber, R. (2000). Some benefits of dichotomization in psychiatric and criminological research. Criminal Behaviour and Mental Health, 10, 100–122. Ferguson, C. J., & Beaver, K. M. (2009). Natural born killers: The genetic origins of extreme violence. Aggression and Violent Behavior, 14, 286–294. Gottesman, I. I., & Gould, T. D. (2003). The endophenotype concept in psychiatry: Etymology and strategic intentions. American Journal of Psychiatry, 160, 636–645. Gullone, E., & Moore, S. (2000). Adolescent risk-taking and the five-factor model of personality. Journal of Adolescence, 23, 393–407. Guo, G., Ou, X.-M., Roettger, M., & Shih, J. C. (2008). The VNTR 2 repeat in MAOA and delinquent behavior in adolescence and young adulthood: Associations and MAOA promoter activity. European Journal of Human Genetics, 16, 626–634. Haberstick, B. C., Lessem, J. M., Hopfer, C. J., Smolen, A., Ehringer, M. A., Timberlake, D., et al. (2005). Monoamine oxidase A (MAOA) and antisocial behaviors in the presence of childhood and adolescent maltreatment. American Journal of Medical Genetics, 135B, 59–64. Harris, K. M., Halpern, C. T., Smolen, A., & Haberstick, B. C. (2006). The national longitudinal study of adolescent health (Add Health) twin data. Twin Research and Human Genetics, 9, 988–997. Kim-Cohen, J., Caspi, A., Taylor, A., Williams, B., Newcombe, R., Craig, I. W., et al. (2006). MAOA, maltreatment, and gene-environment interaction predicting children’s mental health: New evidence and a meta-analysis. Molecular Psychiatry, 11, 903–913. Koller, G., Bondy, B., Preuss, U. W., Bottlender, M., & Soyka, M. (2003). No association between a polymorphism in the promoter region of the MAOA gene with antisocial personality traits in alcoholics. Alcohol and Alcoholism, 38, 31–34. McDermott, R., Tingley, D., Cowden, J., Frazzetto, G., & Johnson, D. D. P. (2009). Monoamine oxidase A gene (MAOA) predicts behavioral aggression following provocation. Proceedings of the National Academy of Sciences, 106, 2118–2123. Meyer-Lindenberg, A., Buckholtz, J. W., Kolachana, B., Hariri, A. R., Pezawas, L., Blasi, G., et al. (2006). Neural mechanisms of genetic risk for impulsivity and violence in humans. Proceedings of the National Academy of Sciences, 103, 6269–6274. Miller, J. D., Lynam, D. R., Widiger, T. A., & Leukefeld, C. (2001). Personality disorders as extreme variants of common personality dimensions: Can the five-factor model adequately represent psychopathy? Journal of Personality, 69, 253–276. Moffitt, T. E. (2005). The new look of behavioral genetics in developmental psychopathology: Gene-environment interplay in antisocial behaviors. Psychological Bulletin, 131, 533–554. Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36, 859–866. Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128, 490–529. Sabol, S., Hus, S., & Hamer, D. (1998). A functional polymorphism in the monoamine oxidase A gene promoter. Human Genetics, 103, 273–279. Samuels, J., Bienvenu, O. J., Cullen, B., Costa, P. T., Easton, W. W., & Nestadt, G. (2004). Personality dimensions and criminal arrest. Comprehensive Psychiatry, 45, 275–280. Turner, M. G., Hartman, J. L., Exum, M. L., & Cullen, F. T. (2007). Examining the cumulative effects of protective factors: Resiliency among a national sample of high-risk youths. Journal of Offender Rehabilitation, 46, 81–111. Udry, J. R. (2003). The national longitudinal study of adolescent health (Add Health), waves I and II, 1994–1996; wave III, 2001–2002 [Data file and documentation]. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill. Viding, E., & Frith, U. (2006). Genes for susceptibility to violence lurk in the brain. Proceedings of the National Academy of Sciences, 16, 6085–6086. Wachs, T. D. (2000). Necessary but not sufficient: The respective roles of single and multiple influences on individual development. Washington, DC: APA. Williams, L. M., Gatt, J. M., Kuan, S. A., Dobson-Stone, C., Palmer, D. M., Paul, R. H., et al. (2009). A polymorphism of the MAOA gene is associated with emotional brain markers and personality traits on an antisocial index. Neuropsychopharmacology, 34, 1797–1809. Yang, J.-W., Lee, S.-H., Ryu, S.-H., Lee, B.-C., Kim, S.-H., Joe, S.-H., et al. (2007). Association between monoamine oxidase A polymorphisms and anger-related personality traits in Korean women. Neuropsychobiology, 56, 19–23.
Contents lists available at ScienceDirect
Journal of Research in Personality journal homepage: www.elsevier.com/locate/jrp
Examining the association between MAOA genotype and incarceration, anger and hostility: The moderating influences of risk and protective factors Kevin M. Beaver ⇑, Joseph L. Nedelec, Meghan Wilde, Courtney Lippoff, Dylan Jackson College of Criminology and Criminal Justice, Florida State University, United States
a r t i c l e
i n f o
Article history: Available online 26 February 2011 Keywords: Add Health Anger Gene-environment interaction Hostility Incarceration MAOA
a b s t r a c t Findings from molecular genetic research have indicated that a polymorphism in the promoter region of the MAOA gene interacts with environmental liabilities to predict antisocial phenotypes. We use these findings as a springboard to examine whether a global protective-risk factor index moderates the effect of MAOA genotype on the probability of being incarcerated and on a measure of anger and hostility. Analysis of data from the National Longitudinal Study of Adolescent Health (Add Health) indicates that exposure to risk and protective factors in adolescence are able to moderate the effect of MAOA genotype on anger and hostility in adulthood for males. The results in relation to the probability of being incarcerated were consistently null. Ó 2011 Elsevier Inc. All rights reserved.
1. Introduction A long line of behavioral genetic research has examined the genetic and environmental underpinnings to virtually every measurable antisocial behavior. The results of these studies, which are based on thousands of kinship pairs, collected in different geographical regions, and at different time periods, have converged to reveal that approximately 50% of the variance in antisocial behaviors is attributable to genetic factors (Moffitt, 2005; Rhee & Waldman, 2002). As a result, there has been increasing interest in moving away from only decomposing phenotypic variance and instead focusing on identifying the specific genetic polymorphisms that are involved in explaining variance in antisocial behaviors. Extant research has indicated that the genes and gene systems that are most likely to contribute to antisocial behaviors are those that are involved in neurotransmission (Ferguson & Beaver, 2009). Of all the genes that have been studied in relation to antisocial phenotypes, the monoamine oxidase A (MAOA) gene has produced the most consistent results. The MAOA gene is located on the X chromosome (Xp11.23-11.4) and is responsible for encoding the MAOA enzyme which degrades neurotransmitters, such as serotonin, dopamine, and norepinephrine. The MAOA gene has a polymorphism (MAOA-uVNTR) that is the result of a 30-base-pair (bp) variable number of tandem repeats upstream in the 50 regulatory region of the gene. This polymorphism has been shown to ⇑ Corresponding author. Address: Florida State University, College of Criminology and Criminal Justice, 634 West Call Street, Tallahassee, FL 32306-1127, United States. Fax: +1 850 644 9614. E-mail address: [email protected] (K.M. Beaver). 0092-6566/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.jrp.2011.02.007
affect the functioning of the MAOA enzyme with some of the alleles encoding a low activity MAOA enzyme and others encoding a high activity MAOA enzyme. Genotyping MAOA via PCR typically produces the following five fragment sizes: 2 repeats (2R), 3 repeats (3R), 3.5 repeats (3.5R), 4 repeats (4R), and 5 repeats (5R). A general consensus has been reached in that the 2R and 3R alleles correspond to low MAOA activity, while the 3.5R and 4R alleles correspond to high MAOA activity. The 5R allele, however, has been shown to produce both low MAOA activity (Sabol, Hus, & Hamer, 1998) and high MAOA activity (Deckert et al., 1999). Human genetic research has examined the direct association between MAOA genotypes and antisocial behaviors, revealing that the alleles that encode the low activity MAOA enzyme confer an increased risk to antisocial phenotypes. For example, the low MAOA activity alleles have been linked to delinquent behavior in adolescents and young adults (Guo, Ou, Roettger, & Shih, 2008) as well as more serious types of violence, such as weapon use and gang membership (Beaver, DeLisi, Vaughn, & Barnes, 2010). While studies have documented the main effects that MAOA might have on violence and aggression, the most consistent evidence linking MAOA genotype to antisocial phenotypes comes from research examining gene-environment interactions. The logic underlying this line of inquiry suggests that MAOA genotype only maintains an association with antisocial phenotypes in the presence of an environmental pathogen. In the first study that tested this possibility, Caspi et al. (2002) examined the interrelationships among MAOA genotype, childhood maltreatment, and violence in a sample of males from New Zealand. Their analysis revealed that MAOA genotype was unrelated to violence for the entire sample. However, they found that MAOA genotype explained a significant amount of
280
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
variance in violence solely among males who had been maltreated as children; MAOA was unrelated to violence for males without a history of maltreatment. Importantly, a relatively recent metaanalysis of the studies examining the MAOA-maltreatment interaction found this interaction to be statistically significant across studies (Kim-Cohen et al., 2006). The link between MAOA genotype and antisocial behaviors and between the MAOA-maltreatment interaction and antisocial behaviors is highly complex and likely involves a long chain of intermediary processes and phenotypes. Known as endophenotypes, these intermediary processes/phenotypes are thought to fall somewhere between genotype and phenotype and are decidedly easier to identify in genetic association studies (Gottesman & Gould, 2003). Imaging genomics has provided evidence that certain neurobiological functions and structures may be endophenotypes that partially explain the association between MAOA genotype and antisocial behaviors. For example, carriers of the low MAOA activity alleles have been shown to have reduced limbic volume, amygdala hyperresponsivity, reduced prefrontal cortex reactivity, and structural changes to the orbitofrontal cortex (MeyerLindenberg et al., 2006). All of these neurobiological markers have been found to be, or have been posited to be, related to antisocial behaviors (Viding & Frith, 2006). Of particular importance is that genomic-imaging research has also drawn attention to the potential for certain personality traits to be endophenotypes in the MAOA-antisocial behaviors association (Alia-Klein et al., 2009). Buckholtz and Meyer-Lindenberg (2008), for instance, showed that some of the neurobiological endophenotypes mentioned previously are associated with higher scores on the personality trait anger and hostility. Even more applicable to the current research are the studies that have examined the association between MAOA genotype and antisocial personality traits. Williams et al. (2009) found, for example, that carriers of the low MAOA activity alleles, in comparison with carriers of the high MAOA activity alleles, scored significantly higher on measures of antisocial personality traits. Similar results were reported by Yang et al. (2007) in their analysis of Korean women. Importantly, however, not all studies have detected an association between MAOA genotype and antisocial personality traits (Koller, Bondy, Preuss, Bottlender, & Soyka, 2003). In general, studies investigating the nexus between MAOA and antisocial personality traits have failed to test for the role of moderating factors. As a result, heterogeneity in these study findings could be the result of differential exposure to risk and/or protective factors, a possibility that has not been fully explored to date. The current study builds off and extends previous research in three important ways. First, consistent with prior research, we examine whether MAOA interacts with certain factors to predict involvement in serious criminal behavior. Second, unlike existing studies, we do not focus on maltreatment as the moderating variable, but instead employ a protective-risk factor index. This index measures exposure to protective and risk factors as a continuum ranging from heavy exposure to protective factors to heavy exposure to risk factors. In this way, we are able to examine whether the presence of protective factors is able to blunt the effects of MAOA and whether the presence of risk factors is able to exacerbate the effects of MAOA (Belsky & Pluess, 2009). Moreover, this protective-risk factor index includes more than only social-environmental factors; instead, it is much more global and examines an array of environmental- and individual-level factors. As a result, we are able to explore the possibility that individual-level characteristics, such as verbal abilities, interact with genotype to affect phenotypic outcomes. Third, instead of only using a measure of antisocial behavior as the outcome of interest, we also employ a measure of anger and hostility. Because anger and hostility has previously been linked to antisocial behaviors (Gullone & Moore,
2000; Samuels et al., 2004), we propose that anger and hostility could be an intermediary phenotype that explains part of the mechanisms by which MAOA interacts with the environment to predict antisocial behaviors. We test these issues by analyzing genotypic and phenotypic data drawn from a longitudinal sample of American youths and adults. 2. Materials and methods 2.1. Participants Participants for this study were drawn from the National Longitudinal Study of Adolescent Health (Add Health; Udry, 2003). The Add Health is a longitudinal and nationally representative sample of American youths who were enrolled in middle or high school during the 1994–1995 school year. Four waves of data have been collected thus far. The first wave of data was comprised of two different components: the wave 1 in-school survey and the wave 1 inhome survey. The wave 1 in-school survey was administered to more than 90,000 youths while they were at school. To gain indepth information about some of the adolescents, a subsample of youths was selected to be re-interviewed at their home along with their primary caregiver. A total of 20,745 adolescents and 17,700 of their primary caregivers participated in the wave 1 in-home component of the Add Health study. About one to 2 years after the wave 1 data were collected, the second round of surveys was administered. Overall, 14,738 adolescents were included in the wave 2 of the Add Health data. Subsequently, between 2001 and 2002, the third wave of data was collected from 15,197 participants. The fourth and final round of surveys was distributed between 2007 and 2008 when most of the respondents were 24– 32 years old. A total of 15,701 respondents participated in the wave 4 component of the Add Health study. During wave 3 data collection, a subset of respondents was asked to submit samples of their buccal cells for genotyping. Respondents who had a sibling who was also participating in the Add Health study were eligible to participate. Overall, more than 2500 subjects submitted usable samples of their DNA, making the Add Health one of the largest samples in the world to include genotypic data (Harris, Halpern, Smolen, & Haberstick, 2006). After removing cases because of attrition and missing data via listwise deletion, the final analytical sample ranged between N = 420 and 493. 2.2. Genotyping Add Health participants were genotyped for the MAOA-uVNTR polymorphism using a variant of a previously developed assay (Sabol et al., 1998). DNA amplification was achieved by using the following primer sequences: forward, 50 ACAGCCTGACCG-TGGA GAAG-30 (fluorescently labeled), and reverse, 50 -GAACGTGACGCTC CATTCGGA-30 . This assay resulted in the PCR products of 291 (2-repeat allele), 321 (3-repeat allele), 336 (3.5-repeat allele), 351 (4-repeat allele), and 381 (5-repeat allele) base pairs. Each genotype was scored by two independent raters. Following previous researchers analyzing the Add Health data (Haberstick et al., 2005), alleles of the MAOA gene were pooled to form two groups: a low MAOA activity group and a high MAOA activity group. The low MAOA activity group was created by pooling together the 2-repeat allele and the 3-repeat allele. The high MAOA activity group was created by pooling together the 3.5-repeat allele, the 4-repeat allele, and the 5-repeat allele. With this coding strategy employed, 17.8% of females were homozygous for the low MAOA activity allele, 43.0% were heterozygous, and 39.1% were homozygous for the high MAOA activity allele. For males, who have only one MAOA allele, 40.2% possessed the low
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
MAOA activity allele, while 59.8% possessed the high MAOA activity allele. Descriptive statistics for the distribution of MAOA alleles as well as the other variables/scales used in the analyses are presented in Table 1. 2.3. Ever incarcerated Previous research has revealed that MAOA interacts with environmental liabilities to predict contact with the criminal justice system (Caspi et al., 2002). In order to examine whether MAOA would interact with the protective-risk factor index, a one-item ever incarcerated variable was employed. During wave 4 interviews, respondents were asked to indicate whether they had ever spent time in a jail, prison, juvenile detention center, or other correctional facility. Responses to this item were coded dichotomously, where 0 = no and 1 = yes. Importantly, alternative measures of contact with the criminal justice system were also used (e.g., ever arrested, ever sentenced to probation) and the substantive results were identical to the ones reported with the ever incarcerated variable. 2.4. Anger and hostility On wave 4 surveys, respondents were asked a range of questions designed to measure personality traits. Four of the available items, drawn from the neuroticism factor, measured individual variation in anger and hostility (Miller, Lynam, Widiger, & Leukefeld, 2001). Specifically, respondents were asked whether they rarely get irritated, whether they keep their cool, whether they lose their temper (reverse coded), and whether they get angry easily (reverse coded). Responses to these items were coded as follows: 1 = strongly agree, 2 = agree, 3 = neither agree or disagree, 4 = disagree, and 5 = strongly disagree. Factor analysis revealed that all of the items loaded on a unitary factor. As a result, responses to these four items were summed together to create the anger and hostility scale, where higher values indicated greater levels of anger and hostility (a = .77). 2.5. Protective-risk factor index To examine the potential role that protective and risk factors play in the moderation of MAOA on anger and hostility, a protective-risk factor index was created using data collected at wave 1. Unlike most gene-environment studies that focus on the negative-side of the environment, this index represents a continuum that ranges from protective/positive factors on the one end to risk/negative factors on the other end. By using such a measure, it is possible to determine whether protective factors can dampen Table 1 Descriptive statistics for Add Health study variables for females and males. Females Mean Ever incarcerated Yes No Anger and hostility Protective-risk factor index Race Caucasian African–American MAOA Low/low Low/high High/high
Males SD
%
N
5.9 94.1
29 464
Mean
SD
10.18
2.72
9.80
2.91
0.97
2.39
0.86
2.21
%
N
21.9 78.1
92 329
77.9 22.1
384 109
79.3 20.7
333 87
17.8 43.0 39.1
88 212 193
40.2
169
59.8
251
281
the effects of MAOA, and whether risk factors can exacerbate the effects of MAOA. Five protective factors and five risk factors were included in the index. The five protective factors included in the index were: verbal skills (as measured by the Peabody Picture Vocabulary Test), a 4-item grade point average (GPA) variable (a = .73), a 5-item attachment to school scale (a = .78), an 8-item social support scale (a = .77), and a 3-item religiosity scale (a = .73). These items, which were all scored such that higher values represent more positive outcomes, were then split at the mean. Respondents who scored at or above the mean for that variable received a value of ‘‘1’’; otherwise, they were assigned a value of ‘‘0.’’ The five risk factors included in the index were: a 3-item neighborhood disadvantage scale (a = .62), an 18-item depression scale (a = .86), a 22-item maternal negativity scale (a = .73), a 3-item delinquent peers scale (a = .75), and a dichotomous broken home variable (0 = raised in a non-broken home, 1 = raised in a broken home). All of these items were coded such that higher values reflect more negative outcomes. The items were then divided at the mean (except for the variable, broken home, which was coded yes/no) and respondents who scored at or above the mean for that variable were assigned a value of ‘‘+1’’; otherwise, they were assigned a value of ‘‘0.’’ All of the protective and risk factors were then summed together to create the protective-risk factor index that ranged between 5 and +5. Values below zero indicate more protective factors in comparison with risk factors, while values above zero indicate more risk factors in comparison with protective factors. Although continuous measures are often viewed as preferable to dichotomous variables, we followed the lead of previous researchers (Turner, Hartman, Exum, & Cullen, 2007) and dichotomized the risk and protective factors at the mean and then summed them together. We employed this measurement strategy for four main reasons. First, there is a significant body of research indicating that the most antisocial adolescents are saturated with multiple risk factors (Wachs, 2000). Focusing on just a single risk factor in isolation fails to delineate those who are highly antisocial versus those who are only marginally antisocial. Second, research has examined the effects that dichotomizing variables have in psychiatric and criminological studies (Farrington & Loeber, 2000). The results reveal that dichotomizing variables has very little effect on the substantive conclusions of the study. In fact, risk factors that are split at the mean tend to maintain a correlation of close to ‘‘1’’ with the original continuously measured variable and tend to have the same magnitude of effects on measures of antisocial phenotypes. Third, dichotomizing risk factors aids in the detection of interaction effects (Farrington & Loeber, 2000). From a statistical and methodological standpoint, if we had retained the original coding of the continuous variables, the data would not have been able to support an interaction analysis. Specifically, the cell frequencies of some genotype/protective-risk factor combinations would have been zero, which would reduce the statistical power needed to detect interactions and would essentially be estimating parameters where no cases exist. Fourth, we reestimated all of the models using different cut-points for the protective and risk factors (e.g., split at the median, split at one standard deviation above the mean) and the pattern of results was virtually identical to those reported when the risk and protective factors were split at the mean. We also reestimated the models by summing the risk and protective factors as continuous items (z-transformed). The substantive results were once again identical to the ones reported with the dichotomous items. As a result, the findings reported here are robust and consistent across different measurement strategies. 2.6. Statistical analysis The potential interaction between MAOA and the protectiverisk factor index was estimated by creating a multiplicative
282
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
interaction term between these two variables. Since the ever incarcerated variable is dichotomous, binary logistic regression models were estimated to test for interactions. Ordinary least squares (OLS) regression models were estimated to test for interactions between MAOA and the protective-risk factor index for the anger and hostility variable. Since MAOA is located on the X-chromosome, all statistical models were estimated separately for males and females. To help avoid population stratification effects due to race/ ethnicity, all statistical models included a dichotomous dummy variable for race (0 = Caucasian, 1 = African American). Finally, because some of the observations lacked independence (i.e., more than one sibling from the same household), all models were estimated using Huber/White standard errors. In no case were both twins from a monozygotic (MZ) twin pair included in the analyses (Haberstick et al., 2005).
3. Results To ensure that any interactions between MAOA and the protective-risk factor index were not the result of gene-environment correlation, we followed a two-step process. First, we tested for gene-environment correlation between MAOA and the protective-risk factor index by calculating difference in means tests (ANOVA for females and a t-test for males). No statistically significant differences in the average scores on the protective-risk factor index for females (F = .603, p > .05) or for males (t = 1.26, p > .05) were detected. Second, we examined whether the untransformed (i.e., not dichotomized) risk and protective factors were correlated with MAOA. As Table 2 shows, MAOA maintained a statistically significant association with two risk-protective factors for males (verbal skills and broken home) and only one for females (neighborhood disadvantage). We next proceeded to test for interactions between MAOA and the protective-risk factor index in predicting the probability of being incarcerated. Table 3 depicts the results of the logit models for females and males separately. For females, the interaction between MAOA and the protective-risk factor index, while nonsignificant, was marginally close to reaching the conventional significance level (p = .061). For males, the interaction between MAOA and the protective-risk factor index was not statistically significant. The results thus far provide little evidence that the protectiverisk factor index interacts with MAOA genotype to predict contact with the criminal justice system. The next models, however, examine whether MAOA interacts with the protective-risk factor index to predict variation in the anger and hostility scale. Table 4 displays the results of these models. The results for females can be seen in
the left-hand column and show that of all the predictor variables, only the protective-risk factor index was significantly related to scores on the anger and hostility scale. Specifically, higher scores on this scale corresponded to higher scores on anger and hostility. The right-hand column of Table 4 presents the results for males. Of particular interest was that the interaction between MAOA and the protective-risk factor index was a statistically significant predictor of scores on the anger and hostility scale. To probe this interaction more closely, we plotted scores on the anger and hostility scale (expressed as z-scores) as a function of MAOA genotype across different values on the protective-risk factor index. As can be seen in Fig. 1, for respondents in the low MAOA activity group, there was a steep increase in anger and hostility scores as scores on the protective-risk factor index increased (b = 0.12, p < .05). For respondents in the high MAOA activity group, the association between the protective-risk factor index and anger and hostility was not nearly as sharp (b = 0.03, p > .05). A post-hoc z-test (Paternoster, Brame, Mazerolle, & Piquero, 1998) confirmed that the slope for the protective-risk factor index was significantly greater for the low MAOA activity group in comparison with the high MAOA activity group (z = 1.91, p < .05, one-tailed test).
4. Discussion Contemporary molecular genetic research has underscored the close interplay between genetic polymorphisms and environmental factors in the production of human phenotypes. This is especially true for antisocial behaviors, where a growing body of research reveals that candidate genes for antisocial phenotypes tend to have their strongest effects when paired with certain environmental liabilities (Moffitt, 2005). The most studied gene-environment interaction in relation to human antisocial behavior involves the MAOA gene, which has been shown to interact with measures of childhood maltreatment to explain variance in measures of antisocial behaviors (Kim-Cohen et al., 2006). The current study expanded on this line of inquiry and examined whether MAOA would interact with a cumulative protective-risk factor index, measured in adolescence, to predict the lifetime probability of being incarcerated and to predict variation in the personality trait anger and hostility, measured in adulthood. Analysis of genotypic data drawn from the Add Health study revealed two key findings. First, for females there was not an association between MAOA genotype and incarceration or between MAOA genotype and anger and hostility. In addition, there was no evidence that MAOA genotype interacted with the protectiverisk factors index to predict the probability of being incarcerated or to predict scores on the anger and hostility scale. This finding
Table 2 Correlation matrix for selected Add Health study variables for females and males. X1 Anger and hostility Ever incarcerated MAOA Verbal skills GPA School attachment Social support Religiosity Neighborhood dis. Depression Maternal negativity Delinquent peers Broken home
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13
X2 .10*
.14* .04 .05 .13* .09* .12* .03 .04 .14* .12* .11* .02
.04 .13* .23* .17* .15* .08* .11* .08* .12* .24* .13*
X3
X4
X5
X6
X7
X8
X9
X10
X11
X12
X13
.02 .05
.05 .05 .04
.18* .11* .04 .37*
.13* .07* .02 .04 .27*
.18* .08* .01 .00 .27* .45*
.12* .07* .03 .02 .17* .22* .22*
.04 .07* .07* .33* .20* .04 .00 .08*
.20* .08* .01 .19* .35* .39* .51* .16* .08*
.11* .05 .05 .02 .21* .31* .61* .22* .00 .37*
.09* .15* .06 .04 .32* .23* .32* .27* .02 .32* .18*
.05 .08* .03 .18* .17* .11* .11* .12* .30* .13* .08* .09*
.07* .03 .01 .01 .05 .05 .01 .00 .00 .10*
.30* .05 .04 .11* .38* .29* .04 .04 .20*
.26* .20* .15* .21* .24* .17* .21* .24*
.45* .18* .07* .31* .24* .24* .12*
Note: Female estimates are above the diagonal and male estimates are below the diagonal. * Significant at the .05 level, two-tailed tests.
.23* .02 .34* .54* .29* .07*
.16* .06 .23* .24* .03
.15* .06 .02 .29*
.18* .20* .10*
.22* .02
.05
283
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284 Table 3 Logit models testing for gene environment interaction in the probability of being incarcerated for females and males. Females
MAOA Protective-risk factor index MAOA protective-risk factor index Race N
Males
B
SE
OR
Z
p
B
SE
OR
Z
.43 .10 .25 .14
.34 .10 .13 .48
1.54 1.10 1.28 .871 493
1.29 0.95 1.87 0.77
.198 .340 .061 .773
.12 .29 .07 .65
.26 .06 .11 .29
1.12 1.34 1.07 1.92 421
0.44 5.11 0.57 2.26
p .658 <.001 .571 .024
Note: Models were estimated using Huber/White standard errors. Race was coded dichotomously, where 0 = Caucasian and 1 = African American. For females, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles, 1 = the presence of one low MAOA activity allele, and 2 = the presence of two low MAOA activity alleles. For males, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles and 1 = the presence of one low MAOA activity allele.
Table 4 OLS regression models testing for gene environment interaction in the prediction of anger and hostility for females and males. Females
MAOA Protective-risk factor index MAOA protective-risk factor index Race N
Males
b
Beta
SE
t
.04 .26 .01 .13
.01 .23 .01 .02
.16 .05 .07 .30 493
0.26 5.12 0.16 0.43
p .792 <.001 .874 .664
b .50 .20 .33 .85
Beta .08 .15 .12 .12
SE
t
p
.29 .07 .14 .36 420
1.74 2.96 2.39 2.36
.083 .003 .017 .019
Note: Models were estimated using Huber/White standard errors. Race was coded dichotomously, where 0 = Caucasian and 1 = African American. For females, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles, 1 = the presence of one low MAOA activity allele, and 2 = the presence of two low MAOA activity alleles. For males, MAOA was coded such that 0 = the presence of zero low MAOA activity alleles and 1 = the presence of one low MAOA activity allele.
Fig. 1. The interaction between MAOA and the protective-risk factor index in the prediction of anger and hostility for males.
is consistent with prior research suggesting that the effect of MAOA on antisocial phenotypes may be gender-specific and may not extend to females (Beaver et al., 2010), but is inconsistent with research showing that MAOA genotype is related to anger in a sample of Korean females (Yang et al., 2007). Second, and in line with the findings garnered for females, there was not a statistically significant interaction between MAOA genotype and the protectiverisk factor index in predicting the odds of being incarcerated for males. However, there was a statistically significant interaction between MAOA genotype and the protective-risk factor index in the prediction of anger and hostility. Post-hoc probing analysis indicated that males who carried the low MAOA activity alleles, in comparison with males who carried the high MAOA activity alleles, scored the lowest on anger and hostility when exposed to the most
protective factors and the highest on anger and hostility when exposed to the most risk factors. This latter finding is directly in line with the differential susceptibility hypothesis which suggests that certain alleles, such as the low MAOA activity alleles, may be more appropriately viewed as plasticity alleles (Belsky & Pluess, 2009). These plasticity alleles act as biological markers that influence the degree to which a person is susceptible to environmental pressures. In the face of adverse environments, carriers of plasticity alleles are more likely to be affected negatively, while in the presence of advantageous environments, carriers of plasticity alleles are more likely to be affected positively. Although the current study was not designed to test directly the differential susceptibility hypothesis, the results are nonetheless consistent with it. Additionally, the interaction between MAOA genotype and the protective-risk factor index is consistent with recent experimental research revealing that carriers of the low MAOA activity alleles employ more aggression in response to a provocation (McDermott, Tingley, Cowden, Frazzetto, & Johnson, 2009). The results of our study indicate that levels of anger and hostility are higher for low MAOA activity allele carriers in comparison with high MAOA activity allele carriers, but only if they were exposed to multiple risk factors. As a result, it is quite possible that anger and hostility serve as an adaptive response to high-risk situations for those with the genetic predisposition for antisocial phenotypes (as measured by the low MAOA activity alleles). Exactly why the interaction between MAOA and the protectiverisk factor index was statistically significant for the anger and hostility scale, but not for the odds of being incarcerated is not immediately apparent. However, we do offer one possible explanation. Being incarcerated is the end result of a complex chain of events that is affected by a host of different factors, including the seriousness of the offense being committed, the number of prior arrests, and being apprehended by law enforcement. In addition, socioeconomic status also plays a role in the ultimate outcome of a criminal case, with those of higher socioeconomic status being able to afford the most high-powered lawyers. Given all of the contributing factors to incarceration, it is possible that the effects of MAOA are
284
K.M. Beaver et al. / Journal of Research in Personality 45 (2011) 279–284
either too small to detect or are overshadowed by other factors. Future research should explore this possibility more closely by estimating the differential effects that MAOA has across a range of antisocial phenotypes. The results of the current study need to be interpreted with caution in light of a number of limitations. First, although the Add Health study was designed to be nationally representative of American youths, the genotypic sample was not. As a result, it is quite possible that the results reported here would not generalize to the larger population of Americans or to populations outside of the United States. Replication studies using different samples that include more cases are needed to determine the robustness of these findings. Second, the measure of anger and hostility was based on only four self-reported items. Ideally, the scale would have been comprised of many more items scored by multiple raters, but because the Add Health data only included four items measuring anger and hostility, we were forced to use a four-item scale. Psychometric analysis of the scale, however, revealed that it was reliable. Nonetheless, follow-up research is needed to ensure that the results generated in the current study would be applicable to other measures of anger and hostility. Lastly, the protective-risk factor index included both environmental and individual-level factors. In this way, we were examining a host of factors that might be able to moderate the association between MAOA and anger and hostility. Therefore, caution should be used when treating this study as a direct test of gene-environment interaction because the ‘‘environment’’ component of this interaction was inclusive of individual-level factors. The factors included in the protective-risk factor index, as a result, were capturing genetic effects. As a result, the interaction between MAOA and the protective-risk factor index may really be detecting gene–gene interactions as opposed to gene-environment interactions. Whether the results of the current study would remain once these shortcomings are addressed is an open-empirical question awaiting future investigation. Acknowledgments This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/ addhealth). No direct support was received from grant P01HD31921 for this analysis. References Alia-Klein, N., Goldstein, R. Z., Tomasi, D., Woicik, P. A., Moeller, S. J., Williams, B., et al. (2009). Neural mechanisms of anger regulation as a function of genetic risk for violence. Emotion, 9, 385–396. Beaver, K. M., DeLisi, M., Vaughn, M. G., & Barnes, J. C. (2010). MAOA genotype is associated with gang membership and weapon use. Comprehensive Psychiatry, 51, 130–134. Belsky, J., & Pluess, M. (2009). Beyond diathesis-stress: Differential susceptibility to environmental influence. Psychological Bulletin, 135, 885–908.
Buckholtz, J. W., & Meyer-Lindenberg, A. (2008). MAOA and the neurogenetic architecture of human aggression. Trends in Neurosciences, 31, 120–129. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., et al. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297, 851–854. Deckert, J., Catalano, M., Syagailo, Y. V., Bosi, M., Okladnova, O., Di Bella, D., et al. (1999). Excess of high activity monoamine oxidase A gene promoter alleles in female patients with panic disorder. Human Molecular Genetics, 8, 621–624. Farrington, D. P., & Loeber, R. (2000). Some benefits of dichotomization in psychiatric and criminological research. Criminal Behaviour and Mental Health, 10, 100–122. Ferguson, C. J., & Beaver, K. M. (2009). Natural born killers: The genetic origins of extreme violence. Aggression and Violent Behavior, 14, 286–294. Gottesman, I. I., & Gould, T. D. (2003). The endophenotype concept in psychiatry: Etymology and strategic intentions. American Journal of Psychiatry, 160, 636–645. Gullone, E., & Moore, S. (2000). Adolescent risk-taking and the five-factor model of personality. Journal of Adolescence, 23, 393–407. Guo, G., Ou, X.-M., Roettger, M., & Shih, J. C. (2008). The VNTR 2 repeat in MAOA and delinquent behavior in adolescence and young adulthood: Associations and MAOA promoter activity. European Journal of Human Genetics, 16, 626–634. Haberstick, B. C., Lessem, J. M., Hopfer, C. J., Smolen, A., Ehringer, M. A., Timberlake, D., et al. (2005). Monoamine oxidase A (MAOA) and antisocial behaviors in the presence of childhood and adolescent maltreatment. American Journal of Medical Genetics, 135B, 59–64. Harris, K. M., Halpern, C. T., Smolen, A., & Haberstick, B. C. (2006). The national longitudinal study of adolescent health (Add Health) twin data. Twin Research and Human Genetics, 9, 988–997. Kim-Cohen, J., Caspi, A., Taylor, A., Williams, B., Newcombe, R., Craig, I. W., et al. (2006). MAOA, maltreatment, and gene-environment interaction predicting children’s mental health: New evidence and a meta-analysis. Molecular Psychiatry, 11, 903–913. Koller, G., Bondy, B., Preuss, U. W., Bottlender, M., & Soyka, M. (2003). No association between a polymorphism in the promoter region of the MAOA gene with antisocial personality traits in alcoholics. Alcohol and Alcoholism, 38, 31–34. McDermott, R., Tingley, D., Cowden, J., Frazzetto, G., & Johnson, D. D. P. (2009). Monoamine oxidase A gene (MAOA) predicts behavioral aggression following provocation. Proceedings of the National Academy of Sciences, 106, 2118–2123. Meyer-Lindenberg, A., Buckholtz, J. W., Kolachana, B., Hariri, A. R., Pezawas, L., Blasi, G., et al. (2006). Neural mechanisms of genetic risk for impulsivity and violence in humans. Proceedings of the National Academy of Sciences, 103, 6269–6274. Miller, J. D., Lynam, D. R., Widiger, T. A., & Leukefeld, C. (2001). Personality disorders as extreme variants of common personality dimensions: Can the five-factor model adequately represent psychopathy? Journal of Personality, 69, 253–276. Moffitt, T. E. (2005). The new look of behavioral genetics in developmental psychopathology: Gene-environment interplay in antisocial behaviors. Psychological Bulletin, 131, 533–554. Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36, 859–866. Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128, 490–529. Sabol, S., Hus, S., & Hamer, D. (1998). A functional polymorphism in the monoamine oxidase A gene promoter. Human Genetics, 103, 273–279. Samuels, J., Bienvenu, O. J., Cullen, B., Costa, P. T., Easton, W. W., & Nestadt, G. (2004). Personality dimensions and criminal arrest. Comprehensive Psychiatry, 45, 275–280. Turner, M. G., Hartman, J. L., Exum, M. L., & Cullen, F. T. (2007). Examining the cumulative effects of protective factors: Resiliency among a national sample of high-risk youths. Journal of Offender Rehabilitation, 46, 81–111. Udry, J. R. (2003). The national longitudinal study of adolescent health (Add Health), waves I and II, 1994–1996; wave III, 2001–2002 [Data file and documentation]. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill. Viding, E., & Frith, U. (2006). Genes for susceptibility to violence lurk in the brain. Proceedings of the National Academy of Sciences, 16, 6085–6086. Wachs, T. D. (2000). Necessary but not sufficient: The respective roles of single and multiple influences on individual development. Washington, DC: APA. Williams, L. M., Gatt, J. M., Kuan, S. A., Dobson-Stone, C., Palmer, D. M., Paul, R. H., et al. (2009). A polymorphism of the MAOA gene is associated with emotional brain markers and personality traits on an antisocial index. Neuropsychopharmacology, 34, 1797–1809. Yang, J.-W., Lee, S.-H., Ryu, S.-H., Lee, B.-C., Kim, S.-H., Joe, S.-H., et al. (2007). Association between monoamine oxidase A polymorphisms and anger-related personality traits in Korean women. Neuropsychobiology, 56, 19–23.