Gendered Pathways: Violent Childhood Maltreatment, Sex Exchange, and Drug Use

Abstract

Objective

Recent work has emphasized the role of violent victimization, along with risky contexts like sex exchange, in pathways to problems of externalizing and substance use in women. Nonetheless, few studies have empirically tested gender differences involving the roles of adversity factors (e.g., childhood violent maltreatment, sex exchange) in drug use patterns. The present study tested a model of gender differences in relationships between childhood physical and sexual abuse, sex exchange, and two indicators of drug use: engagement and symptoms of disorder.

Method

We recruited an ethnically-diverse sample of 304 (130 women) adults with recent histories of violence and/or drug use, who completed a substance use diagnostic interview, the Childhood Trauma Questionnaire, and a sex exchange questionnaire.

Results

First, structural equation modeling revealed that childhood sexual and physical abuse were related to increased drug engagement in women and men, respectively, above the influence of early childhood contextual variables (e.g., neighborhood, family) and age. Second, sexual abuse was related to sex exchange, which in turn was related to drug use symptoms in women but not men.

Conclusions

These data provide empirical support for distinct trauma-related pathways to drug use problems in men and women, which has implications for gendered explanations and prevention approaches.

Social and cultural forces play a substantial role in shaping the emergence and progression of externalizing problems (e.g., antisocial behavior, violence, substance use) in women and men. Greater research attention should focus upon the ways in which strongly gendered contexts, where the experiences of females and males differ, may differentially shape drug use trajectories. In this vein, histories of childhood maltreatment are theorized to affect males and females differently in relation to problem behaviors (Shin, Hong, & Hazen, 2010). This is especially the case for childhood sexual abuse, due to its often gendered nature that may disproportionally affect girls and women. In addition, recent theorists suggest that engagement in the highly gendered context of economic sex exchange (e.g., for money or goods) can contribute adversely to the ways that childhood abuse ultimately impacts externalizing behaviors, especially in women (Javdani, Sadeh, & Verona, 2011). For example, research in sociology and anthropology has implicated sex exchange as mediating the link between child sexual maltreatment and drug use outcomes. However, research to date is limited in that (a) gender differences are rarely explored (or only women are included) and (b) results are not reported for unique drug use outcomes, including episodic engagement in drug use versus addiction.

The goal of the present paper was to examine associations between a history of childhood maltreatment (sexual and physical), commercial sex exchange, and adult drug use in women and men. A novel feature of this paper involves the analysis of intersections between early maltreatment and sex exchange, a combination that may be particularly risky for drug use, and, in addition, provide clarity on distinct pathways for men and women. We capitalize on a sample with high rates of childhood trauma, sex exchange, and histories of violence and drug use to effectively test our hypotheses relative to studies using general population samples.

Gender and Drug Use

As the rates and consequences of drug use and substance use disorders (SUD) have grown among women (e.g., Kloos, Weller, Chan, & Weller, 2009; Chesney-Lind & Pasko, 2013), research attention has increasingly identified differences between males and females in the emergence of SUD. For example, a body of research has revealed important gender differences in the causal chain between first experimenting with drugs and eventually becoming dependent upon them (e.g., Cotto et al, 2010). That is, women seem to progress more rapidly from first use to regular use and addiction, and enter treatment with severity equal to or greater than males (e.g., Hernandez-Avila Rounsaville, & Kranzler 2004; Randall et al, 1999). This suggests that studies on substance use involving men and women should distinguish between different indicators of drug use outcomes (e.g., initiation, regular use and drug addiction) to understand factors unique to these experiences. Research suggests that drug initiation/experimentation is related to, yet distinguishable from addiction (e.g., Martin, Chung, & Langenbucher, 2008; Martin et al, 2014). As such, we investigate the roles of childhood maltreatment and sex exchange in adversely affecting illicit drug use outcomes, with focus on two distinct indices: level of drug engagement (e.g., early initiation, experimentation) and severity of SUD (e.g., drug abuse and dependence).

Early Maltreatment, Gender and Drug Use

Besides differences in the rate at which substance use problems are manifested in women and men, an important body of research has focused on identifying the role of early maltreatment on lifetime rates of illicit drug use (Afifi, Henriksen, Asmundson, & Sareen, 2012). The focus of much of this research has been on childhood maltreatment that can be characterized as “violent”, such as physical abuse (CPA) or sexual abuse (CSA), in that these types often involve coercion or force and violate the person’s physical integrity. These two forms of abuse are especially associated with post-traumatic symptoms (e.g., Farley & Barkan, 1998) and have been most often investigated in terms of gender differences (Javdani et al., 2014; Podgurski et al., 2014).

Results regarding gender differences in links between maltreatment and drug use are mixed, however. For example, a number of studies have indicated that childhood maltreatment has differing impacts on males and females in relation to later drug use, typically with greater impact on females (e.g., Heffner, Blom, & Anthenelli, 2011; Hyman et al, 2008; Verona & Sachs-Ericsson, 2005; Widom, Marmorstein, & White, 2006). Gender differences suggesting a stronger link between early trauma and adverse outcomes for women are particularly prevalent in clinical samples (Heffner et al, 2011; Hyman et al, 2008). In contrast, studies using community samples have yielded no gender differences, including a general population study of more than 30,000 adults (Afifi et al, 2012).

In making sense of the mixed literature, some researchers argue that CSA in particular plays a larger role in drug use outcomes in females relative to males (e.g., Shin et al., 2010), and has been associated with adult SUD in large-sample studies of women (Douglas et al., 2010; Kendler et al., 2000; however, see Hyman et al., 2006). However, there are very few studies in the area of drug use that directly compare men and women, and the studies that exist differ quite widely on sample characteristics and other methodological features. For instance, studies using general population samples typically fail to observe substantial gender differences in the CSA relationship to substance use in general (Molnar et al, 2001; Nelson et al, 2006; for a more nuanced set of findings, see Afifi et al, 2012). In contrast, investigations of clinical samples suggest that CSA is a stronger predictor of SUD in females (Shin et al, 2010: Hyman et al, 2008). Further research is needed to clarify gender differences in relationships between maltreatment involving CPA and CSA and drug use.

Mediating Role of Sex Exchange Relationships

The highly gendered context of sex exchange, where dominance and control reside primarily with male clients or partners, male police officers, and male pimps, is an especially salient area of analysis with respect to the cycle of victimization that marks some women’s lives (Javdani et al., 2011). Thus, this experience is likely to differentiate male and female pathways to drug use. Beyond sex exchange for drugs or money, persons often engage in transactional sex to obtain housing, bill assistance, and other material benefits. Although the latter is a potentially less risky way to obtain basic needs, all sex exchange contexts likely involve similar psychological effects (e.g., shame), power dynamics, and risk of violent victimization (Rekart, 2006).

Research has consistently evidenced links between CSA and later sex work (for a review, see Arriola, Louden, Doldren, & Fortenberry, 2005). For instance, in a longitudinal study of 574 children (half female) aging out of foster care, self-reported CSA was significantly predictive of later sex exchange, especially in females, even after controlling for child physical abuse and neglect (Ahrens, Katon, McCarty, Richardson, & Courtney, 2012). Similarly, cross-sectional research with female-only (Senn & Carey, 2010) and mostly male (Stoltz et al., 2007) young adults suggest that CSA, but not CPA or neglect, was associated with sexually risky behavior and sex work.

Once individuals are engaged in sex exchange, sex workers may use drugs in order to cope with the stress and degradation associated with their work (Romero-Daza, Weeks, & Singer, 2003). Sex workers report very high levels of violent victimization (e.g., rape, assault; Farley & Barkan, 1998; Surratt, Inciardi, Kurtz, & Kiley, 2004). Importantly, research supports taking a gendered focus on this topic, given that women are more likely to be purveyors of sex while males are more likely to be solicitors. That is, engagement in sex exchange relationships is more likely to lead to sexual victimization and other negative consequences in women and girls, relative to men in those relationships (WHO, 2005). Men report feelings of power or excitement in sex exchange relationships, where they are typically soliciting sex, whereas women report economic reasons for engaging in sex exchange (Logan, Cole, & Leukefeld, 2003). The links between sex work, victimization, and drug use are consistent in women, albeit complicated by issues of temporal order (e.g., El-Bassel et al., 2001). Although victimization has been shown to lead to future drug use (Kilpatrick et al., 1997), sex exchange is also an effective or reliable way to obtain drugs or money for drugs among drug-dependent individuals (Morrill, Kasten, Urato, & Larson, 2001). Thus, the relationship between sex exchange and drug use is bidirectional and mutually-reinforcing (see Cusick, 2006).

Finally, theoretical work supports the idea that sex exchange mediates the impact of childhood maltreatment on drug use outcomes. For example, work by Chesney-Lind and Pasko (2013) nicely illustrates the ways in which sexual abuse may motivate adolescents to escape from their homes and spend more time on the streets, where they are at greater risk than those who live at home for engaging in sex exchange to meet their basic needs. In turn, sex exchange is associated with high risk for drug use for all sex workers, most of whom are women (e.g., Surratt et al., 2004). Anthropological studies suggest that drug use is part of the culture of the sex industry, with drugs made accessible and often used as payment, particularly for street workers (Surratt et al., 2004). This work has mostly relied on qualitative analyses, using mostly female participants, providing rich information on women’s lived experiences. The present study will advance this work further by systematically analyzing the pattern of relationships between childhood maltreatment, sex exchange, and drug use as differentially explaining drug use in women and men.

Gaps in Knowledge and Present Study

As noted above, there is a growing literature on links between child maltreatment and sex exchange (Chesney-Lind & Pasko, 2013), as well as between sex exchange and different externalizing behaviors, including drug use problems (Romero-Daza et al., 2003). However, in many cases, studies have not conducted systematic analyses of these relationships nor compared men and women directly. To address these gaps, we examine the influence of two forms of violent child maltreatment (CSA and CPA) and sex exchange, directly comparing women and men, with regard to drug use outcomes. Our study also attempted to optimize design features to help detect gender differences in relationships between childhood maltreatment, sex exchange, and SUD, if they exist. First, as noted above, the review of the literature on CSA and drug use suggested that gender differences were more robustly observed when samples with comparatively high rates of maltreatment and multiple forms of illicit drug use were recruited, relative to studies with low base rate samples (e.g., general population studies; Molnar et al, 2001). Thus, we recruited persons with relatively high base rates of histories and behaviors of interest (e.g., maltreatment, drug use, sex exchange). Second, gender-specific effects of CSA on drug use might be most evident in relation to more severe or chronic forms of SUD (e.g. Shin et al, 2010; Widom et al, 2006). Thus, we moved beyond reporting findings in terms of simple binary categories of drug use problems (such as yes/no presence of SUD), in favor of dimensional assessments of drug use (e.g., Afifi et al, 2012; Nelson et al, 2006). In addition, we separated episodic drug initiation and experimentation (engagement) from SUD or problems (symptoms). As mentioned above, this distinction has relevance for understanding gender differences (Cotto et al, 2010), although it has not been explored in this area. Third, a major limitation in much of the retrospective maltreatment literature has been the failure to account for childhood context variables (e.g. family, neighborhood; Maniglio, 2009) to create a robust test of childhood maltreatment effects (e.g., Rind, Tromovitch, & Bauserman, 1998). Without these covariates, we cannot conclude that violent maltreatment contributes uniquely to drug use. Thus, we included assessments of early socioeconomic status (SES), neighborhood, and family context in our analyses to adjust for broad-level childhood context in examining the unique roles of maltreatment and sex exchange.

Finally, no research of which we are aware has included both childhood abuse and sex exchange simultaneously to explain their unique or shared influences on drug use and SUD symptoms. Although the present study is cross-sectional, and temporal order cannot be firmly established, mediation analyses may provide much-needed quantitative evidence for a model of sex exchange accounting for links between childhood abuse and drug use.

Hypotheses

There are three primary hypotheses based on existing theory and data. First, we expected both forms of childhood abuse to relate to drug use dimensions above the influence of early-childhood context. Second, given previous research with high-risk samples, we expected that CSA would be more strongly related to drug use in women than men. However, we were tentative in this hypothesis given the mixed literature. Third, we expected that the context of sex exchange will mediate the influence of CSA on drug use dimensions, and more so in women than men, as per models implicating sexual abuse with turning to life on the streets (i.e., engagement in sex exchange) and drug use (e.g., Chesney-Lind & Pasko, 2013). Because no research has compared child abuse risk factors for episodic drug engagement versus drug use symptoms in the same study, we did not develop hypotheses in that regard. All hypotheses were tested with a multi-group structural equation modeling approach.

Method

Participants

The data for this study are part of a larger on-going study on gender, violence, and substance use. Three hundred and four participants (130 women) with recent histories of violence and/or substance use were recruited from the community as well as legal or treatment agencies (e.g., parole, substance use treatment). They ranged in age from 18 to 62 (M = 34.6; SD = 11.9), although most were in their 30s and 40s. About half of participants identified as African-American (n = 145; 48%), over a third as Caucasian (n = 110; 37%), followed by Mixed Ethnicity (n = 21; 7%), Asian descent (n = 9; 3%), Hispanic (n = 9; 3%), and Native American (n = 3; 1%), with 3 persons (1%) not reporting. In terms of education history, about 19% of participants dropped out during or before high school, 25% were high school graduates or the equivalent (i.e., GED), and the rest attended at least some college (56%). Over half of the sample (56%) was unemployed at the time of assessment. Men and women did not differ on any of these demographic variables.

Procedures

Participants were recruited through legal or treatment agencies (8%), newspaper and flyer advertisements (56%), and other sources (36%; e.g., word of mouth) targeting individuals with recent histories (last year) of illicit drug use and/or violence perpetration. Participants obtained through these recruitment sources did not differ from each other on any of the primary variables, with the exception that those recruited through flyers/ads reported more CPA than those recruited from other sources, F (2,296) = 4.7, p < .01. Thus, recruitment source was not included in analyses.

Eligible participants were invited to complete a 2.5–3 hour research session and were paid $35–40 for this session. Interviewers were advanced graduate or undergraduate student assistants trained in diagnostic assessment and supervised by a Ph.D.-level licensed clinical psychologist. Using a structured protocol, a short demographic interview was followed by the Structured Clinical Interview for DSM-IV-TR Axis I Disorders (SCID-I; First, Spitzer, Gibbon, & Williams, 2002), which included thorough assessment of drug use history and symptoms. A certificate of confidentiality granted by the National Institute of Drug Abuse provided protection of participants’ data in the case of a subpoena/court order. All staff were thoroughly trained in confidentiality and risk assessment. All participants signed informed consent forms approved by the university institutional review board.

Measures

Drug use dimensions

Drug use information was gathered from the SCID-I substance use module. First, illicit drug engagement was defined according to a) age of initiation and b) level of experimentation across 6 drug categories (i.e., sedative, stimulant, cocaine, cannabis, hallucinogen, opioids). Age of initiation was calculated as the earliest age of use across all drug types, excluding cannabis¹ (range = ages 9 – 52). Broad experimentation was calculated as the sum level of use across all drugs categories (1=never used, 2=used at least twice but not more than 10 times in 1 month, 3=used more than 10 times in 1 month), with high scores indicating frequent experimentation. Although men reported slightly more experimentation than women [Ms = 14.0 and 12.8, respectively; t (302) = -2.9, p < .004], they showed the same range of scores on experimentation (6–24) and reported similar age of drug use initiation [Ms = 18.1 and 19.1; t (254) = .10, ns].

Second, drug use disorder symptoms were assessed according to the DSM-IV-TR (APA, 2000) symptoms for lifetime drug abuse (e.g., failure to fulfill obligations; physically dangerous use; α = .66) and drug dependence (e.g., larger amounts of drug taken, withdrawal, tolerance; α = .79), with respect to self-reported drug of choice (4.2% stimulants, 10.0% opioids, 38.4% cocaine, 42.9% cannabis, and 5.5% other; α = .82). Women and men did not differ in their endorsement of drug of choice [χ (6, N = 289) = 8.4, ns] or a lifetime diagnosis of drug dependence (72% women and 71% men). Secondary ratings of diagnoses of drug abuse and dependence were available for 157 (51.6%) participants, showing good inter-rater reliability (ICC = .95 and .83, respectively, ps< .05). We calculated symptom counts to designate drug use symptoms, a strategy that has shown adequate construct validity in previous research (e.g., correlate with antisocial behavior and personality in expected patterns; Verona & Sachs-Ericsson, 2005; Verona, Sachs-Ericsson, & Joiner, 2004). Women and men did not differ on number of drug use disorder symptoms endorsed [t (300) = 0.4, ns].

A principal axis factor analysis, with promax rotation (for non-orthogonal factors), verified two distinct drug use dimensions. The first factor (eigenvalue = 1.79; 44.71% of the variance) involved high loadings from drug abuse (.79) and drug dependence (.69), and low loadings from the other two variables (< .14). The second factor (eigenvalue = 1.31; 32.84% of the variance) involved high loadings from drug experimentation (.68) and age of initiation (−.67) and low loadings from the symptom variables (< .06). These distinct dimensions are supported by the literature that distinguishes drug experimentation and addiction (Volkow & Li, 2004; Burke, Burke, Regier, & Rae, 1990). Thus, we z-score standardized age of initiation and drug experimentation and averaged their standardized scores to create a drug engagement composite score, and we did the same for drug abuse and dependence symptoms to create a drug use symptoms composite score.

Childhood physical and sexual abuse

Information regarding childhood maltreatment was gathered via two different sources. First, participants were interviewed about a range of traumatic events they may have experienced and their age at the time the trauma occurred within the posttraumatic stress disorder module of the SCID-I diagnostic assessment, based on procedures used in the National Comorbidity Study (Molnar et al., 2001). CPA and CSA were calculated as the number of different types of physical assault (i.e., “attacked”, “physically abused”, “threatened with a weapon”) and sexual assault (i.e., “raped”, “molested”), experienced before the age of 15. This measure of childhood maltreatment has been shown to have adequate validity in terms of correlations with psychopathology and other adversity variables (Sachs-Ericsson et al., 2006; Verona & Sachs-Ericsson, 2005). According to this interview, women reported more CSA than men (43% and 10%), χ (2, N = 304) = 43.6, p < .01, whereas rates of CPA were similar across the genders (35% and 34%, respectively), χ (2, N = 304) = 5.3, ns.

Second, participants completed the Childhood Trauma Questionnaire (CTQ; Bernstein et al., 2003), a 28-item self-report inventory asking participants to report frequency (5-point scale ranging from “Never true” to “Always true”) of experiences indicative of maltreatment when they were “growing up”. The CTQ has evidenced adequate internal consistency and test-retest reliability in previous studies, and strong construct validity including convergence with other trauma assessments (Bernstein et al., 1994). The present study focused on the CTQ physical abuse (e.g., “hit so hard or with belts/objects”, “beaten”; α = .77) and sexual abuse (e.g., “touched in sexual way”; α = .94) subscales. Gender differences on these CTQ subscales paralleled the trauma interview, with women endorsing more sexual abuse than men (Ms = 9.4 and 6.1), t (297) = 6.0, p < .01, whereas scores across the genders were similar on physical abuse (Ms = 8.9 and 8.6, respectively), t (297) = 0.8, ns.

The relationship between the CTQ subscales and respective trauma interview variables were r’s = .53 and .74, respectively. Thus, scores from the interview and CTQ subscales were standardized and averaged to create cross-measure indices of childhood physical abuse (CPA) and sexual abuse (CSA).

Early childhood context

Participants provided information on their early neighborhood, family and socioeconomic context, with specific instructions to consider the time before age 12 (constituting early childhood). First, neighborhood context was assessed using the 18-item Neighborhood Environment Scale (NES; Crum, Lillie-Blanton, & Anthony, 1996), involving perceptions of childhood neighborhood quality (e.g., “there were plenty of safe places to work or play”, “people…always took care of each other”, “there were abandoned or boarded-up buildings”), measured on a True/False scale. Higher total scores reflected higher-quality neighborhoods (α = .64). The NES is an established measure of neighborhood quality and has been associated with community engagement (Tiernan et al., 2013).

Second, family context was measured using a reverse-scored version of the 9-item Conflict subscale of the Family Environment Scale (Moos, 1974), which asked them to rate on a True/False scale the level of anger and conflict in their family growing up (e.g., “we fought a lot in our family”, “family members often criticized each other”). This measure demonstrates adequate construct validity in substance using samples (Sanford et al., 1999). For purposes of consistency with other early childhood context variables, higher scores in our study indicated better family context (α = .81).

Third, SES was based on the occupation and education level of male and female parent/guardians of participants. Occupation level was coded according to Hollingshead (1975) on a 1 to 9 scale (i.e., from service worker to executive and major professional), while education level was coded on a 1 to 6 scale (i.e., from dropped out before high school to graduate degree). The maximum value across the two caregivers (if there were two) was used. The correlation between education and occupation was r = .51, and each was standardized and averaged together to create a composite SES variable, with higher scores representing more affluence. The Neighborhood, Family and SES variables were included to evaluate the role of childhood abuse above early childhood context.

Sex exchange relationships

Participants completed a sex exchange questionnaire meant to measure frequency of sex exchange contacts/relationships (not distinguishing between soliciting or providing sex), adapted from previous research on risky sexual behavior among substance users (Darke, Hall, Heather, Ward, & Wodak, 1991; Lejuez, Simmons, Aklin, Daughters, & Dvir, 2004). Two items asked them to report on the number of partners with whom they had penetrative sex (e.g., vaginal, anal, and/or oral) in exchange for “money or drugs” and “shelter, food, or other basic necessities” in their lifetime (0=none, 7=60+). The average of these two items was used as a measure of sex exchange (α = .75).² This type of assessment of sex work context has demonstrated collateral validity (e.g., items corroborated by sexual partners of participants) and good reliability in previous research (e.g., Lejuez et al., 2004). Men and women did not differ on their involvement in sex exchange relationships, with rates of at least one sex exchange partner at 38% and 33%, respectively.

Planned Analyses

Skew and kurtosis were calculated on model variables to assess the assumption of normality. Sex exchange was highly skewed (skewness = 2.2) and kurtotic (kurtosis = 5.4), justifying the calculation of a normalized variable using Blom’s transformation. Thirteen univariate outliers (+-3 SDs above the mean) were identified across all variables and set to missing. Age was included as a covariate predicting drug dimensions, because older persons had more time to use drugs.

We used Mplus 5.1 (Muthén & Muthén, 2008), full information maximum likelihood (FIML) estimation, for structural equation modeling. A total of 20 participants (12 women) had missing data on, at the most, three variables, which were estimated through FIML. In addition to chi-square fit statistics, we used the standardized root-mean-square residual (SRMR) and the comparative fit index (CFI), with minimum cutoff value of .95 for the CFI and a maximum cutoff value close to 0.06 for the SRMR to indicate good model fit (Hu & Bentler, 1999).

First, we ran a direct effects model (see Figure 1 top panel), containing direct effects from CPA and CSA, early context (SES, family, neighborhood), and participant age (covariate) to the two drug use dimensions to evaluate hypothesis 1 (influence of childhood physical and sexual abuse above other early childhood context variables). The independent variables were allowed to covary with each other, as were the dependent variables (although these correlations are not pictured in Figure 1). Second, multi-group analyses were conducted to evaluate hypothesis 2, regarding gender differences in associations between childhood abuse and drug use.

Figure 1. Parameter Estimates for Direct Effects Model in Women and Men.

Note: Unstandardized regression weights (and standard errors) for women are left of the diagonal and in bold; estimates for men are right of the diagonal and in regular font. Correlations between the independent variables are not shown, but are similar to those reported in Table 1.

* p < .05, ** p < .01

Hypothesis 3 evaluated the role of sex exchange in explaining relationships between childhood abuse and drug use dimensions using the direct-indirect effects model (see Figure 1 bottom panel). This model was the same as the direct effects model except that sex exchange was included, with paths added (a) from the two childhood abuse variables to sex exchange and (b) from sex exchange to the two drug use dimensions. To test hypothesis 3, we analyzed whether sex exchange significantly mediated the relationship between childhood abuse and drug use by bootstrapping 1000 samples to create bias-corrected 95% confidence intervals around indirect effects. This increases power over other methods while controlling Type I error (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002).

Results

Bivariate Correlations and Descriptive Statistics

Bivariate correlations between all key variables are presented in Table 1, separately for women and men (below and above the diagonal), to compare with the multivariate relationships obtained in the structural models below. Descriptive statistics for the whole sample are also displayed in Table 1. Women and men did not differ on most primary variables, except that women reported lower parental SES, more CSA, and less drug engagement, ts (295–302) = −2.4, 7.1, and −3.2, respectively, ps < .05.

Table 1.

Correlations Matrix for Men and Women and Descriptive Statistics for Primary Variables

	Neighborhood	SES	Family	Age	CSA	CPA	Sex Exchange	Drug Engage	Drug Sx
1. Neighborhood	__	.26**	.40**	.08	−.12	−.44**	−.02	.01	−.07
2. SES	.27**	__	.04	−.39**	−.07	−.16*	−.31**	.07	−.09
3. Family	.35**	.03	__	.15*	−.04	−.49**	.08	−.15	−.15
4. Age	.18*	−.31**	.07	__	.11	.01	.35**	.01	.28**
5. CSA	−.02	−.22*	−.28**	.19*	__	.27**	.17*	−.09	.10
6. CPA	−.28**	−.03	−.58**	−.16	.34**	__	.11	.22**	.12
7. Sex Exchange	−.19*	−.26**	−.04	.20*	.30**	.06	__	.06	.24**
8. Drug Engage	.07	.07	−.11	.04	.28**	.06	.08	__	.26**
9. Drug Sx	.05	−.20*	.05	.44**	.22*	−.13	.36**	.38**	__

	Neighborhood	SES	Family	Age	CSA	CPA	Sex Exchange	Drug Engage	Drug Sx
M	30.26	−0.18	14.00	34.64	0.01	0.00	1.34	−0.09	−0.00
SD	5.09	0.88	2.79	11.94	0.94	0.88	2.37	0.88	0.91
Range	17 – 36	−1.70 – 1.71	9 – 18	18 – 62	−0.71 – 3.36	−1.28 – 3.41	0 – 14	−0.37 – 2.45	−1.66 – 1.44

Note. Correlations for women are below the diagonal and in bold, and correlations for men are above the diagonal and regular font. Neighborhood = Neighborhood Environment Scale (Crum et al, 1996); SES = socioeconomic status during early childhood; Family Conflict = Family Environment Scales – Conflict subscale (Moos, 1974); CSA and CPA = childhood sexual and physical abuse, respectively, composites of Childhood Trauma Questionnaire (Bernstein et al, 2003) and trauma history interview; Sex exchange = lifetime number of exchange partners; Drug Engage = drug use engagement (age of initiation and variety of experimentation); Drug Sx = composite of DSM-IV drug abuse and drug dependence symptom counts.

^*p < .05;

^**p < .01

Hypothesis 1

Our first hypothesis was that childhood abuse would explain variance in drug use dimensions above early childhood context (i.e., neighborhood, family, and SES) and current age. This model is just-identified, i.e., χ²(0, N = 304) = 0.00. Multivariate regression weights are displayed in Figure 1 (top panel), separately for men and women. The model accounted for 16% and 23% of the variance in women’s drug engagement and drug symptoms, respectively, and 8% and 12% in men. For women, CSA but not CPA was positively associated with both drug use dimensions, and SES and age were positively related to drug engagement and symptoms, respectively. Among men, CPA but not CSA was positively associated with drug engagement but not symptoms, with age and family positively related to drug symptoms. Thus, our first hypothesis was confirmed, although the relationship between violent maltreatment types and drug use was different for men and women.³

Hypothesis 2

Our second hypothesis was that there would be gender-specific effects in our model. To test for gender differences, we conducted multi-group analyses and computed a chi-square difference test on the chi square values obtained from the unconstrained direct effects model above (i.e., gender variant) and a series of progressively more constrained models (e.g., constraining regression weights and covariances; constraining regression weights, covariances, variances and residuals, etc.). Each model was compared to the previous, less-constrained model. The models with more parameter constraints across gender, including constraining only regression weights, showed progressively poorer fits to the data compared to the unconstrained just-identified model, Δχ²s (8–12) = 24.93 – 73.04, ps < .05 (CFIs = .00 – .86 and SRMR = .04 – .46 for constrained models). Thus, we concluded that there were gender differences in model parameters and compared men and women on particular hypothesized paths.

We compared the gender-unconstrained model with models that constrained the relationship between CSA or CPA and the drug use dimensions across gender. These analyses revealed a significantly poorer fit for the two models, respectively, that constrained the regression weights for CSA and CPA to be equal across gender, Δχ²s (2) = 11.12 and 8.41, ps < .05. For both child abuse variables, this was the case when predicting drug engagement, Δχ²s (1) = 10.94 and 8.46, ps < .05, but not for drug symptoms, Δχ²s (1) = −0.59 and 0.53, ns. Thus, CSA was a stronger correlate of drug engagement (but not symptoms) in women than men, and CPA was a stronger correlate of drug engagement in men relative to women (see Figure 1 top panel). Comparisons of CSA versus CPA within women showed that CSA was a significantly stronger correlate of drug use engagement than CPA in women, Δχ² (1) = 7.87, p < .01, although base rates of CSA and CPA were similar among women (see Table 1).

Hypothesis 3

Our third hypothesis examined the intermediate role of sex exchange relationships in links between childhood abuse and drug use dimensions, which was evaluated using the direct-indirect effects model (see Figure 1 bottom panel). As expected, the model showed adequate fit for women, χ² (4) = 12.18, p = .02, CFI=.91, SRMR=.04, accounting for 17%, 27%, and 9% of the variance in drug engagement, drug symptoms, and sex exchange, respectively. As with the direct effects model, constraining regression weights to be equal across genders produced a detrimental fit to the data relative to the unconstrained model, Δχ² (16) = 29.60, p < .05. This confirms gender differences in this model.

The same path coefficients were significant in this model as in the direct effects model in women (compare to top panel of Figure 1), except that the path from sexual abuse to drug symptoms for women was no longer significant when sex exchange was included in the model. In addition, CSA but not CPA was related to sex exchange in women, and in turn sex exchange was related to drug use symptoms but not engagement. The indirect effect from CSA to drug symptoms via sex exchange was significant, B=.07, SE = .03, p = .02, but not the indirect effect to drug engagement, B=.02, SE = .02, p = .43 nor indirect effects from CPA to drug engagement or symptoms, Bs = −.01 and .02, SEs = .01 and .03, ps > .80. We bootstrapped 95% confidence intervals around the indirect effects in the direct-indirect effects model. The confidence intervals for the indirect effect of CSA to drug symptoms via sex exchange did not include zero, CI [0.03, 0.16], while the other indirect effects did include zero.

Also as expected, this model had poor fit to the men’s data, χ² (4) = 36.05, p < .01, CFI=.57, SRMR=.07, accounting for 9%, 13% and 2% of the variance in the three variables. In men, CPA maintained its significant relationship with drug engagement. However, neither CPA nor CSA was related to sex exchange, although, like in women, sex exchange was associated with drug symptoms. No indirect effects were significant, Bs =.01 – .08, SEs =.01 – .03, ps > .37.⁴

Given the cross-sectional nature of the data, it is possible that childhood abuse leads to drug symptoms, which in turn, increase risk for sex exchange (i.e., drug use leads to sex work). Thus, we evaluated a model reversing directions of the mediator and drug dimensions, in which child abuse predicted drug use engagement and symptoms, with these drug use dimensions in turn predicting sex exchange (covariates included). This model showed poor fit in women, χ² (8) = 32.82, p < .001, CFI=.73, SRMR=.06. Bootstrapping revealed no significant mediation by drug use in the relationship between childhood abuse and sex exchange, although the indirect effect of CSA via drug symptoms to sex exchange was marginally significant in women (B = .04, SE = .03).

Discussion

This study is the first to highlight unique and combined influences of early childhood context, childhood maltreatment, and sex exchange relationships on engagement and symptoms of drug use among men and women. Main findings suggest that CSA and CPA are both associated with drug engagement (i.e., age of drug initiation and level of experimentation), but that CSA is a stronger correlate for women and CPA for men. Sex exchange relationships mediated the relationship between CSA and drug use symptoms in women, supporting a unique pathway to externalizing problems, particularly substance use in women. Ours was a particularly robust test of retrospective childhood abuse relations with drug use, given that we adjusted for the impact of several early childhood contexts.

Research Directions and Implications

Women’s pathways

Our findings refine theoretical pathways implicating early victimization and its reverberating effects in women’s pathways to problem behaviors. First, CSA may impact the earlier initiation and broad experimentation with illicit drugs (the drug engagement pathway) and indirectly influence symptoms of drug abuse and dependence through its association with sex exchange (the drug problems pathway). These patterns make sense developmentally; early sexual abuse likely influences behaviors that occur earlier in life (during teen years), such as initiation and experimentation with drugs. In contrast, sex exchange tends to occur later in life (later teens or early adulthood) and likely influences persistent SUD. An alternate explanation is that drug addiction leads to engagement in sex exchange to finance the addiction. Although it cannot be ruled out with cross-sectional data, post-hoc analyses suggested that reversed-direction model did not fit the data as well. In all likelihood, the relationship is reciprocal and mutually-reinforcing (McClanahan et al., 1999).

Comparing our findings to previous theory and literature, the data are consistent with a model in which a confluence of factors combine in a dynamic way to explain the detrimental effects of sexual abuse and sex exchange for women. Women in our sample, who come from disadvantaged backgrounds, are likely exposed to particular social risks due to sexual abuse, which often precedes running away from home (Chesney-Lind & Pasko, 2013), dropping out of high school (Zierler et al 1991), and teenage pregnancy (Lalor & McElvaney, 2010). In turn, these factors increase the risk that women will rely on sex exchange to support themselves and their families, and the sex exchange context, in turn, places them at risk for further violence (Farlan & Barkan, 1998; Surratt et al., 2004). In tandem, these findings support what has been referred to as an integrated risk-sequelae model of victimization (e.g., Higgins & McCabe, 1994) and suggest that the risk incurred by victimization is often the result of reverberating effects of the violent trauma on survivors’ lives (Blakemore, 2006). This pathway is relevant even when women engage in sex exchange after experimenting with drugs (McClanahan et al., 1999). Further, our findings are consistent with a multiple threshold model, suggesting that women may require a greater level of etiologic liability to express the same amount of externalizing behavior, in this case SUD (Rhee & Waldman, 2002). Our findings point to more than genetic liability but also greater environmental liability in women, potentially by virtue of women’s typically-lower social status (Javdani, Sadeh, & Verona, 2011).

Men’s pathways

Our findings are also consistent with a larger literature implicating specific types of experiences as important to men’s paths. In our study, early family context was related to drug use symptoms in men, but not women, suggesting that a hostile family context combined with CPA are connected to drug use in men. This finding is reminiscent of the “coercion model” advanced by Patterson and colleagues (Patterson, 1982), which posits that chronic antisocial behavior results from a breakdown of family management practices during childhood, with the use of coercion and harsh punishment exacerbating conduct problems in youth, particularly in boys (Patterson, Reid, & Dishion, 1992; Wiesner, Capaldi, & Patterson, 2003). A similar familial pattern may be at work in links between CPA and substance use in men, with both social learning and traumagenic effects of violent abuse influencing the intergenerational transmission of SUD. Indeed, this dynamic is consistent with the “risky families” model (Repetti, Taylor, & Seeman, 2002), which purports that family contexts characterized by conflict-ridden, aggressive, and neglectful relationships create or enhance vulnerabilities in children by disrupting the stress-response system and placing children at risk for later mental health problems, particularly substance abuse. It is important for future longitudinal designs to disentangle the child and parent effects in such dynamics.

Importantly, although sex exchange relationships did not relate to childhood abuse in men, they related to drug use symptoms. There are two important points about this finding. First, our findings are contrary to other research showing that CSA does relate to higher substance use and sex exchange among men who have sex with men (Kalichman, Gore-Felton, Benotsch, Cage, & Rompa, 2004). Thus, our inability to detect the latter relationship among the men in our sample may be due to sampling strategy (e.g., low numbers of men who have sex with men in our sample). Second, our findings linking sex exchange relationships and SUD symptoms in men suggest a need for attention to the connection between sex exchange in men (even as solicitors and not providers of the sex) and drug use. Indeed, scholars studying the effects of hyper-masculine cultural contexts have noted the lack of attention paid to the detrimental effects of soliciting sex work, a social act most likely engaged by male clients and that forms a part of a larger context involving violence, morbidity, and mortality (Nisbett & Cohen, 1996; WHO, 2005). Our findings corroborate these previous theories by suggesting a link between engaging in sex exchange relationships and drug abuse and dependence for adult men.

Limitations and Strengths

The results of this study should be tempered by its limitations. Limitations of the present study include the cross-sectional design and retrospective reports of childhood maltreatment. Given the paucity of research in this area, these cross-sectional data can generate hypotheses for future longitudinal work. In addition, our assessment of sex work did not distinguish between providing versus receiving compensation in exchange for sex. Though this distinction is important, our focus was on examining experiences that were relevant for both men and women, and solicitation and provision of sex work have both been tied with risk for SUD (Messina, Wish, & Nemes, 2000). Also, it is possible that our broad recruitment of individuals from different sources could have affected our results, although we did so to increase the heterogeneity and generalizability of our sample (e.g., not only treatment seekers). Finally, there are several other potentially important variables, such as sexual orientation and protective factors (e.g., resilience) that were not assessed, but are important to investigate in future work as potential intervening variables.

Several strengths of this study were also evident. First, our sample was comprised of individuals with high base rates of key variables, increasing the generalizability of results to clinical populations. Second, this study is one of the first to examine the contributions of childhood abuse and sex exchange on distinct drug use dimensions (i.e., engagement vs. symptoms) in a sample with multiple externalizing problems. Furthermore, highly validated clinical assessments were used to examine drug use outcomes, and sophisticated modeling techniques allowed for thorough examination of holistic and specific gender differences. Results underscore the need to examine gender-specific pathways to problem behaviors informed by models that incorporate gender-relevant risk factors.

Clinical and Policy Implications

The limitations of the study notwithstanding, several clinical implications are suggested by our findings. First, it would be standard practice for clinicians to evaluate clients’ risky lifestyles, including sex exchange, and how these relate to SUD and relapse risk while in treatment. Our results for women suggest that cumulative victimization, including CSA and sex exchange experiences, may be tightly connected to development of SUD and that on-going victimization influences maintenance of substance use. Thus, promising treatments will include trauma-informed care that attends to prior victimization, use of drugs for purposes of coping, and on-going violent interpersonal experiences (including in the context of sex exchange; McClanahan et al., 1999). Other promising approaches include advocacy and community-level interventions that increase access to resources (e.g., to reduce the need for engagement in sex work; increase access to safe sex practices to reduce transmission of STIs; e.g., Sullivan & Bybee, 1999), to mitigate potential risk factors for continued SUD. Finally, policy reform in favor of specialized courts for both youth (e.g., sex trafficking courts) and adults (e.g., drug and domestic violence) have demonstrated particular success in reducing both SUD and recidivism for drug and violent offenses. These courts may be better equipped to target the complexity inherent in women’s pathways to SUD (e.g., Feder, 1997).

Figure 2. Parameter Estimates for Direct-Indirect Effects Model in Women and Men.

Note: Unstandardized regression weights (and standard errors) for women are left of the diagonal and in bold; estimates for men are right of the diagonal and in regular font. Correlations between the independent variables are not shown, but are similar to those reported in Table 1.

* p < .05, ** p < .01