According to the Centers for Disease Control and Prevention, there were an estimated 56,300 new HIV infections [95% confidence interval (CI): 48,200 to 64,500] in the United States in 2006, an estimated incidence rate of 22.8 per 100,000 population (95% CI: 19.5 to 26.1).1,2 Men comprised 73% of incident HIV infections, an estimated 41,400 cases (95% CI: 35,100 to 47,700), with an estimated incidence rate of 34.3 per 100,000 population (95% CI: 29.1 to 39.5).
Stressful or traumatic exposures in childhood during critical periods of cognitive, affective, and physical development have long-term, developmental, psychological, and neurobiological sequelae,3-6 which may have implications for HIV infection in men across the life course. Consistent with an ecological-transactional model developed by Cicchetti and Lynch,7 risk and protective factors and processes at different ecological levels (ie, individual, culture, family, community) not only determine the likelihood of experiencing traumatic stress in childhood and adolescence but also influence the course and development of subsequent psychopathologic outcomes (eg, posttraumatic stress symptoms or negative health behaviors, including HIV risk, and rate of progression of HIV disease characteristics).8,9
Early experiences of violence, including sexual and physical abuse, have been shown to be highly associated with HIV sexual risk behavior across populations, including adolescents,10 persons living with HIV,11,12 and women.13-15 Although less is known about childhood violence and HIV risk in men,16 a growing body of research has also examined the association of violence and HIV risk among men who have sex with men (MSM), particularly in the domain of childhood sexual abuse (CSA).17-25 For example, using data from the EXPLORE Study, a behavioral intervention trial conducted in 6 US cities over 48 months with HIV infection as the primary efficacy outcome, Mimiaga et al21 found that of the 4295 MSM participants enrolled, 40% had a history of CSA. Participants with a history of CSA were at increased risk for HIV infection over study follow-up (adjusted hazards ratio = 1.30, 95% CI: 1.02 to 1.69). A significant association was seen between history of CSA and unprotected anal sex [adjusted odds ratio (aOR) = 1.24, 95% CI: 1.12 to 1.36] and serodiscordant unprotected anal sex (aOR = 1.30, 95% CI: 1.18 to 1.43). Moreover, among participants reporting CSA, the EXPLORE intervention had no effect in reducing HIV infection rates compared to those without CSA. These data provide good initial evidence that the presence of CSA history in MSM interferes with their ability to derive benefit from traditional HIV prevention interventions and hence may attenuate an intervention effect.21
In one of the largest studies conducted to date with US men (not MSM specific), DiIorio et al16 examined the association between unwanted sexual activity during childhood and HIV risk behavior among a sample of 2676 predominantly African American and Latino/Hispanic men recruited for the National Institutes of Mental Health Multisite HIV Prevention Trial. Approximately 25% of men reported unwanted or uninvited sexual activity before 13 years of age; unwanted or uninvited sexual activity during childhood was significantly associated with greater frequency of unprotected sexual acts and a greater number of sexual partners during the past 90 days.16
Much of the child trauma literature and the HIV behavioral risk research that considers the role of childhood traumatic stress takes a “single-event” approach and focuses on individual types of victimization, for example, sexual abuse or unwanted sex.3 However, as Finkelhor et al26 suggest, consideration of “complete victimization profiles” (p 8) offers a more nuanced and contextually relevant understanding of the ecological context in which violence and victimization occur, including the reality that many children experience chronic, complex, or multiple traumatic stressors in their lives, making victimization “more of a ‘condition’ than an ‘event’.” The role of other early life violent experiences beyond CSA, such as physical abuse, verbal abuse, or witnessing parental intimate partner violence, warrants additional consideration, especially in understanding HIV risk among MSM.19 Moreover, understanding the social ecological context in which early experiences of violence occur, such as childhood poverty that represents an additional environmental stressor, remains an area in need of further research in relation to HIV infection.
The aim of this study is two-fold: (1) to evaluate the prevalence of early violent experiences among a nationally representative sample of men in the United States, including co-occurring traumatic stressors, and (2) to investigate the role of early experiences of violence in relation to incident HIV infection, including the potential mediating role of posttraumatic stress disorder (PTSD) in relation to incident HIV infection. Three hypotheses were offered relating to the study's overarching goals (Fig. 1): (1) early life experiences of violence were expected to be associated with incident HIV infection (Fig. 1A); (2) we expected that a significant interaction effect (eg, effect modification) would be observed by factors shown to pattern along disparities in HIV infection among men (sexual minority orientation and race/ethnicity) (Fig. 1B); and (3) we hypothesized that the relationship between early life experiences of violence and incident HIV infection would be mediated by PTSD, shown previously to be highly associated with both HIV sexual risk behavior and HIV infection20,21,27 (Fig. 1C).
Study Sample and Procedures
Cross-sectional data were analyzed from wave 2 of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), 2004-2005. NESARC is a longitudinal, population-based, epidemiologic study implemented to estimate the prevalence of alcohol, drug, and mental health disorders among the general US civilian population (noninstitutionalized) 18 years and older. Study design and methods have been described in detail elsewhere.28-33
Inclusion and Exclusion Criteria
NESARC wave 2 included a total of 34,653 participants from the general US civilian population (noninstitutionalized) 18 years and older. Respondents who reported (1) their sex being male and (2) being white, black (non-Hispanic), or Hispanic/Latino (any race) were included in the present analysis (n = 13,898). Excluded participants were females (n = 20,089) and males who were American Indian/Alaska Native (non-Hispanic) (n = 240) or Asian/Native Hawaiian/Other Pacific Islanders (non-Hispanic) (n = 426).
A complete case analysis was conducted whereby participants who were missing data on any variables included in analyses were excluded from the final data analytic sample. A total of 624 participants were excluded due to missing data or unknown responses as follows: 386 did not report lifetime sexual behavior; 146 did not report family's socioeconomic position (SEP); early childhood violent stressors: 88 physical abuse, 98 sexual abuse, 103 verbal violence, 91 neglect, 120 witness parental violence; and 96 did not self-report HIV serostatus. We chose not to include those participants who reported “unknown” abuse measures in the “no” category as a method of limiting misclassification bias in our analysis.34,35 The final data analytic sample included 13,274 men.
Primary Outcome: HIV Incidence (Past 12 Months)
Participants were asked to self-report whether they had received a diagnosis of HIV in the past 12 months: “Have you tested positive for HIV, the virus that causes AIDS in the past year?” Response options were “yes,” “no,” or “unknown.” If participants answered “yes,” they were coded as an incident HIV case (1); if participants answered “no,” they were not (0). A total of 96 males who responded “unknown” were excluded from this analysis.
Primary Exposure: Violence and Victimization (Before Age 18)
Four domains of violence and victimization in childhood and adolescence were defined as exposures in the present analysis, consistent with the developmental psychopathology literature,6 which defines child maltreatment as follows: (1) physical abuse, which involves the nonaccidental infliction of bodily injury; (2) sexual abuse, which includes sexual contact or attempted contact between a child and caregiver or other adult; (3) neglect, which refers to the failure to provide minimum care and lack of supervision; and (4) emotional maltreatment, which involves persistent foiling of a child's basic emotional needs. Variables included in analyses followed this developmental psychopathological framework6 and provided rationale for our operationalization of domains of violence and victimization.
Domain 1: Physical Abuse
Participants were asked, “Were you physically attacked, beaten, or injured before age 18 by a parent or caretaker?” Response options were “yes,” “no,” or “unknown.” Participants who answered “yes” were considered to have experienced physical abuse (1) and those who answered “no” were not (0). “Unknown” responses (n = 88) were excluded from analyses.
Domain 2: Sexual Abuse
Participants were asked whether they had ever been “sexually assaulted, molested, raped, or experienced unwanted sex” in their lifetime. Response options were “yes,” “no,” or “unknown.” Participants who answered “yes” were considered to have experienced sexual abuse (1) and those who answered “no” were not (0). “Unknown” responses (n = 98) were excluded from analyses. Participants who reported sexual abuse were subsequently asked the age (in years) when this first happened to them. Participants reporting sexual violence at age 18 or below were considered to have experienced sexual abuse as children. The focus on childhood experiences of sexual violence operationalized at age 18 or below was selected because we did not have a measure of who perpetrated the sexual abuse, and research suggests an increased likelihood that early age abuse is perpetrated by a caregiver or adult.36 Adolescent sexual violence experiences were not included because sexual violence experienced during the teenage years has an increased likelihood of being perpetrated by dating/acquaintances,36 and we were interested in family-related developmental exposures.
Domain 3: Neglect
Participants were asked, “Were you seriously neglected before age 18 by a parent or caretaker?” Participants who answered “yes” to the self-defined measure were considered to have experienced neglect (1) and those who answered “no” did not (0). “Unknown” responses (n = 91) were excluded from analyses.
Domain 4: Emotional Abuse
Emotional abuse was assessed using 2 single-item questions about (1) verbal abuse by a caregiver and (2) witnessing paternal violence. (1) Verbal abuse: Participants were asked, “Before age 18, how often did parent/caregiver swear, insult, or say hurtful things to you?” Response options were 1 = never; 2 = almost never; 3 = sometimes; 4 = fairly often; 5 = very often; and 9 = unknown. Participants indicating “1 = never” were considered to have not experienced verbal abuse; participants reporting “2 = almost never,” “3 = sometimes,” “4 = fairly often,” or “5 = very often” were considered to have experienced verbal abuse. Unknowns (n = 103) were excluded. (2) Witnessing paternal violence: Participants were asked, “Before age 18, how often did your father/other adult male push, grab, slap, or throw something at your mother?” Response options were 1 = never; 2 = almost never; 3 = sometimes; 4 = fairly often; 5 = very often; and 9 = unknown. Participants indicating “1 = never” were considered to have not experienced verbal abuse; participants reporting “2 = almost never,” “3 = sometimes,” “4 = fairly often,” or “5 = very often” were considered to have witnessed paternal violence. Unknowns (n = 120) were excluded.
Index of Early Experiences of Violence
A summary index was constructed using the 5 possible early violence exposures (physical abuse, sexual abuse, neglect, verbal violence, and witnessing paternal violence). Participants were classified as having experienced zero (0), one (1), two (2), or 3 or more (3+) events based on their self-reported experiences of early violence.
Potential Effect Modifiers: Sexual Orientation and Race/Ethnicity
Wave 2 of NESARC collected sexual orientation measures along 3 dimensions: identity, attraction, and behavior. For the present study, participants were classified as sexual minority based on lifetime sexual behavior. Participants were asked: “In your lifetime, have you had sex with only males, only females, both males and females, or have you never had sex?” Four response categories were given: only males, only females, both males and females, and never had sex. Men who engage in sexual behavior with other men (whether this is engaging with exclusively men only or with both men and women) represent 53% of new incident HIV cases in 2006 in the United States.1,2 Therefore, sexual behavior was dichotomously operationalized: men who self-reported “only males” and “both males and females” were considered “MSM” (coded as “1”), and those who responded “only females” were considered “MSW” (coded as “0”). A total of 386 males who did not indicate their lifetime sexual behavior were excluded from this analysis.
As mentioned previously, due to the small number of respondents in the American Indian/Alaskan Native, Asian/Asian Pacific Islander, and other race/ethnicity categories, especially stratified by HIV infection, these subgroups were not included in the present data analytic sample (666 males were excluded due to being non-Hispanic American/Indian/Alaska Native, Asian/Native Hawaiian/Other Pacific Islanders, or others). Race/ethnicity was initially coded as (1) white (non-Hispanic), (2) black (non-Hispanic), and (3) Latino/Hispanic (any race). Because of the disproportionate rates of HIV infection among black and Latino/Hispanic individuals on a population level in the United States1,2 and to obtain additional power for subsequent regression analyses, race/ethnicity was dichotomized into non-white that included black (non-Hispanic) and Latino/Hispanic (coded as “1”) and white (coded as “0”).
Potential Mediating Variable: PTSD
Posttraumatic Stress Disorder
The Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV) was used to assess lifetime Diagnostic andStatistical Manual of Mental Disorders (Fourth Edition) mood and anxiety disorders, including PTSD (past 12 months). The AUDADIS-IV is a structured interview schedule administered by lay interviewers (ie, nonclinicians). The reliability and validity of this instrument for mood disorders have been previously tested and described in detail.28 The diagnosis of PTSD using the AUDADIS-IV meets the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) clinical significance criterion: “Symptoms cause clinically significant distress or impairment in social, occupational, or other important areas of functioning.”28,29 Incidence of PTSD was dichotomously operationalized (yes/no).
Participants were asked to report their age in years (continuous).
Low Educational Attainment
Participants who completed less than high school or a high school degree/General Education Development were categorized as low educational attainment (coded as “1”); participants with some college, associate's degree, college degree, or above were classified as high educational attainment (coded as “0”). This variable represents adult low SEP for this analysis.
Low Childhood SEP
Family receipt of government assistance before age 18 was used as a proxy for childhood SEP, operationalized dichotomously as lower SEP vs higher SEP. Participants who reported that their family received money from government assistance programs before age 18 were coded as lower SEP (coded as “1”); respondents who reported that their families did not receive money from government assistance programs before age 18 were coded as higher SEP (coded as “0”). Participants (N = 146) were excluded from this analysis because they did not report family's SEP.
The primary outcome dichotomously assessed was incident HIV infection (yes/no). Statistical analyses were conducted using SAS version 9.2 statistical software. For all analyses, statistical significance was predetermined at the α = 0.05 level.
Descriptive statistics were obtained for all variables included in the analysis. As previously mentioned, a complete case analysis (listwise deletion) was conducted. A bivariate analysis of excluded cases was conducted to assess for statistical patterns by excluded cases. Because cases excluded from this analysis seemed missing at random, it was determined that subsequent analyses would produce unbiased parameter estimates, assuming no unmeasured confounding.
Bivariate associations were obtained for the primary exposures of interest and all covariates by incident HIV infection. Proportional differences were examined using χ2 tests or Fisher exact test when appropriate. The SURVEYLOGISTIC procedure in SAS was used to fit a taxonomy of multivariable regression models to investigate the proposed study hypotheses. PROC SURVEYLOGISTIC estimates binomial logistic regression models for discrete response survey data by the method of maximum likelihood and incorporates complex survey sample designs, including designs with stratification, clustering, and unequal weighting to allow for statistically valid inferences.
Analysis for Hypothesis 1 and Hypothesis 2
To empirically test whether early experiences of violence were associated with incident HIV infection and whether there was an interaction effect by race/ethnicity or sexual behavior, 3 adjusted logistic regression models were fit that included the covariates age, race/ethnicity, sexual behavior, educational attainment, and childhood poverty: (1) model 1: Incident HIV infection was regressed on early experiences of violence to examine the main effects of early violence (Fig. 1A); (2) models 2 and 3: Incident HIV infection was regressed on early experiences of violence, and 2 multiplicative interaction terms were added to examine whether effect modification by sexual orientation or race (MSM*early experiences of violence and non-white*early experiences of violence) was seen in the effects of early violence on risk for HIV (Fig. 1B).
We also calculated the attributable risk fraction (ARF) or the weighted proportion of incident HIV cases among participants exposed to early violence vs not exposed to provide an overall sense of the magnitude of the effects of early life violent events on HIV incidence among US men. The ARF was calculated as (the weighted incidence of HIV in the exposed minus the weighted incidence of HIV in the unexposed)/(the weighted incidence of HIV in the exposed). The ARF was multiplied by 100 to get the percentage of incident HIV cases attributable to early life violence.
Analysis for Hypothesis 3
Our third hypothesis, that PTSD would mediate the relationship between early life stressors and incident HIV infection (Fig. 1B), was tested by fitting a taxonomy of logistic regression models, adjusted for the covariates age, race/ethnicity, sexual behavior, educational attainment, and childhood poverty. We followed the procedures for testing mediator effects as outlined by Baron and Kenny37 and updated by Kraemer et al.38 The adjusted models were as follows: (1) model 1: Incident HIV infection was regressed on early experiences of violence (path c: this was the same model estimated in hypothesis 1 above); (2) model 2: Early experiences of violence were regressed on PTSD (path a); (3) model 3: Incident HIV infection was regressed on PTSD (path b); and (4) model 4: Incident HIV infection was simultaneously regressed on early life events and PTSD (path c).
Weights were included to ensure the sample resembled the general population. The NESARC sample was weighted to adjust for the probabilities of selection of a sample housing unit or housing unit equivalent from the group quarters' sampling frame, nonresponse at the household and person levels, the selection of 1 person per household, and oversampling of young adults. Once weighted, the data were adjusted to be representative of the US population for region, age, sex, race, and ethnicity, based on the 2000 Census.29 The weighting procedure used ratio estimation in which sample estimates are adjusted to independent estimates of the national population by age, race, sex, and ethnicity. This weighting adjustment aimed to correct for bias due to sampling undercoverage, and derivation of weights has been described in detail elsewhere.29
Table 1 presents the characteristics of the study sample (N = 13,274 men), stratified by incident HIV cases vs HIV-uninfected cases. Less than 1% of the sample (0.35%) reported being diagnosed with HIV in the prior 12 months (weighted estimate), which is consistent with the general US population past year HIV incidence.1
Overall, the sample was 23.9% non-white, including 10.8% black (non-Hispanic) and 13.1% Latino/Hispanic. The majority (62.9%) of the sample had a high school education or below. A small proportion (3.9%) of men reported being behaviorally gay or bisexual (ie, had engaged in sex with another man in their lifetime); and 12.7% of men reported that their families received government assistance before the age of 18.
Factors significantly associated with HIV infection in the past 12 months in bivariate analyses were being non-white (P < 0.0001), reporting sexual behavior with men only or with both women and men compared to with women only (P < 0.0001), and reporting lower childhood SEP as a child (P = 0.0004).
Childhood Traumatic Stressors and PTSD Diagnosis
Overall, 3.0% of men reported physical abuse, 2.3% sexual abuse, 2.6% neglect, 28.2% verbal violence, and 13.4% witnessed paternal violence against their mother. More than one-third (33.5%) reported 1 or more childhood traumatic stressors: 21.7% of men reported 1 stressor, 8.6% reported 2 stressors, and 3.2% reported 3 or more early violent stressors. As shown in Table 1, a higher proportion of men with incident HIV infection experienced physical abuse (16.1% vs 2.9%), sexual abuse (11.9% vs 2.3%), neglect (10.8% vs 2.6%), and witnessed paternal violence against their mother (22.8% vs 13.3%) compared with HIV-uninfected men (P < 0.0001). Statistically significant differences were not observed by HIV status for verbal violence (31.2% vs 28.2%; P = 0.20).
Overall, 4.2% of the sample had a Diagnostic andStatistical Manual of Mental Disorders (Fourth Edition, Text Revision) diagnosis of PTSD (Table 1). A significantly higher proportion of men with incident HIV infection received a PTSD diagnosis in the past year compared with HIV-uninfected men (25.7% vs 4.1%; P < 0.0001).
Multivariable Logistic Regression Models: Hypotheses 1 and 2
Table 2 presents adjusted multivariable logistic regression models testing hypotheses 1 and 2.
Number of Early Life Experiences of Violence (Table 2, Model 1)
In model 1, consistent with hypothesis 1, the number of early life violent experiences reported was a significant predictor of 12-month incident HIV infection (aOR = 1.32; 95% CI: 1.16 to 1.50), adjusting for age, race, educational attainment, childhood SEP, and sexual behavior. In other words, for each additional early life violent event, the odds of HIV infection were 1.32 times the odds not becoming HIV infected in the past year, adjusting for relevant covariates. Also significant in this model were being non-white (aOR = 2.09; 95% CI: 1.73 to 2.53), being MSM (aOR = 11.53; 95% CI: 9.45 to 14.06), and having lower levels of education (aOR = 2.72; 95% CI: 2.41 to 3.07).
Interaction Effects of Early Violent Stressors (Table 2, Models 2 and 3)
As expected, a statistically significant interaction effect was observed for MSM with early violent experiences (aOR = 1.65, 95% CI: 1.17 to 2.34). As illustrated in Figure 2A, MSM (behaviorally gay or bisexual) with a higher number of early life experiences of violence were, on average, 65% more likely to report incident HIV infection in the past 12 months compared with men who report having sex with women only with the same number of early life violent experiences.
There was also a significant interaction effect observed for early violent stressors by race/ethnicity (aOR = 0.71; 95% CI: 0.56 to 0.90). As illustrated in Figure 2B showing results from model 3, non-white men with a higher number of early life experiences of violence were, on average, 29% less likely to report incident HIV infection in the past 12 months compared to white men with the same number of early life violent experiences.
ARF Due to Early Life Violent Events
For the calculation of the ARF, we used a dichotomous indicator of experiencing any early life violence, operationalized as none (0 events) compared with any early life violent stressful events (1, 2, 3, or more events). We found that among those exposed to 1, 2, or 3+ violent stressors in childhood (n = 4530), 30 (0.39%) had an incident HIV diagnosis in the past 12 months (proportion using weighted frequencies: 116915/29870247), and among those who were not exposed to violence in childhood (n = 8744), 33 (0.32%) were diagnosed with HIV (proportion using weighted frequencies = 190624/59219188). Therefore, the estimated ARF was 0.1776 [ie, (0.003914096 − 0.003218957)/(0.003914096)], indicating that 17.76% of incident HIV cases in the sample would have been averted if early childhood violent events were eliminated.
Mediation by PTSD: Hypothesis 3
We also sought to assess whether PTSD plays a mediating role in the relationship between early life violence and HIV incidence. Results from the mediational analysis for the whole sample are presented in Table 2 (models 4, 5, and 6). Each additional early life violent event was associated with a change of odds of past year PTSD (model 4; aOR = 1.87; 95% CI: 1.78 to 1.96). Not adjusting for early life violent events, meeting the criteria for PTSD was associated with incident HIV infection (model 5; aOR = 6.22; 95% CI: 4.95 to 7.82). There was evidence that PTSD partially mediated the relationship between and early life events and HIV (model 6). Compared with model 1, the effect of early life violent events was attenuated when adding PTSD to the model (model 6; aOR = 1.14; 95% CI: 1.02 to 1.28), and PTSD remained a significant predictor of HIV infection (aOR = 5.75; 95% CI: 4.76 to 6.95).
Consistent with national HIV surveillance data,1 significant disparities in HIV infection in the past 12 months were observed among racial/ethnic minority men and MSM in the present study, and experiencing early life violent stressors was associated with incident HIV infection. Findings further provide evidence that experiencing early life violence plays a particularly influential role in incident HIV infection for MSM. Moreover, corroborating previous research,18,19,22-24 this study assessed the multiplicative effect of concomitant early life violent stressors and risk of HIV infection among US men, with attention to MSM compared with heterosexual men, and between white and racial/ethnic minority males.
Figure 2A demonstrates that the risk of incident HIV infection is only slightly increased for prototypical heterosexual men of average race, age, educational attainment, and childhood SEP. Those with zero early violent events have a fitted odds of recent HIV infection of 0.0020, whereas those who reported 3 or more violent events have a fitted odds of 0.0024. However, for MSM, we found effect modification of the association between early life violence and risk of HIV, such that prototypical MSM reporting zero early violent events have a fitted odds of recent HIV infection of 0.0022, whereas the fitted odds for those reporting 3 or more violent events is 0.0117. These data indicate that the relationship of early life stressors and HIV infection are particularly robust for MSM, suggesting that the sequelae of childhood violence may play an influential role in population-level disparities in HIV infection among MSM populations. In contrast, although black/Hispanic men were overall more likely to report incident HIV infection, the significant interaction effect observed by race/ethnicity as shown in Figure 2B suggests that the fitted odds of incident HIV infection for the prototypical racial/ethnic minority males reporting early violent events was substantially lower than the fitted odds for white males. This suggests that the role of early life violence and subsequent developmental pathways to HIV infection differ substantially for black/Hispanic men than for white men.
We hypothesized, based on the timing of our exposure variables, that PTSD would likely be a mediator between early life violent experiences and HIV. This hypothesis is consistent with other research, suggesting that both traumatic events18,19,23,24 and PTSD are associated with increased sexual risk behaviors20,21,27 and that PTSD may mediate the primary relationship between trauma and sexual risk behavior. We found evidence of PTSD as a mediator of the relationship between early life violent events and incident HIV infection. Given the interaction effects we observed by sexual orientation and race/ethnicity with the number of early life violent events, we would like to conduct mediational analyses stratified by sexual orientation and race/ethnicity to investigate whether the mediating role of PTSD appeared to differ substantively across these stratified models. For example, it is possible that the pathways and mechanisms involved in early life experiences of violence and HIV infection differ for white men compared with racial/ethnic minority men. However, due to limitations in sample size and the small number of HIV cases, these comparisons were not possible in the present study. Additional research is needed to understand the mechanisms of action, including resilience processes,39 which may be operative by sexual orientation and race/ethnicity and which have implications for public health research, practice, and clinical interventions, including HIV prevention and PTSD treatment.17,40
Finally, an additional noteworthy finding is that when examining childhood poverty as part of the social-ecological context of violence and HIV infection, we found no association with receipt of governmental assistance before age 18 and incident HIV infection, after adjusting for early violent events and trauma. Previous literature has documented that family's economic position may be an important predictor of exposure to early violent events such as community violence, gang violence, etc.41,42 Indeed, in our mediational analysis, family receipt of government assistance was associated with PTSD diagnosis. However, there is mixed evidence as to whether individual-level childhood violence, such as child abuse and exposure to parental violence, is related to SEP.43 There is also evidence that neighborhood-level poverty (ie, living in a poorer compared with higher SEP neighborhood) is predictive of HIV infection, particularly among black men,44 but little evidence concerning individual childhood SEP. As stated previously, we did not find evidence that the receipt of government assistance during childhood was associated with incident HIV in the past year in a national sample of men. The most robust predictors of HIV infection were found to be violent experiences in childhood and PTSD diagnosis, especially for MSM.
Limitations of this analysis should be considered in the interpretation of results. First, participants self-reported whether they had been diagnosed with HIV in the past 12 months (primary outcome), and no biological markers were available to verify self-reported incident infection, which could have resulted in misclassification bias.45 Nonetheless, the weighted HIV incidence of our study sample was consistent with Centers for Disease Control and Prevention national estimates of HIV incidence.2 Second, sexual behavior (ie, MSM vs non-MSM) was defined using lifetime criteria that may not reflect recent sexual behavior patterns. There was no assessment of recent HIV sexual risk behavior (eg, unprotected vaginal or anal intercourse) in the NESARC data; therefore, we could not describe or consider recent behavioral risk patterns that may have led to incident HIV infections in our study.
Third, the dichotomous measure, government assistance, as a blunt proxy for childhood SEP may not fully represent all the ways in which childhood SEP could affect the relationship of early violence experiences and HIV incidence. For example, it was not possible to assess the duration of receipt of government assistance before age 18, which might be a more sensitive indicator of poverty in childhood. Further research utilizing different definitions, conceptualizations, and measures of individual childhood SEP, and alternatively individual adult SEP, is needed to assess the mechanisms through which family socioeconomics may function as a mediator or confounder between early violence and subsequent HIV infection. Multilevel studies would allow for further investigation of both neighborhood-level and individual-level SEP and exposure violence through the life course, allowing for a more nuanced understanding of the socioeconomic determinants of HIV infection.
We note that other studies, for example, those using the Adverse Childhood Experiences Study,46,47 have defined adverse childhood events (before age 18) differently than we have here; however, several questions were excluded from the NESARC survey that precluded us from operationalizing violence and victimization, consistent with prior research to date.48 An additional limitation of the present study is that all experiences of early violence are subject to recall bias that may lead to an underestimation of the association between early violent events and recent HIV infection. Although this is a cross-sectional analysis with limited ability to infer temporality of relationships, we restricted the analysis to violent experiences before age 18 to capture earlier life experiences while HIV incidence was restricted to the last 12 months. We also assumed that the diagnosis of PTSD was the result of at least one of the early childhood violence exposures. Unfortunately, we were not able to consider the specific developmental stage of the child or adolescent when violent stressors were experienced using this large population-level data set. Experiences of violence depend on emotional, cognitive, and physical capacities present in a child or youth's particular developmental period, which influence how the individual makes sense of and responds to traumatic stressors.3-5 More research is warranted that allows for investigation of developmental periods, with attention to whether there are “critical periods” during which exposure to early violent events especially influences development and patterns of HIV risk-taking behaviors, with particular attention to sexual or racial/ethnic heterogeneity.49
In addition, PTSD may not accurately capture other relevant mental health sequelae resulting from traumatic events and influencing development of PTSD symptoms. Furthermore, lack of social support or revictimization in adulthood and nonwestern-based notions of traumatic stress, which may be particularly relevant for immigrant black and Latino populations, warrant additional research attention.
Finally, a strong assumption we made was that any missing data were missing completely at random or that the probability of any variable was missing did not depend on any other variables in the fitted regression models or on the missing values themselves.50 Under this assumption, and assuming no unmeasured confounding, our method of using listwise deletion to handle missing data allows for valid estimation of parameters without directly modeling the missing data (ie, the SEs estimated using listwise deletion are typically accurate estimates of the true SEs).50
Public Health Implications
This nationally representative study found that experiencing early life violence plays a particularly influential role in HIV infection for men, accounting for 17.73% of incident HIV infections. Interventions are urgently needed that address the long-term sequelae of childhood and adolescent violent events. Of particular importance is providing HIV prevention interventions and additional support to (1) all men, regardless of sexual behavior or racial/ethnic minority status, if they meet the criteria for PTSD; and (2) adolescent and young adult men who experienced early life violent events and therefore have a higher odds of acquiring HIV and who may benefit from HIV prevention interventions at earlier ages.
Many validated cognitive behavioral therapy interventions have been shown to be effective for the treatment of mood and anxiety disorders,51 including cognitive behavioral therapy for traumatized children or adolescents52-54 and adults.55 Incorporating HIV prevention into these evidence-based psychotherapeutic treatments for youth, or adapting these validated treatments as part of HIV prevention interventions that target young adult men, represents an important area of future intervention development research, especially given that mental health concerns not only contribute to HIV risk but also likely interfere with the uptake of HIV behavioral interventions for men and for MSM in particular.17
Last, an ecological-transactional model7 posits that negative developmental consequences (eg, development of posttraumatic stress symptoms or emotion regulation deficits that may pattern alongside HIV risk behavior) emerge when vulnerabilities outweigh protective factors. The social determinants (such as early violent stressors) of early behavioral patterns that may place individuals at risk of HIV infection early on in the life course deserve additional investigation. The present study demonstrates the patterning of incident HIV infection by early experiences of violence and mediated through PTSD diagnosis. More research is needed to assess other mechanisms through which early childhood experiences of violence and trauma-related mental health disorder may confer additional HIV risk through increased risk-taking behavior, particularly among minority populations who continue to bear a disproportionate burden of HIV infection in the United States.
2. Hall HI, Song R, Rhodes P, et al. Estimation of HIV incidence in the United States. JAMA. 2008;300:520-529.
3. Chu AT, Lieberman AF. Clinical implications of traumatic stress from birth to age five. Annu Rev Clin Psychol. 2010;6:469-494.
4. Marans S, Adelman A. Experiencing violence in a developmental context. In: Osofsky J, ed. Children in a Violent Society. New York, NY: Guildford Press; 1997:202-222.
5. Lieberman A, VanHorn P. Coping with danger: the stress-trauma continuum. In: Lieberman A, VanHorn P, eds. Psychotherapy With Infants and Young Children: Repairing the Effects of Stress and Trauma in Early Attachment. New York, NY: Guildford Press; 2008:35-64.
6. Cicchetti D, Toth SL. Child maltreatment. Annu Rev Clin Psychol. 2005;1:409-438.
7. Cicchetti D, Lynch M. Toward an ecological/transactional model of community violence and child maltreatment: consequences for children's development. Psychiatry. 1993;56:96-118.
8. Leserman J. Role of depression, stress, and trauma in HIV disease progression. Psychosom Med. 2008;70:539-545.
9. Leserman J, Ironson G, O'Cleirigh C, et al. Stressful life events and adherence in HIV. AIDS Patient Care STDS. 2008;22:403-411.
10. Houck CD, Nugent NR, Lescano CM, et al. Sexual abuse and sexual risk behavior: beyond the impact of psychiatric problems. J Pediatr Psychol. 2010;35:473-483.
11. Henny KD, Kidder DP, Stall R, et al. Physical and sexual abuse among homeless and unstably housed adults living with HIV: prevalence and associated risks. AIDS Behav. 2007;11:842-853.
12. Sikkema KJ, Hansen NB, Meade CS, et al. Psychosocial predictors of sexual HIV transmission risk behavior among HIV-positive adults with a sexual abuse history in childhood. Arch Sex Behav. 2009;38:121-134.
13. Cohen M, Deamant C, Barkan S, et al. Domestic violence and childhood sexual abuse in HIV-infected women and women at risk for HIV. Am J Public Health. 2000;90:560-565.
14. Parillo KM, Freeman RC, Collier K, et al. Association between early sexual abuse and adult HIV-risky sexual behaviors among community-recruited women. Child Abuse Negl. 2001;25:335-346.
15. Wyatt GE, Myers HF, Williams JK, et al. Does a history of trauma contribute to HIV risk for women of color? Implications for prevention and policy. Am J Public Health. 2002;92:660-665.
16. DiIorio C, Hartwell T, Hansen N. Childhood sexual abuse and risk behaviors among men at high risk for HIV infection. Am J Public Health. 2002;92:214-219.
17. Safren SA, Reisner SL, Herrick A, et al. Mental health and HIV risk in men who have sex with men. J Acquir Immune Defic Syndr. 2010;55:S74-S77.
18. Paul JP, Catania J, Pollack L, et al. Understanding childhood sexual abuse as a predictor of sexual risk-taking among men who have sex with men: the Urban Men's Health Study. Child Abuse Negl. 2001;25:557-584.
19. Relf MV, Huang B, Campbell J, et al. Gay identity, interpersonal violence, and HIV risk behaviors: an empirical test of theoretical relationships among a probability-based sample of urban men who have sex with men. J Assoc Nurses AIDS Care. 2004;15:14-26.
20. Kalichman SC, Gore-Felton C, Benotsch E, et al. Trauma symptoms, sexual behaviors, and substance abuse: correlates of childhood sexual abuse and HIV risks among men who have sex with men. J Child Sex Abus. 2004;13:1-15.
21. Mimiaga MJ, Noonan E, Donnell D, et al. Childhood sexual abuse is highly associated with HIV risk-taking behavior and infection among MSM in the EXPLORE Study. J Acquir Immune Defic Syndr. 2009;51:340-348.
22. Arreola SG, Neilands TB, Pollack LM, et al. Higher prevalence of childhood sexual abuse among Latino men who have sex with men than non-Latino men who have sex with men: data from the Urban Men's Health Study. Child Abuse Negl. 2005;29:285-290.
23. Catania JA, Paul J, Osmond D, et al. Mediators of childhood sexual abuse and high-risk sex among men-who-have-sex-with-men. Child Abuse Negl. 2008;32:925-940.
24. Relf MV. Childhood sexual abuse in men who have sex with men: the current state of the science. J Assoc Nurses AIDS Care. 2001;12:20-29.
25. Jinich S, Paul J, Stall R, et al. Childhood sexual abuse and HIV risk-taking behavior among gay and bisexual men. AIDS Behav. 1998;2:41-51.
26. Finkelhor D, Ormrod RK, Turner HA. Poly-victimization: a neglected component in child victimization. Child Abuse Negl. 2007;31:7-26.
27. Reisner SL, Mimiaga MJ, Safren SA, et al. Stressful or traumatic life events, post-traumatic stress disorder (PTSD) symptoms, and HIV sexual risk taking among men who have sex with men. AIDS Care. 2009;21:1481-1489.
28. Grant BF, Dawson DA, Stinson FS, et al. The Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV): reliability of alcohol consumption, tobacco use, family history of depression and psychiatric diagnostic modules in a general population sample. Drug Alcohol Depend. 2003;71:7-16.
29. Grant B, Kaplan K. Source and Accuracy Statement for the Wave 2 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). Rockville, MD: National Institute on Alcohol Abuse and Alcoholism; 2005.
30. Dawson DA, Stinson FS, Chou SP, et al. Three-year changes in adult risk drinking behavior in relation to the course of alcohol-use disorders. J Stud Alcohol Drugs. 2008;69:866-877.
31. Ruan WJ, Goldstein RB, Chou SP, et al. The alcohol use disorder and associated disabilities interview schedule-IV (AUDADIS-IV): reliability of new psychiatric diagnostic modules and risk factors in a general population sample. Drug Alcohol Depend. 2008;92:27-36.
32. Dawson DA, Li TK, Chou SP, et al. Transitions in and out of alcohol use disorders: their associations with conditional changes in quality of life over a 3-year follow-up interval. Alcohol Alcohol. 2009;44:84-92.
33. Grant BF, Goldstein RB, Chou SP, et al. Sociodemographic and psychopathologic predictors of first incidence of DSM-IV substance use, mood and anxiety disorders: results from the Wave 2 National Epidemiologic Survey on Alcohol and Related Conditions. Mol Psychiatry. 2009;14:1051-1066.
34. Widom CS, Shepard RL. Accuracy of adult recollections of childhood victimization: Part 1 childhood physical abuse. Psychol Assess. 1996;8:412-421.
35. Fergusson DM, Horwood LJ, Woodward LJ. The stability of child abuse reports: a longitudinal study of the reporting behaviour of young adults. Psychol Med. 2000;30:529-544.
36. Senn TE, Carey MP, Vanable PA. Childhood and adolescent sexual abuse and subsequent sexual risk behavior: evidence from controlled studies, methodological critique, and suggestions for research. Clin Psychol Rev. 2008;28:711-735.
37. Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol. 1986;51:1173-1182.
38. Kraemer HC, Wilson GT, Fairburn CG, et al. Mediators and moderators of treatment effects in randomized clinical trials. Arch Gen Psychiatry. 2002;59:877-883.
39. Bonanno GA. Loss, trauma, and human resilience: have we underestimated the human capacity to thrive after extremely aversive events? Am Psychol. 2004;59:20-28.
40. Kazdin AE. Mediators and mechanisms of change in psychotherapy research. Annu Rev Clin Psychol. 2007;3:1-27.
41. Kyriacou DN, Hutson HR, Anglin D, et al. The relationship between socioeconomic factors and gang violence in the city of Los Angeles. J Trauma. 1999;46:334-339.
42. Aisenberg E, Herrenkohl T. Community violence in context: risk and resilience in children and families. J Interpers Violence. 2008;23:296-315.
43. Herrenkohl TI, Sousa C, Tajima EA, et al. Intersection of child abuse and children's exposure to domestic violence. Trauma Violence Abuse. 2008;9:84-99.
44. Wiewel EW, Hanna DB, Begier EM, et al. High HIV prevalence and diagnosis rates in New York city black men. J Community Health. 2010;36:141-149.
45. Harrington KF, DiClemente RJ, Wingood GM, et al. Validity of self-reported sexually transmitted diseases among African American female adolescents participating in an HIV/STD prevention intervention trial. Sex Transm Dis. 2001;28:468-471.
46. Dong M, Anda RF, Dube SR, et al. The relationship of exposure to childhood sexual abuse to other forms of abuse, neglect, and household dysfunction during childhood. Child Abuse Negl. 2003;27:625-639.
47. Dube SR, Felitti VJ, Dong M, et al. Childhood abuse, neglect, and household dysfunction and the risk of illicit drug use: the adverse childhood experiences study. Pediatrics. 2003;111:564-572.
48. Afifi TO, Mather A, Boman J, et al. Childhood adversity and personality disorders: results from a nationally representative population-based study. J Psychiatr Res. 2010; [Epub ahead of print].
49. Berkman LF. Social epidemiology: social determinants of health in the United States: are we losing ground? Annu Rev Public Health. 2009.
50. Allison PD. Missing Data. Thousand Oaks, CA: Sage Publications; 2001.
51. Barlow DH. Clinical Handbook of Psychological Disorders: A Step-by-Step Treatment Manual. 4th ed. New York, NY: Guilford Press; 2008.
52. Cohen J, Mannarino AP. Disseminating and implementing trauma-focused CBT in community settings. Trauma Violence Abuse. 2008;9:214-226.
53. Cohen JA, Mannarino AP. Psychotherapeutic options for traumatized children. Curr Opin Pediatr. 2010;22:605-609.
54. Weisz JR, Kazdin AE. Evidence-Based Psychotherapies for Children and Adolescents. 2nd ed. New York, NY: Guilford Press; 2010.
55. Ponniah K, Hollon SD. Empirically supported psychological treatments for adult acute stress disorder and posttraumatic stress disorder: a review. Depress Anxiety. 2009;26:1086-1109.
© 2011 Lippincott Williams & Wilkins, Inc.