Relationship characteristics associated with anal sex among female drug users

Abstract

Background

Anal sex is an important yet little studied HIV risk behavior for women.

Methods

Using information collected on recent sexual encounters, we examined the influence of sex partner and relationship characteristics on the likelihood of engaging in anal sex among women with a high risk of HIV infection.

Results

Anal sex was nearly three times more common among actively bisexual women (OR = 2.96, 95% CI 2.17 – 4.03). Women were more likely to have anal sex with partners who injected drugs (OR = 2.32, 95% CI 1.44 – 3.75), were not heterosexual (OR = 1.85, 95% CI 1.18 – 2.90), and with whom they exchanged money or drugs for sex (OR = 1.79, 95% CI 1.10 – 2.90). The likelihood of anal sex also increased with the number of nights sleeping together (OR = 1.15, 95% CI 1.06 – 1.24). In contrast, emotional closeness and social closeness were not associated with anal sex. Condom use during anal sex was uncommon, and did not vary according to partner or relationship characteristics.

Conclusions

Our findings support the need for HIV prevention interventions that target anal sex among heterosexuals, particularly in drug-using populations residing in neighborhoods with elevated levels of HIV prevalence.

Introduction

AIDS is a leading cause of death among African American men and women under the age of 55.¹ The rate of HIV/AIDS diagnoses among African-Americans greatly exceeds that among other racial/ethnic groups in all age categories; for African-American women, HIV is primarily contracted through sex with an infected male partner.²

It has been known for some time that anal sex increases the risk of male-to-female transmission of HIV^3–7; the relative risk of HIV transmission for women who engage in anal versus vaginal sex ranges from 2.8⁸ to 10.8.⁹ Per-contact probability of HIV transmission is estimated to be 10 times higher for penile-anal than penile-vaginal sex.¹⁰ Yet, many heterosexual men and women may remain ignorant of this risk, as HIV prevention messages targeting this population generally do not address anal sex,¹¹ and open discussion of this practice likely remains somewhat socially unacceptable.^{7, 11}

Recent studies in the United States have reported higher rates of heterosexual anal sex compared to earlier studies.^12–16 Comparing data collected for Project Respect in 1993–1995 versus 1999–2000, Satterwhite and colleagues¹³ found that the proportion of participants reporting any anal sex in the previous three months increased from 9% to 22% among women and 9% to 21% among men. Likewise, the lifetime prevalence of heterosexual anal sex was considerably higher in the 2002 National Survey of Family Growth (37% of men, 33% of women)¹⁴ than in the 1992 National Health and Social Life Survey (NHSLS) (28% of men, 21% of women).¹⁷

Women who are at high risk of exposure to HIV may be more likely to engage in anal sex than women in the general population,^{11, 18–21} compounding their risk. Several studies have found that women who trade sex for money or drugs are more likely to engage in anal sex.^{22, 23} Anal sex has also been associated with injection drug use,²² partner’s injection drug use,^{18, 22} crack use,^{18, 24} and amphetamine use.²³

Condom use appears to be less frequent during heterosexual anal sex than during vaginal sex,^{11, 25} and women who report anal sex are more likely to have unprotected vaginal sex than those who do not report anal sex.^{18, 25, 26} Higher rates of STIs among women who report anal sex ^{12, 18, 19, 23, 27} also indicate more risky sexual behavior.

Much of the theory and research surrounding HIV-related risk behavior has focused on individual-level determinants such as attitudes, self-efficacy, and health beliefs. However, women’s risk for HIV is inextricably linked to their relationships with male partners. As a consequence, the context of sexual activity – particularly partner characteristics – may be especially important to understanding women’s sexual risk.^28–30 This study describes the prevalence and correlates of anal sex among women in a population with a high risk of HIV infection, with particular attention to characteristics of partners and relationships.

Methods

Background and Procedures

The data for this study came from the Chicago site of the Sexual Acquisition and Transmission of HIV Cooperative Agreement Program (SATHCAP). This study used respondent-driven sampling (RDS)³¹ to recruit ‘hard’ drug users, men who have sex with men (MSM), the sex partners of both groups, and sex partners of the sex partners.³² Hard drug use comprised heroin, cocaine, or methamphetamine, or any illicit injection drug use. Interviews were conducted at five community-based sites operated by the School of Public Health, University of Illinois at Chicago using audio computer-assisted self-interview (ACASI) technology. Details on the study methodology are presented by Iguchi and colleagues.³² Blood samples were collected for HIV and syphilis testing, and urine samples were collected for chlamydia and gonnorhea testing. HIV testing was conducted by ELISA and confirmed by Western Blot. Rapid plasma reagin testing was used for the detection of syphilis infections. Urine specimens were tested using nucleic acid hybridization with amplification.

Between August 2005 and December 2008, 1645 women were recruited in Chicago; the records of 121 women (7%) were unusable due to errors in the ACASI instrument. Of the 1326 women who reported sex with a man in the past six months, 97% (N=1282) provided information on sexual activities with at least one male sex partner. Female participants may have provided information on up to six different sex partners, however fewer than 6% reported on more than three partners.

Measures

Participants were asked to classify each sex partner according to a list of partner types. We grouped these partner types into four categories: 1) primary or main, 2) regular, casual, friend or acquaintance, 3) stranger, one-time, or unknown, and 4) trade partner. Participants reported their sexual activity and condom use during the last time they had sex with each reported partner by checking all applicable items on a list of activities. Participants were asked to specify the partner’s sexual identity; we grouped the responses into three categories: homosexual, heterosexual, and bisexual/other. If the respondent indicated that the partner had sex with other people while they were partners, they were asked about the gender of the other partners. We constructed a binary variable to indicate whether the respondent believed the partner had sex with men while they were partners.

Participants were asked how long ago they first met the partner and how long ago they had first had sex. The difference was computed and a three-category measure was constructed indicating first sex occurred 1) within one month, 2) within one year, or 3) more than one year after first meeting. Trading sex was measured with two questions about giving or receiving drugs, money, or other goods in exchange for sex in the past six months. Four measures of closeness of the relationship were examined: 1) emotional closeness (measured on a scale of 1 to 10); 2) nights slept together during the week when they last had sex; 3) free time spent together; and 4) social network closeness (overlap of friendship networks).

Analyses

We compared the rates of condom use during anal and vaginal sex, using sex act as the unit of analysis, both using generalized estimating equation (GEE) logistic regression (a population averaged approach) for a between subjects estimate, and using the Stata xtmelogit³³ procedure to estimate random intercepts on subject and partner for a within subjects estimate.

We used the Stata xtlogit procedure³³ to estimate logistic regression models on anal sex, and condom use during anal sex (among women who reported anal sex). A GEE model was used to obtain estimates on subject characteristics, while a random intercept model was used to obtain estimates on partner characteristics.

These methods control for clustering of responses within subject, and for unequal numbers of observations across subjects. Regression models that ignore the dependency of observations tend to underestimate the standard errors of between-subject covariates (e.g. respondent characteristics), and overestimate the standard errors of within-subject covariates (e.g. partner characteristics). The odds ratios estimated from a random-effects logistic model additionally adjust for heterogeneity of the subjects, which can be considered to be due to unmeasured variables such as unobserved influences of social environmental factors.^34,35,36

We constructed multivariate random intercept logistic regression models using stepwise procedures with p < .01 as the initial entry criteria, and p < .05 the criteria to stay in the model. When competing models emerged we used Akaike’s Information Criteria (AIC) to select the best fitting model.³⁷ Finally, we tested the effect of including an additional level of clustering by recruiter, using xtmelogit. We conducted likelihood ratio tests comparing the two-level model (subject, sex partner) to the three-level model (recruiter, subject, sex partner). A significant clustering effect would indicate that the responses of subjects recruited by the same person are more similar than those of subjects recruited by different people, and that this clustering effect should be included in the model.

Results

Selected characteristics of the sample, and the distribution of sex partners by levels of these characteristics, are shown in Table 1. The women were mostly African American, the median age was 43, and most were residents of Chicago (95%) and unemployed (58%) or disabled (26%). Participants’ residences clustered in the areas surrounding the five recruitment sites, although all of the city’s 77 community areas were represented. Most participants reported using heroin and/or cocaine in the previous 30 days: 60% reported crack use, 45% reported heroin use, 23% reported speedball use, 18% reported powder cocaine, and 17% reported injecting drugs during that period. Close to half (43%) of the women identified as bisexual or something other than homosexual or heterosexual, and 27% reported having sex with both a man and a woman in the past six months. Five percent of the sample were HIV positive, 10.5% tested positive for syphilis, fewer than one percent tested positive for gonorrhea, and 1.4% tested positive for chlamydia.

Table 1.

Respondent Characteristics and Distribution of Sex Partners (N=1292)

	N	%	Sex Partner Observations	% Sex Partner Observations
Race/Ethnicity
White	98	8%	195	7%
African-American	1067	83%	2307	84%
Hispanic	111	9%	218	8%
Other	16	1%	40	1%
Age
18–29	109	8%	214	8%
30–39	295	23%	649	24%
40–49	599	46%	1312	48%
50–75	286	22%	579	21%
Education
less than HS	523	40%	1094	40%
HS grad/GED	425	33%	903	33%
beyond HS	344	27%	763	28%
Sexual Orientation
Homosexual	17	1%	27	1%
Heterosexual	711	55%	1518	55%
Bisexual/Other	553	43%	1195	43%
Sex partners
One	354	27%	397	14%
Two or three	513	40%	1191	43%
More than 3	424	33%	1169	42%
HIV Status
Negative	1226	95%	2625	95%
Positive	66	5.1%	135	5%

The majority of participants reported two or more sex partners in the past six months. The number of sex partners on whom data were collected was not associated with any of the socio-demographic variables, but was (as expected) associated with reported total number of sex partners in the past six months. Women with more than three sex partners comprised 33% of the sample, and contributed 42% of the sex partner observations.

Nearly all of the women (94%) reported vaginal sex with at least one partner, and 71% (76% of those reporting vaginal sex) reported vaginal sex without a condom. Fifteen percent of the women reported anal sex with at least one partner, and 80% of these women (representing 12% of the complete sample) reported anal sex without a condom.

Of the 2334 encounters than involved vaginal sex, 62% were unprotected, while 75% of the 256 encounters that involved anal sex were unprotected. This difference was significant in both GEE (OR = 1.38, 95% CI 1.08 – 1.79) and random effects (OR = 1.78, 95% CI 1.17 – 2.71) regression models. We also tested the interaction of type of sex act and partner type; while condom use did not differ by type of sex act with main partners, anal sex with non-main partners was more likely to be unprotected (GEE: OR = 1.92, 95% CI 1.36 – 2.70; Random Effects: OR = 3.14, 95% CI 1.69 – 5.83). In fact, for anal sex, the rate of condom use did not vary by type of sex partner.

One hundred ninety-four women provided information about their last anal sex partners, representing 256 male partners. Women were most likely to have anal sex with a main or primary partner (63%), followed by a regular or casual partner (43%). Anal sex with a stranger (5%) or trade partner (7%) was uncommon.

Sixty-four percent of the women who reported anal sex received money or goods in exchange for sex from an anal sex partner in the past six months, and 36% reported that they gave money or goods to an anal sex partner in exchange for sex. All but five of the women who gave an anal sex partner money or goods also received money or goods.

Table 2 shows the bivariate associations of anal sex with participant characteristics and behaviors. Odds ratios and robust confidence intervals are reported from GEE population-averaged regression models. Race, age, education, employment status, income, past month drug use, injection drug use, and sexually transmitted infections were unrelated to anal sex. Women who lived somewhere other than a home (e.g. hotel, shelter, squat) were more likely to report anal sex than those who lived in their own home or someone else’s home. Women who identified as something other than homosexual or heterosexual were almost twice as likely to report anal sex, and women who had sex with both men and women were almost three times as likely to report anal sex. Compared to women with three or fewer sex partners, women who reported having more than three sex partners in the past six months were twice as likely to report anal sex. Condom use during anal sex was not associated with any of the subject characteristics.

Table 2.

Subject-Level Correlates of Anal Sex During Most Recent Encounter: GEE Population-averaged Logistic Regressions

	% anal sex	OR	95% CI
Age
under 30	20%	1.00
30 – 39	14%	0.61	0.35	1.08
40 – 49	16%	0.71	0.42	1.18
50 or older	11%	0.52	0.28	0.94
Residence*
Own place	13%	1.00
Someone else’s place	15%	1.15	0.82	1.62
Somewhere else	20%	1.80	1.17	2.76
Sexual Orientation***
Hetero/Homosexual	11%	1.00
Bisexual/Other	20%	1.93	1.41	2.64
Sexual Activity***
Men only	11%	1.00
Men & women	27%	2.96	2.17	4.03
Sex Partners**
One	7%	1.00
Two or three	14%	1.24	0.76	2.03
More than 3	24%	2.04	1.28	3.27

Wald chi-square:

^*p < .05

^**p < .01

^***p < .001

We tested the associations of partner and relationship characteristics with anal sex via bivariate logistic regression analyses with random subject intercepts (Table 3). Partner variables significantly associated with anal sex were age (OR = 0.97, 95% CI 0.95– 0.99), hard drug use (OR = 1.85, 95% CI 1.20 – 2.83), injection drug use (OR = 2.75, 95% CI 1.77 – 4.27), non-heterosexual sexual orientation (OR = 2.47, 95% CI 1.62 – 3.79), and having sex with men while being the respondent’s partner (OR = 2.33, 95% CI 1.39 – 3.93).

Table 3.

Partner-Level Correlates of Anal Sex During Most Recent Encounter (N=2725)

Partner Variable		N	% anal sex
Type **	Primary/Main	1221	11%
	Non-main	1504	8%
Sexual orientation ***	Non-heterosexual	778	14%
	Heterosexual	1921	7%
Age **	< 30	184	14%
	30–39	460	12%
	40–49	1022	9%
	50–59	744	8%
	60 and up	216	2%
Ever injected drugs ***	No	2115	8%
	Yes	546	13%
Ever used hard drugs **	No	866	7%
	Yes	1782	11%
Had sex with male partners **	No	2420	8%
	Yes	292	16%
Time since first met *	< 6 months	475	10%
	6–11 months	295	13%
	1 year – < 3 years	338	13%
	3 yrs or more	1595	8%
Time to first sex	0 to 30 days	366	10%
	30 days to 1 year	1224	10%
	more than one year	989	7%
Nights in same bed *** (week last had sex)	None	977	6%
	One to three	941	10%
	Four or more	745	13%
Free time together ***	None to very little	1021	7%
	Some to about half	738	10%
	Most to all	900	11%
R gave partner money or drugs *** in exchange for sex	No	2184	8%
	Yes	538	16%
Partner gave R money or drugs ** in exchange for sex	No	1303	7%
	Yes	1420	11%

Wald chi-square:

^*p < .05

^**p < .01

^***p < .001

Relationship characteristics having significant associations with anal sex were being a primary/main partner (OR = 1.72, 95% CI 1.21 – 2.45), having first met at least six months ago but less than three years ago (vs. three years or more OR = 1.94, 95% CI 1.25 – 3.01), spending more nights in the same bed during the week last had sex (OR = 1.16, 95% CI 1.08 – 1.25), spending more free time together (all vs. none OR = 3.71, 95% CI 1.79 – 7.67), respondent giving partner money or drugs in exchange for sex (OR = 2.76, 95% CI 1.77 – 4.31), and partner giving respondent money or drugs in exchange for sex (OR = 1.75, 95% CI 1.18 – 2.61). Emotional closeness and network closeness were unrelated to anal sex. None of the associations with condom use during anal sex were significant.

In multivariate analyses incorporating both subject-level and partner-level variables, anal sex was significantly associated with subject bisexual behavior, number of nights slept together, partner injection drug use, partner non-heterosexual sexual orientation, partner age, and giving partner money or drugs in exchange for sex. The likelihood ratio test comparing the two-level and three-level models indicated a significant contribution of the recruiter grouping variable (Approximate LR chi2(01) = 3.77, p < .05). The estimates presented in Table 4 are adjusted for clustering by recruiter as well as by subject.

Table 4.

Multivariate Random Intercepts^a Logistic Regression Predicting Anal Sex From Subject-Level and Partner-Level Variables (N obs = 2449 N = 1242)

Variable	OR	95% CI		p
Respondent:
Bisexual activity	4.30	2.54	7.26	<.001
Partner:
Age	0.97	0.95	0.99	0.013
Not heterosexual	1.85	1.18	2.90	0.008
Injected drugs	2.32	1.44	3.75	0.001
R gave partner drugs or money for sex	1.79	1.10	2.90	0.019
Nights slept together	1.15	1.06	1.24	<.001

^aRandom intercepts on recruiter and subject

Discussion

Despite using a conservative measure of prevalence that asked only about the most recent sexual encounter with each partner, 15% of women in this high-risk sample reported recent anal sex and 12% reported anal sex without a condom. As reported in other studies^{11, 25} condoms were less likely to be used during anal sex than during vaginal sex, but only with non-main partners. Condom use with main partners was low (about 15%) regardless of type of sex act.

We did not anticipate the large proportion of women who identified as bisexual or reported recent sexual activity with a woman. RDS analyses (not shown) suggest that this is related to the targeted recruitment of MSM in this study. Non-heterosexual men were less likely than heterosexual men to recruit heterosexual women. The finding that bisexual identity and behavior are associated with a greater likelihood of heterosexual anal sex is consistent with a considerable body of research on drug users that has found higher levels of HIV risk and infection among drug-using women who have sex with women.^38–41 Young and colleagues⁴² conducted an ethnographic study on sexual minority women who inject drugs, and concluded that experiences of multiple marginalization led to increased HIV risk. Other subject-level variables appear to be associated with anal sex via their associations with bisexual activity. Analyses not shown here confirmed that bisexually active women were more likely to identify as something other than heterosexual, have more than three sex partners in the past six months, and be living somewhere other than a home.

Women were more likely to have anal sex with main or primary partners they spent at least some free time with and with whom they shared a bed more often. In contrast, emotional closeness and social closeness were not related to anal sex. Women also were more likely to report anal sex with partners to whom they had given money or drugs in exchange for sex. Since few studies of substance users address women paying or trading for sex, it is difficult to assess this finding in terms of previous research. In nearly all cases, however, women who gave their male partner money or drugs for sex also received money or drugs from this partner for sex. However, women who only received drugs or money for sex did not have elevated rates of anal sex. These outcomes suggest that anal sex may be more common in relationships characterized by instrumental reciprocity in which partners assist one another with the vagaries of economic marginalization and illicit drug use.

Anal sex was more likely to occur with high-risk partners, specifically non-heterosexual men and injection drug users. In fact 20% of women who had unprotected anal sex said that the partner had had sex with men while they were partners. Condom use during anal sex did not vary significantly according to any of the variables tested, but was uniformly low. In addition to the risk of HIV infection, unprotected anal sex with an injection drug user carries the potential risk of infection with Hepatitis C virus. Although sexual transmission of HCV evidently does not occur at a high rate,^{43, 44} the high background prevalence of HCV among injection drug users should be cause for concern.⁴⁵

Several limitations should be considered when interpreting these results. First, study participants were recruited through respondent-driven sampling (RDS), and debate exists regarding the extent to which RDS minimizes biases in sampling hidden populations.^{32, 46, 47} Study location and enrollment incentives affect recruitment patterns and, in this case, the sample reflects the low-income communities of color around the study sites. While the generalizability of the findings reported here are, therefore, limited, there were participants from all of Chicago’s 77 community areas and the sample represents a population at high risk of acquiring HIV infection. Second, data were self-reported and thus subject to biases associated with the accuracy or completeness of reporting. To minimize recall bias, behavioral questions usually concerned the most recent event, the past 30 days or, to a lesser extent, the past 6 months. Moreover, anal sex may be stigmatized to some extent, leading to underreporting. To minimize socially desirable reporting of illegal or embarrassing behaviors we used ACASI interviews and indigenous study staff.^48–50 Third, the accuracy of participants’ reports about the characteristics of their sex partners is unknown. For this analysis, however, the emphasis is on participants’ perceptions of their partners’ characteristics.

The HIV epidemic in the U.S. increasingly is driven by heterosexual transmission, particularly in low-income African American neighborhoods experiencing multiple socioeconomic problems. The estimated infectivity of HIV via unprotected receptive anal intercourse is similar to the per-act transmission probability for needle sharing.^{10, 51, 52} We believe this comparison is likely to resonate strongly in drug-using populations such as those participating in this study, where needle sharing generally is understood to be the riskiest of HIV related behaviors. We doubt, however, that study participants assign this level of risk to heterosexual anal sex, and we have experienced no evidence of network- or community-level norms supporting the avoidance of anal sex or the adoption of risk reduction in its practice, such as there are with needle sharing among IDUs and those who know them.

Our findings support the need for HIV prevention interventions that target anal sex among heterosexuals, particularly in drug-using populations residing in neighborhoods with elevated levels of HIV prevalence. Interventions that successfully equate the risk of anal sex to that of sharing needles – something widely understood in such neighborhoods – may give women greater power to refuse anal sex, or to bargain for condom use. Since bargaining power is likely to be greatest when men see anal sex as particularly risky to their female partners,⁵³ we believe interventions are needed that involve male partners. Since most women in our sample met their partners in their own neighborhoods (analysis not shown; see also Laumann and colleagues⁵⁴) interventions targeting heterosexual anal sex should aim to change social network norms in these high-risk populations. While we suspect that few U.S. communities are likely to adopt frank public awareness campaigns that promote anal sex risk reduction, interventions could adapt street outreach or respondent-driven sampling strategies that have been used successfully to change norms regarding needle sharing within IDU networks^55–57 and apply them to networks of injecting and non-injecting drug users in HIV high prevalence neighborhoods.

Summary.

Anal sex was common among women in this high HIV risk sample, and was associated with women’s bisexual behavior and high-risk male partners.