Relationship between heterosexual anal sex,injection drug use and HIV infection among black men and women

Abstract

US blacks carry a disproportionate risk of heterosexually transmitted HIV. This study aimed to evaluate the association between self-reported heterosexual anal intercourse and HIV. Using respondent-driven sampling (RDS), we recruited and interviewed 909 blacks from areas of high poverty and HIV prevalence in Houston, Texas, and who reported heterosexual sex in the last year. All individuals were tested for HIV. Weighted prevalence values were calculated to account for non-random recruitment associated with RDS. The weighted population prevalence of HIV infection was 2.4% and 2.5% among men and women, respectively. Education, employment status, income and crack cocaine use were not associated with HIV infection. Lifetime injection drug use (odds ratio [OR] 3.31, 95% confidence interval [CI] 1.31–8.33%) and heterosexual anal intercourse (OR 2.41, 95% CI 1.02–5.73%) were associated with HIV infection. Individuals who reported both injection drug use and heterosexual anal intercourse had 6.21 increased odds of HIV (95% CI 2.47–15.61%). Our results suggest that heterosexual anal sex may be a vector for HIV transmission, especially in the context of injection drug use. Prevention strategies directed at curbing the HIV epidemic among black heterosexuals require that we correctly identify the risks so that appropriate interventions can be developed.

Keywords

African American HIV anal sex risk respondent-driven sampling surveillance

INTRODUCTION

Black men and women carry a disproportionate burden of HIV infection in the United States.^1–4 Blacks comprise 13% of the US population, yet represent nearly half of all individuals living with HIV/AIDS and 45% of new infections.⁵ HIV incidence among black women is 14.7 times that for white women.² Patterns of HIV infection in Houston, TX, are similar: 54% of cumulative HIV cases are black, yet blacks represent only 20% of the Houston population.⁶ Of the cumulative HIV cases reported in Houston since 1999, 33% of male cases and 71% of female cases are blacks.⁶

Factors associated with heterosexual transmission of HIV in the United States are not well understood. Some authors ascribe heterosexual HIV risk among minorities to social and economic factors, including poverty, incarceration, both injection and non-injection drug use, a low male to female ratio and complex social networks associated with high-risk populations.^7–12 Others emphasize the examination of HIV transmission vectors, including sharing of injection drug use equipment and folk-medicine-induced exposure to contaminated equipment.^13–16 Anal intercourse may be an important route of infection among women.^11,13–18

The prevalence of anal intercourse has been reported to be 20–39% in sero-discordant heterosexual couples enrolled in a large European study of heterosexual transmission of HIV.^13,15 Anal intercourse is also associated with increased risk of HIV seroconversion.^15,16

This analysis was designed to evaluate the association between self-reported heterosexual anal intercourse and HIV infection in black men and women in Houston, TX. Specifically, we used data from the National HIV Behavioral Surveillance Project (NHBS) – First Heterosexual Cycle, a study designed and funded by the Centers for Disease Control and Prevention (CDC). We compared the prevalence of anal intercourse in the last 12 months between men and women infected with HIV with those who were not infected.¹⁹

STUDY DESIGN

Data in Houston were collected following the NHBS study protocol.^19,20 NHBS is designed to examine changes in sexual, drug use and HIV prevention behaviours over time by conducting serial cross-sectional surveys among three populations at high risk for HIV: men who have sex with men, injection drug users and heterosexuals who live in high-poverty, high-HIV-prevalence areas. On a three-year cycle, each population will be surveyed and changes in risky behaviours will be assessed. For this analysis, we used data from the first cycle of the heterosexual surveillance component. Houston data were collected from October 2006 to June 2007. Eligibility included recruitment by peers, living within the defined geographic area, age between 18 and 50 years and self-report of heterosexual sex in the last 12 months.

SAMPLING METHODS

Respondent-driven sampling (RDS) was used for recruitment. RDS is a modified chain-referral sampling method that allows for statistical adjustment of network size and recruitment patterns so that population-based prevalence estimates can be calculated.^21–24 The primary RDS procedures are: peers recruit peers, recruitment quotas and a dual-incentive system. First, a group of individuals identified through formative work is selected to serve as seeds. Data from these non-random initial participants are not used in the final analysis. There is ample evidence that the composition of the initial recruits does not bias the composition of the final sample if there are adequate recruitment waves.^21,22 After participating in the study, seeds are offered a limited number of coupons to recruit peers who meet specific eligibility criteria. These new recruits (if eligible) participate in the study and are offered the same limited number of coupons to recruit additional participants. The process is repeated until the sample size is met.

We first identified high-risk geographic areas using the protocol established by the CDC for this cycle of NHBS.^19,20,25,26 Specifically, poverty data from the US census and HIV prevalence data from the Houston Department of Health and Human Services were used to identify census tracts with the highest index for poverty and HIV infection. In Houston, 133 census tracts were included in our high-risk area. Within the high-risk areas, we conducted formative and ethnographic research to describe social network characteristics. Formative work is important in RDS to determine whether the community is sufficiently connected to allow RDS to work, and, if so, to develop community support for the project.²⁷

From key-informant interviews, we identified within our target sampling area several large public-housing complexes with overlapping social networks. Within these areas, we recruited two men and six women to initiate the RDS process, using methodology described by Heckathorn.²² Each seed was instructed to recruit no more than three peers, and each eligible recruit was also invited to recruit three additional individuals until the sample size was met. Recruitment instructions were to recruit friends, relatives or people you are close to and have seen in the last 30 days. No instructions were given to include or exclude sexual partners; however, participants were specifically asked to not recruit strangers. Network size for each participant was estimated by asking them how many people they were close to and had seen in the last 30 days who were between the ages of 18 and 50 years. Our minimum sample size goal, established by the CDC, was 750 eligible participants; we recruited 939 eligible participants with valid responses.

We used a dual-incentive system: recruits were offered remuneration for participating in the questionnaire (US$20) and getting tested for HIV (US$20); in addition, they were offered an incentive (US$10) for each eligible peer they recruited (with a limit of three recruits).

VARIABLES OF INTEREST

We collected sexual, drug use and HIV testing behaviours using an interviewer-administered questionnaire. Our trained interviewers were experienced in HIV prevention activities and with the geographic area of our target population. Questions were read to the recruit from, and responses entered into, a hand-held computer using QDS software (Questionnaire Development System: NOVA Research, Bethesda, MD, USA).²⁸

Participants were asked to recall behaviours over the last 12 months. We explored three types of sex: oral, vaginal and anal. Interviewers read the following statement: ‘Oral sex means mouth on the vagina or penis; vaginal sex means penis in the vagina; and anal sex means penis in the anus or butt.’ We also asked about types of partners. Main partners are partners you have sex with who you feel committed to above anyone else (includes girlfriend, wife, significant other or life partner); casual partners are partners you have sex with but do not feel committed to or don't know very well; and exchange partners are partners you have sex with in exchange for things like money or drugs. Our analysis described women's self-report of anal intercourse with men and men's self-report of anal intercourse with women. All partner types were included.

Because of the uncertainty regarding the validity of self-reported condom use, in the analysis we did not stratify our results by protected and unprotected anal sex.^29,30 Other variables of interest included educational attainment, current employment status, yearly income, HIV testing behaviour (ever tested for HIV) and injection drug use (ever injected drugs and injected in the last 12 months). We also examined self-reported sexual behaviours by partner type over the last 12 months. That is, during the last 12 months, did you have a main, a casual or an exchange partner, and if you did have a specific type of partner, did you engage in oral, vaginal, and/or anal sex with that partner? From this information, we created a variable that included any anal sex with any partner-type in the last 12 months, compared with vaginal and/or oral sex without anal sex.

Among men, we also examined male self-reports of ever having sex with a man, and having sex with a man in the last 12 months.

All consenting individuals were offered HIV testing. We used Oraquick^® as a screening test and Orasure^® as a confirmatory test among those screening positive (Orasure Technologies Inc, Bethlehem, PA, USA). If the participant self-reported being infected with HIV, then Orasure^® was used to confirm the infection.³¹

The CDC Human Subjects Protection Office determined that full CDC International Review Board review was not needed. However, at a local level, this study had the approval of the University of Texas Health Science Center Committee for the Protection of Human Subjects. Verbal consent was obtained to ensure that the responses were anonymous.

STATISTICAL METHODS

RDS requires specific analysis techniques that take into account the individual network size and the sampling characteristics (who recruited whom) to develop unbiased population prevalence estimates.^{21–23,32–34} We used two free programmes to assist in the RDS process. Coupon Manager was used to track recruitment chains and RDS Analysis Tool (RDSAT) generated the weighted population proportion estimates needed to make population-based prevalence estimates from data accumulated through RDS. The weighting was necessary to compensate for non-random recruitment. Both programmes are available online at www.respondentdrivensampling.org.³³

Using RDSAT, we calculated weighted population estimates, stratified by HIV infection status. Next, we generated frequencies for the variables of interest using the weighted population estimates adjusted for the design effect. Samples recruited through RDS have wider variances than simple random samples, and these differences can be quantified by the design effect, which is the ratio of the variance in RDS-adjusted data to the variance associated with a simple random sample.³⁴ Design-effect-adjusted frequencies are calculated by dividing the actual sample by the variable-specific design effect, a procedure necessary before calculating appropriate CIs.³⁵ We calculated ORs and used Fisher's exact test to assess the significance when comparing design-effect-adjusted frequencies.

We compared continuous variables (age and number of sexual partners in the last year) using the t-test, without adjustment. Techniques have not yet been developed for assessing differences in continuous variables collected through RDS.

There is no consensus on how to appropriately weight RDS-collected data when using regression analysis.^33,36 We used logistic regression to assess the multivariable association between our variables of interest and HIV infection status, adjusting for the RDS process by using individual RDS-generated weights generated on the dependent variable (HIV infection status) as a covariate in the multivariable logistic regression.

RESULTS

This analysis included 909 black individuals; 63% were women (n = 569). The mean ages and standard deviation (SD) were 32.6 (10.5) and 31.4 (10.1) years for men and women, respectively (P = 0.07).

There were 9/340 men and 15/569 women with positive Western blot results from Orasure^®. The weighted prevalence of HIV infection in this population was 2.4% (95% CI 1.3–3.9%). By gender, the weighted prevalence of HIV infection was 2.4% (95% CI 0.7–4.6%) for men and 2.5% (95% CI 1.1–4.4%) for women. HIV-infected men and women were older than those uninfected: mean ages and SDs for HIV-uninfected and HIV-infected individuals were, respectively, 32.5 years (SD 10.4) and 36.0 years (SD 13.1) for men (P = 0.33); and 31.1 years (SD 10.1) and 40.3 years (SD 9.3) for women (P < 0.001).

The numbers of sexual partners in the last year were similar between men and women (6.7 for men and 6.2 for women, P = 0.74) and between HIV-infected and HIV-uninfected individuals (6.0 for HIV infected and 6.3 for HIV uninfected, P = 0.95).

Table 1 presents weighted population prevalence values for basic demographic, social, drug use, and sexual risk behaviours, calculated with RDSAT.³³ Table 2 presents frequencies that were developed from the weighted prevalence values and adjusted for design effects (personal communications with Dr Heckathorn). The latter frequencies were used to develop prevalence odds ratios and 95% CIs. Note that the smaller sample size for each variable in Table 2 results from dividing the initial sample by the calculated design effect. These do not represent missing data, but rather data adjusted for the sampling method.

Table 1

Population prevalence of demographic, social, drug use and sexual behaviours of study participants, by HIV infection status. Houston site of the National Behavioural Surveillance Project – First Heterosexual Cycle (NHBS-HET1), 2006–2007. Presented are the actual numbers and the weighted prevalence values, calculated using RDSAT*

	HIV− (n = 885)		HIV+ (n = 24)
Education	No.	Weighted prevalence (%)	No.	Weighted prevalence (%)
12th grade or less	723	85.1	21	88.9
More than 12th grade	162	14.9	3	11.1
Current employment status
Employed full or part-time	325	35.1	5	31.1
Unemployed	560	65.0	19	68.9
Income last year
Less than US$5000	389	42.7	11	58.9
US$5000 to 14,999	207	24.4	9	22.0
US$15,000 and over	289	32.9	4	19.1
Had at least one HIV test (lifetime)	712	79.5	19	90.2
Used crack in the last 12 months	155	25.5	6	25.2
Ever injected drugs	89	10.2	8	20.5
Injected drugs in the last 12 months	29	5.0	4	12.0
Main partner in the past year	784	89.4	18	61.4
Anal sex with main partner	118/784	14.7	4/18	17.4
Casual partner in the past year	464	50.4	17	62.1
Anal sex with casual partner	166/464	42.5	111/17	41.0
Exchange partner in past year	181	21.8	5	13.5
Anal sex with exchange	78/181	57.4	2/5	N/C ^†
Any anal sex in past year	273	31.5	14	42.0

*The weighted prevalence values are generated through RDSAT,³¹ taking into consideration network size and recruitment patterns. They cannot be directly calculated from the numbers provided in this table

^† N/C: the weighted prevalence was not calculable because the sample size was so small

Table 2

Univariate analysis of demographic, social, drug use and sexual behaviour variables of study participants, by HIV infection status. Houston site of the National Behavioral Surveillance project First Heterosexual Cycle (NHBS-HET1), 2006–2007. Presented are the weighted frequencies adjusted for design error; prevalence odds ratios (POR); and 95% confidence intervals (CI). For each category, the first row is considered the risky variable

	HIV−	HIV+	POR	95% CI	P
Education
12th grade or less	428	12
More than 12th grade	79	6	0.37	0.12–1.24	0.06
Current employment status
Unemployed	306	12
Full or part time	161	2	3.16	0.69–29.32	0.15
Income last year
Less than or equal US$5000	196	9
Greater than US$5000	278	10	1.28	0.45–3.57	0.64
Had at least one HIV test (lifetime)
No	92	2
Yes	400	11	0.79	0.08–3.72	1.00
Used crack in the past 12 months
Yes	75	6
No	279	14	1.59	0.48–4.60	0.40
Ever injected drugs
Yes	35	5
No	307	8	5.48	1.32–20.09	0.009
Injected drugs in the last 12 months
Yes	16	3
No	315	12	4.92	1.26 to 19.20	0.04
Main partner in the last 12 months
Yes	445	9
No	50	5	0.20	0.06–0.80	0.01
Anal sex with main partner
Yes	52	4
No	326	5	5.02	1.30–23.99	0.03
Casual partner in the last 12 months
Yes	244	8
No	235	4	1.93	0.51–8.85	0.22
Anal sex with casual partner
Yes	41	3
No	58	0	9.87*		0.08
Exchange partner in the last 12 months
Yes	82	2
No	294	11	0.65	0.07–3.08	0.74
Anal sex with exchange partner
Yes	13	0
No	10	2	0.16*		0.22
Any anal sex in the last 12 months
Yes	130	9
No	281	6	3.24	1.00–11.28	0.05

*With a correction of 0.5 added to each cell. Confidence intervals are not calculable

The majority of our participants had less than a 12th grade education, were unemployed and had incomes less than US$5000 last year. Nearly one-fourth used crack cocaine in the last 12 months. While none of these were statistically significantly associated with HIV infection, there was very little variation among the sample, making the assessment of differences difficult.

On the other hand, while only 10% reported a history of injection drug use, individuals who did report injecting drugs were significantly more likely to be HIV infected (Table 2).

Those who reported a main sexual partner were less likely to be infected with HIV; however, having anal sex with a main partner was associated with HIV infection. Neither having a casual partner or an exchange partner was associated with HIV prevalence. We were unable to calculate adjusted prevalence estimates for anal sex with a casual or exchange partner because of small numbers.

We found that 31.5% of HIV-uninfected and 42.0% of HIV-infected subjects reported any anal sex with any heterosexual partner type in the last 12 months (OR 3.24, 95% CI 1.00–11.28%).

Among men, self-report of male-to-male sex was not a risk factor for HIV infection. Twenty-one men reported having had male-to-male sex at least once in their life (11.0%, 95% CI 2.1–24.5%), but none of these men were infected with HIV. However, 18/21 (88.8%, 95% CI 74.0–100%) of men reporting any lifetime male-to-male sex also reported heterosexual anal sex in the last 12 months. In addition, 11/12 (88.3%, 95% CI 51.6–97.9%) men who reported male-to-male sex in the last 12 months also reported heterosexual anal sex; 5/11 reported male-to-male sex and heterosexual anal sex with a main partner. These five men also reported anal sex with casual and exchange partners in the past year.

Anal sex was more common among both men and women who injected drugs. By sex, 75.7% (95% CI 55.8–91.7%) of men and 57.3% (95% CI 37.5–78.3%) of women who reported injection drug use also reported anal sex, compared with 50.2% (95% CI 41.8–57.4%) of men and 15.1% (95% CI 10.9–19.2%) of women who did not report injection drug use. The number of HIV-infected individuals in this sample size is too small to fully examine risk by sex. However, 8/9 (88.9%) of HIV-infected men reported any anal sex, compared with 52.3% of HIV-uninfected men (OR 7.31, 95% CI 0.89–60.13%). Among women, 40.0% of HIV-infected women reported anal sex compared with 18.0% of the HIV-uninfected women (OR 3.03, 95% CI 1.05–8.74%). By injection drug use, there may be an increased HIV prevalence associated with anal sex among those who never injected drugs (OR 2.06, 95% CI 0.64–6.30), and among those who reported injecting drugs (OR 4.33, 95% CI 0.51–200.63).

We used multivariable logistic regression to examine factors associated with HIV infection. In our initial multivariable model, we included those variables that were significantly associated with HIV infection in the univariate analysis presented in Table 2. In addition, we included crack cocaine use in the past 12 months because other studies have found crack use to be a predictor of HIV in heterosexual populations. The initial multivariable model included gender, education, crack cocaine use, injection drug use ever, main sex partner and any anal sex. We included injection drug use ever rather than injection drug use in the past 12 months because more people reported ever injecting and we wanted as robust a model as possible. Likewise, we included any heterosexual anal sex rather than anal sex for the three partner-specific variables (main, casual and exchange) in order to increase the robustness of the model by keeping cell size as large as possible. Keeping only those variables with a P value of 0.05 or less in the full model, our final model showed that a history of injection drug use (OR 3.31, 95% CI 1.31–8.33%) and any heterosexual anal sex (OR 2.41 95% CI 1.02–5.73%) were the only two variables we examined that were independently associated with HIV infection in this population. The prevalence of HIV among the 62 individuals who were injection drug users and reported any heterosexual anal sex in the last year was more than six times greater than among individuals who did not report both activities (OR 6.21, 95% CI 2.47–15.61%).

DISCUSSION

We found that 50% of men and nearly 20% of women reported heterosexual anal sex in the past 12 months. Individuals with a history of injection drug use were more likely to practise heterosexual anal sex: among injection drug users men were 1.5 times and women were 3.8 times more likely to practise heterosexual anal sex than non-injectors. Even though anal sex was more common among injection drug users, both injection drug use and anal sex were independently associated with HIV infection.

Others have described poverty, low educational attainment and crack cocaine use as risk factors for heterosexual transmission of HIV.^9,17 We could not identify these as risk factors because, by design, our participants were all living in high-poverty areas and had low educational attainment. Nearly a quarter of both those infected with HIV and those not infected used crack cocaine in the past year. By limiting our sample to high-risk areas of the city, we essentially controlled for these potential confounders and could identify factors more proximately associated with HIV infection.

While none of the men in this sample with a history of male-to-male sex were infected with HIV, they were more likely to practise heterosexual anal intercourse than men not reporting male-to-male sex. Even if the prevalence of anal sex does not vary among racial/ethnic groups,¹⁷ if the prevalence of HIV in a population is high, the risk of HIV associated with anal sex will also be high.

In 1998, Leynaert et al. ¹³ suggested that a woman's risk of HIV from anal intercourse was significantly higher than from vaginal intercourse. Chirgwin et al. ¹⁶ found anal intercourse and non-parenteral drug use to be associated with HIV seroconversion in 1999. Brody and Potterat¹⁴ summarized from the literature that penile–vaginal intercourse had one-tenth to one-fiftieth the risk of HIV associated with anal intercourse in 2004. It is sobering to find that individuals who reported both anal intercourse and injection drug use had more than six times the odds of being infected with HIV than individuals who did not report either behaviour. Independently, anal intercourse was associated with a 2.5 increase in odds of HIV infection and injection drug use was associated with a 3.3 increase in odds of HIV infection. Yet, anal intercourse is not prominently discussed as a risk factor for HIV in most interventions.

There are several limitations to our study. The number of individuals with positive Orasure results was relatively small, limiting our ability to fully investigate the potential confounding. However, we did have sufficient power to accept our a priori hypothesis – that HIV prevalence was associated with heterosexual anal intercourse. The population from which we recruited was quite homogeneous in basic demographics, so we could not determine whether education, employment or income was associated with HIV prevalence. Because we used RDS, the analysis was complicated by the need to adjust for network size and recruitment patterns. In addition, the most appropriate method for multivariate analysis from RDS studies has not yet been described. However, we used the best methods published to date.³⁶

To our knowledge, this is the first population-based study describing heterosexual anal sex and injection drug use as independent risks associated with HIV prevalence in US black men and women. We agree with Potterat et al. ¹¹ that it is time to more thoroughly explore heterosexual anal sex as a vector for HIV transmission. Information about HIV transmission, including the risk associated with anal intercourse, should be a part of any prevention strategy designed to curb the disproportionate burden of HIV incidence among black heterosexuals, especially among black women.

Footnotes

ACKNOWLEDGEMENTS

We would like to acknowledge Dr Raouf Arafat, Houston Department of Health and Human Services for his continuing support. This work was partially supported by Cooperative Agreement PS000977 from the Centers for Disease Control and Prevention (CDC). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the CDC.

References

Southern states manifesto 2008. Available at www.southernaidscoalition.org (last accessed 8 January 2008)

Centers for Disease Control and Prevention. Subpopulation estimates from the HIV incidence surveillance system-United States, 2006. MMWR 2008; 57:985–9

Centers for Disease Control and Prevention. Cases of HIV infection and AIDS in the United States and dependent areas, by race/ethnicity, 2002–2006. HIV/AIDS Surveillance Supplemental Report 2008;13(No. 1): Available at http://www.cdc.gov/hig/topics/surviallance/resources/reports/#supplemtal (last accessed 10 June 2008)

Centers for Disease Control and Prevention. New HIV incidence estimates: CDC Responds. CDC HIV/AIDS Facts. Available at http://www.cdc.gov/hiv/topics/surveillance/incidence.htm (last accessed 10 June 2008)

Hall

, Song

, Rhodes

, Estimation of HIV incidence in the United States. JAMA 2008;300:520–9

Houston Department of Health and Human Services. Bureau of HIV/STD Prevention, Quarterly HIV/AIDS Surveillance Report. Available at http://www.houstontx.gov/health/HIV-STD/ (last accessed 8 January 2008)

Laumann

, Youm

. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999;26:250–61

Schulz

, Parker

, Israel

, Fisher

. Social context, stressors, and disparities in women's health. J Am Med Wom Assoc 2001;56:143–9

Adimora

, Schoenbach

, Martinson

FEA

, Heterosexually transmitted HIV infection among African Americans in North Carolina. J Acquir Immune Defic Syndr 2006;41:616–23

10.

Adimora

, Schoenbach

. Contextual factors and the black-white disparity in heterosexual HIV transmission. Epidemiology 2002;13:707–12

11.

Potterat

, Brewer

, Brody

. Blind spots in the epidemiology of HIV in black Americans. Int J STD AIDS 2008;19:1–3

12.

Rothenberg

, Long

, Sterk

, The Atlanta Urban Networks Study: a blueprint for endemic transmission. AIDS 2000;14:2191–200

13.

Leynaert

, Downs

, de Vincenzi

. Heterosexual transmission of human immunodeficiency virus: variability of infectivity throughout the course of infection. European study group on heterosexual transmission of HIV. Am J Epidemiol 1998;148:88–96

14.

Brody

, Potterat

. Establishing valid AIDS monitoring and research in countries with generalized epidemics. Int J STD AIDS 2004;15:1–6

15.

De Vincenzi

. A longitudinal study of human immunodeficiency virus transmission by heterosexual partners. N Engl J Med 1994;331:341–6

16.

Chirgwin

, Feldman

, Dehovitz

, Minkoff

, Landesman

. Incidence and risk factors for heterosexually acquired HIV in an inner-city cohort of women: temporal association with pregnancy. J Acquir Immune Defic Syndr Hum Retrovirol 1999;20:295–9

17.

Halperin

. Heterosexual anal intercourse: prevalence, cultural factors, and HIV infection and other health risks, Part I. AIDS Patient Care STDS 1999;13:717–30

18.

Jaffe

, Seehaus

, Wagner

, Leadbeater

. Anal intercourse and knowledge of acquired immunodeficiency syndrome among minority-group female adolescents. J Pediatr 1988;112:1005–7

19.

Lansky

, DiNenno

. Centers for Disease Control and Prevention. National HIV Behavioral Surveillance System for Heterosexuals at Risk for HIV Infection. Presented at the American Public Health Association Annual Meeting, 2007. Available at http://apha.confex.com/apha/135am/webprogram/Session21687.html (accessed 08/01/08)

20.

Risser

, Padgett

. Using Respondent Driven Sampling in High Risk Aareas to Recruit Heterosexuals at High Risk for HIV: Demographic Results from the Houston site of the National HIV Behavioral Surveillance Project. Presented at the American Public Health Association Annual Meeting, 2007. Available at http://apha.confex.com/apha/135am/webprogram/Session21687.html (last accessed 8 January 2008)

21.

Heckathorn

. Respondent driven sampling: A new approach to the study of hidden populations. Soc Probl 1997;44:174–99

22.

Heckathorn

. Respondent-driven sampling II: deriving valid population estimates from chain-referral samples of hidden populations. Soc Probl 2002;49:11–34

23.

Salganik

, Heckathorn

. Sampling and estimation in hidden populations using respondent-driven sampling. Sociol Methodol 2004;34:193–239

24.

Malekinejad

, Johnston

, Kendall

, Kerr

, Rifkin

, Rutherford

. Using respondent-driven sampling methodology fo HIV biological and behavioral surveillance in international settings: a systematic review. AIDS Behav 2008;12:S105–30

25.

Robinson

. Formative Work and Preliminary Results Using RDS: New Orleans after Katrina. Presented at the American Public Health Association Annual Meeting, 2007. Available at http://apha.confex.com/apha/135am/webprogram/Session21687.html (last accessed 8 January 2008)

26.

Sifakia

, Flynn

. National HIV Behavioral Surveillance System for heterosexuals at risk for HIV – Implementation of Venue-based Sampling in Baltimore, Maryland. Presented at the American Public Health Association Annual Meeting, 2007. Available at http://apha.confex.com/apha/135am/webprogram/Session21687.html (last accessed 8 January 2008)

27.

Schensul

, LeCompte

, Trotter

RTI

, Singer

. Mapping social networks, spatial data, and hidden populations. Fourth ethnogropher's toolkit. Walnut Creek: AltaMira Press, 1999

28.

Coupon Manager. Available at http://www.respondentdrivensampling.org/ (last accessed 8 January 2008)

29.

Geary

, Tchupo

J-P

, Johnson

, Cheta

, Nyama

. Respondent perspectives on self-report measures of condom use. AIDS Educ Prev 2003;15:499–515

30.

Zenilman

, Weisman

, Rompalo

, Condom use to prevent incident STDs: the validity of self-reported condom use. [see comment]. Sex Transm Dis 1995;22:15–21

31.

Oraquick and Orasure. Available at http://www.orasure.com/products (last accessed 8 January 2008)

32.

Abdul-Quader

, Heckathorn

, Sabin

, Saidel

. Implementation and analysis of respondent driven sampling: lessons learned from the field. J Urban Health 2006;83(6 Suppl):i1–5

33.

Heckathorn

. RDSAT, version 6.0. Available at http://www.respondentdrivensampling.org/reports/RDSAT60.htm . Cornell University, Ithaca, New York. 2008

34.

Salganik

. Variance estimation, design effects, and sample size calculations for respondent-driven sampling. J Urban Health 2006;83(6 Suppl):i98–112

35.

Levy

, Lemeshow

. Sampling Populations: Methods and Applications. 3rd edn. New York: John Wiley and Sons, Inc, 1999

36.

Johnston

, Malekinejad

, Kendall

, Iuppa

GWR

. Implementation challenges to using respondent-driven sampling methodology for HIV biological and behavioral surveillance: field experiences in international settings. AIDS Behav 2008;12:S131–41