Prospective Clinical Trials Designed to Assess the Use of Hormonal Contraceptives and Risk of HIV Acquisition

Abstract

Background Hormonal contraceptives are among the most widely used family planning methods worldwide. Over 100 million women use OC and an estimated 20–30 million women use injectable depot medroxyprogesterone acetate (DMPA). Moreover, the use of injectable progestin-only contraception, including DMPA, appears to be among the fastest growing contraceptive methods in many developing countries, where the prevalence of HIV and other STI is high. Because HIV infection primarily affects women of reproductive age, understanding the impact of hormonal contraception on HIV transmission is a critical unanswered public health question. The possible biological mechanisms by which hormonal contraception might facilitate HIV transmission include: (1) increased cervical ectopy caused by OC use1; (2) increased cervical chlamydial infection (and associated purulence)2,3; (3) humoral and cell-mediated immunological changes associated with the use of exogenous steroids4; (4) a direct influence of sex hormones on the virulence of the virus4; and (5) thinning of the vaginal epithelium and irregular uterine bleeding associated with injectable progestin use3. In addition, the particular dose of estrogen contained in OC or the type of progestin contained in hormonal contraceptives may have an impact on the efficiency of HIV transmission. A team of reviewers from Family Health International and the University of Washington recently reviewed the approximately 30 studies addressing a possible association between hormonal contraception and HIV; all studies have been observational. Only 10 studies have been prospective and were thus able to document that hormonal contraceptive use preceded HIV transmission. One prospective study included only four women using hormonal contraception during the defined exposure period and thus provides little relevant information5. Of the remaining nine studies, eight provide data on OC use and HIV6–13 and five on DMPA use and HIV8,10,12–14. The review team judged the studies by Plummer et al.7, Sinei et al.11, Martin et al.12 and Bulterys et al.14 to be of higher quality than the others. Three of the studies reported findings on oral contraceptive use. The study by Plummer et al.7 reported an adjusted odds ratio for OC of 4.5 (95% CI 1.4–13.8), the Sinei study reported a crude OR of 3.5 (95% CI 0.8–21.5), which persisted after control for single confounders at a time11, and the study by Martin et al.12 reported an adjusted hazard ratio of 1.3 (95% CI 0.8–2.2). Both studies examining the DMPA–HIV association reported increased risks associated with DMPA use [Bulterys et al.14 relative risk 1.9 (95% CI 0.8–4.6); Martin et al.12 HR 2.0 (95% CI 1.3–3.1)]. The other five prospective studies of OC use and HIV and the other three prospective studies of DMPA use and HIV all found no significant association between hormonal contraceptive use and HIV acquisition5,6,8–10,13. Nevertheless, the review team concluded that all 10 prospective studies had major methodological weaknesses, including low numbers of hormonal contraceptive users, no assessment of the comparability of the contraceptive exposure groups, crude measurement of hormonal contraceptive exposure, inappropriate comparison groups, unclear timing of hormonal contraceptive use and HIV acquisition, poor follow-up and poor generalizability of study results. Independently, in a recent review of the hormonal contraceptive–HIV association, a British researcher came to the same conclusion15. Our current understanding of a possible hormonal contraceptive–HIV association may thus be inaccurate, and a methodologically sound study must be undertaken. The primary objective of this study is to measure the effect of OC and DMPA use on the risk of HIV-1 acquisition among women. The secondary objectives are to evaluate whether (1) STI modify the effect of hormonal contraceptive use on the acquisition of HIV-1, and (2) an effect of hormonal contraception on HIV-1 acquisition is mediated by cervical ectopy or bacterial vaginosis. Methods We designed a prospective cohort study of HIV-negative women in Uganda, Zimbabwe and Thailand to compare HIV-1 incidence rates among women using OC, DMPA, and women not using hormonal contraception. Women recruited into the study were 18–35 years of age, had used either low-dose OC or DMPA or had not used any hormonal contraception for at least 3 months. All women were HIV seronegative, were sexually active and, if parous, were at least 4.5 months postpartum. Women were excluded if they were currently pregnant (or intending to become pregnant), used an IUD in the previous month, injected illicit drugs, had had a blood transfusion in the previous 3 months, or had had a complete hysterectomy or an abortion in the past month. Participants are followed at 12-week intervals for 15–24 months. At each visit participants undergo a pelvic examination, are tested for HIV and other STI, and complete an interviewer-administered behavioral and contraceptive history questionnaire. Participants also receive pre and post-HIV counseling and are offered either OC or DMPA and condoms or condoms alone. Incident HIV infection is defined as an HIV-positive result (by enzyme-linked immunosorbent assay, two rapid tests, confirmed by Western blot and/or DNA PCR testing) in a previously HIV-seronegative woman. HIV-DNA PCR testing is also conducted on the visit before the identified seroconversion visit to ensure the accurate timing of infection. Proportional hazards modeling will be used to estimate adjusted HR for the effect of OC and DMPA use (compared with no hormonal use) on HIV acquisition. The study is designed to detect a twofold effect of hormonal contraceptive use on HIV-1 incidence with 90% power. Results From October 1999 to September 2002, we enrolled 6171 HIV-uninfected women into the study including 2113 OC users, 2132 DMPA users and 1926 women not using hormonal methods. At baseline, more participants in the non-hormonal than in the hormonal groups were not living with a partner (54 versus 45%), had ever engaged in commercial sex (5 versus 3%), had multiple partners in the past 12 months (13 versus 7%), and had ever used a condom (84 versus 64%). Age, age at first sex, recent oral and anal sex and recent sex under the influence of alcohol or drugs were similar between contraceptive exposure groups. Through January 2003, overall study follow-up is 84%, with 93, 82 and 79% for follow-up rates in Thailand, Uganda, and Zimbabwe, respectively. To date there have been 176 incident HIV-1 infections for an overall incidence rate of 1.9 per 100 woman-years (WY) (Thailand 0.1 per 100 WY, Uganda 1.6 per 100 WY, Zimbabwe 3.7 per 100 WY). Incidence rates for gonorrhea (Uganda 4.4 per 100 WY, Zimbabwe 3.1 per 100 WY, Thailand 1.6 per 100 WY) and syphilis (Uganda 1.9 per 100 WY, Zimbabwe 1.4 per 100 WY, Thailand 0.3 per 100 WY) are higher in Uganda and Zimbabwe than Thailand. Incidence rates for chlamydia are higher in Thailand (6.8 per 100 WY) than in Uganda (4.6 per 100 WY) or Zimbabwe (2.8 per 100 WY). Data collection will be completed in late 2003. Conclusion We have designed a large multicenter prospective cohort study specifically to examine the association between hormonal contraception and HIV-1 acquisition. Although a randomized controlled trial would provide the most rigorous design to test our hypothesis, we chose instead a prospective cohort design in which women select their contraceptive method of choice. This reduces adherence issues, and minimizes ethical issues resulting from randomly assigning women to a method with decreased contraceptive efficacy. We reduced biases by enrolling nearly equal groups of women in each country who use OC, DMPA, and non-hormonal methods, and by collecting detailed information on potential confounding factors. This study will provide critical information regarding the relationship between hormonal contraceptive use and HIV-1 acquisition. Support The National Institute of Child Health and Human Development (NICHD) Contract N01-HD-0-3310 supports this study.