Hormonal contraception and risk of HIV-1 acquisition: results of a 10-year prospective study

The use of hormonal contraception has been associated with an increased risk of HIV-1 in some studies but not in others. We analysed data from a 10-year prospective cohort study of female sex workers in Mombasa, Kenya. In multivariate analysis, women using the injectable contraceptive depot medroxyprogesterone acetate and women using oral contraceptive pills were at increased risk of HIV-1 acquisition compared with women using no contraceptive method.

Almost 150 million women worldwide use hormonal forms of contraception, many of whom are at some risk of HIV-1 [¹]. Unlike barrier methods of contraception, hormonal methods offer no protection against sexually transmitted diseases (STD), including HIV-1, and some studies have suggested that hormonal contraceptive use may even increase the risk of HIV-1. A meta-analysis of 28 studies found a significant association between HIV-1 infection and oral contraceptive pill use, with the strongest effect for studies conducted in Africa [odds ratio (OR) 1.65, 95% confidence interval (CI) 1.09-2.52] [²]. Two prospective studies among sex workers in Kenya and Thailand reported elevated risks of HIV-1 among women using the injectable contraceptive depot medroxyprogesterone acetate (DMPA) [^3,4]. However, other studies have found no relationship between either oral or injectable contraceptive use and incident HIV-1 [^2,5]. Most studies of this topic have been limited by imprecision in the measurement of contraceptive exposure and its relationship to the timing of HIV-1 acquisition, as well as by potential confounding by factors such as concurrent STD and sexual behavior. There thus remain insufficient data to make recommendations to women regarding the effect of contraceptive choices on the risk of HIV-1.

In 1993, we initiated a prospective open cohort study of HIV-1 acquisition among HIV-1-seronegative women attending a prostitute clinic in Mombasa, Kenya. Study procedures have previously been detailed [³]. At approximately monthly follow-up visits, sexual behavior and contraceptive use were recorded, risk reduction counselling was completed, free condoms were provided, and laboratory screening for HIV-1 and STD was performed.

In 1998, we reported that women in this cohort who used DMPA had a twofold (95% CI 1.3-3.1) greater risk of acquiring HIV-1 compared with women using no hormonal method, after controlling for sexual behavior, condom use, and STD [³]. In that analysis, oral contraceptive pill use was associated with increased HIV-1 risk, although this relationship did not reach statistical significance. Here we report an updated analysis from this cohort, now including 10 years of prospectively collected data, including 248 women who seroconverted to HIV-1.

As in our previous analysis, we estimated an exposure interval of 115 days for hormonal contraception in order to account for the time from infection until the detection of HIV-1 antibodies at a clinic visit, as well as for persistence of the effect of contraception if discontinued [³]. The comparison group was women reporting no contraceptive method or who had had a tubal ligation. We used an exposure interval of 60 days for STD and other genital infections [³]. Condom use was analysed as a separate covariate because condoms were used by many in the cohort for STD prevention, often in addition to another method for pregnancy prevention. Multivariate models also controlled for sexual behavior, demographic characteristics, and the occurrence of STD.

Between February 1993 and January 2003, 1498 women were enrolled in the cohort, of whom 1272 (85%) returned for follow-up. The median duration of follow-up was 478 days [interquartile range (IQR) 152-1273], the median number of follow-up visits was six (IQR 2-15), and visits were separated by a median of 35 days (IQR 28-55). A total of 15 428 follow-up visits were accumulated, reflecting 2931 person-years of follow-up, which is more than three times the accumulated follow-up of the previous study.

The median age at enrollment was 26 years (IQR 22-31). Sexual activity was relatively low in this group [median one (IQR 1-2) sexual partner and two (IQR 1-3) sexual encounters per week], because most participants (74%) had primary employment as barmaids and supplemented their income with commercial sex work. None reported injection drug use and only three (< 1%) practised anal sex, making heterosexual vaginal intercourse the principal HIV-1 risk factor for virtually all participants.

We used multivariate Cox proportional hazards models to analyse the association between hormonal contraceptive use and incident HIV-1 infection. DMPA use was associated with a significantly increased risk of HIV-1 acquisition [hazard ratio (HR) 1.8, 95% CI 1.4-2.4; Table 1], similar to our previously reported finding. The use of oral contraceptive pills was also associated with a significantly increased HIV-1 risk (HR 1.5, 95% CI 1.0-2.1). Women who used the implantable contraceptive Norplant were at increased risk of HIV-1, although this was not statistically significant. There was no increased HIV-1 risk among women using an intrauterine device, suggesting that our results were not a result of residual confounding among women using an effective modern contraceptive method. In a separate model, we included only those clinic visits that occurred after July 1997 (the cut-off for our previously published analysis), and found similar results as in the model covering the entire 10-year period (for DMPA, HR 1.9, 95% CI 1.2-2.9, and for oral contraceptive pills, HR 1.8, 95% CI 1.0-3.1).

Table 1

These results suggest that the use of both injectable and oral contraception may increase the risk of HIV-1 acquisition, independent of sexual behavior and STD exposures. Our study was conducted among African prostitutes, and our results may be most applicable to women at high-risk of HIV-1. However, although the rate of partner change for this cohort was high, the average sexual frequency was similar to that reported in surveys among general populations of African women. As the majority of women in this cohort used condoms at least sporadically for STD protection, irrespective of hormonal contraceptive use, our results may be less confounded by patterns of condom use than studies performed among lower-risk populations, such as women attending family planning clinics. Moreover, our data included monthly measurements of HIV-1 status, contraceptive use, sexual behavior, and STD, thus minimizing the potential for bias caused by the misclassification of either outcome or exposures. Given the widespread use of hormonal contraception in areas of high HIV-1 prevalence, our findings are concerning. Regardless of the method women choose for pregnancy prevention, healthcare providers must emphasize that condoms are the only method proved to prevent HIV-1 transmission. Women who use hormonal contraception, especially those at high risk of HIV-1, should be especially encouraged to use condoms consistently.