Human Immunodeficiency Virus, Antiretroviral Therapy, and Fertility in Sub-Saharan Africa: Pieces Still Missing in the Jigsaw Puzzle

(See the Major Article by Johnson et al., on pages 1953–61.)

Quantifying and understanding the effects of human immunodeficiency virus (HIV) infection and antiretroviral treatment (ART) on fertility is essential for accurate global and national HIV estimates—and especially for estimates of key indicators of the HIV epidemic in children, including coverage of prevention of mother-to-child transmission (PMTCT) and pediatric treatment programs, the contribution of long-term survival from mother-to-child transmission to HIV prevalence at adolescent and young adult ages, and the growing numbers of HIV-exposed but uninfected children, for whom exposure to certain ART drugs at conception may increase the risk of neural tube defects [1]. Equally—if not more—important, this information is needed to ensure that women on ART get appropriate counseling and access to family planning and sexual health services.

Prior to ART, lower fertility in women with HIV in sub-Saharan Africa was well-documented [2–5]. The relationship between HIV and fertility varied with age—in the youngest age group (15–19 years), fertility was higher among women with HIV, due to selection of sexually active women, whereas at older ages (≥25 years), the fertility of women with HIV became progressively lower than that of their uninfected counterparts. Biological and sociobehavioral mechanisms were hypothesized to explain HIV subfertility. Biological factors included the physiological and immunological effects of HIV infection, other sexually transmitted coinfections, and lower semen quality in male partners with HIV [6–8]. Sociobehavioral factors included reduced sexual activity in women with HIV compared to uninfected women due to poor health, greater widowhood and marital dissolution, and low rates of remarriage [9–11].

It had been widely anticipated that ART scale-up would attenuate many of these mechanisms, thereby lessening subfertility in women with HIV [12]. For example, the physiological and immunological effects of HIV in lowering fertility could be reduced if ART lessened the progression and severity of women’s HIV disease. Reduced widowhood and divorce, together with increased sexual activity because of improved health, could increase exposure to pregnancy for women with HIV compared to those in the pre-ART era. Fertility desires might also change. However, a systematic review by Yeatman and colleagues [13], which examined data collected largely during periods of ART scale-up, found a mixed picture and concluded that fertility in women with HIV may increase after the first year on ART but remains lower than in uninfected women of the same age.

In this issue of The Journal of Infectious Diseases, Johnson and colleagues [14] employ novel methods to compile and analyze a large dataset drawn from routine program records in the Western Cape in South Africa to provide further insight into the impact of ART on fertility. Using South Africa’s unique patient identifier, they linked data from women’s antenatal care visits with data on HIV status and ART. For women with evidence of at least 1 pregnancy, they used time from date of first pregnancy outcome to date of second pregnancy conception to compare pregnancy rates by HIV infection and ART status. Consistent with the studies conducted in the pre-ART period, the authors found that pregnancy rates in women with HIV not on ART were lower than in uninfected women. However, they also found that pregnancy rates in women on ART were higher than those in uninfected women.

This new study is important in that it strengthens the evidence that, among women with a recent pregnancy, the subsequent pregnancy rate in women on ART is higher than the rate for infected but ART-naive women; it also adds the finding that, in this population, their pregnancy rate also exceeds that in uninfected women. Unfortunately, however, key questions remain regarding fertility differentials in women who do not have a recent history of pregnancy and, therefore, about the generalizability of the study findings to all women in the population. Included in the Yeatman et al review [13], the only clinic-based studies that yielded a similar finding—that among women with HIV, those on ART had higher pregnancy rates than those who were ART naive—were also from populations of women who were recently pregnant [15] or pregnant at enrollment [16]. The second of these studies found no such difference in pregnancy rates for women with HIV who were not pregnant at enrollment [16]. In 1 of 2 population-based studies [17], a higher incidence of pregnancies was found in HIV-infected women on ART compared with those who were ART naive, but women not at risk of pregnancy, including those using hormonal contraceptives, were excluded from the analysis, which may have resulted in selection bias. In the second population-based study, no difference was found between women with HIV on ART compared to those who were ART naive [18]. These studies, therefore, taken together with the new study from South Africa, appear to show that ART raises fertility in women with HIV who have recently been pregnant—and thereby have demonstrated that they are already fecund despite being infected—but that ART has little or no impact on fertility for those who have not been pregnant recently.

For women with HIV who have had a recent pregnancy, as in other studies [15, 16], Johnson and colleagues found that pregnancy incidence rates were higher in women on ART than in ART-naive women even after controlling for CD4 cell count. This suggests that there are mechanisms independent of physiological improvements in health that contribute to increased pregnancy incidence. The authors point to reasons such as increased fertility desires or increased motivation to start ART in women who want more children. Another possibility is that ART may lower the efficacy of some hormonal contraceptives [19] that are used by 31% of women in the Western Cape [20]. Currently, it is unknown whether the high pregnancy rate in women with HIV on ART compared to uninfected women will continue or whether this is a tempo effect, such that completed family size will eventually be the same in the 2 groups.

For HIV-infected women who have not had a recent pregnancy, there are a number of mechanisms that could limit the effect of ART in reversing HIV-associated subfertility. These include that these women may have no sexual partner, may be using condoms or other effective forms of contraception, or may be infertile—perhaps due to a history of other sexually transmitted infections such as gonorrhoea or chlamydia [21].

The impact of ART on fertility at the population level will be a combination of its impact on the fertility of women with HIV who have and have not had recent pregnancies. Although Johnson and colleagues provide evidence that ART increases fertility in women who have recently been pregnant, once women who have not recently had a child or who have never been pregnant are included, the difference in fertility between women on ART and HIV-uninfected women may disappear or be reversed. The result may also depend on the composition of the population on ART in terms of how they enroll into care (via antenatal or counseling and testing services) and overall ART coverage.

Several vital pieces of the jigsaw, therefore, remain missing for modeling the population-level effects of ART on fertility and providing reliable estimates of the impact of HIV, PMTCT, and ART on children. In addition to those mentioned above, further pieces that remain poorly understood include that (1) initially, ART scale-up may increase the fertility of women who are not yet on treatment—in absolute terms and relative to that of uninfected women—due to selective ART initiation in women at more advanced stages of infection; (2) changes in policy over time can change the population on ART; for example, the change in World Health Organization guidelines to option B+, where women are immediately offered treatment for life regardless of their CD4 cell count, may have increased the proportion of healthier, recently pregnant women on ART; (3) the impact of ART may vary by region due to differences in the underlying reasons for HIV subfertility; and (4) ART scale-up may have indirect effects on the fertility of HIV-negative women—for example, through its effects in reducing HIV transmission and in increasing risk compensation and unprotected sex.

The most effective way to fill in these missing pieces of the puzzle and build on the study by Johnson et al could be to collect and analyze more data in prospective general population HIV serosurveys, as has been done in Zimbabwe and Malawi [18, 22] and may be possible in other studies in the Analysing Longitudinal Population-based HIV/AIDS data on Africa network [23], particularly those with links to routine data from local care and treatment (CTC) clinics. This would allow further comparisons of fertility rates between women with HIV—by HIV diagnosis and ART status—and uninfected women, investigation of potential mechanisms of ART impact in these groups, and evaluation of the biases inherent in analyses of clinic-based data. Data from CTC clinics could provide reliable information on patterns and trends in the composition of women on ART (by age, CD4 cell count, pregnancy status at initiation) and on how these women were referred for initiation. This could also help us to understand the reasons for observed fertility differences (or lack of differences) between women on and off ART, and to infer how the population not seen at CTC clinics might be changing. Networks such as the International Epidemiology Databases to Evaluate AIDS (IeDEA) Network [24], an international collaboration between CTC clinics including clinics in sub-Saharan Africa, are ideally placed to look at such data.

Note

Potential conflicts of interest. S. G. reports holding shares in GlaxoSmithKline and Astra Zeneca. M. M. reports no potential conflicts of interest. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References