Several studies have suggested that pregnancy is associated with an increased risk of HIV-1 acquisition [1–3]. In addition, a national HIV survey conducted in South Africa identified pregnancy as a risk factor for HIV acquisition . Only one of those studies, however, adjusted for behavioral factors that might influence HIV-1 risk during pregnancy . After concluding that the risk associated with pregnancy was not the result of changes in sexual behaviors of either the women or their sex partners, these researchers suggested that the high levels of estrogen and progesterone during pregnancy may increase susceptibility to HIV by inducing changes in the genital mucosa or through immunological effects .
The issue of whether pregnancy is associated with an increased risk of HIV-1 acquisition is of considerable importance. Whereas HIV-1 infection of the mother is of serious concern, HIV acquisition is followed by a period of high viremia, which in pregnant women is associated with increased maternal-to-child HIV-1 transmission and more rapid disease progression among HIV-infected infants [5–7]. There is also evidence of an increased potential for adverse birth outcomes among HIV-infected women [8–12]. Moreover, regions with high HIV-1 prevalence are also areas where women experience high levels of fertility. For example, in sub-Saharan Africa, women experience an average of 5.6 births, hence spending a considerable portion of their reproductive years pregnant. Therefore, if pregnancy is a risk factor for HIV-1 acquisition, women need to understand this risk before becoming pregnant (and when considering options for contraception), and antenatal programmes should focus additional efforts on the prevention of HIV-1 acquisition during pregnancy.
We used data from a prospective study of hormonal contraception and HIV-1 acquisition in Uganda and Zimbabwe  to assess the effect of pregnancy on the risk of HIV-1 acquisition and to evaluate whether any increase in risk appears to be caused by behavior or the physiological changes associated with pregnancy.
The study was approved in Uganda by the Office of Research Administration AIDS Research Committee and in Zimbabwe by the Biomedical Research and Training Institute and the Medical Research Council. All women provided written informed consent before joining the study.
The study methods have been described in detail elsewhere . Briefly, between 1999 and 2004, we enrolled and followed women seeking reproductive and general healthcare services in Uganda and Zimbabwe. Participants were aged 18–35 years, were neither pregnant nor intending to become pregnant within the next year, were HIV-1 uninfected, sexually active, and had been using either no hormonal contraceptive method, combined oral contraceptive (COC) pills containing 30 μg ethinyl estradiol and 150 μg levonorgestrel or 150 mg depot-medroxyprogesterone acetate (DMPA) administered every 12 weeks for at least 3 months. Women were ineligible if they had a hysterectomy, or had used an intrauterine device or had a spontaneous or induced abortion within the previous month. Each site attempted to enroll equal numbers of women into the three contraceptive groups.
We also recruited a smaller group of ‘high-risk referral’ women in Uganda (398 women) because of low initial HIV-1 incidence at that study site. These women were recruited from sexually transmitted infection (STI) or primary health care clinics (women with STI symptoms), sex worker networks, or military bases.
The women enrolled were followed quarterly for 15–24 months. Study visits included a physical examination, pregnancy, HIV-1, and STI testing, and administration of a standardized questionnaire.
Pregnancy testing was performed using a urine test for human chorionic gonadotropin (hCG). Clinicians also recorded clinical impressions concerning possible pregnancy on the physical examination form. In the event of discordant results between the pregnancy test and the clinical impression item, the latter variable was used. This occurred primarily when a pregnancy was recently completed and the hCG test remained positive.
We tested for HIV-1 using an algorithm described in detail elsewhere . For confirmed incident HIV-1 infections, HIV-1-DNA polymerase chain reaction (PCR) was performed retrospectively on serial visit specimens in order to identify more precisely the timing of infection. We defined the date of HIV-1 acquisition as the date of the first positive PCR result.
PCR testing for gonorrhea and chlamydia (Amplicor CT/NG; Roche Diagnostic Systems, Branchburg, New Jersey, USA) was conducted on endocervical specimens. We tested serum for herpes simplex virus 2 (HSV-2) antibodies utilizing an enzyme-linked immunosorbent assay (Focus Diagnostics, Cypress, California, USA) . Vaginal infections were diagnosed by microscopy: trichomonas by motile flagellates and yeast infection by yeast forms or pseudohyphae. Bacterial vaginosis was diagnosed according to Amsel criteria .
Analysis population and variable definition
The analysis population consisted of participants with at least one follow-up visit with valid pregnancy and HIV-1 results. The outcome was the number of days from the baseline visit to the earlier of the date of the first positive HIV-1 result or the last study contact.
We differentiated between women who were using and not using hormonal contraception in our primary exposure variable. We defined four exposure categories: (i) pregnant at the current visit; (ii) not currently pregnant but lactating (NP/L) since the last study visit; (iii) not pregnant and not lactating (NP/NL) but using hormonal contraception (either COC or DMPA) since the last visit; and (iv) not currently pregnant or lactating and not using hormonal contraceptives since the last visit (comparison group).
To estimate the effects of time-varying pregnancy exposure, we divided a participant's time into segments corresponding to the period between study visits. This allowed us to capture changes between study visits in pregnancy and lactation status, hormonal contraceptive use, sexual risk behaviors, and the presence of STI, and to account for these changes in the analyses.
We also defined two composite sexual risk variables, one to summarize the study participant's behavioral risk and the other to summarize her primary partner's risk. The participant behavioral risk variable indicated whether study participants had multiple partners, a new sex partner, or had engaged in commercial sex within the past 3 months. The primary partner risk variable indicated whether a participant had a partner who was HIV positive, who had penile discharge or significant weight loss, who had commercial sex partners, or who had spent nights away from home during the past 3 months.
The 211 incident HIV-1 infections diagnosed (including 13 in the pregnancy and 39 in the NP/NL and no hormonal contraception use group) provided 80% power to detect a 2.2-fold difference between the pregnant and the NP/NL and no hormonal contraception use groups.
Comparisons of characteristics among the pregnancy exposure groups were assessed using chi-squared test statistics, which were adjusted for clustering for time-varying covariates .
The relationship between pregnancy status and incident HIV-1 infection was assessed using a Cox proportional hazards regression analysis of time to HIV-1 infection. Variables considered a priori to be important (site, age, condom use) were retained in multivariate models. Participant sociodemographic characteristics, baseline health and reproductive health history, sexual behavior, and primary partner characteristics were assessed as potential confounding factors and retained in multivariate models as previously described .
To assess the robustness of our results, we performed a number of sensitivity analyses on the final multivariate model: (i) testing two variables, site and age, for effect modification of the pregnancy and HIV-1 acquisition relationship; (ii) the addition of time-varying STI including vaginal (trichomoniasis, bacterial vaginosis, candidiasis), cervical (chlamydia, gonorrhea), and HSV-2 infections to assess the effect of adjusting for these infections on the pregnancy effect; (iii) redefining the four-level pregnancy variable into five levels so that pregnancy for the whole segment versus pregnancy for part of the segment could be compared and a dose–response effect evaluated; (iv) redefining the four-level pregnancy variable into five levels so that pregnancy segments were divided into those without any breastfeeding or hormonal contraceptive use compared with those in which either breastfeeding or hormonal contraceptive use also occurred during the segment; (v) redefining the pregnancy variable so that it included not only women who were pregnant at the clinical visit but also women who reported a pregnancy since her last study visit; and (vi) limiting segments used in the analyses to only those in which a women reported unprotected sex since her last study visit.
Of 4531 women participating in the study (2235 in Uganda, 2296 in Zimbabwe), 92 women were excluded because of no follow-up or the exclusive use of non-study contraceptive methods. An additional 24 women were excluded because of missing pregnancy data. A total of 4415 women thus contributed a total of 31 106 visit segments to the analysis.
At enrollment, the median age of study participants was 25 years, median education was 10 years, and most participants lived with a partner (Table 1). Study participants had a median of two lifetime pregnancies and 28% breastfed at enrollment. At baseline, 34.7% of participants used COC, 34.2% used DMPA, and 31.1% did not use a hormonal method. Few participants reported multiple sex partners, commercial sex, or sex while using alcohol or drugs, whereas less than half reported recent condom use.
The 24-month retention rate was 92%; 96% in Uganda and 88% in Zimbabwe. The mean follow-up was 21.9 months; the median time between study visits was 81 days.
Characteristics of pregnancy exposure groups during follow-up
Among the 31 106 visit segments included in the analysis, women were pregnant in 2867 visit segments (9.2%), NP/L in 4699 segments (15.1%), NP/NL but using hormonal contraception in 17 668 segments (56.8%), and NP/NL and not using hormonal contraception in 5872 segments (18.9%). More pregnancy segments were contributed by Ugandan (61%) than by Zimbabwean women (39%) (Table 1). Pregnancy was more common among younger women (18–24 years), women living with a partner and among women with less than three lifetime pregnancies at study enrollment. No important difference was seen in pregnancy occurrence by educational level.
During follow-up, pregnant women were less likely to have multiple partners, to have a new sex partner or to use alcohol or drugs during sex compared with the NP/NL and no hormonal contraception use group (Table 1). Correspondingly, fewer pregnant women had high participant behavioral risk and primary partner risk (2.7 and 38.8%) compared with women in the NP/NL and no hormonal contraception use group (4.8 and 55.6%). On the other hand, pregnant women were much more likely to report some unprotected sex during visit intervals (82.0%) than were women in the NP/NL and no hormonal contraception use group (37.8%). Although chlamydia, gonorrhea, trichomonas and yeast infection were diagnosed somewhat more often during follow-up among pregnant women than among women in the NP/NL and no hormonal contraception use group, little difference was seen in the diagnosis of bacterial vaginosis and HSV-2 infection between the groups (Table 1).
Incident HIV-1 infections by pregnancy exposure groups
There were 211 incident HIV-1 infections, giving an incidence rate of 2.73 per 100 woman years (wy) (4.05 per 100 wy in Zimbabwe and 1.53 per 100 wy in Uganda). Thirteen incident HIV-1 infections occurred among pregnant women (1.64 per 100 wy) and 33 (2.72 per 100 wy), 126 (2.94 per 100 wy) and 39 (2.70 per 100 wy) incident HIV-1 infections occurred among NP/L women, NP/NL women using hormonal contraception, and NP/NL women not using hormonal contraception, respectively (Table 2).
Neither pregnancy nor lactation was significantly associated with the risk of HIV-1 acquisition in either univariate [hazards ratio (HR) 0.56; 95% confidence interval (CI) 0.30–1.05 and HR 1.04; 95% CI 0.65–1.66, respectively] or multivariate analyses (HR 0.60; 95% CI 0.31–1.16 and HR 1.07; 95% CI 0.65–1.76, respectively, Table 3). Covariates significantly associated with increased HIV-1 acquisition included Zimbabwe site and Uganda high-risk referral group, not living with a partner, young age, high participant behavioral risk, high primary partner risk, and recent alcohol use.
Effect modification of the pregnancy and HIV-1 acquisition relationship
We found evidence of statistically significant effect modification by both site (P = 0.010) and by age (P = 0.003) when adjusting for covariates. With respect to site, a statistically significant protective effect existed for pregnancy in Zimbabwe (HR 0.26; 95% CI 0.10–0.68), whereas in Uganda a non-significant increased risk associated with pregnancy occurred in the general population (HR 2.32; 95% CI 0.74–7.27) and a non-significant protective effect in the high-risk referral population (HR 0.50; 95% CI 0.05–4.89) even after adjusting for sexual behavior and sociodemographic covariates.
In younger women, pregnancy had no important effect on HIV-1 acquisition (HR 1.14; 95% CI 0.47–2.80). Among older women (≥ 25 years), however, pregnancy had a protective effect on HIV-1 acquisition (HR 0.37; 95% CI 0.13–1.09), although this did not reach statistical significance.
Results of sensitivity analyses
In general, the sensitivity analyses did not alter our primary study findings. First, adjusting for the effects of time-varying vaginal, cervical and HSV-2 infections in the multivariate model (Table 3) did not have an important impact on the primary results for either the pregnant (HR 0.54; 95% CI 0.26–1.14) or the not pregnant, lactating groups (HR 1.10; 95% CI 0.67–1.83). Second, when pregnancy exposure segments were split into groups in which women were pregnant for the entire compared with part of the segment, little difference existed between the two groups (HR 0.62; 95% CI 0.22–1.77 and HR 0.60; 95% CI 0.28–1.25, respectively). Third, when pregnancy exposure segments were split into groups in which (i) women were pregnant but not breastfeeding or using hormonal contraception (pregnancy only) during the segment, and (ii) women were pregnant but also either breastfed or used hormonal contraception during part of the segment, no statistically significant differences were found between either pregnancy group and the NP/NL and no hormonal contraceptive use group (HR 0.42; 95% CI 0.15–1.19 and HR 0.75; 95% CI 0.35–1.59, respectively), although the hazard ratio was somewhat lower for the ‘pregnancy only’ group. Fourth, redefining pregnancy to include self-reported pregnancies in addition to pregnancies diagnosed at the clinical examination (thus adding 209 pregnancy segments) also resulted in no statistically significant difference between the pregnancy and the NP/NL, no hormonal contraception use groups (HR 0.77; 95% CI 0.43–1.37). Finally, limiting the analysis only to segments in which women reported unprotected sex (resulting in 76% of segments being retained) did not appreciably change the pregnancy effect (HR 0.56; 95% CI 0.26–1.22).
We found that neither pregnancy nor lactation placed women at increased risk of HIV-1 acquisition in this multisite, prospective study of African women. The results held true in various sensitivity analyses, in all site and age subgroups and was similar in univariate and multivariate analyses. These results should provide reassurance for women desiring to become pregnant (and who would subsequently breastfeed) in areas with high HIV-1 prevalence.
There are several possible reasons why our results differ from the results of previous studies examining this relationship [1–3]. First, only one of the previous studies was prospective and adjusted for demographic, reproductive health and sexual risk behaviors , whereas none of the previous studies adjusted for hormonal contraceptive use, an important possible confounder given the hypothesis that high levels of estrogen and progesterone during pregnancy could increase women's susceptibility to HIV acquisition . A second possibility for the differences in the results could be caused by the varying study populations with associated differences in pregnancy rates, behavioral and STI co-factors and HIV-1 exposure risk. For example, although our study results for Zimbabwean women were quite different from those from Rakai , our results for the Ugandan general-population group (the group most closely resembling the Rakai population) were more similar.
The high levels of estrogen and progesterone found during pregnancy are associated with structural and immunological changes that could potentially increase HIV acquisition. Many of these changes are similar to those associated with hormonal contraceptive use that have been described in detail elsewhere . Briefly, the possible mechanisms include increased cervical ectopy probably associated with increased estrogen levels [17–20], increased cervical chlamydial infection (and associated purulence) [19,21–23], increased recruitment of inflammatory and other target cells to the genital tract [24–27], or through a direct effect on the infecting virus inoculum by upregulating HIV-1 gene expression and associated viral replication . In addition, suppression of the local cell-mediated and systemic immune responses that limit the rejection of the developing fetus could also increase HIV-1 susceptibility [8,24–26,28,29].
Although this study suggests that pregnancy appears not to confer additional physiological risks of HIV-1 acquisition, unprotected sex (unless in a known monogamous relationship) is clearly a behavioral risk for HIV-1 acquisition. Similar to women using hormonal contraception (and as illustrated by this study), pregnant women and their partners do not generally use condoms consistently. Because there is evidence of high levels of HIV infection [11,30] among pregnant and lactating women, it remains very important for antenatal and postnatal programmes to stress the need for condom use to protect both mother and baby from HIV infection during the perinatal period. In addition, there is an important need to involve the partners of pregnant women in risk reduction strategies (reduction in numbers of their sexual partners, condom use, etc.) when possible and to understand better the sexual behaviors of pregnant women and their partners that confer HIV acquisition risk to pregnant and lactating women.
Our study had a number of important strengths. First, pregnancy and HIV-1 infection were measured systematically every 12 weeks: pregnancy using urine hCG testing and HIV-1 infection was determined using a standardized testing algorithm that minimized errors and accurately timed the infection. Second, important covariates including hormonal contraceptive use were carefully measured and either used as part of the exposure definition or carefully evaluated for their potential confounding effects on study results. Third, the study was large, with adequate power to detect differences between the pregnant, lactating and non-pregnant, non-lactating groups. Fourth, retention rates were high. Fifth, we conducted sensitivity analyses that confirmed the robustness of our study results under different assumptions. Finally, the study was conducted primarily among women from the general population in two different countries, thus enhancing the generalizability of the results.
Our study also has limitations. Studies of pregnancies and HIV-1 acquisition cannot be randomized and therefore selection and confounding biases cannot be ruled out. Second, although we accurately measured pregnancy every 12 weeks, we were unable to measure the onset of pregnancy more precisely. Third, the measurement of sexual risk behaviors was self-reported and the accuracy of such data is unknown. Inaccurate or incomplete measurement of important co-variates could result in residual confounding. Nevertheless, the measurement of pregnancy, HIV-1 status and hormonal contraceptive use (hormonal contraceptives were distributed or administered by the study and reports were validated against clinic records) to define study exposure groups and outcomes were accurate and provide confidence in the overall study results. Finally, the fact that at enrollment, study participants were not pregnant and did not desire to become pregnant during the following year could limit the generalizability to some groups of women.
In conclusion, neither pregnancy nor lactation appears to increase the risk of HIV-1 acquisition. We believe that these results should provide reassurance for women desiring to become pregnant in areas with a high prevalence of HIV-1. Furthermore, this information is important in contraceptive counseling and in planning interventions to reduce HIV-1 acquisition among women.
Sponsorship: This project was funded with federal funds from the National Institute of Child Health and Human Development, National Institutes of Health, and Department of Health and Human Services through a contract with Family Health International (contract no. N01-HD-0-3310).
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services or Family Health International, nor does the mention of trade names, commercial products, or organizations imply endorsement by the US government.
Conflicts of interest: None.
1. Taha TE, Hoover DR, Dallabetta GA, Kumwenda NI, Mtimavalye LA, Yang LP, et al
. Bacterial vaginosis and disturbances of vaginal flora: association with increased acquisition of HIV. AIDS 1998; 12:1699–1706.
2. Gray RH, Wabwire-Mangen F, Kigozi G, Sewankambo NK, Serwadda D, Moulton LH, et al
. Randomized trial of presumptive sexually transmitted disease therapy during pregnancy in Rakai, Uganda. Am J Obstet Gynecol 2001; 185:1209–1217.
3. Gray RH, Li X, Kigozi G, Serwadda D, Brahmbhatt H, Wabwire-Mangen F, et al
. Increased risk of incident HIV during pregnancy in Rakai, Uganda: a prospective study. Lancet 2005; 366:1182–1188.
4. Shisana O, Rehle T, Simbayi LC, Parker W, Zuma K, Bhana A, et al
. South African national HIV prevalence, HIV incidence, behavior and communication survey, 2005. Cape Town, South Africa: HSRC Press; 2005.
5. Taha TE, Gray RH. Genital tract infections and perinatal transmission of HIV. Ann NY Acad Sci 2000; 918:84–98.
6. Leroy V, Montcho C, Manigart O, Van de Perre P, Dabis F, Msellati P, et al
. Maternal plasma viral load, zidovudine and mother-to-child transmission of HIV-1 in Africa: DITRAME ANRS 049a trial. AIDS 2001; 15:517–522.
7. Ioannidis JP, Tatsioni A, Abrams EJ, Bulterys M, Coombs RW, Goedert JJ, et al
. Maternal viral load and rate of disease progression among vertically HIV-1-infected children: an international meta-analysis. AIDS 2004; 18:99–108.
8. Gray GE, McIntyre JA. Effect of HIV on women. AIDS Reader 2006; 16:365–368, 373–377.
9. Suy A, Martinez E, Coll O, Lonca M, Palacio M, de Lazzari E, et al
. Increased risk of pre-eclampsia and fetal death in HIV-infected pregnant women receiving highly active antiretroviral therapy. AIDS 2006; 20:59–66.
10. Ellis J, Williams H, Graves W, Lindsay MK. Human immunodeficiency virus infection is a risk factor for adverse perinatal outcome. Am J Obstet Gynecol 2002; 186:903–906.
11. Leroy V, Ladner J, Nyiraziraje M, De Clercq A, Bazubagira A, Van de Perre P, et al
. Effect of HIV-1 infection on pregnancy outcome in women in Kigali, Rwanda, 1992-1994. AIDS 1998; 12:643–650.
12. Coley JL, Msamanga GI, Fawzi MCS, Kaaya S, Hertzmark E, Kapiga S, et al
. The association between maternal HIV-1 infection and pregnancy outcomes in Dar es Salaam, Tanzania. Br J Obstet Gynaecol 2001; 108:1125–1133.
13. Morrison CS, Richardson BA, Mmiro F, Chipato T, Celentano DD, Luoto J, et al
. Hormonal contraception and the risk of HIV acquisition. AIDS 2006; 21:85–95.
14. Ashley-Morrow R, Krantz E, Wald A. Time course of seroconversion by HerpeSelect ELISA after acquisition of genital herpes simplex virus type 1 (HSV-1) or HSV-2. Sex Transm Dis 2003; 30:310–314.
15. Schwebke JR, Hillier SL, Sobel JD, McGregor JA, Sweet RL. Validity of the vaginal gram stain for the diagnosis of bacterial vaginosis. Obstet Gynecol 1996; 88:573–576.
16. Research Triangle Institute. SUDAAN user's manual, release 8.0
. Research Triangle Park, NC: Research Triangle Institute; 2002.
17. Jacobson DL, Peralta L, Farmer M, Graham NM, Gaydos C, Zenilman J. Relationship of hormonal contraception and cervical ectopy as measured by computerized planimetry to chlamydial infection in adolescents. Sex Transm Dis 2000; 27:313–319.
18. Brabin L. Interactions of the female hormonal environment, susceptibility to viral infections, and disease progression. AIDS Patient Care STD 2002; 16:211–221.
19. Louv WC, Austin H, Perlman J, Alexander WJ. Oral contraceptive use and the risk of chlamydial and gonococcal infections. Am J Obstet Gynecol 1989; 160:396–402.
20. Bright PL. A longitudinal investigation of cervical ectopy
. UMI Disertation Services, 3086503. Chapel Hill, NC: University of North Carolina at Chapel Hill; 2003.
21. Plourde PJ, Pepin J, Agoki E, Ronald AR, Ombette J, Tyndall M, et al
. Human immunodeficiency virus type 1 seroconversion in women with genital ulcers. J Infect Dis 1994; 170:313–317.
22. Morrison CS, Bright P, Wong EL, Kwok C, Yacobson I, Gaydos CA, et al
. Hormonal contraceptive use, cervical ectopy, and the acquisition of cervical infections. Sex Transm Dis 2004; 31:561–567.
23. Baeten JM, Nyange PM, Richardson BA, Lavreys L, Chohan B, Martin HL Jr, et al
. Hormonal contraception and risk of sexually transmitted disease acquisition: Results from a prospective study. Am J Obstet Gynecol 2001; 185:380–385.
24. Abel K, Rourke T, Lu D, Bost K, McChesney MB, Miller CJ. Abrogation of attenuated lentivirus-induced protection in rhesus macaques by administration of depo-provera before intravaginal challenge with simian immunodeficiency virus mac239. J Infect Dis 2004; 190:1697–1705.
25. Prakash M, Kapembwa MS, Gotch F, Patterson S. Oral contraceptive use induces upregulation of the CCR5 chemokine receptor on CD4(+) T cells in the cervical epithelium of healthy women. J Reprod Immunol 2002; 54:117–131.
26. Martin HL Jr, Nyange PM, Richardson BA, Lavreys L, Mandaliya K, Jackson DJ, et al
. Hormonal contraception, sexually transmitted diseases, and risk of heterosexual transmission of human immunodeficiency virus type 1. J Infect Dis 1998; 178:1053–1059.
27. Ghanem KG, Shah N, Klein RS, Mayer KH, Sobel JD, Warren DL, et al
. Influence of sex hormones, HIV status, and concomitant sexually transmitted infection on cervicovaginal inflammation. J Infect Dis 2005; 191:358–366.
28. Sonnex C. Influence of ovarian hormones on urogenital infection. Sex Transm Infect 1998; 74:11–19.
29. Beagley KW, Gockel CM. Regulation of innate and adaptive immunity by the female sex hormones oestradiol and progesterone. FEMS Immunol Med Microbiol 2003; 38:13–22.
30. Alliegro MB, Dorrucci M, Phillips AN, Pezzotti P, Boros S, Zaccarelli M, et al
. Incidence and consequences of pregnancy in women with known duration of HIV infection. Italian Seroconversion Study Group. Arch Intern Med 1997; 157:2585–2590.