Background: A single recent study has suggested a decrease in HIV risk for women attending antenatal clinics (ANCs) in southern India. Yet, some have questioned the validity of the Indian national surveillance data and analyses. Previous studies suggest that the only major HIV risk factor for married Indian women is the risk behavior of their husbands. Therefore, to address concerns about potential selection bias in the analysis of sentinel surveillance data from multiple sites, we estimated the trajectory of HIV transmission rates among recently married, monogamous, primigravid women attending a single large ANC in Pune, India.
Methods: Participants were self-referred, young, primigravid women from 18 to 27 years of age consenting to HIV screening. Time trends in HIV prevalence over 3.5 years were evaluated by logistic regression adjusted for age. HIV incidence was estimated by dividing the number of HIV-infected mothers by an estimate of exposure person-time, which was an estimate of the average age-specific duration of marriage.
Results: Between August 16, 2002 and February 28, 2006, 30,085 (79.5%) of 37,858 pregnant women consented to HIV screening; 10,982 (36.5%) were primigravid and their age range was from 18 to 27 years. HIV infection risk declined over 3.5 years among primigravid women. An estimated 19,739 person-years (PYs) of exposure yielded an overall HIV incidence rate 1.25/100 PYs (95% confidence interval [CI]: 1.10 to 1.42). Estimated HIV incidence decreased from 2.2/100 PYs (95% CI: 1.6 to 3.0) in 2002 to 2003 to 0.73/100 PYs (95% CI: 0.5 to 1.0) in 2006.
Discussion: HIV infection risk among young primigravid women in Pune seems to have decreased over the past 3.5 years. A decreasing HIV risk among pregnant women in Pune would also decrease the number of HIV-exposed infants. We hypothesize that decreased high-risk sexual behavior among young recently married men is most likely contributing to a decreasing risk to their wives and children in Pune.
From the *BJ Medical College (BJMC)-Johns Hopkins University (JHU) School of Medicine and Bloomberg School of Public Health Mother-to-Child Transmission (MTCT) Clinical Trial Unit, Pune, India; †JHU, Baltimore, MD; ‡Maharashtra Medical University, Nasik, India; and §BJMC, Pune, India.
Received for publication October 25, 2006; accepted May 14, 2007.
Supported by the National Institutes of Health (NIH; grant 1R01 AI45462-04) and an NIH Fogarty International Center Training Grant in HIV/AIDS Research (TW000010-AITRP).
The views expressed do not necessarily reflect the views of the BJMC/Sassoon General Hospital, JHU, National AIDS Research Institute, NIH, Pimpri Chinchwad Municipal Corporation, Pune Municipal Corporation, or Seth GS Medical College/King Edward Memorial Hospital.
Reprints: Robert C. Bollinger MD, MPH, Johns Hopkins Medical School, Ross 1150, 720 Rutland Avenue, Baltimore, MD 21205 (e-mail: firstname.lastname@example.org).
India's more than 5 million persons represent the largest burden of HIV infections in Asia.1 There have been recent controversies about the trajectory and characteristics of India's HIV epidemic, however.2-5 Since the first case of HIV infection in India in 1986, the epidemic has demonstrated wide geographic variability in HIV prevalence rates among the highest risk groups of patients attending sexually transmitted infection (STI) clinics, commercial sex workers (CSWs), and intravenous drug users.1 In addition, HIV prevalence rates among pregnant women have varied geographically and have not increased in India at the same rate as countries, such as South Africa, that demonstrated similar prevalence rates as India in the early 1990s.2 The reasons for the limited increase in HIV prevalence among pregnant Indian women, even in the highest prevalence states, remain unclear. Some have argued that the existing data and surveillance programs are limited and biased and underestimate the true prevalence of HIV infection.3
Until recently, published data on the risk of HIV infection among pregnant women in India have been limited to reports of point prevalence data from antenatal HIV screening programs. It has been reported that in contrast to many other settings with greater HIV prevalence in pregnancy, Indian women who are not CSWs are unlikely to have any other HIV risk factors than the risk behavior of their husband.6,7 Thus, non-CSW Indian women who have been infected with HIV by their husbands are not a major bridging population that would contribute to additional HIV transmissions in the community, except to their own infants. Although use of antenatal seroprevalence data has generally been shown to underestimate the community burden of HIV among women of childbearing age,8-11 temporal changes in antenatal HIV prevalence may reflect changes in risk for HIV transmission rates among women of childbearing age. A recent study has analyzed changes in antenatal clinic (ANC) HIV prevalence from sentinel surveillance data collected from the Indian National AIDS Control Organization (NACO).12 These data suggest a decline in the risk of HIV among pregnant women in southern India. The validity of this conclusion has been questioned because of a number of potential selection biases, including many acknowledged by the authors. During the study period, the number of participating NACO ANCs increased more than 3-fold, and although this recent analysis was adjusted for age, state, and educational level, the NACO data were not adjusted for clinic site. Use of overall antenatal prevalence also may not reflect actual changes in community incidence rates because of an increased death rate of HIV-infected women that would tend to lower the prevalence, and therefore underestimate the incidence, in the community. Because the risk factor for HIV among married Indian women of childbearing age is limited to exposure to a high-risk husband, changes in antenatal prevalence rates over time among recently married primigravid women may more accurately reflect changes in the community risk of incident HIV infection among women of childbearing age.
Therefore, limiting longitudinal ANC prevalence analysis to recently married primigravid women who lack HIV risk factors other than marriage to a high-risk husband may provide an opportunity to estimate temporal changes in HIV transmission rates. In addition, by estimating the HIV exposure period to be equivalent to the duration of marriage, an estimate of HIV incidence from age-adjusted HIV prevalence among primigravid women can be calculated. Such analyses over time, among primigravid women, may inform the ongoing debate about whether HIV prevention efforts are having any impact on the trajectory of the HIV epidemic in India and the validity of recent analyses of national sentinel surveillance data.
Participants were self-referred primigravid women aged 18 to 27 years who consented to HIV screening and presented to the ANC and delivery room of BJ Medical College (BJMC)/Sassoon General Hospital, a large urban public health hospital in Pune, India. After providing informed consent, all pregnant women who were not in the late stages of labor were offered routine predelivery HIV screening. Women who presented in late stages of labor were offered HIV screening within 24 hours of delivery. Pre- and posttest counseling was provided to all screened mothers. In addition, all HIV-infected mothers were offered standard short-course nevirapine or zidovudine for prevention of mother-to-child HIV transmission, according to the local standard of care. Women who consented to HIV screening also consented to collection and analyses of routine demographic and clinical data relevant for their prenatal and peripartum care. A subset of these HIV-screened pregnant women was enrolled into a prospective cohort study, where more in-depth sociodemographic data were collected. This study and the consent processes were reviewed and approved by 2 independent, Office of Human Research Protections (OHRP)-registered Institutional Review Boards at BJ Medical College in Pune, India and in Baltimore, Maryland at Johns Hopkins School of Medicine.
Diagnosis of HIV Infection
HIV infection was diagnosed with an algorithm using a series of commercially available rapid HIV blood tests.13 Fingerprick samples were collected and screened using the HIV-1/2 Determine (Abbott Laboratories, Abbott Park, IL) and the Naked Eye Visual Agglutination (NEVA) HIV-1/2 (Cadila Pharmaceuticals, Ahmedabad, Gurajat, India) rapid kits.
Logistic regression was applied to obtain age-adjusted odds ratios of HIV prevalence, calculated at 6-month intervals, over 42 months. As noted, pregnant women who were referred to the BJMC from outside physicians for confirmation of HIV status or for management of a high-risk pregnancy were excluded from the analyses. Only self-referred primigravidae presenting to the BJMC were included. HIV incidence was estimated by dividing the number of HIV-infected mothers by an estimate of exposure time for all screened primigravidae. The exposure time for the screened mothers was calculated by estimating the average age-specific duration of their marriage. The age-specific duration of marriage was estimated from more detailed demographic data collected from the subset of self-reported monogamous pregnant women who subsequently consented to enroll into 1 of 3 prospective cohort studies. The age-specific mean duration of marriage in this subset of pregnant women was used as an estimate of mean duration of marriage for the screened mothers. Period-specific incidence was calculated by dividing the number of HIV-positive women in that period by the estimated exposure time of women screened in same period. Reflected in our method for estimating HIV incidence among primigravid women was an additional key assumption that the only HIV risk factor for these mothers was sexual exposure to their husbands. In addition, our analysis assumed that the husbands of HIV-infected women were themselves HIV-positive before marriage and that the HIV exposure time for primigravidae was the mean duration of marriage. To determine time trends in HIV incidence, data were compared among 6-month intervals.
Between August 16, 2002 and February 28, 2006, 30,085 (79.5%) of 37,858 pregnant women consented to HIV screening; 10,982 (36.5%) were primigravid and 18 to 27 years of age. Overall, 247 (2.24%) primigravid women were infected with HIV. Figure 1 demonstrates that the risk of HIV infection among young, self-referred, primigravid women presenting to the BJMC antenatal service significantly declined over 3.5 years (P < 0.001). Relative to the HIV prevalence during the first 6-month study interval in 2002, the age-adjusted odds ratio was 0.31 (95% confidence interval [CI]: 0.19 to 0.50) during the last 6-month interval in 2005 to 2006, representing a 69% reduction in HIV risk over 3.5 years for young mothers in Pune who were pregnant for the first time.
Data collected from the subset of 306 screened mothers who were between 18 and 27 years of age and subsequently enrolled in one of our prospective cohort studies (49 HIV-uninfected and 257 HIV-infected women) demonstrated that HIV-infected and uninfected primigravidae were more likely to be from “joint families” than mothers with a parity >0. The term joint family refers to a typical Indian family structure that includes multiple generations, siblings, and related families living in the same household. Among HIV-infected mothers, primigravid women were also more likely to report their own and their husband's occupation as laborer (P < 0.001). Other characteristics were similar between primigravidae and other mothers. The reported age-specific duration of marriage among this subset of HIV-infected and uninfected primigravid women was applied to the entire screening population to calculate the estimated person-time (duration of marriage) for primigravid women, resulting in an estimated 24,216 person-years (PYs) of exposure. This yielded an overall HIV incidence rate of 1.02/100 PYs (95% CI: 0.90 to 1.16). Estimated HIV incidence decreased from 2.21/100 PYs (95% CI: 1.6 to 3.0) in 2002 to 2003 to 0.73/100 PYs (95% CI: 0.5 to 1.0) in 2006.
In Pune, India, the risk for HIV infection among young women who are pregnant for the first time seems to have decreased over the past 3.5 years. The reasons for this decreased HIV incidence cannot be determined from our study, because detailed information on the HIV risk behavior of all pregnant women who were offered screening was not available, nor was any risk information available from their husbands. In light of previous studies demonstrating that the risk for HIV among married Indian women is related to the risk behavior of their husbands,6,7 however, our data are consistent with the hypothesis that over the past 3.5 years, recently married husbands have become less likely to be HIV infected or they have become less likely to transmit HIV to their wives. This hypothesis is also supported by a recent independent study from Pune, which reports that HIV incidence has decreased among male patients with STIs and that this was associated with an increase in male condom use with exposure to female sex workers.14 Thus, young men in Pune have become less likely to become HIV infected over time. An expected consequence of this would be a decrease in risk for HIV infection among recently married primigravid women in Pune. This is exactly what our study has demonstrated. Our study also utilized a method to estimate HIV incidence among primigravid women who agreed to HIV screening at a large urban government hospital in Pune, India. Evidence for a decline in the risk for HIV infection over time among married, pregnant, primigravid women in Pune supports the conclusions of a recent national surveillance study.12
A lower risk for HIV infection among primigravid women would also reduce the burden of new infant HIV infections. These data suggest that HIV prevention efforts may be reducing risk behavior for recently married women in Pune. Although these data are encouraging, it is important to recognize that this study was limited to a single, large, urban public hospital in a single city in Maharashtra. It is not at all clear that these data from Pune are reflective of a more generalized national trend. Analyses of similar data over time among recently married primigravid women from many additional locations in India are necessary to generalize these findings.
Use of HIV prevalence data from ANC populations to generalize about the burden of disease in the general community must be done with caution because of many well-described potential biases.9-11 Our inability to collect detailed risk information on all participating pregnant women is one of a number of potential limitations of our study. Although a previous study in Pune demonstrated that the only risk factor independently associated with incident HIV infection among women with high-risk husbands attending STI clinics was sexual contact with their husbands7, it is possible that our assumption that married primigravid women are only at risk for HIV from their husbands is incorrect. Among the subset of our women with detailed social behavioral data available, 94% of the primigravid women reported that they were married at the time of screening. The remaining 6% reported that they were monogamous but separated, divorced, or widowed. None of this subset of primigravid women reported that they were previously married or never married. Although some pregnant women may have had more than 1 sexual partner, a previous national behavioral survey in India has reported that a small percentage (<2%) of married Indian women reported premarital or extramarital sexual contact, compared with more than 10% of their husbands.6 Again, our analyses assume that the subset of primigravid women for whom detailed marriage status data were available is representative of the entire study group of primigravid women. This is an important assumption and potential limitation. Nevertheless, it is unlikely that any potential bias imposed by this potential limitation on our analyses would have been time dependent. The key point of our study is that the risk of HIV among primigravid women in Pune has declined over time. A key rationale for restricting our analyses to young primigravid women between 18 and 27 years of age was to limit this potential bias. Excluding older primigravid women would limit the chance that we would include primigravid women over time in our analyses who may have had more than 1 lifetime sexual partner or multiple marriages.
Additionally, our assumption that the exposure period for the women was equivalent to the age-specific estimates of marriage duration could have resulted in an overestimation of exposure time. It is possible that their husbands were HIV infected after marriage, thus reducing the exposure period to their wives, resulting in our analysis underestimating HIV incidence. We conducted a sensitivity analysis to address this issue. Under one scenario, we assumed that husbands were infected just before pregnancy and transmitted the virus at the time of conception. The exposure time for these women was assumed to be equal to the gestational age of their baby at the time of HIV diagnosis. This yielded an overall HIV incidence rate of 4.21/100 PYs (95% CI: 3.70 to 4.77). Under these assumptions, our estimated HIV incidence decreased from 9.29/100 PYs (95% CI: 6.60 to 12.69) in 2002 to 2003 to 3.08/100 PYs (95% CI: 2.09 to 4.37). In another scenario, we assumed that 10% of husbands acquired infection slightly before pregnancy and that the remaining husbands were infected before marriage. Thus, for 10% of women, the exposure time was assumed to be the same as their baby's gestational age, and for the remaining 90%, it was the duration of marriage. This yielded an overall incidence rate of 1.13/100 PYs (95% CI: 1.00 to 1.28). A similar analysis was conducted with 50% of husbands being infected just before pregnancy, which yielded an estimated incidence rate of 1.68/100 PYs (95% CI: 1.48 to 1.90). The decreased HIV risk among primigravid women over time remained statistically significant when a sensitivity analysis was applied to our data, using all 3 of these scenarios. Therefore, it is unlikely that such potential biases would have a differential impact on our analyses over time, suggesting our conclusion that HIV infection rates have decreased among primigravid married women in Pune is still valid.
Finally, declines in HIV infection rates among pregnant women attending the BJMC could reflect a decrease in HIV risk for recently married women in the community served, or it could reflect an increase in access to HIV screening over time to lower risk women (ie, a change in the community served by HIV screening over time). The HIV screening acceptance rate among pregnant women also increased from 87% in 2002 to 94% in 2005 in our study overall. One possibility is that more high-risk women accepted voluntary counseling and testing (VCT) because of decreased fear and stigma associated with HIV testing. This would strengthen our conclusions. The alternative is also possible, however, and lower risk women may have become more likely to accept HIV testing later. The BJMC, a large, government-run, public hospital, was used for this analysis because of its >10-year history of providing ANC HIV screening to the lower socioeconomic women of Pune. Our analysis supports the conclusion that the basic social demographics (eg, age, educational level, socioeconomic status) of pregnant women attending the BJMC ANC have not changed over time but that the risk for HIV infection among primigravidae has decreased.
The investigators thank the participants for their help with this study. The authors also acknowledge and thank the members of the BJMC-National AIDS Research Institute (NARI)-Johns Hopkins University Mother-to-Child Transmission Study Group who contributed to the conduct of this study through collection of data, clinical management of study participants, and project support, particularly the project and medical staff at the BJMC/Sassoon General Hospital, NARI, Pimpri Chinchwad Municipal Corporation, Pune Municipal Corporation, Seth GS Medical College/King Edward Memorial Hospital, and Johns Hopkins University (JHU) School of Medicine and Bloomberg School of Public Health.
2. Prasada Rao JVR, Ganguly NK, Mehendale SM, et al. India's response to the HIV epidemic. Lancet
4. Sharma DC. India challenges UN agencies' estimates of HIV prevalence. Lancet
5. Mandavilli A. The coming epidemic. Nature
7. Gangakhedkar RR, Bentley M, Divekar A, et al. Spread of HIV infection in married monogamous women in India. JAMA
8. United Nations Program on HIV/AIDS, Group on Estimates, Modelling and Projections. Improved methods and assumptions for estimation of the HIV/AIDS epidemic and its impact: Recommendations of the UNAIDS Reference Group on Estimates, Modeling and Projections. AIDS
9. Zaba B, Boerma T, White R. Monitoring the AIDS epidemic using HIV prevalence data among young women attending antenatal clinics: prospects and problems. AIDS
10. Glynn JR, Buve A, Carael M, et al, for the Study Group on Heterogeneity of HIV Epidemics in African Cities. Factors influencing the difference in HIV prevalence between antenatal clinic and general population in sub-Saharan Africa. AIDS
11. Gregson S, Terceira N, Kakowa M, et al. Study of bias in antenatal clinic HIV-1 surveillance data in a high contraceptive prevalence population in sub-Saharan Africa. AIDS
12. Kumar R, Jha P, Arora P, et al, for the International Studies of HIV/AIDS (ISHA) Investigators. Trends in HIV-1 in young adults in south India from 2000 to 2004: a prevalence study. Lancet
13. Bhore AV, Sastry J, Patke D, et al. Sensitivity and specificity of rapid HIV testing of pregnant women in India. Int J STD AIDS
14. Mehendale SM, Gupte N, Paranjape RS, et al. Declining HIV incidence among patients attending sexually transmitted infection clinics in Pune, India. J Acquir Immune Defic Syndr
. 2007; June 7 (e-pub ahead of print).