The specific objectives of this study were 1) to compare the baseline characteristics of HIV-infected women who did or did not became pregnant following their initiation of ART, 2) to analyze trends in pregnancy incidence, desire for children and sexual activity over time among these women following ART initiation, 3) to describe the rate of unmet need for family planning and the rate of contraceptive use among these women after 2 years of follow-up, and 4) to assess independent risk factors for pregnancy over 2 years of follow-up post ART initiation.

Women included in this analysis were part of the Home-Based AIDS Care (HBAC) study, a randomized trial of three different ART monitoring strategies involving 1,092 HIV-infected men and women in Tororo and Busia districts in rural Eastern Uganda [15] (Figure 1). HBAC study participants were recruited among clients of the Tororo branch of The AIDS Support Organization (TASO), an indigenous non-governmental organization providing medical and psycho-social care and support for PLHA. Starting in March 2003, TASO clients ≥18 years-old were screened for ART eligibility and were offered first-line ART if they had a CD4 cell count <250 cells/µL or were in WHO disease stage 3 or 4. All households participating in the study received insecticide-treated bed nets to reduce malaria incidence and a water vessel with a supply of chlorine for diarrheal disease prevention. All household members were offered home-based HIV counseling and testing; 99% of them accepted and those diagnosed with HIV were offered couple counseling, were provided with free care and treatment, and were offered enrollment in HBAC [21]. Among HBAC participants, 97% of sexually active women and 98% of sexually active men had disclosed their HIV sero-status to a median of 15 people at baseline, including their sexual partner (Nakayiwa S. et al., unpublished data), reflecting TASO's long-standing ‘positive living’ philosophy which includes empowering PLHA to disclose their HIV sero-status [22].

As part of the enrollment into HBAC, participants underwent physical examinations and answered in-depth questionnaires to evaluate their social, behavioral, and economic baseline characteristics as well as clinical history. Study participants were counseled at enrolment about the potential effects of ART on improving health status, restoring fertility and sexual activity, and were offered referral to the district hospital family planning (FP) clinic that was adjacent to the study clinic if interested in using any FP method other than condoms. The hospital FP clinic provided standard FP counseling and free FP services including any number of clinic visits and contraceptives supplied by the Ministry of Health including combined oral and progesterone-only pills, hormonal injectables and implants, intra-uterine devices, tubal ligation, and condoms. Condoms were also available for free from all HBAC clinicians, counselors and home visitors, both at the study clinic and during home visits.

After enrollment, HBAC participants were visited in their homes on a weekly basis by trained lay workers who delivered drugs, conducted pill counts and monitored participants' clinical status using standardized checklists. No routine clinic visits were scheduled but participants could come to the clinic or hospital when experiencing acute symptoms or showing signs of potential drug toxicity or disease progression. There were no clinic visits fees or any other charge associated with the care provided by HBAC. Home visitors weighed participants and asked women for the date of their last menstrual period on a monthly basis. Women were tested for pregnancy if they had missed their periods for more than 4 weeks. Home visitors also drew blood from participants on a quarterly basis for measurement of CD4 count and viral load. HBAC counselors visited participants quarterly in their homes to provide systematic confidential counseling for psycho-social support, treatment adherence and prevention of sexual transmission of HIV including safer sex and family planning counseling and condoms.

Counseling sessions were carried out in the participants' native language and lasted on average 30 minutes. Counselors also administered in-depth questionnaires related to social and behavioral factors including sexual practices at baseline and quarterly in the first year of follow-up, bi-annually in year 2, and annually in year 3, except for a module on HIV transmission knowledge and desire for children which was administered annually as of year 2. These interviews lasted on average 1 hour and were also carried out in the participants' own home and own language. Women's partners were not interviewed unless they were an HBAC index client. In this study, the partners' desire for children was reported through the women.

The team of home visitors consisted in 30 lay workers trained by HBAC investigators in basic counseling, clinical and adherence monitoring, and referral. The team of counselors consisted in 12 certified and experienced professional counselors trained by HBAC investigators in research methods, behavioral and counseling theory, and family planning in relation to HIV/AIDS. Each home visitor and counselor was assigned to a given household for the entire duration of the study in order to increase the trust of the participants and enhance the validity of the information collected throughout the study. The frequency of home visits and the counseling messages remained the same over the course of the study. The original HBAC study duration was 36 months (median cohort follow-up) but this cohort had taken part in two preceding home-based studies [23], [24], thus most counselors and home visitors had been with their assigned household for over a year at the time the HBAC study started.

All female HBAC participants 18–49 years old who initiated ART between March 2003 and May 2004 were included in the pregnancy sub-study. Pregnancies occurring during follow-up were identified through clinic and home visits and were confirmed by one or more of the following criteria: a positive human chorionic gonadotropin test, a positive obstetrical examination, the presence of fetal heart beat, a history of abortion, or the delivery of a baby.

Pregnant women were educated about giving single-dose nevirapine prophylaxis to their newborn babies for the prevention of MTCT of HIV and were counseled to exclusively breastfeed for 3 to 6 months followed by rapid weaning as per Uganda's national PMTCT guidelines at the time [25]. Pregnancy outcomes were recorded by field and clinical staff. Post-conception counseling did not provide or refer clients for pregnancy termination services as voluntary abortion is illegal in Uganda. However, women who pursued termination on their own were asked about their experiences and were treated for any complications.

All participants provided written consent to participate in the HBAC study in one of six local languages. The HBAC study and the pregnancy sub-study were approved by the Science and Ethics Committee of the Uganda Virus Research Institute, the Uganda National Council of Science and Technology, and the institutional review boards of the Centers for Disease Control and Prevention (CDC), Atlanta, Georgia, USA, and of the University of California, San Francisco, California, USA.

Women with a confirmed pregnancy and whose last menstrual period occurred at least one month after the date of ART initiation were considered to have become pregnant while on ART. Women who were pregnant at the time of starting ART were enrolled 1 month after the end of their pregnancy. Assessment of the last normal menstrual period date was based on self-report, assessment of fundus height, or back dating from actual date of delivery of a live baby. Women who discontinued ART for more than 3 continuous months were excluded from the analysis during the time of treatment discontinuation. Women who did not become pregnant were censored at the date of discharge or on June 30^th 2006. Women who became pregnant were censored at the estimated date of conception. Conception date was estimated at 15 days after the date of the last reported or estimated normal menstrual period. Based on these criteria, 711 of 733 women enrolled in the HBAC trial were eligible for this study (Figure 1). Three of these women were missing baseline data and thus were excluded from the baseline and multivariate analyses.

We compared the baseline characteristics of women who did or did not became pregnant using the Wilcoxon rank sum test for continuous variables and the Chi-squared test for categorical variables. We analyzed trends in pregnancy incidence and desire for children over time. We tested for non-constant incidence of pregnancy by using a test for the shape parameter from the Weibull distribution [26]. We modeled repeated measurements of all variables related to the women's desire for children over time using a generalized estimating equation extension (GEE) to the logistic regression model [27]. The p-values were derived from a chi-square test of a linear trend arising out of the GEE model. We examined the issue of linearity by performing a test for departure from linear trend which was not significant. We also described family planning practices and the unmet need for family planning services in the second year of follow-up.

Risk factors for pregnancy were examined using the Cox proportional hazards model [28]. We assessed the unadjusted effects of potential risk factors for the first incident pregnancy using the log rank test [29]. Subsequent pregnancies were not included in the bi- and multivariate analyses. Adjusted predictor effects were summarized using hazard ratios (HR) from the multivariable models. Most predictor variables were assessed at multiple time points during follow-up. We treated these variables as time-dependent covariates [30] unless otherwise specified. At a given time point, the value of the predictor variable was determined by the most recent prior measurement. All p-values are two-sided. All analyses were conducted in Stata version 9.0 (Stata Corporation, College Station, Texas, USA).

Women enrolment is summarized in Figure 1 and enrolled women's characteristics are described in Table 1. 711 women were enrolled and followed up for a median of 2.44 years (interquartile range (IQR): 2.32–2.62); of these, 11 (1.6%) were lost to follow-up, 5 (0.6%) discontinued ART for >3 months, and 36 (5.1%) died. Among 708 women with baseline data, median age was 37 years (IQR 31–41), 28% (194/702) of all women were married or co-habiting with their partner, 57% (398/702) were widowed, 24% (169/702) had no formal education, and 97% (661/684) had one or more living children (median 4 children (IQR: 1–5). Twenty-four percent (169/694) of women reported being sexually active at baseline, 2% (3/168) of whom reported having more than 1 current sexual partner, and 56% (95/169) reported having always used a condom with their sexual partner in the last 3 months. Forty-five percent (315/697) of the women reported ever having been physically abused by their partner (Table 1).

At baseline, women who became pregnant after starting ART were younger than their non-pregnant counterparts (median age 32 vs. 38 years; p<0.0001), more likely to be married and/or co-habiting with their partner (40% vs 25%; p = 0.002), to have been pregnant in the last 12 months (13% vs 3%; p<0.0001), and to have experienced the death of a child (74% vs 60%; p = 0.004), but less likely to have experienced the death of a spouse or partner (54% vs 66%; p = 0.014). Pregnancy was not associated with the women's allocation to the various ART monitoring arms of the HBAC trial (Table 1).

Desire for children increased significantly over follow-up for both the women (p<0.004) and their partners (p = 0.0001) but remained under 7% for the women and under 26% for their partners throughout follow-up. The proportion of women reporting sexual activity within 3 months of the interview also increased significantly over time (p = 0.0001), which was reflected by an increased frequency of sex in the past month or the past week (both p-values = 0.001).

A total of 120 (17%) women experienced 140 pregnancies after initiation of ART. Median time to first incident pregnancy was 12.4 months (IQR 7.9–18.0) after starting ART. Four of the140 incident pregnancies produced a set of twins and 38 (26%) did not result in a live birth; of these, 4 were spontaneous miscarriages, 8 were stillbirths including one of the twins, and 26 were induced abortions. Overall pregnancy incidence rate was 8.20 per 100 women-years (WY), increasing from 3.5 per 100 WY in the first quarter post-ART to 11.7 per 100 WY in the fifth quarter of follow-up, and ranging between 6.1 and 9.5 per 100 WY thereafter (p<0.0001) (Figure 2).

Among sexually active women, 65% (112/173) reported using condoms 18 months after starting ART and 63% (119/188) at 2 years of follow-up. However, only 14% (24/173) and 13% (25/188) reported using a semi-permanent (hormonal injection or implant) or permanent (tubal ligation) contraceptive method at 18 months, and 3.5% (6/173) and 4.3% (8/188) reported using dual contraception (a hormonal and barrier method in combination) after 24 months on ART. Among 176 sexually active women not desiring children at 2 years of follow-up, 33% (59) did not use any method, 53% (93) used condoms alone, 86% (152) were not using any semi-permanent or permanent method, and nearly 96% (168) did not use dual contraception (Table 2).

Factors associated with an increased risk of pregnancy included young age (HR = 3.16 per 10 year decrease, CI: 2.30–4.34, p<0.001), being married or co-habiting with a partner (HR = 2.72, CI: 1.88–3.95, p<0.001), having a body mass index (BMI)≥18.5 near the time of conception (HR = 2.5, CI: 1.02–5.88, p = 0.037), having a higher CD4 cell count near the time of conception (HR = 1.001 per 50 cells increase, CI: 1.000–1.002, p = 0.045), having experienced the loss of a child during follow-up (HR = 1.63, CI: 1.06–2.50, p = 0.025), and having used condoms inconsistently with one's current sexual partner in the last 3 months (HR = 1.82, CI: 1.05–3.23, p = 0.033). Factors associated with a decreased risk of pregnancy included having experienced the death of a spouse during follow-up (HR = 0.19, CI: 0.04–0.88, p = 0.034), and being in a longer relationship with one's current sexual partner (HR = 0.97 per year of relationship, CI: 0.94–0.99, p = 0.023), (Table 3).

In the adjusted model (Table 3), being younger (HR = 2.71 per 10-year decrease in age, CI: 1.95–3.78, p<0.001), having a BMI≥18.5 (HR = 1.09, CI: 1.01–1.18, p = 0.024) and having inconsistently used a condom with the most recent sexual partner in the last 3 months (HR = 1.79, CI: 1.02–3.13, p = 0.041) remained independently associated with pregnancy. Being married or cohabiting with a partner (HR = 1.59, CI: 0.94–2.69, p = 0.08) had a borderline association with pregnancy. However, time-updated CD4 counts (HR = 1.025; CI: 0.98–1.07, P = 0.26), having experienced the death of a child (HR = 1.94; CI: 0.64–5.95, p = 0.24) or a spouse (HR = 0.49, CI: 0.13–1.83, p = 0.29) or the length of the sexual relationship with the most recent partner (HR = 0.82; CI: 0.65–1.95, p = 0.31) were no longer independently associated with pregnancy.

This study had several limitations: 1) It was not randomized; 2) Spacing between interviews on sexual behavior varied during follow-up; 3) Information about contraceptive use was only collected as of 18 months of follow-up, thus it was not possible to assess contraceptive use as a determinant of pregnancy. However, the low rate of contraceptive use observed after 18 months of follow-up suggests that contraceptive use among these women was low throughout follow-up, as is the case for Ugandan women in general as mentioned above [40]. This possibility is supported by the results of in-depth interviews that were conducted after this study was completed with a subset of both women who did and did not become pregnant after starting ART [37]; 4) We had no baseline or pre-ART data on the fertility rate of the women or men in this cohort, nor was a comparison possible with a matched group of HIV-infected women not on ART. Therefore we could not assess whether the observed increase in pregnancy incidence over follow-up was due to an increase in either fertility or sexual activity among women on ART, or both. However, the older age of this cohort suggests that ART did have at least a partial effect on fertility, since pregnancy incidence increased even though women were older, had a median of 4 children, and thus were expected to have lower fertility rates and desire for children, independent of their sexual activity or HIV status. At any rate, the fact that this increase remained well below the fertility rate of Ugandan women of the same age may indicate a mere return to ‘normal’ levels of fertility rather than a real increase as compared to HIV-uninfected women or HIV-infected women not on ART; 5) Only 27% of all women had a cohabiting partner at the time of their enrolment into the HBAC program. This rate did not change substantially around the time of conception nor did it differ significantly according to the pregnancy outcomes of the women. Moreover, as partners living outside the women's household were not actively traced or tested by the program, the HIV status of the majority of partners was not known. As well, the type and content of discussions around fertility and pregnancy that may have taken place among couples were not assessed in this study. Thus the potential role of the nature of the couples' relationships, of the partners' HIV status and of the women's knowledge of it in predicting pregnancy could not be analyzed in this study. These limitations highlight the need for more in-depth and controlled research on fertility and reproductive intentions and outcomes among people on ART. Furthermore, they underline the importance of incorporating reproductive health measures into ART study designs, as well as ensuring that ART programs provide regular counseling for, and continued access to, reproductive health services including family planning.

The results of this study prompted initiation of an intervention to increase access to family planning services for both female and male HBAC study participants. As of July 2006, HBAC counselors and nurses were retrained in family planning service delivery for PLHA and started to pro-actively counsel all registered clients on family planning on a quarterly basis, to deliver hormonal contraceptives at home, and to actively refer and follow-up women opting for hormonal implants or tubal ligation to the hospital family planning clinic.

In conclusion, our results emphasize the need and importance of making family planning services an integral and continuous part of ART interventions in Africa. While more research is needed to further explore the factors that may place women on ART at increased risk of unwanted pregnancy, women on ART and their partners should be made clearly aware of the effect of ART on fertility. All clients on ART should be given full and free access to family planning services, including dual methods of contraception [47], [48]. Younger women and couples who do not desire children should be targeted in priority and strongly supported to use effective contraception to avoid unwanted pregnancies. Pro-active family planning counseling should be provided not only at the time of screening patients for ART eligibility but also and more importantly, on a regular basis thereafter, once patients resume their normal sexual activity.