Sedgh, Gilda ScD*; Larsen, Ulla PhD†; Spiegelman, Donna ScD*‡; Msamanga, Gernard DrPh§; Fawzi, Wafaie W DrPh*∥
Although many studies have indicated that HIV-infected women experience lower pregnancy and live birth rates than uninfected women,1-5 it seems that women at early clinical stages of infection may experience fertility rates comparable to those of uninfected women.5,6 In fact, there has been little research aimed directly at examining the association between HIV disease progression and fertility. Early follow-up studies in Zaire and the United States observed lower fertility among women with advanced disease compared with women with asymptomatic infection,7,8 but sample size and data limitations precluded control for potential confounders of this association. A small cross-sectional study in Cote d'Ivoire did not find a significant association between disease stage and fertility.9 More recently, a community-based study found an inverse association between disease progression and the prevalence of pregnancy in multivariate analysis. This study used a cross-sectional design and broadly categorized HIV-1-positive women as symptomatic or asymptomatic. Using a prospective cohort study design, none of these studies was able to explore the possible biologic mechanisms of the association between disease progression and fertility or the other factors that independently predict fertility among HIV-infected women. Insights into these associations may enhance our understanding of the natural history of the disease and fertility patterns among infected women.
We seek to shed light on the association between HIV-1 disease progression and fertility by analyzing the association between clinical stage of disease and the incidence of pregnancy and live birth. Multivariate analyses are used to control for other predictors of fertility to study these associations. We also seek to identify other determinants of pregnancy and live birth incidence in this population. These analyses are based on observations of HIV-infected attendees of antenatal clinic services in Dar es Salaam, Tanzania, using a prospective cohort study design.
The data for this cohort study were originally collected for a randomized trial aimed at studying the effects of vitamin supplementation on HIV-1 disease progression and vertical transmission.10 One thousand seventy-five eligible women were recruited in antenatal clinics between April 1995 and July 1997. Eligible HIV-1-positive women were pregnant, at 12 to 27 weeks of gestation, and residing in Dar es Salaam. These analyses are limited to women who were 15 to 39 years old at the time of recruitment.
At the start of the original study, all women were counseled on the risks of HIV transmission to uninfected partners through unprotected sexual transmission and the risk of vertical transmission during pregnancy and breast-feeding. All women were offered a referral to a family planning clinic for contraception. Follow-up data on women's visits to the family planning clinics were not available.
Investigators collected information on women's age, educational level, marital status, source of income, birth histories, middle upper arm circumference (MUAC), and their partners' educational level and income at baseline. Information on marital status was updated annually; information on whether women notified anyone regarding their HIV status was updated semiannually. Laboratory blood tests, including T-lymphocyte counts, were conducted at baseline and reassessed semiannually. Changes in weight, breast-feeding status, and child survival were updated monthly. The study protocol called for infants from the index pregnancy to be tested for HIV infection 6 weeks after birth and at 3-month intervals thereafter.
Women were scheduled for monthly visits with a physician and nurse from the study team; in December 1997, the protocol was revised to call for visits with a physician every third month. Physicians examined women for clinical signs of pregnancy and HIV progression. Nurses measured women's weight and administered a questionnaire asking about other signs and symptoms of illness experienced in the prior month. In the months when a woman was checked by a physician as well as by a nurse, we were able to reassess the woman's clinical stage of disease using a modified version of the World Health Organization's (WHO) recommended staging system (see Appendix).11
Beginning in June 1997, women also reported the date of their last menstrual period (LMP) during their visit with the nurse. This information was used to determine whether women experienced menstrual dysfunction, defined as menstrual cycles longer than 40 days in length. We did not include menstrual irregularities that occurred during lactation or during the use of hormonal contraception among incidents of menstrual dysfunction.
We measured the extent of weight loss, if any, that women experienced during the course of the study. Because follow-up began immediately on completion of a pregnancy, we used the first weight measure taken at least 6 months postpartum as the baseline weight and compared this with subsequent measurements to assess the extent of weight loss during follow-up.
Any changes in marital status and notification of others of participants HIV status were assumed to have occurred at the midpoint between the date at which the new status was reported and that of the most recent prior report.
Pregnancies were identified if (1) the physician recognized a pregnancy during a clinical examination and recorded fetal movement and heart sounds or (2) the woman reported the date of pregnancy completion and the outcome of the pregnancy (live birth, stillbirth, or fetal loss) to a member of the study team. Of the 309 identified pregnancies, 235 were recognized during a clinical examination and 74 were based on women's reports of the dates and outcomes of the pregnancies.
The date of conception was estimated using the woman's reported date of her LMP before conception. When this information was missing, the date of conception was estimated using information on the date of pregnancy completion and type of pregnancy outcome, assuming that all live births had a gestational period of 280 days, stillbirths lasted 200 days, and spontaneous and induced abortions ended after 80 days. In the absence of this information, physicians' assessments of gestational age during clinical examination were used to estimate date of conception. The dates of 211 conceptions were estimated using the reported date of the LMP, 92 were based on the date the pregnancy was completed and the outcome of the pregnancy, and 6 were based on clinical examinations.
We used women's reports of the date of pregnancy completion and the pregnancy outcome to identify live births during follow-up. Of the 309 pregnancies that were recorded, 250 resulted in a reported live birth before the end of follow-up.
We used a prospective cohort study design to examine the association of HIV disease stage with (1) the time to conception of the next clinically recognized pregnancy and (2) the incidence of the next live birth. In separate models, we measured disease progression using clinical stage of disease, CD4 cell count, weight loss, menstrual dysfunction, and MUAC, all measured as time-varying covariates. We used a Kaplan-Meier survival curve and log-rank test to compare the time to pregnancy by clinical stage of disease and Cox proportional hazards models to study multivariate associations of disease stage and other factors with pregnancy and live birth rates. Multivariate models included variables that were significantly associated with the outcome in univariate models. They also included variables representing the woman's educational level and her partner's education as controls for socioeconomic status. We used the likelihood ratio test to test for first-order interactions in our models.
To explore whether the association between clinical disease progression and lower pregnancy incidence was a result of the mediating effects of weight, nutritional status, and menstrual function, we examined whether the measures of effect of clinical stage were attenuated when these potential intermediate factors were added to the multivariate models.
The person-time at risk for pregnancy was assumed to have begun with the completion of the index pregnancy. We additionally constructed models in which the person-time at risk for pregnancy was considered to begin after completion of exclusive breast-feeding and a set of models in which person-time at risk began after completion of exclusive or partial breast-feeding.
Observations on a woman ended when she experienced the outcome of interest (pregnancy or live birth), when she died, when she was last assessed for clinical stage of disease, or at the end of follow-up, whichever came first. For the purpose of these analyses, follow-up ended on September 30, 2001. Observations with missing data were retained in the analyses using the missing indicator method.12
Of 1075 HIV-infected pregnant women originally randomized, 5 were excluded from analysis because they were more than 39 years old, 6 died, and 40 dropped out of the study before or immediately after delivery of the index pregnancy. The remaining 1024 women contributed 32,079 person-months of observation to the analyses of pregnancy rates and 35,415 person-months to the analyses of live birth rates between June 1995 and September 2001.
The study protocol was approved by the Research and Publications Committee of Muhimbili University College of Health Sciences, the Ethical Committee of the National AIDS Control Program of the Tanzanian Ministry of Health, and the Institutional Review Board of the Harvard School of Public Health.
A total of 1024 women were followed for an average of 31.3 months each. Information on HIV disease stage at the time of conception was available for 287 of the 309 pregnancies that were reported during the period of follow-up. Forty-two pregnancies occurred among women at clinical stage I, 148 occurred while women were in stage II, 95 occurred while women were in stage III, and 2 pregnancies occurred while women were in stage IV of the disease. Because of the small number of events in stage IV, we considered person-time spent in stages III and IV together in a single category in further analyses.
Table 1 displays the distribution of the women in the study according to their most recent clinical stage of disease and other health-related and sociodemographic characteristics. Women with advanced disease were older (P < 0.001), had more live births before the start of the study (P < 0.001), were more likely to experience a fetal loss or child death from the index pregnancy (P < 0.001), and were more likely to have revealed their HIV serostatus to a friend or relative during follow-up (P < 0.001). Disease progression was not associated with the educational level of the woman or her partner, her level of financial independence, or her marital status.
Figure 1 displays the Kaplan-Meier estimates of the time to pregnancy by the woman's clinical stage of disease. Women at clinical stage II experienced a longer time to pregnancy compared with women at stage I (P = 0.0009), and women at stages III and IV experienced the greatest delays in time to pregnancy compared with women at stage II (P < 0.0001) or stage I (P < 0.0001).
In univariate analyses, the pregnancy hazard ratio comparing women at clinical stage II with women at stage I was 0.53 (95% confidence interval [CI]: 0.37, 0.75; Table 2), and the hazard ratio for women at clinical stage III or IV was 0.19 (95% CI: 0.13, 0.28). In multivariate analyses controlling for independent predictors of fertility, including age, education, partner's level of education, source of income, number of prior live births, marital status, outcome of the index pregnancy, whether the woman had revealed her HIV status to others, and the randomization arm, the association between clinical stage of disease and fertility remained strong. Women at stage II of the disease experienced a relative pregnancy rate of 0.56 (95% CI: 0.39, 0.82) compared with women at clinical stage I, and women at stages III and IV experienced a relative pregnancy rate of 0.24 (95% CI: 0.16, 0.36) compared with women at clinical stage I.
The association between disease stage and pregnancy incidence varied between women who had lost the child from the index pregnancy and those who had not. Among women with a surviving child from the index pregnancy, the risk of another pregnancy declined to 0.61 (95% CI: 0.34, 1.08) at stage II and to 0.35 (95% CI: 0.20, 0.64) at stage III or IV compared with women at stage I (results not shown). Among women who had delivered a live child from the index pregnancy who subsequently died, the decline in pregnancy incidence with disease progression was more precipitous. The pregnancy risk ratio (RR) associated with clinical stage II was 0.26 (95% CI: 0.13, 0.50), and the RR at stage III or IV was 0.07 (95% CI: 0.03, 0.16) (P < 0.001, test for interaction). The association between disease stage and pregnancy risk did not vary by levels of the other covariates in the model and did not change over time after taking into account this interaction.
We considered other measures of disease progression aside from clinical stage and tested their association with pregnancy incidence in separate models. A CD4 cell count of 250 cells/mm3 or less was also significantly associated with a lower incidence of pregnancy in multivariate models, excluding clinical stage (RR = 0.54, 95% CI: 0.36, 0.81). Women whose body weight fell by more than 10% during follow-up experienced a relative pregnancy rate of 0.45 (95% CI: 0.22, 0.92), and women with MUAC measurements less than or equal to 23 cm were 29% less likely to get pregnant during follow-up than women with MUAC measurements of at least 26 cm (RR = 0.71, 95% CI: 0.51, 0.99). Oligomenorrheic and amenorrheic women experienced a relative pregnancy rate of 0.04 compared with women who did not experience these forms of menstrual dysfunction (95% CI: 0.01, 0.17).
Other predictors of the incidence of pregnancy in multivariate analyses included age, marital status during follow-up, outcome of the index pregnancy, and whether the woman had shared her HIV status with anyone. The pregnancy RR was lowest among women aged 35 to 39 years at the start of follow-up compared with women who were 15 to 19 years old (RR = 0.10, 95% CI: 0.02, 0.43). Women who were married during the period of observation were 54% more likely to get pregnant than those who were single, divorced, widowed or in a nonmarital union (95% CI: 1.08, 2.20), and women who had told someone they were HIV-1-positive experienced a relative pregnancy rate of 0.73 (95% CI: 0.56, 0.96). The outcome of the index pregnancy was a strong predictor of pregnancy incidence. Women who experienced a spontaneous abortion or stillbirth during the index pregnancy were nearly 4 times more likely to get pregnant than women who had a living child from the index pregnancy (RR = 3.9, 95% CI: 2.6, 5.8), and child death was associated with a nearly 3-fold increase in the incidence of another pregnancy (RR = 2.8, 95% CI: 2.1, 3.6).
To explore whether the association between clinical disease progression and lower pregnancy incidence was a result of the mediating effects of weight, nutritional status, and menstrual function, we entered these variables in a model with clinical stage and the determinants of pregnancy listed in the primary model in Table 2. Clinical stage remained significantly and strongly associated with pregnancy incidence. Women at clinical stage II experienced a relative pregnancy rate of 0.57 (95% CI: 0.39, 0.84; P = 0.0004), and women at stages III and IV experienced a relative pregnancy rate of 0.26 (95% CI: 0.17, 0.40; P < 0.0001) (results not shown).
Information on HIV disease stage at the time of birth was available for 224 of the 250 live births that were reported during the period of follow-up. Twenty-three births occurred among women at clinical stage I, 115 occurred while women were in stage II, and 86 births were to women in stage III or IV.
In univariate analyses, the relative rate of a live birth comparing women at clinical stage II with women at stage I was 0.58 (95% CI: 0.37, 0.91; Table 3), and the RR for women at clinical stage III or IV was 0.22 (95% CI: 0.14, 0.35). After controlling for other independent predictors of fertility, the relative birth rate for women at stage II was 0.64 (95% CI: 0.39, 1.03), and women at stages III and IV experienced a RR of 0.26 (95% CI: 0.16, 0.43) compared with women at clinical stage I. A CD4 cell count of 250 cells/mm3 or less was also significantly associated with a lower incidence of birth in multivariate models, excluding clinical stage (RR = 0.36, 95% CI: 0.23, 0.56).
As noted earlier, we additionally fit models in which the person-time at risk for a pregnancy was considered to begin after completion of exclusive breast-feeding or after completion of exclusive or partial breast-feeding. The results were not qualitatively affected by these alternate assumptions about the period of exposure to a pregnancy or live birth.
Our prospective observations of a cohort of HIV-infected women in Dar es Salaam, Tanzania, indicate that clinical progression to WHO stages II through IV is independently associated with a large decline in the incidence of pregnancy and live births, even after controlling for independent predictors of fertility. This pattern was especially marked in women who had lost the child they had delivered at their most recent prior pregnancy. The association between disease progression and a lower pregnancy incidence persisted after additionally controlling for physiologic factors that may represent mechanisms of the association between disease stage and fertility, namely, weight loss, nutritional status, and menstrual dysfunction.
When CD4 cell count, MUAC, and weight loss were each used in lieu of clinical stage as measures of disease progression, only women with the lowest CD4 cell counts and poorest anthropometric measures demonstrated compromised fertility rates.
Estimates of disease stage are based on clinical signs of progression and may therefore be imperfect measures of disease stage. Validation studies in similar settings found that clinical categories closely approximating the WHO staging system corresponded well with laboratory markers of disease stage and survival of HIV-infected subjects, however.5,13
Our measure of pregnancy incidence is vulnerable to misclassification in that we only have information on clinically recognizable pregnancies. If the incidence of preclinical pregnancy loss was greater among women at advanced stages of disease, our effect estimates might be biased away from a null association. The incidence of clinically recognized pregnancies might, however, be considered a distinct outcome from the incidence of all pregnancies.
Our findings of a decline in pregnancy and live birth rates with advancing disease are consistent with findings from other studies that have observed declines in fertility with progression of disease in HIV-infected women.3,7,14
We explored various biologic components of disease progression and their possible roles in the association between advancing disease and declines in fertility. Weight loss is a defining characteristic of disease progression, and we observed an independent association between extreme weight loss and lower pregnancy rates in this cohort. The association of weight loss with subfertility has also been observed elsewhere.15-17 Although excessive weight loss may play a role in declining fecundity with disease progression, it did not explain the declines in the incidence of pregnancy that we observed.
Investigations by others into whether HIV infection is associated with menstrual dysfunction have yielded inconsistent results.18-22 We found that the incidence of long menstrual cycles did not vary substantially between women at clinical stages I and II of HIV infection but did increase with progression to clinical stages III and IV. This factor did not, however, account for the association between disease progression and declines in fertility.
Our estimates of weight loss and the incidence of menstrual dysfunction are susceptible to some degree of misclassification. It is possible that these variables would have demonstrated a role in explaining the associations of interest if we had more complete and accurate information on their values in this population.
Use of reliable contraceptive methods was low in our cohort, and we found that levels of use did not vary significantly according to a woman's stage of disease. Nevertheless, we cannot rule out the possibility that increases in contraceptive use with advancing disease were not fully measured in the study and might explain, at least in part, the observed differentials in pregnancy risk by disease stage.
Factors that may be responsible for an association between disease progression and fertility decline but that we were unable to study in this population include preclinical pregnancy loss, changes in sperm motility and morphology in an infected partner, and changes in behavior (including a decrease in sexual activity with disease progression).
We observed that social factors seem to play an independent role in the incidence of pregnancy in this group of HIV-infected women as a whole. Women who were married, had recently experienced a fetal or child loss, or did not share their HIV status with anyone were most likely to experience a clinically recognizable pregnancy during the period after HIV diagnosis.
Additional studies could help to shed light on the behavioral and biologic dimensions of progression that are responsible for declining fertility during the course of disease progression in HIV-1-positive women. Research of this nature could improve our understanding of the natural history of HIV-1 infection and could inform estimates of the effect of the HIV-1 epidemic on fertility under changing conditions.
The authors are grateful to the mothers who contributed their time and efforts as subjects of this study. Thanks also go to the field teams who implemented the study and to Ellen Hertzmark, who played a substantial role in data management.
1. Barbieri M, LeCoeur S, Prasitwattanasere S, et al. HIV and fertility in Thailand. Presented at: Population Association of America 2000 Annual Meeting; 2000; Los Angeles.
2. Desgrees du LA, Msellati P, Ramon R, et al. HIV-1 infection and reproductive history: a retrospective study among pregnant women, Abidjan, Cote d'Ivoire, 1995-1996. Ditrame Project. Int J STD AIDS
3. Gray RH, Wawer MJ, Serwadda D, et al. Population-based study of fertility in women with HIV-1 infection in Uganda. Lancet
4. Carpenter LM, Nakiyingi JS, Ruberantwari A, et al. Estimates of the impact of HIV infection on fertility in a rural Ugandan population cohort. Health Transit Rev
5. Ryder RW, Batter VL, Nsuami M, et al. Fertility in 238 HIV-1-seropositive women in Zaire followed for 3 years post-partum. AIDS
6. Sedgh G, Larsen U, Spiegelman D, et al. HIV and fertility among ante-natal clinic attendees in Dar es Salaam, Tanzania. Presented at: Population Association of America Annual Meeting; 2000; Los Angeles.
7. Chu SY, Hanson DL, Jones JL. Pregnancy rates among women infected with human immunodeficiency virus. Adult/Adolescent HIV Spectrum of Disease Project Group. Obstet Gynecol
8. Desgreesdu LA, Msellati P, Yao A, et al. Impaired fertility in HIV-1-infected pregnant women: a clinic-based survey in Abidjan, Cote d'Ivoire, 1997. AIDS
9. Fawzi WW, Msamanga GI, Spiegelman D, et al. Rationale and design of the Tanzania Vitamin and HIV Infection Trial. Control Clin Trials
10. Fawzi WW. Proposed World Health Organization staging system for HIV infection and disease: preliminary testing by an international collaborative cross-sectional study. The WHO International Collaborating Group for the Study of the WHO Staging System. AIDS
11. Wilcoxon F. Individual comparisons by ranking methods. Biometrics
12. Kassa E, Rinke de Wit TF, Hailu E, et al. Evaluation of the World Health Organization staging system for HIV infection and disease in Ethiopia: association between clinical stages and laboratory markers. AIDS
13. Malamba SS, Morgan D, Clayton T, et al. The prognostic value of the World Health Organisation staging system for HIV infection and disease in rural Uganda. AIDS
14. Falsetti L, Pasinetti E, Mazzani MD, et al. Weight loss and menstrual cycle: clinical and endocrinological evaluation. Gynecol Endocrinol
15. Reid RL, Van Vugt DA. Weight-related changes in reproductive function. Fertil Steril
16. Green BB, Weiss NS, Daling JR. Risk of ovulatory infertility in relation to body weight. Fertil Steril
17. Grodstein F, Goldman MB, Cramer DW. Body mass index and ovulatory infertility. Epidemiology
18. Minkoff HL, Eisenberger-Matityahu D, Feldman J, et al. Prevalence and incidence of gynecologic disorders among women infected with human immunodeficiency virus. Am J Obstet Gynecol
19. Chirgwin KD, Feldman J, Muneyyirci-Delale O, et al. Menstrual function in human immunodeficiency virus-infected women without acquired immunodeficiency syndrome. J Acquir Immune Defic Syndr Hum Retrovirol
20. Harlow SD, Schuman P, Cohen M, et al. Effect of HIV infection on menstrual cycle length. J Acquir Immune Defic Syndr
21. Shah PN, Smith JR, Wells C, et al. Menstrual symptoms in women infected by the human immunodeficiency virus. Obstet Gynecol
22. Ellerbrock TV, Wright TC, Bush TJ, et al. Characteristics of menstruation in women infected with human immunodeficiency virus. Obstet Gynecol
Criteria used to assign the clinical stage of disease include the following:
Stage I: asymptomatic infection
Stage II: (1) weight loss of 3% to 10%, (2) minor mucocutaneous manifestations, (3) prior or current herpes zoster infection, or (4) acute upper respiratory tract infection in the past month
Stage III: (1) weight loss greater than 10%, (2) chronic diarrhea for at least 4 weeks, (3) prolonged fever for at least 4 weeks, (4) oral candidiasis, (5) oral hairy leukoplakia, (6) pulmonary tuberculosis, (7) severe bacterial infections, or (8) difficulty in swallowing for at least 1 week
Stage IV: (1) weight loss greater than 10% with (a) prolonged fever and weakness for more than 4 weeks or (b) chronic diarrhea for more than 4 weeks, (2) herpes simplex infection; (3) extrapulmonary tuberculosis, or (4) Kaposi sarcoma
© 2005 Lippincott Williams & Wilkins, Inc.