Women of childbearing age constitute the single largest group of people living with HIV in sub-Saharan Africa.1 In South Africa and elsewhere, pregnancy is a common indication for the initiation of highly active antiretroviral therapy (HAART), for the prevention of mother to child transmission2,3; among pregnant women, HIV prevalence has been estimated as high as 40% in some settings.4,5
There are many reasons to hypothesize that response to HAART may be affected by pregnancy. Changes in body mass and blood volume may lead to drug underdosing. Pregnancy may lead to changes in metabolism of nevirapine and lopinavir.6–9 Rising levels of beta-estradiol during pregnancy may attenuate the effect of stavudine.10,11 Finally, social and personal issues including stigma and intimate partner violence (negative) and desire to protect an unborn child (positive) may affect adherence.12
Although a number of studies have examined prevention of mother to child transmission of HIV and subsequent response to HAART,13,14 issues of fertility during HAART,15 and more generally the effects of pregnancy on HIV disease progression in the pre-HAART era,16 there are few published reports examining the impact of pregnancy on maternal response to HAART in sub-Saharan Africa.12,17,18 Here, we examine pregnancy at time of HAART initiation and biological sex as predictors of virologic outcomes of HAART.
Study Population and Design
We performed an observational cohort study in the database of the Themba Lethu Clinic (TLC).3 The TLC Cohort comprises adults initiating HAART in Johannesburg, South Africa. TLC is located within Helen Joseph Hospital in urban Johannesburg and is the largest single clinic providing HAART in South Africa.
Patients starting HAART between April 1, 2004, and September 30, 2009, were selected for study inclusion; end of follow-up was set on March 31, 2010, due to last visit, transfer of care, administrative reasons, dropout, or death. Only antiretroviral therapy–naive patients were selected.19 Subjects were also limited to ages 18–45 because the database included only a single woman older than 45 who initiated HAART although pregnant.18
Typical first-line HAART in South Africa included stavudine, lamivudine, and efavirenz; due to concerns about teratogenicity, women pregnant at baseline are typically placed on the boosted protease inhibitor Kaletra (lopinavir and ritonavir) rather than efavirenz. Nonpregnant women with declared pregnancy intention are typically placed on nevirapine rather than efavirenz due to concerns about teratogenicity of efavirenz.20 Other aspects of the TLC clinical database have been described previously.3,18,21
Definitions and Data
The main exposure in this study was first biological sex (male or female), and then among females, “prevalent” pregnancy: that is, pregnancy present at baseline, the time of HAART initiation. This is distinguished from “incident” pregnancy which occurs subsequent to HAART initiation for a woman's health15,18; prevalent pregnancy is in general the cause for a recognition of an HIV infection requiring treatment, especially for initiation of HAART. Incident pregnancies were ignored in this analysis, but have been addressed elsewhere.18 For this analysis, we conducted additional review of the records of women who were on Kaletra at baseline but were not marked as pregnant in the database.
The main outcome in this study was time to virologic failure defined as either a failure to achieve virologic suppression of plasma HIV-1 RNA to less than 400 copies per milliliter within 6 months of HAART initiation or a viral rebound to above 400 copies at any time after initial suppression.18,22 A second outcome was virologic failure by 6 months (a dichotomous outcome). When possible, confirmation of outcome by a second viral load test within 30 days was obtained, but we included failures from patients missing a confirmatory sample.
We used simple descriptive statistics for baseline characteristics of individuals. We examined time to virologic failure using Kaplan-Meier–type cumulative incidence curves. We estimated relative risk of failure to suppress by 6 months using log-binomial (risk) regression (and logistic regression where noted). Last, we used accelerated failure time models to estimate relative time to virologic failure.
In multivariable analyses, we considered the following confounders based on previous literature and plausible biological mechanism: calendar date of HAART initiation, age, ethnicity, employment status, history of smoking, tuberculosis, WHO stage, weight, body mass index, hemoglobin, CD4 count and CD4 percent, and whether there was any charge for being seen in clinic (discontinued as of October 2006). We used restricted cubic splines to flexibly control for age, body mass index, weight, CD4 count, and time-on-study.
We did not control for baseline viral load because it is collected in less than 25% of participants; instead, we performed a sensitivity analysis restricted to women who had suppressed virus by 6 months of follow-up to eliminate impacts of baseline viral load. We likewise did not control for baseline or time-updated drug regimen in main analysis (but did so in sensitivity analysis) because drug regimen is determined chiefly by pregnancy status (and pregnancy intentions), and is thus may be part of an effect of pregnancy on the outcome.
The study population comprised 9173 men and women at time of HAART initiation, of which approximately two-thirds (n = 5,997) were women. Of the 5997 women, 587 were pregnant at baseline. These 9173 individuals were followed up for a median of 18 (interquartile range: 8–37) months until virologic failure, death, loss-to-follow-up, transfer, or administrative end of follow-up; maximum follow-up time was 72 months. At baseline HAART initiation, about 14% of pregnant women were in their first trimester, 43% were in their second trimester, and 43% were in their last trimester.
Baseline characteristics of all subjects are described in Table 1. In general, men and nonpregnant women were similar at baseline, although men generally had indicators of more advanced disease status. For example, comparing men to nonpregnant women, men had lower median body mass index (BMI; 20.1 vs. 22.2 kg/m2), more active tuberculosis (22.1% vs. 17.7%), and were more likely to have CD4 count ≤ 50 cells/mm3 (40.5% vs. 32.9%). However, as expected, pregnant women were substantially healthier than either men or non-pregnant women. For example, median CD4 cell count at baseline was 74 cells/mm3 among men, 92 cells/mm3 among non-pregnant women, and 156 cells/mm3 among pregnant women. And likewise as expected, only 1.6% of pregnant women had a BMI < 18.5 kg/m2, compared to 27.1% and 17.9% of men and non-pregnant women, respectively.
During all of follow-up, a total of 472 men (14.9%), 822 (15.2%) non-pregnant women, and 70 (11.9%) prevalent pregnant women experienced the outcome of failure to suppress or subsequent virologic failure. The clear majority of these failures (65%) were unconfirmed within 30 days. Over follow-up, 408 (7.5%) nonpregnant women, 297 (9.3%) men, and 17 (2.9%) pregnant women died; whereas 1298 (24.0%) nonpregnant women, 919 (28.9%) men, and 234 (39.9%) pregnant women become lost to follow-up.
The cumulative incidence of virologic failures in nonpregnant women, pregnant women, and men are shown in Figure 1. There was no difference in hazard of failure comparing nonpregnant women to men [crude hazard ratio (HR): = 0.95, 95% confidence limit (CL): 0.85 to 1.07; adjusted HR: = 0.98, 95% CL: 0.84 to 1.14]; accordingly, men were excluded from further analysis.
Results comparing pregnant women to nonpregnant women are reported in Table 2. The crude HR for the outcome of virologic failure comparing only pregnant women to nonpregnant women over all follow-up was 0.73 (95% CL: 0.57 to 0.93), and the adjusted was similar at 0.69 (95% CL: 0.50 to 0.95). The time ratios comparing relative time to virologic failure among pregnant and nonpregnant women were calculated in crude accelerated failure time models under gamma, Weibull, and exponential distributions. All distributions yielded similar inferences; the crude time ratio under an exponential distribution was 1.44 (95% CL: 1.13 to 1.84), and an adjusted time ratio was similar but less precise.
Considering only the first 6 months, 251 (4.6%) of nonpregnant women and only 18 (3.1%) of pregnant women had experienced virologic failure. Again among women only, the crude risk ratio (RR) for the effect of prevalent pregnancy on risk of failure to suppress virus by 6 months was 0.66 (95% CL: 0.41 to 1.06); the adjusted RR was 0.66 (95% CL: 0.35 to 1.22); the adjusted HR was similar.
We performed several sensitivity analyses, all adjusted for confounding and summarized in Table 2. First, we limited analyses to women who had not failed HAART by 6 months of follow-up and examined time to virologic failure from 6 months onward as in the main analysis. We found results very similar to the main “all follow-up” analysis (adjusted HR: = 0.69, 95% CL: 0.47 to 1.02). Second, we analyzed only confirmed virologic failures (295 of 822 in nonpregnant women; 19 of 70 in pregnant women) and found a stronger though less precise effect (HR = 0.54, 95% CL: 0.30 to 0.97) than in main analysis. Third, we analyzed by trimester of pregnancy at HAART initiation; this analysis yielded similar though less precise results overall, with no clear temporal trends. Fourth and fifth, we controlled for whether initial drug regimen contained nevirapine or Kaletra versus efavirenz. In the 6-month analysis (analysis 4), we found RR = 0.41 (95% CL: 0.16 to 1.00) (estimated using logistic regression); in the all-follow-up analysis (analysis 5), we found HR = 0.50 (95% CL: 0.30 to 0.85). Sixth, we controlled for time-updated adherence, to see if this might be a pathway by which baseline pregnancy affected virologic failure; results suggested that it was not (HR = 0.70, 95% CL: 0.50 to 0.97), although residual confounding of the mediator-outcome relationship may be present.23 Seventh, we examined time to virologic failure by baseline CD4 count. In women with baseline CD4 >200 cells per cubic millimeter, the association of pregnancy with virologic failure was stronger (HR = 0.30, 95% CL: 0.10 to 0.87); in women with baseline CD4 ≤200 cells per cubic millimeter, the effect was slightly weaker (HR = 0.76, 95% CL: 0.54 to 1.07), consistent with the overall findings. Last, we analyzed only among women who were indicated to start HAART for their own health as follows: those with CD4 <200 cells per cubic millimeter, or in WHO stage IV, finding HR = 0.73 (95% CL: 0.51 to 1.03) in this group.
The WHO has stated that women's health should be “the overarching priority in decisions about antiretroviral treatment during pregnancy”.24 But despite this, relatively little remains known about how or whether pregnancy predicts response to HAART. Here we found that, among women who remain alive and in care in our cohort, pregnancy is associated with lower risks and hazards of virologic failure over follow-up. Notably, time to virologic failure was 44% longer among (prevalent) pregnant women than among nonpregnant women. These results were generally confirmed in sensitivity analyses addressing likely alternative scenarios (Table 2).
One possible reason for this observed effect might be dynamic observation, or diagnostic, bias. In particular, pregnant women might have fewer viral loads, or have more time between viral loads than nonpregnant women; for example, if antenatal care visits during pregnancy caused visits to our clinic to be scheduled further apart. If this were the case, then the observed longer time to failure might be the result of less detection of failure (rather than less failure, per se). In the first 6 months (the period in which these women were pregnant), the HR comparing timing of new viral loads between pregnant and nonpregnant women was 0.84 (95% CL: 0.75 to 0.95); and after the first 6 months, the HR was essentially null at 0.97 (95% CL: 0.91 to 1.04). Because the association of pregnancy on virologic failure after 6 months (first sensitivity analysis) was identical to the main analysis, diagnostic bias seems unlikely to explain away these findings.
Antiretroviral therapy Guidelines in South Africa during the study period20 stated that pregnant women should initiate HAART at a CD4 count <200 cells per cubic millimeter or a WHO stage IV. More recent guidelines25 recommend starting pregnant women at CD4 ≤350 cells per cubic millimeter. However, in this study, 24% of pregnant women who initiated HAART had both CD4 >200 cells per cubic millimeter and WHO stage of III or lower, suggesting that clinician preferences played some role in determining HAART initiation. Sensitivity analysis which found somewhat different associations of pregnancy with time to virologic failure by baseline CD4 cell count may speak to such clinician preferences; but in neither baseline CD4 cell count stratum did the point estimate indicate an increased risk of virologic failure among pregnant women.
Rates of lost to follow-up were high among these subjects, and indeed were substantially higher among pregnant than nonpregnant women, similar to what was seen in a previous study from South Africa.17 Among nonpregnant women and men, many of these losses are likely to be deaths,19 but among pregnant women, “lost to follow-up” often does not mean “lost from care” (such as from death or leading rapidly to death as antiretroviral therapy is abandoned), but instead transfer to perinatal care centers that were not tracked. The observed relation of trimester of pregnancy to hazard of lost to follow-up is consistent with this, where the earlier in pregnancy women enter care, the more likely they are to be lost to follow-up. Compared with nonpregnant women, the HR for becoming lost is 2.3 (95% CL: 1.7 to 3.1) among women who initiate HAART in their first trimester, 1.9 (95% CL: 1.4 to 2.6) among women who initiate in their second trimester, and 1.2 (95% CL: 0.6 to 2.1) among women who initiate in their third trimester. Thus, comparing rates of loss to follow-up between these 2 populations may be misleading.
Although associated with lower risks of virologic failure, it is likely not the case that pregnancy (in this case, prevalent pregnancy) causes lower rates of failure. Indeed, previous work in this cohort suggests that incident pregnancy increases the risk of virologic failure.18 In this study, many women who initiate HAART although pregnant are initiating because of pregnancy, whereas those who initiate although nonpregnant are, universally, initiating because they are sick. As importantly, pregnant women with low CD4 counts or who are in WHO stage IV were nonetheless healthy enough to become pregnant in the first place. Thus, even exposure groups comparable in measured health status may remain fundamentally noncomparable due to indication, arguing against a causal interpretation of pregnancy in this report. Instead, we argue that current standard of care for women initiating HAART at baseline seems to be adequate in this setting, not associated with increased risks of virologic failure.
Notably, there was little confounding apparent in this analysis as follows: crude and adjusted models were similar, despite control for an extensive list of confounders. This may reflect that there are few factors which affect risk of virologic failure beyond adherence to drugs and drug regimen itself and that neither factor could be a confounder of the prevalent pregnancy–virologic response relationship because both occur after the exposure begins (in this analysis, women were pregnant before initiating a particular HAART regimen and before their adherence can be measured). Nonetheless, we explored whether the effect of pregnancy on virologic response changed although controlling for these postpregnancy factors as a kind of mediation analysis, albeit one whose causal interpretability may be limited (as discussed above). Controlling for adherence did not alter the main effect estimates; the model controlling for drug regimen suggested that increased use of Kaletra or nevirapine during pregnancy (rather than efavirenz) may lead to relatively increased risks of virologic failure among pregnant women. Thus, both models failed to explain the overall reduced risks observed in main analysis; determining why prevalent pregnancy is associated with lower rates of virologic failure should be a priority in future investigations.
In conclusion, pregnancy at time of HAART initiation was not associated with increased hazards or risks of virologic failure over follow-up, leading us to conclude that current clinical management of women who initiate HAART during pregnancy is likely to be adequate. However, there is preliminary evidence that choice of drug regimen associated with pregnancy may be leading to somewhat elevated risks of virologic failure.
The authors thank the patients of the Themba Lethu Clinic.
2. Black V, Hoffman RM, Sugar CA, et al.. Safety and efficacy of initiating highly active antiretroviral therapy in an integrated antenatal and HIV clinic in Johannesburg, South Africa. J Acquir Immune Defic Syndr. 2008;49:276–281.
3. Sanne IM, Westreich D, Macphail AP, et al.. Long term outcomes of antiretroviral therapy in a large HIV/AIDS care clinic in urban South Africa: a prospective cohort study. J Int AIDS Soc. 2009;12:38.
4. Pettifor AE, Rees HV, Kleinschmidt I, et al.. Young people's sexual health in South Africa: HIV prevalence and sexual behaviors from a nationally representative household survey. AIDS. 2005;19:1525–1534.
5. UNAIDS/WHO. AIDS Epidemic Update 2007. Geneva, Switzerland: WHO; 2007.
6. Mirochnick M, Capparelli E. Pharmacokinetics of antiretrovirals in pregnant women. Clin Pharmacokinet. 2004;43:1071–1087.
7. Aweeka F, Tierney C, Stek A, et al.. ACTG 5153s: pharmacokinetic exposure and virological response in HIV-1-infected pregnant women treated with PI. Presented at: 14th Conference on Retroviruses and Opportunistic Infections; February 28, 2007; Los Angeles, CA. Abstract 739S.
8. Floridia M, Giuliano M, Palmisano L, et al.. Gender differences in the treatment of HIV infection. Pharmacol Res. 2008;58:173–182.
9. Roustit M, Jlaiel M, Leclercq P, et al.. Pharmacokinetics and therapeutic drug monitoring of antiretrovirals in pregnant women. Br J Clin Pharmacol. 2008;66:179–195.
10. Prins M, Meyer L, Hessol NA. Sex and the course of HIV infection in the pre- and highly active antiretroviral therapy eras. AIDS. 2005;19:357–370.
11. Zhang M, Huang Q, Huang Y, et al.. Beta-estradiol attenuates the anti-HIV-1 efficacy of Stavudine (D4T) in primary PBL. Retrovirology. 2008;5:82.
12. MacCarthy S, Laher F, Nduna M, et al.. Responding to her question: a review of the influence of pregnancy on HIV disease progression in the context of expanded access to HAART in Sub-Saharan Africa. AIDS Behav. 2009;(suppl 1):66–71.
13. Guay LA, Musoke P, Fleming T, et al.. Intrapartum and neonatal single-dose nevirapine compared with zidovudine for prevention of mother-to-child transmission of HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet. 1999;354:795–802.
14. Chi BH, Sinkala M, Stringer EM, et al.. Early clinical and immune response to NNRTI-based antiretroviral therapy among women with prior exposure to single-dose nevirapine. AIDS. 2007;21:957–964.
15. Myer L, Carter RJ, Katyal M, et al.. Impact of antiretroviral therapy on incidence of pregnancy among HIV-infected women in Sub-Saharan Africa: a cohort study. PLoS Med. 2010;7:e1000229.
16. French R, Brocklehurst P. The effect of pregnancy on survival in women infected with HIV: a systematic review of the literature and meta-analysis. Br J Obstet Gynaecol. 1998;105:827–835.
17. Kaplan R, Orrell C, Zwane E, et al.. Loss to follow-up and mortality among pregnant women referred to a community clinic for antiretroviral treatment. AIDS. 2008;22:1679–1681.
18. Westreich D, Cole S, Nagar S, et al.. Pregnancy and virologic response to antiretroviral therapy in South Africa. PLoS One. 2011;6:e22778.
19. Fox MP, Brennan A, Maskew M, et al.. Using vital registration data to update mortality among patients lost to follow-up from ART programmes: evidence from the Themba Lethu Clinic, South Africa. Trop Med Int Health. 2010;15:405–413.
21. Westreich D, MacPhail P, Van Rie A, et al.. Effect of pulmonary tuberculosis on mortality in patients receiving HAART. AIDS. 2009;23:707–715.
22. Riddler SA, Haubrich R, DiRienzo AG, et al.. Class-sparing regimens for initial treatment of HIV-1 infection. N Engl J Med. 2008;358:2095–2106.
23. Cole SR, Hernan MA. Fallibility in estimating direct effects. Int J Epidemiol. 2002;31:163–165.
24. WHO. Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in Infants. Recommendations for a Public Health Approach. Geneva, Switzerland: WHO; 2010.
© 2012 Lippincott Williams & Wilkins, Inc.