Time to HAART Initiation or Death
The cohort was stratified by CD4 count at delivery into the following groups: <200, 200–349, 350–499, and ≥500 cells/μL. The endpoint was defined as HAART initiation or death (Fig. 2). Surviving women were censored at study close or if they became pregnant again.
Modeling of Time to CD4 ≤350 Cells/μL
Women were stratified by CD4 at delivery into 2 groups: 350–499 and ≥500 cells/μL, and projected time to a decline in CD4 count to 350 cells/μL was calculated for each group. The threshold of 350 cells/μL was chosen because of its current policy relevance. In 2010, GHESKIO increased its threshold for HAART initiation to this level from the previous one of 200 cells/μL. In the 93 multiparous women in the cohort, projections of decline in CD4 count were restricted to decline between the first delivery and 40 weeks before the second delivery, because of uncertainty regarding the effects of HAART or other ART prophylaxis on CD4 count and disease progression during pregnancy.
Time to Subsequent Pregnancy
Birth rates were calculated for years of follow-up without HAART and years after initiation of HAART. They do not include the delivery at GHESKIO, which prompted enrollment into the PMTCT program. We were not able to make a meaningful comparison between the treated and untreated women because of the great variability in time between delivery and HAART initiation. These rates were then used to estimate the average time between pregnancies while off HAART, to estimate the interval between courses of PMTCT prophylaxis in multiparous women.
All incidence rates were calculated using a normal approximation to a Poisson distribution. P values for comparison of all averages were calculated using a 2-tailed t test assuming equal variance.
For the analysis of time to HAART initiation or death, a Kaplan–Meier survival analysis was used to estimate median months from the first delivery at GHESKIO to the combined endpoint of HAART initiation or death. A log-rank test was used to determine the statistical significance of the difference among the 4 groups.
A linear mixed model was used to study the change in CD4 counts in untreated women after the first delivery at GHESKIO. To more closely approximate a Gaussian distribution, a log transformation was applied to the longitudinal record of CD4 counts. A linear trend over time was fit for each stratum. Based on the estimated slope coefficients of the trend in each stratum, an estimate of the number of months it would take to reach a CD4 count of 350 cells/μL was calculated along with approximate 95% confidence intervals (CIs).
These women had previously signed a written agreement to participate in the provision of services at GHESKIO and to allow the physicians to review their records. The GHESKIO Institutional Review Board had previously determined review of these records for study purposes to fall within provision of services. The previous study of the same cohort was approved by the GHESKIO and Weill Cornell Medical College Institutional Review Boards. The need for informed consent was further evaluated by the Dartmouth College Committee for the Protection of Human Subjects and waived according their criteria.7
Five hundred eight pregnant HIV-infected women were enrolled in the program of GHESKIO to limit MTCT between January 1, 1999 and December 31, 2005, at a median age of enrollment of 27 years and interquartile range (IQR) of 23–32. Of the 508 women, 18 were never seen postpartum and were excluded from the analysis. An additional 17 had reached a CD4 count of <200 before the availability of HAART and were therefore excluded for the reasons detailed in the Methods section. Four hundred seventy-three women therefore met criteria to enter into this analysis (Fig. 1 and Table 1).
Of the 473 women who entered into analysis, 160 were put on HAART, with a total of 515 person-years on HAART. Forty-eight of these women were continued on HAART from the time of their delivery for their own health, with the additional 112 put on HAART during their follow-up period.
There were 313 women seen after delivery who had not started on HAART at study close, with a total of 1068 years of follow-up.
Overall mortality in the cohort was 2.8 deaths per 100 person-years of follow-up (95% CI: 2.1 to 3.6). In the women on HAART from delivery, there were 4.0 deaths per 100 person-years (95% CI: 0.5 to 7.6) in the first 3 years after delivery. In those who never received HAART, mortality was 4.9 deaths per 100 person-years (95% CI: 3.2 to 6.5) during this period.
Of the 20 women who died after initiating HAART, 18 had a CD4 count recorded at HAART initiation. All but one started HAART at a CD4 count below 350 cells/μL, with a median CD4 at HAART initiation of 113 cells/μL (IQR: 54–201). By contrast, the median CD4 at HAART initiation in those who remained in active follow-up as of July 17, 2009 was 180 (IQR: 129–229).
Of the 20 given HAART who later died, 11 died within 100 days of starting therapy. The group that died within 100 days had a lower median CD4 at HAART initiation than the group that lived past 100 days: 58 cells/μL (IQR: 33–126) versus 169 (IQR: 100–204). Mean CD4s at HAART initiation were 108 cells/μL, SD = 113, for the group that lived less than 100 days and 198 cells/μL, SD = 147, for the group that lived more than 100 days. In contrast, the median CD4 count at death for 22 women who never received HAART and had a count recorded within 100 days of death was 140 cells/μL (IQR: 53–327); the mean CD4 was 216 cells/μL, SD = 200. For 8 women on HAART who had a CD4 count within 100 days of death, the median was 140 cells/μL (IQR: 29–228); the mean was 143 cells/μL, SD = 121.
Overall, 257 (52%) of the 473 women were followed after delivery, and those who had CD4 counts >200 at the date of introduction of HAART at GHESKIO were alive and in active follow-up at study close. They had between 3.5 and 10 years of follow-up in the MTCT program (Table 1).
The remaining 216 women were censored for the following reasons: 53 died; 32 declined care; 25 transferred care; and the fate of 106 (22% of the cohort) was unknown (lost to follow-up). Follow-up was related to whether the woman was kept on HAART after delivery: 9.8 women were lost per100 life-years (95% CI: 7.4 to 12.1) in the group never put on HAART, and 3.2 women were lost per 100 life-years in those who were on HAART from delivery (95% CI 0.1 to 6.4) (Table 2). This relationship was statistically significant. The group lost to follow-up and the group in active care did not differ significantly in their median age at enrollment (27 for both), their median time of enrollment in the program (May 2004 and March 2004, respectively) or their CD4 counts at delivery (555, SD = 292 and 553, SD = 319, respectively).
Time to HAART Initiation or Death
The median maternal CD4 count at delivery of the child defining entry into the cohort was 467 (IQR: 304–696). The median CD4 count at HAART initiation was 178 (IQR: 106–227). When the cohort was stratified by CD4 at delivery, a significant difference (P < 0.0001) was found in the progression to the combined endpoint of HAART initiation or death (Fig. 2). Estimated median HAART-free survival time for women who had a CD4 of <200 cells/μL was 7 months (IQR: 2–13); for CD4 = 200–349 cells/μL, it was 33 months (IQR 18–49); for CD4 = 350–499 cells/μL, it was 50 months (IQR: 44–57). For CD4 ≥ 500 cells/μL, there was no median available because fewer than 50% of women reached the endpoint.
Modeling of Time to CD4 ≤350 Cells per Microliter
A model of average CD4 decline in untreated women after delivery of their first child yielded times to a CD4 of 350 cells/μL that were significantly different between the groups with the higher CD4s at delivery (CD4 of 350–499 cells/μL and ≥500 cells/μL) (Fig. 3). The group with CD4 counts between 350 and 499 at delivery could be expected on average to decline to 350 cells/μL in 19 (95% CI: 14 to 28) months from delivery of the index child. The group with CD4 counts ≥500 would decline in 71 (95% CI: 59 to 88) months.
Time to Subsequent Pregnancy
Despite the availability of family planning services, 93 of the 473 women had 106 pregnancies and deliveries subsequent to the delivery that established them in the cohort. Seventy-one women (14% of the cohort) had more than one child while not on long-term HAART. The average time between deliveries before HAART initiation was 34.3 months (SD = 14.4 months). While not on HAART, the women of this cohort gave birth to 5.8 children/100 person-years (95% CI: 4.5 to 7.1). After starting HAART, the treated women gave birth to 5.2 children/100 person-years (95% CI: 3.3 to 7.2). Five women had more than 1 child while receiving HAART; 9 women had more than one child while not on HAART.
Haiti currently recommends prophylactic HAART administration for all HIV-positive pregnant women, regardless of CD4 count, and allowing at least a 6-month postpartum period of HAART prophylaxis for breastfeeding. This approach seems to be appropriate for this population, where 97% of women were found to breastfeed, exclusively or with other foods, in a 2006 demographic and public heath survey.3 Additionally, the transition in Haiti to a target CD4 of 350 cells/μL for long-term therapeutic HAART initiation in all HIV-infected individuals further shortens the potential HAART-free interval. In this setting, understanding patterns of mortality on and off HAART after pregnancy, follow-up and success of meeting treatment targets, and the rate at which postpartum women might reasonably be expected to decline in health are all important in formulating new treatment policy for postpartum women. The present study was undertaken to shed light on these areas.
All-cause mortality was not significantly different in HAART and no-HAART groups (Table 2). However, nearly half of the mortality in the group on HAART occurred within 100 days of starting therapy, and mortality was associated with a lower CD4 count at HAART initiation. Recent data from Haiti show that health outcomes are significantly worse when the target for HAART initiation is 200 rather than 350.9 Other recent studies also strongly suggest that initiation of HAART at lower CD4 counts carries an increased risk of complications of AIDS or death.10–12 Our findings—that mortality in those never put on HAART occurred exclusively in women with a CD4 count <350 cells/μL within 100 days of death—are consistent with this literature and supportive of the decision to increase the CD4 threshold for treatment initiation in Haiti.
In addition, more than 25% of women started HAART at a CD4 count less than 106 cells/μL, well below the target of 200 cells/μL, in effect during the study period. Although the threshold for treatment is now 350 cells/μL, the difficulty in beginning treatment at or near the intended CD4 count indicates a need for a follow-up plan specifically targeted at identifying women around the treatment threshold. Findings reported here regarding CD4 progression, if replicable, will help inform this strategy.
A potential relationship was explored between less frequent follow-up intervals and lower CD4 counts at the start of HAART therapy. There was a nonsignificant trend toward lower CD4s at HAART initiation with longer intervals between the 2 follow-up visits previous to HAART initiation (data not shown). Clearly, once a patient is seen and documented to be infected, a rigorous follow-up plan must be in place, so that they are started on HAART at a time that optimizes their clinical outcomes.
In addition to the late initiation of HAART, a higher than anticipated percentage of this cohort was lost to follow-up. It should be noted that being lost to follow-up did not include those who indicated that they had transferred care or declined follow-up. It may have partly represented movement out of or around the country, in which there is now national access to comprehensive HIV treatment. Nevertheless, many of these women had established care for themselves and their children at GHESKIO, and it is concerning that their outcomes are unknown.
The inability to comprehensively follow HIV-infected subjects before the need for HAART is not unique to GHESKIO, as shown by the fact that only 46% of HIV-infected South African adults got a repeat CD4 count with a median time between counts of 236 days,13 and that in another South African study only 15% of people with a CD4 count more than 200 returned for the results of their CD4 count.14 Rates of retention as low as 4% at 6 months have been reported from Malawi for asymptomatic HIV-infected people with higher CD4 counts.15
The loss to follow-up in these 473 women, overall 16% over 3 years, is also concerning because it does not mirror findings in another cohort at the same clinic; adults followed while receiving HAART at GHESKIO have been found to have an 8% rate of loss to follow-up over 5 years.16 However, when the group is broken down into those either on or off HAART from delivery to study close, the 48 women on HAART continuously had loss to follow-up rates similar to other GHESKIO populations on therapy. These 48 also had better follow-up in the first 3 years after delivery than those who were not on therapy, to a statistically significant degree (Table 2). Undocumented mortality, family dynamics, and misunderstanding of the need to follow-up for their own health (rather than just their children's health care) are all possible reasons for the higher rate of loss in those not started on therapy.
The question remains as to whether being placed on therapy was more likely in women who more reliably sought follow-up, or whether being on HAART is a predictor of good follow-up in and of itself. As this could be an important factor in the design of improved follow-up strategies (one of which could be keeping women on HAART indefinitely after delivery regardless of CD4 count). It is an area for future research.
Modeling of CD4 Decline
Our analysis shows that timing of CD4 progression to a critical level is predicted by the initial CD4 count. If the CD4 count is more than 500, it is unlikely to reach the threshold of 350 cells/μL in less than a year (Fig. 3). Therefore, follow-up intervals of 1 year would be appropriate while the CD4 count is above this level. However, with these longer intervals, other measures will need to be put in place to ensure that women remain in active follow-up. For women with lower CD4 counts at delivery, a shorter follow-up interval will be necessary to avoid missing the treatment threshold (Fig. 3).
When to Continue Treatment Past Delivery?
In women whose CD4 counts near delivery are between 350 and 499 cells/μL, our data indicates that it would be reasonable to extend treatment indefinitely past pregnancy and breastfeeding. We have found that without treatment these women will progress to a CD4 of less than 350 cells/μL by an average of 19 months from delivery (Fig. 3).
Recent evidence documents that maternal HAART while breastfeeding for infant prophylaxis will lower MTCT significantly during the postpartum period in resource-poor settings,17–23 and Haiti is therefore currently extending prophylactic HAART to all women up to 6 months postpartum to allow breastfeeding and considering a policy of continuous HAART after delivery. When these 6 additional months of HAART are accounted for, the additional HAART that would be administered to women with a CD4 between 350 and 499 cells/μL at delivery would average 13 months. Therefore, because follow-up is poorer in the cohort not receiving HAART (Table 2), and the average time to progression is relatively short, the benefit of the additional HAART administered in terms of improved health outcomes may well outweigh the costs of an extra 13 months of HAART, although this cost-benefit analysis is outside the scope of the current work.
Notably, for women with a CD4 of >500 cells/μL, the time to a CD4 of 350 cells/μL is significantly longer (71 months; Fig. 3). Therefore, if other studies can identify specific interventions to reduce the rate of loss to follow-up in the group with the highest CD4 count at delivery, a “watch and wait” approach with regards to starting HAART may prove to be safer and more cost-effective in these women than it is in the <500 CD4 group. This is an important area for future research. Recent WHO recommendations have provided the option of continuing all women after delivery on HAART.24
Furthermore, if a woman becomes pregnant a second time before beginning HAART, as 14% of the cohort did, she would begin yet another regime of prophylaxis. With an average birth rate of a child every 3.1 years and, ideally, at least 12 months of HAART per pregnancy (6 months before delivery and 6 months after delivery while breast feeding), a multiparous HIV-positive woman would be projected to spend about one-third of her time on HAART.
Additionally, the rate of multiparity in these women is evidence of continued heterosexual activity. Treating these women would have the additional benefit, not quantified here, of reducing transmission of the virus to partners in the community.25
As a retrospective study, this analysis was limited in its ability to control certain variables that may have affected outcomes or follow-up in the cohort studied. In particular, we were not able to track adherence to HAART therapy or viral loads, without which data the benefits of treatment may not be correctly quantified. As we are suggesting initiation of lifelong HAART, particularly if the CD4 count at delivery is <500 cells/μL, understanding adherence, resistance patterns, and viral rebound in this population would be essential before making any true policy recommendation. The benefits of treating over the longer period depend on resistance patterns and any viral rebound, so these will be key measures to evaluate in any prospective trial. However, adherence in this group is probably representative of usual patterns of adherence for this population, so conclusions as to outcomes made here are still informative.
We also acknowledge that in analyses separating women into HAART and no-HAART groups (mortality, subsequent pregnancy, and follow-up outcomes), women never put on long-term HAART may have been healthier than those who were, introducing bias into these analyses. There was a large, although not statistically significant, difference in average CD4 at delivery between these 2 groups (Table 1). This study, as a retrospective analysis, was not powered to control for that difference, but it may have affected our reported outcomes.
Additionally, the high number of women with unknown outcomes may result in a bias that we were unable to quantify here. It is unclear from this data whether there are specific risk factors for loss to follow-up that could be addressed, or which would change the conclusions of this study. Also, our definition of loss to follow-up (seen in the year before study close) is only a rough indicator of quality of follow-up.
Randomized prospective studies such as Promise (IMPAACT 1077) are needed to elucidate these issues. Finally, analysis of certain outcomes and relationships between certain variables was limited by small sample size.
Summary and Further Research
This study begins to address some of the issues that are central to maternal HIV care. In particular, the present analysis provides a model of CD4 decline after delivery, and adds evidence that missing the target CD4 count for starting HAART was common and likely led to excess mortality in this cohort. With a better understanding of determinants of follow-up and confirmation of the CD4 decline curve modeled here, we will be able to better predict in this setting and others how to best maximize the quality of care for postpartum HIV-positive women.
The helpful comments of Drs Paul Palumbo and Serena Koenig on late drafts of the manuscript are acknowledged with appreciation.
5. World Health Organization. Antiretroviral drugs for treating pregnant women and preventing HIV infection in infants: toward universal access: recommendations for a public health approach—2006 version. 2006. Available at: http://www.who.int/hiv/pub/guidelines/pmtctguidelines3.pdf
. Accessed August 23, 2012.
7. Noel F, Mehta S, Zhu Y, et al.. Improving outcomes in infants of HIV-infected women in a developing country setting. PLoS One. 2008;3:e3723.
8. Deschamps M, Noel F, Bonhomme J, et al.. Prevention of mother-to-child transmission of HIV in Haiti
. Rev Panam Salud Pública. 2009;25:24–30.
9. Severe P, Juste MAJ, Ambroise A, et al.. Early versus standard antiretroviral therapy for HIV-infected adults in Haiti
. N Engl J Med. 2010;363:257–265.
10. Emery S, Neuhaus JA, Phillips AN, et al.. Major clinical outcomes in antiretroviral therapy (ART)–naive participants and in those not receiving ART at baseline in the SMART study. J Infect Dis. 2008;197:1133–1144.
11. Kitahata MM, Gange SJ, Abraham AG, et al.. Effect of early versus deferred antiretroviral therapy for HIV on survival. N Engl J Med. 2009;360:1815–1826. Available at: http://content.nejm.org
. Accessed August 23, 2012.
13. Kranzer K, Zeinecker J, Ginsberg P, et al.. Linkage to HIV care and antiretroviral therapy in Cape Town, South Africa. PLoS One. 2010;5:e13801.
14. Larson BA, Brennan A, McNamara L, et al.. Lost opportunities to complete CD4+ lymphocyte testing among patients who tested positive for HIV in South Africa. Bull World Health Organ. 2010;88:675–680.
15. Tayler-Smith K, Zachariah R, Massaquoi M, et al.. Unacceptable attrition among WHO stages 1 and 2 patients in a hospital-based setting in rural Malawi: can we retain such patients within the general health system? Trans R Soc Trop Med Hyg. 2010;104:313–319.
16. Leger P, Charles M, Severe P, et al.. 5-year survival of patients with AIDS receiving antiretroviral therapy in Haiti
. N Engl J Med. 2009;361:828–829. Available at: http://content.nejm.org
. Accessed August 23, 2012.
17. Kesho Bora Study Group. Eighteen-month follow-up of HIV-1-infected mothers and their children enrolled in the Kesho Bora study observational cohorts. J Acquir Immune Defic Syndr. 2010;54:533–541.
18. Marazzi MC, Nielsen-Saines K, Buonomo E, et al.. Increased infant human immunodeficiency virus-type one free survival at one year of age in sub-saharan Africa with maternal use of highly active antiretroviral therapy during breast-feeding. Pediatr Infect Dis J. 2009;28:483–487.
19. Mofenson LM. Protecting the next generation—eliminating perinatal HIV-1 infection. N Engl J Med. 2010;362:2316–2318.
20. Palombi L, Marazzi MC, Voetberg A, et al.. Treatment acceleration program and the experience of the DREAM program in prevention of mother-to-child transmission of HIV. AIDS. 2007;21(suppl 4):S65–S71.
21. Peltier CA, Ndayisaba GF, Lepage P, et al.. Breastfeeding with maternal antiretroviral therapy or formula feeding to prevent HIV postnatal mother-to-child transmission in Rwanda. AIDS. 2009;23:2415–2423.
22. Shapiro RL, Hughes MD, Ogwu A, et al.. Antiretroviral regimens in pregnancy and breast-feeding in Botswana. N Engl J Med. 2010;362:2282–2294.
23. Taha T, Kumwenda J, Cole S, et al.. Postnatal HIV‐1 transmission after cessation of infant extended antiretroviral prophylaxis and effect of maternal highly active antiretroviral therapy. J Infect Dis. 2009;200:1490–1497.
24. World Health Organization. Use of antiretroviral drugs for treating pregnant women and preventing HIV infection in infants. 2012. Available at: http://www.who.int/hiv/PMTCT_update.pdf
. Accessed August 23, 2012.
25. Montaner JSG, Lima VD, Barrios R, et al.. Association of highly active antiretroviral therapy coverage, population viral load, and yearly new HIV diagnoses in British Columbia, Canada: a population-based study. Lancet. 2010;376:532–539.
Keywords:© 2012 Lippincott Williams & Wilkins, Inc.
Haiti; PMTCT; HAART