Postnatal transmission of HIV-1 after a maternal short-course zidovudine peripartum regimen in West Africa

Background: To assess the postnatal transmission (PT) risk of HIV-1 after a maternal short-course zidovudine regimen in a breastfeeding population.

Methods: Data were pooled from two trials: ANRS 049a DITRAME (Abidjan, Côte d'Ivoire and Bobo-Dioulasso, Burkina-Faso) and RETROCI (Abidjan). Consenting HIV-1 seropositive women were randomized at 36-38 weeks' gestation between September 1995 and February 1998, to receive oral zidovudine or placebo: one tablet twice daily until delivery, and in DITRAME only, for 7 more days. A PT case was infection in a child with a negative HIV-1 PCR at age ≥ 30 days who later became infected as defined by a positive HIV-1 PCR, or if aged ≥ 15 months, a positive HIV serology. Cumulative risks (CR) of PT were computed using a competing risk approach with weaning as a competing event.

Findings: At age 24 months, CR for PT were similar in the zidovudine (9.8%, n = 254) and placebo groups (9.1%, n = 225). In a multivariate model of PT risk factors, the treatment effect was not significant, maternal CD4 cell count < 500 × 10⁶/l at entry tripled the hazard compared to women with CD4 cell counts ≥ 500 × 10⁶/l [hazard ratio (HR), 3.14; 95% confidence interval (CI), 1.31-7.49] as well as an increased maternal plasma viral load at entry (HR, 2.65 for 1 log₁₀ increase; CI, 1.75-4.00).

Interpretation: PT occurred at a similar rate between arms and therefore reduced the long-term overall efficacy of this peripartum zidovudine regimen at age 24 months. The higher risk of PT among women with low CD4 cell count emphasizes the importance of identifying interventions to prevent PT for these women.

Introduction

Postnatal transmission (PT) of HIV-1 through breast milk is of great concern in Africa where breastfeeding is a common practice. Recently, in Kenya, a randomized trial comparing breastfeeding to formula feeding in the absence of any antiretroviral therapy showed an additional PT risk of 16% at age 2 years, with 44% of all HIV infections attributable to PT in the breastfeeding arm [¹]. It has also been hypothesized that subsequent PT could have negative implications for the overall long-term efficacy of peripartum antiretroviral interventions given to mothers to prevent mother-to-child transmission (MTCT) [^2,3]. Subsequent PT that is unaffected by early therapy will reduce the overall long-term efficacy of peripartum antiretroviral regimens, and the higher the PT rate, the greater the impact on efficacy [⁴]. It is also possible that PT risk would be greater in the intervention arm due to a viral load rebound after the antiretroviral treatment discontinuation, leading to a null or negative effect of the antiretroviral peripartum intervention [⁵]. Thus, the subsequent risk of PT through breastfeeding after a peripartum antiretroviral intervention must be assessed and an improved understanding of the timing of PT of HIV and its determinants is important in order to design postpartum interventions complementing appropriate peripartum interventions, including infant feeding strategies and weaning policies [^6,7].

Two randomized trials assessed the efficacy of a peripartum maternal short-course zidovudine regimen to prevent MTCT in African breastfed populations. Despite breastfeeding, these African trials demonstrated a significant reduction of early MTCT, 37% at age 3 months [⁸] and 38% at age 6 months [⁹]. Because these trials individually lacked statistical power to draw conclusions about the impact of PT on the long-term efficacy of the zidovudine peripartum intervention (or the potential effect of other risk factors on transmission), a pooled analysis was conducted as a collaboration between the DITRAME (French National AIDS Research Agency, ANRS) and RETROCI (US Centers for Disease Control and Prevention, CDC) teams. We recently reported that zidovudine efficacy is maintained at age 24 months despite prolonged breastfeeding but is also strongly associated with the maternal immune status [¹⁰]. In the present analysis, we assessed the risk of, timing of, and risk factors for PT of HIV-1 after a maternal short peripartum regimen of oral zidovudine.

Methods

Individual data were pooled from these two randomized double-blind placebo-controlled trials designed to assess the efficacy of short maternal zidovudine regimens in breastfed children: DITRAME ANRS-049a was conducted in Abidjan, Côte d'Ivoire and Bobo-Dioulasso, Burkina-Faso [⁸] and RETROCI in Abidjan [⁹]. Detailed methodology has been published elsewhere [¹⁰]. Briefly, between September 1995 and February 1998, consenting pregnant HIV-1 seropositive women with hemoglobin ≥ 70 g/l were randomized at 36-38 weeks' gestation to receive oral zidovudine (250 mg or 300 mg) or a matching placebo: one tablet twice daily until the beginning of labor, then a single oral dose of 500 mg or 600 mg (DITRAME) or one 300 mg tablet every 3 h until delivery (RETRO-CI), then a 7-day postpartum maternal treatment of 500 or 600 mg per day (DITRAME only). No study drug was given to the neonate.

Baseline maternal lymphocyte subtypes were counted using a flow cytometry (FACScan, Becton Dickinson, Dartford, England). In DITRAME, baseline maternal plasma HIV-1 RNA levels were measured using a branched DNA assay with a detection threshold of 50 HIV RNA copies/μl (Quantiplex 340, version 3.0; Bayer-Chiron, Walpole, Massachusetts, USA). RNA levels were measured when possible for all mothers who transmitted HIV and for a subsample of women who did not (two controls per case, matched on site and treatment group). In RETRO-CI, all maternal plasma HIV-1 RNA levels were measured at enrolment (Amplicor HIV Monitor version 1.5; Roche Diagnostics Systems Inc., Branchberg, New Jersey, USA). Clinical follow-up of and blood collection from each live-born child were scheduled within 1 week of birth, then at 4 weeks (RETRO-CI) or 6 weeks (DITRAME), then at 3 months of age and every 3 months thereafter until 24 months of age. Infant feeding practices and maternal breast lesions (only available in DITRAME) were reported at each visit on standardized questionnaires.

For each child, the sample collected at 3 months (RETRO-CI) or 6 months (DITRAME), or an earlier one when these were not available, was systematically analyzed by PCR. If this sample was positive, all the preceding available samples were analyzed by PCR. PCR was used for samples obtained until 9 months of age for DITRAME and 12 months for RETRO-CI [¹⁰]. HIV-1 DNA nested PCR was used in Abidjan, with primers from the protease gene (DITRAME and RETRO-CI). In Bobo Dioulasso (DITRAME), samples were analyzed first by both DNA PCR and quantitative plasma RNA PCR (Amplicor HIV Monitor version 1.5) and later by RNA PCR only. Serum samples collected between 9 (DITRAME) or 12 months (RETRO-CI) and 24 months of age were screened for HIV-1 and HIV-2 antibodies by a commercial ELISA (Genelavia Mixt, Diagnostics Pasteur, France or Murex ICE 1-O-2, Murex Biotech Ltd, Dartford, UK) and confirmed on the same sample with a commercial synthetic peptide ELISA (Peptilav 1-2, Diagnostics Pasteur).

Pediatric HIV-1 infection was defined by one positive HIV-1 DNA or RNA-PCR test result at any age, or if aged 15 months or greater, at least two positive HIV serologic tests. Early transmission was defined as a positive PCR by age 4-6 weeks, diagnosed either on the first week or the 4-6 week samples. PT was defined as a child with a negative HIV-1 PCR the age of at least 30 days who became infected (as defined above) at older than 4-6 weeks. Period of infection was undefined for those children who could be classified neither as having early infection nor as a PT case. A single negative test result 60 days or more after complete cessation of breastfeeding defined the definitive absence of infection. Children with a negative diagnosis on the last available sample while still breastfed or who had been breastfed within the last 59 days were classified as provisionally uninfected. For all analyses, only the first-born child was used if a woman had a multiple birth.

Differences between treatment groups in mothers' and babies' characteristics at entry and during follow-up were analyzed using Student's t test or the Mann-Whitney test for continuous variables and the Chi-square test or Fisher's exact test (as appropriate) for categorical variables. Probabilities of being breastfed were estimated using the Kaplan-Meier method.

To estimate the cumulative PT risk at 24 months of age, all breastfed children with a negative PCR test result from 30 days were included in a survival analysis comparing groups according to randomization assignment. Cumulative risk was estimated non-parametrically using a modification of Turnbull's method for interval-censored data [¹¹] to handle competing risks [¹²] in which the competing events are complete cessation of breastfeeding and acquisition of infection. Infection was interval-censored between the last negative and first positive test. Definitively uninfected children were regarded as no longer at risk for HIV 60 days after the age of complete cessation of breastfeeding, or right-censored at age 24 months, whichever was earlier. Provisionally uninfected children were right-censored at the age of their last negative test. For each treatment group, cumulative probabilities of PT were estimated within maternal CD4 cell count strata (< 500 × 10⁶ cells/l and ≥ 500 × 10⁶ cells/l, a cut-off point chosen post hoc). Standard errors and risk differences between treatment groups were estimated from a bootstrap with 1000 replications.

PT incidence rates and 95% confidence intervals (CI) were also estimated using the person-time approach. Differences between these estimates according to treatment group and study centers were tested using Poisson regression [¹³]. In these analyses, we assumed that the date of infection was the midpoint between the last negative test and the first positive test.

Estimation of smooth age-specific PT incidence and evaluation of risk factors for PT were performed using the penalized likelihood approach and proportional hazard regression model for interval-censored data [¹⁴]. These analyses do not consider the competing risk from weaning. Upper and lower confidence bounds of the estimate of age-specific incidence were estimated using a Bayesian technique [¹⁴]. The potential determinants of PT risk studied in this analysis were maternal treatment (zidovudine versus placebo), baseline maternal CD4 cell count, baseline maternal plasma viral load, and one time dependent variable, pediatric oral thrush.

In all analyses, deaths among HIV-negative children were regarded as independent right censoring events. All statements about statistical significance are based on tests with a type I error of 0.05.

Results

From September 1995 to February 1998, 701 women were enrolled, of whom 23 (3.3%) were lost to follow-up before delivery (Fig. 1). Characteristics of the 678 women who delivered and their use of assigned treatment were similar between the two treatment groups: in particular, for median duration of prepartum treatment (23 days), the proportion receiving intrapartum treatment (81%), and median length of postpartum maternal treatment (7 days, DITRAME only) [¹⁰]. After exclusion of the second twin from 11 twin pairs, only 21 of the 662 live-born children (3.2%) had no HIV test result.

Fig. 1

Among the 641 remaining children, 12 (1.9%) were formula-fed from birth, eight in the zidovudine group (of whom two were considered infected early) and four in the placebo group (P = 0.23). Age at cessation of breastfeeding was unknown for one child. For the remaining 628 children, the distributions of breastfeeding duration were similar between groups at each site (P ≥ 0.41) but differed significantly between sites (P < 0.001). Median duration of breastfeeding was 8.1 months in Abidjan-DITRAME [interquartile range (IQR), 7-10 months], 15.2 months in Abidjan- RETROCI (IQR, 13-18 months) and 19.4 months in Bobo-Dioulasso-DITRAME (IQR, 18-22 months).

By age 24 months, 68 children born to 319 women assigned to zidovudine and 94 of those born to 322 women assigned to placebo were known to be HIV infected (Fig. 1). After exclusion of the 111 children with early infection, the nine children infected with an unknown timing, the 10 uninfected children not breastfed, and the 32 children with at least one negative HIV test but with no test result after age 4-6 weeks, data from 479 children at risk for PT were analyzed. Baseline characteristics of these children and their mothers were similar between arms (Table 1). Among these 479 children, 23 born to 254 women in the zidovudine group and 19 born to 225 women in the placebo group became infected by PT (Fig. 1). For these 42 PT cases, the median age at first positive test was 9 months (IQR, 6-15 months). The estimated cumulative PT risks in the zidovudine group at ages 6, 12, and 24 months were 3.4%, 9.2%, and 9.8%, respectively (Fig. 2). The corresponding risks in the placebo group were 3.4%, 6.8% and 9.1%. The PT risk at each of these ages was similar in the two arms (difference at 24 months, 0.7; 95% CI, -4.8 to 5.7). PT accounted for at least 26% (42/162; CI, 19-33%) of all HIV infections in this cohort, while 19% of the overall cohort was still uninfected and breastfed and hence at risk for PT at age 24 months.

Fig. 2

Table 1

Despite differences in the duration of breastfeeding among sites, the instantaneous risk of PT did not depend on site. In Abidjan-DITRAME, where the median duration of breastfeeding was about 8 months, the PT risk was estimated to be 10.5 (CI, 4.8-16.2) per 100 child-years of breastfeeding. In Bobo-Dioulasso-DITRAME, where the median duration of breastfeeding was 19.5 months, the PT risk was 7.4 (CI, 3.2-11.6) per 100 child-years of breastfeeding. In Abidjan-RETROCI, where the median duration of breastfeeding was 15 months, the PT risk was 8.5 (CI, 4.4-12.5) per 100 child-years of breastfeeding. Using Poisson regression, neither treatment group nor center had a significant effect on these estimates (P = 0.77 and P = 0.68, respectively). The overall pooled cumulative PT risk was estimated to be 8.6 (95% CI, 6.0-11.2) per 100 child-years of breastfeeding at age 24 months. The estimated age-specific PT incidence rates per 100 child-years of breastfeeding using PHMPL estimates were 9.1 [lower bound (LB), 4.4 to upper bound (UB) 13.9] at age 3 months, 6.9 (LB, 2.2 to UB, 11.7) at age 6 months, 9.5 (LB, 4.0 to UB, 15.3) at age 12 months, and 13.1 (LB, 2.9 to UB, 23.4) at age 18 months.

Cumulative PT risk was much higher among children born to women with CD4 cell count < 500 × 10⁶/l than among those with ≥ 500 × 10⁶ CD4 cells/l (Table 2). Among women with CD4 cell counts < 500 × 10⁶/l, 24-month PT risks were 21.8% in the zidovudine and 16.1% in the placebo group, representing 86% and 68% of the PT cases in the zidovudine and placebo arms, respectively. Within both maternal CD4 cell count strata, the difference in cumulative risk between treatment groups at each age was not statistically significant.

Table 2

Cumulative risks of PT were similar between trials (difference at 24 months, -0.6%; 95% CI, -9.3 to 7.1). Univariate proportional hazard analyses showed that there was no treatment effect on PT but significantly higher risks with decreased maternal CD4 cell count and increased baseline maternal viral load (Table 3). A maternal CD4 cell count < 500 × 10⁶/l was associated with a 7.7-fold increase in the hazard. A diagnosis of pediatric oral thrush before seroconversion was not associated with an increase in PT risk. There was no significant interaction between treatment and CD4 cell count (P = 0.86) and no significant interaction between duration of breastfeeding and maternal CD4 cell count (P = 0.47), the latter suggesting that women with low CD4 cell count did not stop breastfeeding earlier as a result of illness. In a multivariate model with treatment, maternal CD4 cell count, and maternal viral load as risk factors based on 306 children (including 32 PT cases), the treatment effect was not significant [hazard ratio (HR) zidovudine to placebo, 1.15; CI, 0.57-2.31], maternal CD4 cell count < 500 × 10⁶/l at entry tripled the hazard for PT compared to women with ≥ 500 × 10⁶ CD4 cells/l (HR, 3.14; CI, 1.31-7.49), and increased maternal plasma viral load at entry increased also the hazard (HR, 2.65 for 1 log₁₀ increase; CI, 1.75-4.00).

Table 3

Discussion

This is the first analysis of the postnatal transmission risk at 24 months after a maternal peripartum short-course zidovudine regimen to reduce MTCT in breastfeeding populations. Median duration of breastfeeding in our populations was 8.1-19.4 months according to these West African study sites. Our analysis demonstrates that PT risk is substantial but similar between treatment arms and without a clear trend to increase or decrease with time. In addition, we demonstrate that PT risk is strongly associated with advanced maternal HIV disease as indicated by low maternal CD4 cell count and high maternal viral load. Because the cumulative PT risk between ages 4-6 weeks and 24 months (approximately 9%) was substantial but similar in the zidovudine and placebo arms, the estimated overall zidovudine long-term efficacy of 26% at 24 months, including also in utero and intrapartum transmissions, was smaller than the short term-efficacy of 41% at 6 weeks [¹⁰] but still significantly greater than zero. It has been hypothesized that interruption of maternal zidovudine at the end of the peripartum period might cause a rebound in maternal viral load to a level higher than before the treatment at least transiently, resulting in an sustained increased risk of postnatal HIV transmission through breast milk in children exposed to zidovudine [⁵]. This could have led to a loss of the effect of the zidovudine peripartum intervention or even to a negative effect. Although we cannot distinguish between in utero, intrapartum, and early postpartum HIV infection (all resulting in detectable infection by age 4-6 weeks), the fact that the PT transmission risk in this study was essentially the same in the placebo and zidovudine groups both for women with CD4 cell counts of ≥ 500 × 10⁶ cells/l and < 500 × 10⁶ cells/l at enrolment supports the hypothesis that any viral rebound associated with zidovudine monotherapy interruption has little if any effect on PT. It is also corroborated by the lack of a negative effect of zidovudine on CD4 cell count and viral load at 6 months postpartum [¹⁵].

It is likely that we underestimated the cumulative PT risk because some children with early postnatal infections would be PCR-positive at age 4-6 weeks and would be excluded from this analysis. In addition, 19% of the uninfected children were still breastfed at age 24 months and hence remained at risk for PT. Despite these biases, we estimate that breast milk transmission accounts for an additional risk of at least 9%. This is less than the recent estimate of 16% from the Kenya breastfeeding trial considered as the most reliable figure to date [¹].

PT was strongly associated with advanced maternal HIV disease (low maternal CD4 cell count or high maternal plasma viral load at enrolment) reaching 16-22% at age 24 months in children born to women with < 500 × 10⁶ CD4 cells/l CD4, but only 2-5% in women with ≥ 500 × 10⁶ CD4 cells/l. Of all PT cases, 77.5% occurred among the 38% of children born to mothers with < 500 × 10⁶ CD4 cells/l enrolment. In this group, cumulative PT risk at 24 months of age was 1.3-7.5 times higher than in women with ≥ 500 × 10⁶ CD4 cells/l in both treatment arms (Table 3). This suggests that maternal postnatal highly active antiretroviral therapy (HAART) could reduce breast milk transmission by decreasing maternal viral load, particularly in women with more advanced HIV disease. Because the treatment arm-specific cumulative transmission risks at age 4-6 weeks in DITRAME and RETROCI were similar [¹⁰], 1 week of maternal postpartum zidovudine treatment (as used only in DITRAME) is unlikely to have a substantial effect on the risk of early PT. It is also probable that such a short antiretroviral peripartum monotherapy had little effect on the immunovirological response in these women with more advanced disease.

We may underestimate any effect of viral rebound because some PT cases may have occurred before age 4-6 weeks, the age at which we required that children have a negative PCR test. However, a minimal delay of about 4 weeks is needed for PCR test to have a good sensitivity for diagnosing early HIV infection [¹⁶], even in breastfed children. Secondly, this possible underestimate of PT should have been similar in both arms because there no reason to believe that zidovudine monotherapy affects the PCR diagnosis of HIV after therapy. Finally, we found no significant difference between treatment arms for early cumulative PT estimation risks from 1 month of age (Table 2). Thus, our data suggest that any viral rebound after zidovudine therapy does not have a major effect on PT during the early period until 6 weeks and that therefore our assessment of PT at later ages provides a reasonable estimate of this phenomenon.

The strong dependence of PT on maternal CD4 cell count demonstrates that the distributions of maternal CD4 cell counts must be known when comparing PT rates between studies. Our PT rate of 8.6 per 100 child-years of breastfeeding is similar to that observed in an observational cohort in Malawi [¹⁷] but higher than the estimate of 3.2 per 100 child-years of breastfeeding (95% CI, 3.1-3.8) from an international pooled analysis [¹⁸]. In this latter study, children were included from 3 months of age with a median age at entry of 6 months, older than in our study, because early PCR results often were not available. Maternal CD4 cell count distributions were also not published for either study.

Our findings suggest that the age-specific incidence of PT was approximately constant between ages 4-6 weeks and 18 months. Miotti et al. found that PT risk decreased with age [¹⁷], but their estimates may include some intrapartum transmission as their ages at inclusion ranged between 0.7 and 16.8 months, which could have led to an overestimation of early PT rates leading to an artificial decrease over time [¹⁹]. We acknowledge that we also underestimated early postnatal transmission before age 4-6 weeks, but our analysis showed a constant PT risk beyond that age until 18 months, with a reasonable precision.

Our study is the first one to report that PT risk is associated with maternal viral load. We did not estimate risk within viral load strata because a case-control design was used to select non-transmitting mothers for whom viral load information was obtained in DITRAME. We do not believe that this sampling biased our results for two reasons. First, the effect of CD4 cell count on PT risk was similar to that in Table 2 when we restricted the analysis to women with viral load results. Secondly, the CD4 cell count distributions were not different in women with and without baseline viral load measurements (data not shown).

A cohort study in Kenya also found that decreasing maternal CD4 cell count is a risk factor for PT [²⁰]. Infant oral thrush did not appear to be a strong risk factor for PT in our study, in contrast to the Kenyan and Malawi studies [^20,21]. Unfortunately, our data did not document infant feeding modalities well enough to investigate the potential increased risk of PT from mixed feeding as reported by Coutsoudis et al. [²²]. Standardized approaches are needed to explore this issue further [²³].

High rates of PT, especially among women with low CD4 cell counts, suggest the need to evaluate interventions to prevent HIV transmission through breastfeeding, including alternatives to breastfeeding and/or the use of maternal HAART after delivery. In Kenya, formula feeding was found to be safe overall but is not feasible everywhere [²⁴]. A project initiated in 2001 in Abidjan is investigating alternatives to prolonged breastfeeding (formula feeding from birth or exclusive breastfeeding with early cessation) to reduce PT after a peripartum antiretroviral intervention [²⁵]. As we found that PT risk appears to neither decrease nor increase over time, postpartum interventions should be evaluated to cover the entire breastfeeding period. Because PT risk is substantially higher in women with a low CD4 cell count, and these women also have a higher risk of opportunistic infection and death, maternal HAART would be especially beneficial for these women, both for themselves and also to reduce PT. It would be important in this context to analyze the effect of postpartum antiretroviral regimens not only on PT risk but also on HIV viral load in maternal body compartments, particularly breast milk.

The public health implications of our results should be viewed in light of the 2000 WHO/UNAIDS recommendations according to which the safety and effectiveness of antiretroviral regimens to prevent HIV MTCT warrant their use beyond pilot projects and research settings in Africa [²⁶]. If CD4 cell counts could be easily measured for pregnant HIV-1-infected African women, antiretroviral peripartum regimens and infant feeding options [⁷] could be discussed with these women considering their own health and their risk of MTCT, especially PT. These comprehensive and tailored approaches are especially needed for women with advanced HIV disease in order to fulfill the recently agreed United Nations target of reducing MTCT by 20% by the year 2005 [²⁷].

Acknowledgements

The authors thank the women who participated in the study. We thank L. Dequae-Merchadou (Unité INSERM 593) and R. Odum (CDC) for data management. Thanks to R.J. Simonds (CDC) for editorial suggestions.

Sponsorship: The DITRAME-ANRS 049a trial was sponsored by the Agence Nationale de Recherches sur le SIDA (ANRS, Paris, France) and the French Ministry of Cooperation, within the Coordinated Action AC12. The RETRO-CI trial was sponsored by the US Centers for Disease Control and Prevention.