| |
Extended nevirapine prophylaxis to prevent HIV transmission
|
| |
| |
The Lancet, 26 July 2008
Jeffrey SA Stringer a and Benjamin H Chi a
a. University of Alabama at Birmingham, Centre for Infectious Disease Research in Zambia (CIDRZ), Lusaka, Zambia
"....Extended infant prophylaxis with nevirapine is simple enough to be implemented almost anywhere. It represents a long-awaited, if partial, solution to a mother's impossible choice. We should not delay.......It is thus our view that policy makers should recommend the PEPI regimen in settings where it is feasible, and take the results with the SWEN regimen as key supporting evidence that infant prophylaxis with nevirapine is safe and effective. Whether additional protection could be conferred by extending the period of infant prophylaxis throughout the entire breastfeeding period is unknown (but plausible), and currently the focus of at least two studies. The PROMISE study, planned by the IMPAACT network of the US National Institutes of Health, will investigate extended prophylaxis with nevirapine, among other interventions. The PROMISE-PEP study,7 funded by the French Government, will study extended prophylaxis with lamivudine.......The new results with the SWEN regimen may be better understood in comparison with the recently published PEPI study,5 a similar randomised trial undertaken at six sites in Malawi. PEPI compared a 14-week regimen of daily nevirapine in infants with the current standard of care in Malawi: single-dose intrapartum and neonatal nevirapine plus 1 week of infant prophylaxis with zidovudine.6 Compared with the results with the SWEN study, the PEPI intervention was longer (14 weeks), used a higher dose of infant nevirapine, and started prophylaxis immediately after birth (rather than waiting a week). At 9 months of age, 10·6% of liveborn infants uninfected at birth in the standard of care group had become infected, compared with 6·4% in the extended-dose nevirapine group, a highly statistically significant finding. (The PEPI trial also evaluated a third regimen, of the nevirapine intervention plus the addition of 14 weeks of infant zidovudine, which proved no more effective than nevirapine alone but was more toxic.).........The use of nevirapine as a component of any non-suppressive prophylactic regimen is controversial, because the agent rapidly induces resistance to the non-nucleoside reverse-transcriptase inhibitor (NNRTI) class of antiretroviral drugs. A meta-analysis estimated that 52% of infants who become infected with HIV despite a single dose of nevirapine will develop NNRTI resistance.8 It is probable that nearly all infants who become infected while receiving an extended prophylactic course of nevirapine monotherapy will develop such resistance. Should this mitigate enthusiasm for breastfeeding prophylaxis with extended-dose nevirapine? Probably not. In the study reported today, 54 babies randomised to the control group were found to be HIV-infected at 6 weeks of life, compared with 25 babies who received the SWEN regimen. Assuming that 52% of infants in the control group and 100% of infants in the SWEN group develop resistance, we find that the numbers of babies developing resistance are essentially the same (28 vs 25). Thus the SWEN regimen prevented 29 additional infections at the cost of, at most, three additional NNRTI-resistant infections. Clearly, the induction of any resistance should be avoided, but it seems that extended-dose nevirapine regimens will save many babies without conferring any additional risk of NNRTI resistance to the population beyond that of single-dose nevirapine...."
Of the 800_000 new paediatric HIV-1 infections each year, breastfeeding underlies at least a third.1 In developed countries, where water is clean and infant formula affordable, HIV-infected women generally avoid breastfeeding. Yet in many settings where HIV is endemic, infant formula is too costly for most families and its use a stigma of maternal HIV infection. Moreover, replacement feeding can be unsafe; several studies indicate that the HIV-prevention gains made by breastfeeding avoidance or early weaning are largely offset by an increased risk of death from diarrhoea, pneumonia, and malnutrition.2,3 Thus the HIV-infected mother of an African or south Asian newborn baby often faces an impossible choice: breastfeed and risk HIV transmission, or do not breastfeed and risk death from another cause.
In today's Lancet, investigators from the Six Week Extended-Dose Nevirapine Study point towards a solution to this seemingly intractable problem.4 They report three coordinated randomised trials that were done separately in Uganda, Ethiopia, and India, and combined in a preplanned analysis (of more than 2000 infants). SWEN is thus not a trial acronym, but is a convenient abbreviation for the extended-dose nevirapine regimen. In each country, breastfeeding HIV-exposed infants were randomly allocated to receive 5 mg of daily oral nevirapine from days 8 to 42 of life (the SWEN regimen) or no additional drug. All participants received at least intrapartum and neonatal single-dose nevirapine, and some, in India, received other antenatal antiretrovirals as well (mostly zidovudine).
In the primary analysis, which excluded infants already infected at birth, the SWEN regimen failed to achieve a statistically significant reduction in transmission risk at 6 months of age. It did, however, seem to improve key secondary endpoints, such as transmission at 6 weeks of age (2·5% vs 5·3%, intervention vs control; risk ratio 0·54, 95% CI 0·34-0·85) and HIV-free survival at 6 months of age (8·1% vs 11·6%; 0·73, 0·55-0·97). Importantly, the SWEN regimen was well tolerated with no substantially increased risks of rash, liver damage, neutropenia, or other adverse outcomes.
Although soundly designed and executed, the trials had some curious features. First, in Uganda, there was a much lower prevalence of HIV infection at birth in babies randomised to the control group (4·3%) than to the SWEN regimen (9·6%). Because neither the control nor the SWEN regimen would be expected to affect in-utero infection, we must assume that this imbalance occurred by chance, and represents a failure of the randomisation procedure to equally allocate this important confounder between the study groups. The imbalance is unfortunate, because inclusion of infections present at birth in the planned secondary analysis all but negates the protective effect of the SWEN regimen at both 6 weeks and 6 months of follow-up.
Second, although the SWEN regimen seems to have protected against infant death (1·1% of babies died by 6 months of age in the experimental group vs 3·6% in the control group, a statistically significant difference), most of this protection was seen in infants who were HIV-uninfected at their last visit. That visit was on average 2 weeks before death. This finding, too, is a head-scratcher, for it suggests that nevirapine confers a survival advantage unrelated to its HIV-prevention effect. Further, it is the inclusion of these deaths in a combined outcome of HIV-free survival that makes the 6-month trial outcomes statistically significant.
The new results with the SWEN regimen may be better understood in comparison with the recently published PEPI study,5 a similar randomised trial undertaken at six sites in Malawi. PEPI compared a 14-week regimen of daily nevirapine in infants with the current standard of care in Malawi: single-dose intrapartum and neonatal nevirapine plus 1 week of infant prophylaxis with zidovudine.6 Compared with the results with the SWEN study, the PEPI intervention was longer (14 weeks), used a higher dose of infant nevirapine, and started prophylaxis immediately after birth (rather than waiting a week). At 9 months of age, 10·6% of liveborn infants uninfected at birth in the standard of care group had become infected, compared with 6·4% in the extended-dose nevirapine group, a highly statistically significant finding. (The PEPI trial also evaluated a third regimen, of the nevirapine intervention plus the addition of 14 weeks of infant zidovudine, which proved no more effective than nevirapine alone but was more toxic.)
One interpretation of these two studies is that while the SWEN regimen is partly effective, the PEPI regimen is more so. Perhaps, in the SWEN study, the extended dose of nevirapine was too low, or the duration of prophylaxis too short. Whether the PEPI regimen is truly better than the SWEN regimen cannot be known with certainty without a head-to-head comparison. However, the PEPI intervention produced a more definitive result, and covers a longer period of infant risk. It is thus our view that policy makers should recommend the PEPI regimen in settings where it is feasible, and take the results with the SWEN regimen as key supporting evidence that infant prophylaxis with nevirapine is safe and effective. Whether additional protection could be conferred by extending the period of infant prophylaxis throughout the entire breastfeeding period is unknown (but plausible), and currently the focus of at least two studies. The PROMISE study, planned by the IMPAACT network of the US National Institutes of Health, will investigate extended prophylaxis with nevirapine, among other interventions. The PROMISE-PEP study,7 funded by the French Government, will study extended prophylaxis with lamivudine.
The use of nevirapine as a component of any non-suppressive prophylactic regimen is controversial, because the agent rapidly induces resistance to the non-nucleoside reverse-transcriptase inhibitor (NNRTI) class of antiretroviral drugs. A meta-analysis estimated that 52% of infants who become infected with HIV despite a single dose of nevirapine will develop NNRTI resistance.8 It is probable that nearly all infants who become infected while receiving an extended prophylactic course of nevirapine monotherapy will develop such resistance. Should this mitigate enthusiasm for breastfeeding prophylaxis with extended-dose nevirapine? Probably not. In the study reported today, 54 babies randomised to the control group were found to be HIV-infected at 6 weeks of life, compared with 25 babies who received the SWEN regimen. Assuming that 52% of infants in the control group and 100% of infants in the SWEN group develop resistance, we find that the numbers of babies developing resistance are essentially the same (28 vs 25). Thus the SWEN regimen prevented 29 additional infections at the cost of, at most, three additional NNRTI-resistant infections. Clearly, the induction of any resistance should be avoided, but it seems that extended-dose nevirapine regimens will save many babies without conferring any additional risk of NNRTI resistance to the population beyond that of single-dose nevirapine.
Although the science of perinatal HIV prevention develops briskly, its implementation founders. Globally, we continue to do poorly at providing access to even the simplest paediatric HIV-prevention services. A recent report from WHO, UNICEF, and UNAIDS stated that less than a third of at-risk infants worldwide are currently receiving any prophylaxis at all, and that in some regions, such as west and central Africa, this proportion may be as low as 11%.9 Thus, there is just as much room for benefit in expanding coverage of simple services as there is in implementing more efficacious (and complex) interventions in populations that already have access. Extended infant prophylaxis with nevirapine is simple enough to be implemented almost anywhere. It represents a long-awaited, if partial, solution to a mother's impossible choice. We should not delay.
We declare that we have no conflict of interest.
Extended-dose nevirapine to 6 weeks of age for infants to prevent HIV transmission via breastfeeding in Ethiopia, India, and Uganda: an analysis of three randomised controlled trials
Six Week Extended-Dose Nevirapine (SWEN) Study Team
Summary
Background
UNICEF/WHO recommends that infants born to HIV-infected mothers who do not have access to acceptable, feasible, affordable, sustainable, and safe replacement feeding should be exclusively breastfed for at least 6 months. The aim of three trials in Ethiopia, India, and Uganda was to assess whether daily nevirapine given to breastfed infants through 6 weeks of age can decrease HIV transmission via breastfeeding.
Methods
HIV-infected women breastfeeding their infants were eligible for participation. Participants were randomly assigned to receive either single-dose nevirapine (nevirapine 200 mg to women in labour and nevirapine 2 mg/kg to newborns after birth) or 6 week extended-dose nevirapine (nevirapine 200 mg to women in labour and nevirapine 2 mg/kg to newborn babies after birth plus nevirapine 5 mg daily from days 8-42 for the infant). The randomisation sequences were generated by computer at a central data coordinating centre. The primary endpoint was HIV infection at 6 months of age in infants who were HIV PCR negative at birth. Analyses were by modified intention to treat, excluding infants with missing specimens and those with indeterminate or confirmed HIV infection at birth.
Findings
2024 liveborn infants randomised in the study had at least one specimen tested before 6 months of age (1047 infants in the single-dose group and 977 infants in the extended-dose group). The modified intention-to-treat population included 986 infants in the single-dose group and 901 in the extended-dose group. At 6 months, 87 children in the single-dose group and 62 in the extended-dose group were infected with HIV (relative risk 0·80, 95% CI 0·58-1·10; p=0·16). At 6 weeks of age, 54 children in the single-dose group and 25 in the extended-dose group were HIV positive (0·54, 0·34-0·85; p=0·009). 393 infants in the single-dose group and 346 in the extended-dose group experienced grade 3 or 4 serious adverse events during the study (p=0·54).
Interpretation
Although a 6-week regimen of daily nevirapine might be associated with a reduction in the risk of HIV transmission at 6 weeks of age, the lack of a significant reduction in the primary endpoint-risk of HIV transmission at 6 months-suggests that a longer course of daily infant nevirapine to prevent HIV transmission via breast milk might be more effective where access to affordable and safe replacement feeding is not yet available and where the risks of replacement feeding are high.
Funding
US National Institutes of Health; US National Institute of Allergy and Infectious Diseases; Fogarty International Center.
Introduction
Prevention of mother-to-child transmission of HIV can be achieved through administration of highly active antiretroviral therapy (HAART) to pregnant HIV-infected women.1 In settings where access to HAART is limited, other effective strategies have been developed, including administration of single-dose nevirapine to HIV-infected mothers in labour and to their infants within 72 hours of birth.2,3 However, breastfeeding continues to pose a significant risk of HIV transmission to infants born to HIV-infected mothers, accounting for about 150_000 infant infections per year, mainly in low-income settings.4 HIV-infected mothers in communities where safe and affordable infant feeding alternatives are available are advised not to breastfeed their infants.5 In resource-limited settings, infants of HIV-infected mothers who are not breastfed are at very high risk for mortality and morbidity, which can exceed the risk associated with HIV infection itself.6,7 As a result, UNICEF/WHO recommendations advise that infants born to HIV-infected mothers who do not have access to acceptable, feasible, affordable, sustainable, and safe replacement feeding should be exclusively breastfed for at least 6 months.5 Since more than 80% of all HIV-infected mothers are in sub-Saharan Africa and Asia,4 where access to safe and nutritionally adequate alternatives to breast milk are limited, there is a critical need to identify strategies to prevent HIV transmission by breastfeeding. Provision of antiretroviral medications to infants exposed to HIV by breastfeeding is one potential prevention strategy in these settings. A significant risk for breast milk transmission of HIV occurs within the first 6 weeks post partum.8,9 Therefore, targeting interventions to this early breastfeeding period could be particularly important.
This combined analysis of data from three separate, but coordinated, randomised clinical trials in Ethiopia, India, and Uganda provides an opportunity to determine whether giving low-dose nevirapine to breastfed infants born to HIV-infected mothers in the first 6 weeks of life could prevent more early post-partum HIV transmission than the standard single-dose nevirapine regimen given to mothers and infants alone.
Methods
Participants
Pregnant women who presented to antenatal and delivery facilities in Addis Ababa, Ethiopia, Pune, India, and Kampala, Uganda, who were identified as HIV-infected, were offered the local standard of care for prevention of mother-to-child transmission of HIV and provided infant feeding counselling, consistent with WHO/UNICEF guidelines. Women who indicated an intention to breastfeed their infants and provided informed consent were eligible for study enrolment. Written consent was obtained where possible; for eligible women who could not read, consent was obtained orally and documented in writing by a witness according to procedures that were reviewed and approved by the institutional review boards. These boards-in Addis Ababa, Baltimore, Kampala, and Pune-also all reviewed and approved the country-specific study protocols.
Procedures
The three studies were initially designed to be independent clinical trials. However, during protocol development and implementation, investigators from all three trials met to facilitate a future comparison of study results. Key study procedures, primary outcomes, and treatment regimens were coordinated (table 1) and a single data management and analysis centre was established at Johns Hopkins University (Baltimore, MD, USA) for all three studies.
The Ethiopia and Uganda studies were designed from the outset with the intent to combine data to determine the efficacy of the 6-week extended-dose nevirapine regimen. In February, 2004, after initiation of the studies in Ethiopia and India, the US National Institutes of Health (NIH)-established independent prevention data and safety monitoring board recommended that the Indian data also be combined for a three-country merged analysis of primary study endpoints for efficacy and safety, because of the lower than expected accrual of patients. At the time of this recommendation, none of the countries had accumulated enough events to do an interim analysis for efficacy. All data and safety monitoring board discussions related to merging the data occurred before the data and safety monitoring board saw any interim efficacy (HIV transmission) data.
Screening, eligibility criteria, study interventions, randomisation procedures, and primary endpoints were similar across countries (table 1), facilitating the implementation of the recommendation by the data and safety monitoring board. For all three countries, block randomisation was used with treatment assignments generated by computer at a central data coordinating centre at Johns Hopkins University. The randomisation list was provided to study pharmacists only in each country. At initiation of the study and at scheduled follow-up visits during the first 6 weeks post partum, enrolled mothers and their infants met privately with study pharmacists to receive their assigned study drugs and were provided training by the pharmacists about how to properly administer study drugs to their infants. Other study investigators and clinical staff were not provided access to information about treatment assignments. After completion of study participation and before final analyses, treatment arm assignments were verified by the study statisticians and pharmacist at Johns Hopkins to make sure women/infants received the appropriate treatment to which their identification number had been randomised. In India, all study infants were randomised post partum, while in Uganda, randomisation was pre-partum only. Initially in Ethiopia randomisation occurred pre-partum; however, because randomised women were deciding not to breastfeed after delivery, randomisation was subsequently changed to post partum. Although other differences between the three country protocols were limited, there were differences in local standards of care for prevention of mother-to-child transmission as well as changes in access to HAART during the study. In particular, the local standards of care in Ethiopia and Uganda consisted of single-dose nevirapine, while in India it included the option of maternal zidovudine or other antiretroviral therapy, in addition to single-dose nevirapine. After study initiation, there was expanded access to HAART in all three countries for pregnant women with AIDS or a CD4 cell count of less than 200 cells per μL. In Ethiopia and Uganda, women receiving HAART for treatment were not eligible for study enrolment; however, enrolled women who subsequently required and received HAART were continued in the study. Mothers in India who were receiving antiretroviral therapy, including HAART, were eligible for enrolment.
The study in Uganda included a third intervention arm of HIV immunoglobulin for a separate analysis of efficacy compared with the single-dose nevirapine control arm, which will be reported elsewhere. Therefore, the site in Uganda was simultaneously enrolling and randomising into three arms to allow them to do two separate analyses with the same design and control arm. The original design of the Uganda study involved enrolling 300 pairs of mothers and infants into the single-dose nevirapine control arm, 300 pairs into the extended-dose infant nevirapine arm, and about 200 pairs into the HIV immunoglobulin arm. Therefore, the initial randomisation scheme was done with a three to three to two ratio. Subsequently, when changes were made to merge the data from Uganda, Ethiopia, and India to allow more rapid accrual and a shorter time to obtain an answer to the question of efficacy of extended nevirapine treatment, the required Uganda sample size for the nevirapine arms combined was reduced from the initial 600 planned. This then meant that Uganda would not have a sufficient number of participants randomised to the HIV immunoglobulin arm and control nevirapine arms for the HIV immunoglobulin efficacy analysis by the time they reached the revised, lower nevirapine arm targets for the extended nevirapine efficacy analysis. Therefore, Uganda re-randomised the remaining participants into a more appropriate ratio of five to one to four to allow complete and timely enrolment into all three arms. Therefore, in Uganda, a larger number of participants were randomised to the single-dose nevirapine arm than to the extended-dose arm.
The two study arms for comparison here were identical in all three countries (table 1). Following standard guidelines for use of the HIVNET 012 regimen, the single-dose nevirapine arm consisted of a single 200 mg dose of nevirapine for mothers self-dosed at labour onset and a 2 mg/kg oral dose of nevirapine to newborn babies.2 The extended-dose nevirapine arm consisted of the single-dose regimen plus 5 mg oral nevirapine to infants daily from 8 to 42 days of age. The prophylactic extended dose of nevirapine was chosen to be much lower than the standard dose of nevirapine recommended for the treatment of HIV-infected infants (7 mg/kg twice per day). The daily extended-dose nevirapine regimen was initiated at day 8 because a previous study (HIVNET 006) showed that the level of nevirapine in infants at one week of life after administration of the maternal/infant single-dose nevirapine regimen in labour and after birth maintained the plasma concentration of the drug above 100 ng/mL (ten times the 50% inhibitory concentration of nevirapine) in all infants.10Therefore, it was thought that a regimen of additional nevirapine beginning at day 8 could still be effective and minimise toxicity during the first week of life. To limit the potential exposure of HIV-infected infants to nevirapine alone, in all three countries, study drugs in both arms were discontinued as soon as possible if an infant was found to be HIV PCR positive within 6 weeks of age.
Study products were administered by pre-filled amber oral syringes in Ethiopia and Uganda and by opaque dropper bottles in India, after training of mothers by study pharmacists. This study was not blinded. However, a number of specific study procedures were implemented to introduce some degree of masking of study staff and care givers. Specifically, none of the study investigators or staff, with the exception of the study pharmacists and a member of the central data management team, had access to the randomisation assignments for study participants. Because liquid vitamins and liquid nevirapine have different colours, smells, and tastes, opaque dropper bottles and syringes were used for administration of study drugs. Additionally, infants in both study arms received 1 mL multivitamins from day 8 to day 42 post partum. Using a double-dummy technique, infants received study product in two separate syringes or dropper bottles. For infants randomised to the single-dose nevirapine arm, both syringes or dropper bottles contained multivitamins. For infants randomised to the extended-dose nevirapine arm, one syringe or dropper bottle contained multivitamins and the other contained nevirapine. Study products were only handled by the study pharmacists before dispensing to the care giver, so that study personnel involved in participant assessments would remain masked to the assigned study arm.
The primary endpoint was HIV infection at 6 months of age among liveborn infants who were determined to be HIV PCR negative at birth (ie, in a modified intention-to-treat population). Infants in all three countries with a positive HIV result within 7 days of birth were deemed to have been infected at birth. The median and 90th percentile of day of post-natal testing was 0 (ie, within 24 h) and 1 day, respectively. Mortality and HIV-infection-free survival at 6 months among HIV-uninfected infants at birth were secondary endpoints, as were infant HIV status and mortality among HIV-uninfected infants at birth, at 6 and 14 weeks. For the purposes of this analysis, the visit window was up to 54 days post partum for the 6 week visit, up to 111 days for the 14 week visit, and up to 202 days for the 6 month visit.
Infant HIV infection status was determined by HIV DNA PCR or quantitative HIV-1 RNA PCR testing. In Ethiopia, HIV infection determination was made with a variety of commercially available RNA PCR tests during the study period. An initial positive HIV RNA PCR result was confirmed by a repeat RNA PCR test at a subsequent study visit. In India, HIV infection determination was made with an in-house, externally validated HIV DNA PCR assay developed at the National AIDS Research Institute in Pune. A positive DNA PCR test was confirmed by an HIV-1 RNA PCR test at the next available study visit using the commercially available Roche Amplicor HIV-1 Monitor test, version 1.5 (F Hoffmann-La Roche Ltd, Basel, Switzerland). In Uganda, the HIV infection determination was made with the commercially available Roche Amplicor DNA PCR test and was confirmed by a repeat DNA PCR test at the next available study visit. HIV RNA PCR tests were considered positive if the viral load was over 5000 copies per mL. The choice of this threshold was based on a modification of the Pediatric ACTG definition for infant HIV diagnosis, generally requiring at least 10_000 HIV RNA copies per mL before deemed to be positive.11 HIV-infected infants typically have HIV RNA copy numbers in the hundreds of thousands or millions of copies per mL. At the time of the study no HIV RNA assay had been approved by the US Food and Drug Administration for HIV diagnosis. Although the specificity of these assays improved over the years, there was still concern about the possibility of false positive results. However, there was also concern that viral load might be partly suppressed by prophylactic nevirapine. Therefore, the protocol team decided to use a threshold of 5000 copies per mL instead of 10_000 copies per mL.
For the primary study analyses, an infant was determined to be infected with HIV if two independent HIV PCR tests were positive at different timepoints or if one test was HIV PCR positive and there were no subsequent infant samples available for testing. All HIV PCR testing was done by laboratories in the host country. The reliability of the HIV PCR results were assessed through a centralised quality assurance programme, coordinated by the laboratory of one of the authors (JBJ) at Johns Hopkins School of Medicine, which included blinded retesting of samples from HIV-infected infants and a random sample of HIV-uninfected infants. Final infant infection status was assigned for each country by a blinded study investigator and results from all three countries were subsequently reviewed by a fourth blinded investigator (JBJ) to ensure consistency across the countries.
In all countries, adverse events were detected, documented, and graded based on the NIH Division of AIDS Toxicity Table for Grading Severity of Adverse Experiences, April, 1994. Serious adverse events were reported to the respective country institutional review boards, the institutional review boards of the Johns Hopkins University, and the US NIH, NIAID, Division of AIDS Medical Officer according to their respective guidelines.
Statistical analysis
The data and safety monitoring board, including representative experts from the participating countries, provided ongoing and interval review of efficacy and safety data from this study. In response to a recommendation by the data and safety monitoring board to combine the analysis of key safety and efficacy endpoints for the three studies, a sample size calculation was done with the methods of Nam12 for sample size estimation for stratified fourfold tables, as implemented in NQuery 5.0 (Statistical Solutions, Cork, Ireland). We calculated that 822 evaluable infants would be required in each study arm to achieve 80% power with a type I error of 5%. Inflating for losses to stillbirth, mortality to 6 months, and loss to follow-up (about 16% in total), we arrived at 982 enrolled women in each arm, for a total of 1964 pairs of women and infants across all three countries. This sample size was calculated on the basis of overall assumed transmission risks of 18% in the single-dose nevirapine arm versus 13% (ie, a proportionate 28% reduction) in the extended-dose nevirapine arm at 6 months of age in all infants, whether HIV positive or not at birth (ie, in the whole intention-to-treat population). Although the primary analysis presented here is among those infants who are determined to be HIV negative at birth, we wanted to ensure sufficient power to assess rates of HIV infection at 6 months in the whole intention-to-treat population; we knew the primary outcome, with greater specificity from excluding those infants who were HIV positive at birth, would have even greater power. We have also reported some results of country-specific analyses; they are underpowered, and given mainly for descriptive purposes.
For completeness of reporting, we note that at the October, 2004, meeting of the data and safety monitoring board, partly unblinded data (coded as groups X and Y) on overall 6-month transmissions (70 events) were made available to the board as part of the ongoing safety review of the study. The board made no suggestion that the data be reviewed with the possibility of early stopping for efficacy, since it was still early in the accrual of patients for these trials. By use of a modified implementation of the O'Brien-Fleming design,13 we did one formal interim analysis in January, 2006, when 105 transmissions up to 6 months of age had occurred among those infants who tested HIV-negative at birth. The size of the test was 0·0045, leaving a final test size of 0·0494 for a total of 149 such transmissions. The test did not meet the statistical criterion, and no changes were recommended at that time.
Infant HIV testing was done at 6 weeks, 14 weeks, and 6 months in all three countries. Therefore, these time points were selected for combined analyses. The three countries' product-limit estimates of the analyses endpoints were combined with inverse-variance weights. For each country, a risk ratio for cumulative transmission was calculated; the overall risk ratio was an inverse variance-weighted mean of the separate countries' risk ratios, where the weighting took place on the logarithmic scale.13 The Kaplan-Meier analyses of time-to-first-HIV detection censored infants at the earliest of documented official study termination, last HIV determination, or death. Similar analyses were done to examine mortality as an outcome, and the earliest of HIV infection or death (the complement of infection-free survival) as an outcome. Cox regression analyses, stratified by country, were fitted that adjusted for antenatal antiretroviral use and receipt of the maternal nevirapine dose. Cox models were also used to test for varying treatment efficacy (interaction) across countries. The numbers of infants with serious adverse events were compared across treatment arms by comparing risk ratios. Adjustment for study country was done in a log-binomial regression model, with two dummy variables to account for the three countries.
Stata versions 8, 9, and 10 were used for analyses.
Role of the funding source
Medical officers from the NIH Division of AIDS contributed to the initial study design and early protocol development, as well as review of adverse events reports. Sheryl Zwerski (acting director of the Prevention Sciences Program, Division of AIDS) contributed to the preparation of this manuscript. The NIH had no other role in data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication, which was based on a majority vote of the members of the SWEN Study Writing Team.
Results
Figure 1 shows the trial profile. 2024 liveborn infants with at least one specimen tested before 6 months of age were randomised (1047 infants in the single-dose group and 977 infants in the extended-dose group). 137 infants were excluded from the modified intention-to-treat population: 36 did not have birth specimens available for testing, eight were of indeterminate HIV infection status at birth or 6 months, and 93 were found to be infected at birth. Inverse variance-weighted HIV infection rates at birth were 4·1% in the single-dose group and 4·7% in the extended-dose group (p=0·157). Country-specific birth infection rates were similar between groups in Ethiopia (4·2% in the single-dose group vs 4·5% in the extended-dose group; p=0·83) and in India (3·8% vs 3·6%; p=0·88), but significantly different in Uganda (4·3% vs 9·6%; p=0·02).
The modified intention-to-treat population thus consisted of 986 infants in the single-dose group and 901 infants in the extended-dose group. Among these infants, endpoint status information was available at 6 months of age for 928 (94·1%) infants in the single-dose group and 831 (92·2%) in the extended-dose group. There were 30 sets of twins; in no set did both twins die or both become infected with HIV; there was thus no within-set correlation to account for in the corresponding analyses.
In the modified intention-to-treat population, maternal characteristics at enrolment and delivery, including the proportion of caesarean deliveries, maternal CD4 cell counts, and plasma viral load at enrolment, were similar in the two study groups (table 2). Some women in India, who presented in the late stages of labour and were not eligible for pre-delivery HIV screening and administration of single-dose nevirapine, were eligible for post-partum HIV screening and enrolment. Additionally, some women in India were prescribed prenatal zidovudine by their personal physicians, according to local standards of care. Access to affordable HAART was limited in all countries during this study and most enrolled pregnant and post-partum women did not meet local clinical guidelines for initiation of antiretroviral therapy. However, 30 women with infants in the single-dose group and 32 women with infants in the extended-dose group did receive HAART. With the exception of three enrolled women with infants subsequently randomised to the single-dose group and nine women with infants subsequently randomised to the extended-dose group who initiated maternal HAART before delivery in India, all other enrolled women who received HAART before 6 months in Ethiopia, India, and Uganda initiated antiretroviral therapy at some point after delivery.
Likewise, the characteristics of infants in the modified intention-to-treat population-in particular gestational age, duration of breastfeeding, and duration of exclusive breastfeeding-were similar across the two study groups (table 2). The proportion of male infants was higher in the single-dose group than in the extended-dose group in each country (p=0·02). Infants in the single-dose group received more doses of study product (ie, multivitamins) than did infants in the extended-dose group (p=0·03). However, this reflected a difference found only in India, where the method of determining adherence to study drug dosing differed from Ethiopia and Uganda. 6-month mortality or HIV status was unknown for 58 (5·9%) infants in the single-dose group and 70 (7·8%) in the extended-dose group infants; there were no significant differences in baseline characteristics for these children (data not shown).
Although fewer children in the extended-dose group than in the single-dose group were positive for HIV at 6 months, the difference in transmission rate was not statistically significant (table 3). By contrast, the rate of transmission in the extended-dose group was significantly lower than that in the single-dose group at 6 weeks (table 3).
Table 3. Risk of HIV transmission and death among infants at 6 weeks and 6 months by study regimen and country (modified intention-to-treat population) __
Data are risk (number of events) or risk ratio (95% CI).
Combined mortality was lower in the extended-dose group than in the single-dose group at 6 weeks of age, but not significantly so (table 3); however, by 6 months of age, the mortality rate was significantly lower in the extended-dose group than in the single-dose group (table 3). 16 infants in the extended-dose group and 37 in the single-dose group had died by 6 months of age. 15 (28%) of the 53 deaths occurred among infants known to be HIV infected (four in the extended-dose group and 11 in the single-dose group). Of the 38 HIV-uninfected infants who died, there were fewer deaths in the extended-dose group than in the single-dose group (12 [1·4%] vs 26 [2·9%]). The (non-mutually exclusive) reasons for death included respiratory conditions (n=22), gastrointestinal conditions (14), neonatal conditions (five), infections (15), congenital anomalies (two), malnutrition and poor growth (five), and other causes (five). Respiratory conditions contributed to a lower proportion of overall mortality in the extended-dose group (37·5%, six deaths) than in the single-dose group (43·2%, 16 deaths). Similarly, gastrointestinal conditions contributed to a lower proportion of overall mortality in the extended-dose group (12·5%, two deaths) than in the single-dose group (32·4%, 12 deaths). All other conditions were responsible for a similar proportion of deaths (data not shown).
There were significantly fewer instances of post-natal HIV transmission or death in the extended-dose group than in the single-dose group at both week 6 and after 6 months (table 3). The lower risk of HIV infection or death among infants in the extended-dose group was also seen when analysed by country, although this difference was only statistically significant in India (table 3). A lower cumulative probability of death or HIV transmission was seen in the extended-dose group than in the single-dose group at each of three time points assessed (6 weeks, 14 weeks, and 6 months of age; figure 2).
Figure 2. Risk of infant HIV infection or death _
At each time point (6 weeks, 14 weeks, and 6 month of age), for each study arm, the three countries' product-limit estimates were combined with inverse-variance weights. Events were recorded at the time they occurred, but the cumulative Kaplan-Meier estimates at the specific points 6 weeks, 14 weeks, and 6 months are used in these calculations.
The hazard ratios obtained from multivariate Cox regression analyses, including adjustments for maternal receipt of nevirapine and use of other antiretrovirals for prevention of mother-to-child transmission in India, were qualitatively similar to the inverse variance-weighted risk ratios obtained from cumulative Kaplan-Meier estimates. Specifically, the 6-month hazard ratio estimates for the extended-dose group were 0·78 (95% CI 0·57-1·09; p=0·15) for transmission, 0·43 (0·24-0·79; p=0·006) for death, and 0·71 (0·53-0·95; p=0·02) for transmission or death. We found no evidence of interaction by country (Cox regression transmission model 2 df interaction test p=0·88). Country-specific Kaplan-Meier estimates of HIV infection, as well as HIV infection or death, are shown in figure 3.
Secondary analyses in the full intention-to-treat population (ie, including infants at birth found to be HIV positive, with indeterminate infection status, or lacking birth specimens for testing) showed that the risks of transmission, death, or death or transmission were reduced in the extended-dose group compared with in the single-dose group at both 6 weeks and 6 months, although not significantly so (table 4).
Grade 3 and 4 serious adverse events were common. Overall, 1150 serious adverse events were reported in the first 6 months of life in 739 of the 1887 infants in the modified intention-to-treat population (table 5). However, the frequencies of specific serious adverse events were similar between groups. 521 (45%) of the serious adverse events were categorised as laboratory abnormalities. The next most frequent events were gastrointestinal conditions (n=184) and respiratory illnesses (n=158). The proportion of infants who experienced any grade 3 or 4 event in the first 6 months of life was much the same between both treatment groups; fewer infants in India had events than did those in either Ethiopia or Uganda (table 5).
Serious rashes and abnormal aminotransferase concentrations were rare events and the proportion of infants with rash, abnormal aminotransferase concentrations, and neutropenia were similar in both groups (table 5). A higher proportion of Ethiopian infants had grade 3 and 4 neutropenia than did those in the other two countries. However, episodes of neutropenia were equally distributed in both study groups, suggesting they were not specifically related to the extended dosing of nevirapine and could reflect ethnic variations. In multivariate log binomial regression models adjusting for study country, the risk of having adverse events was not significantly different between the two groups (risk ratio 0·93, 95% CI 0·85-1·09; p=0·17).
Discussion
The results of this study indicate that, among breastfeeding infants who were HIV uninfected at birth, extended use of oral nevirapine 5 mg/day did not significantly reduce the risk of infection with HIV at 6 months compared with single-dose nevirapine. This lack of significance could have been due to a loss of statistical power from underlying HIV transmission rates that were lower than those used for our sample size calculation, particularly in the control group (estimated rate of 18·0% vs measured rate of 12·8%). By contrast, the risk of infection with HIV was reduced at 6 weeks of age in those infants who received extended-dose nevirapine, although this was not the primary endpoint of the study.
The extended-dose regimen significantly reduced the risk of overall infant death at both 6 weeks and 6 months of age, compared with the single-dose regimen. The reported causes of death among infants were similar by treatment arm, as well as among HIV-infected and HIV-uninfected infants. The reasons for this reduction in mortality are not clear. Delay of HIV infection to an older age among infants receiving the extended-dose regimen might have had a role, since HIV-positive newborn babies have a higher overall mortality than do infants older than 6 weeks of age. However, this would not explain the entire effect, since there were also fewer deaths among HIV-uninfected infants receiving the extended-dose regimen than among those receiving single-dose nevirapine. A number of the infant deaths could have occurred among infants who became HIV-infected after their last study visit (ie, after their last assessment of HIV status). Since many infant deaths occurred at home or outside of the study clinics, information about the cause of these deaths was limited to verbal autopsy interviews of mothers and other family members. Thus, HIV status at the time of death was not available for infants who could have become HIV positive between their last study visit and their death. 38 (72%) of the deaths occurred among infants for whom their last HIV screening test was negative by HIV PCR. However, the mean time from this last study visit until death was 13·8 (SD 17·7) days, decreasing the likelihood that we failed to identify HIV-infected infants before their death. We assessed the distribution of mortality by sex, because of an imbalance between study groups at baseline, but there was no significant difference between males and females in terms of the distribution of overall mortality (data not shown). We found no differences between groups in other factors that might be associated with infant mortality, including maternal age, maternal CD4 cell count, maternal viral load (available in India and Uganda), maternal education level, parity, gravidity, and gestational age at enrolment (data not shown). Other unavailable or unmeasured factors could have been different in the two arms that could explain the difference in mortality, such as differences in maternal feeding practices that might affect infant mortality. Overall infant mortality was lower than might have been expected for infants born to HIV-infected mothers generally in the participating communities, possibly because infants who were HIV-infected at birth were excluded from the analysis set and thus did not contribute to these mortality figures. Additionally, participating infants had access to close medical supervision by study staff in the event of illness. Further studies will be necessary to understand the reasons for the infant mortality benefit seen here.
Although country-specific sample sizes were not powered to show statistically significant differences by treatment group, there were fewer instances of HIV transmission in the extended-dose groups than in the single-dose groups in each country, particularly at 6 weeks. HIV transmission rates in the control groups varied among the three study populations, with Uganda having the highest transmission rates of the three countries. One should note that overall rates of HIV transmission seen among participating infants may not be generalisable to their local communities. For instance, we may have enrolled mothers with a risk of HIV transmission lower than that seen in HIV-infected mothers in the general communities. As noted in figure 1, a large number of HIV-infected mothers identified in the screening process in the three countries were not enrolled. A frequent reason for non-enrolment in India and Ethiopia was because a mother expressed her intent not to breastfeed her infant. These mothers would be expected to have an overall transmission rate that is less than that in enrolled breastfeeding mothers. Country-specific differences in HIV transmission levels could also be related to unmeasured maternal, infant, or virological factors (eg, viral load, clinical stage, mixed feeding, viral subtype) in the study populations. Differences in duration of breastfeeding could also have contributed to country-specific differences in overall HIV transmission, as well as country-specific differences in HIV transmission by treatment group. As with all maternal and infant factors assessed (except infant sex and study product dosing), the proportion of mothers receiving additional antiretroviral therapy for either treatment or prevention of mother-to-child transmission was equally distributed between the two arms of the study. Therefore, while these factors suggest that enrolled mothers might not be completely representative of all HIV-infected mothers in these communities, they are unlikely to explain the observed differences in efficacy between the two treatment arms.
Previous studies have demonstrated that single-dose nevirapine for prevention of mother-to-child transmission is safe for use in HIV-exposed infants.14-17 Extended-dose nevirapine for breastfed infants born to HIV-infected mothers seems to be equally well tolerated. Nevirapine use has been associated with a number of toxicities, including rash, liver toxicity, and haematological abnormalities.18 Such adverse events were reported at similar rates in both the single-dose and extended-dose groups. Some differences in the occurrence of adverse events were evident across countries, including a greater frequency of neutropenia in infants in both treatment arms in Ethiopia. However, the fact that the extended-dose nevirapine regimen was not associated with a greater risk of adverse events than was the single-dose regimen overall or within countries is reassuring.
The single-dose nevirapine regimen has been associated with an increased risk of selection of resistance to the drug in mothers and HIV-infected infants,19-21 which could decrease the effectiveness of subsequent treatment with HAART containing other non-nucleoside reverse transcriptase inhibitors.22,23 The clinical consequences for infants found to have evidence of early and often transient post-partum nevirapine-resistant virus remains unclear. Studies of the patterns of viral resistance among mothers and infants enrolled in our study are ongoing. Infants who receive single-dose nevirapine show signs of resistance to the drug, and evidence suggests that levels of resistance are even higher in those who received the extended dose regimen (unpublished data). With the recent results of the CHER study24 encouraging early initiation of HAART in HIV-infected infants, alternative regimens of HAART are likely to be needed in infected infants who receive extended-dose nevirapine. However, any risk of nevirapine resistance in HIV-infected infants needs to be considered in the context of the reduction in the risk of HIV transmission and the survival benefit seen among HIV-uninfected infants receiving extended-dose nevirapine.
Few other studies have compared extended infant prophylactic regimens to single-dose nevirapine or other regimens. In the multicentre PETRA study, infant HIV-1 transmission at 6 weeks of age was 5·7% with zidovudine plus lamivudine starting at 36 weeks' gestation, plus intrapartum and 7 days post-partum dosing of mothers and infants, compared with 8·9% with the same regimen without the pre-partum maternal dosing, 14·2% with intrapartum zidovudine and lamivudine only, and 15·3% for a placebo group.25 A study from Malawi in which infants whose mothers had not received any prophylaxis were randomly assigned to receive either single-dose nevirapine alone or single-dose nevirapine plus 1 week of zidovudine showed that, at 6-8 weeks of age, among infants who had been HIV negative at birth, the combination regimen was associated with a 36% reduction in transmission.26 However, a subsequent study done among infants whose mothers received single-dose nevirapine found similar transmission rates at 6-8 weeks among infants who had received single-dose nevirapine and in those who received the combination regimen.27 In the SAINT study, mother-infant pairs were randomly assigned to receive either one or two doses of nevirapine or multiple-dose zidovudine and lamivudine; mothers and infants in the second group received prophylaxis for 1 week post partum. There were no significant differences in the estimated transmission rates at 8 weeks by regimen.28 A recent open-label, non-randomised cohort study in which infants were given prophylactic lamivudine during breastfeeding reported a rate of HIV infection of 3·8% at 6 weeks of age and of 4·9% at 6 months of age among breastfed infants born to HIV-infected mothers in Tanzania.29 Another open label, non-randomised observational study recently reported post-partum HIV transmission to only one of 238 breastfed infants born to HIV-infected mothers on HAART in Rwanda.30 None of these studies were randomised trials specifically designed to compare the efficacy of interventions to prevent breast milk transmission of HIV. One such study-the PEPI study in Malawi31-showed a reduction in postnatal HIV transmission at 14 weeks and 9 months of age with the daily dosing of nevirapine to breastfed infants for 14 weeks, compared with single-dose nevirapine, generally consistent with our findings.
A number of limitations to our study exist. Although, before initiation, the three study country teams coordinated key procedures and the three countries used the same data and safety monitoring board, data coordination centre, and pharmacy centre, the studies in the three countries were originally designed to be independently powered. The lower than expected overall transmission rates and subsequent loss of statistical power to detect a significant reduction in risk of HIV infection at 6 months of age has already been mentioned. The unbalanced rate of HIV infection at birth in the two arms in Uganda cannot be explained by any identifiable differences in the two arms to date, and might have occurred by chance. However, one cannot rule out the possibility that the cause of this imbalance might not have been uncovered in our analyses. This imbalance does not affect our efficacy assessment because infants who were positive at birth (ie, before the start of the intervention) were not included in the modified intention-to-treat population. However, in addition to the expected reduction of statistical power resultant from inclusion of HIV-infected infants at birth, the unbalanced birth HIV infection rate probably contributed to the loss of benefit of the extended-dose regimen seen in the secondary analysis in the full intention-to-treat population.
Our study also had a number of strengths. First, the study was randomised, controlled, and partly masked, with a large sample size of nearly 2000 mother-infant pairs. Second, maternal baseline covariates including CD4 cell count and viral load-strong predictors of mother to child transmission at baseline-were well balanced between the two arms. Third, rates of infant loss to follow-up were low and HIV infection and mortality status were determined for a very high proportion of infants. Fourth, initial breastfeeding rates were high and not significantly different between the two arms overall. Lastly, although not statistically significant in each country individually, the differences in mortality and transmission at 6 weeks seen with extended-dose nevirapine compared with the single-dose regimen were consistent across countries, giving confidence in the efficacy of the intervention in different geographic locations with different HIV subtypes and breastfeeding practices.
These results, together with those of the PEPI study,31 suggest that, in communities where access to affordable and safe replacement feeding is not yet available and where the risks of replacement feeding are high, and where access to HAART for HIV-positive mothers is limited, extended post-partum nevirapine should be considered for HIV-uninfected infants being breastfed by HIV-infected mothers to improve survival and to reduce the risk of HIV transmission. Such regimens seem to be as safe-in terms of adverse events-as and more effective than the current single-dose nevirapine regimen for prevention of post-partum HIV transmission via breast milk in the first 6 weeks of life. The optimum duration of post-partum infant nevirapine is, however, unclear, and is probably dependent on the duration of breastfeeding and other maternal and infant factors that vary among communities at risk for transmission of HIV via breast milk. The HPTN 046 study, currently underway in multiple countries, is comparing the 6-week extended-dose nevirapine regimen with a 6-month extended-dose regimen for breastfed infants.32 Furthermore, additional interventions to optimise the safety of breastfeeding should be assessed, particularly in light of recent WHO recommendations that infants born to HIV-infected mothers in resource-limited settings who opt to breastfeed should exclusively breastfeed for at least 6 months post-partum before assessing whether they can safely stop breastfeeding at 6 months or should continue breastfeeding with introduction of other foods.5
|
|
| |
| |
|
|
|
