MATERIAL AND METHODS

Study setting and design of the Ditrame project, enrolment, treatment allocation, follow up procedures, and treatments tested in the Agence Nationale de Recherches sur le Sida (ANRS) 049b trial have been already described elsewhere.^18,19 The enrolment was performed in neighbourhood maternal/infant care units in the cities of Abidjan, Côte d’Ivoire, and Bobo-Dioulasso, Burkina Faso, from September 1996 to May 1997. The study protocol was approved by the ethics committees in Burkina Faso, Côte d’Ivoire, and France and by the scientific committee of the ANRS. In brief, women aged 18 years or more with confirmed HIV-1 positive serostatus who returned for post-test counselling were offered enrolment in the trial. Written consent was obtained individually, after the study objectives and content had been explained by a specially trained counsellor, in French or in a vernacular language.

Women self administered one daily vaginal capsule of either BC (1% concentration) or placebo from 36 weeks of pregnancy until labour. Another vaginal capsule (BC or placebo) was administered at the beginning of the delivery process in the maternity ward under supervision of the study team. Finally, the neonate was bathed with a BC solution (1% concentration) or placebo within 30 minutes after birth, in the delivery room. The trial drugs, vaginal capsules, and bath powder were provided by Innothera Laboratory (Paris, France).

Women were followed weekly for tolerance and compliance visits from enrolment through delivery.¹⁸ After delivery, women and children were seen each month until 15 months after delivery/birth, and medical care was provided free of charge. Special attention was paid to counselling on infant feeding practices to reduce duration of breast feeding when possible.²⁰

Study outcomes

The outcome measures in the present analysis were HIV-1 infection and child mortality through 15 months. The diagnosis of paediatric HIV-1 infection was made on the basis of one positive polymerase chain reaction (PCR) test, according to international standards.²¹ All samples collected at 6 months, or earlier when not available, were systematically processed by a DNA (Abidjan) or RNA (Bobo-Dioulasso) PCR test.¹⁹ PCR was then applied to all the preceding available samples for a given child if the first one tested was positive, and the timing of transmission was estimated on the basis of the timing of the first positive test.²¹ Absence of infection was defined by a negative PCR on the latest available sample beyond the age of 3 months or by a negative serological test between 9 and 15 months of age.²² Those children who had no sample available for PCR and could not be followed beyond 6 months of age were considered of definitive unknown HIV status.

Statistical analysis

We chose the intent to treat approach for analysis. Group comparisons used Student’s t test or non-parametric Mann-Whitney test for quantitative variables, χ² test, or Fisher’s exact test for qualitative variables when appropriate. The probability of infection at a given age, breast feeding, and infant death were estimated using the Kaplan-Meier survival technique and comparisons used the log rank test. For the analysis of risk factors of MTCT HIV transmission, we excluded the second child in each pair of liveborn twins. We used as explanatory variables sociodemographic characteristics of the women (age, gestation, and parity) and CD4 lymphocyte count collected at enrolment, duration of labour and rupture of membranes, birth weight and gestational age at birth. All these data were used in a Cox multivariate proportional hazards model to study BC relative efficacy, expressed as one minus the hazard ratio (HR) of the transmission rate in the BC arm versus the transmission rate in the placebo arm. Point estimates of summary statistics are reported with their 95% confidence interval (CI). The same approach was also used to look for determinants of infant mortality.

RESULTS

Of the total of 107 HIV infected women enrolled, 53 were allocated to the BC group and 54 to the placebo group. Maternal age (mean 24.6 years), parity (mean 1.8), education level, marital status, clinical characteristics, and distribution of CD4 lymphocyte counts (mean 589 10⁶/l) at baseline were balanced between the two groups. In both groups, the mean duration of antenatal treatment was 3 weeks and 97% of the mothers delivered in the hospital maternity unit.

A total of 111 children were born, including four sets of twins (two in each treatment group). There were 107 livebirths and four stillbirths (one in the BC group and three in the placebo group, all singletons). Stillbirths were excluded from the transmission study as well as the second twins. Thus, 103 children were eligible for this analysis. Mean birth weight was 2845 g, the percentage of low birth weight (<2500 g) was 20%, and prematurity rate (<37 weeks of pregnancy) was 7.2% without differences between the two groups. The median follow up of the children was 450 days in both groups. The median duration of breast feeding was 324 days without differences between the two groups.

Five children remained of indeterminate status for HIV infection (two died at day 2 and day 57, respectively and three were lost to follow up at day 1, day 7, and day 34, respectively, before any interpretable blood sample was available). Three of these were in the BC group and two in the placebo group. At 15 months of follow up, 11 children were diagnosed as HIV infected in the BC group and 12 in the placebo group (table 1). The probability of transmission of HIV was 20.4% and 21.0% at 6 months of age in placebo and BC group respectively and 24.8% and 23.5% at 15 months of age. There was no significant difference between the two treatment groups (log rank test, p=0.84). There was no difference between the transmission rates in the two trial sites (data not shown). The transmission rate differed significantly according to maternal immune status (40.5% among mothers with CD4 lymphocytes counts below 500 × 10⁶/l, versus 10.7% among those with counts of 500 × 10⁶/l or more, p=0.001).

In the multivariate analysis of risk factors for HIV transmission, none of the factors studied was significantly associated with the infection status of the infants: trial arm, study site, birth weight, gestational age, maternal age, number of previous pregnancies, maternal CD4 lymphocyte count, duration of membrane rupture. For maternal CD4 lymphocyte counts, the hazard ratio for HIV transmission was 0.88 for each increase of 100 CD4 × 10⁶/l (95% CI: 0.729 to 1.068; p=0.20).

Overall mortality at 15 months was 10 of 55 infants in BC group and six of 56 in the placebo group, in addition to two and three stillbirths, respectively (table 2). The difference in mortality over the study period was not significantly different between the two treatment arms (log rank test, p= 0.31). Perinatal mortality (stillbirths and early neonatal deaths in the first week of life) was 5.5% in the BC group and 5.4% in the placebo group, (p=0.49). Twelve of the 17 children who died during the first 15 months of follow up were HIV infected; the causes of death were lower respiratory tract infections (six), meningitis or encephalitis (three), diarrhoea (two), and unknown cause (one). There were three deaths among uninfected infants, one neonatal death related to intrapartum asphyxia and two cases of meningitis. Two children died with indeterminate HIV infection status, one of intrapartum asphyxia at day 2 and the other of lower respiratory infection at day 57. HIV free survival probability at 15 months (1 − the risk of HIV infection or death) was 77.6% in the placebo group versus 74.9% in the BC group (log rank test p=0.79).

In the Cox model for risk factors for infant mortality, the only factor associated with the risk of death by 15 months was HIV infection of the child (hazard ratio: 16.7, 95% CI: 4.8 to 58.4). Age of the mother, number of previous pregnancies, maternal CD4 lymphocyte count, birth weight, low gestational age, duration of rupture of membranes, site of the study, and trial arm were not associated with the risk of child death.

DISCUSSION

In this randomised clinical trial, we failed to demonstrate a difference in mother to child HIV transmission after vaginal cleansing and neonatal bathing with BC, compared to placebo. However, this trial was designed as a phase II, and therefore the sample size was too small to rule out a clinically significant treatment difference between the two groups. None the less, the transmission rate in both groups was of the same order of magnitude as that observed in the placebo arm of the Ditrame/ANRS 049a trial, which was conducted in the same study sites.^19,23 In this study, which evaluated the efficacy of short course zidovudine, the transmission rates in the absence of zidovudine were 27.5% at 6 months of age and 30.6% at 12 months. To date, there are no clinical data suggesting the efficacy of vaginal microbicide use in decreasing mother to child HIV transmission. In large clinical trials evaluating vaginal disinfection with chlorhexidine in Malawi and Kenya,^11,13 no reduction in transmission was observed. However, the interventions were performed in advanced labour, possibly too late to inactivate the virus and eliminate it from the birth canal. Although the rationale for our intervention was to begin the use of the vaginal microbicide well before the onset of labour or membrane rupture, we did not find evidence of any benefit in reducing transmission.

Furthermore, in our study, BC did not appear to reduce the rate of infant mortality, including during the perinatal period. The distressingly high mortality we observed was in the same range as that reported in the infant feeding trial from Nairobi²⁴ and the causes of death were the same as in other African cohorts of children born to HIV infected women.^25,26 As mentioned above, the statistical power of our study may not have been sufficient to detect a benefit in terms of mortality, but there was not even a trend suggesting that the intervention may be useful. Furthermore, the majority of deaths occurred in HIV infected children, beyond the neonatal period. Finally, the babies in the placebo group were bathed at birth with water containing no BC, which was not done in previous studies. The hypothesis that bathing, even without adding an antiseptic, may have a beneficial effect on neonatal infectious morbidity has not been evaluated, but could be speculated.

Although our study failed to demonstrate any beneficial effect of vaginal cleansing with BC, we believe that further research on vaginal microbicides in pregnancy is warranted. In this trial, we previously demonstrated that BC, used in vaginal capsules for a median of 3 weeks in late pregnancy and in a neonatal bath, was well tolerated by HIV infected pregnant women and their neonates.¹⁷ The feasibility and acceptability were good in both west African cities. Antiseptics such as benzalkonium chloride, as well as new microbicides presently in preliminary research phases, have the potential to decrease perinatal, neonatal, and maternal morbidity and mortality.¹⁰ In the Malawi study, chlorhexidine use was associated with significant decreases in postpartum fever in mothers and perinatal mortality in children.¹³ Further clinical trials may be considered to evaluate the efficacy of other vaginal microbicides in reducing perinatal infectious morbidity, beyond the specific issue of mother to child HIV transmission and the crude outcome of infant mortality. Any such intervention should be evaluated in addition to prevention of MTCT using antiretroviral drugs, which have proved efficacy in decreasing mother to child HIV transmission, and should therefore be considered today as the standard of care.³ The place for vaginal microbicides may be as part of a public health package, including STD management, malaria, and anaemia prevention in rural Africa, where access to HIV testing and antiretroviral chemoprophylaxis is limited and is likely to remain difficult to implement for many years.

Appendix

The ANRS 049 Group is organised as follows: Biostatistics: L Dequa-Merchadoux³, R Lassalle³, V Leroy³, R Salamon³. Epidemiology: M Cartoux², F Dabis³ (coordinator of the ANRS 049 trial/Ditrame programme), N Meda² (coordinator of Bobo-Dioulasso Center), P Msellati¹ (coordinator of Abidjan Center), R Ramon¹. Gynaecology-Obstetrics: A Bazié², B Dao², R Likikouet¹, L Mandelbrot⁴ (principal investigator), C Welffens-Ekra¹ (principal investigator). Microbiology: D Bonard¹, P Combe¹, M Dosso¹, L Gautier-Charpentier², FD Ky², A Ouangré², T Ouassa¹, O Sanou², F Sylla-Koko¹, Y Traore², P Van de Perre². Molecular biology: AM Cassel-Beraud², JB Kottan¹, O Manigart², C Montcho¹, C Rouzioux⁴, A Simonon², D Valea², B You¹. Paediatrics: R Camara¹, N Elenga¹, B Nacro², F Tall², M Timité¹. Trial monitoring: G Gourvellec¹, O Ky-Zerbo², V Noba¹, I Sombié², S Tiendrebeogo², I Viho¹, S Yaro².

Abidjan health facilities

Anonkoua-Koute, Ouassakara, Yopougon and Yopougon-Attie, Centre Hospitalier Universitaire de Yopougon.

Bobo-Dioulasso health facilities

Accart-Ville, Farakan and Social Security, Centre Hospitalier National Souro Sanou.

1 Abidjan Centre, 2 Bobo-Dioulasso Centre, 3 Bordeaux Coordination Unit, 4 Paris.