Coffie, Patrick A MD, PhD; Kanhon, Serge K MD; Touré, Hapsatou MD, MPH; Ettiegne-Traoré, Virginie MD; Stringer, Elizabeth MD; Stringer, Jeff S MD; Dabis, François MD, PhD; Ekouevi, Didier K MD, PhD
In 2001, during a special session on HIV/AIDS, the United Nations General Assembly set the ambitious goal of reducing the proportion of infants infected with HIV by 50% by 2010 by ensuring that 80% of pregnant women accessing antenatal care (ANC) would receive information, counseling, and access effective antiretroviral prophylaxis for the prevention of mother-to-child transmission (PMTCT) of HIV-1.1
Over the past 2 decades, considerable progress has been made in the scaling up of PMTCT services. Despite the increasing availability of antiretroviral drugs for PMTCT and large-scale investments for implementation of these services, pediatric AIDS remains a largely uncontrolled epidemic. In 2009, approximately 370,000 children became newly infected in the world, and 91% of these contaminations occurred in sub-Saharan Africa through mother-to-child transmission.2 The achievement of the United Nation's goal is clearly limited by many factors, including underdeveloped health care infrastructures and the poor availability of trained providers.
In Africa, the coverage of antiretrovirals for PMTCT was estimated to be 54% in 2009.3 There is however a lack of consensus on the method to use to document PMTCT coverage and thus effectiveness4 and a standardized methodology is needed to optimize services evaluation, identify gaps and possible solutions. Two methods have been proposed so far. First, a commonly described metric for program effectiveness is program coverage, that is, the proportion of women in a population that receive a PMTCT intervention according to health facility reporting.5 The second method is the direct measurement of the drugs in cord blood samples. Among all of the drugs used for PMTCT, nevirapine (NVP) is a suitable candidate for this kind of evaluation as it is a component of all the standard PMTCT regimens6 and can be detected in the cord blood within minutes of its absorption.7
Our objective was to estimate the coverage of NVP use for PMTCT among HIV-infected women in Côte d'Ivoire using these 2 methods and identifying factors that could help improve the coverage.
The PMTCT Effectiveness in Africa: Research and Linkages to Care study is a multicountry (Cameroon, Côte d'Ivoire, South Africa, and Zambia) evaluation of the effectiveness of services for the PMTCT at both the community and facility level.8 In this report, we provide in-depth results of the Côte d'Ivoire component of this evaluation. This protocol was approved by the national ethics committee of the Ministry of Health from Côte d'Ivoire; by the US Centers for Diseases Control and Prevention, Atlanta; and by the University of Alabama at Birmingham.
Population and Sampling
This cross-sectional survey of mother-infant pairs was conducted in the country between November 2007 and September 2008 in randomly selected facilities providing delivery services. The sample size calculation (8000 samples) was based on the expected country estimate of the HIV prevalence with a given precision and was reported in details elsewhere.8
We obtained the list of all delivery facilities that were known to be providing PMTCT services from the Ministry of Health. We limited the sampling to the coastal provinces due to civil unrest in the northern part of the country. Among the 144 selected PMTCT sites available in 2007, 6 were randomly selected first. Then, 4 additional PMTCT facilities located around one site in the Koumassi district were added because the target sample of HIV-exposed neonates could not be reached over a 1-year period. Therefore, the study was undertaken in 10 PMTCT facilities in total. They were located in Abidjan, the economic capital of the country,6 Bouaké (the country's second largest city), Grand-Lahou, Sassandra, and San Pedro (1 each).
Each selected site had an established PMTCT program. The antiretroviral prophylaxis used varied among sites, but maternal and infant administration of NVP was a standard component of all of the drug regimens.
All pregnant women who were admitted for delivery in one of the selected facilities during the study period were included.
At each facility, midwives anonymously collected umbilical cord blood specimens from every live-born delivery that occurred during the study period. Approximately, 2 tubes of 7 ml of whole blood were collected after delivery from the umbilical cord and placed in a serum-separating tube affixed with a unique identification number. The cord blood samples were tested on site for HIV antibodies using a rapid test (Determine, Abbott Laboratories, Chicago, IL) and confirmed by Genie II HIV/HIV-2 (Biorad, Marnes-La-Coquette, France). The seropositive samples were transferred by pipette to filter paper. The resulting dried blood spots were then shipped to the University of Cape Town (Cape Town, South Africa) for NVP testing. The detection of NVP in the dried blood spot was performed using minor modifications of the method described previously.8,9
We also recorded nonidentifying information from maternal antenatal records onto anonymous standardized forms and linked these forms to the cord blood sample using a unique identification number. The data collected from each maternal antenatal record were the followings: age, gravidity, number of ANC visits, location of the first ANC visit, proposal and acceptance of HIV testing, mode of delivery, and PMTCT regimen (if she received prophylaxis).
At the time of discharge, the clinical staff recorded on the surveillance form whether the infant was given NVP prophylaxis (direct observation). Babies who were born before the arrival at the delivery ward and those who were stillborn were excluded from this analysis.
We estimated NVP coverage among HIV-infected women and/or HIV-exposed infants. The true maternal NVP coverage was defined as the proportion of HIV-infected women with confirmed NVP ingestion, which was measured by HIV and NVP testing of umbilical cord blood specimens from all live-born deliveries that occurred in the 10 selected facilities during the study period. Infant NVP coverage was determined by chart reviews (direct observation) to assess whether NVP had been administered prior to discharge.
Group comparisons used Student t tests, nonparametric Mann-Whitney U tests for continuous variables, and χ2 tests or Fisher exact tests for categorical variables. Results of maternal NVP coverage were expressed as percentages with their 95% confidence intervals (95% CI). Univariable and then multivariable logistic regression analyses were performed with a stepwise descending selection procedure to study the relation between maternal NVP coverage and variables collected on the maternal antenatal record. Adjusted odds ratios (aOR) and their 95% CI are reported, and P values are 2-sided. All analyses were performed using the STATA software, version 9 (College Station, TX).
Characteristics of Women at Delivery
During the study period, 9953 deliveries were registered at the 10 selected facilities. The median age of women was 25 years [interquartile range (IQR), 20-30], and the median gravidity was 3 (IQR: 1-4). A total of 9463 women (95.1%) had made at least 1 ANC visit before delivery, and the median number of ANC visits was 3 (IQR: 2-4). The vast majority of women delivered vaginally (96.1%). The median birth weight was 3000 g (IQR: 2700-3300), and 1212 babies (12.2%) had a birth weight of <2500 g (Table 1).
Cord Blood Surveillance and NVP Coverage
Cord blood specimens were collected from 9646 (98.1%) of the 9836 live-born deliveries. In 190 cases (1.9%), we were not able to collect cord blood samples, and the reasons were the following: mother transferred to a referral center before the blood sample collection (n = 76), the cord blood was unhealthy (n = 73), the cord blood snapped (n = 18), and other reasons (n = 23). Mothers from whom we were unable to obtain cord blood specimens did not differ from those from whom we obtained specimens for the following variables: age, gravidity, ANC visit, and HIV testing proposal (data not shown).
HIV antibodies were detected among 570 (5.9%, 95% CI: 5.5% to 6.4%) of the 9646 women who had a cord blood specimen available.
According to the maternal antenatal records of these 570 HIV-infected women, 229 (40.2%) were HIV seropositive, 58 (10.2%) were HIV seronegative, and 283 (49.6%) had no HIV information reported. Twenty-two women who were identified as HIV seropositive during an ANC visit according to their record were negative when their cord blood specimens were tested.
Among the 570 HIV-infected women, 138 (24.2%; 95% CI: 20.7% to 27.9%) had NVP detected in the umbilical cord blood. NVP syrup was administered to 208 (36.5%; 95% CI: 32.5% to 40.6%) neonates before their discharge from the facilities. Total NVP coverage (both maternal and infant dosing) was thus achieved in 102 (17.9%; 95% CI: 14.8% to 21.3%) (Fig. 1).
Standard PMTCT Cascade
In most mother-infant pairs, sufficient information was available in the maternal antenatal record to retrospectively describe the PMTCT cascade. Of the 9747 women (97.9%) who made at least 1 ANC visit, 5880 (60.3%) received HIV testing proposal; 5135 (87.3%) accepted it; 251 (4.9%) women were HIV infected; and 176 of them (70.1%) received antiretroviral prophylaxis (Fig. 1).
Among the 176 women who received antiretroviral prophylaxis during their ANC visits, 142 were seropositive based on cord blood surveillance and 85 (60.0%) had NVP concentration in cord blood specimens (Fig. 1).
Factors Associated With NVP Coverage
In a multivariate analysis after controlling for age and gravidity, maternal NVP coverage was higher when the woman had 2 or 3 ANC visits (aOR: 2.61; 95% CI: 1.27 to 5.39), or ≥4 (aOR: 3.84; 95% CI: 1.86 to 7.90) (ref. 0-1 ANC visit), and giving birth in clinic of first ANC visit (aOR: 2.21; 95% CI: 1.43 to 3.40). Maternal and infant NVP coverage was associated with the attendance of ≥4 ANC visits (aOR: 3.26; 95% CI: 1.45 to 7.34) and giving birth in the same clinic (aOR: 3.97; 95% CI: 2.29 to 6.88) (Table 2).
In this survey, we evaluated the effectiveness of a PMTCT program in Côte d'Ivoire using on one hand a standard PMTCT cascade approach and on the other hand an anonymous cord blood surveillance system. We found that 60% of the 570 HIV-infected women were identified with anonymous cord blood surveillance, and among them, 24% had NVP detected in their umbilical cord blood samples. The NVP coverage decreased to 18% when both mother and infant who received NVP were considered. We also found that 40% of the HIV-infected women who received NVP during ANC visits did not swallow the drug. In the multivariable analysis, a higher frequency of ANC visits (≥2) and giving birth in the center where the first ANC visit was performed were strongly associated with maternal and/or infant NVP coverage.
In Côte d'Ivoire, where the opt-out approach was recommended since 2006, we found that only 60% of pregnant women received an HIV test proposal despite completing a median of 3 ANC visits. Based on the World Health Organization report, the percentage of pregnant women in Côte d'Ivoire who received HIV testing was 47% in 2009.3 Thus, HIV-testing proposals seem to be the first major obstacle for creating an effective PMTCT program in this country. We could not collect information on the reasons that could explain this low rate of HIV testing. However, based on our field experience, the workload in the PMTCT facilities, the limited staff, and the insufficient training of health care workers involved in the PMTCT activities may be major reasons. In-depth interviews of health care workers could help to better describe the precise reasons for incomplete HIV testing proposals and be used as a mean to develop a specific strategy to correct this weakness.
Regarding anonymous cord blood surveillance, we observed that 60% of the 570 HIV-infected women were not identified as HIV infected during ANC visits and 10% of HIV-infected women were classified as HIV seronegative during ANC visits. In addition, 22 women identified as HIV seropositive during ANC visit were negative with anonymous cord blood surveillance. These observations confirm the importance of repeat HIV testing during ANC visits and in delivery wards by strictly applying the validated algorithm for HIV diagnosis, especially for at-risk pregnant women. This approach is now recommended by World Health Organization for generalized HIV epidemic areas10 and should be enforced in Côte d'Ivoire.
Anonymous cord blood surveillance also revealed that maternal NVP coverage was 24%, whereas it was estimated at 70% with standard monitoring methods where antiretroviral coverage is defined as the proportion of women who received antiretroviral prophylaxis among those infected with HIV according to the available records.4 Most countries use in fact this latter definition to estimate PMTCT coverage. In 2009, 53% of pregnant women living with HIV in low-income and middle-income countries were reported to receive antiretrovirals for PMTCT using this indicator.3 Our results show clearly that the data recorded in ANC visits are not always accurate for the evaluation of PMTCT effectiveness and may substantially overestimate antiretroviral PMTCT coverage. This overestimation may be explained by the fact that HIV seroprevalence may differ between women who are tested during ANC visits and women who are not tested11 and by the possible acquisition of new infections by HIV-seronegative women during pregnancy.12 Therefore, anonymous cord blood surveillance, including both HIV tests and NVP concentration measurements, is more reliable for the documentation of the effectiveness of PMTCT and should be considered as a useful tool for program evaluation at country level.8,11
Overall, the low NVP coverage observed in Côte d'Ivoire is explained by several reasons. First, the rate of HIV testing proposal is too low with only 60% of pregnant women receiving an HIV test proposal in our sample. Second, 70% of HIV-infected pregnant women only receive antiretroviral prophylaxis. Third, drugs are momentarily lacking in some clinics, an event which was reported during the study. Finally, a low rate of maternal NVP adherence is likely as 40% of HIV-infected pregnant women who received NVP during ANC visits did not swallow the tablet during labor. A similar observation was already reported in Lusaka, where 32% of women who received a single-dose of NVP did not take it.11 In Kenya, among 348 HIV-infected pregnant women, only 106 took NVP during labor, which resulted in a coverage rate of 30%.13 Adherence may be improved in a number of ways, such as promoting couples' HIV testing and communication during ANC visits, repeat counseling at each ANC visit, and the training of the labor ward staff to dispense again NVP to HIV-infected women before or during labor.
Regarding the factors associated with NVP coverage, we did not observe that a younger age was associated with a better NVP coverage, a finding of the multicountry study.8 However, we confirmed that performing at least 2 ANC visits was associated with a better NVP coverage. The location of the delivery center was another factor that we investigated and found associated with NVP coverage in our analysis but was not available in the multicountry study. Indeed, maternal NVP coverage was 2 times higher among women who delivered at the same health facility where their first ANC visit took place. Therefore, health care workers should recommend whenever feasible to pregnant women to stay in the same center from first ANC visit until delivery.
This study does have a few limitations. First, although the facilities were randomly selected, they are only a representation of the facilities providing PMTCT services in the country. Second, we could not include women who delivered at home in this study. These women and infants may be less likely to receive prophylaxis. Third, because no national identification of pregnant women tested for HIV exists, we probably underestimated the women who received an HIV-test proposal. Finally, we only tested NVP in the cord blood sample, and therefore, we probably underestimated maternal antiretroviral coverage with zidovudine or other antiretroviral drugs used for PMTCT although they were rarely used at that time without NVP.6
In conclusion, our data indicate that the estimates of NVP coverage based only on registries or maternal antenatal records are not always appropriate for the evaluation of the effectiveness of PMTCT programs and they were insufficient for Côte d'Ivoire at the time of the survey. Such data should often and regularly be linked with anonymous cord blood surveillance to facilitate PMTCT program planning and the identification of means of improvement. Our study identified 3 major obstacles as follows: (1) the low rate of HIV-testing proposal; (2) the low rate of dispensation of maternal antiretroviral drugs; and (3) the low maternal adherence. At the end of the study, we recommended the national authorities to take appropriate actions to improve PMTCT coverage, including: to reinforce the HIV testing opt-out approach and to propose a national identification system to facilitate the tracking of the testing status of pregnant women during ANC care. In the context of the implementation of the 2010 PMTCT recommendations,14 additional operational research should extend the approach used in this survey to better document the PMTCT effectiveness with more complex drugs that may no longer include NVP.
The authors acknowledge the Elizabeth Glazer Pediatric AIDS Foundation (EGPAF) staff in Côte d'Ivoire (Joseph Essombo, Anthony Tanoh and Patricia Fassinou-Ekouevi) and the PMTCT coordinators as well as the midwives working on the selected sites for their important contribution to this study. Hapsatou Touré is a fellow of the French charity Sidaction.
Copyright © 2011 Wolters Kluwer Health, Inc. All rights reserved.