Introduction Initiating and retaining pregnant women on antiretroviral therapy (ART) to prevent mother-to-child HIV transmission (PMTCT) remains a major challenge facing African HIV programs, particularly during the critical final months prior to delivery. In 2013, South Africa implemented its “Option B” PMTCT regimen (three-drug ART throughout pregnancy and breastfeeding, regardless of maternal CD4 count) and introduced once-daily fixed-dose combinations and lifelong ART. Currently, the uptake of Option B and its possible impact on adherence to PMTCT during the critical final months of pregnancy is unclear. Materials and methods We prospectively collected visit data from a cohort of adult, HIV-infected, pregnant women between July 2013-August 2014 to estimate three models of adherence to PMTCT during the final 16 weeks immediately preceding delivery. Adherence was defined according to possession of antiretroviral drugs, which was inferred from clinic visit records under varying assumptions in each model. We describe uptake of the PMTCT regimen, gestational age at initiation, and model possible scenarios of adherence through delivery after the implementation of Option B. Results Among 138 women enrolled (median (IQR) age 28 years (24–32), median CD4 count 378 cells/mm3), median (IQR) gestational age at initiation was 22 weeks (16–26). Estimates of adherence during the final 16 weeks of pregnancy prior to delivery ranged from 75% (52–89%) under the best case scenario assumptions to 52% (30–75%) under the worst case scenario assumptions. Estimates of the proportion of women who would achieve 80% adherence to PMTCT were <50% across all models. Conclusions Despite the switch to Option B and once-daily dosing, South African women continue to initiate PMTCT late in pregnancy, and estimations of regimen adherence, as modelled using PMTCT visit attendance data, is poor, with <50% of women reaching 80% adherence during final months of pregnancy across all models. Further guideline changes and interventions are needed to achieve vertical transmission goals.
The data utilized in this study were collected at a public-sector primary health clinic (PHC) in Johannesburg, South Africa. It is located in an informal settlement on the edge of the city with a large population of immigrants from neighboring countries and patients from other South African provinces. It provides a full range of primary healthcare services, including antenatal care (ANC), maternal and child health, HIV testing and counseling, ART initiation, and ART management. Under South African guidelines], women are advised to make at least four antenatal care visits prior to delivery, starting at 14 weeks’ gestation. PMTCT guidelines increased the required number of visits, however, as clinics standard practice was to provide only one month of ARV medication at a time during the study period. The clinic does not have a delivery ward; referral for delivery was to either a nearby hospital or one of three maternal obstetric units (MOUs) located 4, 5.5, and 7 kilometers from the clinic, respectively. At least one antenatal clinic visit was needed for a woman to obtain an antenatal care card, which is required to assure access to a hospital or MOU for delivery. We sequentially enrolled adult (≥18 years old), HIV-infected, pregnant women at the earliest clinic visit for a positive HIV test, provision of a blood sample for a CD4 count, or first ANC visit between July 4, 2013 and August 14, 2014. Women who were already on ART for their own health or who indicated that they did not intend to seek the remainder of their antenatal care at the study clinic were excluded. After informed consent, the women were administered a questionnaire eliciting information about demographic and socioeconomic characteristics; previous pregnancies, health-seeking behavior and exposure to PMTCT; normal activities and employment, and the costs the patient incurred per clinic visit. Study patients were then followed by passive medical record review, using the site’s and delivery facility’s routine patient records, until delivery. These records included antenatal and labour ward registries as well as an electronic medical record system, TherapyEdge-HIV™. Fields collected included dates of antenatal clinic visits, gravidity and parity, gestational age at first visit and predicted delivery date, date of HIV test, ARV regimen prescribed, and baseline CD4 count. Not all data for all study subjects could be found in the routine records maintained by the study site, nor could confirmation of all deliveries be found at the referral facilities. Viral load test results were not available for the majority of study participants, as most had spent fewer than 6 months on ART at the time of delivery. The primary defined outcome for this study was proportion of patients achieving 80% adherence to PMTCT in the 16 weeks immediately preceding date of delivery. For this analysis, we assumed 16 weeks to be a reasonable minimum duration of ART needed to achieve a high rate of viral suppression during pregnancy based on published South African data [12,13] and approximately 80% adherence to PMTCT to be the minimum required for successful viral suppression [14,15]. We focused the analysis on the final four months (16 weeks) before delivery on the assumption that viral suppression during this later period is more important for reducing the risk of MTCT [13] than earlier in the pregnancy. Adherence to PMTCT was estimated using clinic visit attendance under three different models with varying assumptions. Date of delivery was unknown for some of the women in the study. We instead assumed that for those without a known delivery date, delivery took place at either 1) 36 weeks’ gestation; 2) 38 weeks’ gestation; or 3) 40 weeks’ gestation. Based on these assumptions, delivery dates were imputed accordingly based on the gestational age at which a woman presented at her first antenatal visit. Using the data described above, we then estimated three models with varying assumptions about medication dispensed at each observed antenatal clinic visit (Table 1). Because we cannot observe actual adherence, we used medication possession to proxy adherence, assuming that if a 28-day supply of medication was dispensed, then the patient could be considered fully adherent for the next 28 days [16]. In Model A (100% dispensing), we assumed that all women with a documented clinic visit received 28 days’ worth of medication at each visit (i.e., that the stipulated clinic practice was adhered to perfectly) except the last documented visit, as explained below. For Model B (90% dispensing) we assume that 90% of women who attended a visit received 28 days’ worth of medication while the remaining 10% attended a visit but were not dispensed medication. To model this, 90% of women were randomly allocated 28 days’ medication while the remaining 10% were allocated 0 days’ medication at each visit except the last documented visit, and women who were allocated 0 days’ medication at a given visit were not permitted to be randomized to the 0 days’ medication allocation at the next visit, so that that no woman attended two consecutive visits without receiving medication at one of the visits at a minimum. Model C (80% dispensing) followed a similar structure to Model B: at each visit except the last documented visit, 80% of women were randomly allocated 28 days’ medication while the remaining 20% were allocated 0 days’ medication. Again, we assumed that no women attended two consecutive visits without receiving medication during at least one of the visits. For any visit that occurred one month unobserved time prior to delivery date or estimated delivery date—assume 40% of these women remain in care during unobserved time ((they get sufficient pills up to known or estimated delivery date) while 60% get the allocated 28 days’ worth of pills dispensed at their last visit according to the model allocation (100%, 90% or 80% dispensing). ** Assumption applied to n = 34 women with unknown delivery date We handled the last observed visit differently. For women with observed clinic visits within one month of their delivery date or estimated delivery date, medication dispensing continued according to the primary assumption for each model (Model A—100% women got 28 days’ allocation, Model B—90% women got 28 days’ allocation, or Model C—80% women got 28 days’ allocation). For women with more than one month of unobserved time prior to their delivery date (i.e., those who were considered lost from care more than one month prior to their known or estimated delivery date), however, several possibilities exist for medication adherence during the unobserved time. First, a woman may truly be lost from care and not taking any medication during the entire period. Second, the woman may have transferred to another facility, such as the maternity unit where she intended to deliver, and continued treatment at the new facility. Finally, the woman may have been in possession of some medication from her previous visits, due to less than full adherence earlier in the pregnancy, and continued to take the leftover medications until that supply was depleted. A previous report from South Africa estimated that almost 40% of women who appeared lost from care at the original antenatal care facility were in fact receiving on-going care at another facility where they ultimately delivered]. To account for this, in all models we adjusted dispensing at the last observed visit among women with more than one-month unobserved time prior to delivery date or estimated delivery date (i.e. those apparently lost to care). We randomly selected 40% of these women and assumed that they remained in care during the entire period of unobserved time prior to delivery, while the remaining 60% got only the medications dispensed at their last visit according to the model allocation (Model A—100% women got 28 days’ pill dispensing, Model B—90% women got 28 days’ pill dispensing, or Model C—80% women got 28 days’ pill dispensing). We initially summarized characteristics at study enrolment with descriptive statistics. Next, for each of the three models’ set of assumptions, we estimated the proportion of women considered adherent to PMTCT during the final 16 weeks immediately preceding delivery using medians and interquartile ranges (IQR). We then categorized adherence for each woman in the last 16 weeks of pregnancy (weeks 24 to 40) as >80% or not and described the proportion of women achieving 80% adherence during this period for each of the levels in the three models. Finally, we fitted a log-binomial regression model to estimate the risk ratios (and 95% confidence intervals) of potential baseline predictors of achieving 80% adherence to PMTCT during the final 16 weeks of pregnancy. ClinicalTrials.gov {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01710397″,”term_id”:”NCT01710397″}}NCT01710397 South African National Clinical Trials Register DOH-27-0213-4177 This study was nested within the original RapIT randomized controlled trial which was registered on ClinicalTrials.gov (Registration number) {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01710397″,”term_id”:”NCT01710397″}}NCT01710397 on September 7, 2012 and the South African National Clinical Trials Register DOH-27-0213-4177. The first participant in the RapIT study was enrolled on May 8th, 2013. All study participants provided written informed consent and the consent procedure as well as all other study procedures received ethics approval from the institutional review boards of Boston University and the University of the Witwatersrand.