A retrospective study of Human Immunodeficiency Virus transmission, mortality and loss to follow-up among infants in the first 18 months of life in a prevention of mother-to-child transmission programme in an urban hospital in KwaZulu-Natal, South Africa

Globally, approximately 370 000 children were newly infected with Human Immunodeficiency Virus (HIV) in 2009; the vast majority in sub-Saharan Africa [1], mostly due to mother-to-child transmission. In resource-constrained settings, with little or no antiretroviral treatment, approximately one-third of HIV-infected children die before one year and more than half die before two years of age [2]. In South Africa (SA), where the 2008 antenatal HIV seroprevalence was 29% [3], the child mortality rate increased from 56 per 1000 live births in 1990 to 67 per 1000 in 2008 [4]; HIV was the main contributor to the rising child mortality, responsible for approximately 40% of under-five mortality in 2000 [5]. Between 2001 and 2006, in a rural area of northern KwaZulu-Natal, improving maternal access to antiretroviral therapy (ART) and Prevention of Mother-to-Child Transmission (PMTCT) interventions resulted in a 57% decline in infant mortality [6].

While PMTCT interventions in SA prior to 2010 focused only on the perinatal period [7], the updated guidelines now recommend continuing ART during the postnatal period while breastfeeding and the use of oral nevirapine for all infants born to HIV-infected women who are not on ART for life and still breastfeeding (until one week after cessation of breastfeeding) [8]. With the successful implementation of the current South African antenatal PMTCT programmes aiming to reduce vertical transmission of HIV to less than five percent [9], poor follow-up of HIV-exposed infants remains a major weakness, undermining the ability to support safe infant feeding practices and to measure infant HIV-free survival at 18 months. Hence, understanding the determinants of retention is important. In a public-funded urban hospital in Johannesburg, South Africa, almost half of infants born to HIV-infected mothers in a routine PMTCT service were lost to follow-up (LTFU) by two weeks of age [10]. At the time this study was conducted, the national guidelines in South Africa did not include routine HIV testing for infants at six weeks and recommended testing for infants at 12 months of age [10]. Similarly, studies from Zimbabwe, Uganda, and Malawi reported LTFU of infants born to HIV-infected mothers in PMTCT programmes of 41% at 12 months, 30% at 18 months, and 30% at 24 months, respectively [11‐13]. In a hospital in Malawi 88% of infants who received nevirapine at the hospital were seen at the postnatal clinic [14]. More than half of these infants (n = 122, 54%) returned for their six month follow-up visit [14]. Early recognition of HIV-exposed infants is crucial given the higher incidence of infectious disease morbidity and mortality observed in HIV-infected and uninfected infants [2, 15‐18]. Further, early identification of HIV-infected infants is essential for initiation of ART, which can reduce early infant mortality by 76% and HIV disease progression by 75% [19].

In response to an increasing concern about LTFU of infants born to HIV-infected women in the PMTCT programme following delivery, in May 2008, McCord Hospital implemented a postnatal clinic for all HIV-infected mothers (with their infants) who had attended the PMTCT programme at the hospital. This new postnatal service provided an opportunity to characterize the frequency and correlates of LTFU of infants born to women attending the PMTCT programme at McCord Hospital.

In total, 272 HIV-infected mothers had deliveries between 1 May 2008 and 31 May 2009. Fourteen infants were excluded from the study due to missing records (n = 4), stillbirth deliveries (n = 4), duplicate entry (n = 1) and not meeting the inclusion criteria (n = 5). There were 258 deliveries resulting in 264 infants (251 singletons, five sets of twins and one set of triplets).

Median maternal age was 28.0 years (inter quartile range (IQR): 25.0 to 33.0 years). Almost two-thirds of mothers were employed and the majority were single. The baseline median maternal CD4⁺ cell count was 308 cells/mm³ (IQR: 192 to 420 cells/mm³). The median baseline viral load was 4320 copies/ml (IQR: 450 to 15 500 copies/ml). Twelve of 127 (9.4%) women with data had pulmonary tuberculosis and were receiving treatment (Table 1). There was only one maternal death during the study period (Table 1). The cause of the maternal death was reported as disseminated Kaposi’s Sarcoma (data not shown).

Nearly half of the women (n = 128, 49.6%) included in the study were on ART for life (Table 1). Of the 106 women for whom the date of ART initiation was available, 97 (91.5%) were initiated on ART for life in the current pregnancy. The other 130 women (50.4%) were prescribed ARV prophylaxis for PMTCT; 85 (32.9%) were given interrupted triple therapy (zidovudine, lamivudine and efavirenz) from 28 weeks gestation until delivery, 40 (15.5%) were prescribed zidovudine from 28 weeks, and four (1.6%) received sd-nevirapine at delivery. Only one (0.4%) patient had sd-nevirapine alone due to late attendance in labour. All ARV prophylaxis for PMTCT was followed by a seven day tail of zidovudine and lamivudine. Most women (n = 155, 63.0%) attended their first antenatal visit at McCord Hospital between 13 to 27 weeks of gestation.

The median length of infant follow-up at McCord Hospital in this study was 9.1 months (IQR: 1.6 to 12.0 months) (Table 2). Of the 89 infants delivered through elective caesarean sections, 44.9% (n = 40) of caesarean sections were for women with viral loads above 1000 copies/ml at 36 weeks. The median birth weight was 3080 grams (IQR: 2780 to 3380 grams). There were 32 low birth weight infants (< 2500 grams) (12.8%; 95% CI: 8.9% to 17.6%). The median gestational age at birth was 38.0 weeks (IQR: 38.0 to 39.0 weeks). There were 189 infants born at term at or after 37 weeks (89.2%, 95% CI: 84.2% to 93.0%). Most infants (n = 223, 98.7%) received nevirapine and zidovudine as antiretroviral prophylaxis following delivery.

In an urban South African setting, the integrated mother-child postnatal clinic at McCord Hospital achieved good outcomes in the first 13 months of its inception. These outcomes included infant HIV testing of 83% and HIV transmission risk of 2.7% at six weeks in returning HIV-exposed infants whose mothers received PMTCT at McCord Hospital. However, LTFU by six months was substantial, limiting our ability to determine HIV-free survival at 18 months.

In developed countries vertical transmission of HIV to infants occurs in 1-2% of pregnancies in HIV infected women, achieved through a combination of interventions, including antiretroviral therapy regimens that optimally suppress viral load, elective Caesarean section and complete avoidance of breastfeeding [25‐30]. In developing countries the caesarean section proportion ranges from 3% to 12.6% [11, 12, 31]. The high proportion of caesarean sections at McCord Hospital may reflect the ability of the women in this setting to choose the best possible care for their infants and the hospital's capacity to deliver a comprehensive package of interventions to reduce the vertical transmission of HIV.

Although the rate of LTFU at McCord Hospital was comparable to other PMTCT programmes in public settings in sub-Saharan Africa [11, 14, 32‐34], the high rate of early dropout in our setting may in part be due to women changing to service providers nearer to their residences. The period of greatest attrition was in the first week of life. This data may support the explanation of mothers returning to clinics in closer proximity to their homes. A similar finding was noted in a PMTCT programme at a centralized hospital in Malawi [14]. The authors suggested that the LTFU of mother-infant pairs may have been due to women from rural areas returning to their peripheral clinics following delivery [14]. The revised South African PMTCT guidelines may improve the follow-up of mothers and infants in this setting as HIV-infected women attend for their own health, and infants who are breastfed beyond six weeks require repeated access to nevirapine until one week after cessation of breastfeeding.

Late antenatal presentation (28 weeks of gestation or later) was a strong predictor of LTFU. There may be several reasons for late antenatal clinic attendance at McCord Hospital: women may have been minimizing expenses related to antenatal care; lack of awareness of the value of early antenatal care; or poor maternal health seeking behaviours. These results highlight the importance of identifying late antenatal attendees to determine the risk factors which contribute to attrition following delivery. Moreover, the fee-paying nature of the hospital needs to be taken into account when assessing the attrition of late antenatal attendees.

In prior studies, poor socio-demographic circumstances have been associated with patient attrition. In Malawi HIV-exposed infants born to parents who were less educated and in farming occupations were more likely to be LTFU [11]. In our multivariable model, there was no association between socio-demographic characteristics and infant LTFU (maternal employment and marital status). Since the parental level of education was not routinely collected at the time this study was conducted, the association with LTFU could not be assessed. In the Ugandan study lack of understanding of the importance of follow-up was noted as a reason for attrition [12]. In Johannesburg, maternal unemployment, geographical relocation and lack of paternal support were noted as reasons for poor retention [10]. Maternal education and support regarding the importance of follow-up care for their own and their infants’ health may improve patient retention and facilitate early diagnosis of infant morbidity and HIV transmission risk.

This study demonstrated that relative to infants of HIV-infected women with CD4 ⁺ counts 200 cells/mm³ or less, infants born to mothers with CD4⁺ counts above 200 cells/mm³ were more likely to be lost to follow-up. This finding was not statistically significant. Maternal well-being may be a risk factor for poor retention in our setting as mothers do not perceive themselves or their infants to be at risk for disease. These mothers may be less likely to seek health care. A similar correlation between infant wellbeing and loss to follow-up was shown in a study in rural Uganda, where infant illness was a protective against loss to follow-up in the PMTCT programme [12].

A key strength of this study was the ability to determine if patients were not returning for care. Clinic staff at McCord Hospital routinely contacted patients if they missed a scheduled appointment at the clinic, noting reasons for the missed appointment and rescheduling another. If these patients decided to attend other health care facilities, they were not considered LTFU as the reason for their non-attendance was known. Moreover, routine monitoring allowed early detection of infants requiring follow-up care.

A limitation of this study was the selection of the study population. According to the selection criteria, only infants whose mothers attended the PMTCT programme at McCord Hospital, and/or were delivered at McCord Hospital, and/or were brought back to the hospital for follow-up care were included in the study. Accordingly, the outcomes of infants who were not brought back to McCord Hospital for further care remained unknown. In addition, since the service was not available, the study could not include a comparison group of infants born to HIV-uninfected women from McCord Hospital. Inclusion of this comparison group would have made it possible to determine the background risk of HIV-exposed uninfected infants in terms of loss to follow-up. Finally, errors in the routine maternal and infant records and assignment of LTFU status may have constituted information bias.

The results of this study may not be generalizable to the HIV-exposed infant population in the public sector in South Africa. Moreover, formula feeding was reported as the predominant infant feeding modality. The impact of feeding practices on infant mortality requires further consideration. The results from this new integrated postnatal mother and baby clinic at McCord Hospital provide important lessons for other PMTCT services in developing settings. The attrition of HIV-exposed infants, particularly those born to women who attended antenatal services at McCord Hospital after 28 weeks highlights the need to identify late antenatal attendees for follow up care and to determine the HIV transmission risk and mortality of HIV-exposed infants.

As efforts are made to improve rates of early HIV detection at six weeks of age and subsequent provision of care and treatment to HIV-exposed children in sub-Saharan Africa, it is essential that we develop better understanding of the determinants of retention and refine our strategies for retaining children who are at the greatest risk of morbidity and mortality in care. This preliminary study assessed a new postnatal service with its accompanying problems, further study is recommended to determine changes in the follow-up rates of infants.

TC FCPHM. SK FCPHM. JG MFamMed. TLC PhD. LMB PhD. MLN PhD.

The Africa Centre for Health and Population Studies receives core funding from the Wellcome Trust.