Introduction
Deaths from HIV/AIDS are rapidly eroding child survival gains in southern Africa [1]. Programmes for the prevention of mother-to-child transmission of HIV (PMTCT) consisting of a package of HIV testing and provision of antiretroviral drug treatments (including mono- or multidrug therapy in short course or single-dose regimens) have been shown to be one of the most cost-effective interventions in preventing vertical HIV transmission [2,3]. Governments in Africa have committed to scale up this intervention to reduce the proportion of infants infected with HIV by 50% by 2010 [4]. To achieve these ambitious targets requires a rapid PMTCT programme expansion in countries most impacted by the HIV pandemic.
However, there are few data concerning the effectiveness of these programmes in operational, rather than clinical trial, settings in Africa. Operational studies have so far been centred around pilot programmes [5,6], mostly hospital based [5–7]; did not measure HIV transmission and/or infant death [5,6,8]; or had short follow-up (6–16 weeks) [7,9,10]. One small study on operational effectiveness of a programme in Kenya showed no reduction in vertical HIV transmission from mother to child at 14–16 weeks when compared with the preintervention period [7]. Longer-term outcomes such as HIV-free survival beyond 6 months have not been documented in operational large-scale programmes in Africa. In particular, the importance of measuring longer-term postnatal HIV transmission is underlined by the fact that breastfeeding is responsible for a high proportion of mother-to-child transmission in developing countries.
In 2001, the South African National PMTCT Programme was initiated in 18 pilot sites and then rapidly expanded to all districts country-wide. The national protocol [11] initially consisted of (a) antenatal voluntary counselling and testing for HIV; (b) single-dose nevirapine to the mother in labour and to the baby within 72 hours of delivery; and (c) antenatal infant feeding counselling with recommendations for mothers to chose either exclusive breastfeeding (with rapid weaning between 3 and 6 months after birth) or exclusive formula feeding; (d) free infant formula from birth to 6 months for mothers choosing exclusive formula feeding; and (e) co-trimoxizole prophylaxis to the newborn infant from 6 weeks until infant HIV testing at age 9–12 months.
This study reports the findings of an evaluation of the routine operational effectiveness of the South African National PMTCT Programme, primarily measured by HIV-free survival at 36 weeks post-delivery across three districts.
Methods
Study design
This was a prospective cohort study of mother–infant pairs participating in the national PMTCT Programme in South Africa. HIV-positive women (and their infants) were recruited prior to, or at the time of, delivery and followed until the infants were 36 weeks of age. Ethical approval was obtained from the University of Kwa-Zulu-Natal and permission to conduct the study was obtained from participating institutions and the Departments of Health. Signed informed consent was obtained at enrolment, with verbal informed consent at each home visit. Compensation for interview time was based on local norms in the form of food vouchers (Umlazi), cash (Paarl) or food parcels (Rietvlei).
Study settings
The three study sites (Paarl, Rietvlei and Umlazi) were among the 18 original national pilot sites and were deliberately selected to reflect different socioeconomic contexts, rural–urban locations and HIV prevalence rates. Paarl (Western Cape province) is a periurban/rural area, with a relatively higher socioeconomic profile, a relatively well-functioning public health system and an antenatal HIV prevalence of 9% during the study period. Rietvlei (Eastern Cape province) is a rural area in one of the poorest regions of South Africa, with 28% antenatal HIV prevalence. Umlazi (KwaZulu-Natal province), a periurban area with formal and informal housing, is considered to be intermediate with regard to health resources compared with the other two sites. The antenatal HIV prevalence was 47% [12].
Participants: cohort composition
The sample size for each site was estimated to provide a precision in the HIV transmission (or death) rate at 9 months of ±4.3% in Umlazi, ±6.5% in Rietvlei and ±7.5% in Paarl. Patient volumes and HIV rates varied across sites. Study length was determined to achieve desired minimum precision in smallest site (Paarl), resulting in better estimated precision in the two larger sites.
HIV-positive mothers were recruited from the local hospital or clinic offering the PMTCT programme by a qualified field researcher. Refusal rates both for the PMTCT programme (0–19%) and study participation were low in all sites. Recruitment over a period of 12 months yielded a total sample size of 665 mother–infant pairs as follows: Paarl 149, Rietvlei 192 and Umlazi 324. Loss to follow-up at 36 weeks overall was lower than expected, at 21%, and was not statistically different across sites, yielding a final sample of 525.
Main outcome measures
The main outcome measure for this study was HIV transmission from mother-to-child and/or infant death by 36 weeks of age. An infant age of 9 months was chosen as this was the original age for HIV testing and exit from the PMTCT programme in the South African PMTCT protocol.
Maternal HIV status was initially determined from routine PMTCT medical records. HIV viral load in mothers and HIV infection in infants was assessed from dried blood spots collected on Guthrie cards by means of a heel/finger prick during home visits at 3, 24 (infants only) and 36 weeks. HIV infection in infants was determined by detectable viral load using quantitative HIV-1 RNA NASBA (Nuclisens ECL; bioMérieux, Marcy l'Etoile, France) and/or positive test using qualitative HIV-1 DNA polymerase chain reaction assay (Amplicor version 1.5; Roche Molecular Systems, Branchburg, New Jersey, USA). In cases where a mother recorded as being HIV positive had no detectable viral load, a repeat laboratory enzyme-linked immunosorbent assay was carried out [Uniform 2 HIV-1 Assay (bioMérieux) followed by Biorad HIV-1 Assay (Hercules, California, USA)]. If this also confirmed negative (12 mothers; 1.9%) these mother–infant pairs were deleted from any analysis including HIV transmission.
Data collection
Data were collected during the period October 2002 to November 2004 by either trained field researchers (postpartum in the delivery facility, and in the home at 3 weeks to measure early feeding and HIV transmission; at 24 weeks as the age by which breastfeeding should have been discontinued; and at 36 weeks, which was the original PMTCT programme exit point) or trained community health workers (in the home at 5, 7, 9, 12, 16, 20, 28 and 32 weeks). Semistructured interviews with the infant's mother or caregiver collected data on sociodemographics, infant feeding practices, infant and maternal morbidity, and disclosure of HIV status. Interviews were in the preferred language of the subject (Xhosa, Zulu, Afrikaans or English). The field researcher also reviewed the perinatal medical record of both mother and infant to extract data on antenatal, intrapartum and postpartum care, PMTCT programme care and the occurrence of pregnancy or newborn complications.
Data management and analysis
Quantitative data were entered into MS ACCESS using double data entry at a central site (MRC Durban). After validation, the database was exported to Stata version 8.0 (Stata Corp., College Station, Texas, USA) for data management and analysis. Comparisons of variables across sites were carried out using χ2 tests for categorical variables and one-way analysis of variance for continuous variables (with the exception of income, which was compared using a Kruskal–Wallis test). There were no adjustments made for multiple testing as the comparison between sites was viewed as descriptive rather than inferential. In analysing the data, cognisance was taken of papers published by the Ghent group [13,14]. This led to the choice of HIV-free survival as the primary outcome (in fact the composite endpoint of HIV transmission or death was used). The Ghent group have pointed out that the methods they advocate are not appropriate for exploratory analyses to investigate risk factors for HIV transmission or death, and for this reason Turnbull's adaptation of the Kaplan–Meier estimate of overall HIV transmission or death was not used. In order to investigate potential risk factors for HIV transmission, Cox's proportional hazards regression models were fitted, using the midpoint between the last negative and first positive test as the time of infection and using Efron's method for adjusting for tied survival times (since there were a high proportion of tied survival times). The proportional hazards assumption (between sites) was checked using log cumulative hazards plots and the test of Grambsch and Therneau [15]. In addition, multiple logistic regression models were fitted using the method of Hogan et al. [16] to inversely weight for the probability of not being lost to follow-up. The results from these multiple logistic regression models were consistent with those of the Cox regression models [on converting the hazard ratios (HR) from the Cox regression to odds ratios]. Consequently, only the results from the Cox regression models will be presented here, as HR values are easier to interpret than odds ratios. All factors as described below (Table 1) were examined as potential risk factors. Variables were retained in the final models if they were either of at least marginal significance or played a confounding role, in that including the variable had a noticeable effect (> 10% change), on the between-site HR values.
;)
Table 1: Characteristics of HIV-positive mothers and their infants: quality of care, HIV transmission and infant mortality across study sites.
Results
There were substantial differences across the three study sites with regard to maternal and infant characteristics, quality of maternity and PMTCT programme care, and infant feeding (Table 1). Maternal sociodemographics showed significant differences for all variables, with Rietvlei being the most disadvantaged. The mean gestational age at birth was lower in Umlazi, but birthweight was similar across the sites. For antenatal and postnatal care, Rietvlei showed the poorest indicators for quality of care as measured by antenatal visits, syphilis testing, postnatal visits and immunization rates. With regard to infant feeding, Paarl had the lowest and Umlazi the highest proportion of infants who reported exposure to any breastmilk between birth and 36 weeks of age (‘ever breastfed’). For PMTCT programme indicators, Rietvlei showed the poorest performance with regard to nevirapine per protocol, counselling and disclosure.
With regard to outcomes of interest, the differences seen in cumulative HIV transmission at 36 weeks across sites were not statistically significant (P = 0.07; Table 1). However, the differences in infant mortality rates across the three study sites by 36 weeks were statistically significant (P = 0.0005; Table 1), with Rietvlei showing substantially higher infant mortality.
The composite indicator of infant HIV transmission and/or death was also significantly different across the sites (Table 2). This translates into HIV-free survival rates of 84%, 64% and 73% in Paarl, Rietvlei and Umlazi, respectively (P = 0.0003; Fig. 1).
Table 2: Infant HIV positivity and/or death by 36 weeks of age by study site amongst infants born to HIV-positive mothers.
Fig. 1: Kaplan–Meier plots of infant HIV positivity and/or death by 36 weeks of age amongst infants born to HIV-positive mothers. Symbol, Paarl;
Symbol, Umlazi;
Symbol, Rietvlei. Differences across sites by log rank χ
2 test for equality of survivor functions (two degrees freedom) χ
2 = 16.42;
P = 0.0003.
SymbolSymbolSymbolAll factors in Table 1 were examined as risk factors for HIV transmission and/or infant death both individually and in a multivariable analysis. In the multivariable analyses (Table 3), the most important risk factor for HIV transmission and/or infant death overall was maternal viral load: an increase of 1 on the log scale increased the risk of HIV-transmission and/or death to the infant of by approximately 1.5 times. Being premature also increased the risk (marginal significance in adjusted analyses: P = 0.058–0.093) of the infant becoming HIV positive or dying. There was also some evidence that being ‘ever breastfed’ increased the risk of HIV positivity and/or death, although this effect was not statistically significant (P = 0.13 and P = 0.16, respectively; Table 3).
Table 3: Risk factors for HIV transmission or death by 36 weeks of age: Cox's proportional hazards regression models.
The data presented in Table 3 also provides a comparison of the HR across sites using Paarl as the referent. Model 1 shows significant differences in the crude comparison across sites (Rietvlei: HR, 2.27; 95% CI, 1.36–3.77; Umlazi: HR, 1.74; 95% CI, 1.06–2.84). Model 2 suggests that even after controlling for maternal viral load and prematurity (the only two independent risk factors as noted above), the site differences remained statistically significant. In model 3, adjusting for ‘ever breastfed’ resulted in a reduction in the magnitude of the Umlazi–Paarl comparison (HR, 1.46; 95% CI, 0.84–2.53), which was no longer statistically significant, while the Rietvlei–Paarl hazard ratio was essentially unchanged. The final model, 4, which included two health system variables (having fewer than four antenatal visits and not having an antenatal syphilis test), resulted in a reduction in the Rietvlei–Paarl HR to 1.81 (95% CI, 0.93–3.50) and loss of statistical significance, but no appreciable difference in the HR for the Umlazi–Paarl relation compared with model 3. These results suggest that health systems variables in Rietvlei and exposure to breastmilk feeding in Umlazi could explain a portion of the differences seen in HIV-free survival rates across these PMTCT programme sites.
Discussion
While nearly all (95%) pregnant women attend an antenatal clinic at least once in South Africa [17], and antenatal HIV testing uptake was high in these three study sites (81–100%) [12], this study found very different outcomes across the sites. An important finding of this research is that established risk factors do not explain these substantial differences in HIV-free survival between sites. Maternal viral load, the strongest independent risk factor for HIV transmission and/or death, was actually highest in Paarl, the site with the best HIV-free survival.
The biggest difference was seen between Paarl, which had rates of HIV transmission and/or death comparable with the treatment group in clinical trials of single-dose nevirapine [2], and Rietvlei, which showed 36 week rates of HIV transmission and/or death similar to placebo in the Petra trial [18]. The regression analyses suggest that a mother in Rietvlei with a similar viral load, gestational aged baby and infant feeding practice as a mother in Paarl is still more than twice as likely to experience her child becoming HIV infected or dying by 9 months than the mother in Paarl. Our analyses suggest that differences in the quality of healthcare services explain a portion of the difference in HIV transmission and/or death between Paarl and Rietvlei (Table 3). Despite relatively high utilization of health services across all three sites (the mean number of antenatal visits even in Rietvlei was more than three), there were significant differences in the uptake of nevirapine and in indicators of quality of care, such as proportion of mothers having a syphilis test, immunization coverage and recall of counselling (Table 1).
Rietvlei also has the poorest socioeconomic and environmental conditions of the three sites [19]. However, the use of fairly crude measures of socioeconomic status, the strong relation of socioeconomic status with site and the homogeneity of socioeconomic status within sites probably explains why we did not find socioeconomic status to be an independent risk factor in the multivariable analyses. Painter and colleagues in Côte d'Ivoire found sociodemographic factors influenced PMTCT programme participation, and they suggested that increased attention to ‘broader structural factors’ will also be needed to increase the effectiveness of PMTCT interventions [20].
In contrast, the difference in HIV-free survival between Umlazi and Paarl was smaller and was not explained by differences in quality of care. This is not surprising as the health system variables in Umlazi and Paarl were similar, as were the socioeconomic conditions (Table 1). However, exposure to HIV via breastfeeding does explain a portion of the difference seen between the sites (Table 3). Notably, Umlazi had the highest rate of breastfeeding among the three sites (Table 1).
Our findings with regard to risk factors for HIV-free survival are consistent with findings of many other studies. For example, maternal viral load, baby birthweight and breastfeeding were found to be risk factors for HIV transmission in a recent pooled analysis of short-course drug efficacy trials across Africa [21]. The analysis also found combination short-course therapy with zidovidine and lamivudine to be more efficacious than single-dose therapy. Expanded access to such short-course combination therapy, increased access to HAART for eligible mothers, and postnatal support for optimal infant feeding could considerably reduce HIV transmission and infant mortality rates. However, expanded interventions would need to take cognisance of existing operational challenges, as demonstrated by this study.
Generalizations from these results should be cautious as the sites were deliberately selected to include diverse settings in the South African context. The health system variables were highly correlated with each other and with site, which may have led to multi-colinearity within statistical models. However, the correlation of health systems variables with site probably acted as markers for the extreme inequities found in health systems across these sites. Finally, it is important to note that this study evaluated the PMTCT programme as a whole. It was not designed or powered to examine individual components within the national protocol.
Our findings support the suggestion by McCoy et al.[22] and Peeling et al. [23] that the addition of new clinical interventions, such as HIV treatment and prevention programmes, to already under-resourced and poorly functioning health systems may not lead to improved HIV-related health indicators. McCoy et al. [22] suggested that already overburdened systems may have trouble coping with the additional workload involved in the new programme. Peeling et al. [23] reported no significant reduction in infant mortality in Haiti even with the introduction of significant new PMTCT resources. They partly attributed this lack of improvement to the neglect of the broader health system, as reflected by continuing high levels of congenital syphilis as a result of the low levels of syphilis screening and treatment in the antenatal services.
Our findings suggest that if the benefits of PMTCT interventions are to be realised then simultaneous attention to underlying socioeconomic conditions and healthcare infrastructure is needed, including the provision of additional resources such as staff and funding in disadvantaged areas with poorly functioning health services. Most national PMTCT plans, however, do not address such inequities within the broader healthcare system. In South Africa, annual district health expenditure ranges from under R50 (US$8) per capita to R389 (US$55) per capita, with the most deprived districts with the greatest health needs generally receiving the least resources [24]. It is hardly surprising that PMTCT service outcomes vary so considerably given the lack of an accompanying programme of health resource redistribution.
Although the HIV-free survival rates in Rietvlei are disappointing, this does not support a conclusion that PMTCT services should be withheld until socioeconomic and healthcare conditions improve. Rather it necessitates greater efforts to improve socioeconomic conditions and healthcare infrastructure in order to ensure that PMTCT interventions actually result in improved child health outcomes. Furthermore, the scale-up and implementation of new programmes such as PMTCT could be designed from the beginning for active catalysis of broader health system development, rather than as a narrow single-issue intervention.
While well-functioning healthcare systems and reasonable socioeconomic conditions are axiomatic to improved child health outcomes, the findings from this research also point to the ongoing and vexing challenge of reducing postnatal transmission from breastfeeding. Our results suggest that breastfeeding is still a major source of mother-to-child transmission of HIV, even in the context of a PMTCT programme that offers free commercial infant formula. More research is urgently needed on how pregnant and postpartum HIV-positive women can be supported to make the most appropriate infant feeding choice for their personal circumstance and how they can be supported to adhere to this choice.
Finally, this paper highlights the importance of designing and funding studies to evaluate the long-term outcomes of new interventions within operational settings. Indicators of coverage or uptake of new interventions are easier to measure, but they can result in a false sense of achievement. In under-resourced areas of Africa, with poorly functioning healthcare systems, cohort studies to evaluate operational effectiveness are necessary.
Acknowledgements
The Good Start Study Group would like to thank the study sponsors: the South African National Department of Health, the US Centers for Disease Control and Prevention South Africa, UNICEF and the Swedish-South Africa Cooperation (SIDA/NRF).
References
1. Zaba B, Whitworth J, Marston M, Nakiyingi J, Ruberantwari A, Urassa M,
et al.
HIV and mortality of mothers and children: evidence from cohort studies in Uganda, Tanzania, and Malawi. Epidemiology 2005; 16:275–280.
2. Guay LA, Musoke P, Fleming T, Bagenda D, Allen M, Nakabiito C,
et al. Intrapartum and neonatal single-dose
nevirapine compared with zidovudine for prevention of mother-to-child transmission of
HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet 1999; 354:795–802.
3. Moodley D, Moodley J, Coovadia H, Gray G, McIntyre J, Hofmyer J,
et al. A multi-center randomized controlled trial of
nevirapine versus a combination of zidovudine and lamivudine to reduce intrapartum and early postpartum mother-to-child transmission of human immunodeficiency virus type 1. J Infect Dis 2003; 187:725–735.
4. United Nations.
A World Fit for Children. [
Outcome document of the United Nations Special Session on Children 8–10 May 2002.] New York: United Nations; 2002. Accessed 20 June 2006 at
http://www.unicef.org/specialsession/docs_new/documents/A-RES-S27-2E.pdf.
5. Perez F, Orne-Gliemann J, Mukotekwa T, Miller A, Glenshaw M, Mahomva A,
et al. Prevention of mother to child transmission of
HIV: evaluation of a pilot programme in a district hospital in rural Zimbabwe. Br Med J 2004; 329:1147–1150.
6. van't Hoog AH, Mbori-Ngacha DA, Marum LH, Otieno JA, Misore AO, Nganga LW,
et al. Preventing mother-to-child transmission of
HIV in Western Kenya: operational issues. J Acquir Immune Defic Syndr 2005; 40:344–349.
7. Quaghebeur A, Mutunga L, Mwanyumba F, Mandaliya K, Verhofstede C, Temmerman M. Low efficacy of
nevirapine (HIVNET012) in preventing perinatal
HIV-1 transmission in a real-life situation. AIDS 2004; 18:1854–1856.
8. Welty TK, Bulterys M, Welty ER, Tih PM, Ndikintum G, Nkuoh G,
et al. Integrating prevention of mother-to-child
HIV transmission into routine antenatal care: the key to program expansion in Cameroon. J Acquir Immune Defic Syndr 2005; 40:486–493.
9. Coetzee D, Hilderbrand K, Boulle A, Draper B, Abdullah F, Goemaere E. Effectiveness of the first district-wide programme for the prevention of mother-to-child transmission of
HIV in
South Africa. Bull World Health Org 2005; 83:489–494.
10. Stringer JSA, Sinkala M, Maclean CC, Levy J, Kankasa C, De Groot A,
et al. Effectiveness of a city-wide program to prevent mother-to-child
HIV transmission in Lusaka, Zambia. AIDS 2005; 19:1309–1315.
11. South African National Department of Health. Protocol for Providing a Comprehensive Package of Care for the Prevention of Mother to Child Transmission of
HIV (PMTCT) in
South Africa. Pretoria: National Department of Health; 2001.
12. Doherty T, Besser M, Donohue S. An Evaluation of the PMTCT Initiative in
South Africa. Durban: Health Systems Trust; 2003.
13. Alioum A, Cortina-Borja M, Dabis F, DequaeMerchadou L, Haverkamp G, Hughes J,
et al. Estimating the efficacy of interventions to prevent mother-to-child transmission of human immunodeficiency virus in breastfeeding populations: comparing statistical methods. Am J Epidemiol 2003; 158:596–605.
14. Alioum A, Dabis F, Dequae-Merchadou L, Haverkamp G, Hudgens M, Hughes J,
et al. Estimating the efficacy of interventions to prevent mother-to-child transmission of
HIV in breast-feeding populations: development of a consensus methodology. Stat Med 2001; 20:3539–3556.
15. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994; 81:515–526.
16. Hogan JW, Roy J, Korkontzelou C. Handling drop-out in longitudinal studies. Stat Med 2004; 23:1455–1497.
17. South African Department of Health, Medical Research Council and Measure DHS.
South Africa Demographic and Health Survey 1998, Full Report. Pretoria: Department of Health; 2002.
18. Petra Study Team.
Efficacy of three short-course regimens of zidovudine and lamivudine in preventing early and late transmission of HIV-1 from mother to child in Tanzania, South Africa, and Uganda (Petra study): a randomized, double-blind, placebo-controlled trial.
Lancet 2002;
359:1178–1186.
19. Equity Project.
Primary Health Care in the Eastern Cape Province 1997–2000. Bisho: Equity Project and Eastern Cape Department of Health; 2001.
20. Painter TM, Diaby KL, Matia DM, Lin LS, Sibailly TS, Kouassims MK,
et al. Sociodemographic factors associated with participation by
HIV-1 positive pregnant women in an intervention to prevent mother-to-child transmission of
HIV in Côte d'Ivoire. Int J STD AIDS 2005; 16:237–242.
21. Leroy V, Sakarovitch C, Cortina Borja M, McIntyre J, Coovadia H, Dabis F,
et al. Is there a difference in the efficacy of peripartum antiretroviral regimens in reducing mother-to-child transmission of
HIV in Africa? AIDS 2005; 19:1865–1875.
22. McCoy D, Chopra M, Loewenson R, Aitken J, Ngulube T, Muula A,
et al. Expanding access to antiretroviral therapy in sub-Saharan Africa: avoiding the pitfalls and dangers, capitalizing on the opportunities. Am J Public Health 2005; 95:18–22.
23. Peeling RA, Mabey D, Fitzgerald D, Watson-Jones D. Avoiding
HIV and dying of syphilis. Lancet 2004; 364:1561–1563.
24. Blecher M, Thomas S. Health care financing. In
South African Health Review 2003/4. Edited by Injumba P, Day C, Ntuli A. Durban: Health Systems Trust, 2004: 269–290.
