The HIV epidemic threatens to undermine the hard-won gains in child mortality in sub-Saharan Africa [1–3]. Whereas the impact of the epidemic on adult mortality in Africa is well documented [4–7], impacts on child mortality are harder to measure. Two community-based studies [8,9] and clinical data from follow-up of post-natal cohorts [10–14] suggest substantial mortality in children born to HIV-positive mothers.
Whereas infection may directly affect child mortality, uninfected children may also experience higher mortality if they are orphaned or if family resources are diverted to care for HIV-infected parents . Thus, studies seeking to measure the impact of the HIV epidemic on child mortality have to account for both the direct effects due to child infection and indirect effects due to parental mortality. The direct effects are difficult to assess, partly because of the presence of transmitted maternal antibodies in children of seropositive mothers that make antibody results in young children unreliable . Polymerase chain reaction (PCR) tests  are too expensive for routine use in large epidemiological studies especially since they must be repeated to capture breast feeding transmission .
Verbal autopsy data on childhood deaths are less sensitive than in adults [19,20], because of high background mortality , and the difficulty in distinguishing other causes of death, such as infectious diseases and malnutrition, that are common in the region  from HIV.
In the absence of vital registration systems, African countries rely on indirect estimation of child mortality using mothers’ responses to census questions about the survival of their children [23,24]. If mortality is higher among orphans  these methods under-estimate mortality because of the high correlation between maternal and child mortality.
We used data from a well-characterized longitudinal study cohort followed for 11 years (1989–2000) by the Medical Research Council (MRC) Programme on AIDS in Uganda to study the relationships between HIV status of mothers, their mortality, and the mortality of their children.
The MRC General Population Cohort resides in 15 neighbouring villages in Masaka district, Southwest Uganda and consists of about 10 000 people who have been studied since 1989. Detailed description of the study population is given elsewhere . In brief, after each annual census in which all residents including in-migrants and births since the last survey round are recorded, survey staff obtain informed consent to interview and collect a blood sample from adults (13+ years old). From 1991, local leaders were asked to record births and deaths in their villages to supplement annual census data.
Blood specimens are tested at the Uganda Virus Research Institute laboratory in Entebbe for HIV-1 antibodies using two enzyme linked immunoassays (EIAs): Recombigen HIV-1/2 EIA (Trinity Biotech, Wicklow, Ireland) and Wellcozyme HIV-1 Recombinant EIA (Murex Biotech Ltd, Dartford, UK) following set algorithms  and rigorous quality control procedures . First time seropositive and discordant EIA results were confirmed by HIV-1 western blot (Calypte Biomedical Corporation, Rockville, Maryland, USA).
Records for all resident adult females seen between rounds 1 and 11 were analysed, linking children born in this period to their mothers. Follow-up time for each child was defined as time from birth to death, leaving the study area, or end of study period, whichever was earliest.
Maternal HIV status at birth of a child was determined using all available HIV test results. Figure 1 shows that women were classified as HIV negative at all times prior to their last negative test, positive at all times after a first positive test, and as sero-converters in the interval between last negative and first positive test. The interval following the last negative test is described as ‘post-negative’ and initial exposure in women who never had a negative test is described as ‘pre-positive'. Mothers who never had an HIV test have been categorized as ‘never tested’ throughout. For analytical purposes, the intervals ‘post-negative', ‘pre-positive’ and ‘sero-converters’ were grouped into one ‘ambiguous’ category since the HIV status of births during these intervals are unknown. Children were classified according to their mother's HIV status at birth, a classification that never changed during the child's lifetime.
Maternal survival status was divided into four categories. A mother who survived to the last study round, or who migrated away from the study area was classified as ‘living’ for the whole period in which she was under observation. Mothers who died were also classified as ‘living’ up to 1 year before their date of death, ‘terminally ill’ from a date 1 year before their death to the date of death, and as ‘recently dead’ from their date of death to 1 year later. After that, they were classified as ‘dead for over a year'. Children whose mothers died were exposed to changing maternal survival status. A 1-year period was chosen to represent terminal illness as this approximately covers the median survival time between the appearance of AIDS symptoms and death in this population .
Life table survivorship probabilities and risk of dying were summarized at single years of age based on the Kaplan–Meier survival function, which allows for censoring. The proportion of live births dying by age 1 year is the infant mortality rate (IMR), and by age 5 years is the child mortality rate (CMR). Life tables were also constructed for sub-populations, some based on attributes fixed at birth, such as the child's sex, others, such as mother's co-residence, varied over time. For fixed characteristics, exposure time starts from birth for all children, and finishes at exit from the study. For time variant characteristics, exposure time may start at different ages, and may finish when the child ceases to be classified in a particular category.
A piecewise exponential hazards model was the basis of the statistical analyses. This assumes that forces of mortality were constant and independent for any segment of time in which the independent risk variables did not change. Whereas the life table functions were expressed as risks or probabilities, hazard ratios (HRs) were based on central mortality rates calculated for the segments of time when the risk variables assumed their specified values. Although both univariate and multi-variate results are presented, some of the risk variables can only be interpreted when the confounding effects of others have been fully allowed for, the most important confounder being the child's age.
Other potential child mortality risk factors included calendar year, co-residence of mother and child, mother's age, whether the child was a singleton or one of twins, and child's sex .
Of the 3727 children born between 1989 and 2000, 415 (11%) died, 716 (19%) left the study area, and 2596 were still alive in 2000. The HIV status of the mother at birth of the child was ascertained for 3004 (81%) children. Of these 2786 were children born to mothers who subsequently tested negative for HIV, and 218 (6%) were born to mothers who were HIV-positive on a previous round. There were 723 children in the study whose mother's exact serostatus at birth could not be ascertained, either because the mother was never tested (271 births), or tested negative for HIV prior to birth (365 births), or seropositive after birth (59 births), or the birth occurred during a seroconversion interval (28 children). Of the 716 children who left the study area, 39 (5%) were born to HIV-positive mothers, 433 were born to HIV-negative mothers and 244 were born to mothers with unknown status at birth.
Table 1 shows numbers of children ever exposed for each category of risk factor, the child-years of exposure, the death rate, and compares observed mortality rates for each variable category. Whereas child-years of exposure always sum to 42 110 across variable categories, the total number exposed across time-variant categories can be greater than the total number of births (3727) since children may be exposed to more than one condition during the course of their lives.
Mortality declined rapidly with age, from infancy to age 3 years and older. It was higher among boys, twins, in the 1989–1995 period compared to girls, singletons, and 1996–2000. Children of teenage mothers had higher mortality whereas mother's absence appeared to be protective.
Maternal and child mortality were highly correlated. In the year leading up to and following a mother's death, children experienced mortality rates that were five times higher than those with ‘living’ mothers. Only 43 children survived for more than a year after their mother's death, but no further deaths were observed in 97 child-years of observation accumulated.
Mortality rates of children born to HIV-positive mothers and those born during the interval when their mothers sero-converted were four times those born to HIV-negative women. Children of ‘post-negative', ‘pre-positive’ and ‘never tested’ women experienced intermediate levels of mortality.
The first panel of Table 2 shows the life table cumulated proportions of children dying by exact ages, categorized by the HIV status of the mother at birth of the child. IMR among children born to HIV-positive mothers was 225 per thousand live births, [95% confidence interval (95% CI), 174–289], and CMR was 313 per thousand (95% CI, 251–385). The corresponding IMR and CMR for children of HIV-negative mothers were 53 (95% CI, 45–62) and 114 (95% CI, 101–128), respectively. Intermediate risks were observed for children born to mothers who had never been tested and for those with ambiguous HIV status.
The second panel of Table 2 shows cumulated mortality risks for mothers up to 5 years after birth of their children. Mortality risks for mothers are much lower than those of their children, and maternal deaths do not cluster in the year immediately after birth, as is the case with infant deaths. The cumulative mortality of HIV-positive mothers (all of whom were already infected at childbirth) only overtook that of their children by age 6 although only 25 to 40% of their children were themselves likely to become HIV infected . By age 9, one-third of the children of infected mothers were dead compared to over half of the mothers.
A multivariate hazard analysis is presented in Table 3. Two models are presented: the first includes all the variables from the univariate analysis, treating them as independent factors. The second includes only the significant factors from model 1, groups categories that are indistinguishable in terms of HR, and adds significant interaction terms.
According to model 1, age remained independently associated with child mortality; but the level of significance diminished after 3 years of age. Girls had insignificantly lower mortality than boys whereas multiple births had twice the hazard of singleton births. Both mother's absence and having a teenage mother increased mortality risk by about 70%. Hazards associated with a mother's terminal illness and death were significantly higher in the year immediately following the death of the mother (HR = 5.5) than the year before the mother's death (HR = 3.2). The excess hazard associated with having an HIV-positive mother were somewhat reduced (from 3.7 to 3.2) when maternal mortality was allowed for.
Interactions between maternal mortality and HIV status, and between child's age and maternal HIV status and survival were not significant. However, an interaction was found between maternal HIV status and child's age. If we allow for this interaction, as shown in model 2, the HR for maternal HIV infection in infancy rises to 4.43 (= 1.69 × 2.62); at subsequent ages it falls to 1.69. Including this interaction term in the regression model caused only a small change in the HRs, increasing HRs for ages 2 and 3+, and the HRs associated with maternal death.
Figure 2 illustrates the independent effects of HIV status and maternal death at each age, showing separate curves for children born to HIV-positive and HIV-negative mothers, sub-classified by maternal survival. Figure 2a shows the age-specific death rates faced by the children plotted at the mid-point of each single-year age group. The low baseline hazards at ages 3+, and the small numbers of children with HIV-positive and/or dead mothers followed up at these ages, make it difficult to draw firm conclusions concerning the increased hazards faced by children in these categories at later ages. Indeed, confidence limits on the rates for the high-risk categories at ages 3+ are so wide, that the differences are not statistically significant.
Figure 2b shows the cumulated proportion dying from the original birth cohort, with children classified by whether the mother has survived or died by the end of the age interval. The cumulative effect of maternal HIV status constitutes a hazard for child survival almost equal to that posed by a mother's death. This runs counter to the impression given by the relative size of the HRs presented in Table 3. Children whose mothers eventually died had a few years of relatively low mortality prior to their mothers’ terminal illnesses, and if they survived the years immediately following the deaths of their mothers, their mortality hazards reverted to the levels experienced by children whose mothers remained alive. By contrast, children infected with HIV in utero or through breastfeeding can be expected to experience high mortality rates at all ages post-infection , without reversion to the mortality patterns of the children of HIV-negative women. Since the highly differentiated infant death rates contribute to the cumulated death rate at all ages, the statistical significance of the differences from the low-risk category is maintained up to age 5 years.
Mothers’ HIV and survival status were independently associated with child mortality rates in this population. Our study with 14 110 child-years of observation and 11 years of follow-up is the largest and longest to study this relationship. Our results are consistent with those from a much shorter study  from the Rakai population in which, children of HIV-positive women who were aged less than 2 years experienced a 2.8-fold increase in mortality, compared to 3.9 by age 2 years in our study. The overall HR for deaths to children of all ages for HIV-positive mothers in comparison with children of HIV-negative mothers was 3.2 but the numbers at risk are small making the population level effects smaller than might be expected.
The high child mortality risks associated with maternal death could arise because household resources are channelled into care of the mother, or the household may be facing reduced economic circumstances because of the earlier illness and death of the child's father. Elevated mortality could also be a result of forced weaning, due to malnutrition or diarrhoeal disease associated with cessation of breast-feeding. A contributory factor may be the increased likelihood of HIV transmission during weaning that has been reported [33–35], especially during periods of mixed feeding.
The hazards analysis showed that maternal HIV status and maternal survival were highly correlated but had independent effects. This is somewhat surprising as transmission is thought to increase for mothers with advanced disease . The absence of an interaction does not preclude a relationship between viral load and HIV transmission, since viral load, is high immediately following infection as well as in the advanced stages of the disease. Mothers with advanced HIV illness, although likely to transmit infection because of a high viral load, are also less likely to become pregnant [37–39], so relatively few children are born at this stage.
Maternal HIV status, but not maternal mortality, showed an interaction with child's age, with a much stronger HIV hazard in infancy. As HIV-infected children are bound to become an ever-diminishing proportion of the cohort of children born to HIV-infected mothers, we might expect the HR to continue to decline, even though the number of child-years observed in our study do not allow us to capture such a trend. However, clinical data  show that the mortality of those HIV-positive children who survive infancy increases at older ages, in the same way that mortality in adults increases with time since infection. These two effects (declining proportion of infected children and increasing mortality among the infected) could interact to produce a constant age specific HR at ages over 1 year.
Mother's absence appeared to be protective in univariate analysis, because of age confounding; infants, who have very high mortality rates are rarely separated from their mothers. When age was controlled for, mother's absence was associated with increased child mortality. Deaths of absent mothers would not be recorded in the surveillance system, and HIV-positive mothers were more likely to leave the study area compared to those who were HIV negative (odds ratio, 1.4; P = 0.008), because of the association of HIV with mobility. However, the association of child mortality with mother's absence remained when mothers’ HIV status was controlled for. Although HIV-positive mothers were more likely to leave the study area, the proportion of children born to infected mothers leaving the area (5%) was slightly lower than the overall proportion of children born to infected mothers (6%). The mean age at which the children born to HIV-positive mothers left the study area was 2.5 years, which was not significantly different from the age at which other children left, but considerably older than the mean age at death (0.9 years) of those children who died. It is therefore unlikely that the results are biased by a disproportionate amount of unobserved deaths occurring among those whose survival status cannot be ascertained because they left the area.
The high mortality of children of teenage mothers in the univariate analysis could be due to age confounding, since a high proportion of these children would be infants. However, when child's age was controlled for, this high child mortality remained.
Increases in neo-natal mortality associated with HIV have been described  but we could not distinguish neonatal from post-neonatal mortality, because dates of birth and death are only accurate to the nearest month.
As the majority of infants in this study have not been tested for HIV, we are unable to provide associations and estimates based on individual children's HIV status, or estimate the expected reduction in numbers of infected children from use of antiretrovirals during pregnancy or birth . Changes in intrapartum transmission could affect the mortality rate ratio for children classified by maternal HIV status at birth and the survival patterns of HIV-infected children through a shift in timing of acquisition of infection.
The very high mortality of mothers who die within a few years of giving birth suggests that simply reducing vertical transmission might not proportionately reduce the mortality risks in children of infected mothers . There is an urgent need for support services to be provided for families in which mothers are terminally ill, or have recently died to forestall further deaths occurring among the children.
The majority of child deaths are in children of HIV-negative mothers and the majority of such deaths are preventable with currently available public health interventions. In the present study 65% of all child deaths were children of HIV-negative mothers, and 42% of deaths of children whose mothers died were children whose mothers were HIV-negative at birth, or of unknown status. Amongst surviving maternal orphans, 37% had had an HIV-positive mother, 32% had mothers who had been HIV negative, 31% had mothers whose HIV status was unknown. The high mortality risks faced by orphans in general, and the large number of orphans whose mothers were not HIV positive suggest that programmes directed at improving HIV-related child mortality should be provided in a frame-work of integrated management of childhood illness, and not merely targeted at HIV-infected families.
The authors thank the population of the study area, the field and laboratory staff, the Director of the Uganda Virus Research Institute, and Director-General of Medical Services, Ministry of Health, Uganda.
Sponsorship: Work on this part of the study was supported by a grant from UNICEF. The study cohort is funded by the Medical Research Council (MRC) of the United Kingdom, and the Department of Foreign and International Development (DFID) of the United Kingdom.
1. UNAIDS. Children and Young people in a World of AIDS
. Geneva: UNAIDS; 2001.
2. Mirsky J, Radlett M, AIDS and Children: A Family Disease. WorldAIDS 1989. London:
PANOS Institute; 1989, p. 6.
3. Walker N, Swhartlander B, Bryce J. Meeting international goals on child survival and HIV-AIDS
4. Mulder DW, Nunn AJ, Nakiyingi J, Wagner HU, Kengeya-Kayondo JF. Two Years HIV-1-associated mortality in a Ugandan rural population.Lancet
5. Nunn AJ, Mulder DW, Kamali A, Ruberantwari A, Kengeya-Kayondo JF, Whitworth J. Mortality associated with HIV infection over Five years in a rural Ugandan population: cohort study.BMJ
6. Sewankambo NK, Wawer MJ, Gray RH, Serwadda D, Li C, Stallings RY, et al
. Demographic impact of HIV infection in rural Rakai District, Uganda: results of a population-based cohort study.AIDS
7. Boerma JT, Ngalula J, Isingo R, Urassa M, Senkoro KP, Gabone R, et al
. Levels and causes of adult mortality in rural Tanzania with special reference to HIV/AIDS.Health Trans Rev
8. Sewankambo NK, Gray RH, Ahmad S, Serwadda D, Wabwire-Mangen F, Nalugoda F, et al
. Mortality associated with HIV infection in Rakai district, Uganda.AIDS
9. Urassa M, Boerma JT, Isingo R, Ngalula J, Ng'weshemi J, Mwaluko G, et al
. The impact of HIV/AIDS on mortality and households in rural Tanzania.AIDS
10. Taha ETT, Dallabetta GA, Canner JK, Chiphangwi JD, Liombo G, Hoover DR, et al
. The effect of human immunodeficiency virus on birthweight, infant and child mortality in urban Malawi.Int J Epidemiol
. 1995, 24
11. Ryder RW, Nsuami M, Nsuami W, Kamenga M, Badi N, Utshudi M, et al
. Mortality in HIV-1 seropositive women, their spouses and newly born children during 36 months of follow-up in Kinshasa.AIDS
12. Marum LH, Tindyebwa D, Gibb B. Care of children with HIV infection and AIDS in Africa.AIDS
(Suppl B): S125–S134.
13. de Cock K, Zadi F, Adjorlolo G, Diallo MO, Sassan-Morokro M, Ekpini E, et al
. Retrospective study of maternal HIV-1 and HIV-2 infections and child survival in Abidjan, Côte d'Ivoire.BMJ
14. Ryder R, Temmerman M. The effect of HIV_1 infection during pregnancy and the perinatal period on maternal and child health in Africa.AIDS
15. Bledsoe C, Ewbank D, Isiugo-Abanihe U. The effect of child fostering on feeding practices and access to health services in rural Sierra Leone.Soc Sci Med
16. Ades A, Newell M,
Peckham C. Children born to women with HIV-1 infection: natural history and risk of transmission. European Collaborative Study.Lancet
17. Rollins NC, Dedicoat M, Danaviah S, Page T, Bishop K, Kleinschmidt I, et al
. Prevalence, incidence, and mother-to-child transmission of HIV-1 in rural South Africa
18. Embree JE, Njenga S, Datta P, Nagelkerke NJD, Ndinya-Achola JO, Mohammed Z, et al
. Risk factors for postnatal mother-child transmission of HIV-1
19. Adult Mortality and Morbidity Project. Cause-specific adult mortality: evidence from community-based surveillance – selected sites, Tanzania, 1992–1998.Morb Mortal Wkly Rep
20. Kamali A, Wagner HU, Nakiyingi J, Sabiiti I, Kengeya-Kayondo JF, Mulder DW. Verbal autopsy as a tool for diagnosing HIV-related adult deaths in rural Uganda
. Int J Epidemiol
21. Boerma JT, Nunn AJ, Whitworth J. Mortality impact of the AIDS epidemic: evidence from community studies in less developed countries.AIDS
22. Timaeus I. Impact of the HIV epidemic on mortality in sub-Saharan Africa: evidence from national surveys and censuses.AIDS
23. Hill K. Approaches to the measurement of childhood mortality: a comparative review.Population Index
24. United Nations. Step-by-Step Guide to the Estimation of Child Mortality.
New York: Department of International Economic and Social Affairs, United Nations; 1990.
25. Ward P, Zaba B. The effect of HIV on the estimation of child mortality using the children surviving /children ever born technique.
Presented at meeting of IUSSP Scientific Committee on AIDS, Copenhagen, June 1999, Conference proceedings session 7.
26. Mulder DW, Nunn AJ, Wagner HU, Kamali A, Kengeya Kayondo JF. HIV-1 incidence and HIV-1-associated mortality in a rural Ugandan population cohortAIDS
27. Nunn AJ, Biryahwaho B, Downing RG, Van Der Groen G, Ojwiya A, Mulder DW. Algorithm for detecting antibodies to HIV-1: results from a rural Ugandan cohort.AIDS
28. Nunn AJ, Biryahwaho B, Downing RG, Ojwiya A, Mulder DW. Computer-assisted quality assurance in an HIV serology laboratory.Methods Inf Med
29. Morgan D, Malamba SS, Orem J, Mayanja B, Okongo M, Whitworth JA. Survival by AIDS defining condition in rural Uganda.Sex Transm Infect
30. Mosley WH, Chen LC. An analytical framework for the study of child survival in developing countries.
In: Mosley WH, Chen LC (editors). Child Survival: Strategies for Research
. New York: New York, Population Council; 1984.
31. Dabis, F , Msellati P, Dunn D, Lepage P,Newell ML, Peckham C, Van de Perre P. Estimating the rate of mother-to-child transmission of HIV. Report of a workshop on methodological issues Ghent (Belgium), 17-20 February 1992. The Working Group on Mother-to-Child Transmission of HIV.AIDS
32. Nduati R, John G, Mbori-Ngacha D, Richardson B, Overbaugh J, Mwatha A, Ndinya-Achola J, et al
. Effect of breastfeeding and formula feeding on transmission of HIV-1: a randomized clinical trial.JAMA
33. Van de Perre P, Simonon A, Msellati P, Hitimana DG, Vaira D, Bazubagira A, et al
. Postnatal transmission of HIV-1 from mother to infant.N Engl J Med
34. UNAIDS. HIV and Infant Feeding: A Policy Statement Developed Collaboratively by UNAIDS, WHO and UNICEF
. Geneva: UNAIDS; 1997.
35. Coutsoudis A, Pillay K, Kuhn L, Spooner E, Tsai W, Coovadia H. Method of feeding and transmission of HIV-1 from mothers to children by 15 months of age: prospective cohort study from Durban, South Africa.AIDS
36. John GC, Nduati RW, Mbori-Ngacha DA, Richardson BA, Panteleeff D, Mwatha A, et al
. Correlates of mother-to-child human immunodeficiency virus type 1 (HIV-1) transmission: association with maternal plasma HIV-1 RNA load, genital HIV-1 DNA shedding, and breast infectionsJ Infect Dis
37. Carpenter L, Nakiyingi J, Ruberantwari A, Malamba S, Kamali A, Whitworth J. Estimates of the impact of HIV infection on fertility in a rural Ugandan population cohort.Health Trans Rev
38. Gray RH, Wawer MJ, Sarwadda D, Sewankambo N, Li C, Wabwire-Mangen F, et al
. Reduced fertility in women with HIV_1 infection: a population_based study in Uganda.Lancet
39. Hunter S, Isingo R, Boerma JT, Urassa M, Ng'weshemi JZL, Zaba B, et al
. The association between HIV and fertility in a cohort study in rural Tanzania.J-Biosoc-Sci
40. Dunn D, Simonds R, Bulterys M, , Kalish LA, Moye J, de Maria A, et al
. Interventions to prevent vertical transmission of HIV-1: effect on viral detection rate in early infant samples.AIDS
41. Brocklehurst P, French R. The association between maternal HIV infection and perinatal outcome: a systematic review of the literature and meta_analysis.Br J Obstet Gynaecol
42. Dabis F, Leroy V, Castetbon K, Spira R, Newell ML, Salamon R. Preventing mother-to-child transmission of HIV-1 in Africa in the year 2000.AIDS
43. Gregson S, Zaba B, Garnett G. Low fertility in women with HIV and the impact of the epidemic on orphanhood and early childhood mortality in sub-Saharan Africa. AIDS
1999; 13 (Suppl A):S249–S258.