|Home | About | Journals | Submit | Contact Us | Français|
Few prospective studies have measured survival rates among human immunodeficiency virus (HIV)–infected children in sub-Saharan Africa prior to the availability of antiretroviral therapy.
In the context of an observational study of the immunogenicity of measles vaccine in Zambia, we prospectively followed up children from approximately 9 months of age and assessed survival rates, risk factors for mortality, and circumstances at the time of death according to HIV-infection or HIV-exposure status.
There were 56 deaths among 492 study children during follow-up to 3 years of age. Thirty-nine percent of the 105 children with HIV infection died during the study period, compared with 5.0% of the 260 HIV-seropositive but uninfected children and 1.6% of the 127 HIV-seronegative children. Estimated survival probabilities from 9 through 36 months of age were 52% among HIV-infected children, 95% among initially HIVseropositive but uninfected children, and 98% among HIV-seronegative children. In multivariable analyses, history of a clinic visit within the 4 weeks prior to study entry (adjusted hazard ratio, 4.6; 95% confidence interval, 1.5–13.5), hemoglobin level <8 g/dL at study entry (adjusted hazard ratio, 4.4; 95% confidence interval, 1.5–12.6), and CD4+ T lymphocyte percentage <15% at study entry (adjusted hazard ratio, 3.2; 95% confidence interval, 1.1–9.5) were associated with mortality among HIV-infected children.
Only approximately one-half of HIV-infected Zambian children who were alive at 9 months of age survived to 3 years of age, supporting the urgent need for the prevention of mother-to-child transmission of HIV and the early diagnosis and treatment of HIV infection in children in sub-Saharan Africa.
In 2007, an estimated 330,000 children died of AIDS, and an additional 2.5 million children were living with HIV infection or AIDS . Most of these children lived in sub-Saharan Africa . In 2003, the Joint United Nations Program on HIV/AIDS, the World Health Organization, and their partners made a commitment to provide antiretroviral therapy to 3 million people by the end of 2005. Although this target was not achieved, an increasing number of HIV-infected persons in subSaharan Africa have access to treatment, with the goal being to achieve universal access for those in need by 2010. Improving access to treatment for HIV-infected children has been slow , although encouraging results recently were reported among HIV-infected children in Lusaka, Zambia .
Because of limited access to care, poor nutritional status, and frequent exposure to endemic coinfections, HIV-infected children residing in sub-Saharan Africa have extremely high mortality rates. Approximately one-third of HIV-infected children in sub-Saharan Africa who have no access to antiretroviral therapy are estimated to die by 1 year of age, and approximately one-half are estimated to die by 2 years of age . However, these estimates are based on a limited number of observational and prospective studies . Longitudinal data on the survival of HIV-infected infants in sub-Saharan Africa are sparse, in part because the diagnosis of HIV infection in infancy requires the detection of HIV DNA or RNA using expensive PCR-based assays. In a prospective, observational study of the immunogenicity of measles vaccine in HIV-infected and uninfected children , we assessed survival rates from 9 months of age, risk factors for mortality, and circumstances at the time of death.
The study was conducted at Chawama Clinic, a public health care facility located in Lusaka, Zambia. At the time of the study, antiretroviral therapy was not readily available for the treatment of HIV-infected children in Zambia. Children 2–8 months of age who attended the clinic for routine vaccination from May 2000 through November 2002 and who had not been vaccinated against measles were eligible for screening for study entry. All children who were HIV seropositive, plus a random sample of HIV-seronegative children, were invited to participate . If caregivers gave written informed consent, they were asked to return to the clinic for measles vaccination of their child at 9 months of age. HIV testing was conducted anonymously, and caretakers were offered the opportunity to take their child for voluntary counseling and testing. At the time of measles vaccination (i.e., study entry), care-givers were interviewed using a standard questionnaire to obtain information on basic demographic characteristics and the child's medical history. The child was examined for signs of illness, and the child's length and weight were recorded. Blood was obtained by venipuncture, placed in EDTA tubes, and transported to the laboratory. We randomly assigned children to follow-up at 1 or 3 months after vaccination, and we asked mothers of all children to return at 15 and 27 months after vaccination and to seek care from the study team any time that the child was ill. Outpatient care for common childhood illnesses, including diarrhea and pneumonia, was provided by the study team. A blood sample was obtained by venipuncture at each scheduled follow-up visit. The study was approved by the University of Zambia, the London School of Hygiene and Tropical Medicine, and the Johns Hopkins University Bloomberg School of Public Health.
A community health worker visited the child's home if they failed to return for scheduled follow-up visits, and home visits were performed every 3 months until the child was at least 3 years of age to assess the child's vital status. The date of death was recorded for children who died. A verbal autopsy was performed by a study clinical officer using a standard verbal autopsy questionnaire  at the child's home ~3 months after the death of the child. Verbal autopsies were reviewed by 2 pediatricians who listed as many as 4 contributing causes of death per child. One of the authors (W.J.M.) assigned the causes of death on the basis of interpretations of the verbal autopsies.
Children who died before measles vaccination for whom verbal autopsies were performed were included in an analysis of the circumstances at death but were not included in the survival analyses, because the HIV infection status was not ascertained for all children who were not vaccinated against measles or those lost to follow-up prior to study entry.
Blood samples were tested for antibodies to HIV by rapid immunoassay (Determine HIV-1/2; Abbott Laboratories). HIV infection was determined in each HIV-seropositive child by detection of HIV RNA by RT-PCR (Amplicor HIV-1 Monitor, version 1.5; Roche Molecular Systems). Hemoglobin measurements were made with the QBC Autoread Plus System (QBC Diagnostics). Severe anemia was defined as a hemoglobin concentration <8 g/dL . The percentage of CD4+ T lymphocytes in freshly isolated PBMCs was measured by flow cytometry (Becton Dickinson FACScan) using Cell Quest software, version 1.2, and directly conjugated monoclonal antibodies to CD3 and CD4 (Becton Dickinson).
Data were double-entered, checked, and validated using EpiInfo, version 6.04 (Centers for Disease Control and Prevention). SAS, version 9.1 (SAS Institute), and Stata, version 9 (Stata), were used for analysis.
We classified the HIV infection or exposure status of children as HIV infected if HIV RNA was detected in any plasma sample, as HIV seropositive but uninfected if antibodies to HIV but no HIV RNA were detected, and as HIV seronegative if no antibodies to HIV were detected. Mothers of children who were HIV seropositive were assumed to be HIV infected. For HIV-seropositive children who became infected with HIV during follow-up, the timing of acquisition of HIV was estimated as the midpoint between the last visit at which the child had undetectable HIV RNA and the first visit at which the child was determined to be HIV infected on the basis of detection of HIV RNA. These children contributed person-time to both the HIV-seropositive and HIV-infected groups.
Weight-for-height, weight-for-age, and length-for-age Z scores were calculated with growth reference curves developed by the National Center for Health Statistics and Centers for Disease Control and Prevention . Z scores of less than −2 were considered to represent poor anthropometric status. Infants with poor length-for-age scores were considered to be stunted, those with poor weight-for-age score were considered to be underweight, and those with poor weight-for-length scores were considered to be wasted.
Survival after vaccination was evaluated using Kaplan-Meier survival curves stratified by HIV infection and exposure status. Age was used as the time axis, with staggered entry times, because children did not enter the study at the same age. The last contact with the study team (either a scheduled study visit or community health worker home visit) was used to estimate survival. For children for whom the exact date of death could not be ascertained, the date of death was estimated as the midpoint between the last study visit and either the date of the next missed visit, the date that the vital status was last ascertained, or the date of the verbal autopsy, depending on available data. Survival curves between groups were compared using the log-rank test.
Mortality rates over the 27-month follow-up period were calculated by HIV infection and exposure status, with person-time accumulating from the time of study entry (which was the date of measles vaccination for the majority of children) until death, loss to follow-up, or administrative censoring at the last study visit. We estimated 95% CIs for the proportion of children surviving to 24 and 36 months of age using Greenwood's formula. Risk factors for mortality were evaluated using Cox proportional hazards regression models.
From May 2000 through November 2002, 4748 mothers were invited to participate, and 2190 children were screened. Of these 2190 children, 696 were eligible for the study and had informed consent provided by their caregivers. Of the eligible children, 492 were vaccinated against measles at ~9 months of age (including 21 children who received measles vaccine outside the study and were not included in the published report on the immunogenicity of measles vaccine ), had known HIV infection status, and were included in the survival analysis (figure 1). Of these 492 children, 127 (26%) were HIV seronegative, 292 (59%) were HIV seropositive but uninfected at study entry, and 73 (15%) were HIV infected. Thirty-two (11%) of the children who were HIV seropositive but uninfected at study entry became HIV infected during follow-up; therefore, 105 children (21% of all study children) were HIV infected for at least part of the follow-up period.
The median age at study entry was 9.1 months (interquartile range, 9.0–9.3 months) and did not differ by HIV infection status (P = .4). Maternal mortality prior to study entry was low in all 3 groups of children, but paternal mortality was higher among HIV-seropositive but uninfected and HIV-infected children (i.e., higher among children born to HIV-infected women) (table 1). Almost one-half of mothers were <25 years of age, although the mothers of HIV-infected children tended to be older (table 1). A higher proportion of fathers of HIV-infected children and of seropositive but uninfected children had >7 years of education, compared with fathers of HIV-seronegative children (table 1). HIV-infected children were more likely to have a history of hospitalization, to have visited a clinic in the 4 weeks prior to study entry, to be ill (e.g., to have cough) at study entry, and to be stunted, wasted, or underweight (table 1). Hemoglobin levels and CD4+ T lymphocyte percentages were lower among children who were HIV infected at the time of study entry (table 1). Eighteen percent of HIV-infected children had a CD4+ T lymphocyte percentage <15%, which is a marker of severe immunodeficiency, and 14% of HIV-infected children were severely anemic (hemoglobin level, <8 g/dL) at ~9 months of age.
There were 56 deaths among the 492 study children during the follow-up period. Mortality was highest among HIV-infected children (figure 2). Forty-one (39.0%) of the 105 children who were infected with HIV by the end of the follow-up period died during the study period, compared with 13 (5.0%) of the 260 children who were HIV seropositive at study entry but remained uninfected and 2 (1.6%) of the 127 HIV-seronegative children. The crude mortality rate for HIV-infected children (290 deaths per 1000 person-years; 95% CI, 213–393 deaths per 1000 person-years) was >35 times higher (hazard ratio [HR], 36.3; 95% CI, 8.8–150; P < .001) than that among HIV-seronegative children (7.9 deaths per 1000 person-years; 95% CI, 2–32 deaths per 1000 person-years). The crude mortality rate among HIV-seropositive but uninfected children (25 deaths per 1000 person-years; 95% CI, 15–43 deaths per 1000 person-years) appeared to be higher than that among HIV-seronegative children (HR, 3.2; 95% CI, 0.7–14.0; P = .13). Mortality among HIV-infected children differed by the timing of infection. The crude mortality rate among children infected after vaccination (125 deaths per 1000 person-years; 95% CI, 52–301 deaths per 1000 person-years) was >3 times lower (HR, 0.3; 95% CI, 0.1–0.8) than it was among children infected at vaccination (354 deaths per 1000 person-years; 95% CI, 255–491 deaths per 1000 person-years).
Survival at 24 and 36 months of age was 69% (95% CI, 57%–78%) and 52% (95% CI, 41%–62%), respectively, among HIV-infected children, 95% (95% CI, 92%–97%) and 95% (95% CI, 91%–97%) among HIV-seropositive but uninfected children, and 99% (95% CI, 94%–100%) and 98% (95% CI, 93%–100%) among HIV-seronegative children. Among HIV-infected children, survival at 24 and 36 months of age was 65% (95% CI, 52%–75%) and 45% (95% CI, 32%–57%), respectively, among children infected with HIV at vaccination and was 90% (95% CI, 65%–97%) and 79% (95% CI, 56%–91%) among children infected with HIV after vaccination.
Characteristics of children at study entry that were associated in univariable analyses with mortality among HIV-infected children included prior hospitalization (HR, 1.8; 95% CI, 1.0–3.4), a clinic visit within the 4 weeks prior to study entry (HR, 4.3; 95% CI, 1.7–10.9), being underweight (HR, 2.0; 95% CI, 1.1–3.7), wasting (HR, 3.1; 95% CI, 1.2–8.1), hemoglobin <8 g/dL (HR, 3.1; 95% CI, 1.3–7.3), CD4+ T lymphocyte percentage of 15%–24% (HR compared with a CD4+ T lymphocyte percentage 25%, 3.0; 95% CI, 1.2–7.3), and CD4+ T lymphocyte percentage <15% (HR compared with CD4+ T lymphocyte percentage 25%, 3.0; 95% CI, 1.0–8.8) (table 2). In multivariable analysis, risk factors for mortality among HIV-infected children included history of a clinic visit within the 4 weeks prior to study entry (adjusted HR, 4.6; 95% CI, 1.5–13.5), underweight at study entry (adjusted HR, 2.1; 95% CI, 1.0–4.5), hemoglobin <8 g/dL at study entry (adjusted HR, 4.4; 95% CI, 1.5–12.6), CD4+ T lymphocyte percentage, 15%–24% at study entry (adjusted HR compared with a CD4+ T lymphocyte percentage 25%, 2.6; 95% CI, 1.1–6.6), and CD4+ T lymphocyte percentage <15% at study entry (adjusted HR compared with CD4+ T lymphocyte percentage 25%, 3.2; 95% CI, 1.1–9.5) (table 2).
Because there were few deaths among HIV-seronegative children and HIV-seropositive but uninfected children, these groups were combined to assess risk factors for mortality among HIV-uninfected children. Only CD4+ T lymphocyte percentage <15% was associated with mortality among the HIV-uninfected children (table 2), although few children were in this category (11 [5%] of the uninfected children). Seven of these 11 HIV-uninfected children were seropositive for HIV, 6 were underweight, and 1 was wasted.
Verbal autopsies were conducted for 48 (86%) of the 56 children who died after measles vaccination and for an additional 44 children who died prior to their scheduled vaccination visits. Three HIV-seropositive children for whom verbal autopsies were conducted were excluded from analysis, because their HIV infection status could not be confirmed. Among the remaining 89 children, verbal autopsies were conducted most frequently with a parent (52 children; 63%). Most children (81 children; 91%) had been seen by a health care worker during the illness preceding death, and 68 (76%) had been admitted to a health care facility, with no differences by HIV infection status of the child. However, the duration of final illness was significantly longer for HIV-infected children (median duration, 30 days; interquartile range, 7–153 days), compared with HIV-seropositive but uninfected children (median duration, 8 days; interquartile range, 4–21 days) and HIV-seronegative children (median duration, 7 days; interquartile range, 5–14 days; P = .02) (table 3).
Most children (73%) died in the hospital or another health care facility. Six HIV-infected children (10%) were reported to have died on the way to a health care facility, and a substantial proportion of children born to HIV-infected women (17%) died at home, compared with none of the children born to HIV-uninfected women (P = .35) (table 3).
Contributing causes of death were assigned for 80 children with known HIV infection status, including 6 HIV-seronegative children, 20 HIV-seropositive but uninfected children, and 54 HIV-infected children. The most common illnesses contributing to death were diarrhea (in 49 [61%] of cases), acute respiratory tract infection (39 cases; 49%), malnutrition (31 cases; 39%), tuberculosis (15 cases; 19%), meningitis (8 cases; 10%), and malaria (7 cases; 9%). Both HIV-seropositive and HIV-infected children were more likely than seronegative children to have acute respiratory tract infection as a contributing cause of death (in 1 [17%] of the HIV-seronegative children, 7 [35%] of the HIV-seropositive but uninfected children, and 31 [57%] of the HIV-infected children; P = .02, by test for trend). A similar trend was observed for malnutrition as a contributing cause of death (present in 1 [17%] of the HIV-seronegative children who died, 5 [25%] of the HIV-seropositive but uninfected children who died, and 25 [46%] of HIV-infected children who died; P = .05, by test for trend).
More than 1 illness was assigned as a contributing cause of death for two-thirds of children (54 children). HIV-infected children had more illnesses that contributed to death than did HIV-uninfected children (P = .03). Specifically, HIV-infected children were more likely to have diarrhea, acute respiratory tract infection, and malnutrition contribute to death than were HIV-uninfected children (P = .04) (figure 3).
HIV-infected Zambian children who survived to 9 months of age had a crude mortality rate of 290 deaths per 1000 person-years, which was <35 times that among HIV-seronegative children, and approximately one-half died by 3 years of age. This may have been an underestimate of mortality, because many children were lost to follow-up, especially at the 15-month and 27-month scheduled visits. Although community health workers tried to ascertain reasons for losses, deaths may have been missed. Our results are consistent with those of a study performed in Malawi, which examined mortality among children who survived the first year of life. Seventy percent of HIV-infected children were alive at 2 years of age, and 55% were alive at 3 years of age . These findings support the urgent need for the prevention of mother-to-child transmission of HIV and the early diagnosis and treatment of HIV-infection in children in sub-Saharan Africa.
Whether HIV-uninfected children born to HIV-infected women have higher mortality than do children born to HIVuninfected women remains unclear. We and others have found a tendency for increased mortality among uninfected children born to HIV-infected women, although, as in our study, the increase was not always statistically significant [10, 11].
We identified immune suppression, anemia, and illness requiring a clinic visit within 4 weeks after study entry as independent risk factors for mortality among HIV-infected children. Immune suppression, as measured by percentage or number of CD4+ T lymphocytes, is a well-established predictor of mortality among HIV-infected children in sub-Saharan Africa [2, 12-14]. In a previous study involving HIV-infected Zambian children, low hemoglobin level (<8 g/dL) also was identified as an independent predictor of mortality .
Few studies have determined the immediate cause of death among HIV-infected children in sub-Saharan Africa, but reported causes of death were similar to those reported for HIV-uninfected children and included diarrhea and pneumonia [10, 15, 16]. Using verbal autopsies, we found the duration of illness to be longer and the combination of diarrhea, acute respiratory tract infection, and malnutrition to be more common during the final illness among HIV-infected children than among uninfected children. A combination of diarrhea, pneumonia, failure to thrive, and neurological abnormalities was identified as a marker of rapid disease progression and death among HIV-infected children in South Africa . The multiple contributing causes of death, among HIV-infected children in particular, highlight the limitations of verbal autopsies in assigning a single cause of death.
Even in the absence of nucleic acid–based tests to diagnose HIV infection during early infancy, a substantial reduction in mortality could be achieved by diagnosing and treating HIV infection in children at 9–12 months of age. Affordable and accessible diagnostic tests for HIV infection during early infancy have the potential to further reduce mortality .
We thank the children and their parents for participation in the study and the clinical officers and nurses who assisted with the study.
Financial support. Wellcome Trust-Burroughs Fund Infectious Disease Initiative (GR059114MA) and the National Institutes of Health (AI23047). C.G.S. was supported by a Doctoral Research Award in the Area of HIV/AIDS Research from the Canadian Institutes of Health Research. T.C.Q. was supported by the Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health.
Potential conflicts of interest. All authors: no conflicts.