This study was conducted at Nyanza Provincial General Hospital (NPGH), a large publicly funded hospital in Kisumu (population 300,000) in western Kenya. Malaria transmission within Kisumu is perennial, and P. falciparum is the predominant species, accounting for 98% of malaria cases (19). Chloroquine resistance is prevalent in the area: 75%–80% of P. falciparum strains show a RII/RIII resistance pattern (20). The prevalence of HIV infection among pregnant women is approximately 25% (6,21).

Pregnant women were enrolled from June 1996 through May 2000. Screening procedures have been described previously (21). Briefly, women at the antenatal clinic were eligible for participation if they had an uncomplicated singleton pregnancy of >32 weeks’ gestation (based on the fundal height estimation), resided within the Kisumu municipality, and had no known underlying chronic illness. Following informed consent, a structured questionnaire was administered in the local language (Dholuo or Kiswahili) to obtain information on sociodemographic, health, and obstetric factors.

A trained HIV counselor then counseled each woman, and a posttest counseling appointment was made. A blood sample was taken for HIV antibody testing, hemoglobin level, and malaria thick blood film. All screened women were encouraged to deliver at NPGH. In addition, nonscreened women who delivered at NGPH were eligible for participation if they met study inclusion criteria. Routine use of zidovudine or nevirapine was not the Kenyan Ministry of Health policy during the study period, and these drugs were not available in Ministry of Health facilities.

At delivery, information was collected on mode and outcome of delivery and any illness and treatment in the previous 2 weeks. Within 24 hours of birth, infants were weighed (±1 g) on an electronic balance (Ohaus, Florham Park, NJ), and gestational age was assessed by using the Ballard method (22). All live, singleton, vaginally delivered infants of HIV-positive mothers were eligible for study. Infants were seen monthly until 12 months of age.

Informed consent was obtained from all women before they were enrolled in this study. Human subjects guidelines of the Centers for Disease Control and Prevention (CDC) and the Kenya Medical Research Institute ethical review committee were strictly followed. Mothers of enrolled infants signed an additional informed consent form for infant participation.

The study protocol was approved in 1995 by the institutional review boards of the Kenya Medical Research Institute; CDC, Atlanta, Georgia, USA; and the Academic Medical Center (AMC), University of Amsterdam, Amsterdam, The Netherlands, and was reviewed annually by the participating institutions. This study occurred during a changing environment in preventing perinatal HIV transmission, with results of studies in other countries demonstrating the benefits of short-course AZT use (23,24) or nevirapine use (25) in late pregnancy and perinatally to reduce the risk for MTCT. Similarly, studies of malaria prevention in pregnancy during the study interval were demonstrating the benefit of intermittent antimalarial treatment to prevent maternal anemia and low birth weight (6,26). The participating institutions and the Kenyan Ministry of Health were engaged in discussions about these findings regarding the ethical considerations for this investigation. The study continued with full ethical approvals during these discussions. Investigators and institutions have supported the transition to a system that supports providing intermittent preventive antimalarial therapy during pregnancy, according to the adopted Kenyan national policy. The MTCT prevention program now offers HIV counseling and testing, community education, and antiretroviral drugs to prevent MTCT in the study setting (4).

At delivery, maternal peripheral and placental thick blood films were prepared, stained with 10% Giemsa, and examined under oil immersion for malaria parasites. Placental blood was obtained by cutting into the maternal side of the placenta and placing collected blood on a slide. A thick film was considered negative if 100 microscopic fields showed no parasites. Malaria parasites and leukocytes were counted in the same fields until 300 leukocytes were counted. Parasite densities were estimated by using an assumed count of 8,000 leukocytes/μL blood.

Blood samples were collected into EDTA tubes. At delivery, blood was collected from mothers of enrolled infants to assess viral load, whether syphilis was present, and hemoglobin level. One month postpartum, maternal venous blood was collected to determine counts of CD4- and CD8-positive T-lymphocytes (CD4+ and CD8+). Capillary blood was collected from infants by a heel prick on the day of delivery and then monthly thereafter for HIV testing. Plasma was separated from blood samples and stored at –70°C.

Infant HIV testing was done by polymerase chain reaction (PCR) of proviral DNA extracted from peripheral blood mononuclear cells (27). HIV testing of pregnant women used two rapid test methods: an initial Serostrip HIV-1/2 (Saliva Diagnostic Systems Pte Ltd, Singapore) and a confirmatory Capillus HIV-1/HIV-2 (Cambridge Diagnostics, Wicklow, Ireland) on all Serostrip-positive samples. Sequential testing of samples using both methods has a high sensitivity and specificity (Richard W. Steketee, pers. comm.). Western blot was performed on discordant samples.

Maternal CD4+ and CD8+ counts were assessed by using commercial, dual-label monoclonal antibodies (Becton-Dickinson Immunocytometry, San Jose, CA) and standard fluorescent-activated cell sorting (FACScan, Becton-Dickinson) analysis following whole-blood lysis (28). Maternal HIV-1 viral load was determined by the Roche Amplicor HIV-1 monitor test version 1.0 (Roche Diagnostics, Indianapolis, IN), with a quantification limit of 400 viral copies per milliliter.

Syphilis antibodies were detected by venereal disease research laboratory (VDRL) slide test (EUROTEX-VDRL, Euromedi equip ltd., West Harrow, UK). Hemoglobin was measured to the nearest 0.1 g/dL using a Hemocue machine (Mission Viejo, CA).

An uncomplicated pregnancy was defined as a pregnancy without the presence of AIDS-defining symptoms, hypertension, preeclampsia, polyhydramnios, abnormal fetal presentation, history of a cesarean section, hemorrhage, or repeated spontaneous abortions (>2). Placental parasitemia was defined as any plasmodial asexual form detected on a thick film. Maternal HIV infection was defined as a positive result on both rapid tests; women not reactive with the initial Serostrip HIV-1/2 test were considered HIV uninfected. Women whose serostatus could not be determined (i.e., those with discordant results on the two rapid tests and an indeterminate status with Western blot) were excluded from analysis. Newborns were classified as normal birth weight if they weighed >2,500 g, regardless of gestational age, and low birth weight if they weighed <2,500 g. Preterm delivery was defined as occurring at <37 weeks of gestation. Small for gestational age was defined as sex-specific birth weight <10th percentile for weight-for-gestational-age (29).

An algorithm was developed to describe the perinatal HIV infection status of infants. Infants were classified as having acquired HIV infection perinatally (e.g., in utero or during labor and delivery) if they met the following conditions: 1) died and by the time of death had >2 consecutive positive HIV PCR tests, the first of which was at <4 months of age; 2) were lost to follow-up but had had >3 consecutive positive PCR results, the first of which was at <4 months of age; 3) remained alive with continued and consistent positive PCR results, the first of which was at <4 months of age. Infants were considered negative for perinatally acquired HIV if they had >3 PCR tests performed on them, and all tests were negative, with at least one of the negative tests at >4 months of age. Because infants had to be >4 months old to be classified as uninfected, those who died or were lost to follow-up before 4 months of age were excluded. Infants who acquired HIV at >5 months of age were considered to have acquired HIV postnatally and were included as nontransmitters from the perinatal perspective. Infants for whom we had insufficient PCR data to determine their status were classified as indeterminate and were excluded. Mothers of infected infants were classified as transmitters and those of uninfected infants as nontransmitters.

To evaluate the effect of placental malaria on perinatal MTCT, a Poisson log-linear model containing placental malaria and maternal viral load as primary predictor variables was constructed by using backward elimination; adjusted RRs were computed. An interaction term between placental malaria density and maternal viral load was significant (p = 0.02) and was retained in the model. Because women with and without placental malaria may have different risk factors for perinatal MTCT, we fit three separate multivariate models: all study women, women without placental malaria, and women with placental malaria. All tests were two-sided; p values <0.05 were considered significant. Analysis was done using STATA (StataCorp. 2001. Stata Statistical Software: Release 7.0 College Station, TX) and SAS (Version 8.0, SAS Institute, Cary, NC).

A total of 829 mother-infant pairs were enrolled; 317 (38.2%) infants with incomplete follow-up or indeterminate HIV status were excluded, leaving 512 mother-infant pairs. Included and excluded women did not differ in age, level of education, mean maternal viral load, mean CD4+ counts, malaria rates, high-density malaria rates, proportion of low birth order (i.e., gravida <2 versus >3), and rates of episiotomy or perineal tear. However, excluded women were more likely to have infants born with low birth weight (9.1% versus 5.5%, p = 0.04). The baseline characteristics of the 512 women included in the analysis are shown in Table 1. Among these 512 women, 128 (25.0%) had placental malaria and 353 (84.4%) had anemia (hemoglobin <11 g/dL).

Of 512 mothers, 102 (19.9%) transmitted HIV to their infants perinatally. High maternal viral load (>10,000 viral copies/μL) and low CD4+ count (<200 cells/μL) were significantly associated with perinatal MTCT (Table 2). When analyzed as continuous variables, transmitting mothers had higher geometric mean viral loads than nontransmitting mothers (7,083 versus 1,378 copies/μL; p<0.001) and low mean CD4+ count (511 versus 657 cells/μL; p<0.001). Other characteristics significantly associated with perinatal MTCT included episiotomy or perineal tear, low birth weight or small-for-gestational-age infants, and being of low birth order (primigravid or secundigravid).

Maternal peripheral parasitemia at delivery was not associated with MTCT; however, placental malaria was associated with a 40% reduction in the risk for perinatal MTCT (RR 0.6, 95% CI 0.4 to 1.0, p = 0.05) (Table 2). Compared to women without malaria (perinatal MTCT rate = 21.9%), women with placental parasitemia at lower densities (<10,000 parasites/μL) had a lower rate of perinatal MTCT (11.5%); this was not true for women with higher-density placental parasitemia (>10,000 parasites/μL, perinatal MTCT rate = 25.0%), who represented approximately one fifth of women with malaria. Among women with viral load below the detection level, perinatal MTCT occurred in none of 35 mothers with malaria, compared to 10 (8.7%) of 115 mothers without malaria (p = 0.12).

Among all study women, high maternal viral load and episiotomy or perineal tear were independent risk factors, and low-density (but not high-density) malaria was an independent protective factor for perinatal MTCT. Among women without microscopically detectable malaria, high maternal viral load, gravidity <2, and low birth weight in the newborn were significant risk factors or markers for perinatal MTCT. Among women with malaria (where perinatal MTCT was overall lower than in the malaria-negative women), high maternal viral load, episiotomy or perineal tear, and high-density placental infection were significant risk factors for perinatal MTCT (Table 3). In both the multivariate models for the entire population and women with malaria, significant interaction was found between viral load and placental malaria density.

In further examination of the interaction between maternal viral load and placental malaria, we found no significant differences in the frequency of episiotomy or perineal tear, mean CD4+ count, mean maternal hemoglobin level, mean birth weight, and frequency of small-for-gestational-age neonates among women without malaria, those with low-density placental malaria, and those with high-density placental malaria. Women with placental malaria were of lower mean gravidity than women without malaria (2.1 vs. 2.4, p = 0.03). Geometric mean maternal viral load was slightly higher in women with low-density placental malaria (2,226 copies/μL) and was nearly twofold higher in women with high-density placental malaria (3,390 copies/μL) than the viral load in women without placental malaria (1,774 copies/μL); however, these differences were not significant (p = 0.14). Viral load was significantly higher among women with peripheral malaria parasitemia at delivery (2,979 copies/μL) than among women without (1,725 copies/μL, p = 0.03). As shown in Table 4, the geometric mean viral load was fivefold higher in transmitters than in nontransmitters. While in women without malaria the geometric mean viral load in transmitters was fourfold higher than in nontransmitters, among women with malaria, the geometric mean viral load was 25-fold higher in transmitters than in nontransmitters. Among transmitting women, those with malaria had an eightfold higher geometric mean viral load than those women without malaria.

These associations were further evaluated in models comparing the relative risk for perinatal transmission between the three groups with placental malaria at various levels of maternal viral load. Low-density placental parasitemia was associated with significant protection for perinatal MTCT at the lower viral load levels, but not at higher viral load levels (Figure 1). When women with low- and high-density parasitemia were compared (Figure 2), high-density placental parasitemia was associated with increased risk for perinatal MTCT but only at the higher viral load levels.