The association between malnutrition and the incidence of malaria among young HIV-infected and -uninfected Ugandan children: a prospective study

This study was conducted in the district of Tororo, a rural area in south-eastern Uganda. Malaria transmission in Tororo is holo-endemic, occurring perennially with an entomologic inoculation rate estimated to be 562 infective bites per person-year [23]. Data were collected from a cohort of HIV-unexposed, HIV-exposed, and HIV-infected children as has been described elsewhere [24]. Briefly, convenience sampling was used to enrol children referred to a dedicated study clinic from an adjacent post-natal clinic at Tororo District Hospital. Eligibility criteria included: 1) age six weeks to 12 months; 2) documented HIV status of mother and child by DNA polymerase chain reaction (PCR); 3) agreement to come to the study clinic for any febrile episode or other illness; 4) residence within a 30 km radius of the study clinic; 5) absence of active medical problem requiring in-patient evaluation at the time of screening; 6) provision of informed consent; and 7) currently breastfeeding if HIV-exposed. At enrolment, all study participants received long-lasting insecticide-treated bed nets (ITN) and all HIV-exposed and HIV-infected children were given daily TS prophylaxis as per Uganda's Ministry of Health (MOH) guidelines. HIV-unexposed children were not prescribed TS prophylaxis. The study protocol was approved by the Uganda National Council of Science and Technology and the institutional review boards of the University of California, San Francisco, Makerere University, the University of Washington, and the US Centers for Disease Control and Prevention.

Subjects were followed for all medical problems at a dedicated study clinic open seven days a week from 8 am to 5 pm. Parents (or guardians) were encouraged to bring their children to the study clinic whenever they were ill. After-hours care was available through the Tororo District Hospital. Children who presented with new medical problems underwent a standardized medical evaluation using algorithms to guide therapy for common illnesses that were developed locally by the study investigators. Medications with anti-malarial activity were avoided for the treatment of non-malarial illnesses. Monthly routine assessments were performed in the study clinic to ensure adherence with the study protocol. HIV-exposed children were re-tested for HIV by DNA PCR six to eight weeks after breastfeeding cessation. Those children who remained HIV-uninfected after breastfeeding cessation were then randomized to continue or discontinue TS prophylaxis as has been described elsewhere [25]. HIV-infected children, including those who seroconverted during breastfeeding, were continued on TS prophylaxis for the duration of the study. All HIV-infected children were anti-retroviral therapy (ART) naïve at enrolment and ART (nevirapine + lamivudine + zidovudine or stavudine) was initiated for those participants who met standardized WHO criteria. Study participants' compliance to TS, ART and ITN was assessed through questionnaires administered to their mothers at monthly visits. Study participants were not provided supplemental feeding. Study participants were withdrawn from the study for: 1) movement out of the study area; 2) inability to be located for > 60 consecutive days; 3) withdrawal of informed consent; 4) inability to adhere to the study schedule and procedures; or 5) inability to tolerate the drugs used for malaria treatment.

Weight and length/height measurements were taken at every visit to the study clinic at least once a month by trained study nurses in accordance with internationally recommended procedures [26]. Participants were weighed using a designated baby scale (Seca- 310, Hamburg, Germany) precise to the nearest 0.1 kg. Recumbent length measurements precise to the nearest centimetre were taken for children < 1 year of age using a stadiometer board and standing height was measured for older children precise to the nearest centimetre using TALC height measuring chart (Teaching Aids at Low Cost, St Albans, UK).

Subjects who presented to the study clinic with a documented fever (tympanic temperature ≥ 38.0°C) or history of fever in the previous 24 hours, had blood obtained by finger prick for a thick smear. If the thick smear was positive, the patient was diagnosed with malaria regardless of parasite density. Study participants ≥ 4 months and ≥ 5 kg were randomly assigned to receive either open label artemether-lumefantrine (AL) or dihydroartemisinin-piperaquine (DP) at the time their first episode of uncomplicated malaria was diagnosed. Study participants received the same treatment regimen for all subsequent episodes of uncomplicated malaria. Episodes of uncomplicated malaria in children < 4 months of age or weighing < 5 kg, as well as episodes of complicated malaria and treatment failures occurring within 14 days of initiating treatment were treated with quinine.

Thick and thin blood smears were stained with 2% Giemsa for 30 minutes. A thick smear was considered negative if no parasites were seen after review of 100 high-powered fields. Speciation of parasites was based on readings of thin smears by a trained microscopist. A final diagnosis of malaria used in the analysis was based on a rigorous quality-control system that included re-reading of all blood smears by a second microscopist and resolution of any discrepancies between the first and second readings by a third microscopist within 24 hours of the initial smear reading.

Assessment of infants HIV status prior to enrolment and at six to eight weeks after cessation of breastfeeding involved collecting a heel, toe or finger-prick blood sample on filter paper which was sent to a reference laboratory for HIV DNA PCR using Roche Amplicor HIV-1 DNA PCR test v1.5 kits (Roche Diagnostic Systems, Inc., Branchburg, New Jersey, USA).

Data were double entered in Access (Microsoft Corporation, Redmond, Washington, USA), and statistical analysis was performed using STATA, version 10 (Stata Corporation, College Station, Texas, USA). The observation period for each study participant began the day after enrolment and ended when a study participant was either prematurely withdrawn from the study or reached 2.5 years of age.

Nutritional status was assessed using the following standardized anthropometric z-scores as per WHO guidelines [27, 28]: height-for-age z- score (HAZ) for stunting and weight-for-age z-score (WAZ) for underweight. Measures of malnutrition were categorized as mild (z-score between -1 and -2) and moderate-severe (z-score < -2). Wasting (based on weight-for-height z-scores) was not included as a measure of nutritional status because values < -1 were relatively uncommon and therefore lacked statistical power for analyses. Repeated cross-sectional associations between variables of interest (breastfeeding, HIV-status, use of TS chemoprophylaxis and location of residence) and measures of malnutrition (dichotomized using z-scores < -1 as a cut-off and expressed) were estimated as a relative risk using generalized estimating equations with exchangeable correlation and robust standard error, binomial distribution, and adjustment for repeated measures within individuals.

Malaria incidence was defined as the number of new episodes of malaria per time at risk. Episodes of malaria occurring within 14 days of a previous episode were not considered incident events. Longitudinal associations between variables of interest and malaria incidence were estimated as an incidence rate ratio using generalized estimating equations with exchangeable correlation, negative binomial distribution, and adjusted for repeated measures. For longitudinal associations, measures of malnutrition were categorized into none, mild, and moderate-severe based on the cut-offs above for mean z-scores during observation periods of interest. Interaction among potential confounders was also examined by including interaction terms into the multivariate models. A p-value < 0.05 was considered statistically significant for all analyses.

A total of 100 HIV-unexposed children, 203 HIV-exposed children, and 48 HIV-infected children were enrolled between August 2007 and April 2008 (Figure 1). Two participants with no follow-up information were not included in this analysis. During follow-up, nine HIV-exposed children seroconverted during breastfeeding resulting in a total of 57 HIV-infected children. Of the 349 study participants included in the study, a total of 307 (88%) reached 2.5 years of age with the remaining 42 prematurely withdrawn.

Study participants were followed for mean of 16 months, resulting in a total observation period of 662 person-years. HIV-exposed children were prescribed TS for a total of 73% of the observation period (Table 1). Mothers of the HIV-infected and HIV-exposed children prescribed TS reported that their children took > 99% of their doses. HIV-unexposed children breastfed until a median age of 20 months compared to 17 months among the HIV-infected children and seven months among the HIV-exposed children, whose mothers were counseled to rapidly wean after six months of age (per national recommendations at the time). During routine follow-up visits, mothers reported that over 98% of children slept under an ITN the previous night. All 57 HIV-infected children were ART naïve at enrolment. Fifty-two (91%) were initiated on ART, two (3.5%) were eligible but withdrew from the study before ART could be initiated, and three (5.2%) never met eligibility criteria for ART initiation.

Overall, 36% of study participants were mildly stunted and 34% were moderate-severely stunted. In contrast, 25% of study participants were mildly underweight and 13% were moderate-severely underweight. Mean HAZ scores decreased from six to 18 months of age (the age at which most children were stopping breastfeeding) and then rose slightly to 30 months of age (Figure 2). HAZ scores were consistently lower for HIV-infected children compared to HIV-uninfected children at all ages (Figure 2). Mean WAZ scores were fairly constant from six to 30 months of age and lower for HIV-infected children, particularly in younger children.

In multi-variate analysis, both breastfeeding (RR = 0.75, 95% CI 0.69-0.82, p < 0.001) and being prescribed TS chemoprophylaxis (RR = 0.82, 95% CI 0.74-0.90, p < 0.001) were associated with a lower risk of stunting of any severity (Table 2). HIV-infection was associated with a higher risk of stunting (RR = 1.38, 95% CI 1.18-1.62, p < 0.001). The only significant risk factor for being underweight of any severity was HIV-infection (RR = 1.41, 95% CI 1.02-1.96, p < 0.001).

By the end of the observation period, 288 (83%) of the study participants had experienced at least one episode of malaria. A total of 2,414 new episodes of malaria were diagnosed, resulting in an overall incidence of 3.64 episodes per person year. In multi-variate analysis, children with mild stunting (IRR = 1.24, 95% CI 1.06-1.46, p = 0.008) and moderate-severe stunting (IRR = 1.24, 95% CI 1.03-1.48, p = 0.02) had a similarly higher incidence of malaria compared to children without stunting (Table 3). The association between stunting and an increased incidence of malaria was similar for HIV-infected and uninfected children (p = 0.52 for interaction term in multi-variate model). As expected, being prescribed TS prophylaxis and living in an urban residence were associated with a significantly lower incidence of malaria. Notably, breastfeeding was associated with a lower incidence of malaria in the uni-variate analysis, but not in the multi-variate analysis. Increasing age was associated with a significantly higher incidence of malaria. On the other hand, HIV infection was not associated with the incidence of malaria in the multi-variate analysis (Table 3). In contrast, being mildly underweight (IRR = 1.09, 95% CI 0.95-1.25, p = 0.24) and moderate-severe underweight (IRR = 1.12, 95% CI 0.86-1.46, p = 0.39) were not associated with a significant difference in the incidence of malaria compared to children who were not underweight.