Study Design

We chose a case–control design for this study because severe malaria is relatively rare in areas of stable transmission (13). Each adult with severe malaria was matched with 2 controls: 1 adult who had uncomplicated malaria and 1 asymptomatic adult in the community. We matched for major confounding variables: age group (15–19, 20–29, 30–39, and 40–49 years), sex, area of residence, and seasonal variation. Controls were recruited within 4 weeks after identification of case-patients.

Study Site and Patients

Figure

Figure. Malaria endemicity in Zambia. This map is a product of the Mapping Malaria Risk in Africa/Atlas du Risque de la Malaria en Afrique (MARA/ARMA) collaboration (www.mara.org.za), July 2005. Malaria distribution...

The study was conducted at Thompson Hospital, a government district hospital serving Luanshya district in the Copperbelt province of Zambia. In this district, 99% of the population is at risk for malaria (Figure), and 30% of women attending voluntary counseling and testing at the antenatal care department are infected with HIV-1 (15). Luanshya is not a site of sentinel surveillance for HIV, but sentinel surveillance conducted during late 2004 and early 2005 in the adjacent district of Ndola showed an HIV prevalence rate of 30% (16). From October 2005 through April 2007, all patients 15–49 years of age who sought treatment for symptoms and signs of febrile illness were screened for falciparum malaria by thick and thin blood smear. Patients in whom severe malaria was diagnosed were enrolled after informed consent was given by the patient or by his or her legally authorized representative. We defined severe malaria (16) as fever (body temperature >37.5°C ) or history of fever in the previous 48 hours, Plasmodium falciparum monoinfection with a density of at least 100 parasites per 200 leukocytes (assumed to be >4,000 parasites/µL) in the absence of any other evident causes of illness, and at least 1 of the following signs: impaired consciousness (Glasgow Coma Score <10), multiple grand mal convulsions, jaundice, hypoglycemia (glycemia <2.5 mmol/L), hyperparasitemia (parasite density >100,000/µL), renal impairment, and cardiorespiratory distress. We collected information about anemia but did not include anemia in the case definition because it can be HIV-1 related, resulting in possible bias, and because uncomplicated and severe malaria episodes can be misclassified in HIV-1–infected persons.

Patients with severe malaria were treated according to the national guidelines for severe malaria in Zambia (i.e., quinine 10 mg/kg/8 h for 7 d), with adequate supportive treatment (17). On the basis of information about areas of unstable malaria and of HIV surveillance data in Zambia, we predicted that HIV-1 increased the risk for severe malaria 5-fold and assumed a 30% HIV-1 prevalence (primary exposure) among controls. Using a 1:1 case:control ratio, we determined that we needed to recruit 30 case–control pairs to detect this risk with 80% power. Recruitment of cases continued until the required sample size was reached.

Controls with uncomplicated malaria were selected from the clinic closest to the homes of the case-patients. The first patient attending the clinic and exhibiting fever or having a history of fever in the previous 48 hours, without any other disease, with a positive rapid diagnostic test, residing in the catchment area of the health center, matching sex and age group of the case-patient, and willing to participate was recruited as a control. Pregnant women were excluded.

Field trips were organized to case-patients’ residences, where, following the random sampling method, a community control was identified and independently assessed. The first healthy asymptomatic adult fulfilling the matching criteria and willing to participate was recruited as a community control. Pregnant women were excluded. All community controls were assessed for possible confounding factors and screened for falciparum malaria by thick and thin blood smear. We sought to identify controls within 4 weeks after we recruited a case-patient.

Laboratory Tests

Laboratory technicians were blinded to patient identity and to patient-related parameters. We used microscopy to screen case-patients and community controls for malaria using Giemsa staining 10% for 5 min. A thin blood film was examined to identify type of parasite; a thick film was taken for quantification. Computation of parasite density took into consideration the actual leukocyte count. In asymptomatic controls, we screened for P. falciparum infection by using a rapid diagnostic test, Malaria Pf immuno-chromatographic test (ICT) (ICT Diagnostics, Cape Town, South Africa). For each participant, a blood sample (0.5 mL) was collected for HIV-1 testing and CD4 cell count, and for hemoglobin (1 mL), by venipuncture. Neither the study staff nor the patient had access to the HIV-1 test results. HIV-1 testing followed an unlinked anonymous procedure: blood samples were sent to the hospital laboratory, where they were processed anonymously. Determine HIV1/2/O (Abbott Laboratories, Abbott Park, IL, USA) was the first test. If test results were negative, the patient was considered not infected. If results were positive or indeterminate, the blood sample was tested with Unigold Recombinant HIV-1/2 (Trinity Biotech PLC, Bray, Ireland); if results were positive, the patient was classified as HIV-1 infected. If results were negative, the outcome was considered indeterminate, and the sample was further tested with ELISA and Western blot. Patients, after recovering from malaria, and controls were counseled about HIV-1 and offered the opportunity for voluntary counseling and testing.

CD4 counts were determined by flow cytometry on a CyFlow (Cyflow Counter, Partec, Germany) within 5 hours after collection, and multiset software was used to obtain the absolute counts and the CD4+ lymphocyte ratio (18). A FACSCount machine (Becton Dickinson, Sparks, MD, USA) was used as a quality control to validate the accuracy of the Cyflow data over time and served as backup.

Ethics and Consent

The trial was approved by the ethical committee of the Institute of Tropical Medicine, Antwerp, Belgium, and by the Research Ethics Committee of the Tropical Disease Research Centre, Ndola, Zambia. The written informed consent signed by study participants or their legally authorized representatives described the purpose of the study, procedures followed, and risks and benefits of participation. The consent form was in both English and the local language. Patients were counseled on HIV and offered the opportunity to undergo voluntary counseling and testing in conjunction with the study; this procedure was followed to ensure adherence to the standard of counseling and testing stipulated in the guidelines for HIV/AIDS counseling in Zambia (17).

Statistical Analysis

Crude associations between the primary outcome measures and potential risk factors, including HIV-1 status and HIV-1–related immunosuppression (HIV-1+ with CD4 >350/µL and HIV-1+ with CD4 <350/µL), were described by using summary statistics, means, and count and were tested by using Wilcoxon signed rank test for matched variables.

Risk Analysis

Case-patients differed from controls in use of other drugs during the previous week, HIV-1 infection, and CD4 count. No parasitemia was detected in any asymptomatic controls. Because of the matching, controls with uncomplicated malaria and asymptomatic controls were similar to case-patients for all other assessed risk factors (Table 1).

HIV-1 infection was detected in 45% of asymptomatic controls, 52% of controls with uncomplicated malaria, and 93% of case-patients. HIV-1 infection was not a risk factor for uncomplicated malaria (odds ratio [OR] 1.3, 95% confidence interval [CI] 0.5–3.7, p = 0.59) (Table 3). Case-patients were more likely to be infected with HIV-1 than were controls with uncomplicated malaria and asymptomatic controls (OR 12.6, 95% CI 2.0–78.8, p = 0.0005, and OR 16.6, 95% CI 2.5–111.8, p = 0.0005, respectively).

Eighty-three percent of case-patients had a CD4 count <350/µL, compared with 79% of controls with uncomplicated malaria (p = 0.76) and 8% of asymptomatic controls (p<0.0001). Controls with uncomplicated malaria were more likely than asymptomatic controls to have a CD4 count <350/µL (OR 7.67, 95% CI 1.78–33.01, p = 0.001). Case-patients were more likely to have a CD4 count <350/µL than were asymptomatic controls (OR 23.00, 95% CI 3.35–158.00, p<0.0001) but not controls with uncomplicated malaria (OR 3.00, 95% CI 0.83–10.83, p = 0.32).

The extremely high proportion of low CD4 count in both case-patients and controls with uncomplicated malaria might be confounded by a temporary malaria-induced reallocation of specific T-cells (19). Therefore, 28 days after successful treatment, the absolute CD4 count was measured in 10 HIV-infected case-patients. During this period, the mean CD4 count increased >2-fold, from 142 (95% CI 76–269) to 320 CD4/µL (95% CI 169–607) (data not shown). However, the proportion of case-patients with CD4 count <350/µL remained substantial (70%), although slightly less than at admission (90%).