Background: Clinical trials of interventions designed to prevent severe falciparum malaria in children require a clear endpoint. The internationally accepted definition of severe malaria is sensitive, and appropriate for clinical purposes. However, this definition includes individuals with severe nonmalarial disease and coincident parasitaemia, so may lack specificity in vaccine trials. Although there is no "gold standard" individual test for severe malaria, malaria-attributable fractions (MAFs) can be estimated among groups of children using a logistic model, which we use to test the suitability of various case definitions as trial endpoints.
Methods and findings: A total of 4,583 blood samples were taken from well children in cross-sectional surveys and from 1,361 children admitted to a Kenyan District hospital with severe disease. Among children under 2 y old with severe disease and over 2,500 parasites per microliter of blood, the MAFs were above 85% in moderate- and low-transmission areas, but only 61% in a high-transmission area. HIV and malnutrition were not associated with reduced MAFs, but gastroenteritis with severe dehydration (defined by reduced skin turgor), lower respiratory tract infection (clinician's final diagnosis), meningitis (on cerebrospinal fluid [CSF] examination), and bacteraemia were associated with reduced MAFs. The overall MAF was 85% (95% confidence interval [CI] 83.8%-86.1%) without excluding these conditions, 89% (95% CI 88.4%-90.2%) after exclusions, and 95% (95% CI 94.0%-95.5%) when a threshold of 2,500 parasites/mul was also applied. Applying a threshold and exclusion criteria reduced sensitivity to 80% (95% CI 77%-83%).
Conclusions: The specificity of a case definition for severe malaria is improved by applying a parasite density threshold and by excluding children with meningitis, lower respiratory tract infection (clinician's diagnosis), bacteraemia, and gastroenteritis with severe dehydration, but not by excluding children with HIV or malnutrition.