Background: Multiple clinical trials evaluating therapies for cerebral malaria (CM) have failed to demonstrate improved outcomes. This may derive from inclusion of children at all risk levels, including those at low risk of mortality or neurologic morbidity, limiting power to detect significant differences between intervention arms. One solution is enrichment, enrolling clinical trial participants at higher risk of adverse outcomes. We assessed if demographic, physical examination and point-of-care laboratory testing results in combination could identify children with CM at higher risk of death or neurologic disability.
Methods: Retrospective case-control study of 1674 children hospitalized with CM in Blantyre, Malawi. We used univariate and multivariate analyses of admission factors to find the most parsimonious model associated with death or neurologic disability. To assess the clinical utility of the models, we evaluated derived probability density curve separation.
Results: Blantyre Coma Score (BCS), deep breathing and high blood lactate were independently associated with mortality. The derived receiver operating curve yielded an area under the curve of 0.7118. There was poor separation of derived probability density curves predicting death or survival, indicating limited clinical utility of this model. On multivariate modeling of neurologic sequelae in CM survivors, only BCS was associated with adverse outcomes (area-under-the-curve = 0.6151). Probability density curves again largely overlapped, demonstrating limited utility of BCS alone in outcome prediction.
Conclusions: Combinations of admission demographic, clinical and point-of-care laboratory factors are inadequate to predict prognosis in children with CM. Higher technology assessment methods are necessary for clinical trial enrichment.