(P-248) Modeling Parasite Clearance Curves in a Cohort to Compare Quinine and Artesunate for the Treatment of Pediatric Cerebral Malaria

Background: In the treatment of malaria, clearance curves measure an antimalarial drug’s ability to clear parasites after treatment initiation. While a clearance curve for each subject can be partitioned into three periods: a time-lag, an exponential decay, and a plateau, the duration of each period or the rate of decay vary across subjects.

Objectives: Our objectives were to (1) Develop models to estimate a cohort central tendency and between subject variability in different summary features including clearance rate constant, half-life, and time to limit of microscopy detection (2) Compare these features between drugs and for different cohort characteristics.

Methods: We performed a retrospective cohort study of 465 children with cerebral malaria treated with either intravenous quinine (2011-2013) or artesunate (2014-2019). All children were enrolled in a longstanding study of cerebral malaria pathogenesis located in Blantyre, Malawi. Blood parasite count was measured on admission and every 6 hours thereafter until two consecutive measurements were negative. We used mixed effect models incorporating non-linearity and accounting for varying number of measurements on each subject to estimate the cohort clearance rate constant and half-life. We investigated change in the time to limit of microscopy detection using time to event analyses. We explored how these features varied with exposure, admission parasite count, and demographic and medical history characteristics using regression adjustments or propensity score weighting.

Results: Our methods for estimating features and evaluating change across characteristics showed that compared to quinine, artesunate had a shorter clearance lag time and faster rate of decay, and thus reached the limit of microscopy detection faster. The latter feature was easier to estimate and analyze using time-to-event analyses which showed a separation of clearance curves as early as 6 hours post-admission. All analyses adjusted for admission parasite count, an important predictor of most features.

Conclusions: We developed methods to estimate multiple cohort features using longitudinal data on parasite count during hospitalization, with unequal number of measurements on each subject. Analyses of time to reaching the limit of parasite detection were the most informative in showing superiority of artesunate over quinine in treating malaria.