ABSTRACT
The distribution and range of 50% inhibitory concentrations (IC50s) of doxycycline were determined for 747 isolates obtained between 1997 and 2006 from patients living in Senegal, Republic of the Congo, and Gabon and patients hospitalized in France for imported malaria. The statistical analysis was designed to answer the specific question of whether Plasmodium falciparum has different phenotypes of susceptibility to doxycycline. A triple normal distribution was fitted to the data using a Bayesian mixture modeling approach. The IC50 geometric mean ranged from 6.2 μM to 11.1 μM according to the geographical origin, with a mean of 9.3 μM for all 747 parasites. The values for all 747 isolates were classified into three components: component A, with an IC50 mean of 4.9 μM (±2.1 μM [standard deviation]); component B, with an IC50 mean of 7.7 μM (±1.2 μM); and component C, with an IC50 mean of 17.9 μM (±1.4 μM). According to the origin of the P. falciparum isolates, the triple normal distribution was found in each subgroup. However, the proportion of isolates predicted to belong to component B was most important in isolates from Gabon and Congo and in isolates imported from Africa (from 46 to 56%). In Senegal, 55% of the P. falciparum isolates were predicted to be classified as component C. The cutoff of reduced susceptibility to doxycycline in vitro was estimated to be 35 μM.
A Bayesian mixture modeling approach for assessing the effects of correlated exposures in case-control studies
