Robust Algorithm for Systematic Classification of Malaria Late Treatment Failures as Recrudescence or Reinfection Using Microsatellite Genotyping

Abstract

Routine therapeutic efficacy monitoring to measure the response to antimalarial treatment is a cornerstone of malaria control. To correctly measure drug efficacy, therapeutic efficacy studies require genotyping parasites from late treatment failures to differentiate between recrudescent infections and reinfections. However, there is a lack of statistical methods to systematically classify late treatment failures from genotyping data. A Bayesian algorithm was developed to estimate the posterior probability of late treatment failure being the result of a recrudescent infection from microsatellite genotyping data. The algorithm was implemented using a Monte Carlo Markov chain approach and was used to classify late treatment failures using published microsatellite data from therapeutic efficacy studies in Ethiopia and Angola. The algorithm classified 85% of the Ethiopian and 95% of the Angolan late treatment failures as either likely reinfection or likely recrudescence, defined as a posterior probability of recrudescence of <0.1 or >0.9, respectively. The adjusted efficacies calculated using the new algorithm differed from efficacies estimated using commonly used methods for differentiating recrudescence from reinfection. In a high-transmission setting such as Angola, as few as 15 samples needed to be genotyped in order to have enough power to correctly classify treatment failures. Analysis of microsatellite genotyping data for differentiating between recrudescence and reinfection benefits from an approach that both systematically classifies late treatment failures and estimates the uncertainty of these classifications. Researchers analyzing genotyping data from antimalarial therapeutic efficacy monitoring are urged to publish their raw genetic data and to estimate the uncertainty around their classification.