BACKGROUND: The objective of this study was to evaluate the distribution of a series of independent doxycycline inhibitory concentration 50% (IC50) values to validate the trimodal distribution previously described and to validate the use of the pftetQ and pfmdt genes as molecular markers of decreased in vitro doxycycline susceptibility in Plasmodium falciparum malaria. METHODS: Doxycycline IC50 values, from 484 isolates obtained at the French National Reference Centre for Imported Malaria (Paris) between January 2006 and December 2010, were analysed for the first time by a Bayesian mixture modelling approach to distinguish the different in vitro phenotypic groups by their IC50 values. Quantitative real-time polymerase chain reaction was used to evaluate the pftetQ and pfmdt copy numbers of 89 African P. falciparum isolates that were randomly chosen from the phenotypic groups. RESULTS: The existence of at least three doxycycline phenotypes was demonstrated. The mean doxycycline IC50 was significantly higher in the group with a pftetQ copy number >1 compared to the group with a pftetQ copy number = 1 (33.17 muM versus 17.23 muM) and the group with a pfmdt copy number >1 (28.28 muM versus 16.11 muM). There was a significant difference between the combined low and medium doxycycline IC50 group and the high IC50 group in terms of the per cent of isolates with one or more copy numbers of the pftetQ gene (0% versus 20.69%) or pfmdt gene (8.33% versus 37.93%). In the logistic regression model, the pfmdt and pftetQ copy numbers >1 (odds ratio = 4.65 and 11.47) were independently associated with the high IC50 group. CONCLUSIONS: Copy numbers of pftetQ and pfmdt are potential predictive molecular markers of decreased susceptibility to doxycycline.