|Home | About | Journals | Submit | Contact Us | Français|
Bacon et al. describe polymorphisms in a series of drug resistance genes or candidates before and after the introduction of artemisinin combination therapy (ACT) in Peru (1). Longitudinal studies of this type offer a dynamic view of molecular resistance determinants, but confounding factors that change over the course of the study must be removed during analysis. As noted in their paper, isolates collected in 1999 were obtained from two clinics. Sixty-four came from Padre Cocha (5 km from Iquitos) and 40 from Caballococha (311 km from Iquitos); data were pooled for statistical calculations. The 2006-2007 isolates were collected in Iquitos only. This change in composition of the study population between the two time points introduces a significant confounder that appears not to have been recognized in the analyses undertaken, undermining the value of these interesting data.
The issue appears critical in the case of the Plasmodium falciparum ATP6 (PfATP6) sequence analysis. In Table 6 of the article, the 1999 data were expressed as a percentage with a total denominator of 104. Using the information provided in the tables in the supplemental material, we undertook a reanalysis of the PfATP6 sequence data, calculating percentages with the number of samples per region as the denominator, thereby removing this confounding factor. (Note that, given the data available, we expressed haplotypes 4 and 5 as a combined total; the single haplotype 5 sample in the study differs from haplotype 4 in the synonymous polymorphism at C1031.)
Our reanalysis reveals that, contrary to the reported findings, the frequencies of both the G844 deletion allele and haplotype 6 (G844 deletion in isolation) did not change during the study period in the Iquitos samples (Table (Table1).1). The reported increase in haplotype 6 stemmed from the failure to take into account the confounding factor noted above. In Caballococha, the G844 deletion was completely associated with a second mutation, S466N (rendering it haplotype 4/5), unlike the majority of Iquitos area G844 deletion mutants that had the deletion in isolation (haplotype 6). Hence, the reason haplotype 6 became relatively more common is simply because the composition of the study population changed between the two time points; no Caballococha samples entered the 2006-2007 analysis. The result represents an artifact introduced by the analytical approach chosen.
There is still no consensus as to how artemisinins exert their antimalarial action, a situation that has not been helped by the absence of a paper describing PfATP6 transfection data presented at a conference in early 2008 (2). The problems in data analysis in the paper by Bacon et al. are unfortunate since the authors' carefully collected longitudinal data are a precious resource that can inform this controversial area of research. Reanalysis of the data, taking into account the confounding factor of location, shows that allelic and haplotype frequencies were highly comparable before and after ACT introduction in this area and indicates that PfATP6 was not subjected to selection pressure over this time.
Woodrow and Bustamante raised an important issue, and we therefore reanalyzed and modified Table 6 (see the erratum on page 2282 of this issue) so that the genotype allele frequencies at the 1999 time point are listed separately for the two geographical sites. We agree that Padre Cocha and Iquitos are more likely to have similar patterns of parasite population than Padre Cocha and Caballococha. As pointed out by Woodrow and Bustamante, our reanalysis indicates that there are no differences in the frequencies of G844 deletion in PfATP6 (combining the frequencies from rows 4 to 7 in the revised Table 6 gives 73% for Padre Cocha and 77% for Iquitos) or in the frequencies of genotype 623 when Padre Cocha (1999) and Iquitos (2006-2007) isolates are compared. We regret this oversight and appreciate Woodrow and Bustamante's contribution. We would also like to point out that, due to this reanalysis, the frequencies of genotypes 431 and 769 at the two study sites are different from those in our original report but with no obvious implications.