TABLE 1.

Molecular correction algorithms proposed to decide whether a patient presenting again with a recurrent malaria infection during follow-up is a recrudescence or a reinfection based on the WHO-recommended genetic markers of msp-1, msp-2, and glurp^a

Algorithm	Reference	Definition	Consequences (identified in the model)
No correction		All recurrent infections are classified as recrudescence.	Algorithm grossly overestimates failure rate at higher FOI.b
WHO/MMV	2	Initial and recurrent samples must have shared alleles at all 3 markers to be classified as recrudescence.	Stringent conditions for recurrences to be classified as recrudescence mean that around 50% of true recrudescences are misclassified as reinfections, resulting in greatly underestimated failure rates. Most reinfections are correctly classified, so FOI has little impact on estimated failure rate.
No glurp	11	Results are as for the WHO/MMV algorithm but based on 2 loci (i.e., msp-1 and msp-2). (glurp is omitted as it is prone to genotyping errors.)	Results are largely identical to the WHO/MMV method.
≥2/3 markers	11	Results are as for the WHO/MMV algorithm, but initial and recurrent samples must share alleles at least at 2 out of 3 markers to be classified as recrudescence.	Results are generally intermediate between no-glurp and allelic family switch algorithms.
Allelic family switch	11	Comparison was initially based on msp-1 and msp-2. Identical alleles observed at both markers indicate recrudescence. Absence of shared alleles at both markers indicates reinfection. If 1 marker shares alleles and 1 does not (i.e., the sample is “discordant”), a complete allelic family shift in the nonsharing marker is required to classify a recurrence as a reinfection.	A tendency to misclassify reinfections as recrudescences leads to a dependency on FOI and results in large overestimates of failure rates at higher FOI, although the algorithm produces accurate failure rate estimates at low FOI.

^aWe also summarize the consequences of applying these algorithms for the analysis of clinical trials as quantified by our methodology: the failure rate estimates obtained from each algorithm are shown in Fig. 2 and Fig. 4.

^bFOI, force of infection, our measure of transmission intensity. FOI is the mean number of malaria infections that emerge in an individual and would become patent in the absence of drug killing over the course of a year.