The impact of this strong selection has been revealed at many different levels. Most important, as the use of chloroquine increased, drug resistance evolved in parasite populations and childhood mortality from malaria increased, even as all-cause mortality in children decreased [3-5]. The sequence of the P. falciparum genome has recently been published [6] and this has made it possible to trace the ancestry of highly drug-resistant parasites. These studies show that parasites resistant to chloroquine and sulfadoxine-pyrimethamine have arisen relatively rarely, but they have spread widely from a few initial foci in "selective sweeps" of the parasite population [7-11]. This new view affects many of the assumptions that underlie models of the speed at which resistance evolves [12] and inform practical decisions about changes in drug policy. Parasites without borders make it absolutely essential that the emergence of drug resistant populations be "tracked" worldwide; a resistant parasite that arises in Southeast Asia may travel rapidly to East Africa.
This improved understanding of the evolution of drug resistance has come from a relatively simple situation. Until recently, the number of antimalaria drugs in common use was small: chloroquine and sulfadoxine-pyrimethamine in Africa and the Americas, with mefloquine and more recently, mefloquine-artesunate in Southeast Asia[13]. As chloroquine and sulfadoxine-pyrimethamine have lost their efficacy, combination drugs have been strongly endorsed as the most effective next step [14]. In response to this emphasis, many different combination drugs, most containing an artemisinin derivative are being used in various countries, especially in East Africa (Figure 1. Many of these combinations have shown excellent initial efficacy in drug trials [13], but only mefloquine/artesunate has a long enough history to allow a strong prediction of the useful therapeutic life of these combinations [15]. It is particularly important to establish a baseline for effectiveness of new drugs and combinations so that any subsequent changes can be seen. This complex situation underlines the importance of regional surveillance of drug use, efficacy and effectiveness as these new combinations are tried in a variety of demographic and ecological settings. What has worked well for a long time in Thailand may not be so long lived in Tanzania [16]!
Appropriately, the gold standard for drug efficacy has been the outcome of clinical treatment. When patients are treated with the drug, do they recover? The substantial expense and logistical difficulty to change the recommended drug treatment have led most countries to rely on a large increase in clinical treatment failure before initiating a change [17]. Systematic studies have shown repeatedly that assessment in vitro of drug efficacy in local parasite isolates can give an early warning of rising drug resistance in vivo [18-20]. In addition, when molecular correlates of drug resistance are known, the prevalence of resistant alleles can also give early warning of evolving resistance in the parasite population [21-23]. In all three approaches, the temporal and geographic patterns of resistance are most informative. When the in vitro tolerance of parasites to a drug is rising or when the prevalence or the geographic range of resistant alleles is increasing, clinical drug failure is likely to increase as well. Figure 2 shows a small example of the linkage among the three parameters. In this data set, the increase in the in vitro IC50 values and the increased prevalence of the triple mutant allele of P. falciparum dhfr preceded by several years the increase in sulfadoxine-pyrimethamine treatment failure among young children in Coastal Kenya. Similar studies will be needed to determine whether the lags between these parameters observed in Kilifi will be similar in other sites or for other drugs, but it is clear that the in vitro increase in IC50 values and the increase in the molecular marker can provide an early warning of the onset of clinical treatment failure. The community will need similar data sets in many different settings for all of the drugs in use to manage effectively the current drugs and any novel drugs that are introduced in the future.
