Declining Efficacy of Artemisinin Combination Therapy Against P. Falciparum Malaria on the Thai–Myanmar Border (2003–2013): The Role of Parasite Genetic Factors

The pivotal factor leading to the declining efficacy of the artemisinin-based combination on the Thailand–Myanmar border (mefloquine–artesunate) to a clinically unacceptable level is the increasing local prevalence of K13 mutations superimposed onto a long-standing background of Pfmdr1 amplification.

The Thailand-Myanmar border is endemic for malaria with high-grade antimalarial drug resistance [1,2]. After the failures of chloroquine and sulfadoxine-pyrimethamine, mefloquine was introduced as the treatment of uncomplicated Plasmodium falciparum malaria in 1985. However, resistance developed rapidly [2,3], mediated by amplification of the multidrug-resistance gene 1 (Pfmdr1) [4].In 1992 the cure rate of high-dose mefloquine monotherapy had fallen to 50% [5]. The first artemisinin-based combination treatment (ACT), mefloquine plus artesunate (MAS3), was deployed in 1994 in camps for displaced people [2]. The new treatment was highly efficacious and accompanied by a large and sustained reduction in P. falciparum incidence [6,7] despite continued presence of effective malaria vectors. A similar impact was documented in the population of migrant workers living along the border [8]. The diminishing proportion of infections carrying multiple parasite clones also suggested reduced transmission [9].
Resistance to artemisinin derivatives, characterized by delayed parasite clearance in artesunate-treated patients, was first recognized in western Cambodia in 2007 [10,11] and later on the Thailand-Myanmar border where it was found to be heritable [12] and associated with a selective sweep on chromosome 13 of the P. falciparum genome [13] due to mutations in the propeller region of the Kelch gene (K13) [14,15]. However, the link between K13 polymorphism and treatment failure was not clearly established, and some authors have contested whether use of the term "artemisinin resistance" is justified [16][17][18]. This uncertainty may have contributed to the failure to contain artemisinin resistance in the greater Mekong area.
Two recent studies have provided some evidence that K13 mutations are involved in combination with other mutations in the declining efficacy of 3-day dihydroartemisinin (DHA)-piperaquine in Cambodia [19]. An observational cohort study reported that treatment failure rates were higher in patients infected with parasites carrying the K13 C580Y mutation compared with the R539T mutation, but there were only 4 wild-type infections in that study. Two additional mutations (MAL13:1718319 and MAL10:688956), associated with parasite clearance rate in a genome wide association (GWAS) study [20], were in strong linkage disequilibrium with C580Y. The authors concluded that this triple mutant genotype might have contributed to the failures observed. A second study of 241 patients showed falciparum malaria recrudescence was associated with increased piperaquine in vitro 50% inhibitory concentrations and the presence of K13 mutations at 3 Cambodian sites [21] where isolates with multiple copies of Pfmdr1 are rare [21].
In contrast to Cambodia, P. falciparum malaria parasites along the Thailand-Myanmar border frequently have multiple copies of Pfmdr1 [4,22]. The MAS3 combination has been deployed successfully as a first-line regimen for more than 15 years, but failure rates in recent years have risen. To determine the factors contributing to declining ACT efficacy, we studied 1005 falciparum malaria patients treated with MAS3 between 2003 and 2013.

Patient Recruitment
The study was designed for the longitudinal monitoring of MAS3 efficacy in prospectively enrolled patients presenting to the clinics of the Shoklo Malaria Research Unit [23] with uncomplicated P. falciparum malaria, excluding pregnant women, patients with severe malaria [24] or >4% infected red blood cells, and those who had been treated with mefloquine in the previous 60 days. All patient treatments were supervised. Oral artesunate (Guilin Pharmaceutical Co., People's Republic of China), was given at a dose of 4 mg/kg/day for 3 days. Mefloquine (Roche, Switzerland, or CIPLA, India) was given either as split doses of 15 mg/kg and 10 mg/kg or 8 mg/kg once a day for 3 days [25]. On the first treatment day (day 0) a full clinical examination was performed; parasitemia and hematocrit were measured from capillary blood. Blood smears were microscopically examined daily until negative. Patients were seen weekly for 6 weeks. At each visit symptoms were recorded and a capillary blood sample was obtained for malaria smear and hematocrit. Dried blood spots for parasite genotyping were collected on grade 3 MM Whatman paper (Whatman, United Kingdom) at the first visit and in case of recurrence. Recurrence was defined by the occurrence of parasitemia during follow-up due either to a recrudescence or a reinfection. In 2013 single-dose primaquine as a gametocytocide (0.25 mg/kg) was added routinely on the first day of treatment.

Parasite Genotyping
Parasite DNA was extracted and genotyped at 3 polymorphic loci (MSP1, MSP2, and GLURP) to distinguish recrudescence from reinfection [26]. We determined Pfmdr1 copy number using quantitative polymerase chain reaction (PCR) and K13 polymorphisms by direct sequencing of PCR products (see Supplementary Methods). Pfmdr1 and K13 sequencing were done retrospectively after patient recruitment was completed.

Statistical Analyses
Data were analyzed using Stata 14 (StataCorp, College Station, Texas). Normally distributed data were compared using Student t test and nonnormally distributed data were compared using the Mann-Whitney rank sum test. Categorical variables were assessed using χ 2 tests. Analysis of variance was used to compare 3 or more groups. We used logistic regression to examine the association between each potential risk factor and used outcome and multiple logistic regression to analyze resulting risk factors. Survival time data were assessed using Cox regression. The population attributable fraction (PAF) for falciparum malaria recrudescence was calculated for K13 mutations and Pfmdr1 amplification individually and then for the marker combination using the formula 1 -(1 -PAF K13 ) × (1 -PAF Pfmdr1 ). We censored patients with indeterminate genotypes, new infections with P. falciparum, or those lost to follow-up.

Ethical Approval
The Oxford Tropical Research Ethics Committee (OXTREC 562-15) and the faculty of Tropical Medicine, Mahidol University (MUTM 2015-019-01) gave ethical approval for the study.

Patients
Between January 2003 and December 2013, 1022 patients were recruited, and 1005 remained in the study ( Figure 1). Overall, 290 (28.8%) patients did not complete the full 42 days but were included in survival analyses. There were significant changes in gender ratio, age, duration of symptoms, and admission temperatures over the recruitment period, with an increasing proportion of older febrile male patients over time (Supplementary Table 1). The proportion of patients with fever at admission (tympanic temperature ≥37.5°C) rose significantly from 26.2% in 2003 to 65.4% in 2013 (P = .005). There was also an increase in the median duration of symptoms before presentation from 2 days in 2003 to 3 days in 2013 (P = .04).
Our study did not characterize the single nucleotide polymorphisms of the Pfmdr1 gene because these relevant singlenucleotide polymorphisms (SNPs) are rare in Thailand [27] and mefloquine resistance is driven by copy number changes on a Pfmdr1 wild-type background [4,[28][29][30]. Isolates carrying wild-type K13 were associated with single-copy Pfmdr1, while K13 mutations were associated with amplified Pfmdr1 (Fisher exact test, P < .001) (Supplementary Figure 1).
The success rate for genotyping of K13 and Pfmdr1 was lower in recurrent infections because of significantly lower parasitemia (P = .0004) and thus much lower parasite DNA concentrations.

Parasite Clearance
Clearance data were available for 957 patients (95.2%). There was a significant increase in the proportion of patients who were parasitemic at day 3 (P < .001, test for trend; Table 1).    Multivariate analysis showed that mutation in the K13 gene was the strongest risk factor for day 3 positivity, with later year of treatment, higher parasitemia, higher hematocrit, and fever on admission (but not Pfmdr1 amplification) as independent risk factors ( Table 2). K13 propeller mutations were stronger predictors of day 3 positivity, and the 3 most common propeller mutations (C580Y, N458Y, and R561H) were each significantly associated with day 3 positivity (Table 2).

Cure Rates
Of the 186 patients with recurrent P. falciparum infections, there were 117 (62.9%) recrudescences and 53 (28.5%) reinfections, with 2 indeterminate results, 1 amplification failure, and 13 missing samples. PCR-adjusted parasitological efficacy at day 42 remained above or close to 90% from 2003 to 2009 but declined sharply thereafter (test for trend, P < .001; Table 1, Figure 3). There was a similar trend for recurrence (PCR unadjusted) rate. The lowest cure rates were recorded in 2011 when PCR-adjusted and unadjusted cure rates fell to 43.6% and 50.9%, respectively.

Predictors for Recrudescence
In a Cox regression model, increased Pfmdr1 copy number and K13 mutation were significant independent predictors of falciparum malaria recrudescence (with a multiplicative effect when in combination) along with year of recruitment and age ( Table 3). The risk of recrudescence was even higher for K13 propeller mutations as a group and for the 3 most common individual propeller mutations (C580Y, N458Y, and R561H; Table 3). Cure rates were highest for infections with isolates with single-copy Pfmdr1 gene and wild-type K13 (97.8%) and lowest in patients with multiple copies of Pfmdr1 and any K13 propeller SNP (57.8%; Table 4, Figures 4 and Supplementary Figure 2). Cure rates declined markedly in recent years as the prevalence of K13 mutations rose (Supplementary Figure 2).
The PAFs (equivalent to the percentage reduction in recrudescent infections that would occur if these mutations were not present) for K13 and Pfmdr1 amplification were 69.0 and 41.9%, respectively. The PAF for the 2 factors in combination was 82%. Hence these 2 factors alone explain almost entirely the increased ACT treatment failure rate observed in this population.

Pre-and Post-treatment Gametocytemia
The proportions of patients presenting with pre-treatment gametocytes or with post-treatment gametocytemia and the associated risk factors are shown in the Supplementary Materials (Supplementary Tables 1, 1.1 and 1.2). There was no association

DISCUSSION
MAS3 for the treatment of uncomplicated P. falciparum malaria has had a remarkable therapeutic longevity and a dramatic impact on morbidity and mortality on the Thailand-Myanmar border. When artesunate was introduced in 1991, mefloquine was a failing drug, the incidence of falciparum malaria was rising, and most parasite isolates had multiple copies of Pfmdr1 associated with mefloquine resistance [4]. After deployment of this ACT in 1994, mefloquine recovered its efficacy, and less fit isolates with multiple Pfmdr1 copies were replaced by single copy-containing isolates [31]. High cure rates were then sustained for more than a decade [22,32]. In 2006, 53% of patients presenting with uncomplicated falciparum malaria had infections with multiple copies of Pfmdr1 [22]. The rise in prevalence of P. falciparum parasites with K13 propeller mutations was the definitive event that led to the demise of MAS3. The temporal sequence of K13 selection is informative. The E252Q mutation (which is located on the stem region of the K13 gene; Figure 2A), associated with moderate slowing of parasite clearance [15], predominated before 2010. Then parasites with the E252Q mutation were progressively overtaken by parasites with K13 propeller mutations, conferring greater reductions in parasite clearance rates. The main propeller mutation was C580Y, a polymorphism common in Cambodia [14,15]. The E252Q mutation was associated with a moderate increase in risk of treatment failure, suggesting that the selective advantage of the K13 mutation is proportional to the degree of parasite clearance prolongation. It is possible that genetic changes outside the K13 locus have contributed to the emergence of artemisinin resistance and the selection of K13 mutations [20]. It remains to be seen whether parasite genotypes with even slower clearance will become established while ACTs remain the mainstay of antimalarial treatment. The large body of data presented here strongly support the hypothesis that there is a substantial impact of K13 mutations on treatment failure. K13 mutations in admission samples are associated with a failure rate of 21.5%. In combination with multiple copies of Pfmdr1, this rises to 42.2%. The sharp decline in MAS3 cure rates temporally corresponds with the emergence of K13 propeller mutant isolates on a long-standing genetic background of Pfmdr1 amplification. The proportion of treatment failures that can be attributed to K13 mutations and multiple copies of Pfmdr1 (ie, PAF) is 82%. Hence, these 2 factors alone explain the majority of increase in treatment failure observed. Other parasite genetic factors that our study was not designed to detect or changes in the demographics of the patient population may have contributed to the 18% of unexplained variation in treatment failure.
We highlight 3 additional features of particular interest. First, K13 mutations and Pfmdr1 amplification have a multiplicative (rather than an additive) effect on risk of treatment failure. Synergy between these 2 resistance determinants may help to explain why failure rate declined precipitously in 2009parasites carrying both markers only became common as K13 mutations rose in frequency. Second, in contrast to the 2 Cambodian studies, we observed multiple K13 mutations in this study. These data demonstrate significant heterogeneity in the impact of different K13 mutations on treatment failure. A mutation outside the K13 propeller (E252Q) increased treatment failure relative to wild-type parasites, but the 3 common propeller mutations had a much greater effect (Table 3). These data suggest that surveillance for emergence of K13 mutations should be expanded to include K13 regions outside the propeller domain. Third, 2 studies [15,19] provide evidence that gametocyte carriage is elevated in parasites bearing K13 mutations, leading to the suggestion that such parasites may have a transmission advantage. In contrast, in this large study from a single location, we observed no association between gametocyte carriage and K13 mutations.
What are the implications of these data for the therapeutic life span of ACTs in Southeast Asia and beyond? The most widely deployed ACT is the coformulation of artemether and lumefantrine (AL), which has a resistance mechanism that is similar to that of mefloquine, involving Pfmdr1 amplification [29,33,34]. In most areas where AL is used, there is no evidence for resistance to either component. However, in Myanmar where AL has recently been introduced, K13 propeller mutants and Pfmdr1 amplification are widespread [28,[35][36][37][38]. The therapeutic lifetime of AL in Myanmar and other areas with a high prevalence of K13 mutations may be relatively short.
ACTs containing an alternative partner drug should provide relief in the short term. In 2012 the first-line treatment of uncomplicated P. falciparum malaria in our treatment centers on the Thailand-Myanmar border was changed from MAS3 to DHA-piperaquine, which is currently highly efficacious in this area but relies increasingly on the piperaquine component. The recent emergence of piperaquine resistance in Cambodia [39] and associated rising failure rates with DHA-piperaquine [19,21,40] also cast doubt over the long-term future of this combination. Alternatives are needed desperately. With new antimalarials still years from deployment, there is an urgent need to eliminate P. falciparum from the area before the recent and substantial gains in malaria control are reversed.

Supplementary Data
Supplementary materials are available at http://cid.oxfordjournals.org. Consisting of data provided by the author to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the author, so questions or comments should be addressed to the author.