Genetic surveillance of Plasmodium falciparum populations following treatment policy revisions in the Greater Mekong Subregion

Categories: Disease & Virus

May 21, 2025

Between January 2017 and December 2022, GenRe-Mekong collected and processed a total of 5982 Pf samples from 21 provinces in three countries (Fig. 1b and Supplementary Fig. 1). Sequencing coverage was high across the genotyped samples, with a mean of 747 reads across barcode loci and 565 reads across drug resistant loci (Supplementary Fig. 2). In Vietnam, the project processed 3340 samples in 2017–2022 (32.7% of WHO-reported Pf infections); in Laos, it processed 2400 samples over the same period (16.2%); while in Cambodia, collections took place between July 2020 and December 2022, producing 242 samples (21.0% of the WHO-reported infections in the country) (Supplementary Fig. 3)^28,29,30. A rapid decline in national Pf prevalence occurred sequentially across the three neighboring countries: first in Cambodia starting in 2018, then in Laos from 2019, and finally in Vietnam from 2020 (Fig. 1a). This decline in prevalence reflected a decrease in the number of confirmed cases rather than a decline in the number of tests (Supplementary Fig. 4). The longitudinal prevalence trend, as reported by WHO^28,29,30, has a strong correspondence with the number of samples collected by the project in Laos and Vietnam, indicating that routine surveillance data can serve as an indicator of Pf prevalence in these countries (Fig. 1a). As the number of cases declined over time, Pf populations became more spatially patchy. By 2022, the majority of samples were concentrated in one province in each country: Attapeu in Laos, Gia Lai in Vietnam, and Pursat in Cambodia (Supplementary Table 2); these provinces had consistently exhibited the highest incidence within their respective countries in previous years. Lack of coverage in certain provinces towards the end of our sampling period reflects underlying epidemiological trends rather than a lack of collection effort: although sampling sites across the whole region were engaged, the marked reduction in Pf cases led to some provinces not collecting any samples in the final period.

**Fig. 1: *P. falciparum* (Pf) samples genotyped by GenRe-Mekong during the study period.**

Given this evolution of the Pf epidemiological landscape, we opted to conduct longitudinal analyses by partitioning our sample set into three time periods (2017–2019, 2020–2021, and 2022), which reflect different epidemiological scenarios. Specifically, the period from 2020–2021 was a unique phase characterized by policy changes and major external events (e.g., pandemic-related movement restrictions), which differentiates it from both the earlier years and the year 2022. By separating these three periods, we sought to capture and highlight temporal trends more effectively, and assess how change factors have influenced Pf epidemiology.

DHA-PPQ resistance decline after changes in first-line treatments

We tracked the spread of DHA-PPQ resistant strains, KEL1/PLA1, by mapping the frequency of two key markers: pm23 amplifications conferring resistance to piperaquine (PPQ-R) and mutations in kelch13 gene associated with resistance to artemisinin (ART-R) as listed by the WHO (Supplementary Table 1)³¹. Although several crt mutations are also associated with PPQ-R, they rarely occur in field samples unless accompanied by the pm23 amplification; therefore pm23 amplification is deemed sufficient for defining resistance due to its strong association with treatment failure^4,12. In the 2017–2019 period, 62% (1957/3132) of all collected samples were predicted to be resistant to DHA-PPQ, as they carried both markers (Fig. 2a). Prior to that period, DHA-PPQ was used as frontline treatment for uncomplicated malaria in Cambodia, Thailand and Vietnam, which led to cross-border spread of the KEL1/PLA1 strain, resistant to both ACT components, across large areas of the eastern GMS¹³. This strain also spread into the southern Lao provinces of Attapeu and Champasak, where it introduced high levels of DHA-PPQ resistance, despite DHA-PPQ never being selected as a national treatment^9,32. Starting in 2017, Cambodia gradually switched its frontline treatment from DHA-PPQ to AS-MQ^17,18,19,22, while Thailand and Vietnam switched from DHA-PPQ to other ACTs by 2020^18,19,22,23. Following these changes, regional levels of DHA-PPQ predicted resistance fell steeply, from 62% (1957/3132) to 30% (204/690) in 2020–2021, and 1% (2/278) in 2022 (Z = −23.9, p < 0.001) (Fig. 2a). This was underpinned by a decline in predicted PPQ-R, dropping to 31% (228/737) in 2020–2021, and 2% (5/316) in 2022 (Z = −24.5, p < 0.001) (Fig. 3a and Supplementary Fig. 5).

**Fig. 2: Spatiotemporal patterns of predicted resistance to DHA-PPQ and artemisinin, from 2017 to 2022.**

**Fig. 3: Predicted regional antimalarial resistance from 2017 to 2022.**

Regional Pf prevalence shows a highly significant positive correlation with levels of predicted PPQ-R estimated from the proportions of pm23 amplifications (R = 0.64, p < 0.001), where periods of lower prevalence were generally characterized by lower PPQ-R proportions, suggesting that switching away from the use of DHA-PPQ as the frontline ACT may have had an impact on prevalence (Supplementary Fig. 6). However, there are other factors that may have contributed to the decline in the number of cases: in particular, the COVID-19 pandemic resulted in travel and trade restrictions that may have affected the transmission and spread of malaria parasites. To tease apart the impact of these two factors, we applied segmented regression to analyze shifts in observable trends at each intervention point; this analysis could only be conducted in Vietnam, where our data supported trend analyses before and after the two interventions. Following the change of frontline ACT, a significant reduction in prevalence was observed (β = −0.1069, p = 0.017), reversing the upward trend to a negative slope. Although the pandemic-related lockdown policy also shifted the trajectory from positive to negative, its immediate impact on prevalence was more modest and statistically non-significant (β = −0.0526, p = 0.111). These results suggest that, although both interventions impacted the decline in prevalence in Vietnam with robust explanatory power (R² ≈ 0.70), the switch of frontline ACT had a more pronounced and statistically significant impact on reduction (Supplementary Fig. 7c). Data from Vietnam also revealed a strong correlation between predicted PPQ-R prevalence–measured by pm23 amplification–and changes in first-line treatment policy. Following the replacement of DHA-PPQ with AS-PYR in five endemic Vietnamese provinces, we observed a reduction in Pf prevalence and parasites with pm23 amplifications within 3 months, ultimately leading to the complete disappearance of PPQ-R parasites within 18 months (Fig. 4 and Supplementary Fig. 7c). The effect of frontline ACT changes could not be assessed in Cambodia, since their policy change took place near the beginning of our study; Laos did not change its treatment policy during this period (Supplementary Fig. 7a, b). In both of these countries, lockdown restrictions during the pandemic showed no significant effect on prevalence, nor on PPQ-R decline (all p > 0.05).

**Fig. 4: Predicted levels of piperaquine resistance in five endemic provinces of Vietnam.**

In addition to pm23 amplifications, seven crt mutations have been associated with reduced susceptibility to piperaquine, either in vivo or in vitro³¹. To confirm the negative trend of PPQ-R frequency in the region, we analyzed the frequencies of these crt mutations, to investigate whether they are present in parasites lacking the pm23 amplification marker, and therefore predicted to be sensitive to piperaquine. One associated mutation (C350R) was completely absent from our sample set, while the remaining six mutations circulated at very low frequencies throughout the study period (Supplementary Figs. 8 and 9). No increase in the regional prevalence of these crt mutation was observed as pm23 amplifications decreased in frequency. Rather, the frequencies of the two most common mutations (I218F and T93S) declined in parallel with those of the pm23. Moreover, crt mutations were significantly more likely to be detected in the presence of pm23 amplifications (χ² = 245.2, p < 0.001), suggesting these resistance markers typically co-occur rather than circulate independently (Supplementary Fig. 10). The only exception was the H97Y mutation in Cambodia, which occurred more frequently in samples without pm23 amplification; however, the declining trend and low prevalence of H97Y (<10% frequency in 2022) are consistent with the waning of PPQ-R (Supplementary Fig. 11). The concordant evidence from multiple validated molecular markers- pm23 amplifications and crt mutations- provides support for a regional decline in PPQ-R prevalence during this study period.

Resistance to other antimalarials

ART-R levels remained high throughout the period analyzed (Fig. 3), except in some provinces at the periphery of the endemic region (Savannakhet in Laos, Quang Tri and Ninh Thuan in Vietnam), where most parasites remained susceptible to both artemisinin and piperaquine (Fig. 2b and Supplementary Fig. 5), suggesting that local parasite populations were isolated from the spread of DHA-PPQ resistant strains in the eastern GMS. Predicted mefloquine resistance (determined by detection of amplifications of the mdr1 gene³³) remained low in the region throughout the study periods (Fig. 3 and Supplementary Fig. 12), even after Cambodia adopted AS-MQ as frontline therapy in 2017¹⁷; in 2022, we detected no samples with an mdr1 amplification. Markers of resistance to the historical drugs chloroquine, sulfadoxine and pyrimethamine remained high across the region throughout the study period (Fig. 3 and Supplementary Fig. 13).

Distribution of kelch13 allele variants across fragmented populations

It has been shown that ART-R strains in the eastern GMS originated from multiple founder populations carrying different kelch13 alleles^34,35. Parasite strains carrying kelch13 haplotypes inherited from those early resistant populations have persisted over time³², even though their geographical distribution and prevalence often changed (Fig. 5). In 2017–2019, the majority of ART-R parasites in the eastern GMS carried the kelch13 C580Y mutation (98%, 2482/2539), and 79% of these mutants (1957/2482) were classified as KEL1/PLA1 (possessing both kelch13 C580Y and pm23 amplification) (Supplementary Fig. 14). However, since 2020 the dominance of KEL1/PLA1 waned, and other kelch13 variants, previously circulating at low frequency, expanded in the region. In the western provinces of Cambodia, kelch13 Y493H increased in frequency, and eventually dominated the population in Pursat province in 2022 (Fig. 5 and Supplementary Table 3). In Laos, kelch13 R539T mutants expanded in 2020–2021 in Attapeu and Champasak provinces, causing an outbreak³², and subsequently subsided in 2022 to be replaced by parasites possessing the kelch13 C580Y mutation without pm23 amplification in Attapeu. In Vietnam, C580Y remained the dominant kelch13 variant. We observed that the presence of these ART-R kelch13 alleles was strongly associated with the NFD mdr1 haplotype (characterized by N86, Y184F, and D1246 mutations; β = 3.44, p < 0.001). In contrast, wild-type kelch13 parasites predominantly carried either the NYD (N86, Y184, D1246; wild-type) or YYD (N86Y, Y184, D1246Y) mdr1 haplotypes (β = −5.14/−6.81, p < 0.001) (Supplementary Fig. 15).

**Fig. 5: Prevalence of *kelch13* alleles between 2017 and 2022.**

Clustering of parasite populations

Expecting the changes in kelch13 allele distributions to be a result of the change in prevalence of circulating populations, we sought to examine population structure by clustering Pf barcodes by similarity and examining the geographical distributions of the clusters. Clustering identical or near-identical parasites (sharing at least 95% barcode similarity) identified 27 clusters with ≥20 members (Table 1). In the period 2017–2019, four of the five largest clusters (KLV01, KLV02, KLV03 and KLV05), accounting for 30% of samples (1398/4632), were found to carry both the kelch13 C580Y mutation and a pm23 gene amplification, and thus likely to have emerged from the KEL1/PLA1 strain. Two of these large clusters (KLV01 and KLV05) had considerable geographic spread across the eastern GMS (Fig. 6). Following the implementation of new frontline treatment policies, however, notable changes occurred in the parasite populations. From 2020, clusters remained confined within single countries, often restricted to single provinces, and populations of non-KEL1/PLA1 clusters began to expand reaching high frequencies, as was the case for KLV04 and KLV07 in southern Laos and KLV12 in western Cambodia (Fig. 6 and Table 1). In Vietnam, the KEL1/PLA1-derived clusters KLV01 and KLV03 continued to circulate but lost their pm23 amplification (Supplementary Table 4). By the third quarter of 2021, none of the circulating clusters possessed the pm23 amplification, and the Central Highlands of Vietnam were dominated by the piperaquine-susceptible KLV03 population (Fig. 6). Taken together, these results suggest that the observed reduction in DHA-PPQ resistance in 2020–2021 can be explained by the disappearance of KEL1/PLA1 clusters, or their loss of pm23 amplification, occurring after changes in first-line treatment policies.