Mapping the prevalence of molecular markers of <i>Plasmodium falciparum</i> artemisinin partial resistance in Africa: a spatial-temporal modelling study

A bstract Background Plasmodium falciparum kelch13 ( k13 ) mutations in Africa signal emerging artemisinin partial resistance (ART-R), endangering malaria control by undermining artemisinin-based combination therapies (ACTs). Sparse surveillance obscures whether rising k13 ART-R prevalence reflects local emergence or geographic expansion. We aimed to model and infer high-resolution spatial-temporal prevalence of k13 ART-R and ACT partner-drug markers to inform public health policy. Methods We conducted a systematic literature review (PROSPERO-ID CRD42024593923) spanning the years 2014–2025, complementing existing data from WWARN, MalariaGEN Pf7, and the WHO Malaria Threats Map. This integrated dataset, comprising 3,806 distinct molecular epidemiology surveys and 182,071 genotyped samples, was harmonized using a standardized data schema. We applied a spatial-temporal Gaussian process model to estimate the continuous prevalence of WHO k13 ART-R mutations, mdr1 86Y, and crt 76T. Findings ART-R increases were driven by distinct emergences of k13 561H in Rwanda, k13 675V in Uganda, and k13 622I in Ethiopia and Eritrea. From 2012 to 2024, predicted k13 prevalence rose steadily in Northern Province, Uganda (1.81% per year) and Northern Province, Rwanda (3.49% per year), reaching 26.36% in Uganda and 39.44% in Rwanda by 2024. Modelling indicated a rapid transition from localized k13 ART-R mutation emergence to entrenched regional hotspots centred on Uganda–Rwanda and the Ethiopia–Eritrea border. Partner drug amodiaquine marker mdr1 86Y is fading, but crt 76T remains prevalent in the Horn of Africa. Interpretation The rapid and multicentric expansion of k13 ART-R mutations in East Africa threatens ACT efficacy, especially where ART-R k13 and partner drug markers co-occur, mirroring early patterns observed before ACT failure in Southeast Asia. This study provides an updated k13 ART-R mutation database and high-resolution resistance maps with uncertainty quantification, demonstrating a crucial need to prioritize targeted molecular surveillance across greater East Africa to safeguard ACT efficacy. Funding US National Institutes of Health. 1 Research in Context Evidence before this study We searched OVID MEDLINE and PubMed databases: Pubmed was searched from database inception to July 9th 2025, using the following search terms: (((((((plasmodium falciparum) OR (falciparum))) AND (Africa)) AND (((resistance) OR (kelch13) OR (pfkelch13) OR (k13) OR (crt) OR (pfcrt) OR (mdr1) OR (pfmdr1)))) AND (((Sequencing) OR (markers) OR (genotyping))). This search identified 922 publications, of which 116 meet our inclusion criteria. We extracted the prevalence of all WHO-validated and candidate k13 mutations (441L, 446I, 449A, 458Y, 469F, 469Y, 476I, 481V, 493H, 515K, 527H, 537I, 537D, 538V, 539T, 543T, 553L, 561H, 568G, 574L, 580Y, 622I, 675V), mdr1 86Y, and crt 76T for these studies. In Africa, ART-R resistance mutations have been increasingly detected beginning in 2013-2015, particularly in East Africa (Rwanda, Uganda) and the Horn of Africa (Ethiopia, Eritrea). Therapeutic efficacy studies (TES) generally show high ACT efficacy across the continent; nevertheless, the presence of k13 resistance mutations is viewed as a high-priority public health threat. Prior efforts to map these molecular markers have been hampered by sparse and heterogeneous sampling, resulting in prevalence maps that rely on simple aggregation with wide, uninformative uncertainty intervals, limiting utility for national programs seeking to identify true increases in resistance. Furthermore, sub-continental estimates have often been static or limited in their ability to model the spatial and temporal dynamics of distinct mutation lineages. Added value of this study This study provides the largest and most current harmonized dataset of k13 ART-R and partner drug mutations in Africa with 182,071 samples and 3,806 surveys, integrating systematic literature review findings with major public databases (WWARN, MalariaGEN Pf7, and the WHO Malaria Threats Map). We used a spatial-temporal Gaussian process model to overcome the limitations of sparse sampling. This modelling approach enables the generation of continuous prevalence maps across space and time, simultaneously predicting prevalence in unsampled areas while rigorously quantifying the statistical uncertainty of those predictions. We provide the first contemporary, high-resolution maps that illustrate the concurrent selection dynamics of key partner drug markers ( mdr1 86Y and crt 76T) alongside the expanding focus of ART-R. Implications of all the available evidence The dramatic rise in k13 ART-R mutations, coupled with the steady geographic expansion identified by our model, confirms that ART-R is now firmly established in East Africa. The colocalization of high prevalence k13 ART-R mutations in Uganda and Rwanda with near-fixation of markers related to partner drug tolerance creates a molecular resistance profile highly reminiscent of the conditions preceding widespread treatment failure in Southeast Asia. This suggests the need to more broadly shift away from reliance on artemether-lumefantrine and necessitates greater molecular surveillance in sub-Saharan Africa, prioritizing the identified high-prevalence regions (Uganda, Rwanda, Ethiopia, Eritrea) and neighboring countries. Our integrated approach provides the foundational, evidence-based tools needed by national malaria control programs (NMCPs) and the WHO to refine drug policy decisions before ART-R develops into clinical treatment failure.