Malaria antibody responses augment surveillance in low-transmission settings in the Upper River Region, the Gambia

In low-transmission malaria settings, routine surveillance often fails to detect asymptomatic infections in partially immune populations. P(Detect), defined as the proportion of infections captured by surveillance systems, serves as a proxy for both population immunity and surveillance completeness. We evaluated whether antibody responses to Plasmodium falciparum antigens could classify low vs. high P(Detect) populations and identify surveillance blind spots.

We analyzed data from 5,300 seropositive individuals across 32 villages in The Gambia following mass drug administration. Random forest classification models were used to predict low vs. high P(Detect) groups based on antibody responses to 19 P. falciparum antigens. Analyses were stratified by transmission intensity (<5% and <10% PCR prevalence). Model performance was assessed using precision-recall area under the curve (PR AUC) in held-out test data, and permutation-based variable importance identified key predictive antigens.

Models demonstrated strong discrimination between detectability groups, with PR AUCs of 0.92-0.93 in the <5% stratum and 0.89-0.90 in the <10% stratum using a reduced 8-antigen panel. PfSEA, PfMSP1-19, gSG6, and Etramp4.Ag2 were consistently important predictors, though rankings varied by transmission context. Individual antigen models outperformed combined immune score approaches.

Antibody profiles can identify populations where surveillance underestimates transmission due to immunity. Integrating serological data into surveillance systems may improve detection of hidden reservoirs and support more targeted malaria elimination strategies.