Serological markers validated for detecting recent Plasmodium vivax exposure in Cambodia

Cambodia has driven its malaria burden down to a fraction of what it was less than a decade ago, but the parasite that now dominates the country’s remaining transmission is also the one its diagnostic toolkit is least equipped to catch. Plasmodium vivax, the species responsible for the vast majority of infections left in the Greater Mekong Subregion, hides in the liver in a dormant form that no blood smear, rapid diagnostic test, or molecular swab currently in routine use can detect. A new study published in The Lancet Regional Health – Western Pacific reports the first Cambodian validation of a blood-test panel designed to flag exactly those hidden carriers, using nothing more than the fingerprints of antibodies circulating in a finger-prick sample of blood — and the results suggest that serology could become a central weapon in the final push toward elimination.

The biological challenge is unlike anything posed by Plasmodium falciparum. P. vivax produces dormant liver stages known as hypnozoites, which can lie silent for weeks, months, or even years after the original mosquito-borne infection before reactivating to trigger a clinical relapse. These relapses are thought to account for more than 80 percent of all P. vivax blood-stage infections, meaning that most of what clinicians actually see is the echo of an older, invisible exposure. The only drugs capable of killing hypnozoites are the 8-aminoquinolines, primaquine and tafenoquine, but they carry a serious caveat: in people with glucose-6-phosphate dehydrogenase (G6PD) deficiency, they can provoke severe, potentially life-threatening haemolysis, and their efficacy may be blunted in individuals with low cytochrome P-450 2D6 activity. Radical cure therefore demands careful G6PD testing — a capability Cambodia’s National Centre for Parasitology, Entomology, and Malaria Control has only recently begun rolling out nationwide.

Compounding the problem, up to 90 percent of P. vivax infections are asymptomatic, submicroscopic, or both, slipping past routine case detection entirely. Individuals carrying only dormant liver-stage parasites — who may make up more than half of all infected people — cannot be identified by any current diagnostic. That blind spot is why mass screening and treatment has repeatedly failed to reduce transmission in low-endemic settings: blood-stage screening simply misses the reservoir that matters. Mass drug administration with 8-aminoquinolines could theoretically reach that reservoir, but without widespread G6PD testing it risks harm, and for most recipients it offers limited benefit. What elimination programmes need is a way to identify people with recent exposure — the group most likely to be harbouring hypnozoites — and treat them selectively.

The new research, led by an international team including first author Costanza Tacoli of the Institut Pasteur and collaborators from the Walter and Eliza Hall Institute and the Institut Pasteur du Cambodge, exploits a quirk of the immune response to P. vivax. Even low-density, asymptomatic blood-stage infections provoke strong and durable IgG antibody responses against a broad range of parasite proteins, and those responses persist for months after the parasite itself has been cleared. Because nearly all hypnozoite carriers experienced a primary blood-stage infection within the preceding six to nine months, the antibody landscape in a person’s plasma effectively encodes a readable record of recent exposure — a biomarker window that aligns almost perfectly with the period in which hypnozoite carriage is most likely.

Building on earlier work that developed an eight-protein panel tested in Thailand, Brazil, the Solomon Islands, and Peru, the team evaluated its performance against two cohorts drawn from Kaev Seima, a district in Cambodia’s eastern province of Mondulkiri, where asymptomatic PCR-detected P. vivax infection had previously reached 36 percent. The longitudinal arm enrolled 950 individuals aged five and above from 391 randomly selected households across ten villages, following them monthly from January 2019 to mid-February 2020 with blood sampling and clinical assessment regardless of symptoms, then adding three quarterly visits through December 2020. Of these, 471 individuals met the strict inclusion criteria for the accuracy analysis, having completed at least the first twelve months of follow-up. Notably, no mixed Plasmodium species infections were detected, so the results reflect pure P. vivax epidemiology.

The second arm was a dry-season cross-sectional survey conducted between mid-December 2017 and mid-April 2018, enrolling 4,200 people from 1,147 households across 17 villages, with census-based random sampling and statistical weights applied so that smaller villages were oversampled relative to their population size. Around 500 microlitres of capillary blood were collected by finger-prick from each participant, separated in the field into plasma and cell pellets, and shipped under cold chain to the central laboratory at the Institut Pasteur du Cambodge in Phnom Penh. Every sample — 9,928 longitudinal specimens across twelve time points and 4,200 survey samples — was screened by real-time polymerase chain reaction for parasite DNA, providing the molecular ground truth against which the antibody-based classifier would be judged.

The serological engine of the study was a multiplex Luminex immunoassay measuring IgG responses to fourteen recombinant P. vivax proteins, most produced in a wheat germ cell-free expression system and based on the Salvador-I reference strain. The panel included the eight top-performing markers from the original global algorithm — merozoite surface proteins MSP3a, MSP1-19, MSP5 and MSP8, the rhoptry-associated membrane antigen RAMA, two constructs of the Pv-fam-a protein, and the erythrocyte binding protein PvEBP — plus six additional antigens with strong individual performance. Protein-coupled magnetic microspheres were incubated with plasma diluted 1:100, tagged with PE-conjugated anti-human IgG, and read on a Bio-Plex 200 instrument. Median fluorescence intensities were converted to relative antibody units using a five-parameter logistic model anchored by a serially diluted hyperimmune plasma standard, correcting for plate-to-plate variation.

Those antibody units then fed into PvSeroTAT, an R-Shiny classification tool built on a Random Forest algorithm trained on serological data from Thailand, Brazil, the Solomon Islands, and negative controls. The model assigns each person a probability of recent infection — defined, in line with prior work, as at least one PCR-confirmed infection within the preceding nine months — and allows operators to adjust the vote threshold to trade sensitivity against specificity. In the longitudinal cohort, the original ‘global’ algorithm achieved a specificity of 83.3 percent, closely matching its designed 80 percent target: of 330 individuals who never tested PCR-positive over a full year, 275 were correctly called seronegative. Sensitivity, however, ran below the 80 percent benchmark at 69.5 percent, and the data revealed a clear decay curve: classification accuracy was highest, at 82 percent, for infections within the last month, and fell to 55 percent for infections six to nine months old and 41 percent for those nine to twelve months prior, consistent with the slow waning of antibody titres.

To squeeze more performance out of the Cambodian data, the team trained a country-specific Random Forest model, exhaustively testing combinations of eight antigens drawn from the fourteen measured candidates and validating each with ten-fold cross-validation repeated five times. Individually, the reticulocyte binding protein RBP2b was the strongest classifier with an area under the ROC curve of 0.82, while MSP8 and PvEBP lagged at 0.72. The best Cambodia-optimised combination — which swapped in an alternative C-terminal construct of RBP2b and the related RBP2a — nudged the AUC to 0.86, with a balanced sensitivity and specificity of 78 percent, only marginally better than the off-the-shelf global panel. The modest gain, the authors note, reflects substantial redundancy among correlated serological markers: adding more antigens increases model complexity and overfitting risk without buying much accuracy.

Perhaps the most striking demonstration came from the cross-sectional survey, where serology exposed a transmission landscape that PCR alone could barely see. Molecular testing found P. vivax in just 6.73 percent of the 3,538 participants with paired data, but the serological classifier labelled 21.6 percent as recently exposed — more than three times as many. Seropositivity was concentrated in forest villages such as Ohkra, where 68.3 percent of residents tested seropositive, Beng-Gaty at 63.3 percent, and Ohtrone at 50.5 percent. Because serological outcomes capture cumulative exposure over a roughly nine-month window rather than a single point-in-time blood-stage snapshot, they dramatically increased the statistical power to detect behavioural and environmental risk factors. Multivariable mixed-effects logistic regression, with random intercepts for village and household and model selection by Akaike information criterion, confirmed that men carried roughly three times the infection odds of women, that children aged five to ten were substantially less exposed, and — an insight only serology could resolve — that rubber plantation workers had significantly lower risk than cassava field workers, with an adjusted odds ratio of 0.15, a comparison statistically impossible against the sparse PCR data.

Taken together, the findings validate serological exposure markers in one of Asia’s last P. vivax strongholds and sketch out their dual role: as a trigger for targeted anti-hypnozoite treatment through the PvSeroTAT strategy, and as a high-resolution surveillance instrument for mapping residual transmission in places where parasite genomic data are scarce and cases are too few for conventional metrics to catch. With Cambodia targeting elimination of human malaria species by 2025 and reporting only 313 cases in the first nine months of 2024, the margin of error is vanishingly small, and every undetected hypnozoite carrier is a potential relapse waiting to reignite transmission. A test that reads the immune system’s memory of recent infection may prove to be exactly the tool that lets programmes see — and finally clear — the reservoir that has always been one step ahead of them.