Assessing the effect of parasite antigen genetic diversity on performance of Plasmodium vivax serological exposure markers for malaria: a multicentre observational diagnostic accuracy study

Plasmodium vivax poses a major obstacle to malaria elimination because this parasite can lie dormant in the liver for weeks to months before reactivating and causing a relapse of infection. These dormant forms (hypnozoites) cannot be detected using standard diagnostics, but P vivax exposure in the previous 9 months and, by proxy, hypnozoite carriage, can be inferred using serological markers. In this study, we aimed to examine how genetic variation in P vivax affects the utility of these markers and whether redesigned antigens could improve performance.

In this observational diagnostic accuracy study, we analysed global P vivax genetic data to assess variation in leading serological markers (n=14). Accordingly, we expressed new haplotypes that better reflect global sequence diversity for eight antigens, compared with the commonly used reference strain (Sal-1). Antibody responses against these were tested using samples from cohorts in Brazil and Thailand, with magnitude assessed in relation to how recently participants had a qPCR-detectable blood-stage P vivax infection. We compared the ability of the haplotypes versus the reference to correctly identify individuals infected within the previous 9 months.

Extensive global genetic diversity was identified in two P vivax antigens, MSP5 (π=14·8 × 10^-3) and DBPII (π=7·7 × 10^-3). Several antigens had large numbers of circulating haplotypes, with the percentage with similar sequence identity to the reference Sal-1 ranging from 0·4% (MSP5) to 99% (S16). Samples for immune analysis were previously collected between April 2013 and June 2014, with 774 and 923 participants included in the current analysis from Thailand and Brazil, respectively. Two antigens showed strong differences in immunogenicity by region and construct (RBP2a and DBPII). However, for most proteins (five of eight: MSP5, RiPR, PTEX150, Pv-fam-a, and RBP2b), these differences had no significant effect on the accuracy of identifying recent exposure. Affected performance (eg, RBP2a) was overcome by adding multiple antigens into the classification model.