Antibody dynamics after melioidosis infection in Thailand: a retrospective cohort study.
BACKGROUND Melioidosis is a life-threatening infectious disease caused by Burkholderia pseudomallei, endemic to southeast Asia and northern Australia. Serological tests targeting haemolysin co-regulated protein 1 (Hcp1) and O-polysaccharide (OPS) antigens have shown promise for diagnosis and surveillance, but their interpretation is limited by an incomplete understanding of antibody persistence following infection. To address this gap, we aimed to characterise post-infection decay dynamics of Hcp1 and OPS-specific IgG and IgA antibodies to improve diagnostic interpretation and population-level seroepidemiological inference. METHODS We conducted a retrospective cohort study using previously collected samples from two cohorts of patients with culture-confirmed B pseudomallei infection and healthy blood donors recruited from three tertiary hospitals in northeast Thailand (Mukdahan Hospital, Roi-Et Hospital, and Surin Hospital). Inclusion criteria were age 15 years or older for cohort 1 and 18 years or older for cohort 2, and hospitalisation history with microbiologically confirmed B pseudomallei infection from any clinical specimen. Exclusion criteria for both cohorts included pregnancy, receipt of palliative care, or incarceration. Whole blood samples were collected at enrolment and on days 5, 12, and 28 and at months 3 and 12. Clinical and demographic data were collected through medical record review and patient interviews. We measured IgA and IgG antibody responses to Hcp1 and OPS antigens using kinetic and endpoint ELISAs. Diagnostic accuracy was evaluated at enrolment and day 28 using receiver operating characteristic curves, with optimal cutoffs determined using Youden's J statistic. Joint cutoffs for antigen-isotype pairs were optimised by evaluating all pairwise threshold combinations and selecting those maximising balanced accuracy. Antibody dynamics were modelled using Bayesian hierarchical two-phase modified predator-prey models that captured the exponential rise during active infection and non-exponential decay following pathogen clearance, both in the overall cohort and in analyses stratified by diabetes status. FINDINGS We tested 656 samples from 142 patients with melioidosis and 76 healthy blood donors enrolled between Dec 30, 2015, and July 13, 2022. The median patient age was 54 years (IQR 46-61); 99 (70%) of 142 patients were male, and 98 (69%) of 142 had diabetes. Areas under the receiver operating characteristic curves (AUCs) for each antigen-isotype combination at enrolment ranged from 0·89 to 0·98 and were comparable between kinetic and endpoint assays (difference ≤0·1 AUC); the kinetic ELISA provided a broader dynamic range at high antibody concentrations, where endpoint assay values reached saturation. Modelled median antibody responses remained above the Youden diagnostic threshold for more than 1 year for all antigen-isotype combinations. OPS IgG had the highest diagnostic accuracy at enrolment (AUC=0·93, 95% CI 0·90-0·96); however, modelled median antibody responses remained elevated above the Youden threshold for 902 days (approximately 2·5 years). A higher joint cutoff for OPS IgA and IgG improved accuracy and reduced the modelled duration of positivity to less than 1 year. People with diabetes had higher peak responses and slower decay rates across most antigen-isotype combinations. INTERPRETATION Modelled OPS-specific and Hcp1-specific antibody responses show rapid increases following infection and gradual decay over the subsequent year; these dynamics can be used to estimate population-level incidence rates. Clinically, persistent antibody positivity may reduce diagnostic specificity in endemic settings by detecting past rather than recent infections. Joint antigen cutoffs might improve sensitivity and specificity at clinical presentation. FUNDING US National Institutes of Health, Defense Threat Reduction Agency, and Wellcome Trust UK.