[摘要]：Background. Currently, the only tests capable of determining the serotype associated with dengue virus (DENV) infection require sampling during the period of acute viremia. No test can accurately detect the serotype associated with past DENV infections. The standard assay for determination of serotype-specific antibody against DENV is the plaque reduction neutralization test (PRNT), although performance of this test continues to be evaluated. Methods. From a cohort study among schoolchildren in Thailand, PRNT values were determined in serum samples collected before and after infection. A multinomial logistic regression model was used to infer the serotype associated with intercurrent DENV infections. Models were validated based on polymerase chain reaction identification of DENV serotypes. Results. The serotype associated with DENV infection inferred by the model corresponded with polymerase chain reaction in 67.6% of cases, and the kappa statistic was 0.479. A model for 35 cases with primary seroconversion correctly identified the DENV serotypes causing infection in 77.1% of cases, compared with 66.9%, using a model for 169 cases with secondary seroconversion. The best model using only postinfection PRNT values correctly inferred the DENV serotype causing infection in 60.3% of cases. Conclusions. A statistical model based on both pre- and postinfection PRNT values can be used to infer the serotype associated with DENV infections in prospective studies and vaccine trials.
