A clinical tool to predict Plasmodium vivax recurrence in Malaysia

BACKGROUND Recurrence rates of Plasmodium vivax infections differ across various geographic regions. Interestingly, South-East Asia and the Asia-Pacific region are documented to exhibit the most frequent recurrence incidences. Identifying patients at a higher risk for recurrences gives valuable information in strengthening the efforts to control P. vivax infections. The aim of the study was to develop a tool to identify P. vivax- infected patients that are at a higher risk of recurrence in Malaysia. METHODS Patient data was obtained retrospectively through the Ministry of Health, Malaysia, from 2011 to 2016. Patients with incomplete data were excluded. A total of 2044 clinical P. vivax malaria cases treated with primaquine were included. Data collected were patient, disease, and treatment characteristics. Two-thirds of the cases (n = 1362) were used to develop a clinical risk score, while the remaining third (n = 682) was used for validation. RESULTS Using multivariate analysis, age (p = 0.03), gametocyte sexual count (p = 0.04), indigenous transmission (p = 0.04), type of treatment (p = 0.12), and incomplete primaquine treatment (p = 0.14) were found to be predictors of recurrence after controlling for other confounding factors; these predictors were then used in developing the final model. The beta-coefficient values were used to develop a clinical scoring tool to predict possible recurrence. The total scores ranged between 0 and 8. A higher score indicated a higher risk for recurrence (odds ratio [OR]: 1.971; 95% confidence interval [CI]: 1.562-2.487; p ≤ 0.001). The area under the receiver operating characteristic (ROC) curve of the developed (n = 1362) and validated model (n = 682) was of good accuracy (ROC: 0.728, 95% CI: 0.670-0.785, p value < 0.001, and ROC: 0.766, 95% CI: 0.700-0.833, p-value < 0.001, respectively). In both the developed and validated models, area under the ROC curves showed no significant difference in predicting recurrence based on the constructed scoring mechanism (p = 0.399; Z-value: -0.8441; standard error: 0.045). CONCLUSIONS The developed model to predict recurrence was found to be of good accuracy and could be a useful tool in targeting patients at a higher risk for recurrence for closer monitoring during follow-up, after treatment with primaquine.