Background: Melioidosis is an important cause of morbidity and mortality in East Asia. Recurrent melioidosis occurs in around 10% of patients following treatment either because of relapse with the same strain or re-infection with a new strain of Burkholderia pseudomallei. Distinguishing between the two is important but requires bacterial genotyping. The aim of this study was to develop a simple scoring system to distinguish re-infection from relapse.
Methods: In a prospective study of 2,804 consecutive adult patients with melioidosis presenting to Sappasithiprasong Hospital, NE Thailand, between 1986 and 2005, there were 141 patients with recurrent melioidosis with paired strains available for genotyping. Of these, 92 patients had relapse and 49 patients had re-infection. Variables associated with relapse or re-infection were identified by multivariable logistic regression and used to develop a predictive model. Performance of the scoring system was quantified with respect to discrimination (area under receiver operating characteristic curves, AUC) and categorization (graphically). Bootstrap resampling was used to internally validate the predictors and adjust for over-optimism.
Findings: Duration of oral antimicrobial treatment, interval between the primary episode and recurrence, season, and renal function at recurrence were independent predictors of relapse or re-infection. A score of < 5 correctly identified relapse in 76 of 89 patients (85%), whereas a score > or = 5 correctly identified re-infection in 36 of 52 patients (69%). The scoring index had good discriminative power, with a bootstrap bias-corrected AUC of 0.80 (95%CI: 0.73-0.87).
Conclusions: A simple scoring index to predict the cause of recurrent melioidosis has been developed to provide important bedside information where rapid bacterial genotyping is unavailable.