Melioidosis is associated with significant mortality in countries in which it is endemic. Previous studies have demonstrated that quantitative Burkholderia pseudomallei counts in blood are predictive of mortality. Here we examine the relationship between outcomes and quantitative B. pseudomallei counts in urine. A total of 755 patients presenting to Sappasithiprasong Hospital, Ubon Ratchathani, northeast Thailand (in the northeast part of the country), with melioidosis between July 1993 and October 2003 had quantitative urine cultures performed within 72 h of admission. Urine culture results were divided into the following groups: (i) no growth of B. pseudomallei from a neat sample or pellet, (ii) positive result from a centrifuged pellet only (< 10(3) CFU/ml), (iii) detection of between 10(3) CFU/ml and 10(5) CFU/ml from a neat sample, or (iv) detection of > or = 10(5) CFU/ml from a neat sample. The overall in-hospital mortality rate was 45%. Patients with negative urine cultures had the lowest death rate (39%). Mortality rates rose with increasing B. pseudomallei counts in urine, from 58% for those with positive spun pellets only to 61% for those with between 10(3) CFU/ml and 10(5) CFU/ml and 71% for those with > or = 10(5) CFU/ml. This was independent of age, presence of bacteremia, known risk factors for melioidosis such as diabetes, and the prior administration of antibiotics. The presence of B. pseudomallei in urine during systemic infection is associated with a poor prognosis.