Melioidosis is a significant cause of morbidity and mortality in east Asia and northern Australia (8). Early diagnosis is vital if patients are to receive effective treatment. Definitive laboratory diagnosis is based on the isolation of Burkholderia pseudomallei, but culture is often slow or not available in these regions. Techniques such as direct immunofluorescence (DIF) (9) and urinary antigen enzyme-linked immunosorbent assays (5) allow a rapid provisional diagnosis to be made with acceptable levels of accuracy but are expensive and not widely available. Rapid, cheap diagnostic methods are still needed. Direct microscopy using Gram staining may be helpful if typical slender, pale, bipolar-staining bacilli are seen, but this technique lacks specificity (7). The Wayson stain is a simple, cheap, modified methylene blue stain, originally used for diagnosing murine plague (6). Wayson staining gave better sensitivity than Gram staining did in suspected meningitis cases in one study (3). U.S. Army pathologists in Vietnam recommended the use of the Wayson stain (with the Gram stain) in cases of melioidosis (1) but did not report a formal assessment. We report a randomized, blind comparison of the Wayson and Gram stains for rapid diagnosis of melioidosis in an area of endemicity. Clinical specimens were collected from patients with suspected melioidosis who had been admitted to Sappasithiprasong Hospital, Ubon Ratchathani, Thailand (Table (Table1).1). Air-dried, heat-fixed smears were prepared for staining or DIF. B. pseudomallei was isolated and identified as described previously (4, 7). Paired slides were stained either with a modified Gram stain (with a neutral red counterstain) (7) or with the Wayson stain (2) for 10 seconds, rinsed in water, and air dried. Randomized slides were examined blind by three separate observers. Interobserver measures of agreement were assessed by means of the kappa statistic. TABLE 1. Types of specimen examined One hundred thirty-two consecutive specimens from 95 patients were included (Table (Table1).1). B. pseudomallei was isolated from 40 specimens (30%). The results for sensitivity, specificity, positive predictive value, negative predictive value, and Youden's J index (10) based on determinations by three observers for each stain were generally poor compared to the results of culture and are shown in Table Table2.2. Excluding sputum specimens or specimens with visible gram-positive diplococci because of a possible false-positive bipolarity did not enhance the diagnostic accuracy of either stain (data not shown). TABLE 2. Comparison of Gram and Wayson stain results with culture resultsa DIF results were available for 110 specimens. The results of comparisons between Gram and Wayson stain results and DIF results were similar to those of comparisons made between stain results and those of culture (data not shown). There was satisfactory agreement between the three observers regarding the results of the Gram stain (Κ = 0.72) but not regarding those of the Wayson stain (Κ = 0.36), suggesting considerable variability in interpretation of the latter. Our results suggest that the Gram and Wayson stains are both insufficiently sensitive and specific, compared to either culture or to DIF, to be relied upon for rapid presumptive diagnoses in cases of suspected melioidosis. Even when corrected for the presence of gram-positive diplococci, the sensitivity approached approximately 50% at best. Combining Gram and Wayson stain results did not significantly improve the diagnostic sensitivity. We also found a wide variation in the reporting of the presence of bipolar rods based on the Wayson stain, which may represent a lack of familiarity with the stain but which further limits its potential utility.