Comparative performance characteristics of the urine lipoarabinomannan strip test and sputum smear microscopy in hospitalized HIV-infected patients with suspected tuberculosis in Harare, Zimbabwe.

BACKGROUND

In Zimbabwe, sputum smear microscopy (SSM) is the routinely used TB diagnostic tool in hospitalised HIV-infected patients. However, SSM has poor sensitivity in HIV-infected patients. We compared performance of urine lipoarabinomannan strip test (LAM) and SSM among hospitalized HIV-infected patients with suspected TB.

METHODS

Hospitalized HIV-infected patients with suspected TB were randomized to LAM plus SSM or SSM alone groups as part of a larger multi-country parent study. Here we present a comparison of LAM versus SSM performance from the Zimbabwe study site. LAM analyses (grade 2 cut-off) were conducted using (i) a microbiological reference standard (MRS; culture positivity for M.tb and designated definite TB) and (ii) a composite reference standard (CRS; definite TB plus probable TB i.e. patients with clinical TB excluded from the culture negative group). CRS constituted the primary analysis.

RESULTS

82/457 (18%) of the patients randomized to the LAM group were M.tuberculosis culture positive. Using CRS, sensitivity (%, 95% CI) of LAM was significantly higher than SSM [49.2 (42.1-56.4) versus 29.4(23.2-36.3); p < 0.001]. Specificity and PPV were 98.1%, and 95.8%, respectively. By contrast, using MRS, LAM sensitivity was similar to SSM and specificity was significantly lower, however, the combined sensitivity of LAM and SSM was significantly higher than that of SSM alone, p = 0.009. Using CRS, LAM sensitivity (%, CI) was CD4 count dependent [60.6(50.7-69.8) at ≤50 cells/μL; 40.0(22.7-59.4) at 51-100 cells/μL, and 32.8(21.0-46.3) at >100 cells/μL. The combined sensitivity of LAM and SSM was higher than SSM alone being highest at CD4 counts <50 cells/μL [67.6(57.9-76.3); p = <0.001]. Specificity of LAM or SSM alone, or of combined LAM and SSM was >97% in all the 3 CD4 strata.

CONCLUSION

Among hospitalized HIV-infected patients with suspected TB, the sensitivity of LAM is significantly higher than that of SSM, especially at low CD4 counts. LAM and SSM are complimentary tests for diagnosis of TB in HIV-infected patients. We recommend a combination of LAM and SSM for TB diagnosis in HIV-infected patients with low CD4 counts in HIV/TB co-endemic countries, where alternative methods are unavailable.