Point-of-Care Urine Test for TB Screening Among HIV+ Adults
Among the 411 newly-diagnosed HIV-infected adults enrolled, 342 (83.2%) participants completed urine LAM and sputum culture testing, and were not receiving anti-tubercular therapy (Figure 1). Forty-five (13.2%) were urine LAM positive, with an estimated population prevalence of 9.8–17.0%. Based on this sample, the estimated sputum culture-positive tuberculosis prevalence was 17.5% (95% CI 13.7–22.0%). The mean age was 35.6 years (S.D. ±9.8 years), and 190 (55.6%) participants were male ( Table 1 ). Twenty-nine (8.5%) participants reported a history of previous tuberculosis disease. Three (0.9%) participants were taking diuretic medications at the time of LAM testing. Two hundred-eighty (81.9%) underwent CD4 testing; the median CD4 cell count was 182/mm (IQR 70–298/mm). Among the entire cohort, 44 people (11%) were unable to provide an expectorated or induced sputum sample, while 13 people (3%) were unable to provide a urine sample.
(Enlarge Image)
Figure 1.
Participant flow diagram.
Among this cohort, 24 (7.0%) participants were sputum smear microscopy (AFB) positive, with an estimated population prevalence of 4.6–10.3%. Sixty-three (18.4%) participants had either a positive sputum AFB or urine LAM test, with an estimated population prevalence of 14.5–22.9% ( Table 1 ). Among urine LAM positive participants, 36 (80.0%) had two positive urine LAM tests, and 9 (20.0%) had one positive urine LAM test. We performed urine LAM testing on the same day as specimen collection for 294 (86.0%) participants. We performed urine LAM testing on frozen samples for 48 (14.0%) participants, as a result of an interruption in LAM test supply. Among the samples tested in the clinic, the mean time from urine specimen collection to obtaining a LAM test result was 40.0 minutes (95% CI 34.2–45.5 minutes).
Urine LAM test estimated sensitivity and specificity for diagnosing culture-confirmed pulmonary tuberculosis were 28.3% (95% CI 17.5–41.4%) and 90.1% (95% CI 86.0–93.3%), respectively ( Table 2 ). Urine LAM test sensitivity was 37.5% (95% CI 21.1–56.3%) for those with CD4 cell count <100/mm, which did not differ significantly from those with a CD4 count > l00/mm (sensitivity =25.0%, 95% CI 9.0–49.0%; p = 0.38). In addition, a test for trend did not find significantly greater sensitivity at lower CD4 count strata (p = 0.46). Estimated test sensitivity and specificity for sputum AFB were 18.3% (95% CI 9.5–30.4%) and 95.4% (95% CI 92.3–97.5%), respectively. The urine LAM test identified 10% more cases of culture-positive pulmonary tuberculosis than sputum AFB. Likelihood ratio (LR) positive and LR negative values for urinary LAM testing were 2.85 (95% CI - 1.67–4.87) and 0.80 (95% CI - 0.68–0.94), respectively. The LR positive and negative for sputum AFB were 3.98 (95% CI 1.87–8.44) and 0.86 (95% CI 0.76–0.97), respectively. Among the 9 LAM discordant results, 2 people were culture-positive for tuberculosis and 1 additional person was clinically suspected of having active tuberculosis. There were no significant differences in baseline characteristics or urine LAM test results among the 48 frozen urine samples compared to those with a fresh urine sample, or participants without a CD4 result compared to those with a CD4 result.
When using a combined screening strategy of either urine LAM or sputum AFB positive, test sensitivity and specificity were 38.3% (95% CI 26.0–51.8%) and 85.8% (95% CI 81.1–89.7%). The addition of urine LAM to sputum AFB testing detected an additional 20% of culture-positive pulmonary tuberculosis cases, which was a statistically significant improvement (p = 0.02). With this combined screening strategy, the LR positive and negative values were 2.70 (95% CI 1.76–4.16) and 0.72 (95% CI 0.59–0.88).
Within our cohort, 92 (26.9%) participants met our definition of "clinically suspected tuberculosis", which gives an estimated population prevalence of 22.3–31.9%. Nine (2.6%) participants were diagnosed with extrapulmonary TB, which gives an estimated population prevalence of 1.2–4.9%. Among those 32 participants started on therapy with a negative sputum culture, 9 (28.1%) had positive AFB testing, 6 (18.8%) were diagnosed with extrapulmonary tuberculosis, and 16 (50.0%) were diagnosed based on clinical signs, symptoms, and/or chest radiography. In addition, 1 (3.1%) participant was diagnosed as sputum culture-positive on a sputum sample sent through the hospital's clinical laboratory, and was started on anti-tubercular therapy, but the sputum culture at our reference laboratory remained negative.
We assessed diagnostic performance characteristics of the urine LAM test using our secondary case definition of clinically-suspected active tuberculosis ( Table 3 ). In these analyses, urine LAM test sensitivity and specificity were 25.0% (95% CI - 16.6–35.1%) and 91.2% (95% CI - 87.0–94.4%), respectively. The LR positive and negative were 2.84 (95% CI - 1.67–4.48) and 0.82 (95% CI - 0.73–0.93). These results were almost identical to the culture-confirmed pulmonary tuberculosis, and CD4 strata had little impact on test characteristics. In addition, diagnostic accuracy of the urine LAM test was not associated with gender, prior tuberculosis infection, or diuretic use for both culture-confirmed and clinical suspected tuberculosis.
Among the 24 sputum AFB positive participants, 6 (25%) were urine LAM positive and 18 (75%) were urine LAM negative. Among the 60 participants with culture-confirmed pulmonary tuberculosis, urine LAM testing identified 12 (20%) participants more than sputum AFB, and sputum AFB testing identified 6 (10%) participants more than urine LAM (Figure 2a). Only 5 of 60 (8%) sputum culture-positive participants were identified by both sputum AFB and urine LAM testing.
(Enlarge Image)
Figure 2.
Urine LAM and sputum smear microscopy (AFB) to diagnose tuberculosis among those who were either culture-confirmed (Figure 2a) or diagnosed with clinically suspected (Figure 2b) tuberculosis. a. Culture-confirmed Pulmonary Tuberculosis (N = 60). b. Clinically Suspected Tuberculosis (N = 92). Any participant who were LAM + but not diagnosed with culture-confirmed or clinically suspected tuberculosis are not depicted in these figures.
Among the 92 participants with clinically suspected tuberculosis, urine LAM testing identified 17 (18%) participants more than sputum AFB, and sputum AFB testing identified 11 (12%) participants more than urine LAM (Figure 2b). Only 6 of 92 (7%) participants with clinically suspected tuberculosis were identified by both sputum AFB and urine LAM testing.
A representation of post-test probability of tuberculosis when using a combined screening strategy of sputum AFB and urine LAM testing is shown in Figure 3. If the baseline prevalence of tuberculosis is 10% (column A), negative sputum AFB and urine LAM decreased the post-test probability to 7.4%, while either positive test increased the post-test probability to 23.1%. When assuming a prevalence of tuberculosis of 17.5% (column B), as observed for culture-positive pulmonary tuberculosis in our cohort, positive sputum AFB or urine LAM increased post-test probability to 36.4%. When assuming a prevalence of tuberculosis of 26.9% (column C), as observed for clinically suspected tuberculosis in our cohort, positive sputum AFB or urine LAM increased post-test probability to 49.8%. When assuming a higher pre-test probability of tuberculosis (40%; column D), negative sputum AFB and urine LAM decreased the post-test probability of tuberculosis to 32.4%, while either positive test increased post-test probability to 64.3%.
(Enlarge Image)
Figure 3.
Post-test probability of tuberculosis when using a combined screening strategy of sputum AFB and urine LAM testing for various pre-test probabilities. Column A represents a cohort with 10% prevalence of tuberculosis; columns B and C represent the baseline prevalence of culture-positive pulmonary tuberculosis (17.5%) and clinically suspected tuberculosis (26.9%) that we observed in our cohort. Column D represents a cohort with a 40% prevalence of tuberculosis, which might occur among tuberculosis suspects in a highly endemic region.
Results
Among the 411 newly-diagnosed HIV-infected adults enrolled, 342 (83.2%) participants completed urine LAM and sputum culture testing, and were not receiving anti-tubercular therapy (Figure 1). Forty-five (13.2%) were urine LAM positive, with an estimated population prevalence of 9.8–17.0%. Based on this sample, the estimated sputum culture-positive tuberculosis prevalence was 17.5% (95% CI 13.7–22.0%). The mean age was 35.6 years (S.D. ±9.8 years), and 190 (55.6%) participants were male ( Table 1 ). Twenty-nine (8.5%) participants reported a history of previous tuberculosis disease. Three (0.9%) participants were taking diuretic medications at the time of LAM testing. Two hundred-eighty (81.9%) underwent CD4 testing; the median CD4 cell count was 182/mm (IQR 70–298/mm). Among the entire cohort, 44 people (11%) were unable to provide an expectorated or induced sputum sample, while 13 people (3%) were unable to provide a urine sample.
(Enlarge Image)
Figure 1.
Participant flow diagram.
Among this cohort, 24 (7.0%) participants were sputum smear microscopy (AFB) positive, with an estimated population prevalence of 4.6–10.3%. Sixty-three (18.4%) participants had either a positive sputum AFB or urine LAM test, with an estimated population prevalence of 14.5–22.9% ( Table 1 ). Among urine LAM positive participants, 36 (80.0%) had two positive urine LAM tests, and 9 (20.0%) had one positive urine LAM test. We performed urine LAM testing on the same day as specimen collection for 294 (86.0%) participants. We performed urine LAM testing on frozen samples for 48 (14.0%) participants, as a result of an interruption in LAM test supply. Among the samples tested in the clinic, the mean time from urine specimen collection to obtaining a LAM test result was 40.0 minutes (95% CI 34.2–45.5 minutes).
Diagnostic Accuracy of LAM for Culture-Confirmed Pulmonary Tuberculosis
Urine LAM test estimated sensitivity and specificity for diagnosing culture-confirmed pulmonary tuberculosis were 28.3% (95% CI 17.5–41.4%) and 90.1% (95% CI 86.0–93.3%), respectively ( Table 2 ). Urine LAM test sensitivity was 37.5% (95% CI 21.1–56.3%) for those with CD4 cell count <100/mm, which did not differ significantly from those with a CD4 count > l00/mm (sensitivity =25.0%, 95% CI 9.0–49.0%; p = 0.38). In addition, a test for trend did not find significantly greater sensitivity at lower CD4 count strata (p = 0.46). Estimated test sensitivity and specificity for sputum AFB were 18.3% (95% CI 9.5–30.4%) and 95.4% (95% CI 92.3–97.5%), respectively. The urine LAM test identified 10% more cases of culture-positive pulmonary tuberculosis than sputum AFB. Likelihood ratio (LR) positive and LR negative values for urinary LAM testing were 2.85 (95% CI - 1.67–4.87) and 0.80 (95% CI - 0.68–0.94), respectively. The LR positive and negative for sputum AFB were 3.98 (95% CI 1.87–8.44) and 0.86 (95% CI 0.76–0.97), respectively. Among the 9 LAM discordant results, 2 people were culture-positive for tuberculosis and 1 additional person was clinically suspected of having active tuberculosis. There were no significant differences in baseline characteristics or urine LAM test results among the 48 frozen urine samples compared to those with a fresh urine sample, or participants without a CD4 result compared to those with a CD4 result.
When using a combined screening strategy of either urine LAM or sputum AFB positive, test sensitivity and specificity were 38.3% (95% CI 26.0–51.8%) and 85.8% (95% CI 81.1–89.7%). The addition of urine LAM to sputum AFB testing detected an additional 20% of culture-positive pulmonary tuberculosis cases, which was a statistically significant improvement (p = 0.02). With this combined screening strategy, the LR positive and negative values were 2.70 (95% CI 1.76–4.16) and 0.72 (95% CI 0.59–0.88).
Diagnostic Accuracy of LAM for Clinically Suspected Tuberculosis
Within our cohort, 92 (26.9%) participants met our definition of "clinically suspected tuberculosis", which gives an estimated population prevalence of 22.3–31.9%. Nine (2.6%) participants were diagnosed with extrapulmonary TB, which gives an estimated population prevalence of 1.2–4.9%. Among those 32 participants started on therapy with a negative sputum culture, 9 (28.1%) had positive AFB testing, 6 (18.8%) were diagnosed with extrapulmonary tuberculosis, and 16 (50.0%) were diagnosed based on clinical signs, symptoms, and/or chest radiography. In addition, 1 (3.1%) participant was diagnosed as sputum culture-positive on a sputum sample sent through the hospital's clinical laboratory, and was started on anti-tubercular therapy, but the sputum culture at our reference laboratory remained negative.
We assessed diagnostic performance characteristics of the urine LAM test using our secondary case definition of clinically-suspected active tuberculosis ( Table 3 ). In these analyses, urine LAM test sensitivity and specificity were 25.0% (95% CI - 16.6–35.1%) and 91.2% (95% CI - 87.0–94.4%), respectively. The LR positive and negative were 2.84 (95% CI - 1.67–4.48) and 0.82 (95% CI - 0.73–0.93). These results were almost identical to the culture-confirmed pulmonary tuberculosis, and CD4 strata had little impact on test characteristics. In addition, diagnostic accuracy of the urine LAM test was not associated with gender, prior tuberculosis infection, or diuretic use for both culture-confirmed and clinical suspected tuberculosis.
Comparison Between Sputum AFB and Urine LAM Testing
Among the 24 sputum AFB positive participants, 6 (25%) were urine LAM positive and 18 (75%) were urine LAM negative. Among the 60 participants with culture-confirmed pulmonary tuberculosis, urine LAM testing identified 12 (20%) participants more than sputum AFB, and sputum AFB testing identified 6 (10%) participants more than urine LAM (Figure 2a). Only 5 of 60 (8%) sputum culture-positive participants were identified by both sputum AFB and urine LAM testing.
(Enlarge Image)
Figure 2.
Urine LAM and sputum smear microscopy (AFB) to diagnose tuberculosis among those who were either culture-confirmed (Figure 2a) or diagnosed with clinically suspected (Figure 2b) tuberculosis. a. Culture-confirmed Pulmonary Tuberculosis (N = 60). b. Clinically Suspected Tuberculosis (N = 92). Any participant who were LAM + but not diagnosed with culture-confirmed or clinically suspected tuberculosis are not depicted in these figures.
Among the 92 participants with clinically suspected tuberculosis, urine LAM testing identified 17 (18%) participants more than sputum AFB, and sputum AFB testing identified 11 (12%) participants more than urine LAM (Figure 2b). Only 6 of 92 (7%) participants with clinically suspected tuberculosis were identified by both sputum AFB and urine LAM testing.
Post-test Probability of Tuberculosis
A representation of post-test probability of tuberculosis when using a combined screening strategy of sputum AFB and urine LAM testing is shown in Figure 3. If the baseline prevalence of tuberculosis is 10% (column A), negative sputum AFB and urine LAM decreased the post-test probability to 7.4%, while either positive test increased the post-test probability to 23.1%. When assuming a prevalence of tuberculosis of 17.5% (column B), as observed for culture-positive pulmonary tuberculosis in our cohort, positive sputum AFB or urine LAM increased post-test probability to 36.4%. When assuming a prevalence of tuberculosis of 26.9% (column C), as observed for clinically suspected tuberculosis in our cohort, positive sputum AFB or urine LAM increased post-test probability to 49.8%. When assuming a higher pre-test probability of tuberculosis (40%; column D), negative sputum AFB and urine LAM decreased the post-test probability of tuberculosis to 32.4%, while either positive test increased post-test probability to 64.3%.
(Enlarge Image)
Figure 3.
Post-test probability of tuberculosis when using a combined screening strategy of sputum AFB and urine LAM testing for various pre-test probabilities. Column A represents a cohort with 10% prevalence of tuberculosis; columns B and C represent the baseline prevalence of culture-positive pulmonary tuberculosis (17.5%) and clinically suspected tuberculosis (26.9%) that we observed in our cohort. Column D represents a cohort with a 40% prevalence of tuberculosis, which might occur among tuberculosis suspects in a highly endemic region.
SHARE
