Fig. 2.

The effect of a negative Xpert result on post-test probability and decision to treat, applied to HIV infected patients with smear-negative presumptive TB, in Uganda. For patient A, with a pre-test probability above the treatment threshold, but below the test-treat threshold, a negative Xpert results in withholding TB treatment. In Patient B, with a pre-test probability above the test-treat threshold, a negative Xpert result has no effect on the decision to treat, as the post-test probability is still higher than the treatment threshold. Thus, when the pre-test probability of the target disease is higher than the test-treat threshold, there is no need to test. The pre-test probability of TB was 17% in this setting. As the treatment threshold was not yet calculated for Uganda, we use in this example the 12%, calculated in Rwanda [5]. The test-treat threshold is the probability of disease at which there is no difference between testing and treating without testing. The distance between the two thresholds depends on the excluding power of negative test result. Given the weak power of negative Xpert result, the test-treat threshold is 28% (12% corresponds to an odds of 0.14, and a log₁₀ odds of −0.9. For a weak power-level 0.5 is added, on a log₁₀ odds scale. -0.4 log₁₀ odds corresponds with a 0.4 odds or 28% test-treat probability. The use of the log₁₀ odds scale has been published elsewhere [13])