Background: South Africa is rapidly implementing the World Health Organization’s recommendation to adopt Xpert MTB/RIF technology (Xpert) for first-line diagnosis of pulmonary tuberculosis (TB). With a high burden of HIV/TB co-infection and a high proportion of patients with smear-negative TB, however, determining a diagnostic algorithm for Xpert negative TB suspects remains important. South Africa’s current Xpert algorithm calls for HIV-infected TB suspects with an initial negative Xpert result to provide a consecutive sputum sample for culture (X/C). We used a previously developed model to estimate the difference in cost and number of TB cases diagnosed and initiated on treatment if the algorithm incorporated a second Xpert (X/X) instead of culture. Methods: We estimated the incremental cost per TB patient treated of replacing culture with a second Xpert test for known or suspected HIV-positive TB suspects whose first Xpert test was negative, using a population-level decision model we developed for the South African government. Xpert test costs were calculated using data from a pilot evaluation. Public-sector salaries, drug and laboratory prices from 2011 were used for all other costs. The number of patients requiring diagnostics, TB treatment uptake, and loss to follow-up at each clinic visit were estimated using results from Xpert demonstration studies and TB and HIV positivity rates calculated from the national TB register and public-sector laboratory databases. Results: Xpert is more expensive per test than culture ($25 v $13) but X/C requires more non-laboratory resources (clinic visits, chest x-rays and antibiotics). At national programme scale, X/X ($136 million/year) is less expensive than X/C ($153 million/year). Because Xpert is less sensitive than culture for smear-negative suspects, the X/X algorithm diagnoses 3% fewer TB cases. This is partly offset by higher expected treatment uptake with X/X due to the faster availability of results, resulting in <1% fewer patients initiating treatment in X/X than X/C. The cost per TB patient initiated on treatment under X/X is $401, 10% less than under X/C ($448/ patient initiated). Conclusions: Modifying the diagnostic algorithm to use a second Xpert test for HIV-infected TB suspects who have a negative Xpert test at their first visit could provide rapid results, simplify the diagnostic process, and generate cost savings, without negatively impacting the number of patients initiating TB treatment.
Conference: CROI 2012, Seattle, USA