Xpert MTB/RIF for diagnosis of tuberculosis : performance variability and impact on patient-important outcomes

Accurate diagnosis and early treatment are key in achieving global targets to end tuberculosis (TB). Between 2000 and 2017, nearly 54 million deaths were averted due to timely diagnosis and treatment initiation. Xpert MTB/RIF (“Xpert”¸Cepheid, Sunnyvale, CA, US) is a molecular diagnostic test with integration of an automated sample processing system and hemi-nested real time polymerase chain reaction (PCR) in a single cartridge. It provides results within two hours and can be used at point-of-care. Xpert was endorsed by WHO in 2010 and currently is strongly recommended for adults and children suspected for MDR-TB or HIV and conditionally for all suspected with TB, should resources allow. By 2016, approximately 34 million Xpert cartridges were procured globally. Despite considerable development in evaluation and roll out of Xpert, important questions still remain to be addressed. Xpert has shown varibility in performance and it remains a question whether the observed variability is entirely explained by differences in populations and epidemiological burden or methodological differences in how Xpert is being evaluated such as effect of time to positivity (TTP) on a reference standard (culture) on sensitivity and how previous history of TB affects the specificity of Xpert. Accurate TTP can only be determined in automated systems such as MGIT culture system. Furthermore, studies that have assessed patient-important outcomes have shown inconsistent results. This doctoral thesis provides evidence on the overall population level pooled effect of Xpert on important patient outcomes based on the Cochrane review and evidence on effect of TTP and history of TB on sensitivity and specificity Xpert, respectively.
This thesis is based on three manuscripts:
Manuscript 1: Effect of time to culture positivity as a reference standard on Xpert MTB/RIF sensitivity for diagnosis of pulmonary tuberculosis
This manuscript shows the effect of TTP on the sensitivity of Xpert based on the analysis of data from a database of 16 different studies that evaluated Xpert with a total of 11,870 participants from ten different countries. The TTP was divided up in categories of five days up to 42 days. In all culture positive, sensitivity of Xpert in categories up to 15 days were: ≤ 5, 99.3% , (95% CI: 95-100, N=351) , >5≤10, 99.3 % (95%CI, 98.4-99.7, N=2231) and >10≤15, 96.8% (95%CI,94-98.3,N=1885). In smear negative-culture positive: ≤ 5, 99%, (95% CI: 62-100, N=37) , >5≤10, 98% (95%CI, 80-100, N=237) and >10≤15, 92% (95%CI,83-96 ,N=413). In HIV positive: ≤ 5, 96% , (95% CI: 63-100, N=51) , >5≤10, 98% (95%CI, 90-100, N=199) and >10≤15, 94% (95%CI,82-98 ,N=168) The sensitivity of Xpert in the first fifteen days was high in all three groups of analysis: all culture positive, smear positive-culture positive and smear negative-culture positive and irrespective of HIV status. A significant drop in sensitivity of Xpert when TTP of more than 15 days were assessd across all groups of analysis. Xpert sensitivity decreased with increasing TTP. This may explain some of the variation seen across different studies published on the diagnostic performance of Xpert. In settings with better TB control efforts, it is expected that patients are diagnosed earlier and TTP will be increased. Thus, reporting TTP of MGIT culture is important to ensure comparability of studies on Xpert as a diagnostic TB test.
Manuscript 2: Effect of a previous history of tuberculosis on the specificity of Xpert MTB/RIF
This manuscript shows the effect of history of TB based on the analysis of data from a database of 16 different studies that evaluated Xpert with a total of 11,870 participants. A total of 1,630 participants had a history of TB. The median time since previous TB treatment was 3 years (Interquartile range (IQR), 0.0-6). Among the 803 patients with a TB episode within two years of testing, the specificity of Xpert was 92.2% (95% CI 81-97). The specificity increased with time since previous TB. Between two and five years (373 participants), the specificity was 99.0% (95% CI; 86-100) and above five years (454 participants), 98.6% (95% CI; 85.4-99.8).This manuscript shows that a history of TB negatively affects Xpert specificity, and this effect is increased among those with TB in the last two years. This implies for a need of algorithms in patients with history of TB and positive Xpert to guide interpretation and management. The use of chest X-ray and clinical judgement remain relevant.
Manuscript 3: Impact of the diagnostic test Xpert MTB/RIF® on health outcomes for tuberculosis
This manuscript is based on a Cochrane review and summarizes the effect of Xpert on patient-important outcomes. The review included ten studies, seven of which were randomized controlled trials and three pre-post intervention studies. In the domains of the Cochrane risk of bias tool, most randomized studies had a low risk of bias. There was evidence of a positive effect of Xpert on tuberculosis confirmation in treated patients (RR 1.29 95%CI 1.11, 1.51 in the randomized trials) and reduction in pre-treatment loss to follow up (RR 0.59 95% CI 0.42-0.84). Overall there was a reduction on all-cause mortality of 12% (RR 0.88 (95% CI 0.73, 1.05) and 24% (RR 0.76 95% CI 0.58-1.00) among HIV positive participants. There was evidence that Xpert lead to an increase in the overall proportion of patients treated for TB (RR 1.10 95%CI 0.98, 1.23), the likelihood of being cured (OR 1.09 95% CI 1.02, 1.16), and that the proportion of those treated who were not microbiologically confirmed was reduced (RR 0.59 95%CI 0.41 0.85. This manuscript shows that compared with smear microscopy, Xpert reduces all-cause mortality by 12% although uncertainty around the effect estimate was high and the data was also compatible with reduction of up to 27% and an increase of up to 5%. The mechanisms by which Xpert could affect mortality is likely at least in part related to the reduction in pre-treatment loss to follow-up as well as the increase in the proportion of patients cured. Further studies should assess the role of empirical treatment on the impact of Xpert on patient outcomes.