Xpert TB test has important advantages but does not reduce morbidity
24 March 2014. Related: Conference reports, TB coinfection, CROI 21 (Retrovirus) 2014.
Nathan Geffen, CSSR
South Africa consumes more than half the Xpert MTB/RIF cartridges used worldwide and the device has been rolled out widely with the support of the Minister of Health.
It was hoped that Xpert would reduce morbidity and mortality because it is faster and more sensitive than sputum microscopy. However, although two recently published trials indicate that the Xpert has brought no mortality or morbidity benefit, there are still advantages to using this diagnostic.
In February, Grant Theron and the TB-NEAT team published the results of a randomised trial in The Lancet.  Participants with suspected TB (one or more symptoms according to WHO criteria) were randomised to what was effectively an onsite TB diagnostic test using the Xpert or smear microscopy. The primary outcome was TB morbidity measured using Karnofsky performance and a scale developed in 2008 by a Scandanavian team called TBscore. 
The TBscore scale is from 0 to 13, with lower scores being healthier, in contrast to Kanofsky performance. All the facilities either had laboratories on site or close by. The sites were located in Cape Town, Durban, Harare, Lusaka and Mbeya. Two sputum samples were collected from patients, one used either for the Xpert or microscopy test and the other for lab culture confirmation using MGIT.
The study was powered to detect a one-point difference in TBscore and 10 point difference in Kanofsky score.
From April 2011 to 30 March 2012, over 1,500 patients were recruited, of whom 758 were assigned to microscopy and 744 to Xpert. At baseline the median age was 37 (IQR: 30-46), 43% of participants were women, 26% had previous TB and 60% were HIV-positive, of whom 26% were on ART. Baseline Kanofsky score was 70 (IQR: 60-90) and TBScore was 5 (IQR: 4-7).
In the microscopy group, 114 were smear-positive of whom 91 were culture-positive, 20 were culture-negative and three culture-contaminated. 111 began treatment. Of the 643 smear-negative cases, 91 were culture-positive, 540 culture-negative and 12 culture-contaminated. (One patient had no microscopy done and was culture-negative.) Of these 212 were put on treatment after chest radiography and other considerations.
In the Xpert group, 184 were positive for TB. Of these 154 were culture-positive, 27 culture-negative and three culture-contaminated, and 182 participants began treatment. Of the 559 Xpert negative cases, 31 were culture-positive, 517 were culture-negative and 11 were culture-contaminated. (One patient had no Xpert result and was culture-negative.) After chest radiography and other considerations a further 139 participants began treatment.
Sensitivity of the Xpert [83%, 95%CI: 77-88] was significantly better than microscopy [50%, 95%CI: 43-57] with p < 0.0001. Specificity was similar on both tests: 97% [95%CI: 95-98] for microscopy vs 95% [95%CI: 93-97] for Xpert.
There was no difference in median [IQR] TBscore between the two groups at two months (TBscore: 2 [0-3] in microscopy vs 2 [0.25-3] in Xpert) or six months (1 [0-3] vs 1 [0-3]), nor for Kanofsky performance (80 [70-90] vs 90 [80-90]) at two months, or six months (100 [90-100] vs 100 [90-100]).
Nevertheless, there was a measured benefit to the Xpert. People in the Xpert group were more likely to get a same-day diagnosis (24% vs 13%; p <0.0001) and same-day treatment initiation (23% vs 15%; p=0.0002).
At CROI 2014, Gavin Churchyard presented the primary endpoint results of the XTEND study that looked at a similar question in a different way. 
This was a randomised cluster-controlled trial. Twenty South African clusters were randomly chosen either to use the Xpert or smear microscopy (10 clusters each). A cluster was defined as two clinics served by the same lab. In contrast to the TB-NEAT study, the diagnostic test was done at a supporting laboratory located away from the clinics. The primary outcome was mortality in adults investigated for TB.
Participants in this study provided one sputum sample if they were at an Xpert site and two sputum samples if they were at a microscopy site. Health workers could upon request take additional sputum samples for culture and perform X-rays.
Just under 5,000 participants were screened, 2,541 in the Xpert arm and 2,431 in the microscopy one, of whom 2,344 and 2,368 were enrolled respectively. In the Xpert arm, 20 people withdrew and 36 withdrew from the microscopy arm leaving 2,324 and 2,332 participants for analysis.
At baseline the median age was 36 and 62% were female. There were differences in self-reported HIV status between the arms. In the Xpert clusters, 73% said they knew their status vs 79% in the microscopy arm, but in both arms 62% of those with known status said they were HIV-positive. Median CD4 count was just over 300 in both arms.
The participants in the microscopy arm appear on average to have been less healthy at enrollment with significantly more having a body mass index (BMI) less than 18.5 (8.7% in the Xpert clusters vs 12.4% in microscopy). TB symptoms were also significantly worse in the microscopy arm with, for example, 9.8% reporting no symptoms in the Xpert arm vs 6% in the microscopy arm.
Deaths were high across the study with 4% of participants on the Xpert arm dying vs 5% on the microscopy arm but adjusted for age group, sex, BMI, number of TB symptoms and HIV status there was no difference between the arms. A Kaplan-Meier graph shows a constant risk of death in both arms across the six month follow-up period. Having multiple TB symptoms, low BMI, being male and older were all predictive of risk of death (p < 0.001 for all these).
Reporting being HIV positive and not being on ART conferred an adjusted odds ratio of death of 3.32 (95%CI: 2.03-5.41) over reporting being HIV negative. For participants who reported being HIV positive and on ART this was 1.79 (95%CI: 0.99-3.21). Interestingly, for participants who reported unknown status the adjusted odds ratio of death over those reporting being HIV negative was 2.41 (95%CI: 1.47-3.98; p <0.001).
Katherine Fielding presented the secondary endpoint results from the study. Diagnostic results were obtained for 97% of participants. On the Xpert arm 9.2% tested positive for TB vs 7.8% on the microscopy arm. Adjusted for age, sex, BMI and number of TB symptoms the prevalence ratio was 1.49 [95%CI: 1-2.23] times higher in the Xpert arm, suggesting much greater sensitivity by the Xpert diagnostic.
Loss-to-follow-up might be expected to be higher in the microscopy arm because of the longer time to diagnose patients and get them on treatment. But it was 17% in the Xpert arm by day 28 compared to 15% in the microscopy arm. When adjusted for BMI and number of TB symptoms there was no difference between the study arms.
Of the 200 participants with positive results in the Xpert arm, 4% (n=8) were positive for rifampicin resistance.
Overall, 541 participants (11.6%) were treated for TB over the six month follow up period, 10.8% on the Xpert arm and 12.5% on the microscopy one. The adjusted risk ratio showed no difference between the arms (1.04: 95%CI: 0.76-1.43).
A post hoc analysis reported that 71% people treated for TB were microbiologically confirmed to have TB, either by Xpert or culture (385/541), again without any significant difference between arms.
During question time it was suggested that nurses in the microscopy sites had to make a greater effort to diagnose their patients and keep them in care, which compensated for the Xpert’s better sensitivity and faster turnaround time.
While it is disappointing that the Xpert did not confer a morbidity or mortality benefit, it’s advantages include quicker and easier diagnosis with less work for health workers. Churchyard’s group will publish a cost-effectiveness analysis later this year.
The need for a highly sensitive and specific affordable point-of-care TB test remains urgent.
- Theron G et al. Feasibility, accuracy, and clinical effect of point-of-care Xpert MTB/RIF testing for tuberculosis in primary-care settings in Africa: a multicentre, randomised, controlled trial. The Lancet, 383 (9915); 424-435. (1-7 February 2014).
- Wejse C et al. TBscore: Signs and symptoms from tuberculosis patients in a low-resource setting have predictive value and may be used to assess clinical course. Scandinavian Journal of Infectious Diseases 2008; 40(2): 111-20.
- Churchyard G et al. Effect of Xpert MTB/RIF on early mortality in adults with suspected TB: a pragmatic randomized trial. 21st Conference on Retroviruses and Opportunistic Infections, 3-6 March 2014, Boston. Oral Abstract 95.
- Fielding KL et al. Xpert as the First-Line TB Test in South Africa: Yield, Initial Loss to Follow-Up, Proportion Treated. 21st Conference on Retroviruses and Opportunistic Infections, 3-6 March 2014, Boston. Oral Abstract 96LB.