Performance of the G4 Xpert® MTB/RIF assay for the detection of Mycobacterium tuberculosis and rifampin resistance: a retrospective case-control study of analytical and clinical samples from high- and low-tuberculosis prevalence settings

Background The Xpert® MTB/RIF (Xpert) assay is a rapid PCR-based assay for the detection of Mycobacterium tuberculosis complex DNA (MTBc) and mutations associated with rifampin resistance (RIF). An updated version introduced in 2011, the G4 Xpert, included modifications to probe B and updated analytic software. Methods An analytical study was performed to assess Xpert detection of mutations associated with rifampin resistance in rifampin-susceptible and -resistant isolates. A clinical study was performed in which specimens from US and non-US persons suspected of tuberculosis (TB) were tested to determine Xpert performance characteristics. All specimens underwent smear microscopy, mycobacterial culture, conventional drug-susceptibility testing and Xpert testing; DNA from isolates with discordant rifampin resistance results was sequenced. Results Among 191 laboratory-prepared isolates in the analytical study, Xpert sensitivity for detection of rifampin resistance associated mutations was 97.7% and specificity was 90.8%, which increased to 99.0% after DNA sequencing analysis of the discordant samples. Of the 1,096 subjects in the four clinical studies, 49% were from the US. Overall, Xpert detected MTBc in 439 of 468 culture-positive specimens for a sensitivity of 93.8% (95% confidence interval [CI]: 91.2%–95.7%) and did not detect MTBc in 620 of 628 culture-negative specimens for a specificity of 98.7% (95% CI: 97.5%–99.4%). Sensitivity was 99.7% among smear-positive cases, and 76.1% among smear-negative cases. Non-determinate MTBc detection and false-positive RIF resistance results were low (1.2 and 0.9%, respectively). Conclusions The updated Xpert assay retained the high sensitivity and specificity of the previous assay versions and demonstrated low rates of non-determinate and RIF resistance false positive results. Electronic supplementary material The online version of this article (doi:10.1186/s12879-016-2039-4) contains supplementary material, which is available to authorized users.

of positive specimens. Two sputum samples were collected from each study participant; the 23 specimen provided for Xpert testing was based on sufficient volume available. Symptoms suggesting pulmonary TB were defined as: persistent cough generally ≥2 25 weeks and at least one other symptom suggestive of TB such as fever, malaise, recent weight 26 loss, night sweats, contact with an active TB case, hemoptysis, chest pain, loss of appetite. 27 Samples in Collection three and four were partially processed prior to Xpert testing by 28 homogenizing for one minute with a vortexer and glass beads and split into 500 µL aliquots. 29 Aliquots underwent microbiologic testing for MTB and were frozen at -70ºC until shipment on 30 dry ice to NJMS for testing by Xpert. Although partially processed, these specimens were treated 31 in the Xpert as if they were raw sputa.

33
Clinical Study 2 (CS2) 34 Up to 3 specimens were collected from each patient for MTB diagnosis. Subjects with at 35 least one culture positive specimen were consecutively selected for Xpert testing using the first 36 specimen with sufficient volume. MTB negative specimens were selected by choosing the first 37 available study participant with all MTB negative baseline cultures that were collected after an 38 eligible MTB positive case. An additional 25 MTB negative cases were randomly selected and 39 included for testing throughout the study.  Laboratory testing -Analytic study 55 The culture isolates were spiked at known concentrations of colony forming units 56 (CFU)/mL into aliquots of a pool of sputa known to be negative for MTB and for nontuberculous 57 Mycobacterium. Forty percent of the isolates were spiked at low concentrations near the assay 58 cutoff (1500 CFU/mL), 30% at moderate concentrations (15,000 CFU/mL) and 30% at high 59 concentrations (150,000 CFU/mL).
Which is the probability of passing given the Null Hypothesis is true. The actual probability calculations are as follows: Which is the probability of passing given the Null Hypothesis is true.  The four clinical studies were tested for homogeneity by analyzing the results across 122 multiple parameters using the Fisher's Exact Test. A critical p-value was set to 0.01 due to 123 multiple testing in several categories (Bonferroni principle). No differences were identified 124 across the following parameters (p>0.01): protocol, archived or fresh samples, processed or 125 unprocessed samples, gender, age group and site (data not shown). Some difference was found 126 for sensitivity across site (p=0.08), specificity across site (p=0.06) and specificity across US vs. 127 non-US samples (p=0.04) but these differences did not meet the cutoff for significance.

129
However, it was felt that the small number of induced samples relative to expectorated and the 130 relative large percentage of false negatives in the induced samples yielded the small p-value.

131
Forty percent (10/25) of the MTB positive culture induced specimens were AFB smear negative.

132
Of these ten specimens, six were false negative by Xpert MTB/RIF Assay. The sensitivity may 133 therefore have been lower in induced specimens due to the large percentage of AFB smear 134 negative specimens in this population and the potential dilutional effect of the induction 135 7 process[1]. The studies were therefore determined to be poolable and the results of testing from 136 all four clinical studies were combined and are presented together.