GeneXpert demonstrates good sensitivity and specificity but at high cost
1 February 2011. Related: Conference reports, TB coinfection, World Conference on Lung Health 41st 2010.
Nathan Geffen, TAC
Catherine Boehme of FIND and her colleagues have published the test results of Cepheids GeneXpert TB diagnostic technology in the NEJM. 
Previously, in the April 2010 edition of HTB South, we published a report on this test. 
This test aims to diagnose TB and and also determine rifampicin resistance in less than two hours. Preliminary results have been good. This study confirms that this device has high sensitivity and specificity in a variety of settings in both HIV-positive and HIV-negative patients and in culture-positive sputum-negative patients.
Over 1,800 participants were screened at five sites located in Lima, Baku, Cape Town, Mumbai and Durban. 1,730 participants were able to provide three sputum samples with sufficient volume and were consequently eligible for the study.
Of these, 268 were excluded from final analysis, 115 because they were culture-negative and suspected of MDR TB while receiving treatment, 28 because three or more of their cultures were contaminated, 23 because they had growth of non-MTB only, 10 because they had indeterminate phenotypic rifampicin results, 39 because they were smear-positive but culture-negative, seven because they had suspected culture cross-contamination and 46 because they died or were lost to follow-up.
Of the 1,462 participants included in the main analysis 741 were culture-positive, of whom 567 were smear-positive and 174 were smear-negative. Of the 721 culture-negative cases, 105 had clinical TB and 616 did not have TB (as determined by a clinical review committee).
As explained in our previous article, the GeneXpert consists of a computer installed with Cepheids proprietary software and a machine the smallest of which is about the size of a desktop computer that takes cartridges loaded with sputum and reagents. The cartridges consist of a syringe barrel, a sonicator dome, a reverse-transcriptase PCR tube and a rotary valve. The smallest version of the machine takes four cartridges. The highest capacity one apparently contains 100 cartridges. As explained below, two or even three cartridges might be needed for a patient.
The screening results and consequent inclusion and exclusion criteria of patients in various analyses is complicated in this study. Table 1 presents an overview that readers can refer to when reading the remainder of this summary.
Table 1: Screening results. Adapted from Boehme et al.
|Number of patients screened||1,843|
|Number of patients eligible||1,730|
|Number of eligible patients excluded||268|
|– Excluded because culture-negative suspected MDR TB while receiving therapy||115|
|– Contamination of =3 of 4 cultures||28|
|– Had growth of non-MTB only||23|
|– Indeterminate phenotypic rifampicin result||10|
|– Smear-positive sample with all cultures negative||39|
|– Suspected Cross culture contamination||7|
|– Died or lost-to-follow up||46|
|Included in main analysis||1,462|
|– Clinical TB||105|
|– No TB||616|
TB sensitivity and specificity
With one sputum sample, the Gene Xpert had a sensitivity of 92% for all culture-positive specimens. This increased to 96% for two samples and 98% for three. Specificity on non-TB cases was 99% with one sputum sample, declining marginally to 98% with three samples. However, for culture-positive, sputum-negative specimens, sensitivity using one sputum sample was 73% rising to 90% with three samples. No site had a sensitivity lower than 83% for culture-positive, sputum-negative specimens.
Further details including confidence intervals are provided in Table 2.
Table 2: Sensitivity and specificity of Gene Xpert on culture-positive patients and culture-negative patients not treated for TB. Adapted slightly from Boehme et al.
|Site and No of Tests||All culture-positive
Number correct/Total (%)
|Culure-positive and smear-positive
Number correct/Total (%)
|Culture-positive and smear-negative
Number correct/Total (%)
Number correct/Total (%)
|Lima||209/211 (99.1)||199/199 (100)||10/12 (83.3)||102/102 (100)|
| 95% CI||96.699.7||98.1100.0||55.295.3||96.4100.0|
|Baku||144/149 (96.6)||80/80 (100.0)||64/69 (92.8)||68/70 (97.1)|
|Cape Town||142/148 (95.9)||95/96 (99.0)||47/52 (90.4)||186/189 (98.4)|
|Durban||43/45 (95.6)||30/30 (100.0)||13/15 (86.7)||213/219 (97.3)|
|Mumbai||185/188 (98.4)||162/162 (100.0)||23/26 (88.5)||35/36 (97.2)|
|Three sputum samples used||723/741 (97.6)||566/567 (99.8)||157/174 (90.2)||604/616 (98.1)|
|Two sputum samples used||1423/1482 (96.0)||1127/1134 (99.4)||296/348 (85.1)||1215/1232 (98.6)|
|One sputum sample||675/732 (92.2)||551/561 (98.2)||124/171 (72.5)||604/609 (99.2)|
Sensitivity was 94% in HIV-positive patients with pulmonary TB versus 98% in HIV-negative patients (p=0.02). Of the 105 patients with culture-negative samples excluded from the main analysis but who had clinical signs of TB, 29.3% had positive results on the Gene Xpert.
Rifampicin sensitivity and specificity
Of the 723 culture-positive patients correctly identified as having TB by the Gene Xpert, 720 were tested phenotypically for rifampicin resistance (for the remaining three, the Gene Xpert gave indeterminate resistance results). The Gene Xpert identified 200 out of 205 rifampicin resistant specimens correctly for a sensitivity of 98%. It identified 505 out of 515 rifampicin sensitive specimens correctly for a specificity of 98%.
Details of resistance testing with confidence intervals are presented in table 3.
Table 3: Sensitivity and specificity of Gene Xpert on phenotypically determined rifampicin susceptibility. Adapted from Boehme et al.
|Site||Sensitivity – number of specimens correctly identified as rifampicin resistant (%)||Specificity – number of specimens correctly identified as rifampicin sensitive (%)|
|Lima||16/16 (100)||190/193 (98.4)|
|Baku||47/49 (95.9)||90/94 (95.7)|
|Cape Town||15/16 (93.8)||126/126 (100)|
|Durban||3/3 (100)||38/38 (100)|
|Mumbai||119/121 (98.3)||61/64 (95.3)|
The authors also did a second analysis that included the results of gene sequencing of the 15 discrepant results between phenotyping and the Gene Xpert. After three of these were excluded from analysis because sequencing gave indeterminate results, sensitivity was 99.1% [95%CI: 96.6-99.7] (209/211 correct) and specificity was 100% [95%CI: 99.2100.0] (506 correct).
Importantly, 195 out of 200 of the rifampicin resistant specimens were also resistant to isoniazid. This suggests that rifampicin resistance is a good predictor of MDR TB in practice.
In 115 patients, excluded from the main analysis in the study, who were culture-negative but who were diagnosed with MDR TB and consequently received treatment, 51 had positive results on the Gene Xpert. Rifampicin resistance was detected in eight. Interestingly, the authors note that all eight patients were later started on second-line therapy by physicians unaware of the results of the Gene Xpert results.
These results are promising. The Gene Xpert is much easier to use than sputum microscopy. It has a high sensitivity and specificity and appears to be better than culture in a subset of patients who are culture-negative but nevertheless have TB. It has high sensitivity and specificity for detecting rifampicin resistance. The diagnostic can be used in facilities that offer consistent electricity supply. One drawback, as with most TB diagnostics, is that patients need to provide sputum and preferably as many as three samples.
But the main obstacle to wider use of the Gene Xpert will be its price. The cheapest machine reportedly costs $20,000. Each cartridge costs approximately $20. There is a great need for better TB diagnostics primarily in poor communities. Pressure needs to be exerted on Cepheid to bring down the price of this system, which was in any case developed with substantial public investment. Conversely pressure needs to be placed on international TB bodies to fund the implementation of diagnostics such as this one in resource-poor settings.
- Geffen N. 2010. Cepheid Gene Xpert diagnostic technology for TB. HTB South April 2010.
- Boehme C et al. 2010. Rapid Molecular Detection of Tuberculosis and Rifampin Resistance. NEJM September 2010.