Portal de Periódicos da CAPES

Blood transcriptional signatures for tuberculosis diagnosis: a glass half-empty perspective – Authors' reply

2016; Elsevier BV; Volume: 4; Issue: 6 Linguagem: Inglês

10.1016/s2213-2600(16)30039-x

2213-2619

Timothy E. Sweeney, Purvesh Khatri,

Diagnosis and treatment of tuberculosis

Independent, prospective testing is key to understanding the strengths and weaknesses of any new diagnostic test, but all available evidence should be considered when drawing conclusions. Therefore, although we agree with Nicholas Walter and colleagues that sarcoidosis is an uncommon alternative diagnosis, we included all publicly available data in our analysis,1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google Scholar without selecting what to report. In this context, we are puzzled by the choice of Walter and colleagues to focus solely on the worst available result in the smallest comparison (GSE34608;2Maertzdorf J Weiner J Mollenkopf HJ et al.Common patterns and disease-related signatures in tuberculosis and sarcoidosis.Proc Natl Acad Sci USA. 2012; 109: 7853-7858Crossref PubMed Scopus (253) Google Scholar N=26) out of 2572 samples in 14 datasets that we analysed, and ignore other results. For example, Walter and colleagues do not acknowledge that the three-gene test was significantly better at distinguishing active tuberculosis from non-sarcoid other diseases.1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google Scholar Furthermore, Walter and colleagues state that our signature is “substantially less accurate” than previous transcriptional tests, which is contradicted by evidence that the three-gene set is superior to most previous tests in independent validation.1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google ScholarHere, we have updated our validation summary receiver operating characteristic (ROC) curves for latent versus active tuberculosis and other diseases versus active tuberculosis with the new dataset that Walter and colleagues have contributed (GSE73408; figure),3Walter ND Miller MA Vasquez J et al.Blood transcriptional biomarkers for active tuberculosis among patients in the United States: a case-control study with systematic cross-classifier evaluation.J Clin Microbiol. 2016; 54: 274-282Crossref PubMed Scopus (32) Google Scholar which profiled adults in the USA with latent tuberculosis, bacterial pneumonia, and acute tuberculosis. The summary ROC in validation cohorts for latent versus acute tuberculosis is 0·91 (95% CI 0·81–0·96), and for other diseases versus acute tuberculosis is 0·81 (0·67–0·91; which increases to 0·83 [0·68–0·93] if patients with sarcoidosis are removed). Thus, for comparisons of other diseases versus active tuberculosis in the validation data, the three-gene set can achieve 95% sensitivity, with a specificity of 45–59% (lowest for sarcoidosis; figure). WHO has specifically called for a rule-out test for active tuberculosis; used in this manner, the three-gene set would have a negative likelihood ratio of 0·09–0·11; at 10% prevalence, this translates into a negative predictive value of 99% (99 of 100 patients classified negative for acute tuberculosis would be true negatives).4WHOHigh-priority target product profiles for new tuberculosis diagnostics: report of a consensus meeting. World Health Organization, Geneva2014Google Scholar Thus, whereas standard culture or GeneXpert MTB/RIF might be appropriate in this setting for patients with adequate sputa, the three-gene test could fill a valuable niche as a rule-out test for patients without adequate sputa.4WHOHigh-priority target product profiles for new tuberculosis diagnostics: report of a consensus meeting. World Health Organization, Geneva2014Google Scholar To summarise, a glass half empty is also half full.The authors' declarations of interest remain the same as those declared in the original Article. Independent, prospective testing is key to understanding the strengths and weaknesses of any new diagnostic test, but all available evidence should be considered when drawing conclusions. Therefore, although we agree with Nicholas Walter and colleagues that sarcoidosis is an uncommon alternative diagnosis, we included all publicly available data in our analysis,1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google Scholar without selecting what to report. In this context, we are puzzled by the choice of Walter and colleagues to focus solely on the worst available result in the smallest comparison (GSE34608;2Maertzdorf J Weiner J Mollenkopf HJ et al.Common patterns and disease-related signatures in tuberculosis and sarcoidosis.Proc Natl Acad Sci USA. 2012; 109: 7853-7858Crossref PubMed Scopus (253) Google Scholar N=26) out of 2572 samples in 14 datasets that we analysed, and ignore other results. For example, Walter and colleagues do not acknowledge that the three-gene test was significantly better at distinguishing active tuberculosis from non-sarcoid other diseases.1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google Scholar Furthermore, Walter and colleagues state that our signature is “substantially less accurate” than previous transcriptional tests, which is contradicted by evidence that the three-gene set is superior to most previous tests in independent validation.1Sweeney TE Braviak L Tato CM Khatri P Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis.Lancet Respir Med. 2016; 4: 213-224Summary Full Text Full Text PDF PubMed Scopus (223) Google Scholar Here, we have updated our validation summary receiver operating characteristic (ROC) curves for latent versus active tuberculosis and other diseases versus active tuberculosis with the new dataset that Walter and colleagues have contributed (GSE73408; figure),3Walter ND Miller MA Vasquez J et al.Blood transcriptional biomarkers for active tuberculosis among patients in the United States: a case-control study with systematic cross-classifier evaluation.J Clin Microbiol. 2016; 54: 274-282Crossref PubMed Scopus (32) Google Scholar which profiled adults in the USA with latent tuberculosis, bacterial pneumonia, and acute tuberculosis. The summary ROC in validation cohorts for latent versus acute tuberculosis is 0·91 (95% CI 0·81–0·96), and for other diseases versus acute tuberculosis is 0·81 (0·67–0·91; which increases to 0·83 [0·68–0·93] if patients with sarcoidosis are removed). Thus, for comparisons of other diseases versus active tuberculosis in the validation data, the three-gene set can achieve 95% sensitivity, with a specificity of 45–59% (lowest for sarcoidosis; figure). WHO has specifically called for a rule-out test for active tuberculosis; used in this manner, the three-gene set would have a negative likelihood ratio of 0·09–0·11; at 10% prevalence, this translates into a negative predictive value of 99% (99 of 100 patients classified negative for acute tuberculosis would be true negatives).4WHOHigh-priority target product profiles for new tuberculosis diagnostics: report of a consensus meeting. World Health Organization, Geneva2014Google Scholar Thus, whereas standard culture or GeneXpert MTB/RIF might be appropriate in this setting for patients with adequate sputa, the three-gene test could fill a valuable niche as a rule-out test for patients without adequate sputa.4WHOHigh-priority target product profiles for new tuberculosis diagnostics: report of a consensus meeting. World Health Organization, Geneva2014Google Scholar To summarise, a glass half empty is also half full. The authors' declarations of interest remain the same as those declared in the original Article. Blood transcriptional signatures for tuberculosis diagnosis: a glass half-empty perspectiveIn The Lancet Respiratory Medicine, Timothy E Sweeney and colleagues1 used an innovative approach based on secondary data analysis to identify a novel whole-blood transcriptional signature for diagnosis of active tuberculosis. The investigators classified blood samples as tuberculosis or non-tuberculosis using a score based on expression levels of only three genes. Development of a simple, accurate, and parsimonious transcriptional signature for active tuberculosis in blood would represent a stunning achievement of enormous public health importance. Full-Text PDF Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysisOverall, our integrated multicohort analysis yielded a three-gene set in whole blood that is robustly diagnostic for active tuberculosis, that was validated in multiple independent cohorts, and that has potential clinical application for diagnosis and monitoring treatment response. Prospective laboratory validation will be required before it can be used in a clinical setting. Full-Text PDF