User login
A new point-of-care assay designed with machine learning offers improved accuracy for rapid identification of active tuberculosis (TB) infection, according to investigators.
, reported lead author Rushdy Ahmad, PhD, of the Broad Institute of MIT and Harvard in Cambridge, Mass., and colleagues. When fully developed, such a test could improve interventions for the most vulnerable patients, such as those with HIV, among whom TB often goes undiagnosed.
“Rapid and accurate diagnosis of active TB with current sputum-based diagnostic tools remains challenging in high-burden, resource-limited settings,” the investigators wrote. Their report is in Science Translational Medicine.
They went on to explain the gap that currently exists between microscopy, which is operator dependent and insensitive, and newer technologies, such as nucleic acid amplification, which are more sensitive but heavily resource dependent. “Furthermore, two of the most vulnerable and highly affected groups – young children and adults with HIV infection – are unlikely to be diagnosed using sputum because of difficulty obtaining sputum and low bacillary loads in the sample.”
To look for a more practical option, the investigators drew blood from 406 patients with chronic cough. Then, using a bead-based immunoassay with machine learning, the investigators identified four blood proteins associated with active TB infection: interleukin-6 (IL-6), IL-8, IL-18, and vascular endothelial growth factor (VEGF). Blind validation of 317 samples from patients with chronic cough in Asia, Africa, and South America showed that the four biomarkers offered a sensitivity of 80% and a specificity of 65%. By adding a fifth biomarker, an antibody against TB antigen Ag85B, the investigators were able to raise accuracy figures to 86% sensitivity and 69% specificity.
Adding even more biomarkers could theoretically raise accuracy even further, according to the investigators. The WHO minimal performance thresholds are 90% sensitivity and 70% specificity, with optimal targets slightly higher, at 95% sensitivity and 80% specificity. Although these standards have not yet been met, the investigators plan on testing the existing assay in real-world scenarios while simultaneously aiming to make it better.
“A near-term goal is ... to incrementally improve the marker panel up to an anticipated 6- to 10-plex assay,” the investigators wrote. “However, given the urgency of the problem, the possibility of incremental improvements will not delay platform refinement and field testing.”
The Bill and Melinda Gates Foundation funded the study. The investigators reported additional relationships with Quanterix Corporation and FIND.
SOURCE: Ahmad et al. Sci Transl Med. 2019 Oct 23. doi: 10.1126/scitranslmed.aaw8287.
A new point-of-care assay designed with machine learning offers improved accuracy for rapid identification of active tuberculosis (TB) infection, according to investigators.
, reported lead author Rushdy Ahmad, PhD, of the Broad Institute of MIT and Harvard in Cambridge, Mass., and colleagues. When fully developed, such a test could improve interventions for the most vulnerable patients, such as those with HIV, among whom TB often goes undiagnosed.
“Rapid and accurate diagnosis of active TB with current sputum-based diagnostic tools remains challenging in high-burden, resource-limited settings,” the investigators wrote. Their report is in Science Translational Medicine.
They went on to explain the gap that currently exists between microscopy, which is operator dependent and insensitive, and newer technologies, such as nucleic acid amplification, which are more sensitive but heavily resource dependent. “Furthermore, two of the most vulnerable and highly affected groups – young children and adults with HIV infection – are unlikely to be diagnosed using sputum because of difficulty obtaining sputum and low bacillary loads in the sample.”
To look for a more practical option, the investigators drew blood from 406 patients with chronic cough. Then, using a bead-based immunoassay with machine learning, the investigators identified four blood proteins associated with active TB infection: interleukin-6 (IL-6), IL-8, IL-18, and vascular endothelial growth factor (VEGF). Blind validation of 317 samples from patients with chronic cough in Asia, Africa, and South America showed that the four biomarkers offered a sensitivity of 80% and a specificity of 65%. By adding a fifth biomarker, an antibody against TB antigen Ag85B, the investigators were able to raise accuracy figures to 86% sensitivity and 69% specificity.
Adding even more biomarkers could theoretically raise accuracy even further, according to the investigators. The WHO minimal performance thresholds are 90% sensitivity and 70% specificity, with optimal targets slightly higher, at 95% sensitivity and 80% specificity. Although these standards have not yet been met, the investigators plan on testing the existing assay in real-world scenarios while simultaneously aiming to make it better.
“A near-term goal is ... to incrementally improve the marker panel up to an anticipated 6- to 10-plex assay,” the investigators wrote. “However, given the urgency of the problem, the possibility of incremental improvements will not delay platform refinement and field testing.”
The Bill and Melinda Gates Foundation funded the study. The investigators reported additional relationships with Quanterix Corporation and FIND.
SOURCE: Ahmad et al. Sci Transl Med. 2019 Oct 23. doi: 10.1126/scitranslmed.aaw8287.
A new point-of-care assay designed with machine learning offers improved accuracy for rapid identification of active tuberculosis (TB) infection, according to investigators.
, reported lead author Rushdy Ahmad, PhD, of the Broad Institute of MIT and Harvard in Cambridge, Mass., and colleagues. When fully developed, such a test could improve interventions for the most vulnerable patients, such as those with HIV, among whom TB often goes undiagnosed.
“Rapid and accurate diagnosis of active TB with current sputum-based diagnostic tools remains challenging in high-burden, resource-limited settings,” the investigators wrote. Their report is in Science Translational Medicine.
They went on to explain the gap that currently exists between microscopy, which is operator dependent and insensitive, and newer technologies, such as nucleic acid amplification, which are more sensitive but heavily resource dependent. “Furthermore, two of the most vulnerable and highly affected groups – young children and adults with HIV infection – are unlikely to be diagnosed using sputum because of difficulty obtaining sputum and low bacillary loads in the sample.”
To look for a more practical option, the investigators drew blood from 406 patients with chronic cough. Then, using a bead-based immunoassay with machine learning, the investigators identified four blood proteins associated with active TB infection: interleukin-6 (IL-6), IL-8, IL-18, and vascular endothelial growth factor (VEGF). Blind validation of 317 samples from patients with chronic cough in Asia, Africa, and South America showed that the four biomarkers offered a sensitivity of 80% and a specificity of 65%. By adding a fifth biomarker, an antibody against TB antigen Ag85B, the investigators were able to raise accuracy figures to 86% sensitivity and 69% specificity.
Adding even more biomarkers could theoretically raise accuracy even further, according to the investigators. The WHO minimal performance thresholds are 90% sensitivity and 70% specificity, with optimal targets slightly higher, at 95% sensitivity and 80% specificity. Although these standards have not yet been met, the investigators plan on testing the existing assay in real-world scenarios while simultaneously aiming to make it better.
“A near-term goal is ... to incrementally improve the marker panel up to an anticipated 6- to 10-plex assay,” the investigators wrote. “However, given the urgency of the problem, the possibility of incremental improvements will not delay platform refinement and field testing.”
The Bill and Melinda Gates Foundation funded the study. The investigators reported additional relationships with Quanterix Corporation and FIND.
SOURCE: Ahmad et al. Sci Transl Med. 2019 Oct 23. doi: 10.1126/scitranslmed.aaw8287.
FROM SCIENCE TRANSLATIONAL MEDICINE
Key clinical point: A new point-of-care assay designed with machine learning offers improved accuracy for rapid identification of active tuberculosis (TB) infection.
Major finding: The assay had a sensitivity of 86%.
Study details: A machine learning and validation study involving patients with chronic cough from multiple countries.
Disclosures: The Bill and Melinda Gates Foundation funded the study. The investigators reported relationships with Quanterix Corporation and FIND.
Source: Ahmad et al. Sci Transl Med. 2019 Oct 23. doi: 10.1126/scitranslmed.aaw8287.
