Sunday, February 12, 2012

Testing Times for TB: Statistics Distinguish

Multivariate statistical data processing has been used to create a model from gas chromatography-mass spectrometry (GC-MS) data of metabolite profiles of the various types of Mycobacterium species tuberculosis (TB). The model could allow diagnosticians and biomedical researchers to quickly and easily distinguish between various infectious Mycobacterium species.

Tuberculosis is relatively rare in the developed world, although there are concerns that pockets of incidence are on the rise in certain poverty-stricken sectors of society, among the homeless, HIV/AIDS patients and drug user communities. Needless to say, there are millions of cases of TB in the developing world and multiple-drug resistant species have recently been reported. TB most commonly attacks the lungs and is spread through coughs and sneezes by those with active infection. Most infections are asymptomatic but about one in ten latent cases progress to active disease and untreated will kill half of those people. The most recent data for 2010 says there were 8.8 million new cases, and 1.45 million deaths. Most cases are seen in sub-Saharan Africa, according to the World Health Organisation.

Current TB detection assays are very sensitive, spotting bacteria at concentrations as low as 10-100 cells per millilitre in the original patient sample. However, the test requires two to six weeks to incubate colonies by which time a patient's infection may have progressed or been passed on to others. Moreover, this so-called gold standard also has a 15-20% rate of false negatives in adults tested. Pre-processing also makes the approach susceptible to 1-4% false positives. There have been many improvements that help overcome these shortfalls but they are expensive systems and require the training of highly skilled staff.

Rapid Response Required

A straightforward, sensitive and rapid test that primarily identify TB easily but also distinguish between potentially lethal and benign Mycobacterium bacteria.

Ilse Olivier and Du Toit Loots of the Centre for Human Metabonomics, at North-West University, in Potchefstroom, South Africa, used a modified Bligh-Dyer extraction procedure to obtain lipid components from Mycobacterium tuberculosis, M. avium, M. bovis, and M. kansasii cultures. They applied principle component analyses (PCA) to the GC-MS data on these extracts.

The team then identified the twelve compounds that best show the variation between the sample groups, suggesting that these might be the preferred metabolite markers for the pathogenic species. PCA and partial least-squares discriminant analysis (PLS-DA) was applied and the metabolite markers used to build a Bayesian statistical classification model to discriminate between the metabolite marker profile of novel samples. Tests of this model correctly identified two blinded samples for each of the Mycobacterium species analysed. The team reports a certainty ranging from 72 to 100% in the Journal of Microbiological Methods.

Sensitive and Speedy

The team adds that the test procedure requires less than 16 hours to carry out and can be used on samples with as low a concentration of bacteria as 1000 cells per millilitre. The smear microscopy test widely used the world over as an alternative to the sophisticated laboratory techniques is much less sensitive at 5000 to 10000 bacteria per millilitre, which means it can only identify between 60 and 70 percent of adult cases from cultured samples.

The new technique is, the team says, the first of its kind to use "a true GC-MS, metabolomics research approach" for the identification of the requisite lipid markers and can be fully automated. The approach has the potential to detect the metabolic markers repeatedly over time. "This study proves the capacity of a GC-MS, metabolomics pattern recognition approach for its use in TB diagnosis and disease characterisation," the researchers conclude.

"We are in the process of testing this method on a variety of other strains of M. avium, M. bovis, M. kansasii and M. tuberculosis, in addition to expanding its capacity to detect other infectious Mycobacterium species simultaneously," Loots explains. "Further validations using larger populations in a typically clinical scenario would probably follow, testing this methods sensitivity and specificity alongside the currently used diagnostic procedures, before various stake holders will be approached to investigate implementation on a larger scale," he adds.


Post a Comment