Predictive test for tuberculosis

A predictive blood test for tuberculosis (TB) is "a step closer", according to BBC News. The article said that a “DNA fingerprint in the blood shows promise in identifying which carriers of TB will go on to get symptoms and spread the infection”.

This study is important and illustrates the power of a relatively new technique called “genomic transcriptional profiling”, but it is too early to know in practice how many of the TB patients identified by the test will go on to develop the disease.

After some fine-tuning in London, the test was repeated in patients from South Africa, which increases confidence in its accuracy. However, the study did not look at how many patients identified by the test later went on to develop the disease.

As tests perform differently in different populations, another step in checking its accuracy will involve assessing its performance in developing areas of the world where TB is more common. As the test requires complex expensive machines, this may be easier said than done.

Where did the story come from?

The study was carried out by researchers from the Medical Research Council, National Institute for Medical Research and St Mary’s Hospital in London along with American researchers. The study was funded by the MRC and The Dana Foundation, and no competing financial interests are declared. The study was published in the peer-reviewed journal Nature .

The media covered this complex study accurately and emphasised both the early nature of the research and its potential promise. The BBC quotes experts who say the test is "remarkable" but needs to be proven by further work.

What kind of research was this?

The researchers aimed to investigate biological markers that have the potential for diagnosing and predicting the outcome of latent TB. They explain that TB is mainly a disease of the lungs, which kills up to 1.7 million people a year worldwide. About one-third of the world's population has been exposed or infected with the TB bacteria (Mycobacterium tuberculosis), but only 10% of these people with latent TB become ill with the active form of the disease. It is thought that reduced immunity plays a part in whether or not a person becomes ill, but the exact reasons are poorly understood.

This was a laboratory study, using the technique of “genomic transcriptional profiling”. The study had three main parts:

• A training set of 42 blood samples from London were used to develop the test.
• A test set of 54 blood samples were used to identify differing pattern of the biomarkers in people with the active and latent forms of TB compared to healthy controls (from London).
• A validation set of 51 samples from South Africa were used to independently assess the accuracy of the test in identifying latent compared to active TB.

The researchers were interested, among other things, in how well they could tell active TB apart from other inflammatory diseases and how many patients with active disease were identified correctly by the test (its sensitivity).

Studies of the diagnostic accuracy of new tests need to be repeated many times in different settings. This is to work out the best cut-off points for diagnosing latent disease that will become active and the usefulness of the test in real-life settings. This study provides a useful starting point for this process.

What did the research involve?

Genome transcription profiling is a technique that measures the activity (the expression) of thousands of genes at once. Simply put, the technique gives an idea of what cells do. It is different from sequencing the actual genetic code of a cell, as instead of looking at the DNA on chromosomes, it creates a picture of what the cell is actually doing with this DNA (which genes are active and how active they are). This gene activity is assessed by how much RNA (or “transcripts”) cells produce. These RNA molecules carry instructions on how to make different proteins to the protein making machinery of the cell, or play other roles in the protein making process.

In the test set, the researchers compared the transcriptional profiles of expressed genes in the blood samples from three groups of patients. They had 21 samples from people with latent TB, 21 with active TB before treatment and 12 healthy controls.

In the validation set where the accuracy of the test set and its cut-offs were checked in a second set of samples, there were 31 latent TB samples, 20 active TB samples and no healthy samples.

By comparing the profiles from people in these test and validation sets, the researchers aimed to identify a gene transcript pattern that was similar in people with active TB and “high-risk” latent patients.

They tested the transcriptional pattern further in blood taken from people with other diseases such as bacterial infections, and an immune disease called lupus to see if they could identify a transcription signature specific for TB and not other diseases.

The analysis appears comprehensive and has been thoroughly reported.

What were the basic results?

The researchers identified a 393-transcript signature that was characteristic of active TB, and that returned to normal once a person had been successfully treated for TB.

They report that the transcriptional profiles of 10%–25% of patients with latent TB (five out of 21 from the test set and three out of 31 from the validation set) were similar to those of patients with active TB. This means that 75% to 90% of patients with latent TB did not have the characteristic “active” or high-risk profile they were looking for.

Using the 393-transcript signature in the test set of people, the sensitivity quoted was 61.67%, which means that 61.67% of people with active TB were correctly identified by the test. The test also had a specificity of 93.75%, so it correctly identified 93.75% of people who did not have active TB. It had an indeterminate rate of 1.9% for the test set, where the status (active, latent or healthy) could not be determined. Five patients with latent TB were classified as having active TB by the test and four patients with active TB were classified not having active TB by the test.

In the validation set, the sensitivity was 94.12%, specificity 96.67% and the indeterminate rate was 7.8%.

The researchers also identified an 86-gene transcript signature test that was able to discriminate active TB from other inflammatory and infectious diseases.

How did the researchers interpret the results?

The researchers say that their research has implications for vaccine and therapy development. They claim theirs is the first complete description of the human blood transcriptional signature of TB.

The signature of active TB, also observed in 10%–20% of patients with latent TB, may help to identify people who will develop active disease. They say this will make it easier to direct preventative therapy. However, they caution that further prospective studies conducted with patients followed over time are needed to assess this possibility.

Conclusion

This study uses a relatively new and complex genomic test to see if it is possible to identify people with active TB. The researchers also aimed to see whether the test could identify people who have latent TB and who are at risk of developing active TB in the future.

Diagnostic tests obviously need to be accurate and this is measured in several ways. How good the test is at identifying people with a disease (called sensitivity), and how good it is at identifying people who do not have the disease (called specificity) are two commonly used measures.

In this study:

The test had good results for sensitivity and specificity in the highly selected samples tested, suggesting that when disease status is already known the test (the pattern) is good at confirming that a person has active TB and identifying a pattern in those without the disease too. However, it is important to point out that in the test set, there was a sensitivity of only 62%, which means that 38% of samples with latent TB were identified as having active TB by the test (about six out of 16 in absolute terms).

All the people in the validation set were already known to have TB (active or latent) and so were “selected”. It is important to also measure the test’s accuracy in a population that has not been selected, a later stage of testing that will require following a set of people over time. This is because testing samples taken from people known to have latent TB or active TB before treatment will give better results than when the same test is used as a diagnostic tool in real-life populations with lower rates of active or latent TB.

Studies of a test’s accuracy in its ability to predict future disease also depend on how many people in the population tested have the condition. The researchers did not test a random sample of people with latent TB to see how well the test does in predicting who goes on to develop active disease. This will be a further step in the research. It is for this reason that the researchers wisely advise further testing of their new exciting technique.