WHO TB catalogue case study

Mycobacterium tuberculosis and drug resistance

M. tuberculosis causes approximately 10 million infections in humans each year, causing the disease tuberculosis. Around 0.5M of these are due to drug-resistant strains. Only 13 drugs are approved for clinical use in tuberculosis so resistance to even one of these drastically reduces the treatment options.

This bacterium gains resistance to anti-tuberculosis drugs primarily through point mutations in the genes encoding proteins targeted by the drug. For instance, rifampicin resistance is almost wholly driven by mutations in the 81bp region of rpoB referred to as the rifampicin resistance determining region (RRDR). No horizontal gene exchange occurs in M. tuberculosis so the prediction of resistance relies on identifying the SNV mutations in target genes or associated promoters or gene regulation genes which can lead to resistance.

The WHO catalogue 1st edition

A global effort was made several years ago to collate all mutations thought to be associated with drug resistance in M. tuberculosis. This catalogue would then serve as a reference standard for the community to allow for comparable and statistically supported geno/pheno correlation and prediction. The first catalogue was published in 2021 and gathered data from over 41000 strains gathered from 45 countries where both the genome and the phenotypic resistance information was known for at least 1 of the 13 drugs.

A simple yet effective set of cascading descriptive and inferential statistical tests were used to determine if a certain mutation was associated with resistance. This is outlined in figure 1.

Click to expand

Figure 1. WHO TB catalogue, edition 1, grading of mutations that confer resistance to drugs. This flowchart shows how each mutation is assessed to be associated with resistance (passes all tests) or separated into other groups. Taken from figure 5 of the Catalogue of mutations in Mycobacterium tuberculosis complex and their association with drug resistance, edition 1

In brief, the steps are as follows, with the following steps only performed if the answer is ‘yes’ at that stage:

1. Is the mutation found in at least 5 strains (SOLO_SR>=5)?
2. Does the lower bound of the confidence interval of the positive predictive value of this mutation being associated with resistance >25% (PPV|SOLO_lb+=25%)?
3. Is the odd ratio (OR) of this mutations presence in phenotypically resistant strains compared to sensitive strains >1 (OR SOLO >1)?
4. Is the statistical significance of this OR difference <0.05 based on Fischers false discovery ratio (FDR) test, after correcting for FDR (OR SOLO_FE-sig TRUE)?

If the mutation in question is positive for all of these criteria, then it is strongly supported to be associated with resistance. Based on this flow, mutations were then categorised into 5 groups:

• Associated with resistance

   Mutations from high quality data (WHO dataset) that passed all tests
  

• Associated with resistance – interim

   Mutations from high and medium/high quality data that pass all tests
  

• Not associated with resistance

   Mutations found to be neutrally evolving
  

• Not associated with resistance – interim

   Mutations found to be neutrally evolving or listed as such by experts
  

• Uncertain significance

   No strong statistical support for association with resistance but 
   also no support for neutrality
  

This system worked very well for drugs where there were many strains available for testing such as rifampicin, izoniasid and fluouroquinalones. It did not do so well in this first edition for newer drugs where little to no resistant strains were available. This highlights the need for community input and collective contributions of clinical data from around the globe.

The WHO catalogue 2nd edition

In 2023 the WHO released a 2nd version of the catalogue which boosted the strain numbers to over 52000 isolates, many of these with resistance to the newer drugs to address the problems with the 1st edition. This community input to the large scale statistical analysis allowed for better sensitivity, specificity and range of mutations catalogued in this version. Future iterations and editions are planned for released every few years to further enhance this resource.

These efforts by the WHO and TB community stand as a strong example of what can be achieved when data contributions happen at a global scale to tackle a problem such as AMR. The hope is that similar efforts will be undertaken, and indeed have begun, for other pathogens, allowing the geno/pheno correlations to be well supported.