In many settings, smear microscopy, chest x-ray and bacterial culture remain the mainstay of TB diagnosis. While recent advances in molecular diagnostics are changing the landscape for the diagnosis of TB, the accuracy of such tests in many subpopulations such as children is still poor. The response to emergence of drug resistance in an individual, or indeed at a programmatic level, requires robust information about the drug resistance profile of the Mtb strain involved. Obtaining this information by conventional microbiology is too slow to be of use, but the advent of molecular tools has facilitated a rapid, though imperfect picture of resistance. The importance of an indication of drug sensitivity is highlighted by the paper by Martin and colleagues, which demonstrates that the selective use of molecular tools can lead to a significant under identification of drug resistance and the concomitant failure to treat adequately [10]. The authors call for investment in services to allow the microbiological diagnosis to be completed. Of course tools such as the Xpert MTB/RIF, have a lot to offer in this respect; but this provides evidence of rifampicin resistance and presumptive evidence of multi-drug resistance.
The increased accessibility of WGS has begun to change the diagnostic landscape, at least in high income settings. Thus, Witney and colleagues describe how WGS methods are being adapted for diagnostic use, including rapid identification of single nucleotide polymorphisms (SNPs) associated with known drug resistance [6]. This approach is already in use to inform the management of patients with complex drug resistance patterns in the absence of timely phenotypic tests. However, the authors rightly point to the areas that require development not least the need for a robust and user friendly bioinformatics pipeline. A key piece of information that is required to underpin this is the ability to correlate drug resistance with a SNP. It is not always possible to identify the necessary phenotype for sequence analysis and so Phelan and colleagues present a modelling approach to map potential resistance genotypes [11]. Using protein structure prediction they are able to predict polymorphisms associated with resistance. They propose that this approach can be used to inform the development of novel diagnostics with improved predictive value for phenotypic drug resistance.
Thus, we see that appropriate diagnosis of drug resistance, like TB diagnosis generally, is an area in development. The power of molecular tools needs to be applied carefully, but the emerging use of WGS and the associated analysis offers a real opportunity to provide rapid and robust drug sensitivity data in a timely fashion.
Treatment for drug sensitive TB remains stubbornly at six months despite efforts to improve upon this. Microbiological indicators of treatment response that can discriminate between different regimens and also predict long term clinical outcome at an individual patient level would be of great value. The REMoxTB trial, which compared two 4 month regimens containing moxifloxacin to standard 6 month treatment and where cultures were obtained weekly for 8 weeks and monthly thereafter, provides an opportunity to assess the performance of a variety of different measures of sputum culture conversion. Phillips and colleagues show that whilst there is an advantage in using continuous endpoints (which capture time on treatment) better than measuring the proportion of negative cultures at a given time, none were discriminatory enough to be used as either a trial surrogate endpoint or as an indicator of individual patient outcome [12].
Phillips and colleagues showed that patient factors were often most important in predicting outcome. The study by Ranzani and colleagues, based in Sao Paulo State, Brazil, highlights this. Here, successful treatment outcome was directly related to an individual’s social circumstances – with less than half of the homeless population studied having a favourable outcome (43 % compared to 83 % in the comparator group) [13]. This was mainly due to large numbers being lost to follow up or dying. Although the number of homeless subjects was relatively small at 3 %, they were a population with high rates of HIV co-infection, and hence had considerable health needs beyond TB. As is frequently the case, this study emphasises the importance of an engaged and co-ordinated societal response if we are to control TB. In this collection of articles, Phillips and colleagues describe an innovative approach, the “STEP trial design”, which may accelerate regimen development by ensuring a move towards simpler and shorter later phase evaluation of drug regimen [14].