If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Australian Institute of Tropical Health and Medicine, James Cook University, 1 James Cook Drive, Townsville, QLD 4811, AustraliaDepartment of Medicine, Royal Melbourne Hospital, University of Melbourne, Melbourne, Victoria, Australia
School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, AustraliaVictorian Tuberculosis Program at the Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia
Multidrug-resistant tuberculosis is a significant global health problem.
Person-to-person transmission is a more important driver of drug resistance than acquisition.
Increased efforts to directly address multidrug-resistant tuberculosis are urgently needed.
Multidrug-resistant tuberculosis (MDR-TB) is a threat to tuberculosis (TB) control. To guide TB control, it is essential to understand the extent to which and the circumstances in which MDR-TB will replace drug-susceptible TB (DS-TB) as the dominant phenotype. The issue was examined by assessing evidence from genomics, pharmacokinetics, and epidemiology studies. This evidence was then synthesized into a mathematical model.
This model considers two TB strains, one with and one without an MDR phenotype. It was considered that intrinsic transmissibility may be different between the two strains, as may the control response including the detection, treatment failure, and default rates. The outcomes were explored in terms of the incidence of MDR-TB and time until MDR-TB surpasses DS-TB as the dominant strain.
Results and conclusions
The ability of MDR-TB to dominate DS-TB was highly sensitive to the relative transmissibility of the resistant strain; however, MDR-TB could dominate even when its transmissibility was modestly reduced (to between 50% and 100% as transmissible as the DS-TB strain). This model suggests that it may take decades or more for strain replacement to occur. It was also found that while the amplification of resistance is the early cause of MDR-TB, this will rapidly give way to person-to-person transmission.
This phenomenon necessitated the use of multidrug combination therapies, with strong health programmes − the directly observed treatment short-course (DOTS) strategy. DOTS, as it was originally implemented, focused on drug-susceptible (DS) TB. Its aim was to increase successful treatment outcomes and reduce drug resistance by ensuring all TB patients were treated with multiple agents for at least 6 months.
Despite these measures, since 1985, the world has seen a constant rise in the levels of multidrug-resistant (MDR) TB, defined as M. tuberculosis with in vitro resistance to at least isoniazid and rifampicin, the two most potent first-line anti-TB drugs. The most recent report suggests that these isolates accounted for at least 480 000 incident cases and 210 000 deaths worldwide in 2014.
Extensively drug-resistant (XDR) strains, which are MDR-TB strains with additional resistance to fluoroquinolones and a second-line injectable agent (kanamycin, amikacin, or capreomycin), have now been found in every region of the world.
Initially, public health agencies looked at the soundness of the DS-TB control programmes as a means to fix the problem: “While it is important, on a clinical basis and epidemiologically in some contexts, to care optimally for patients with MDR-TB, it is more important to address the cause of MDR-TB and to fix the programme generating MDR-TB”.
However, a new paradigm has gradually emerged in MDR-TB control, with greater acknowledgement that MDR-TB must be addressed directly. This paper examines why this shift in thinking is necessary to control MDR-TB, as well as the consequences of neglecting to address the problem of MDR-TB directly. Evidence from genomics, pharmacokinetics (PK), and epidemiology was reviewed. Mathematical modelling was then used to synthesize the evidence into scenarios in which MDR-TB and DS-TB vie for dominance.
In the last decades, major advances in molecular biology have increased our knowledge of the mechanisms of resistance to the main anti-TB drugs, with the identification of specific gene mutations that are associated with drug resistance.
This has also allowed the detailed mapping of M. tuberculosis transmission pathways, which has indicated typical epidemic spread of drug-resistant strains of M. tuberculosis in most settings where this has been evaluated.
Unlike other bacteria, in which acquired drug resistance is generally mediated through horizontal transfer of mobile genetic elements, M. tuberculosis acquires drug resistance through spontaneous chromosomal mutation, typically resulting in a fitness cost seen as a reduced growth rate in vitro.
Indeed, other processes, such as compensatory evolution and genetic co-selection, complicate the picture. As an example, the genetic background of each strain in which a specific resistance-conferring mutation occurs can modulate the fitness impact of this mutation, such interaction between genes being called epistasis.
Resistance to rifampicin is the most pressing concern in TB management, because it necessitates very long, expensive and relatively toxic drug schedules and leads to poorer outcomes. The identification of specific compensatory mutations among clinical strains of M. tuberculosis has improved our understanding of drug resistance and fitness.
Rifampicin resistance is nearly always caused by one of several possible point mutations to the rpoB gene, which encodes a small part of the β-subunit of RNA polymerase close to the catalytic centre of the enzyme.
The so-called rifampicin resistance-determining region (RRDR) covers 81 base pairs encoding amino acids 507–533 in the β-subunit. Compensatory mutations that ameliorate the fitness costs of the common rifampicin-resistance mutation rpoB R529C have been described in the rpoA, rpoB, and rpoC genes, coding for different subunits of RNA polymerase (α, β- and β′ subunits, respectively). Some clinical rifampicin-resistant M. tuberculosis isolates have mutations outside the detection regions (leading to false-negative rpo-β tests), while other isolates reveal no rpo-β mutation at all. Two efflux pumps (Rv2936 and Rv0783) over-expressed in the resistant isolates are postulated to cause the rifampicin resistance phenotype in these M. tuberculosis strains.
Drug resistance in M. tuberculosis is now recognized to result from complex drivers, rather than simply from weak programmes and inadequate adherence to therapy. Recent evidence suggests that the emergence of drug resistance can occur despite better than 98% treatment completion.
There is increasing evidence that variability in PK profiles between individuals (i.e., inter-individual variability) is a more likely cause of the emergence of drug resistance than non-completion of treatment.
exhibit marked inter-individual variability. Such variability in PK occurs as a result of demographic characteristics such as sex, age, ethnicity, and body weight, comorbidities, drug interactions, and genetic polymorphisms affecting drug absorption, metabolism, and elimination. PK variability in turn may lead to inadequate drug exposure at the site of infection, facilitating the emergence of drug resistance.
For example, Calver et al. found in their clinical study that low drug exposure (as measured by area under the concentration–time curve (AUC) and peak concentrations (Cmax)) was the main driver of drug resistance despite meticulous DOTS.
MDR-TB is now found in most countries around the world and the proportion of new TB cases showing multidrug resistance is increasing. Currently, the highest absolute numbers of MDR-TB cases occur in the most populous countries: India and China.
However, epidemiology studies now provide evidence to the contrary. For example, studies have shown that the transmission of drug-resistant strains (i.e., primary resistance) rather than amplification from susceptible strains (acquisition of resistance-conferring mutations, i.e., acquired resistance) is the dominant source of MDR-TB.
Drug-resistant M. tuberculosis strains have several survival advantages owing to the way in which TB is managed globally. Firstly, because only a fraction of new cases in the world are tested for resistance,
MDR-TB has the opportunity to spread in the community before detection. Universal use of GeneXpert for new cases is being advocated to improve the detection of resistant cases and hence inhibit their spread. Additionally in many countries MDR-TB treatment has a large backlog of patients, delaying specific therapy.
For this reason, it is possible to speculate (and this will be shown with mathematical modelling) that all else being equal, MDR-TB is likely to take over from DS-TB.
It is now possible to estimate the likely contribution of the various pathways to MDR-TB, i.e. (1) primary transmission of MDR-TB resulting in new or retreatment cases, or (2) the development of drug resistance in patients infected with DS-TB following inadequate drug exposure. Global MDR-TB rates and available modelling data suggest that the primary transmission of MDR-TB strains from person to person has become more frequent than acquisition following treatment.
If MDR-TB were not transmissible, its emergence would pose at worst a small increased cost for health systems. MDR-TB transmissibility, the relative likelihood of person-to-person transmission, is an important determinant of whether it will come to dominate over DS-TB. The early hypothesis that resistance is always associated with a loss of bacterial fitness, and hence leads to lower case fatality rates and decreased transmission of such strains, has been disproved.
Indeed, some isolates of MDR M. tuberculosis appear to have no reduction in fitness.
Observations and simple calculations show that most MDR-TB is transmitted rather than mutated from pre-existing DS strains. Figure 1 gives a simplified summary of the sources of MDR-TB found in new and pre-treated cases. Much is often made of the high ratio of MDR-TB isolates in retreated cases (global estimate >20%) compared with new cases (global estimate 3.3%) of TB.
However the ratio does not necessarily indicate high rates of amplification, as illustrated in Figure 1.
These arguments combined provide a framework to consider MDR-TB as an infection that can arise whenever first-line agents are used against DS-TB. Although the amplification of resistance is made worse by poor programmes, even well-functioning programmes should expect some MDR-TB as a result of natural variability in population PK. Once MDR-TB has emerged, its potential to displace DS-TB as the dominant phenotype depends on features other than those that drive the original amplification. These may depend on the transmission rates in the community and the detection and cure rates of the TB programme. The following section uses mathematical models to explore these ideas and elicit the key drivers of MDR-TB dominance.
Modelling drivers of MDR-TB burden
Despite the evidence presented above, there remains considerable debate as to the key drivers of the emerging drug-resistant TB epidemic. Mathematical modelling can illuminate this discussion by simulating the transmission dynamics in high-risk communities and identifying the primary factors that most strongly contribute to the incidence and prevalence of drug-resistant strains. In this section, the conditions under which one would expect MDR-TB to become dominant and the time taken to transition to the dominant type are examined.
For this purpose, the two-strain TB transmission model introduced by Trauer et al. was used.
The two-strain model allows for a modified transmissibility, and detection rate, of the MDR-TB strain relative to the transmissibility and detection rate of the DS-TB strain. Additionally, the model allows for different treatment failure rates according to the resistance profile. A diagram of the model, along with the ordinary differential equations and all parameter values used, is provided in the Appendix.
Baseline parameters for the model were chosen to simulate high prevalence conditions, in which TB is always endemic regardless of the values of the modifiable parameters. In doing this it was aimed to capture transmission dynamics seen in global hotspots;
however, the model was not specifically calibrated to a particular country. Hence the aim was to make general qualitative and semi-quantitative conclusions about model outcomes, rather than specific quantitative predictions.
An amplification pathway is included in the model structure, representing acquired resistance, so that as long as DS-TB exists and is treated, it will continue to supply MDR-TB patients. The resultant equilibrium scenarios are one of two possibilities: either both DS-TB and MDR-TB remain in circulation (this represents DS being the dominant driver of TB transmission), or MDR-TB outcompetes DS-TB, driving the latter to extinction.
The results, displayed in Figure 2, show the incidence of MDR-TB and time to replacement of DS-TB with MDR-TB as a function of the relative transmissibility and relative detection rate of MDR-TB compared with DS-TB (time of replacement is defined as the point at which the incidence of MDR-TB exceeds the incidence of DS-TB). Predictably, the higher the transmissibility of the isolate (moving along the x-axis from zero to twice as fit as DS-TB), the higher the incidence of MDR-TB at equilibrium and the faster its progress to replace DS-TB strains as the dominant strain population. In this model, the MDR-TB burden depends more sensitively on the relative strain transmissibility than it does on the relative detection rate. Figure 2 indicates a clear delineation, such that if the relative transmissibility of MDR-TB is more than approximately 80% of DS-TB, MDR-TB always comes to dominate, regardless of the relative detection rate. Conversely, if the relative transmissibility of MDR-TB is less than approximately 50%, DS-TB remains dominant regardless of the relative detection rate.
The relative detection rate does influence the incidence and dominance of MDR-TB in the interval in which MDR-TB transmits 50% to 80% as efficiently as DS-TB. A detection rate of MDR-TB equal to that of DS-TB (as may be expected when GeneXpert is the first-line diagnostic), would lead to very low levels of MDR-TB, whereas no MDR-TB detection would lead to high incidence levels (Figure 2).
The equilibrium MDR-TB incidence and replacement time, respectively, as functions of the unsuccessful live outcome rate (i.e., default or failure) and the relative detection rates are shown in Figure 3. In these graphs the impact of changes in transmissibility (relative transmissibility is set at a fixed value of 70%) is not considered. The upper left-hand corner of each panel in Figure 3–corresponding to the best case control scenario − finds DS-TB dominates; however, as we move towards the lower right-hand corner, MDR-TB begins to dominate. Surprisingly, in this case, the MDR-TB burden is more sensitive to the relative detection rate than it is to the treatment outcome. However, as the relative detection rate increases, the treatment outcome becomes more important. This calculation was repeated for a relative transmission fitness of 100% and it was found that MDR-TB dominates under all conditions, except zero treatment.
Laboratory findings suggest that compensatory mutations occur in some isolates of MDR M. tuberculosis that potentially allow the growth and transmissibility fitness cost to be minimized or completely overcome.
Furthermore, modelling suggests that strain replacement can occur even if the basic reproduction number of the MDR-TB strain is less than that of the DS strain; that is, even when there is a fitness cost to drug resistance.
Therefore, the growing risk of the MDR-TB epidemic cannot be dismissed by assuming that the relative fitness of the mutant strains is diminished, which in itself may not be a valid assumption.
The shortest time period for MDR-TB to overcome the less resistant co-circulating strain is of the order of decades. Given this, it would be possible to mistake the slow emergence of the resistant strain for an absence of significant burden at equilibrium. This may help to explain the historical assumption that MDR-TB was a temporary phenomenon resulting directly from non-completion of therapy in individual patients, rather than a persistent challenge to global control. Moreover, as the replacement time is of the order of centuries under most proposed conditions (Figure 2, Figure 3), this could also explain the observation that some countries have very low rates of MDR-TB while others have extremely high rates, at 50 years following introduction of the antibiotics.
In Figure 4, the model output of the proportion of MDR-TB that arises through transmission from person to person as a fraction of all incident MDR-TB is measured. It can be seen that, initially, all MDR-TB arises through amplification, as would be expected, given that starting conditions are that there is no MDR-TB. After 10 years, the proportion of MDR-TB is nearing its equilibrium state and depends on the relative transmissibility of the MDR-TB.
In this model, the amplification of MDR-TB from DS-TB serves as a trigger, bringing MDR into existence, while fitness cost and the favourable conditions for replacement inadvertently brought about by control programmes determine whether MDR will dominate. Factors that drive MDR-TB replacement include low levels of detection of MDR compared with DS-TB and higher default/failure rates of MDR-TB than DS-TB. However, relative detection rates are much more influential than relative default/failure rates.
Earlier models, such as those of Dye and Espinal, had a similar structure but did not explore as broad a parameter space.
In fact, they did not allow that MDR-TB may be harder to detect than DS-TB, or be more likely to fail therapy, effectively exploring only the top left corner of the heat maps in Figure 2, Figure 3. As can be seen, this is not a very interesting part of the parameter space and one in which the MDR-TB transmission environment is unfavourable.
Further modelling work defining the conditions under which strain replacement may occur will be very useful in determining the risk of additional or accumulated resistance, for example the emergence of XDR-TB and resistance to newer agents such as bedaquiline and delamanid.
Modelling demonstrates that TB with a resistant phenotype may thrive even in the presence of some transmissibility fitness cost − a concerning possibility that needs to be appreciated by public health policymakers. Current surveillance and reporting systems are inadequate to estimate the extent of MDR-TB,
which is a critical methodological flaw that needs to be addressed urgently. Without a renewed focus on the prevention, early diagnosis, and effective treatment of MDR-TB cases, we are likely to witness MDR-TB epidemic replacement in the coming decades, which could derail progress towards global TB control and ultimate elimination.
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.