Modelling the effect of short-course multidrug-resistant tuberculosis treatment in Karakalpakstan, Uzbekistan

Abstract

Background: Multidrug-resistant tuberculosis (MDR-TB) is a major threat to global TB control. MDR-TB treatment regimens typically have a high pill burden, last 20 months or more and often lead to unsatisfactory outcomes. A 9-11 month regimen with seven antibiotics has shown high success rates among selected MDR-TB patients in different settings and is conditionally recommended by the World Health Organization.

Methods: We construct a transmission-dynamic model of TB to estimate the likely impact of a shorter MDR-TB regimen when applied in a low HIV prevalence region of Uzbekistan (Karakalpakstan) with high rates of drug resistance, good access to diagnostics and a well-established community-based MDR-TB treatment programme providing treatment to around 400 patients. The model incorporates acquisition of additional drug resistance and incorrect regimen assignment. It is calibrated to local epidemiology and used to compare the impact of shorter treatment against four alternative programmatic interventions.

Results: Based on empirical outcomes among MDR-TB patients and assuming no improvement in treatment success rates, the shorter regimen reduced MDR-TB incidence from 15.2 to 9.7 cases per 100,000 population per year and MDR-TB mortality from 3.0 to 1.7 deaths per 100,000 per year, achieving comparable or greater gains than the alternative interventions. No significant increase in the burden of higher levels of resistance was predicted. Effects are probably conservative given that the regimen is likely to improve success rates.

Conclusions: In addition to benefits to individual patients, we find that shorter MDR-TB treatment regimens also have the potential to reduce transmission of resistant strains. These findings are in the epidemiological setting of treatment availability being an important bottleneck due to high numbers of patients being eligible for treatment, and may differ in other contexts. The high proportion of MDR-TB with additional antibiotic resistance simulated was not exacerbated by programmatic responses and greater gains may be possible in contexts where the regimen is more widely applicable.

Keywords: Epidemiology; Extensively drug-resistant tuberculosis; Modelling; Multidrug-resistant tuberculosis; Public health; Treatment; Tuberculosis; Uzbekistan.

Fig. 1

Model structure. Spontaneous recovery for patients in the detected compartments and all death flows are not depicted. Brown arrows represent case detection flows, the total of which are set equal for all strains. Hollow arrows represent treatment commencement flows, which are determined by the total number of persons awaiting treatment with that regimen and the availability of the regimen for each of the three regimens. Individual compartment names are explained in Additional file 1: Table S1 and summarised as follows: blue text and _s subscript, drug-susceptible TB; red text and _m subscript, multidrug-resistant TB; green text and _x subscript, XDR-TB (including also MDR-TB strains with resistance to fluoroquinolones or second-line injectable agents). S, susceptible to TB (_A and _B subscripts refer to fully susceptible and partially immune, respectively); L, latent infection (_A and _B subscripts refer to early and late latent infection, respectively); I, active TB disease in the community not yet detected; D, detected (first subscript refers to the actual resistance pattern of the infecting strain, second subscript refers to the strain thought to be present at diagnosis); T, on treatment (subscripts are as for D compartments for those incorrectly diagnosed, while for those correctly diagnosed _I subscript indicates still infectious on appropriate regimen, while _N subscript indicates no longer infectious). For simplicity, the model assumes no I_s patients are incorrectly detected as drug-resistant

Fig. 2

Implementation of main intervention and comparators. Model of the implementation of short-course MDR-TB and of the four comparator programmatic interventions. Increased flows highlighted by thick purple arrows, with indirect effects indicated through dashed purple arrows. For Scenario 4, the flows that are decreased are illustrated with thin purple arrows. Reinfection omitted

Fig. 3

Scenario outcomes. Strains are presented by columns of panels and disease burden outcomes are presented by rows. Legend for all plots is presented in the lower left panel