Repurposing metformin to prevent and treat tuberculosis

Current antimicrobial therapy against tuberculosis (TB) requires a long and complex treatment regimen, which is related to the ability of the causative pathogen, Mycobacterium tuberculosis, to persist in a semi-dormant state despite prolonged antimicrobial exposure.1 Relapse rates remain conceivable after apparently successful treatment with 6-month short-course regimens and attempts to shorten the treatment with introduction of fluoroquinolones have so far been futile.2 Poor treatment adherence and suboptimal programmatic performance are breeding drug resistance, threatening TB control worldwide.3 Host defence appears to be a crucial factor in containing the activity of the invading pathogen as only a small fraction of infected individuals will ever develop TB in their lifetime, and the TB risk is strongly influenced by many host factors, such as HIV infection, silicosis, nutritional status and diabetes mellitus (DM), among others.4

In a 1:1 propensity score-matched analysis using data from a large longitudinal health insurance database in Taiwan, Lin et al. showed 76% reduction in TB risk among 5026 diabetic patients put on metformin as compared with 5026 non-metformin users, after adjusting for co-morbidities, diabetic complications, anti-diabetic therapy type and statin use.5 Being one of the largest longitudinal studies published to date, the study also found progressively larger protective effect with higher cumulative daily dose of metformin. Consistent association was observed between metformin use and active TB risk in most subgroup analyses but the protective effect diminished with increasing age. Direct data on diabetic control, a key confounder in relation to TB risk,6 were not available. However, in another longitudinal study from the same locality,7 metformin use was associated with 44% reduction in mortality during TB treatment, even though metformin users had a higher haemoglobin A1c level than non-users, thus suggesting a protective effect independent of diabetic control.

The sizeable protective effects of metformin in the two studies mentioned above suggest a potential role of metformin as host-directed therapy in the treatment of both latent TB infection5 and active TB.7 However, confounding by indication is a common pitfall in observation studies. As expected, substantial differences were seen in the baseline characteristics between metformin users and non-users in either study. Despite rigorous attempts to control potential confounders by propensity matching and/or adjustment in multivariate regression, residual confounding is not easy to exclude, especially in view of the considerable inter-correlation among the large number of predicting variables and only modest numbers of disease endpoints. Although most of the available observational studies did show some protective effects of metformin on TB,5, 7-9 there were also notable discrepancies in their observations. For example, while a case–control study in an Indian tertiary hospital8 showed a 3.9-fold protective effect of metformin against TB, no difference in protective effect was found between the 1000- and 500-mg doses, in sharp contrast with the dose-dependent effect found in the study by Lin et al.5 In another retrospective cohort study undertaken in Korea on patients with culture-positive pulmonary TB, metformin use had no significant effect on sputum culture conversion at 2 months and recurrence within 1 year of treatment completion in the overall analysis, even though there was improved sputum culture conversion rate in a subgroup analysis of patients with cavitary pulmonary TB.9 Randomized trials are therefore indispensable in delineating of the exact role(s) of metformin before its introduction into clinical practices.

Metformin is commonly used as first-line therapy in DM. Its well-established pharmacokinetic profiles and relatively good safety records are likely to expedite its entry into clinical trials as an adjunctive host-directed therapeutic agent against TB.10 Although the drug has also been investigated in pre-diabetic and other non-diabetic conditions, all available observation studies supporting its protective effect against TB were restricted to patients with DM.5, 7-9 It therefore remains to be determined whether such protective effect, if confirmed, is generalizable to non-diabetic subjects. Furthermore, preclinical studies on the adjunctive role of metformin in the treatment of TB in two different non-diabetic mouse models showed rather disparate results.11, 12 In the study by Singhal et al.,11 metformin ameliorated lung pathology, reduced chronic inflammation and enhanced the specific immune response and the efficacy of isoniazid or ethionamide monotherapy in C57BL/6 mice, while in the study by Dutta et al., metformin did not reduce lung bacillary burdens or the relapse rate of chronic TB in BALB/c mice when given in combination with the standard short-course regimen of isoniazid, rifampicin and pyrazinamide.12 Further animal studies will be needed to resolve this critical discrepancy. As a proof-of-concept study on host-directed therapy, it may also be particularly relevant to examine the effects of metformin, if any, in animal models (e.g. C3HeB/FeJ Mice) showing histological lesions with necrotic granuloma formation resembling those of human TB.13

For any therapy directed towards the host, potential influences of the baseline host status on treatment efficacy cannot be easily excluded. Careful stratification of patients by diabetic and other host factors would therefore be desirable in the design of future clinical trials to avoid incidental imbalance in any of the key potentially confounding variables during the process of randomization. Although long-term administration of metformin to prevent TB is unlikely to be an attractive option in non-diabetic individuals, the adjunctive role of short-term metformin use in either increasing the efficacy or shortening the treatment duration of conventional treatment regimens for latent TB infection may worth further exploration, especially when suitable surrogate markers for future TB development are found to allow substantial reduction of the required sample sizes and/or follow-up periods for such clinical trials.14