Mathematical modeling of the cause of tuberculosis during tumor necrosis factor blockade

Abstract

Objective: Tumor necrosis factor (TNF) blockade increases the risk of tuberculosis (TB). The purpose of this study was to use Markov modeling to examine the contributions of reactivation of latent tuberculous infection (LTBI) and the progression of new infection with Mycobacterium tuberculosis to active TB due to TNF blockade. These 2 pathogenic mechanisms cannot otherwise be readily distinguished.

Methods: Monte Carlo simulation was used to represent the range of reported values for the incidence of TB associated with infliximab (TNF monoclonal antibody) and etanercept (soluble TNF receptor) therapy. Iterative methods were then used to identify for each pair of incidence rates the Markov model parameters that most accurately represented the distribution of time to onset of TB as reported to the Food and Drug Administration.

Results: Modeling revealed an apparent median monthly rate of reactivation of LTBI by infliximab treatment of 20.8%, which was 12.1 times that with etanercept treatment (P<0.001). In contrast, both drugs appeared to pose a high risk of progression of new M tuberculosis infection to active TB. Progression of new infection appeared to cause nearly half of the etanercept-associated cases; it became the predominant cause of infliximab-associated cases only after the first year.

Conclusion: Despite sharing a common therapeutic target, infliximab and etanercept differ markedly in the rates at which they reactivate LTBI. Confirmation of these findings will require the application of molecular epidemiologic tools to studies of TB in future biologics registries. Hidden Markov modeling and Monte Carlo simulation are powerful tools for revealing otherwise hidden aspects of the pathogenesis of TB.