Optimizing the Design of Latent Tuberculosis Treatment Trials: Insights from Mathematical Modeling.

Published

Journal Article

Rationale: Noninferiority trials of treatment for latent tuberculosis infection (LTBI) are challenging because of imperfect LTBI diagnostic tests.Objectives: To assess the effect on study outcomes of different enrollment strategies for a noninferiority trial of LTBI treatment.Methods: We mathematically simulated a two-arm randomized clinical trial of LTBI in which the experimental therapy was 50% efficacious and the control was 80% efficacious, with an absolute 0.75% noninferiority margin. Five enrollment strategies were assessed: 1) enroll based on no LTBI diagnostic test; 2) enroll based on a positive tuberculin skin test (TST); 3) enroll based on a positive IFN-γ release assay (IGRA); 4) enroll based on either a positive TST or IGRA; and 5) enroll regardless of test result, assuming 70% had negative TSTs, 20% positive TSTs, and 10% unknown results.Measurements and Main Results: Under most LTBI prevalence assumptions, enrolling based on a positive IGRA was least likely to result in falsely declaring noninferiority of the experimental regimen. Enrolling based on no test or regardless of test result led to falsely declaring noninferiority unless LTBI prevalence in the underlying population was higher than 45%. Enrolling based on a mix of TST and IGRA substantially reduced the likelihood of falsely declaring noninferiority over enrolling based on TST alone, even if as many as 70% of participants were enrolled based on positive TST.Conclusions: Noninferiority trials of LTBI should enroll based on the most specific diagnostic tests available (i.e., IGRAs) to avoid misclassifying inferior treatment regimens as noninferior.

Full Text

Duke Authors

Cited Authors

  • Stout, JE; Turner, NA; Belknap, RW; Horsburgh, CR; Sterling, TR; Phillips, PPJ

Published Date

  • March 1, 2020

Published In

Volume / Issue

  • 201 / 5

Start / End Page

  • 598 - 605

PubMed ID

  • 31711306

Pubmed Central ID

  • 31711306

Electronic International Standard Serial Number (EISSN)

  • 1535-4970

Digital Object Identifier (DOI)

  • 10.1164/rccm.201908-1606OC

Language

  • eng

Conference Location

  • United States