121. Improving Predictive Value of Phase 2a TB Drug Development Models Through PET/CT Imaging (NexGen EBA)

Abstract

BackgroundCurrent Phase 2a methods in tuberculosis (TB) drug development are severely limited in capturing a drug candidate’s sterilizing activity (ability to prevent TB relapse). We hypothesize that sputum-based measurements of current methods fail to capture drug activity in deep TB lesion compartments, where recalcitrant TB populations are sequestered. In our recent marmoset study, we found complementary 14-day PET/CT delta signatures across different TB lesion compartments that distinguish drugs with long-term sterilizing activity (e.g., rifampin and pyrazinamide) versus drugs with primarily early bactericidal activity (e.g., isoniazid) (Figure 1). We proceeded to evaluate this novel PET/CT Phase 2a approach among TB patients, benchmarked by drugs that are well-characterized clinically and pharmacokinetically.MethodsHIV-negative patients with smear-positive pulmonary tuberculosis from Cape Town, South Africa were enrolled and randomized to receive a single drug [isoniazid (H), rifampin (R), pyrazinamide (Z), moxifloxacin (M)], a 2-drug regimen (HZ, RZ), or a 4-drug regimen (HRZE, MRZE) for 14 days in an inpatient facility. Primary data points were PET/CT scans evaluated at pretreatment baseline and treatment day 14 (and day 28 in the HRZE arm), and daily overnight sputum for early bactericidal activity measurements.ResultsFrom December 2015 to September 2017, 160 participants completed the study, from whom 340 PET/CT scans and 2,238 overnight sputum samples were collected. PET/CT scans from participants demonstrate diversity of lesion architecture and inter-individual variation in PET/CT changes over 14 days of treatment (Figure 2). A preliminary computational algorithm and first- and second-order PET/CT statistics have been developed for initial segmentation and TB lesion categorization of all study scans. Derivation of radiologic signatures by analysis of 14-day PET/CT changes across each lesion type is ongoing; blinded prediction of each participant’s treatment arm based on these signatures is expected by third-quarter 2018.ConclusionIf successful, these signatures may be applied within a 14-day Phase 2a framework to evaluate new TB drug candidates and construct TB drug regimens with complementary lesion pharmacokinetics. Disclosures All authors: No reported disclosures.