Center for Infectious Diseases Research and Experimental Therapeutics, Baylor Research Institute, Dallas, Texas.
Division of Pharmacology, Department of Medicine, University of Cape Town, Observatory, South Africa.
Clin Infect Dis. 2018 Nov 28;67(suppl_3):S349-S358. doi: 10.1093/cid/ciy623.
A major challenge in medicine is translation of preclinical model findings to humans, especially therapy duration. One major example is recent shorter-duration therapy regimen failures in tuberculosis.
We used set theory mapping to develop a computational/modeling framework to map the time it takes to extinguish the Mycobacterium tuberculosis population on chemotherapy from multiple hollow fiber system model of tuberculosis (HFS-TB) experiments to that observed in patients. The predictive accuracy of the derived translation transformations was then tested using data from 108 HFS-TB Rapid Evaluation of Moxifloxacin in Tuberculosis (REMoxTB) units, including 756 colony-forming units (CFU)/mL. Derived transformations, and Latin hypercube sampling-guided simulations were used to predict cure and relapse after 4 and 6 months of therapy. Outcomes were compared to observations, in 1932 patients in the REMoxTB clinical trial.
HFS-TB serial bacillary burden and serial sputum data in the derivation dataset formed a structure-preserving map. Bactericidal effect was mapped with a single step transformation, while the sterilizing effect was mapped with a 3-step transformation function. Using the HFS-TB REMoxTB data, we accurately predicted the proportion of patients cured in the 4-month REMoxTB clinical trial. Model-predicted vs clinical trial observations were (i) the ethambutol arm (77.0% [95% confidence interval {CI}, 74.4%-79.6%] vs 77.7% [95% CI, 74.3%-80.9%]) and (ii) the isoniazid arm (76.4% [95% CI, 73.9%-79.0%] vs 79.5% [95% CI, 76.1%-82.5%]).
We developed a method to translate duration of therapy outcomes from preclinical models to tuberculosis patients.
医学面临的一个主要挑战是将临床前模型的研究结果转化为人类，尤其是治疗持续时间。最近结核病的短疗程治疗方案失败就是一个主要的例子。
我们使用集合理论映射来开发一个计算/建模框架，将从多个中空纤维系统结核模型（HFS-TB）实验中消除结核分枝杆菌群体所需的时间映射到患者中观察到的时间。然后使用来自 108 个 HFS-TB 快速评估莫西沙星治疗结核病（REMoxTB）单元的数据，包括 756 个菌落形成单位（CFU）/mL，来测试所得转换的预测准确性。从 REMoxTB 临床试验中，对 1932 名患者进行了衍生转换和拉丁超立方抽样引导模拟，以预测 4 个月和 6 个月治疗后的治愈和复发情况。结果与 1932 名 REMoxTB 临床试验患者的观察结果进行了比较。
在推导数据集的 HFS-TB 系列细菌负荷和系列痰数据形成了一个结构保持的图谱。杀菌效果用单个步骤的转换来映射，而灭菌效果用 3 步转换函数来映射。使用 HFS-TB REMoxTB 数据，我们准确地预测了 4 个月 REMoxTB 临床试验中治愈患者的比例。模型预测与临床试验观察结果一致：（i）乙胺丁醇组（77.0%[95%置信区间{CI}，74.4%-79.6%]比 77.7%[95% CI，74.3%-80.9%]）和（ii）异烟肼组（76.4%[95% CI，73.9%-79.0%]比 79.5%[95% CI，76.1%-82.5%]）。
我们开发了一种从临床前模型将治疗持续时间结果转化为结核病患者的方法。
