Hollow-fibre system model of tuberculosis reproducibility and performance specifications for best practice in drug and combination therapy development

Abstract

Background: The hollow-fibre system model of tuberculosis (HFS-TB) has been endorsed by regulators; however, application of HFS-TB requires a thorough understanding of intra- and inter-team variability, statistical power and quality controls.

Methods: Three teams evaluated regimens matching those in the Rapid Evaluation of Moxifloxacin in Tuberculosis (REMoxTB) study, plus two high-dose rifampicin/pyrazinamide/moxifloxacin regimens, administered daily for up to 28 or 56 days against Mycobacterium tuberculosis (Mtb) under log-phase growth, intracellular growth or semidormant growth under acidic conditions. Target inoculum and pharmacokinetic parameters were pre-specified, and the accuracy and bias at achieving these calculated using percent coefficient of variation (%CV) at each sampling point and two-way analysis of variance (ANOVA).

Results: A total of 10 530 individual drug concentrations, and 1026 individual cfu counts were measured. The accuracy in achieving intended inoculum was >98%, and >88% for pharmacokinetic exposures. The 95% CI for the bias crossed zero in all cases. ANOVA revealed that the team effect accounted for <1% of variation in log10 cfu/mL at each timepoint. The %CV in kill slopes for each regimen and different Mtb metabolic populations was 5.10% (95% CI: 3.36%-6.85%). All REMoxTB arms exhibited nearly identical kill slopes whereas high dose regimens were 33% faster. Sample size analysis revealed that at least three replicate HFS-TB units are needed to identify >20% difference in slope, with a power of >99%.

Conclusions: HFS-TB is a highly tractable tool for choosing combination regimens with little variability between teams, and between replicates.

Figure 1.

Mean drug concentration–time profiles across all replicates, teams, regimens and growth conditions. Error bars are standard deviations. (a) Ethambutol. (b) High and standard dose rifampin. (c) Isoniazid. (d) High and standard dose moxifloxacin. All replicates for all experiments are included, e.g. for isoniazid, 81 systems are shown at each timepoint. In many instances, the error bars are smaller than the symbol size, demonstrating the very low variance. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 2.

Variation in log-phase growth bacteria in HFS-TB replicates within and across teams. Shown are the log₁₀ cfu/mL counts for each replicate by sampling day with colour coding by team. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 3.

Variation in intracellular bacteria HFS-TB replicates within and across teams. Shown are the log₁₀ cfu/mL counts for each replicate by sampling day with colour coding by team. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 4.

Variation in semidormant bacteria HFS-TB replicates within and across teams. Shown are the log₁₀ cfu/mL counts for each replicate by sampling day with colour coding by team. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 5.

Comparison of slopes between regimens and teams, showing mean slope (log₁₀ cfu/mL/day) and standard deviations. ***P value <0.0001. (a) Log-phase growth bacilli revealed very little variation of replicates within each team (standard deviation) and between teams (different colours). There was no statistically significant different slope between the regimens. The non-treated controls slope is not shown since these grew (and thus slope is in the opposite direction) at 0.07 ± 0.01 log₁₀ cfu/ml/day. (b) Slopes for intracellular bacteria also showed little variability between replicates within each team, and little variation between teams. The slopes for the 3 days of isoniazid followed by high-dose rifampicin, pyrazinamide and moxifloxacin, and the high-dose rifampicin, pyrazinamide and moxifloxacin were significantly higher than standard therapy and experimental REMox-TB arms. (c) For sterilizing effect, the REMox-TB arms were similar to standard therapy, while both high-dose regimens were higher. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.