Junjie Ding (1,2,3), Nguyen Thuy Thuong Thuong (4), Pham Van Toi (4), Dorothee Heemskerk (4), Thomas Pouplin (1,3), Tran Thi Hong Chau (5), Nguyen Thi Hoang Mai (5), Nguyen Hoan Phu (4,5), Phan Phu Loc (5), Nguyen Van Vinh Chau (5), Guy Thwaites (1,4), Joel Tarning (1,2,3)
(1) Centre for Tropical Medicine and Global Health, Nuffield Department of Clinical Medicine, University of Oxford, Oxford, United Kingdom; (2) The WorldWide Antimalarial Resistance Network, Oxford, United Kingdom; (3) Mahidol-Oxford Tropical Medicine Research Unit, Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand; (4) Oxford University Clinical Research Unit, Centre for Tropical Medicine, Ho Chi Minh City, Vietnam; (5) Hospital for Tropical Diseases, Ho Chi Minh City, Vietnam.
Objectives: The optimal anti-tuberculosis chemotherapy of tuberculous meningitis (TBM) is poorly defined (1). The current choice of drugs and their dosing is based upon the treatment of pulmonary tuberculosis, but the distribution of some of the drugs into the brain is restricted by the blood-brain barrier, which may result in sub-optimal drug exposures, bacterial killing, and treatment outcomes. Thus, it is hypothesised that bacterial killing and clinical outcomes from TBM can be improved by increasing the doses of some drugs, in particular rifampin, and by using drugs which cross the blood-brain barrier freely (2).
Methods: We conducted a randomised controlled trial (ISRCTN61649292) comparing the standard 4-drug anti-tuberculosis regimen (using 10 mg/kg/day rifampin) with a 5-drug ‘intensive’ regimen that included higher dose rifampin (15 mg/kg/day) and the addition of levofloxacin (1000 mg/kg/day) for the first 2 months of TBM treatment (3). The trial randomised 817 Vietnamese adults with TBM 1:1 to the two regimens, but did not demonstrate any benefit of the intensive regimen on survival or any other endpoint. Part of this was a nested pharmacokinetic/pharmacodynamic study in 237 trial participants to define exposure-response relationships. Both blood and CSF samples were collected at steady state for drug measurements using LC-MS/MS. Population PK/PD analyses were performed using nonlinear mixed-effects modelling in NONMEM. PD outcomes were modelled using a time-to-event approach. A classification and regression tree (CART) analysis, using a recursive partitioning algorithm, was also employed to verify the covariates identified by NONMEM and their optimal cut-off points. CART analyses were performed using the rpart package in R.
Results: Both plasma and CSF concentration-time data were available for rifampin, isoniazid, and levofloxacin and modelled simultaneously for each drug. Only plasma data were available for ethambutol and pyrazinamide. The pharmacokinetic properties of each drug were described successfully, incorporating allometric scaling by fat-free mas. In addition, the rifampin model comprised both enzyme auto-induction and dose-dependent absorption, and isoniazid elimination was best described by a mixture model for slow and fast acetylators. The best base hazard model for time-to-death was a log-normally distributed model. Inclusion of Glasgow coma score (GCS) improved the model fit (p <0.001). Inclusion of either HIV co-infection, or albumin, or AST improved model fit further (p <0.001). Considering the correlation between these 3 clinical covariates, and the importance of HIV co-infection to outcome, we retained HIV co-infection as a covariate in the model. Inclusion of individual rifampin exposures did not significantly affect the hazard (p>0.05; Emax model). In contrast, individual isoniazid exposures were found to significantly affect the hazard, with a higher exposure leading to a decreased probability of death. The estimated EC50 for CSF Cmax and AUC0-24 were 1.37 mg/L and 7.03 mg∙h/L, respectively. The significant pharmacodynamic covariates identified by population modelling, as well as the drug exposures were evaluated in the CART analysis. The final full CART tree included 5 terminal nodes; GCS, HIV co-infection, isoniazid CSF Cmax and rifampin CSF Cmax. However, higher rifampin exposures led to the prediction of worse outcomes and were therefore discarded as clinically implausible.
Conclusions: Our clinical trial failed to show that the addition of higher dose rifampin and levofloxacin to standard anti-tuberculosis treatment for the first 2 months of therapy has any impact on clinical outcomes. The current study showed that 15 mg/kg/day rifampin increased plasma exposures substantially, with AUC0-24 similar to those associated with improvements in survival in other studies. However, we were unable to find any significant relationship between increased rifampin exposure and survival in our cohort. In contrast, we found that isoniazid exposure was associated with survival, with low exposure predictive of death and linked to the fast metabolizer phenotype. Whilst phase III trials of high dose (>30 mg/kg) rifampin for TBM remain justified, consideration should also be given to exploring higher doses of isoniazid for the treatment of adults with TBM, especially fast acetylators.
References:
[1] Wilkinson, R.J. et al. Tuberculous meningitis. Nature reviews Neurology 13, 581-98 (2017).
[2] Cresswell, F.V. et al. Intensified antibiotic treatment of tuberculosis meningitis. Expert review of clinical pharmacology 12, 267-88 (2019).
[3] Heemskerk, A.D. et al. Intensified Antituberculosis Therapy in Adults with Tuberculous Meningitis. N Engl J Med 374, 124-34 (2016).
Reference: PAGE () Abstr 9423 [www.page-meeting.org/?abstract=9423]
Poster: Oral: Drug/Disease Modelling