Time-dependent pharmacokinetics of hydromethylthionine mesylate (HMTM), a tau aggregation inhibitor, in healthy volunteers and patients with Alzheimer's Disease
Eline van Maanen (1), Antonio Goncalves (1), Lu Chen (1), Floris Fauchet (1), Sinziana Cristea (1), James Smith (1), Tom Baddeley (2,3), Fiona Hewitt (2), Bjoern Schelter (2,3), Claude Wischik (2,3)
(1) Certara Inc., Princeton, New Jersey, USA (2) TauRx Therapeutics Ltd, Aberdeen, UK (3) University of Aberdeen, Aberdeen, UK
Objectives:
Hydromethylthionine mesylate (HMTM) is an orally administered tau aggregation inhibitor (TAI), that also disaggregates pathological tau oligomers and filaments. Thus it has potential for treatment of tauopathies, including Alzheimer’s disease (AD). This work used population PK (PopPK) modeling to quantify the PK of parent hydromethylthionine (HMT) and investigate sources of its variability in healthy volunteers and AD patients (mild to moderate AD and mild cognitive impairment (MCI)-AD).
Methods:
Data from five Phase 1 (TRx-237-021, -023, -024, -035, -036) and one Phase 3 study (TRx-237-039) were used for PopPK modeling. The model described the active HMT moiety, whether delivered as HMTM or as the control methylthioninium chloride (MTC; 4 mg biweekly). Single doses of HMTM ranged between 4 and 100 mg and multiple dose regimens ranged between 8 and 80 mg/day. The maximum duration was 104 weeks (TRx-237-039).
A total of 710 participants with 6440 plasma parent HMT concentrations, including 3342 concentrations from 557 AD patients, were available. Concentrations below the limit of quantification (BLQ) were excluded (<20%). The inter-individual random effects on the parameters were included assuming a log-normal distribution. Parent HMT concentrations were log-transformed and an additive residual error was included within the log domain.
Dose nonlinearity and time-varying PK were investigated. Covariates of interest included demographics, food status, smoking history, erythrocyte count and markers of renal and liver functions. Continuous and categorical covariates were incorporated into the model using a power function and proportional function, respectively.
The model fit was assessed using the objective function value (OFV) and (stratified) goodness-of-plots. The covariate analysis followed forward inclusions and backward deletion steps, using an OFV difference of >6.64 (P <0.01 χ2 distribution) and >10.83 (P <0.001) for statistical significance, respectively. The predictive performance of the final model was assessed by stratified prediction correction visual predictive checks (pcVPCs) based on 500 replicates of the analysis data.
The first-order conditional estimation method with interaction of NONMEM (version 7.5.1) was used during model development. R 4.2.0 was used for (graphical) analysis, diagnostics, and statistical summaries.
Results:
Observed dose normalized non-BLQ trough concentrations indicated a time-varying PK, with accumulation at week 52 that was reversed at week 104. A linear two-compartmental disposition model, with delayed first-order absorption using two transit compartments into the central compartment, and a time-varying (parabolic) elimination best described all parent HMT data. The time-varying elimination was described by a parabolic function on CL
CL (t) = CL0 * (1 – MaxDrop * t / Tmax * (2 – t / Tmax))
with CL0 the initial clearance, MaxDrop the maximal reduction of the time-varying parabolic clearance, Tmax the time at which CL reached the minimum.
The typical CL decreased from 1660 L/h at time zero and reached the minimum of 782 L/h (47.1% of CL0) on week 54, and then rebounded to 1550 L/h on week 104 (93.0% of CL0). Simulations indicated there is ~ 3-fold more accumulation at week 52 as would be expected based on linear PK.
Body weight on central volume of distribution, prandial status on absorption rate constant, creatinine clearance (CrCL), smoking status and sex on CL0 were retained in the final model as statistically significant covariates. However, univariate simulations indicated that only CrCL and smoking status had a clinically relevant impact (fold change beyond 0.8-1.25) on parent HMT exposure. A subject with a creatine clearance of 41.8, showed a 38% increased exposure compared to a subject with a creatine clearance of 72.4 mL/min. In a subject currently smoking exposure decreased by 23% compared to a non-smoking subject.
The pcVPC showed that the final PopPK model adequately predicted the healthy volunteer studies TRx-237-021, -023, -024, -035, -036. For the patient study TRx-237-039, the pcVPC showed that the median trend of all treated HMTM groups is adequately captured but may be underestimated at Visit 3 and Visit 7 in the control group MTC.
Conclusions:
A descriptive, robust popPK model for parent HMT was developed. Important covariates influencing pharmacokinetics and a time-varying clearance were identified which might guide the further development of HMTM and optimize its long-term use in the treatment of AD.
References:
[1] Text for reference 1.
[2] Text for reference 2, etc etc