By: Nastya Kassir
In my previous blog post, I introduced how modeling and simulation (M&S) is transforming how drugs are developed, and how software uses existing data to improve efficiency and productivity for drug development at sponsor companies.
Clarus Therapeutics, a small pharmaceutical company that specializes in men’s health, was developing an oral testosterone (T) replacement product for use in men with T deficiency. Clarus needed to optimize their Phase 3 study design to meet FDA requirements for efficacy and safety. Learn how Certara scientists used M&S to pick the Phase 3 trial design that would help Clarus achieve FDA approval for their product.
Use of Population PK Modeling to Build a Solid Model
Certara used data from prior Phase 2/3 studies in Phoenix NLME to develop a population PK model to determine the covariates affecting T concentrations. The model was best described with a 1-compartment PK model with first-order absorption and lag time. Also included in the model was the allometric function on CL/F (clearance) and Vc/F (volume of distribution) that was driven by the body weight of patients. Figure 1, a diagram of the model shows between subject variability, or how concentrations vary from one individual to another and inter-occasion variability, how concentrations vary from one visit to another in the same individual.
To validate the population PK model, Certara used the visual predictive check, where the performance was evaluated and the 95% confidence intervals (CIs) of the 5th, 50th, and 95th prediction percentiles overlaid with the corresponding percentiles of the observed data. The higher concentrations (95th) and the average concentrations (50th) were robustly predicted, suggesting adequate predictions of maximum concentration (Cmax) related to safety and adequate predictions of Cavg related to efficacy, respectively. The lower concentrations (5th) were associated with a higher uncertainty, which was not a big issue in this case because we were more interested in efficacy and safety. This provided enough evidence to use the population PK model for the next step of running trial simulations.
Use of Trial Simulator to Modify the Original Titration Scheme
The population PK model was then fed into Trial Simulator to simulate different trial designs.
The first step was to confirm that the imported model was valid. For the confirmation simulation, Certara used rich PK sampling on Days 30 and 60 with the titration based on the 4-6 hour post dose collection sample. The frequency of outlying Cmax and Cavg were accurately predicted with the proposed PK model (comparing observed vs predicted).
The second step was to make a direct comparison between the current adaptive design (Figure 2) and the competing designs to validate the Day 90 endpoint. The simulated Cmax was higher than the observed, but this means that the model is on the conservative side (over-predicting), which works for the purposes of this example.
Next, Certara evaluated the differences in outcomes (efficacy and safety) at Day 90 when four aspects of the proposed study design were varied (Figure 2). An important feature of Trial Simulator is the ability to change individual properties of the trial design by creating different scenarios to see differences between properties without altering the original design. This makes comparing designs very easy and also allows for identifying which properties should be modified.
Impact of Four Aspects on Study Design
1) Starting dose
The proposed study design had a drug starting dose of 200 mg T twice daily (BID) at Day 0. When the starting dose was decreased from 200 mg T BID to 150 mg T BID, this change resulted in only a slight improvement in the predicted ability to meet the efficacy objective and a large improvement in the predicted ability to meet the safety objective. By contrast, a starting dose of 100 mg T yielded a substantial proportion of subjects with serum T Cavg (efficacy) below the lower normal limit of 300 ng/dL.
2) Pharmacokinetic sampling time window
The results of the pharmacokinetic collection timeframe for a single sample suggested that changing the 4 to 6 hours sample collection time window to 3 to 5 hours can improve the ability to meet the safety objective (Cmax) without affecting the likelihood of achieving the efficacy objective.
3) Range of T concentrations that would prompt dose titration/adjustment
Reducing the 250 –1100 ng/dL bounds for titration to 250-700 ng/dL resulted in a marked improvement in safety with no change in efficacy. Decreasing the upper bound (eg, to 700 ng/dL) was critical to improve the proportion of subjects that met the safety objective but did not reduce the proportion of subjects meeting the efficacy objective.
4) Dose adjustment for titration at Days 42 and 84
The original dose adjustment at Days 42 and 84 was in 100 mg steps. However, when evaluated in the context of the other 3 aspects, 50 mg steps for both dose titration timepoints resulted in improved safety while maintaining efficacy close to a desirable level.
Conclusion
The optimized study design resulted in the approval of the oral T product, and Clarus Therapeutics has now gone on to market the drug. M&S plays an increasingly important role in drug development and regulatory submissions. In this example case, Phoenix NLME and Trial Simulator supported the approval of a novel oral T product. Complex adaptive design features were tested by simulation to select a design with the best chance of success for the Phase 3 study meeting FDA requirements and gaining approval.
Watch this webinar for more details about how Clarus Therapeutics utilized M&S to improve their Phase 3 study design and ultimately gain FDA approval for their product.