1. Using extrapolation to support a pediatric investigational plan: an application in liver transplantation development
Thomas Dumortier, Martin Fink, Ovidiu Chiparus
Pharmacometrics
Novartis Pharma AG, Basel, Switzerland
Background
Conclusion
Given design differences between adult and paediatric studies, this assessment could not be obtained via such a s simple comparison.
The Paediatric Investigation Plan (PIP) for everolimus (EVR) in liver transplantation (Tx) included an extrapolation analysis to obtain a rational interpretation of the limited paediatric evidence in the context of existing adult data.
A goal of this analysis was to assess (‘validate’) the hypothesis of similar efficacy between paediatric and similarly treated adult patients.
Generally, a validation is achieved by a simple comparison of the efficacy results between the paediatric study and an adult study.
Using a pharmacometric approach tailored to account for the differences between study designs, a valid assessment of similarity was obtained.
This assessment supported similar efficacy between paediatric and adult populations
Objectives
To validate the hypothesis of similar efficacy between paediatric and adult patients similarly treated with EVR while accounting for design differences between adult and paediatric studies
Methods
Results
The incidence of treated biopsy proved acute rejection (tBPAR) after randomization in the adult study (Fig. 1) was modeled by a hazard model (Fig. 2). The effect of tacrolimus (TAC) and EVR minimum concentration (Cmin) on the hazard was explored. The baseline hazard was considered to represent the Tx-induced immunological risk. One- compartment population PK models were used to predict the time-course of those concentrations (not described further in this poster; see [1] and [2] for an example).
Step 1 – Build the hazard model in adults
There was a strong TAC conc. effect, no differential EVR effect in the range of observed conc., and a higher immunological risk (i.e., baseline hazard) in the first 2-3 months post Tx (Fig. 3). The final model was accordingly estimated (Fig. 4)
Step 2 (Cont’d)
Accordingly, the predictive distribution for the incidence of tBPAR in the paediatric study could be calculated (Fig. 6).
Figure 5. Survival curves for each EVR + rTAC patients of the adult and paediatric studies, assuming the adult model holds
Figure 3. graphical representation of the concentration-tBPAR relationship in adult data
Figure 1. Design of the paediatric and adult studies
Paediatric:
One line represents the survival function of one actual patient of the EVR + rTAC arm (N=461) of the adult study and of the paediatric study (N=22) as predicted from the final adult model.
Step 3 – VALIDATION
There were no events observed among the 22 patients of the paediatric study.
This observed efficacy is at the mode of the predictive distribution (Fig. 6).
This supports validation of the similarity assumption.
Adult:
The boxplots summarize the distribution of individual predicted Cmin on a day with at least one event in that treatment group . The red dot ( ) is the Cmin of the patients with event on that day.
Methods
Figure 4. Final adult model
Figure 7. Predictive distribution* of the number of patients of the paediatric study with tBPAR event (assuming the adult model holds), and number of observed events
sTAC and rTAC = TAC at standard and reduced exposure
Figure 2. Schematic representation of the model
Effect of TAC concentration and EVR treatment on the probability of being tBPAR-free
Estimated baseline hazard (immunological risk)
Population PK
Time-to-event
EVR dose
EVR Cmin.
tBPAR
TAC dose
TAC Cmin
sTAC
EVR + rTAC
Immunological risk
This model was hypothesized adequate to predict the tBPAR incidence for adults treated like the patients of the paediatric study (Fig. 1), in spite of differences in design between the 2 studies, such as the immunological risk* relative to EVR starting time. (*expected higher for 2-3 months after Tx, see [3]).
Analysis strategy:
Build the hazard model in adults.
Predict efficacy for adults with similar concentration time-course as the paediatric patients during the analysis period (i.e., from the EVR start time). This constitutes the predictive distribution for the incidence of tBPAR in the analysis period in the paediatric study under the assumption that the adult model holds.
Validation: Compare this predictive distribution to the efficacy observed in the paediatric study.
* Assuming the adult model holds in children
Discussion
Step 2 – Predict efficacy for adults similarly treated as the patients of the paediatric study
The paediatric study included 22 patients.
The probability of staying event-free for adults treated like the paediatric patients was larger than that of the patients of the adult study (Fig. 5).
This was mostly driven by the lower immunological risk (baseline hazard) at the time paediatric patients enter the analysis (Fig. 1, 4).
By contrast, the TAC concentration had no differential influence on the probability of staying event-free, as the TAC concentration of the paediatric patients was similar to that of adults’ treated with EVR + rTAC (data not shown).
Validating the assumption of similar efficacy between paediatric and adult patients is an important prerequisite to allow paediatric extrapolation.
Differences in study designs between adults and paediatric studies can confound the similarity assessment.
Depending on the study design difference, adult and paediatric data can be ‘bridged’ using an adequate model, such as in this work with a pharmacometric approach, but in other instances, it may not be possible to bridge the data of the 2 studies.
It is thus important to keep those limitations in mind when designing a paediatric study.
References: 1. Kovarik JM et al. (2001) Population pharmacokinetics of everolimus in de novo renal transplant patients: impact of ethnicity and comedications. Clin Pharmacol Ther; 70:247–54.
2. Staatz CE and al. (2004) Clinical Pharmacokinetics and Pharmacodynamics of Tacrolimus in Solid Organ Transplantation. Clin Pharmacokinet; 43 (10):
3. Dumortier, T. et al. Estimating the contribution of everolimus to immunosuppressive efficacy when combined with low-dose tacrolimus in liver transplant recipients: an innovative model-based approach. Clin. Pharmacol. Ther. (2014)
This study was supported by Novartis Pharma AG, Basel, Switzerland.
Copyright © 2016 Novartis Pharma AG, Basel, Switzerland.
All rights reserved.
Poster presented at the Population Approach Group Europe (PAGE), June 7-10, 2016, Lisbon, Portugal.