An Application of Modelling and Simulation to Type 2 Diabetes: Development of a general drug-disease model based on a meta analysis of over 40 studies investigating 5 PPAR drugs
Alan Maloney(1), Klaas Zuideveld(2), Karin Jorga(2), Cornelia Weber(2), Nicolas Frey(2), Per Olsson(1), Eric Snoeck(1)
(1) Exprimo Consulting LLP, Colchester, Essex, United Kingdom
(2) Clinical Pharmacology, F. Hoffmann-La Roche Ltd., Basel, Switzerland
The PPARγ / PPARαγ class of drugs represent an excellent addition to the pharmacotherapeutic choices for the management of type 2 diabetes.
The primary goal of this project was to describe this pharmacotherapeutic area with a drug/disease model, such that an investigational new drug completing Phase 2 could be incorporated into the model, and various potential phase 3 designs could be prospectively simulated and evaluated.
To date, 42 studies investigating 5 drugs (Actos, Avandia, Rezulin, Ragaglitazar and GI262570), with over 130 treatment arms, have been considered in a complex longitudinal model that describes the changes in glycosylated haemoglobin (HbA1c) and fasting plasma glucose (FPG) as a function of drug, dose, dosing regimen, treatment duration, concomitant therapy, and baseline patient characteristics. Unlike classical meta analysis, the use of a longitudinal model, combined with multiple study level random effects, permitted the aggregation of studies with various designs, and the resulting model effectively described all the key features of how the glycaemic responses changed with time, and the similarities and differences between studies.
Both standard model qualification and additional PPC’s via simulation showed that the general drug-disease model for Type 2 diabetes was adequate for the purpose for which it was developed. With this knowledge, new potential study designs could be investigated. A range of important metrics were defined and then determined from the simulations. These included:
- Dose Response (predicted differences) versus placebo (and baseline).
- Predicted differences versus comparator Y (at any dose).
- Likelihood of achieving superiority/non-inferiority in a given active control phase 3 design.
- Impact of different inclusion criteria (e.g. patient population).
- Extrapolation of phase 2 results to longer treatment durations.
- Associated prediction intervals on all of the above figures.
Key results were also subject to a range of sensitivity analyses, and showed than the main conclusions where robust.
In conclusion, a prospectively planned development of a general drug disease model for Type 2 diabetes allowed the fast synthesis of the results from an investigational new drug completing Phase 2 into the broader pharmacotherapeutic model, thus allowing an extensive range of potential phase 3 designs to be predicted and evaluated shortly after the data was available.