Investigation of a Data Transformation Procedure to Improve Identifiability of Efficacy and Potency: Application to Continuous Elevated Plus-Maze Data in Rats
Eline van Maanen (1), Oliver Ackaert (1), Nelleke Snelder(1), Kirsten Bergmann (1), Carla Maciag (2), Mike Quirk (2), Carin Wallsten (3), and Bart Ploeger (3)
(1) LAP&P Consultants BV, Leiden, The Netherlands; (2) CNSP iMed, AstraZeneca R&D Wilmington, Delaware, USA; (3) Modelling & Simulation, CNSP iMed, AstraZeneca R&D, Södertälje, Sweden
Objectives: Many biological variables are not normally distributed, which makes visual data exploration challenging and affects the performance of non-linear mixed effect modeling. Our objective was to develop a data transformation method to facilitate analysis of non-normally distributed data. The effect of the low-trapping, NMDA channel blocker, AZD6765, on the time that prenatally stressed (PNS) animals spent in the open arms of an elevated plus maze (EPM) is used as a case study. The EPM is a plus-shaped maze with 2 open and 2 closed arms. An increase in exploration of the open arms by PNS rats may reflect antidepressant effects.
Methods: Dose ranging, continuous data obtained in the EPM test with male PNS rats receiving a single dose of AZD6765 or saline ip were used with 5-6 repeated measurements up to 9 weeks postdose. The saline treatment group consisted of a non- PNS and a PNS subgroup. An iterative data transformation procedure was followed by calculating: 1) the cumulative sum of time in open arms of the EPM at each time point; 2) the natural logarithm of cumulative sum of time in open arms at each time point. For the analysis of the transformed data, a stepwise modeling approach was followed: 1) the baseline curve (non-PNS) was described; 2) PNS effect on baseline curve was evaluated; 3) the drug effect was assessed.
Results: After the first step in the data transformation, a consistent trend in the data for all treatment groups was obtained: the number of 0 values in the data was reduced and the variability in the data was decreased. Following the second step a close to normal distribution was obtained and a comprehensible PNS effect and dose response was observed. Model development using the transformed data resulted in an adequate description of the dose-response relationship and the observed variability. Potency and efficacy of AZD6765 were quantified.
Conclusion: Data transformation calculations enable data to be converted into a more readily available format that provides a visual representation and allows additional analysis. Data transformation can be applied iteratively and each transformation can produce a different perspective that may provide greater insight and understanding. With the data transformation procedure presented here, potency and efficacy of AZD6765 in the EPM data could be estimated.