William S. Denney (1) and Simon Kirby (2)
(1) Pfizer, Cambridge, Massachusetts, USA (2) Pfizer, Cambridge, UK
Objectives: To summarise all adverse events (AE), vital signs, electrocardiograms (ECG), and lab measurements for healthy subjects receiving placebo in First in Human (FIH) and Multiple Ascending Dose (MAD) studies in Pfizer’s Phase 1 Management System (PIMS) to aid in the interpretation of ‘What is Normal?’ and to provide informative prior distributions for Bayesian analyses.
Methods: All AE, vital sign, ECG and lab measurement data for healthy subjects receiving placebo in FIH and MAD studies were selected from PIMS. AEs were classified using MedDRA 15.0, lab names and units were standardized and baseline was selected as nominal time equal to zero or any point a multiple of 24 hr prior in the same treatment period. AEs were summarised by numbers and percentages of events, subjects, and studies with events. Additionally the distribution of percentage occurrence by study was summarised for common AEs. For vital signs, ECGs, and lab measurements, baseline, raw values, and change from baseline were summarised using distribution quantiles, histograms, and empirical distribution functions. Any numerical measurements with at least 100 subjects were modelled with a linear mixed effect model testing demographic parameters as fixed effects and random effects on intercept by study and subject within study using the lme4 function in R [1]. Model fits for single and multiple covariates were obtained; the latter were estimated using an automated stepwise modelling procedure.
Results: The final data summarised were for 1204 subjects from 82 FIH and MAD studies. Updated ranges for extreme values of labs, vitals, and ECG measurements have been generated, and the importance of demographic parameters on measurements (or lack thereof) has been estimated with many subjects and dense measurements. The results were summarized and posted to an internal website allowing rapid queries without requiring specialized tools.
Conclusion: The analysis has allowed classification of potentially abnormal measurements incorporating the large data set of placebo subjects in similar populations—the placebo population within the current study can be augmented and anchored by historical data. It has also provided data that can be used for the formation of informative prior distributions for future Bayesian analyses. The analysis of healthy subjects has enabled a more thorough estimate of what is normal and quantification of potentially abnormal signals in early clinical development.
References:
[1] Bates, D., Maechler, M., Bolker, B., Walker, S. (2013) lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0-4.
Reference: PAGE 23 () Abstr 3190 [www.page-meeting.org/?abstract=3190]
Poster: Drug/Disease modeling - Safety