Biostatistics in nonclinical and preclinical drug development

Abstract

Compared to biostatistics in clinical drug development, much less papers were published in the area of nonclinical and preclinical drug development. There seem to be several reasons for this. First, the "preclinical" and "nonclinical" are contain a wide range of topics. These include, but are not limited to, the candidate drug search process, toxicology and pharmacological studies, drug stability assays, and bioassays. This broad range of subjects causes the definition of this area of statistics to be much vaguer than that for the phase I-IV clinical trials in drug development. Second, due to the interest of the regulators, most notably in the clinical area, the statistical approaches tend to be more standardized. Refer, for example, the detailed description on baseline adjustments in the "CPMP points-toconsider" by Grouin and Lewis (2004). In the nonclinical area only a few statistics-related guidance documents can be found, for example for carcinogenicity studies in toxicology (anonymous, 2001). The number of staff in statistics dedicated to the respective areas was described as directly related to the number of guidelines (Lendrem, 2002). Thus, as a third effect, clinical topics are dominating at international biostatistics conferences. Fourth, in the clinical area is an external pressure for objectivity and correctness, and therefore eventually for statistical presence; because a mistaken decision in nonclinical research does mainly harm the pharmaceutical company, whereas a mistaken decision in clinical development may harm the public primarily.On April [22][23] 2004, the 4th International DIA workshop on statistical methodology in nonclinical research and development took place in Dublin, Ireland. A total of 20 presentations were given in five sessions: statistics for genomics, in vitro pharmacology, statistics for QSAR, advanced modeling in pharmacology, and statistics of quantitative assays. In this special issue of Biometrical Journal, two of the contributions are printed, together with two additional papers on this topic. Straetmans et al. Design and Analysis of Drug Combination ExperimentsA parametric log-logistic Hill-type model for claiming synergy by calculation the confidence intervals of the interaction indices in the ray design (fixed ratios) is discussed. For in vitro assays on microtiter plates a bias correction for plate-location effects is included. The analysis of developmental toxicity studies is performed by random effects threshold doseresponse model for clustered binary-response data with the assumption of an additional hormetic effect to the threshold effect. A score tests derived from a random effects hormetic-threshold dose-response model was compared with the likelihood ratio test in a simulation study. For experimental in vivo data on DEHP evidence for a hormetic effect was concluded, which was not be as strong as the threshold effect. Further to these papers, I would like to discuss the recent development of statistics in the respective areas.Many approaches for the search of poten...

Year 2005 – Report

“…Issue 3 of the Year 2005 had a discussion paper on the drop-the-losers design by Sampson and Sill (2005) again followed by comments of four experts. The other highlight of this issue was the special topic on Statistical Issues in Pre-Clinical Research edited by Ludwig Hothorn (Hothorn, 2005).…”