|
 |
|
|
|
||
 |
 |
|
||||
 |
||||||
|
|
Session B7: Experimental Design: A Tool to Promote Refinement and ReductionChairs: Derek Fry (UK) and H. van Wilgenburg (The Netherlands) B7: Reduction by Well-Defined Objectives Examination of programs of animal work indicates the scope for reduction by identifying clear objectives. Having "to optimize the methods" as a defined first objective for new or transferred methodology, rather than an exploratory "try it with a few animals and modify" approach, is more likely to obtain the conditions for best "signal to noise" ratio. With repeated use of the same methodology, even small gains in this may be worthwhile, as the reduction is cumulative. Sometimes subdivision to more precise objectives helps concentrate the animal usage. For example, separating "to determine the time of a distinct peak" from "to determine the size of the peak" avoids using the numbers needed to achieve the latter for every time point at which the peak might occur. For a program, a series of objectives relating to decision points focuses thought on decision criteria and the most efficient design for each stage. Examples of reduction by such means will be presented. B7: Optimizing Resources via Reduction: The Frame Reduction Committee The FRAME Reduction Committee, formed in 1998, comprises experienced professionals in the fields of statistics, experimental design, animal welfare, and alternatives. Literature surveys suggest that often the number of laboratory animals used in experiments could have been reduced, while still resulting in the generation of statistically valid data. The FRAME Reduction Committee undertakes projects designed to contribute to reduction by addressing the implementation of recommendations made at the ECVAM workshop, "The Three R's: The Way Forward." The projects include ongoing reviews of computer statistical packages and the creation of a directory of training material on experimental design and statistical analysis. The poster highlights one practical example of a reduction strategy to address the logistical problem posed by an anti-cancer drug development screening system. B7: Computer Simulation for Improving the Precision of an Experiment Poorly designed animal experiments cannot be ethically justified. Understanding variation in order to control variability and to minimize experimental error should be a requirement for designing an experiment that will find reliable, efficient results. Realistic experimental conditions can be simulated with computer simulations. A computer aided learning program has been developed that allows formal design, such as completely randomized, randomized block, crossover, and sequential design, and their statistical analysis. Unwanted variability will increase the number of animals used in subsequent experiments. This can be simulated in examples. Accurate data can then be collected. With visualized mathematical methods, the appropriate sample size can be determined and a final choice of an optimal experimental design can be selected. By paying careful attention to the factors that affect variability and efficient experimental design, the quality of information provided by real experiments will increase, resulting in a reduction in animal use. B7: Reduction by Use of Factorial Experimental Design In the pharmaceutical industry, compounds are passed through a cascade of screens for testing biological activity against a specific target. In vivo models are developed in later stages of these cascades. Factorial Experimental Design (FED) provides a set of tools for rapid learning about processes, enabling a structured approach and efficient and effective decision-making. The techniques have been widely and successfully applied in a variety of fields outside research, including optimization of manufacturing processes. Historically, its use within drug discovery has been more limited; however, recent experience in AstraZeneca has shown significant benefits in reducing time scales for developing in vitro and in vivo assays and for establishing optimum conditions that benefit the life of the screen. In the case of animal work, it is possible to use significantly fewer animals by adopting FED in comparison to more conventional approaches. Furthermore, the conditions developed for in vivo screens are more likely to be optimal, ensuring maximum benefit from animals used within the screen itself. The presentation will introduce the concepts of FED, how the techniques can be applied effectively, and what the benefits are, with real examples from AstraZeneca R&D. B7: Adequate Statistical Methods to Reduce the Number of Animals Used in Behavioral Experiments: The Analysis of the Behavioral Sequences In ethological and behavioral toxicological studies, elaborate behavioral patterns shown by the animals in well established experimental paradigms or naturalistic conditions are routinely observed and split in single behavioral items. Subsequently, these items are analyzed in terms of their frequencies and/or durations. Behavioral observations are usually videotaped and scored by dedicated software, which collect the sequences of behavioral items together with frequencies and durations. So far, the Cox proportional hazards model, a method originally developed for the analysis of time-to-event data, has been employed for the analysis of the time-structure of behavior, but its usefulness has been limited, because of not allowing the inclusion of random effects in the model. However recent developments in mixed models for the analysis of time-to-event data may overcome these limitations and improve the analysis of behavioral patterns. Data on the effects of exposure to some toxicants on social interactions in mice will be presented to illustrate the use of these new statistical methods. The study of behavioral sequences may highlight the role of the investigated conditions in setting behavioral organizations. In addition the refinement of this statistical approach may contribute to a reduction in the number of animals used in this field of the life sciences.
|
|
|
|
|
