Optimal Allocation of Prognostic Factors in Randomized Preclinical Animal Studies
Positive results observed in preclinical animal studies rarely translate into the clinical setting, one of the major reasons for which is the subop-timal design of animal experiments. It is of vital importance to the success of a preclinical study to maintain the balance of potentially prognostic factors in all of the treatment cohorts. Various randomization methods have been proposed to increase the likelihood of having balanced prognostic factors at the end of a clinical trial where patients are randomized sequentially, whereas few approaches exist in the literature for preclinical studies where all the animals need to be randomized simultaneously prior to the start of the study. A flexible approach to allocating animals to treatment cohorts in a preclinical study such that the optimal balance of multiple prognostic factors can be achieved is proposed, which allows the prognostic factors to be ranked as needed and can handle both continuous and categorical factors. In the proposed approach an overall imbalance score for the study is first defined based on pairwise comparisons of rank statistics among the treatment cohorts, and then the imbalance score is optimized via a well-known local optimization method. Simulation results show that the proposed method significantly outperforms the standard randomization approach.
KeywordsRandomization Optimization Prognostic factor Preclinical Animal study
Unable to display preview. Download preview PDF.
- 5.Weaver J. Animal studies paint misleading picture. Nature. March 20, 2010; doi:10.1038/news.2010.158.Google Scholar
- 12.Aarts EHL, Korst JHM, van Laarhoven PJM. Simulated annealing. In: Aarts EHL, Lanstra JK, eds. Local Search in Combinatorial Optimization. New York: John Wiley; 1997.Google Scholar
- 14.Hoos HH, Stutzle T. Stochastic Local Search: Foundations and Applications. New York: Elsevier; 2004.Google Scholar