Abstract
Recent reports have called into question the reproducibility, validity and translatability of the preclinical animal studies due to limitations in their experimental design and statistical analysis. To this end, we implemented a matching-based modelling approach for optimal intervention group allocation, randomization and power calculations, which takes full account of the complex animal characteristics at baseline prior to interventions. In prostate cancer xenograft studies, the method effectively normalized the confounding baseline variability, and resulted in animal allocations which were supported by RNA-seq profiling of the individual tumours. The matching information increased the statistical power to detect true treatment effects at smaller sample sizes in two castration-resistant prostate cancer models, thereby leading to saving of both animal lives and research costs. The novel modelling approach and its open-source and web-based software implementations enable the researchers to conduct adequately-powered and fully-blinded preclinical intervention studies, with the aim to accelerate the discovery of new therapeutic interventions.
Highlights
We demonstrate the benefits of these tools over conventional analysis in two applications of orthotopic xenografts of VCaP prostate cancer cells in immune deficient mice as disease models for castration-resistant prostate cancer (CRPC) (Supplementary Fig. S1)
The importance of controlling for individual variation is well-acknowledged in human clinical studies, with the aim to increase the study validity and reproducibility[26]
In preclinical studies, reproducibility of the findings is associated with transparent reporting and paying careful attention to the experimental issues, including balancing, randomization and blinding[2,3]
Summary
We developed and implemented a novel methodology to improve the experimental design and statistical analysis of preclinical studies carried out with experimental animals. The benefits gained by such matching-based paired testing became more evident with the MDV3100 case, where we observed that the animal body weight at baseline was inversely associated with the final PSA level (correlation coefficient ρ =−0.607, p = 0.021, Supplementary Fig. S2d) Such multivariate, longitudinal relationship between the baseline variables and treatment responses cannot be captured by the conventional, unmatched model, leading to reduced statistical sensitivity (Fig. 3c, left). The paired inference, enabled by the matching information, was able to capture these pairwise response differences, leading to subtle yet significant intervention-specific effect sizes (Table 1) These results support the improved statistical sensitivity gained by the baseline matching information in the detection of true treatment effects, especially when studying more complex and subtle intervention effects. A more fair comparison point for the matched approach would require a new study, conducted without using the baseline matching information, but this was not carried out within our preclinical studies because of ethical reasons
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
