Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 Aug 2:6:30723.
doi: 10.1038/srep30723.

Optimized design and analysis of preclinical intervention studies in vivo

Teemu D Laajala  1   2   3   4 , Mikael Jumppanen  3   5 , Riikka Huhtaniemi  3   4   6   7 , Vidal Fey  3   6   8 , Amanpreet Kaur  5   9   10 , Matias Knuuttila  3   4   6 , Eija Aho  7 , Riikka Oksala  4   7 , Jukka Westermarck  5   9 , Sari Mäkelä  3   11 , Matti Poutanen  3   6   12 , Tero Aittokallio  1   2   3
Affiliations

Optimized design and analysis of preclinical intervention studies in vivo

Teemu D Laajala et al. Sci Rep. .

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.

PubMed Disclaimer

Conflict of interest statement

M.P. is the director of Turku Center for Disease Modeling (TCDM), providing preclinical mouse models including statistical analyses of drug interventions. All other authors have declared no competing interests.

Figures

Figure 1
Figure 1. Benefits of the modelling framework over the course of the study period.
The animal baseline matching improves the statistical analysis and design of preclinical animal studies in terms of power calculations, balanced allocations, and intervention blinding (pre-intervention period), as well as through the use of matching information in the statistical testing of the intervention effects (post-intervention period).
Figure 2
Figure 2. Optimal matching of animals in the case of orthotopic VCaP mouse xenografts.
The original task was to randomly assign 75 animals into five balanced intervention groups (one control and four treatment groups, each consisting of 15 animals), but here we focus on two of the treatments only (ARN-509 and MDV3100), using a sub-sample of the complete data matrix (see Supporting Fig. S3). (a) Bivariate observations sampled from the VCaP study, illustrating the two selected baseline variables (body weight and PSA). (b) 15 × 15 dimensional distance matrix D calculated based on the baseline variables was used as an input to the matching procedure, which solves the optimal animal matching matrix X. (c) The optimal submatches from the branch and bound algorithm, which guarantees a globally optimal solution (see Supporting Fig. S7). (d) The optimally matched animals were randomized into the intervention groups via blinded treatment label assignments (coloured points). The baseline matching information was used in the statistical testing of the treatment effects, mainly through paired comparisons between the treated and control animals (solid lines). Alternatively, the model also allows for direct comparisons between the two treatments (dotted lines).
Figure 3
Figure 3. Statistical testing of the treatment effects using pairwise matched inference.
(a) The matched inference makes use of the baseline matching information when testing the intervention effects by pairing the observed responses according to the optimal submatches at equal time points. (b) An example of the submatch-based pairing in the MDV3100 vs vehicle comparison, where the example trajectory was previously shown as a single estimate value in the original study. Complex response differences are better captured when additional baseline information is incorporated into the statistical inference. The paired differences from the longitudinal observations (left panel) construct a single treatment curve for the pairwise matched mixed-effects modelling (right panel). (c) Comparison of the matched and unmatched statistical inference approaches in the MDV3100 vs vehicle comparison. Even if both inference approaches yield rather similar conclusion about the possible intervention effects, the matched approach improves the sensitivity of the detection (right panel). Different aspects of the mixed-effects modelling are visualized based on the observed data (top panel): the full model fit combining both the random and fixed effects (middle panel), and the population inference depicting only the fixed effects along with their interpretation (bottom panel). In the matched inference, the population of paired differences in the intervention effects (βintervention) is tested against a null hypothesis of no paired differences (y = 0 line). The statistical inference results of the intervention effects are summarized in Table 1, and the full model fits for the four treatment cases are shown in Supplementary Figs S5 and S6.
Figure 4
Figure 4. Model-based power calculations for sufficient sample size estimation.
Statistical power (the likelihood that a true treatment effect is detected) as a function of the sample size (animals per treatment arm). Power calculations were computed by bootstrap re-sampling, either without the matching information (unmatched) or using the information from the optimal pairs of matched samples (matched). The estimated sample sizes (N) are defined based on the conventional threshold of 0.8 power. (a) ARN-509 and MDV3100 intervention effects in the VCaP mouse xenografts. (b) ORX and ORX+Tx intervention effects in the orchiectomized (ORX) VCaP mouse xenografts.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Collins F. S. & Tabak L. A. Policy: NIH plans to enhance reproducibility. Nature 505, 612–613, doi: 10.1038/505612a (2014). - DOI - PMC - PubMed
    1. Henderson V. C. et al.. A meta-analysis of threats to valid clinical inference in preclinical research of sunitinib. Elife. 4, 1–13, doi: 10.7554/eLife.08351 (2015). - DOI - PMC - PubMed
    1. Begley C. G. & Ellis L. M. Drug development: Raise standards for preclinical cancer research. Nature 483, 531–533, doi: 10.1038/483531a (2012). - DOI - PubMed
    1. Freedman L. P., Cockburn I. M. & Simcoe T. S. The economics of reproducibility in preclinical research. PLoS Biol 13, e1002165, doi: 10.1371/journal.pbio.1002165 (2015). - DOI - PMC - PubMed
    1. Singh M. & Ferrara N. Modeling and predicting clinical efficacy for drugs targeting the tumor milieu. Nat Biotechnol 30, 648–657, doi: 10.1038/nbt.2286 (2012). - DOI - PubMed

Publication types