What makes an accurate and reliable subject-specific finite element model? A case study of an elephant femur

J. R Soc. Interface 9, 351–361 (7 February 2012; Published online 13 July 2011) (doi:10.1098/rsif.2011.0323)

Subsequent to publication of ‘What makes an accurate and reliable subject-specific finite element model? A case study of an elephant femur. J. R Soc. Interface 9, 351–361 (7 February 2012; Published online 13 July 2011, doi:10.1098/rsif.2011.0323)’, the authors (O. Panagiotopoulou, S. D. Wilshin, E. J. Rayfield, S. J. Shefelbine and J. R. Hutchinson) uncovered problems with the methods used in the paper and therefore some of the final data presented therein.

In the analysis presented in this paper, finite element model nodes other than those shown in the electronic supplementary material, figure S4, in the paper were used to select data from for comparison to the experimental (laser speckle interferometry) data. While that model itself appears to be valid, the data taken from it are therefore invalid, as they do not match the location of the region that was experimentally sampled.

Furthermore, the difficult process (due to differences in model versus experiment resolution, for example) of comparing the finite element data to average ‘nodes’ from regions of the experimental data was conducted in a subjective way that resulted in an inappropriate match of the experimental data to the model's data. This subjective procedure was not noted in the Methods (p. 355), and on reflection and discussion among the authors after realizing its usage, we agreed that it was invalid as it biased the analysis to favour the conclusion that the model and experiment matched well, sometimes within 5–7%. This original conclusion did not seem implausible to us, as we conducted a detailed characterization of femoral material properties, but on intense scrutiny of the methods after becoming aware of potential flaws well after publication, we now see that it was unjustified.

Considering these problems with the selection of the data from the models and experiments, we judge that the analysis in the paper is flawed and the major conclusions about the accuracy of the model versus experiment (and homo- versus heterogeneous models) are not reliable in the form presented in the paper. We therefore recommend to the editors and journal that the paper is retracted, while we re-design and re-execute the analysis more appropriately for a new paper that lacks these problems. We apologize to the journal, reviewers and readers for this serious oversight.

O. Panagiotopoulou

School of Biomedical Sciences, The University of Queensland, Brisbane,

Australia, o.panagiotopoulou@uq.edu.au

S. D. Wilshin

Structure and Motion Laboratory, Department of Comparative Biomedical

Sciences, The Royal Veterinary College, University of London, London, UK,

swilshin@rvc.ac.uk

E. J. Rayfield

School of Earth Sciences, University of Bristol, Bristol, UK,

e.rayfield@bristol.ac.uk

S. J. Shefelbine

Department of Mechanical & Industrial Engineering, Northeastern University, Boston, MA, USA,

s.shefelbine@neu.edu

J. R. Hutchinson

Structure and Motion Laboratory, Department of Comparative Biomedical Sciences, The Royal Veterinary College, University of London, London, UK,

JHutchinson@rvc.ac.uk

Professor Leslie Dutton FRS

Editor-in-Chief, Interface