“Potential error” leads to Expression of Concern for macaque paper

janat414The Journal of Anatomy has expressed concern about a 2011 paper on primate jaws.

The article, “The mechanical function of the periodontal ligament in the macaque mandible: a validation and sensitivity study using finite element analysis,” by a group from the University of York, in the United Kingdom, purported to find that:

Whilst the periodontal ligament (PDL) acts as an attachment tissue between bone and tooth, hypotheses regarding the role of the PDL as a hydrodynamic damping mechanism during intraoral food processing have highlighted its potential importance in finite element (FE) analysis. Although experimental and constitutive models have correlated the mechanical function of the PDL tissue with its anisotropic, heterogeneous, viscoelastic and non-linear elastic nature, in many FE simulations the PDL is either present or absent, and when present is variably modelled. In addition, the small space the PDL occupies and the inability to visualize the PDL tissue using μCT scans poses issues during FE model construction and so protocols for the PDL thickness also vary. In this paper we initially test and validate the sensitivity of an FE model of a macaque mandible to variations in the Young’s modulus and the thickness of the PDL tissue. We then tested the validity of the FE models by carrying out experimental strain measurements on the same mandible in the laboratory using laser speckle interferometry. These strain measurements matched the FE predictions very closely, providing confidence that material properties and PDL thickness were suitably defined. The FE strain results across the mandible are generally insensitive to the absence and variably modelled PDL tissue. Differences are only found in the alveolar region adjacent to the socket of the loaded tooth. The results indicate that the effect of the PDL on strain distribution and/or absorption is restricted locally to the alveolar bone surrounding the teeth and does not affect other regions of the mandible.

But according to the Expression of Concern:

The above article from Journal of Anatomy, published in Volume 218, Issue 1, pages 75–86 (and first published online on 24 June 2010 in Wiley Online Library, www.onlinelibrary.wiley.com), is subject to an Expression of Concern by agreement between the authors, the Journal’s co-Editors-in-Chief, Julia Clarke, Thomas Gillingwater, Anthony Graham and Stefan Milz, and John Wiley & Sons Ltd.

The Expression of Concern has been agreed due to a potential error in the validation results of the experimental model published in a Journal of Biomechanics paper (Panagiotopoulou et al., 2010, 43: 1603–1611) and referenced in this paper.

The Journal of Anatomy paper has been cited 23 times, according to Thomson Scientific’s Web of Knowledge, while the Journal of Biomechanics article has been cited 20 times.

We notice that the earlier paper in the Journal of Biomechanics has not been corrected, nor is it the subject of its own expression of concern — which we find, um, concerning. We’ve emailed the editor of that journal for an explanation and will update this post if we learn more.

5 thoughts on ““Potential error” leads to Expression of Concern for macaque paper”

  1. Here’s the text of what was reported on SF in fall 2012 (as reported to me by “Monkey Alert”, a pseudonymous individual with connections to the case)…

    The study is a finite element analysis (FEA) of macaque mandible. FEA is a method to simulate how complex structures such as bones deform under load. The results are visualized with color-coded plots that show the distribution of stresses and strains.


    Figure 5 (above) shows plots of a model in which biting on one tooth was simulated. According to the legend, the figure shows the model with and without periodontal ligament (PDL), which is the connective tissue that connects each tooth root with the surrounding bone. In the PDL model, PDL was modeled as a layer of soft material surrounding the tooth root. In the no PDL model, PDL material was assumed to have the same mechanical properties as the surrounding bone.

    The result (see also Table 1 & Fig. 6) is remarkable… strain in the two models was almost identical, so the authors claim it is not necessary to include PDL in their model. This assumption then impacts the main part of the study, a comparison of strains in a model system with or without crypts of the developing incisors.The authors state (p. 104): “This remarkable similarity in the strain magnitude patterns and orientation indicated that the presence of the PDL has no effect on the mechanical behaviour of the mandible, and its exclusion will not affect the mechanical performance of the anterior symphysis for the testing of hypotheses regarding the presence of the crypts of the developing incisors.”

    However, it seems there’s another reason underlying the decision to exclude PDL… the modeling software they used is limited to only two materials: (p. 100) “In addition, due to restrictions in the material properties imposed by the FEA software used in this study, only two isotropic and homogeneous properties could be assigned to the FE models.” Thus, if the PDL had affected bone strain, a third material (soft PDL) would have to be included in the crypt/no-crypt model, in addition to two hard materials (1. trabecular bone and 2. cortical bone/teeth), but that was not possible with this software.

    How convenient that PDL had no effect on bone strain. Wouldn’t it be unfortunate if the images in Figure 5 used to support this conclusion were themselves problematic? Well, despite the claim that Figure 5 shows plots of a PDL and a no PDL model, in-fact it appears to only show models with PDL. This is evident because soft materials deform under load more than hard materials, and the PDL material was modeled as a softer material than bone, so strains are much higher in the PDL. In models without PDL (i.e. where the material in that region has the same properties as bone) the strains around the teeth are not elevated (as in the models in Fig. 7 of the paper)…


    The images in the lower half of Fig. 5 (panels b/d/f) clearly show elevated strains in the PDL region and thus undoubtedly show screenshots of a model with PDL. So, of course there’s no difference in strain because all images are from a PDL model.

    In addition, the data in Figures 5 and 7 contradict each other… the PDL model in Fig. 5 and the Crypt model in Fig. 7 should show very similar bone strains if the results of the paper were valid (i.e. that the inclusion/exclusion of the PDL and the way in which the trabecular bone is modeled have only a negligible effects on the strains), but the bone strains in the Crypt model are considerably higher than those in the PDL model (also note that the strain display ranges in the two figures are not the same: e.g. 42.42-391.93 µε versus 48.71-454.97 µε for max. shear strain). Something is wrong here.

    As an aside, the remarkable finding regarding the PDL is also the main result of another paper in the same issue of Journal of Anatomy by the same authors (J Anat 2011;218;75-86; PMID20584094). This has created quite a bit of controversy in the field, since there are similar studies (including another study with a macaque jaw) that have found exactly the opposite, i.e., that including PDL in jaw models has a large effect on the modeling results. (J Biomech 2012;45;1749-1751; PMID22349114).

    1. Whoooaaahhh!!! That’s some deep analysis right there Monkey Alert!

      With respect to Professor Fagans work (J Biomech 2012;45;1749-1751; PMID22349114) – This is a superb example of science at its best and should in no way be overshadowed by the potentially erroneous work of the others.

      note: Both in Hull, UK.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.