Bad science: Part 2
In this post I would like to highlight another study to illustrate how individuals affiliated with Galgo Medical present misleading results to try to “sell” 3D-DXA as a valid research tool. This is with regards to the following study by Jorge Malouf on assessing the bone density changes in response to a PTH (1-84) treatment1. Now let’s be honest. No one is particular interested in the presented changes in bone density. Everything is already known from more elaborate clinical trials and retrospective studies on real CT scans. What the study by Jorge Malouf is really about, is to attempt to convince other researchers that 3D-DXA can be used for research into drug responses. You should be aware that several of the co-authors are affiliated with Galgo Medical and have a financial interest in the 3D-DXA software. I will have more to say about financial disclosures in a later blog post.
When presenting results from studies in scientific publications, you are always expected to provide a significance value, i.e., a p-value. This value indicates how likely your changes are due to mere chance alone, and thus how reliable your numbers are. To say it bluntly, if the p-value is higher than 0.05 your reported changes are considered meaningless. Let’s now have a look at the abstract in question which can be found here. The text is in Spanish but you can still have a look at the presented numbers. I would like to highlight the cortical measurements in particular, of which none of the presented changes are in fact statistically significant. This means that it is very likely that they show mere random variations. This is not surprising since I already explained that 3D-DXA can not measure cortical parameters. Nonetheless the decrease in cortical density is sensible and has been reported in previous studies using CT scans. But what about the other parameters? Galgo Medical claims to be able to extract a vast about of measurements from a 3D-DXA reconstruction. Why limit this study only to the cortical density and some volumetric densities? There is no good reason to not also present the cortical thickness change, right? Well, unless this change is not sensible and might make it appear that 3D-DXA in fact can not measure the cortical parameters at all. Are we perhaps the victim of selective reporting? It certainly does appear that the authors did all the measurements using the 3D-DXA software, which mostly gave some random changes, and then chose to report only those parameters which gave sensible results. This, by an unsuspecting reader or conference attendee might seem to validate the 3D-DXA software, but in fact proves nothing. Hiding negative results or selectively reporting positive results is unfortunately a widespread problem in scientific publications and particularly damaging in medical research.
Jorge Malouf has now been made aware of the illegal activities and otherwise questionable practices by members of Galgo Medical and I’m confident that he will continue his research using only the highest of standards. Which means staying far far away from 3D-DXA.
- Jorge Malouf, Roger Fonollà Navarro, Y. Martelli, L. Del Rio, S. Di Gregorio, L. Humbert. Cortical and Trabecular Bone Mineral Content changes at the Proximal Femur assessed by 3D-DXA after PTH (1-84) treatment. WCO-IOF-ESCEO 2015.