Molecular Psychiatry commentary: Fried & Kievit 2015

On December 15th, Molecular Psychiatry published our commentary “The volumes of subcortical regions in depressed and healthy individuals are strikingly similar: a reinterpretation of the results by Schmaal et al”. You can find the full text PDF in the above link if you have a subscription to the journal, otherwise see the project’s open science repository where you can find a full text (PDF) along with the R-code and data we used for the simulation analysis.

The commentary is a response to the insightful paper by Schmaal et al. (2015) published in Molecular Psychiatry a few months ago: a large team of collaborators meta-analyzed the volumes of a number of brain regions, and found the hippocampal volume to be smaller in 1728 depressed patients compared to 7199 healthy participants. We greatly commend the authors to the huge amount of very thorough work they did; analyzing a number of large datasets really moves the field forward by providing much more accurate estimates than the prior literature.

We wrote the commentary for a number of reasons:

  • The authors have the primary hypothesis that depression causally decreases hippocampal volume.
  • The authors had the goal to “robustly discriminate MDD patients from healthy controls” (MDD stands for Major Depressive Disorder), but do not actually calculate a discrimination parameter; such statistics lie between 50% (there is no discrimination above chance between the healthy and the depressed group) and 100% (every depressed person can be sorted with absolute accuracy to the depressed group).
  • One of the authors said in an interview that “this resolves for good the issue that persistent experiences of depression hurts the brain”, which seems at best very premature, considering the data. While it’s important to mention the possibility that this is a misquote — which happens regularly in such interviews — this quote reflects what we consider a general over-interpretation of the cross-sectional results as causal, and serves as one of many examples.
  • Other groups have started to develop novel antidepressant drugs to stimulate neurogenesis in the hippocampus of patients, based on the strong assumptions that (1) hippocampal volume is reduced in depressed patients and (2) causally related to depression.

In the commentary, we simulated hippocampal volume data based on the sample and effect sizes reported by the authors. We find that the hippocampal volume differences are so small that knowing a person’s hippocampus size does not predict group membership much above chance (the discrimination accuracy is ~52.6%). While the volume is smaller in depressed patients than healthy participants, it certainly does not “robustly discriminate”.


We also discuss why hippocampal volume may not be related to depression at all (for reasons such as lack of disease specificity and confounds), and even if it would be related, why it may not be reasonable to consider it a biomarker.

Comments are very welcome!

» Fried, E. I., & Kievit, R. (2015). The volumes of subcortical regions in depressed and healthy individuals are strikingly similar: A reinterpretation of the results by Schmaal et al. Molecular Psychiatry. (URL) (preprint PDF)

Leave a Reply

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