Classifying yin and yang using MRI

Zad Chow writes:

I wanted to pass along this study I found a while back that aimed to see whether there was any possible signal in an ancient Chinese theory of depression that classifies major depressive disorder into “yin” and “yang” subtypes. The authors write the following,

The “Yin and Yang” theory is a fundamental concept of traditional Chinese Medicine (TCM). The theory differentiates MDD patients into two subtypes, Yin and Yang, based on their somatic symptoms, which had empirically been used for the delivery of effective treatment in East Asia. Nonetheless, neural processes underlying Yin and Yang types in MDD are poorly understood. In this study, we aim to provide physiological evidence using functional magnetic resonance imaging (fMRI) to identify altered resting-state brain activity associated with Yin and Yang types in drug-naïve MDD patients.

They didn’t really have much prior evidence to go on with this study, so a lot of the analyses seemed exploratory,

The aim of this exploratory study is to provide physiological evidence, using functional magnetic resonance imaging (fMRI), to identify altered resting-state brain activity associated with Yin and Yang types in drug-naïve MDD patients. Previous studies using functional connectivity (FC) method of resting-sate fMRI demonstrated altered inter- and intral-regional brain connectivity, including local functional connectivity in the medial prefrontal cortex and frontoparietal hypoconnectivity in MDD brains (14, 15). As proposed in Drysdale’s work (8), differential brain function at resting-state may be a useful physiological marker to identify specific subpopulation of MDD patients. Thus, we hypothesize that resting-state brain activity and FC in MDD patients with Yin type are altered when compared to those with Yang type. To test this hypothesis, we examined resting-state functional activities across the entire brain in MDD patients in both Yin and Yang groups as well as matched healthy controls.

The authors ended up finding a few differences that were corrected for using the AlphaSims approach (a method to correct for multiple comparisons in fMRI studies) and a few exploratory comparisons that weren’t corrected for because those were considered exploratory. The authors state,

To the best of our knowledge, this is the first study demonstrating [emphasis added] biological differences in brain function associated with Yin and Yang types characterized by somatic symptoms.

I think the conclusions the authors draw here are fairly interesting because it seems there wasn’t that much evidence to go on with this theory besides ancient Chinese traditions. They acknowledge that a lot of the study is exploratory, but they’re able to say so confidently that they’ve demonstrated biological differences between participants that are classified as “yin” and those classified as “yang”.

I personally believe that subtypes of depression likely do exist. We’ve had some interesting discoveries using data-driven clustering (which is a method that obviously has problems of its own) and it would be in our best interest to discover accurate subgroups so we could tailor therapies for them, but the idea of depressed patients being classified as yin and yang doesn’t seem to sound very realistic to me.

And the conclusions of studies like this, even when correcting for multiple comparisons (which I know you think is unnecessary when using multilevel modeling), make me incredibly skeptical of fMRI studies. Would love to hear your thoughts.

My reply: I took a very quick look at the article. It seems that there are 48 people in the study, and it’s not clear at all how we are supposed to draw conclusions based on the general population. The groups identified as “yin” and “yang” are different in systematic ways—something about somatic symptoms and responses to a questionnaire—so you’d expect to see some differences in other measures too. But, again, I don’t know what this really tells us about people not in the group.

The point of the study can’t be just to demonstrate that the two groups are different. We already knew they were different in some systematic ways, even before doing a single MRI scan. The real question is what are the systematic differences. And, for that, statistical significance is not so useful.

I guess they could consider a preregistered replication. But I share your concern, as it does seem like a bit of a fishing expedition. And I don’t think the researchers would have much of a motivation to do a replication study, as the potential losses from a failed replication are greater than the potential gains from a successful replication.

Just to be clear: I know nothing about yin and yang and I only skimmed the article, I did not read it carefully. So I’m just giving my general impression, which is that I’d be cautious about generalizing beyond these 48 people in the particular setting of the study.

The post Classifying yin and yang using MRI appeared first on Statistical Modeling, Causal Inference, and Social Science.