Brain predictors of fatigue in rheumatoid arthritis: A machine learning study, 2022, Goni et al

[Sly Saint]

Abstract
Background
Fatigue is a common and burdensome symptom in Rheumatoid Arthritis (RA), yet is poorly understood. Currently, clinicians rely solely on fatigue questionnaires, which are inherently subjective measures. For the effective development of future therapies and stratification, it is of vital importance to identify biomarkers of fatigue. In this study, we identify brain differences between RA patients who improved and did not improve their levels of fatigue based on Chalder Fatigue Scale variation (ΔCFS≥ 2), and we compared the performance of different classifiers to distinguish between these samples at baseline.

Methods
Fifty-four fatigued RA patients underwent a magnetic resonance (MR) scan at baseline and 6 months later. At 6 months we identified those whose fatigue levels improved and those for whom it did not. More than 900 brain features across three data sets were assessed as potential predictors of fatigue improvement. These data sets included clinical, structural MRI (sMRI) and diffusion tensor imaging (DTI) data. A genetic algorithm was used for feature selection. Three classifiers were employed in the discrimination of improvers and non-improvers of fatigue: a Least Square Linear Discriminant (LSLD), a linear Support Vector Machine (SVM) and a SVM with Radial Basis Function kernel.

Results
The highest accuracy (67.9%) was achieved with the sMRI set, followed by the DTI set (63.8%), whereas classification performance using clinical features was at the chance level. The mean curvature of the left superior temporal sulcus was most strongly selected during the feature selection step, followed by the surface are of the right frontal pole and the surface area of the left banks of the superior temporal sulcus.

Conclusions
The results presented evidence a superiority of brain metrics over clinical metrics in predicting fatigue changes. Further exploration of these methods may support clinicians to triage patients towards the most appropriate fatigue alleviating therapies.

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0269952

 

[rvallee]

It detect changes!

This is the Scientology e-meter, which "detects" Thetans, ghosts that inhabit the bodies of people and cause illness. Using the CFQ here is just as much of a joke as this machine, or polygraphs, or Tarot cards, or anything really.

 

[MEMarge]

In this study, we identify brain differences between RA patients who improved and did not improve their levels of fatigue based on Chalder Fatigue Scale variation (ΔCFS≥ 2), and we compared the performance of different classifiers to distinguish between these samples at baseline.

The great CFS once more!

 

[alktipping]

what are these mysterious fatigue relieving therapies.

 

[Hutan]

The general idea of the study was ok - collect a lot of data at baseline and at 6 months and see if there were any correlations between the data and fatigue improvement. But the researchers seem to be in thrall to their methodologies and machines, using a chromosome/gene metaphor to add an air of complexity to what was basically an exercise in fishing for correlations.

In much the same way that the recent study of fatigue in women after giving birth ignored important real life factors that might account for fatigue, this study also gets carried away with the various medical machinery, and doesn't adequately account for lifestyle and illness changes which have to be at least as likely to affect reports of fatigue as the change in, for example, the "curvature of the left superior temporal sulcus".

The paper is not written well with, as I said, a lot of extraneous waffle. It's not clear to me if there was another clinical assessment at the end of the 6 months. So, I can't tell if the change in fatigue levels correlates with a change in disease status. Fatigue level might also have changed due to an increase or decrease in work hours or support. The possibility that medication might be a factor is raised as a limitation of the study in the discussion - but it's a big one.

On the CFS/CFQ:

An inherent limitation of this work lies in using the Chalder Fatigue Scale (CFS) to stratify subjects into improvers and non-improvers of fatigue, which is a self-rated score questionnaire. This tool gathers both physical and mental scores. Nonetheless, it may give a general overview rather than capturing the severity of fatigue in depth. Additionally, since CFS is a self-assessment questionnaire, the final fatigue score may be biased for the subjective individual perception of fatigue. Still, CFS has been the chosen tool in this work as it has been widely applied and validated in the current literature.

They aren't fully aware of the problems with the Chalder Fatigue Questionnaire. It pretty much invalidates the whole study, as it's impossible to know if they have captured variation in fatigue improvements or just variation in the way people interpret the questions.

Our results show the superiority of MR metrics in the prediction of fatigue changes, compared with commonly used clinical measures.

That's a brave call. They found some measures that were correlated with a possible fatigue improvement in their very small sample. That certainly doesn't mean the measures will be of any use in subsequent studies.

 

[Jonathan Edwards]

It isn't clear from the abstract whether the initial MRI findings predicted fatigue improvement or that the change in MRI over 6 months correlated with fatigue improvement. I cannot be bothered to try to find out because the whole thing looks like gibberish.

Surely asking the patient how they feel (a clinical measure) is more use, as a way of finding out how they feel, than any of this?

 

[Hutan]

Surely asking the patient how they feel (a clinical measure) is more use, as a way of finding out how they feel, than any of this?

I guess if a change in the curvature of the left superior temporal sulcus really was correlated with an improvement in fatigue, that might give a clue as to what treatment should be offered (with that part of the brain's links to 'audiovisual processing and motion perception' I predict it will be a course of jazzercise).

But yeah, it really is gibberish. All those words, and although they say the procedure was repeated at 6 months, I still don't know if they used baseline and/or 6 months data for their model.

They had 596 structural features and 304 diffusion features. Quite a lot to pick from. And yet the predictive value of their models is really poor (less than 70% accuracy, even when they use 10 variables).

It would be essential to validate our model on an independent dataset. To do this, it is important to bear in mind that all patients from our dataset were recruited locally from Scotland, UK. This means that these results might not be reproducible in other countries, given environmental differences.

They seem to be suggesting that their model is great, and if it isn't reproducible, it's because the environment of other countries changes which bits of the brain are related to fatigue.

 

[rvallee]

It sounds a bit exaggerated, hyperbolic, when I say that most of how biopsychosocial happens in evidence-based medicine is a popularity contest, an issue of fashion, rather than reliability. Nope, it literally is. It's bad but fashionable so it's used. Same as all the stupid labels invented for us over the years. We didn't invent them, they did. And they continue to be used because they are fashionable.

Still, CFS has been the chosen tool in this work as it has been widely applied and validated in the current literature.

It's like those psychological experiments they do with a fake group, where only one person is the actual subject and everyone else is part of the experiment, and they're asked simple things like which line is shorter and everyone part of the experiment gives the wrong answer because the goal is to check how people react to situations like this.

Everyone says the longer line is shorter? Well I don't want to rock the boat here so it must be. This questionnaire is obviously bad but it's very fashionable? Let's go right ahead and use it. That's how you build a career.

 

[Hutan]

It sounds a bit exaggerated, hyperbolic, when I say that most of how biopsychosocial happens in evidence-based medicine is a popularity contest, an issue of fashion, rather than reliability. Nope, it literally is.

I don't find that hyperbolic at all. Much of the way life is done is a popularity contest - part of it is fashion, part of it is doing or saying what you think will be acceptable to those who have power. Careful logic often comes a fair bit of the way down the list, hopefully not so much when building bridges, sadly all too often in medicine and medical research.