By David Tuller, DrPH
As I wrote in a post the other day, the journal NeuroImage: Clinical has just published a letter from a group I organized about the misrepresentation of findings regarding the prevalence of functional neurological disorder (FND). They have also published a response from the authors of the article we criticized. The findings in question were from the Scottish Neurological Symptoms Study (SNSS), a major project that yielded multiple papers a dozen or so years ago.
Since then, more than 50 peer-reviewed articles have asserted, citing the SNSS, that FND prevalence at outpatient neurology clinics is 16% and/or that FND is the second-most-common diagnosis in those venues. In fact, as our letter explained, the only acceptable rate for FND to cite from the SNSS is 5.5%, which represented the patients diagnosed with conversion disorder symptoms–gait and motor disorders, non-epileptic seizures, and sensory deficits–and categorized in the study as “functional.” (Conversion disorder is the former name for FND.) The higher rate was for a group that included many participants who were labeled collectively as having “psychological” diagnoses, yielding a much larger combined “functional/psychological” category.
When we initially wrote to NeuroImage: Clinical about the inflated prevalence claim in one of those 50+ articles, a 2021 paper called “Neuroimaging in functional neurological disorder: state of the field and research agenda,” the journal and the authors quickly agreed to a correction. Our group felt that the proposed correction, as written, was an inadequate remedy. But we accepted it as the best that could be achieved.
After the correction had been agreed upon, I sent a letter to the lead and senior authors of the NeuroImage: Clinical paper. In the letter, I thanked them for agreeing to correct the citation and then urged them to correct the related misrepresentations of the SNSS in several other papers for which one or the other was either the lead or senior author. (This number did not cover the many other papers in which one or both served as a co-author but not the lead or senior author.)
I didn’t hear back. Not long after, however, the journal informed me that the authors had withdrawn their agreement to a correction. What?? I found this hard to understand. What kind of investigators would withdraw their agreement to a correction? Were they surprised or pissed that I would urge them to correct their other papers? Wouldn’t they have realized or assumed that a correction in one paper would require a correction in other papers with the same error?
The other co-signatories of the letter were as flabbergasted as I was. None of us had previously heard of anyone withdrawing an agreement to issue a correction. Presumably the authors found the prospect of contacting many other journals about the need to correct false assertions too unappealing to contemplate.
Instead, the journal invited us to submit a letter as correspondence. Under the circumstances, we agreed. The letter went through two rounds of peer-review; both our initial and later versions were also posted on a pre-print server. (They can be read here and here. It is easy to discern some of the changes that occurred during this process. For example, our initial letter highlighted the authors’ agreement to correct the paper, and their subsequent withdrawal of this agreement; the final version did not.)
Complications continued even after acceptance. For some reason, the journal ended up publishing the authors’ response to our letter before publishing the letter itself. That has been rectified, and both our letter and the response are now online.
Our point is pretty simple. FND is not synonymous with the SNSS’s category of “functional/psychological” and should not be presented as if it were. As our letter noted, what happened here is an example of the phenomenon known as diagnostic creep. The authors made a bold assertion, and then cited a study that very clearly did not support it. It is self-evident, or should be, that misrepresenting the findings from a study being cited is unacceptable.
Not surprisingly, the paper’s authors don’t agree with us. But these are among the leaders in the field of FND, so they have a lot at stake in debates about prevalence—as is clear from the very extensive list of “declared conflicts of interest” appended to the article itself and their response to our letter. (Of course, I have my own “declared conflicts of interest” related to my crowdfunding support.)
Trying to prove FND is the same as “functional/psychological”
In the authors’ response, they acknowledge what is indisputable—the claim that FND is the second-most-common diagnosis was based on a reinterpretation of the data from the study being cited. They then offer four paragraphs of rationalizations and justifications for why, in their view, it was legitimate to equate FND with the “functional/psychological” group in the SNSS. Let’s review.
Paragraph #1: The sentence was a “scene setting statement” meant to make a “brief comment,” and the epidemiology of FND wasn’t “the focus of our paper.”
It is hard to see the relevance of these points. That the intent was to make general comments about an issue that was not the focus of a paper is not an acceptable rationale for violating appropriate reporting and citation practices. A statement should never flagrantly misrepresent the reference being cited in support, whatever the context. Furthermore, the scene-setting nature of the sentence is part of what makes the citation so offensive and ethically untenable, since the authors are presenting their distortion of the data in the article’s opening paragraph as a blanket assertion of fact.
Paragraph #2: The “functional/psychological” category in the SNSS is pretty much equivalent to what we would call FND now.
This long paragraph is the heart of the authors’ response, and it is one big fudge. Here is the core of the defense, in which the authors explain who fell into the “functional/psychological” group besides those with conversion disorder symptoms:
“Others were labelled as…“non-organic” and others received a psychiatric diagnosis like anxiety and depression, but would have had neurological symptoms in order to be referred to a neurology clinic. Some other patients had problems like dizziness and cognitive difficulties likely equivalent to modern diagnoses of functional dizziness and functional cognitive disorder that are now considered part of FND.”
This explanation makes our point for us. Having unexplained symptoms labeled “non-organic,” or having anxiety or depression plus neurological symptoms, does not equate to an FND diagnosis; neither does being dizzy or having cognitive difficulties. Just because some people with those complaints might now be diagnosed as having FND by exuberant proponents of FND does not excuse the wholesale reinterpretation of the diagnoses actually received by all the anxious, depressed, dizzy and cognitively impaired participants in the SNSS.
Then, the authors state this: “So, the category refers to a broader grouping of functional neurological symptoms that clinical neurologists will recognize.” Again, this statement confirms our criticism. In the sentence we questioned from the 2021 paper, the authors authoritatively declared that FND was the second-most-common diagnosis. Here, they appear to concede that they were using FND as a short-hand for this “broader grouping”—whether or not those in this “broader grouping” meet the required FND criteria. In other words, the same experts who insist that FND is not a diagnosis of exclusion but requires rule-in clinical signs are willing to overlook that requirement and identify a “broader grouping” when they want to claim higher FND prevalence rates. They apply one set of criteria when discussing diagnosis, and a much looser one when discussing prevalence.
Paragraph #3. Subsequent papers have endorsed the higher rates.
Like paragraph #1, this point is irrelevant to our concerns about the authors’ decision to misrepresent the findings from a seminal paper in the field. They didn’t cite other studies; they cited a specific study. Furthermore, the SNSS remains the most extensive study of its kind, including many more participants and involving many more neurologists than either of the other two studies mentioned by the authors.
4. Where diagnoses rank in these kinds of lists depends on lots of factors.
Like paragraphs #1 and #3, this point is irrelevant. The concern expressed in our letter did not involve general issues about the complexities of ranking a list of diagnoses. Whatever those complexities are, they do not justify misrepresenting findings from a major study. The authors themselves chose to rank FND as the second-most common diagnosis at outpatient neurology clinics. Now they want to lecture and opine about the various factors that impact rankings.
The bottom line here is this: A group of distinguished researchers misrepresented key data from a prominent study in their field. They got called out. Instead of forthrightly acknowledging their mistake and correcting it, they have issued an unconvincing response—in the process documenting the validity of the criticism.