In latter day STAR*D I…, I was looking at a study [Depression is not a consistent syndrome: An investigation of unique symptom patterns in the STAR*D study] where the authors demonstrated the wide diversity of symptom profiles among the subjects enrolled in the STAR*D Clinical Trial, yet all carried the same diagnosis [Major Depressive Disorder]. The author, Eiko I. Fried, kindly commented and added a link to the full text PDF. So here’s the update:
by Eiko I. Fried and Randolph M. NesseJournal of Affective Disorders. 2014 172C:96-102.
Also in the comments, Bernard Carroll pointed out that the problem came when they initially chose a "disjunctive format," saying:
At my last count there were 19 possible symptoms in DSM-IV for major depressive disorder, with only 5 needed for the diagnosis. You do the math…
And then Charles Olbert also commented, adding another interesting article from his group coming at the same issue from a different direction. And they actually did a lot of math. But first, a brand new word for me:
po·ly·the·tic polí’θetik
adjective
Relating to or sharing a number of characteristics which occur commonly in members of a group or class, but none of which is essential for membership of that group or class.
-
A monothetic class is defined in terms of characteristics that are both necessary and sufficient in order to identify members of that class. This way of defining a class is also termed the Aristotelian definition of a class.
-
A polythetic class is defined in terms of a broad set of criteria that are neither necessary nor sufficient. Each member of the category must possess a certain minimal number of defining characteristics, but none of the features has to be found in each member of the category. This way of defining classes is associated with Wittgenstein’s concept of "family resemblances."
It’s a general term that describes the kind of classification chosen for the DSM diagnostic system – one where there is a symptom list, and the diagnosis is made by meeting some preset number or configuration of those symptoms:
by Charles Olbert, Gary Gala, and Larry TuplerJournal of Abnormal Psychology. 2014 123[2]:452-62.
Heterogeneity within psychiatric disorders is both theoretically and practically problematic: For many disorders, it is possible for 2 individuals to share very few or even no symptoms in common yet share the same diagnosis. Polythetic diagnostic criteria have long been recognized to contribute to this heterogeneity, yet no unified theoretical understanding of the coherence of symptom criteria sets currently exists. A general framework for analyzing the logical and mathematical structure, coherence, and diversity of Diagnostic and Statistical Manual diagnostic categories [DSM-5 and DSM-IV-TR] is proposed, drawing from combinatorial mathematics, set theory, and information theory. Theoretical application of this framework to 18 diagnostic categories indicates that in most categories, 2 individuals with the same diagnosis may share no symptoms in common, and that any 2 theoretically possible symptom combinations will share on average less than half their symptoms. Application of this framework to 2 large empirical datasets indicates that patients who meet symptom criteria for major depressive disorder and posttraumatic stress disorder tend to share approximately three-fifths of symptoms in common. For both disorders in each of the datasets, pairs of individuals who shared no common symptoms were observed. Any 2 individuals with either diagnosis were unlikely to exhibit identical symptomatology. The theoretical and empirical results stemming from this approach have substantive implications for etiological research into, and measurement of, psychiatric disorders.
When I say they actually did a lot of math, I wasn’t kidding. They calculated the theoretical numbers of symptom clusters possible and the number of possible disjoint pairs [two diagnosable cases that shared no symptoms] and came up with some impressive numbers. Then they looked at two large databases and showed that this kind of heterogeneity was present in nature, not just in a math book. There’s no way to summarize their study in a simple blog post, but I was impressed. Here’s their lone figure illustrating a disjoint in MDD:
And I mangled their theoretical table to fit, primarily to show that the possible disjointed diagnoses aren’t uniform, but hit some pretty important disorders [eg MDD]:
It’s not hard to figure why the DSM framers went for a system like this. Psychiatry has always been criticized for its problems with definition. My take on that is that it is in the nature of the beast. But in the 1970s, they were out to get objective, get medical. The downside of this polythetic classification is obvious. It gives the illusion of descriptive unity where, in many cases, none exists. And that pseudo-unity has been exploited to the hilt in the Clinical Trial era where drugs are approved for specific conditions – most obvious in the area of Major Depressive Disorder. Fried et al and Olbert et al have done a good job in different ways to put this heterogeneity down in black and white. My guess is that one could show the same heterogeneity with a factor analysis of the other metrics like HAM-D, BDI, MADRS, CDRS, IDS, QIDS, etc.
Long ago, before the blood pressure cuff, doctors approximated blood pressure readings by feeling the "hardness of the pulse." Before the thermometer, fever was measured with the back of the hand. Anemia was diagnosed by looking at the color on the back of the eyelid, and diabetes by tasting the urine. You do what you can do until something better comes along. What’s important is to know the precision and limits of the instrument, and not make more of it than it has to offer. The symptomatic, descriptive system worked in a few places, but was heavily overvalued in others. That was apparent on day one, yet it got reified and deified early on for reasons other than clinical accuracy. Worse, it was used by many to imply unproven unity and even etiology.
I appreciate the commenters adding to this discussion…
polythetic polymorphism = polymorphous perversity. Sorry I just couldn’t resist!
Mickey, I appreciate the publicity for the paper! I’ve been reading your blog for years and so it’s quite an honor.
When I got turned on to the issue of within-disorder heterogeneity, the problem had been mentioned for over 30 years (since DSM-III) but, surprisingly, had received very little attention in terms of empirical research. A few papers had calculated theoretically the number of ways to meet criteria for specific disorders, but virtually nobody had examined any disorder empirically, let alone provided a comprehensive treatment of the issue.
Perhaps the main contributions I see our paper making are (1) providing a framework for allowing cross-disorder comparisons as well as (2) documenting that the most extreme types of heterogeneity (disjoint pairs — when two individuals share the same diagnosis yet share no common symptoms) occurs in empirical reality, whether you use self-report data (a private dataset from the Jed Foundation) or structured-interview data (from the National Comorbidity Study).
I’d be happy to answer questions or discuss the issue further — but right now I have to run: believe it or not, this is the morning of my doctoral comprehensive examinations, and so I’ve got bigger fish to fry!
All the best,
Charles
Charles,
Good luck! And thanks for your article. It’s a real contribution…
Comorbid morbidity— why have just one diagnosis when there are so many overlapping possibilities and drugs to go with them.