Model to Explain Persistence of Fatigue in CFS (not recent but very influential)

[oceanblue]

The model put forward in this 1998 paper has been much cited by the Dutch CBT school and was explicitly used by them as the basis for their version of CBT for CFS. Yet it is a fundamentally flawed piece of work.

The persistence of fatigue in chronic fatigue syndrome and multiple sclerosis: Development of a model, Vercoulen 1998

JHMM Vercoulen, CMA Swanink, OR Hommes, JMD Galama, G Bleijenberg, PJH Jongen, JFM Fennis, and JWM van der Meer

Abstract

Objective. The cause of chronic fatigue syndrome (CFS) is unknown. With respect to factors perpetuating fatigue, on the other hand, a model has been postulated in the literature in which behavioral, cognitive, and affective factors play a role in perpetuating fatigue. In the present study this hypothesized model was tested in patients with CFS and in fatigued patients with multiple sclerosis (MS).

Method. 51 patients with chronic fatigue syndrome and 50 patients with multiple sclerosis matched for age, sex, and education. The hypothesized model was formulated in terms of cause and effect relationships and an integral test of this model was performed by the statistical technique "structural equation modeling".

Results. Attributing complaints to a somatic cause produced low levels of physical activity, which in turn had a causal effect on fatigue severity. In contrast, depression had to be deleted from the model. Two new causal relationships were found: sense of control over symptoms and focusing on bodily symptoms each had a direct causal effect on fatigue. The model showed excellent fit for CFS patients, but was rejected for MS patients. Therefore, a new model for MS patients had to be developed. Sense of control had a causal effect on fatigue. In the MS model no causal relationship was found between the somatic substrate as measured by the Expanded Disability Status Score and fatigue or functional impairment. Causal attributions, low level of physical activity, and focusing on bodily symptoms played an important role in the persistence of fatigue in CFS, but not in MS.

Conclusions. The etiology of CFS is unknown, but the present study shows that cognitive and behavioral factors are involved in the persistence of fatigue. Treatment should be directed at these factors. The processes involved in the subjective experience of fatigue in CFS were different from the processes related to fatigue in MS.

note this abstract come from JHHM Vercoulen's Doctoral Thesis (see Chapter 9) but is just a tiny edit different from the published article so I'm also assuming the paper is not materially changed from the Thesis version.

The resulting model looks like this:



There are several fundamental flaws to this stuy:

1. This is a cross-sectional study (no 'before' and 'after' data) - and correlation does not mean causation, which is Game Over for the study, really. Still:
2. The patients had Oxford-criteria defined CFS and a quarter of them did not have daily fatigue
3. Some of the measures eg of 'functional ability' are dubious at best
4. The authors hypothesised model failed and they had to modify it 3 separate times before it 'fitted' the data - taking post-hoc chicanery to extremes.

 

[peggy-sue]

No need to go any further than their assumed "cause" merely being a correlate.
I hope this prat does not get awarded his PhD - but if he does- more proof positive that psychology isn't science.

 

[oceanblue]

No need to go any further than their assumed "cause" merely being a correlate.
I hope this prat does not get awarded his PhD - but if he does- more proof positive that psychology isn't science.

The doctorate was awarded and the paper published so draw your own conclusions. The supervisor of the PhD and senior author on the paper was Professor Gijs Bleijenberg.

 

[oceanblue]

More detail on those 4 flaws (if anyone is still reading):

1. The study is not suitable to draw conclusions about causation
The big claim of this study is that it has been able to draw conclusions about what causes persistence of fatigue in CFS, because it uses Structural Equation Modelling (SEM):

the statistical technique we used, "structural equation modeling", allows to make causal inferences from such data...
...The strength of this study was that the model on the relationships between behavior, affect, cognitions, and fatigue was formulated in terms of cause and effect relationships and that an integral test of this model was performed.

This is plain wrong. There is no statistical technique on earth that can draw causal conclusions from cross-sectional data. You need at least 'before' and 'after' data, and preferably 'during' as well to have a chance of concluding that A causes B which causes C. Here's an explanation from a relevant recent paper: Structural Equation Modeling in Medical Research: A Primer (2010)

[using the example of a schizophrenia study]
Because this is a cross-sectional model, it is unknown whether the language processing deficit existed before, at the same time, or after the onset of schizophrenia. Direction of cause in the model is, thus, unknown.
...
[from Strengths and Weakensses section]
SEM has sometimes been referred to as causal modeling; however, caution must be taken when interpreting SEM results as such. Several conditions are deemed necessary, but not sufficient for causation to be determined. There must be an empirical association between the variables - they are significantly correlated. A common cause of the two variables has been ruled out, and the two variables have a theoretical connection. Also, one variable precedes the other, and if the preceding variable changes, the outcome variable also changes (and not vice versa). These requirements are unlikely to be satisfied; thus, causation cannot be definitively demonstrated. Rather, causal inferences are typically made from SEM results.

2. The sample of patients wasn't really suitable
The study used 51 patients diagnosed according ot Oxford Criteria, which only requires 6 months of fatigue and no other symptoms. Tellingly, 26% of the sample did not report daily fatigue. Which means that any small effects (and this study only found small-ish effects) could be explained by inclusion of inappropriate patients. Also, the sample size of 51 is rather small. Sample size in SEM is a controversial issue, but, as the primer above says:

To determine the models goodness-of-fit, sample size is an important consideration. It must be large enough to obtain stable estimates of the parameters... the suggestion that at least 100 but preferably 200 cases are needed to obtain stable results [16]. [16. Kline RB: Structural Equation Modeling New York: Guilford; 1998]

3. Several of the questionnaires used as 'measures' are not appropriate
"Somatization" was measured by the SCL-90 Somatization subscale, which actually simply asks about symptoms experienced, which might be due to CFS (eg weakness and soreness of muscles) and is not a proper measure of excessive symptom-focusing.

"Functional Impairment" was measured soley by the Sickness Impact Profile (SIP) Home Management subscale (10 questions). The exact question list is not available but the original SIP paper gives as examples: 'I am not doing any of the maintenance or repair work around the house that I usually do' and 'I am not doing heavy work around the house'. Later studies from the same group more reasonably used a broader measure of functional impairment including 'Social', 'Rrecreation' and 'Work' (25% of patients still worked); there are 27 questions in these extra categories, giving a broader and more robust measure.

"Physical Activity" was measured only by the SIP-Mobility subscale (10 questions, with none on walking, which is covered by the ambulation subscale). A study by the same authors on the same patients using actometers showed that SIP-mobility did not correlate well with activity measured by actometer (r=0.35).

Again, from the primer:

Care must be taken in using variables that provide a valid and reliable indicator of the constructs under study

4. The model only fitted the data after tweaking it 3 separate times
One of the strengths of a modelling approach is that the model is developed 'a priori' i.e. before the data is seen, then tested agains the data. Unfortunately, the hypothesised model failed badly. The data from the model test suggested modification i.e. dropping 5 of the original 9 'mediators' and adding 3 new ones. This radical change still didn't fit that data so they added another new mediator, which still didn't cut the mustard - so they added a 5th new mediator and finally got a fit. This is post-hoc tweaking in the extreme, with the strong possibility that the model has just been bent and twisted to fit that partiuclar group of patients. It would need replicating on an independent sample to have any meaning.

From the primer again:

Re-specify the Model if Necessary
To obtain improved fit results, the above sequence of steps is repeated until the most succinct model is derived (i.e., principle of parsimony)... Once the model is re-calculated, its fit may show improvement and residual may be reduced. These results then need to be confirmed on an alternate sample, and through further studies. This replication strengthens confidence in the inferences, and provides implications for theoretical development and practical application.

Conclusion
It's hard to understand how this study ever got published. The fatally flawed methodology was unsuitable to detect causal relationships, the sample was very small with a dubious diagnosis, the questionnaires were not suitable to measure what was claimed and the model was twisted and manipulated several times to get a 'fit' to the data.

One last comment from the primer:

SEM cannot correct for weaknesses inherent in any type of study... In addition, poor research planning, unreliable and invalid data, lack of theoretical guidance, and over interpretation of causal relationships can result in misleading conclusions.

How on earth could this study be used as the basis for the type of CBT promoted in the Netherlands for over a decade?

 

[Snow Leopard]

I don't mind the publishing of such a method, so long as the discussion/conclusion is reasonable (ie far more tentative than what they actually wrote).
The bigger question is the question in the end of your post - an excellent and perhaps alarming question.

 

[oceanblue]

Vercoulen model undone

Given the fundamental flaws in the methodology, it's not surprising that the model's findings haven't held up well in studies looking at how CBT/GET 'works', as this diagram shows:


So most of the proposed mechanisms haven't been supported by further studies. In particular, the large Wiborg 2010 study showed that CBT led to no change in physical activity measured by actometer (though self-reported activity did increase), while the Moss-Morris study showed that physical conditioning similarly did not improve with GET.

Significance of 'Sense of Control' finding?
The 'confirmatory' finding that 'sense of control' can affect fatigue less robust. However, a study of CBT for pain found a similar role for sense of control over symptoms, as did an earlier prospective study of CFS by Vercoulen. (n=298 [questionnaire validation], 18-month follow-up). Rather than 'Sense of Control' reflecting 'correct' illness beliefs it might be linked to appropriate adaptation to the illness - why else do people pace? i.e. The modest effect of 'Sense of Control' may simply be a finding that strategies like pacing are modestly helpful.


Diagram references:
Deale A, Chalder T, Wessely S. Illness beliefs and treatment outcome in chronic fatigue syndrome. 1998

Moss-Morris R, Sharon C, Tobin R, Baldi JC. A Randomized Controlled Graded Exercise Trial for Chronic Fatigue Syndrome: Outcomes and Mechanisms of Change. 2005

Wiborg JF, Knoop H, Stulemeijer M, Prins JB, Bleijenberg G. How does cognitive behaviour therapy reduce fatigue in patients with chronic fatigue syndrome? The role of physical activity. 2010

Wiborg JF, Knoop H, Frank LE, Bleijenberg G. Towards an evidence-based treatment model for cognitive behavioral interventions focusing on chronic fatigue syndrome. 2012

 

[Enid]

Well it was 1998 - worst possible time for this cabal off in "la-la" land. "Correct" illness belief seems to be their problem. (No excuses for rudeness and contempt having been told it's "all in your mind" 10 years ago, much in the way of pathologies discovered since).

Good to see your analysis here ocean.

 

[Esther12]

Nothing to add, but I just wanted to add a 'thanks' for the thread and analysis.

 

[oceanblue]

I don't mind the publishing of such a method, so long as the discussion/conclusion is reasonable (ie far more tentative than what they actually wrote).
The bigger question is the question in the end of your post - an excellent and perhaps alarming question.

That is indeed the question; not much chance of an answer from the Dutch school, though. You are more generous than I am: I don't mind SEM on before/after data like the 2012 Wiborg study I quote below, but I don't think any paper using SEM on cross-sectional data as a basis for causal claims merits publication.

@e12: thanks.

 

[roxie60]

Great another 'objective' opionion telling us if we would just stop focusing on our symptoms we would feel better...yada yada