Wessely studies from the 90s that keep cropping up

[oceanblue]

Paper that makes the case for a psychological explanation for CFS (and its flaws)

Psychological symptoms, somatic symptoms, and psychiatric disorder in chronic fatigue and chronic fatigue syndrome: a prospective study in the primary care setting, 1996 (pdf).

The study may be 15 years old but Simon Wessely and Peter White still like to quote its findings to support a psychological explanation for CFS.

This paper strongy links CFS with psychiatric disorders, implies that both fatigue and symptoms of the illness are primarily due to psychiatric disorder, and suggests that CFS is not a specific illness, just one end of a spectrum of chronic fatigue. Essentially, it says that both fatigue and psychosomatic symptoms accumulate with psychogical morbidity until 'hey presto', some people cross an arbitrary threshold and are defined as having CFS.

Abstract Highlights
Subjects with chronic fatigue were at greater risk than those without chronic fatigue for current psychiatric disorder assessed by standardized interview (60% versus 19%)and were more likely to have experienced psychiatric disorder in the past.

Most subjects with CFS also had current psychiatric disorder when assessed by interview (75%). Both the prevalence and incidence of chronic fatigue syndrome were associated with measures of previous psychiatric disorder. The number of symptoms suggested as characteristics of chronic fatigue syndrome was closely related to the total number of somatic symptoms and to measures of psychiatric disorder.

CONCLUSIONS: Most subjects with chronic fatigue or chronic fatigue syndrome in primary care also meet criteria for a current psychiatric disorder. The symptoms thought to represent a specific process in chronic fatigue syndrome may be related to the joint experience of somatic and psychological distress.

Read the full abstract

Method is as described in post #7 and note that 'somatic symptoms', including CDC-94 symptoms, are assessed too.

The fundamental flaw in this study is that while CFS is defined by unexplained fatigue, the paper fails to distinguish betweeen fatigue that is unexplained and fatigue that is explained. In fact:

nearly all of the subjects recruited in primary care did not present with fatigue as the principal complaint.

Since fatigue is common to many illnesses, both physical and psychological, it seems likely that in many cases the principal complaint would explain the chronic fatigue. For example, depression is common in primary care setting and one of the main symptoms of depression is fatigue. In contrast to other studies, only 1.5% of chronic fatigue patients were excluded for explained fatigue, which might be because they weren't given a proper clinical evaluation and because the authors don't seem to consider any psychological disorder an explanation for fatigue.

The authors say:

we have confirmed previous community and primary care reports of strong associations between chronic fatigue and... psychiatric disorders.

And that's the problem. This study tells us little about CFS where the fatigue has to be unexplained. The strong correlation between fatigue and psychological morbidity does not explain CFS. I will post later in more detail on the correlation between symptoms and 'psychological morbidity'. However, the main point is that the correlation might be interesting if it was for those with unexplained fatigue, but without that qualificiation it tells us little.

CFS cases
The study also finds a strong link between psychological morbidity and full-blown CFS. To show a causal link they point out that 5 out of the 6 NEW cases of CFS (ie patients who were not fatigued at the start of the study) had previously had a psychiatric diagnosis. Similarly, 5 out of 6 new CFS cases had previously been prescribed psychoactive medication. This can be seen as further evidence of a strong link between pyschological morbidity and CFS. Alternatively, it can be seen as further evidence of poor diagnosis leading to inclusion of too many patients with psychologically caused fatigue and a false prevalence rate of 2.5%.

Conclusion
Failure to properly define the patients by excluding cases of explained chronic fatigue makes the findings of this study unreliable. The paper merely confirms the known link between chronic fatigue and psychological morbidity, while throwing little light on the nature of CFS.

 

[Dolphin]

1. CFS was primarily diagnosed via the 'Chronic Fatigue Questionnaire'; this unpublished, unvalidated questionnaire came from Trudie Chalder's 1990 MSc Thesis. The study fieldwork for all these papers was carried out in 92/93 - yet the Fukuda/CDC-94 criteria the study uses weren't published until 1994 (obviously). Did they use time travel to make sure they had the right criteria for their fieldwork? I emailed Trudie Chalder to request a copy of her questionnaire but as yet have received no reply..

Well spotted about when they started and how the questionnaire might then not be suitable.

2. They refer to 9 (or more, Wessely 1996) CDC symptoms when there are only 8.

If you include fatigue, then you can say there are 9 symptoms in the Fukuda criteria. Although the graphs actually say 9+ which is confusing (unless mental and physical fatigue were split up - I don't see that; or more of the questions they set they felt were like Fukuda symptoms that they counted them as counting - although again none jump out of me that would qualify)

Table 5 of this paper list 9 symptoms first that appear to be their CDC ones yet includes 'muscle weakness' and 'fever/chills' that are NOT CDC symptoms, but misses out 'sore throats' that is a CDC symptom. They may have diagosed some case of CFS on the basis of the wrong symptoms.

"Sore Throat" is in Table 5. I'm not sure we can read too much into the order.

For example, perhaps they pasted the symptoms from the 1988 (Holmes et al.) criteria in first:

Holmes et al. (1988)

Minor:

6 or more of the following symptom criteria.

.. Mild fever
.. Sore throat
.. Painful lymph nodes
.. Unexplained muscle weakness
.. Myalgia
.. Prolonged fatigue (>24 hr) after exercise
.. Headaches
.. Migratory arthralgias
.. Neuropsychologic symptoms
.. Sleep disturbance
.. Onset of symptom complex from a few hours to a few days.

And 2 or more of the following physical criteria:

.. Low grade fever
.. Nonexudative pharyngitis
.. Palpable or tender lymph nodes.

 

[Dolphin]

The authors say:
And that's the problem. This study tells us little about CFS where the fatigue has to be unexplained. The strong correlation between fatigue and psychological morbidity does not explain CFS. I will post later in more detail on the correlation between symptoms and 'psychological morbidity'. However, the main point is that the correlation might be interesting if it was for those with unexplained fatigue, but without that qualificiation it tells us little.

Good point. Look at people who have headaches on a Sunday morning (say) and you might find a strong correlation with how much they had to drink the previous day even though that might have no relevance for some of the cases. But usually in medicine, you would exclude such a group with explained headaches; or alternatively one would look at them specifically e.g. looking for hangover cures. Lumping them altogether is probably not going to tell you very much.

CFS cases
The study also finds a strong link between psychological morbidity and full-blown CFS. To show a causal link they point out that 5 out of the 6 NEW cases of CFS (ie patients who were not fatigued at the start of the study) had previously had a psychiatric diagnosis. Similarly, 5 out of 6 new CFS cases had previously been prescribed psychoactive medication. This can be seen as further evidence of a strong link between pyschological morbidity and CFS. Alternatively, it can be seen as further evidence of poor diagnosis leading to inclusion of too many patients with psychologically caused fatigue and a false prevalence rate of 2.5%.

As well as the fact that we can't be sure they are all proper new cases, I imagine - they could have been undiagnosed, more mildly affected cases for example (who maybe had a relapse) - being undiagnosed would put a psychological strain on most people.

Put another way, it could be a bit like "regression to the mean": if you look at all people who have CFS, at any one time some might be very mildly affected (here: not satisfy the definition for fatigue caseness) - however, there's a good chance some of them might experience a worsening of their fatigue and then become a "case" and be seen as the authors as a new case. As I said before (maybe in this thread?) we need to be cautious about whether people were truly new onset cases until easy tests become available and they are also actually used.

 

[oceanblue]

If you include fatigue, then you can say there are 9 symptoms in the Fukuda criteria. Although the graphs actually say 9+ which is confusing (unless mental and physical fatigue were split up - I don't see that; or more of the questions they set they felt were like Fukuda symptoms that they counted them as counting - although again none jump out of me that would qualify)

"Sore Throat" is in Table 5. I'm not sure we can read too much into the order.

Thanks for looking at this. Let me clarify:
If you look at Table 5 the first 9 symptoms are ranked in descending prevalence order for non-fatigued patients (47% down to 7%, column 1). The remaining symptoms are again ranked in descending order of prevalence from 41% down to 3% (column 1 again). You would expect the CDC symptoms to be listed together and the real ones almost are - but not quite. The alternative would be a more random distribution of symptoms and we don't have that. The Holmes list would have included sore throat with the other symptoms. Agreed, it's not conclusive but I'd be interested in another explanation for the ranking used in col 1.

That's a very interesting point about 'regression to the mean' ie CFS patients's fatigue being sub-threshold when they were assessed at the start of the study. Especially as the study didn't take a full medical history that might have more accurately identified the start of the illness.

 

[oceanblue]

Correlation between CDC symptoms and psychological morbidity

Simon Wessely and Peter White are fond of dispalying a graph that purports to show a strong correlation between how psychologically ill someone is and how many CDC symptoms they have. CDC symptoms increase with psychological morbidity.

For now I will put to one side the fact that the data is based on patients who may well have fatigue explained by psychological illness.

The study shows there is a strong correlation (r=0.53) between psychological morbidity as measured by the CIS-R interview and CDC symptoms for patients without chronic fatigue. However, it's interesting that for patients with chronic fatigue the correlation is only moderate (r=0.41), which isn't so impressive.

Yet this overstates the correlation because - as first noted by Dolphin - several questions of the CIS-R would be affected by CDC symptoms:

• Somatic Symptoms - max 4 points
  This relates to aches and pains and 6 CDC symptoms could score here: headaches, myalgia, joint pain, sore throat, sore glands, post-exertional malaise
• 'Concentration and forgetfulness' (CDC impaired memory/concentration)- max 4 points
• Sleep problems (CDC sleep disturbance)- max 4 points

So CDC symptoms could directly add up to 12 points to the CIS-R score. The average CIS-R score for those with 4+ CDC symptoms ranges from 12 points (5 symptoms) to 21 points (9+ symptoms), so the CIS-R scores directly attributable to symptoms could play a large role in the correlation.

All this despite the authors claming "we... used questionnaires that avoided the somatic symptoms associated with psychiatric disorder and chronic fatigue syndrome.", which is clearly not the case.

To put it simpy, one reason CIS-R scores correlate with CDC symptoms is that the CIS-R awards points for CDC symptoms.

So we start with a large correlation between CIS-R and CDC symptoms for those without chronic fatigue. The correlation weakens substantially for those with chronic fatigue. Because CIS-R scores points for CDC symptoms the true correlation is even weaker, probably only a small correlation. This is nothing to make a song and dance about, though Wessely and White do.

The fact that this correlation is based on all types of chronic fatigue, including that due to psychological causes, means that even this weak correlation has rather limited meaning.

 

[oceanblue]

If you include fatigue, then you can say there are 9 symptoms in the Fukuda criteria.

True, but that's not the normal way of counting symptoms, and fatigue isn't listed on the symptom list in table 5.

Fig 3 shows data for 189 individuals, presumably all those with fatigue who completed the symptom checklist. If fatigue counts as a symptom then the minimum symptom count for any of these is 1, not zero. Yet the bottom line of the figure shows one patient scoring 0, suggesting fatigue is not being counted as a symptom.

 

[Dolphin]

True, but that's not the normal way of counting symptoms, and fatigue isn't listed on the symptom list in table 5.

Fig 3 shows data for 189 individuals, presumably all those with fatigue who completed the symptom checklist. If fatigue counts as a symptom then the minimum symptom count for any of these is 1, not zero. Yet the bottom line of the figure shows one patient scoring 0, suggesting fatigue is not being counted as a symptom.

I've seen now that Figure 2 relates to those with "chronic fatigue" so fatigue isn't a symptom. By the way, in the PACE Trial protocol, one of the secondary outcome measures was:
"An operationalised Likert scale of the nine CDC symptoms of CFS" - so at least occasionally, one sees the Fukuda criteria being described as having 9 symptoms.

Anyway, I think I've found how there can be 9+ symptoms: The graph is looking at the Holmes symptoms:

we found that the symptoms included in the
original CDC definition of chronic fatigue syndrome
(18) were associated with psychological symptoms,
again in both the subjects with chronic fatigue (r=0.41,
N=l85, p<O.OOl) (figure 2)

18. Holmes G, Kaplan J, Gantz N, Komaroff A, Schonberger L,
Straus 5, Jones J, Dubois R, Cunningham-Rundles C, Pahwa J,
Tosato G, Zegans L, Purtilo D, Brown N, Schooley R, Brus I:
Chronic fatigue syndrome: a working case definition. Ann Intern
Med 1988; 108:387-389

Presumably Figure 3, which is talked about just below Figure 2 in the text, relates to the Holmes criteria CFS symptoms also.

 

[Dolphin]

Simon Wessely and Peter White are fond of dispalying a graph that purports to show a strong correlation between how psychologically ill someone is and how many CDC symptoms they have. CDC symptoms increase with psychological morbidity.

For now I will put to one side the fact that the data is based on patients who may well have fatigue explained by psychological illness.

The study shows there is a strong correlation (r=0.53) between psychological morbidity as measured by the CIS-R interview and CDC symptoms for patients without chronic fatigue. However, it's interesting that for patients with chronic fatigue the correlation is only moderate (r=0.41), which isn't so impressive.

Yet this overstates the correlation because 3 questions of the CIS-R would be affected by CDC symptoms:

• Somatic Symptoms - max 4 points
  This relates to aches and pains and 6 CDC symptoms could score here: headaches, myalgia, joint pain, sore throat, sore glands, post-exertional malaise
• 'Concentration and forgetfulness' (CDC impaired memory/concentration)- max 4 points
• Sleep problems (CDC sleep disturbance)- max 4 points

So CDC symptoms could directly add up to 12 points to the CIS-R score. The average CIS-R score for those with 4+ CDC symptoms ranges from 12 points (5 symptoms) to 21 points (9+ symptoms), so the CIS-R scores directly attributable to symptoms could play a large role in the correlation.

All this despite the authors claming "we... used questionnaires that avoided the somatic symptoms associated with psychiatric disorder and chronic fatigue syndrome.", which is clearly not the case.

To put it simpy, one reason CIS-R scores correlate with CDC symptoms is that the CIS-R awards points for CDC symptoms.

So we start with a large correlation between CIS-R and CDC symptoms for those without chronic fatigue. The correlation weakens substantially for those with chronic fatigue. Because CIS-R scores points for CDC symptoms the true correlation is even weaker, probably only a small correlation. This is nothing to make a song and dance about, though Wessely and White do.

The fact that this correlation is based on all types of chronic fatigue, including that due to psychological causes, means that even this weak correlation has rather limited meaning.

Well said. (I'm saying that as I've said something similar myself before - but it needs to be re-iterated regularly).

 

[oceanblue]

Thanks, dolphin - I like to think I built on your earlier excellent work

And I see what you mean about CDC symptoms referring to the superceded Holmes definition rather than the CDC-94 definition they used to define CFS cases. I wonder if they did that because Holmes symptoms gave them a stronger correlation with CIS-R than using CDC-94?

 

[Dolphin]

Thanks, dolphin - I like to think I built on your earlier excellent work

Just re-reading, I don't believe I noticed the correlation getting weaker for the "chronic fatigue" group anyway.

And I see what you mean about CDC symptoms referring to the superceded Holmes definition rather than the CDC-94 definition they used to define CFS cases. I wonder if they did that because Holmes symptoms gave them a stronger correlation with CIS-R than using CDC-94?

Yes, seems plausible that they cherry-picked like that.

 

[oceanblue]

Letter from Professor Michael Bland criticising Pawlikowska statistics

Thanks to Angela Kennedy, who mentioned this on another thread

Wow, a CFS paper getting some serious scrutiny, and from Professor Michael Bland, who co-authors that BMJ's famous 'Statistics Notes' series: Fatigue and psychological distress - Statistics are improbable.

He points out (in 1999) that the means given in the 1994 Pawlikoska paper are plain wrong. He says

But the quoting of impossible means should be enough to show that this paper is flawed.Why has nobody noticed, in refereeing, editing, reading the paper (several authors have cited it but seem to accept it uncritically)?

It's probable that the issue with the means appears was down to a production error, in that a reviewer of the original papers states the correct means were given in the review copies but nontheless it was sloppy by the authors to miss this, and by those that later uncritically cited this study.

Chalder and Wessely then give the correct means in a reply to Bland

The mean Likert score for fatigue was 13.72 (95% confidence interval 13.65 to 13.79); in men 13.13 (13.03 to 13.24); in women 14.16 (14.06 to 14.25).

Nothing else in the paper has changed in any way, including the figures, and the conclusions are unaltered.

These figures tie in with my calculations from the given data of 13.79.

However, Bland makes further criticisms of the study statistics:

There are several more subtle statistical problems: the histograms with unequal interval sizes shown as the same length on the graph; the statement that with such large numbers the distributions of responses to the fatigue and the general health questionnaires follow a normal distribution (the shape of the distribution is not related to the sample size); the ignoring of the cluster sampling; the use of two different scoring systems for the questionnaires.

Maybe not earth shattering, but it is nice to see that just occaisionally there's some proper scrutiny of CFS papers that underpin psychoogical explanations of the illness.

 

[Enid]

Yes very nice to see - thanks oceanblue.

 

[oceanblue]

Thanks, Enid, but with hindsight I realise I've lost it - one informed critical letter 12 years ago and I'm like an excited puppy. I suppose that shows just how little proper scrutiny these papers normally get.

 

[Battery Muncher]

I also find this kind of thing exciting. It seems that any publicity is good publicity really; it gets people thinking about ME, and it gets them interested in reading the research behind the controversies. This, I hope, will eventually lead to higher levels of scrutiny although, having said that, the scientific establishment is as conservative as any on earth. I don't expect major change any time soon.

And of course, as you say, must not go overboard. There has been no fundamental change in attitudes so far. But over the past few months a seed of doubt seems to have been planted in certain scientist's minds, and hopefully the more open minded are really considering the possibility that ME/CFS research is flawed.

 

[Astrocyte]

Hi Oceanblue,
Like the vision of you as an excited puppy. Puppies tend to wee on the carpet when things get really exciting but I'm guessing that you dont have a problem with that.

I wonder whether the scientific community haven't bothered critiquing ME psychological research cos they dont actually consider psychology proper science and therefore psychology papers are not worth bothering with?

 

[oceanblue]

Hi Oceanblue,
Like the vision of you as an excited puppy. Puppies tend to wee on the carpet when things get really exciting but I'm guessing that you dont have a problem with that.

I wonder whether the scientific community haven't bothered critiquing ME psychological research cos they dont actually consider psychology proper science and therefore psychology papers are not worth bothering with?

I was pretty excited about that letter but I'm relieved to report there were no accidents this time.

You could be right about why psychological papers are not critiqued. The people who should be interested in critiquing them are fellow psychology researchers; I'm not sure if they lack the inclination or the skills. On another thread, someone quoted a book about virology research which stated that most participants apparently regard it as a 'combat sport' - savage criticism is the norm. Be nice to see some of that in this field too.

Welcome to the forum.

 

[Dolphin]

I'm not sure if the post by the reviewer has been linked to:
http://www.bmj.com/content/320/7245/1343.full?sid=19f252e7-380a-4b97-9f3f-0470da038685
(if that link doesn't so up, search the BMJ site for articles with author Pelosi in 2000)

the reviewer shows that the gremlins might have attacked on several fronts
Anthony Pelosi, consultant psychiatrist
+ Author Affiliations

Department of Behavioural Sciences in Medicine, POB 1111 Blindern, 0317 Oslo, Norway
Academic Department of Psychological Medicine, Guy's, King's and St Thomas's School of Medicine and Institute of Psychiatry, London SE5 8AF
Hairmyres Hospital, East Kilbride G75 8RG
BMJ
EDITORBland found that the analysis was flawed in a paper on fatigue and psychological distress that had been published by the BMJ in 1994.1 2 He quite properly asks why nobody noticed this at the refereeing stage. The authors are unable to account for their errors and have tried, as they say, totally ungallantly to transfer the blame to the BMJ.3 They also wonder how the referee could have failed to detect the mistakes.

I refereed the manuscript for that paper. I think I can explain the most serious discrepancy identified by Bland. The authors originally coded the four possible responses in each item of their main questionnaires as 1 to 4. Therefore total scores for the general health questionnaire ranged from 12 to 48 and those for the fatigue questionnaire from 11 to 44. I told them they had to code the responses as 0 to 3. It looks as if they gave the correct scores when describing the total sample but did not recode the responses when examining males and females separately. I made various other suggestions, but I was not going to rewrite the results section of the paper and I certainly draw the line at proof reading.

I remember this manuscript well for several reasons. At first I could make neither head nor tail of some of the analyses. I ended up being angry with myself for spending too much timeon a sunny Sunday afternoontrying to understand the results instead of just recommending rejection. My irritation melted away when I got a note back thanking me for an excellent referee's report. At Christmas I received my first ever invitation to the BMJ's friends of the journal party, and I have always believed it was because of the high standard of that report.

May I take this opportunity gently to point out to the editors at the BMJ that they have never again invited me to their Christmas party. Am I doing anything wrong?

References
1.? Bland M. Fatigue and psychological distress. BMJ 2000; 320: 515. (19 February.)
2.? Pawlikowska T, Chalder T, Hirsch SR, Wallace P, Wright DJM, Wessely SC. Population based study of fatigue and psychological distress. BMJ 1994; 308: 763766.
3.? Chalder T, Wessely S. Fatigue and psychological distress. BMJ 2000; 320: 515. (19 February.)

Anthony Pelosi has said some forthright things about ME and CFS over the years (perhaps particularly in the 1990s - I don't remember him much from the 2000s). He would be of the "Wessely school of thought". One has to wonder if his decision as a reviewer might have been different with a different group of individuals.

 

[oceanblue]

Thanks, Dolphin, I hadn't seen that concluding correspondence. And how nice that the BMJ editor decided to invite everyone to dinner to celebrate their conclusion that it didn't matter after all.

No one ever mentioned again the other errors that Bland pointed out in his initial critique

There are several more subtle statistical problems: the histograms with unequal interval sizes shown as the same length on the graph; the statement that with such large numbers the distributions of responses to the fatigue and the general health questionnaires follow a normal distribution (the shape of the distribution is not related to the sample size); the ignoring of the cluster sampling; the use of two different scoring systems for the questionnaires.

Maybe that's why PACE felt at liberty to treat the CFQ as a normal distribution (it isn't, so 'mean-SD' should not be applied as they did, to define 'normal'). PACE didn't cite this Pawlikowska study, but did cite the 2010 Cella study which was based on the same underlying data and similarly was not normally distributed.

 

[Esther12]

Thanks OB. I was just looking at some Wessely slides that had him presenting some of the data you picked apart here.

It's a shame that we often don't get transcripts, so it's hard to bust them over it, as they may have been saying: "Here is a misleading graph that perfectly illustrates why CFS research needs to be viewed cautiously..."