The pathway from glandular fever to CFS: can the CBT model provide the map?

[oceanblue]

The odds ratio for gender (for CFS at 6 months) is 6.46 (95% C.I. 0.77-54.10) p=0.09.
For All-or-nothing behaviour it's 1.92 (1.07-3.39) p=0.03*

What they say, that gender isn't an issue, seems to me to be debatable. It looks numerically that it is likely to be a big factor (female gender increases risk a lot) but they don't have a big enough sample to be sure. While they can be more sure All-or-nothing behaviour is a factor but it's not as big a factor (odds ratio is lower so less than double the risk).

I'm a bit confused here with the gender stuff. Looking at the raw data, only 1/75 men developed CFS vs 16/142 (1 in 9) women: in a simple-minded way, surely that's got to be significant? The pattern could hardly be stronger (zero men with CFS, 1 in 9 women?).

The multivariate logistic regression analyses table gives an enormous SD for Gender, but isn't it possible to calculate an OR from the basic data? Or to put it another way, could their calculations be flawed? They don't give the raw data for all-or-nothing (numbers with/without A-o-N attitude who do/don't develop CFS), but I'd be very surprised if they showed a stronger pattern than that for gender. Is there any way of checking this from the data presented?

 

[Dolphin]

I'm a bit confused here with the gender stuff. Looking at the raw data, only 1/75 men developed CFS vs 16/142 (1 in 9) women: in a simple-minded way, surely that's got to be significant? The pattern could hardly be stronger (zero men with CFS, 1 in 9 women?).

The multivariate logistic regression analyses table gives an enormous SD for Gender, but isn't it possible to calculate an OR from the basic data? Or to put it another way, could their calculations be flawed? They don't give the raw data for all-or-nothing (numbers with/without A-o-N attitude who do/don't develop CFS), but I'd be very surprised if they showed a stronger pattern than that for gender. Is there any way of checking this from the data presented?

If one calculates the Odds Ratio for gender, it's 9.4.
We are told there is a significant difference with regard to gender.

I believe there could be situations with multivariate logistic regression where the variable with the highest odds ratio wouldn't remain.

In this case, only one item remains after multivariate logistic regression:

"a final single-item factor, all-or-nothing behaviour."

(that's factor 5)

"The five factor scores were entered as covariates
along with gender, age and GF symptoms (eight predictors
in total) into two separate logistic regression
equations, with CF and CFS outcomes as the dependent
variables. Table 4 shows that the first four
factors predicted the onset of CF at 3 months, while
only factor 5, all-or-nothing behaviour, predicted CFS
at 6 months. Age, gender and GF symptoms were not
significant predictors in these equations. Interestingly,
patients with fewer GF symptoms were more at risk of
CF symptoms at 3 months."

So, for CFS, according to the authors, only "all-or-nothing behaviour" predicted CFS at 6 months based on the multivariate logistic regression.

But in the abstract they say:

"Results. Of the participants, 9.4% met the criteria for CF at 3 months and 7.8% met the criteria for CFS at 6 months. Logistic regression revealed that factors proposed to predispose people to CFS including anxiety, depression, somatization and perfectionism were associated with new-onset CFS. Negative illness beliefs including perceiving GF to be a serious, distressing condition, that will last a long time and is uncontrollable, and responding to symptoms in an all-or-nothing behavioural pattern were also significant predictors. All-or-nothing behaviour was the most significant predictor of CFS at 6 months. Perceived stress and consistently limiting activity at the time of GF were not significantly
associated with CFS."

They don't point out that once one controls for "all-or-nothing", the others drop out - the others only showed up on individual logistic regression, after controlling for age, gender and glandular fever symptoms, which they don't mention!

So I can see your point oceanblue. But also in general, the abstract is misleading.

 

[lucy]

I do not understand the point of such 'research'. Logically, it seems they want to be able to identify people, who are not accepting the illness easily, as they think it leads to longer ilness. Let us say, they manage to identify them. Then what? Will they tell them 'Do not worry, you will be fine, there is no cure for glandular fever and 95 percent of population get over with it in their childhood, acquiring the immunity for life. You're are one of the few who had the virus reactivated and we do not know why. We also have convincing research, that if you do not care about it, you will heal faster'. Or, will they get a psychologist assigned, where patients attend a group and have to stand up and say "Hello, my name is Bob, I have mono, and I am worried about it".
In some other diseases hope and fighting strength is important or at least highlighted in our culture. So far I have seen that dignity and acceptance is valued only in the face of death in our society. Now it will be death, mono and CFS. We will hear saying "Oh, she faced CFS with such a dignity!".

 

[oceanblue]

'Odd' Odds Ratios

If one calculates the Odds Ratio for gender, it's 9.4.
We are told there is a significant difference with regard to gender.

I believe there could be situations with multivariate logistic regression where the variable with the highest odds ratio wouldn't remain.

In this case, only one item remains after multivariate logistic regression (that's factor 5)

Thanks for that. I'm beginnig to suspect their numbers. Seems strange that gender is significant in its own right but drops out in the multivariate analyisis (does this mean that all-or-nothing attitudes predict gender?).

Anyway, i had a chat with a friend who works with Odds Ratios in his work on genomics. He hasn't seen the data for this study but did comment that the problem with small studies like this is that they have very wide confidence intervals. Now, the CI for gender - using raw data, not the regression stuff is huge (1.22-73.21 at 95%, OR 9.4). But the 95% CI for the BRIQ All-or-nothing variable is tiny, OR 1.14 (1.02-1.26), see table 2, and this looks suspicious. Certainly my friend said he would only expect such a small CI in a very large study.

Then I played around with numbers using an online Odd Ratio with CI calculator, at all times using 17 cases, 200 non-cases total, but trying different hypothetical splits for All-or-nothing. This wasn't an exhaustive exercise, but every combination I tried gave very large CIs. According to one explanation I found,

it turns out that the smallest count in the two by two table plays the largest role in determining the size of the standard error. http://www.childrens-mercy.org/stats/model/confidence.asp

So I did lots of trials around small numbers in the Cases (from1/16 split to 8/9 split). Again, wide CIs.
Also, the OR given is only 1.1.4, which again suggests we are looking at a small effect so surely we would expect large confidence intervals in such a small trial?

I'd be interested to hear your views on this. Am I barking up the wrong tree or might the published stats be suspect? (my friend did comment that it wasn't that unusual to see mistakes in published OR statistics).

thanks

 

[oceanblue]

I'm beginnig to suspect their numbers...

Then I played around with numbers using an online Odd Ratio with CI calculator, at all times using 17 cases, 200 non-cases total, but trying different hypothetical splits for All-or-nothing. This wasn't an exhaustive exercise, but every combination I tried gave very large CIs.

Update: I've now done this exhaustively using all possible permutations for AoN, from 1-16 CFS cases, in each scenario adjusting the number of AoN non-cases to give an OR of close to the 1.14 quoted in the study. The best Confidence Interval I got was 0.42-3.13 (based on 7 AoN cases, 76 non-cases), compared with their CI of 1.02-1.26. If my calculations are right, then the finding would be non-significant. I've also looked at the OR and CIs in a few other studies of similar or larger size and they all much larger confidence intervals than in this study.

The confidence intervals quoted for BRIQ AoN in this study look wrong. If they are, maybe those for the other BRIQ factors are wrong too.

 

[Dolphin]

Hi Oceanblue,

Well done on keeping at it. In situations like this, I prefer journals which allow online comments/rapid responses which can be simply a few sentences or whatever.

You could try writing to the authors. Just a simple question about what was the breakdown of A-or-N. Like you say, it seems difficult to imagine in this situation that it could be correct. If you don’t get a reply from the corresponding author after a certain period, you could write to the other authors. Or alternatively write to them all together – one can probably find E-mail addresses for them with a quick search.
One could send a reminder after 4 (?) weeks.

If still no reply, you could write to the editor and see if he/she will look into it.

Just brainstorming here.

 

[Dolphin]

Original abstract in 2006 was different

New research form the biopsychosocial fan club:

The pathway from glandular fever to chronic fatigue syndrome: can the cognitive behavioural model provide the map?
by: R. Moss Morris, M. J. Spence, R. Hou

Psychological Medicine First View, 1-9. URL http://dx.doi.org/10.1017/S003329171000139X

Apparently, developing CFS after glandular fever is down to patient attitudes:

Abstract

Background: The cognitive behavioural model of chronic fatigue syndrome (CFS) suggests that the illness is caused through reciprocal interactions between physiology, cognition, emotion and behaviour. The purpose of this study was to investigate whether the psychological factors operationalized in this model could predict the onset of CFS following an acute episode of infectious mononucleosis commonly known as glandular fever (GF).

Method: A total of 246 patients with GF were recruited into this prospective cohort study. Standardized self-report measures of perceived stress, perfectionism, somatization, mood, illness beliefs and behaviour were completed at the time of their acute illness. Follow-up questionnaires determined the incidence of new-onset chronic fatigue (CF) at 3 months and CFS at 6 months post-infection.

Results: Of the participants, 9.4% met the criteria for CF at 3 months and 7.8% met the criteria for CFS at 6 months. Logistic regression revealed that factors proposed to predispose people to CFS including anxiety, depression, somatization and perfectionism were associated with new-onset CFS. Negative illness beliefs including perceiving GF to be a serious, distressing condition, that will last a long time and is uncontrollable, and responding to symptoms in an all-or-nothing behavioural pattern were also significant predictors. All-or-nothing behaviour was the most significant predictor of CFS at 6 months. Perceived stress and consistently limiting activity at the time of GF were not significantly associated with CFS.

Conclusions: The findings from this study provide support for the cognitive behavioural model and a good basis for developing prevention and early intervention strategies for CFS.

An interesting comparinson with the Dubbo Studies, which found that developing CFS was predicted by the severtity of the initial illness rather than psychological factors.

"Enjoy"

Oceanblue and anybody else interested:

I came across the following abstract (it was originally posted on the MEActionUK yahoogroup and I happened to be clearing some stuff out):

As one can see, the results section is a bit different.

http://neuroscience.mahidol.ac.th/9icbm-2006/

Ninth International Congress of Behavioral Medicine

Timeless - Thai Elegance - Amazing

Bridging Behaviour and Health

Connecting the Hemispheres

November 29th - Decenber 2nd 2006

Bangkok Thailand

9th ICBM-MANAGER


http://neuroscience.mahidol.ac.th/9icbm-2006/manager/index.php


Home / Somatoform Disorders/Chronic Fatigue/Medically Unexplained
Symptoms / Symposium

[..]

1st Abstract

Author Dr Rona Moss-Morris

Address School of Psychology, University of Southampton, Highfield,
Southampton SO17 1 BJ, United Kingdom of Great Britain and Northern
Ireland
Email remm @ soton.ac.uk
Tel +44 2380597549
Fax



Abstract Topic THE PATH FROM GLANDULAR FEVER TO CHRONIC FATIGUE
SYNDROME: CAN THE COGNITIVE BEHAVIOURAL MODEL PROVIDE THE MAP?


Authors Spence M1, Moss-Morris R*2 1Psychological Medicine,
University of Auckland, New Zealand; 2School of Psychology,
University of Southampton, UK



Body of abstract Purpose: The purpose of this study was to
investigate whether psychological variables operationalised from the
cognitive behavioural (CB) model of chronic fatigue syndrome (CFS)
could predict the development of CFS following an acute episode of
infectious mononucleosis.


Method: Participants included 246 primary care patients with
serological evidence of infectious mononucleosis, and no previous
history of CFS or related disorders. They completed an initial
questionnaire at the time of acute infection which included the
Hospital Anxiety and Depression scale, Perceived Stress Scale, the
Negative Perfectionism Scale, the Illness Perception Questionnaire-
Revised (IPQ-R), and the Behavioural Responses to Illness
Questionnaire (BRIQ).

Six months after the initial infection participants completed a
second questionnaire which was used to determine whether they met
published diagnostic criteria for CFS.

Results: 7.8% of the sample met diagnostic criteria for new onset
CFS six months post infection. Logistic regression controlling for
age, gender, and severity of glandular fever symptoms found that
those who went on to develop CFS had significantly higher levels of
negative perfectionism (p < .05), anxiety (p < .05), depression (p
< .01), and negative illness beliefs (p < .01) at the time of
infection than those who did not develop CFS.

Conclusions and reflections: High personal expectations, negative
mood and interpreting symptoms in a negative fashion were associated
with the onset of post viral CFS, but perceived stress and
behavioural responses were not.

These findings lend some support to the CB model of CFS and suggests
that early interventions which target both personal and illness
beliefs may help to prevent the onset of the illness. Future
prospective studies need to focus on the interaction between post
viral biological changes and psychological factors in the onset of
the illness.

Contributed support No
Conflict of interest No

 

[Dolphin]

Update: I've now done this exhaustively using all possible permutations for AoN, from 1-16 CFS cases, in each scenario adjusting the number of AoN non-cases to give an OR of close to the 1.14 quoted in the study. The best Confidence Interval I got was 0.42-3.13 (based on 7 AoN cases, 76 non-cases), compared with their CI of 1.02-1.26. If my calculations are right, then the finding would be non-significant. I've also looked at the OR and CIs in a few other studies of similar or larger size and they all much larger confidence intervals than in this study.

The confidence intervals quoted for BRIQ AoN in this study look wrong. If they are, maybe those for the other BRIQ factors are wrong too.

The numbers in the paper are the adjusted odds ratios. I am didn't get as far as studying logistic regression in college - I have worked out what it is but I don't know really know what happens too much when you start adjusting for three things (age, gender and glandular fever symptoms).

 

[oceanblue]

The original abstract was different.

Ah, well if it was first published in late 2006, that would explain why they didn't look at illness severity properly, since the Dubbo study linking acute illness severity to CFS in a similar prospective wasn't published unitl that year.

More on this later.

 

[oceanblue]

There is a worrying inconsistency in the main findings of the study

The study concludes that all the psychological variables they think predict CFS simplify into 5 different factors . Just one of these factors, "all-or-nothing" behaviour, predicts CFS. Unfortunately this factor doesn't predict CF, fatigue after 3 months, - but the other 4 factors do! So the factors predicting fatigue after 3 months are completely different from the factors predicting fatigue after 6 months. Oh.

What's really worring about this is that the 17 CFS cases (6 months) are presumably counted within the 21 cases of CF (3 months). This makes it look like dropping the 4 cases of CF who didn't make it to CFS completely changes the predictors of illness. Which is a sign that the model might not be very good.

And what would have predicted fatigue at 12 months or 2 years (both of which would count as CFS) - the same factors as at 6 months, or different factors again? We don't know because the research stopped after 6 months.

It's beginning to look like All-or-nothing behaviour might not be such a great predictor of CFS after all.

 

[richvank]

Hi, all.

If I developed glandular fever (mononucleosis) I would raise my intake of lysine relative to arginine and also take NAC and vitamin C. Taking Tagamet (cimetidine) is another thing I would likely try. I recently developed shingles, another disorder caused by a herpes-family virus, and treatment with lysine, NAC and vitamin C cleared it up in about a week.

I don't think anyone knows for sure why raising the lysine to arginine ratio helps to put herpes-family viruses back into latency, but it seems to do so. It has been used quite a lot for herpes simplex I (the cold sores or fever blisters virus), but seems to work for the other herpes-family viruses, also. I've heard suggested explanations in terms of stimulating micro RNA, affecting histones, or disrupting the viral protein synthesis, but I don't know which if any of these explanations is valid.

The vitamin C and NAC work by making the redox conditions more reducing, particularly through raising the concentration of reduced glutathione inside the cells where the viruses reside. Palamara et al. at the University of Rome showed a few years ago that this will prevent herpes simplex I from completing the formation of the disulfide bonds in glycoprotein B, which it needs to do to be able to move from one cell to another. All the human herpes-family viruses have this glycoprotein, so this should work for all of them.

The Life Extension Foundation magazine a few years ago explained that Tagamet, which is a histamine H2 receptor blocker that is sold over the counter in the U.S. now to lower stomach acid production, works by blocking the signal that the Epstein--Barr virus sends to immune cells to shut down their response.

Rather than studying the psychology of these unfortunate patients after they become chronically ill, I think it would be more productive to give them effective treatment early on, before the mononucleosis has a chance to produce chronic illness.

Best regards,

Rich

 

[oceanblue]

Illness severity & unreliability of many psychological measures used in this study

Probably the biggest flaw with this paper is that it fails to address the impact of the severity of Glandular Fever on both the risk of developing CFS and on the psychological variables.

The 2006 Dubbo study, a prospective study the same size as this, did look closely at the severity of the intial infection and concluded that the main factor predicting CFS was the severity of the inital infection (Glandular Fever, but also 2 different infections as well). So it's a pity this new study doesn't look properly at illness severity. Maybe it would have been a bigger predictor of developing CFS than psychological variables.

Just as importantly, the severity of illness could account for many of the changes seen in the psychological variables

The CBT study did attempt to measure illness severity by the COUNT of the number of GF symptoms. Yes, not the severity of fever, headache etc, the number of different symptoms so a lot of minor symptoms is registered as more severe illness than a few but very severe symptoms. As the authors admit: "In terms of the measure of GF severity, clinical examination and physiological measures may have provided more objective markers than simply the overall number of symptoms." Exactly. And what about asking the patient to grade the severity of their own symptoms?

The problem is the severity of Glandular Fever is likely to influence many of the 'psychological' variables, as theses 3 IPQ-R times show:

- Consequences (what impact patients believe the illness will have on their daily lives)
- Timeline (how long patients believe their illness will last)
- Emotional representations (perceived emotional impact of the illness)

'Consequences' is directly related to illness severity, while more severely ill patients are likely to score higher for both 'Timeline' and 'Emotional representations' as a the iller you are the more it will affect your life and the more you are likely to believe the illness will last longer than if you were mildly ill.

Remember, patients are completing the questionnaires early in the study while they probably sitll have the acute infection.

If the study had a valid measure of illness severity, it could use statistical analysis to see if illness severity correlates with these psychological measures and adjust the figures accordingly. But because they didn't measure severity properly they can't do this, and so we can't rely on the data they present.

 

[Dolphin]

Looks like you have got to the heart of it, Oceanblue.

I have read at least one other study by Moss-Morris (and Petrie) that used the IPQ-R that I was not happy with.
They were also the crowd that started the theory that people with ME/CFS catastrophize - I can't recall the details now but I think the data could have been interpreted in other ways.

 

[oceanblue]

Looks like you have got to the heart of it, Oceanblue.

Thanks, Dolphin. There's more too, I'm afraid.

 

[oceanblue]

More problems with their variables: Anxiety and Depression

HADS Anxiety and Depression scores

The problem is that this questionnaire is being used to assess anxiety and depression in the month BEFORE the patients got ill. This leads to recall bias as the patient is being questioned after the event, and more importantly probably while they are still ill with Glandular Fever. In healthy subjects you would expect some recall bias, but for ill patients it's quite likely that patients who are more severely ill will overreport anxiety and depression simply because they are feeling lousy and that colours their memory. So recall bias is likely to make these values unreliable.

 

[oceanblue]

More problems: "Non-Glandular Fever somatic symptoms "

Unfortunately this is a variable that was basically made up by the authors with no validation provided and so this probably shouldn't be included in the study at all. The authors wrote:

"A number of non-specific symptoms (e.g. sore eyes, loss of strength, dizziness, racing heart beat) [were recorded] as a measure of general somatization".

I think this means imagined symptoms rather than real symptoms as a result of Glandular Fever. The problem with this is that the authors don't provide any evidence that these really are somatic symptoms in the case of GF, or that is realy is a valid "measure of somatization". What's notable in their data is that most patients appear to have both 'GF' and 'Somatic' symptoms, suggesting that the distinctioin is arbritary. To include Non-GF symptoms in the model requires a leap of faith, or as the authors put it, an assumption:

"Non-GF somatic symptoms were included as a psychological risk factor rather than a control variable as they were assumed to be independent of GF."

And that's the point: the authors offer no evidence that Non-GF symptoms are independent of GF. So I can't see how they can include it in their model.

 

[oceanblue]

The analysis assumes a random sample and accurate diagnosis: both are suspect

All the statistical analyis rests on these 217 participants being a random sample. The problem is that the intial selection was of 737 people with a lab diagnosis for Glandular Fever; only 35% (260) of the original selection provided questionnaires. With 65% of the selection effectively 'lost', it's entirely possible that those who made it into the study are not properly representative of the original selection.

2. The "CFS" cases were diagnosed by Questionnaire ie without lab test or a clinician's assessment, so they haven't been properly diagnosed. The rate of CFS cases found at 8% after 6 months is in line with other prospective studies of Glandular Fever, though it's possible that a clinical analysis would have excluded some of the cases and included some of the non-cases, and that could change the results. We'll never know.

The low inclusion rate of participants, giving plenty of scope for a non-random sample, and lack of robust CFS diagnosis casts a shadow over the study that calls into question any marginal findings (e.g. small difference, or only just significant).

To compensate, perhaps the required confidence level for significance should be increased from 95% to, say, 99% (interestingly, this would render most of their findings insignigicant).

In any event, this is a small study; all the conclusions are based on how the 17 'CFS' cases differed from the 200 non-cases (217 total that completed the whole study). Additional and larger studies would be needed to confirm any findings.

 

[Dolphin]

HADS Anxiety and Depression scores

The problem is that this questionnaire is being used to assess anxiety and depression in the month BEFORE the patients got ill. This leads to recall bias as the patient is being questioned after the event, and more importantly probably while they are still ill with Glandular Fever. In healthy subjects you would expect some recall bias, but for ill patients it's quite likely that patients who are more severely ill will overreport anxiety and depression simply because they are feeling lousy and that colours their memory. So recall bias is likely to make these values unreliable.

This is interesting and seems perfectly plausible. I wonder is there any general research on this area. I know since I became ill, I've recalled all sorts of relatively minor negative-ish things that happened to me before I became ill - things that I had completed put out of my mind. I had a relatively carefree existence so you are talking about minor things like people telling me I was pronouncing their name wrong. Also, I have heard it said that their can be a prodormal period with some infections and conditions where people are starting to get ill without fully being ill.

 

[oceanblue]

This is interesting and seems perfectly plausible. I wonder is there any general research on this area

I can't find anything specific on HADS, but recall bias is often mentioned in papers as an issue so I'm guessing it's not particularly contentious to raise the issue.

 

[oceanblue]

I have read at least one other study by Moss-Morris (and Petrie) that used the IPQ-R that I was not happy with.
They were also the crowd that started the theory that people with ME/CFS catastrophize - I can't recall the details now but I think the data could have been interpreted in other ways.

I think the whole thing with the IPQ-R is rather dodgy, mainly because it asseses patients views without relating them to reality eg how treatable an illness really is, how much disability it really causes. It takes us to a fantasy land where the illness itself has no bearing on the experiences of the patient.