1946 Birth Cohort. Etiology of CFS: Testing Popular Hypotheses (2008)

[oceanblue]

One of two papers on sef-reported CFS in the 1946 Birth Cohort

Etiology of Chronic Fatigue Syndrome: Testing Popular Hypotheses Using a National Birth Cohort Study
Harvey, Wadsworth, Wessely & Hotopf 2008

It concludes:

Individuals who exercise frequently are more likely to report a diagnosis of CFS in later life. This may be due to the direct effects of this behavior or associated personality factors. Continuing to be active despite increasing fatigue may be a crucial step in the development of CFS.

but not for the first time these conclusions are not very robust.

 

[oceanblue]

Commentary

Method
This survery is based on a random sample of 5,362 single births in one week of March, 1946. In 1999, at 53 years, the 3,035 remaining participants were asked if they had ever had a diagnosis of CFS/ME with 2,983 giving an answer. This makes it about 1/3 the size of the other 2 cohorts, which might explain the unreliable findings.

CFS prevalence
There were only 37 cases of CFS reported, and of these only 22 (65%) reported the diagnosis had been made by a doctor. Hospital records (primary care records were not checked) ruled out 3 of the 22 cases due to exlusionary medical conditions leaving 34 cases in total, giving a lifetime prevalence of 1.2% (ie people ever having CFS, no data was given for those with a current diagnosis). Age of onset was between 41 and 53 years ie the first case started in 1987, which ties in with the time CFS/ME was first being recognised. No doubt many particpants had CFS before this time but didn't know that's what it was. The same pattern of cases not begining until the 80s in seen in the other 2 birth cohorts too.

Ruling out hypotheses
This paper aimed to test a number of hypotheses, most of which were ruled out, including:

• Atopy (allergy)
• Specific childhood symptoms (persistent abdominal pain had been shown to be a risk factor for unexplained hospitalisations and the authors hoped it would be linked to CFS too; it wasn't)
• Childhood Chronic Illness (age 0-15). This contradicts the finding of the much larger 1970 cohort. More to the point, this 1946 study lost a lot of those with childhood chronic illness making it an unrepresentative sample and so these findings are pretty meaningless:
  

  Those followed up to the end of our study were not totally representative of the sample taken at the beginning. In particular, males, those from poorer families, and those who suffered ill health as children were less likely to be followed up. This attrition bias must be considered when assessing our results, especially in regard to our negative findings relating to childhood illness experiences

Frequent exercise a risk factor for CFS... ....or is it?
The biggest finding of this study was that exercising frequently is a significant risk factor for CFS. A later and large study contradicts this finding but here are the results (full table):

Extreme Energy levels reported by teacher (age 13) significant, p=0.006
Sports more than weekly (age 31) NOT significant
Self-rated as very fit (age 31) NOT significant
Very active at age 36 NOT significant
Sport weekly or more at 41 significant, p=0.006
Vigorous activity at least weekly at 31 and 43 significant, p=0.01, OR=10.8 (2.7-43.8)

Let's just look at that last finding - an odds ratio of 10.8? Woohoo! This must be the answer. But what the authors don't mention is that this figure is based on under half (43%) of all participants ie the data is missing for most of the participants (and the finding is based on just 4 energetic people with self-reported CFS), which makes it rather meaningless. This leaves me wondering if the reviewers of this paper were lazy, useless or biased, but this slightly harsh view may be coloured by Wales going out of the Rugby World Cup today ("it was never a red card, ref!").

Other than that, it's a mixed bag of significant and non-signigicant findings. The extreme energy at age 13 only led to 5 excess cases and would only explain 17% of all CFS cases, suggesting it's a minor factor.

More to the point, these findings were not replicated in the much bigger 1958 cohort, which found that high levels of activity were not a risk factor for CFS at any age.

Main conclusions
The findings on childhood illness are, according to the authors, questionable because some children with ill health were lost from the adult survey. The findings on exercise were mixed, based on small numbers and were not replicated in the larger 1958 cohort. It doesn't amount to much.

 

[justy]

Thanks Ocean blue for posting this ive been following it with interest.

Why oh why didnt they just put him in the sin bin for 10 minutes - arrgggghhh!
Wales played like heroes.

 

[Enid]

Played far better than we did - off and on the field !

 

[oceanblue]

Thanks, Justy and Enid. Yup, 10 minutes in the bin would have been fair, but I still think we should have sneaked a win even one man down; oh, those missed kicks... I don't fancy France's chances in the final next week.

There is more commentary to come (on the paper, not the game) but I'm a tad tired at the moment.

 

[Esther12]

Individuals who exercise frequently are more likely to report a diagnosis of CFS in later life. This may be due to the direct effects of this behavior or associated personality factors. Continuing to be active despite increasing fatigue may be a crucial step in the development of CFS.

I wonder if they've thought about the harm caused by earlier claims of CFS being the result of deconditioning on the psychosocial setting faced by patients?

When I was still feeling exhausted after glandular fever I was told by my GP that it was deconditioning, and that, while it will feel like a lot of work, it's important that I push through to maintain a normal level of activity. At the time, my fatigue was taken as evidence of a lack of motivation, now it's being seen as evidence of a dysfunctional personality which pushes too hard. I really hate the psychosocial quacks that surround CFS.

 

[oceanblue]

I've now updated my post #2 above.

Esther12: yes, it does sometimes seem that psychosocial proponents start from the position that the patients cause their own problems and then come up with different -even contradictory - arguments to suit different situations. However, the findings of this paper that patients are pushing on regardless are - like the rest of the findings - not very robust:

Evidence of prodromal fatigue
Interestingly, this study provides some evidence of rising fatigue levels before the onset of CFS. The psychological questionnaires used at age 36 and 43 included questions related to energy levels and fatigue and the answers to these questionnaires suggested prodromal fatigue:

However, by the age of 43 years, participants who were to report a diagnosis of CFS at the age of 53 years were beginning to report significantly more fatigue than the rest of the sample (OR adjusted for gender = 2.62; 95% CI = 1.1.55.96; p = .02).

Given all the problems with inaccuracy of diagnosis I'm not sure how much I trust this finding but it does tally with comments I've read on this forum from quite a few people.

Evidence of continuing to exercise despite fatigue?
The authors then noted that this group of 43-year olds who would go on to report CFS and were - as a group - showing significantly raised levels of fatigue were continuing to exercise at a high level:

...those who were later to report a diagnosis of CFS were still persisting in exercising more frequently than those who did not go on to report CFS.

This is the basis for the authors claim that

The observation that those who went on to report a diagnosis of CFS continued to exercise more frequently, even after they had begun to report increased levels of fatigue, is of particular interest. It may be that this persistence of strenuous activity, despite subjective fatigue, is an important initial step in the emergence of a CFS state.

The problem with this conclusion is that it depends on those exercise levels findings (shown above) that were not replicated in the larger 1958 cohort. So maybe the fatigued pre-patients weren't really exercising this hard. Furthermore, Figure 1, which summarises this data, itself looks odd. At age 31, around 14% of non-CFS particpants were exercising at least weekly; 5 years later at age 36 this had mysteriously more than doubled to around 37% exercising weekly - so can we trust the data? Just as odd, the data shows that at age 53, the CFS cases were exercising weekly at the same rate as healthy controls (and I don't believe they were all doing GET in 1999...). And if the exercise data is dodgy then so too are the conclusions about over-exercising.

The stuff about it being down to personality types not behaviour was plucked out of the air, and some of the studies that they cited re physical activity found no correlation with personality types.

 

[Esther12]

Darn you OB - you've tamed my righteous recriminations. Nothing's more satisfying than a bit of indignation, and you've now made it quite clear that these findings have little to no bearing on my own experiences.

Oh well. I still hate them, so at least I have that to enjoy.

 

[Enid]

Apart from your marvellous analysis oceanblue - would one trust any of this lot - they are psychiatrists looking for psychiatric conclusions only - not real medicine.

 

[oceanblue]

Esther12
Keep on righteously recriminating - the authors made those very claims, I just suggested they were rather suspect (the claims, not the authors).

Enid
Thanks, and i can tell ou're going to love my forthcoming thread on a paper from the same authors looking at prior psychiatric disorder as a risk factor for CFS.

 

[eric_s]

I think we can be happy about this study. It seems to directly contradict what some of the same people said earlier. So it shows how credible their theories are. Also i like this conclusion better.

22 subjects with a diagnosis made by a doctor. And when it comes to ME/CFS i don't especially trust UK doctors... Can you draw a valid conclusion from such a sample?

 

[oceanblue]

Individuals who exercise frequently are more likely to report a diagnosis of CFS in later life. This may be due to the direct effects of this behavior or associated personality factors.

I don't know how such speculation makes it into the abstract, especially as the authors offered no good evidence that personality factors were linked to the claimed excessive exercising. More to the point, the authors failed to point out evidence against their personality hypothesis: two months before submitting this paper, the same authors submitted another paper analysing the same cohort that concluded

Personality factors were not associated with a self-reported diagnosis of CFS/ME.

 

[Dolphin]

Thanks for doing this, oceanblue. I read the studies at different times while it is interesting to consider them all in and around the same period to see how findings fit together.

Ruling out hypotheses
This paper aimed to test a number of hypotheses, most of which were ruled out, including:
[..]
[*]Childhood Chronic Illness (age 0-15). This contradicts the finding of the much larger 1970 cohort. More to the point, this 1946 study lost a lot of those with childhood chronic illness making it an unrepresentative sample and so these findings are pretty meaningless:
Childhood Chronic Illness (age 0-15). This contradicts the finding of the much larger 1970 cohort. More to the point, this 1946 study lost a lot of those with childhood chronic illness making it an unrepresentative sample and so these findings are pretty meaningless:

Those followed up to the end of our study were not totally representative of the sample taken at the beginning. In particular, males, those from poorer families, and those who suffered ill health as children were less likely to be followed up. This attrition bias must be considered when assessing our results, especially in regard to our negative findings relating to childhood illness experiences

[..]

Main conclusions
The findings on childhood illness are, according to the authors, questionable because some children with ill health were lost from the adult survey. The findings on exercise were mixed, based on small numbers and were not replicated in the larger 1958 cohort. It doesn't amount to much.

Is there any reason to believe the missing data would bias the findings one way or another rather than simply reduce the power? If it is simply reducing the power, if one looks at table 2 http://www.psychosomaticmedicine.org/content/70/4/488/T2.expansion.html , it doesn't look like it would make much difference.

 

[Esther12]

I don't know how such speculation makes it into the abstract, especially as the authors offered no good evidence that personality factors were linked to the claimed excessive exercising. More to the point, the authors failed to point out evidence against their personality hypothesis: two months before submitting this paper, the same authors submitted another paper analysing the same cohort that concluded

lol. That's the sort of thing that makes me love those cheeky chaps.

 

[Dolphin]

Furthermore, Figure 1, which summarises this data, itself looks odd. At age 31, around 14% of non-CFS particpants were exercising at least weekly; 5 years later at age 36 this had mysteriously more than doubled to around 37% exercising weekly - so can we trust the data? Just as odd, the data shows that at age 53, the CFS cases were exercising weekly at the same rate as healthy controls (and I don't believe they were all doing GET in 1999...). And if the exercise data is dodgy then so too are the conclusions about over-exercising.

Could that be due to different questions being asked at 31 and 36 years? It looks to me that the list at 36 years is broader, or else why say what give the full list for the 31 years group, which comes earlier in the paragraph:

Physical Activity

Participants level of physical activity was ascertained at several points throughout their life. Teachers were asked to rate the participants energy level and their ability in sports when they were 13 years old. At age 31 years, participants were asked to describe how often they engaged in various physical activities, such as swimming, cycling, squash, tennis, badminton, fitness classes, or other activities. Using the frequencies of these various activities, we were able to estimate how often participants would engage in any exercise, and whether or not this occurred more than once a week. At age 31 years, participants were also asked to rate their perceived level of fitness. At age 36 years, participants were again asked about various activities; although on this occasion, they were asked to rate both the frequency and total time in the last month they had spent on numerous different forms of physical activity. We inquired about a total of 27 different sports and recreational activities, together with the level of physical activity at work, the amount of cycling/walking, and ten different heavy gardening and do-it-yourself (DIY) tasks. Participants responses for each of these four broad types of exercise were divided into inactive, less active, and most active according to detailed criteria previously published (26). A combined estimate of overall physical activity at age 36 years was made according to the following classification: "Inactive"inactive in three or more of the four broad types of exercise; "Less Active"less active in two or three types of exercise or more active in one category; "More Active"more active in two types or less active in all four; "Very Active"more active in at least three of the four types of exercise. At ages 43 and 53 years, participants were asked to estimate how many times in the previous 4 weeks they had taken part in sports, exercise, or vigorous leisure activities.

Participants heights and weights were also measured by nursing staff during home visits when they were aged 36 and 43 years, thus allowing their body mass index (BMI) to be calculated and the interaction between exercise levels, weight, and CFS to be examined. Weight measurements were also available from birth and at age 7 years.

To me, the findings of increased activity in the group that go on to develop CFS seem more trustworthy because of the lower BMIs at 36 and 43 years http://www.psychosomaticmedicine.org/content/70/4/488/T5.expansion.html.

The stuff about it being down to personality types not behaviour was plucked out of the air, and some of the studies that they cited re physical activity found no correlation with personality types.

Interesting. (if you recorded the dodgy refs, feel free to post them).

-------
I imagine that doctors spend most of their lives incl. during medical school feel fatigued to some extent. I would think they might not like it seen as evidence of dysfunctional behaviour/personality type to exercise when fatigued, or they and lots of others could never exercise. Of course, it may be the case that people shouldn't; however, at this stage, people in society certainly don't get a clear message one shouldn't.

 

[oceanblue]

Thanks for doing this, oceanblue. I read the studies at different times while it is interesting to consider them all in and around the same period to see how findings fit together.

Is there any reason to believe the missing data would bias the findings one way or another rather than simply reduce the power? If it is simply reducing the power, if one looks at table 2 http://www.psychosomaticmedicine.org/content/70/4/488/T2.expansion.html , it doesn't look like it would make much difference.

If we're talking about those with childhood illness being lost from the study then - and we are looking at childhood illness as a risk factor - then yes, I think it could well bias the findings: those missing people could have gone on to develop CFS (if there really is a link between chronic illness and CFS) and the 'no risk' finding could have been changed to a finding of significant risk.

Is that what you were talking about? I'm not sure I've definitely understood your point.

In more general terms, there are presumably a lot of missed CFS cases in the study (as shown by no cases before the mid-80s in all 3 cohorts). I think these missed cases may well be missing at random and so would simply have the effect of reducing the power. Consequently, findings that appear marginally significant/non-significant could in fact be significant, eg a marginal finding of p=0.06 might change to a more convincing one of p=0.01 if the missing data could somehow be recovered.

 

[Dolphin]

This paper aimed to test a number of hypotheses, most of which were ruled out, including:
[..]

Specific childhood symptoms (persistent abdominal pain had been shown to be a risk factor for unexplained hospitalisations and the authors hoped it would be linked to CFS too; it wasn't)

In fact, 0 of the subsequent CFS cases had it http://www.psychosomaticmedicine.org/content/70/4/488/T2.expansion.html (although it was relatively rare - only around 2% of non-CFS group had it also).

 

[Dolphin]

Thanks for doing this, oceanblue. I read the studies at different times while it is interesting to consider them all in and around the same period to see how findings fit together.

Is there any reason to believe the missing data would bias the findings one way or another rather than simply reduce the power? If it is simply reducing the power, if one looks at table 2 http://www.psychosomaticmedicine.org...expansion.html , it doesn't look like it would make much difference.

If we're talking about those with childhood illness being lost from the study then - and we are looking at childhood illness as a risk factor - then yes, I think it could well bias the findings: those missing people could have gone on to develop CFS (if there really is a link between chronic illness and CFS) and the 'no risk' finding could have been changed to a finding of significant risk.

Is that what you were talking about? I'm not sure I've definitely understood your point.

In more general terms, there are presumably a lot of missed CFS cases in the study (as shown by no cases before the mid-80s in all 3 cohorts). I think these missed cases may well be missing at random and so would simply have the effect of reducing the power. Consequently, findings that appear marginally significant/non-significant could in fact be significant, eg a marginal finding of p=0.06 might change to a more convincing one of p=0.01 if the missing data could somehow be recovered.

It's unclear to me what you are saying in the first part - you seem to be saying a loss of power but actually what I was asking about was bias. What you say in your last paragraph is what I mean by loss of power.

But I'm trying to get at is if there is a loss of power, but no biasing effect, the OR would still be in the same direction. If one looks at the data for childhood illness we have in Table 2, it isn't consistently showing that Childhood Illness Measures increase the risk. And the magnitude for the specific item "any chronic illness age 0-15" with an odds ratio of 1.20 isn't particularly interesting, even if it did become statistically significant with a bigger sample size.

----
Here is info on where people were generally lost:

Despite this, there were significant numbers of individuals who were not able to be followed up, with the greatest attrition rates occurring in the early adult years (25).

If we are assuming people don't have CFS, and looking at pre-morbid factors, it is not clear to me why of people who have a chronic illness as a child, the ones who are likely to go on to develop CFS are less likely to return the form. But then I don't know what is the supposed mechanism of childhood illness subsequently causing CFS.

There could be situations (outside the CFS) where I can see how bias could affect who of a particular group does and does not return a questionnaire.

 

[oceanblue]

It's unclear to me what you are saying in the first part - you seem to be saying a loss of power but actually what I was asking about was bias. What you say in your last paragraph is what I mean by loss of power.

But I'm trying to get at is if there is a loss of power, but no biasing effect, the OR would still be in the same direction. If one looks at the data for childhood illness we have in Table 2, it isn't consistently showing that Childhood Illness Measures increase the risk. And the magnitude for the specific item "any chronic illness age 0-15" with an odds ratio of 1.20 isn't particularly interesting, even if it did become statistically significant with a bigger sample size.

----
Here is info on where people were generally lost:

If we are assuming people don't have CFS, and looking at pre-morbid factors, it is not clear to me why of people who have a chronic illness as a child, the ones who are likely to go on to develop CFS are less likely to return the form. But then I don't know what is the supposed mechanism of childhood illness subsequently causing CFS.

There could be situations (outside the CFS) where I can see how bias could affect who of a particular group does and does not return a questionnaire.

Ah, good point, I now see what you mean. The net result of loss of children with chronic illness should be on power, not on the size of the effect, and you're right, OR 1.2 is uninteresting whether statistically significant or not.

Nonetheless, this 1946 cohort was only 1/3 the size of the 1970 cohort that did find chronic childhood illness was a risk factor - and crudely combining the data from the 2 cohorts still gives a sizeable and significant effect for childhood chronic illness.