Premorbid risk markers for chronic fatigue syndrome in the 1958 British birth cohort

[oceanblue]

The abstract gives the impression that the "parental physical abuse" finding was a prospective one i.e. the data was collected when the individuals were children. However, it is actual based on a retrospective question, asked when the individuals were 45. This leaves open the possibility of what is called "recall bias". ..

Even in the results in Table 1 (i.e. only adjusted for gender), for the CFS/ME group (which I find the most interesting), there were only three significant odds ratios. Two of them are parental physical abuse (reported at age 45) and parental sexual abuse (reported at age 45). The third one, cumulative adversity, appears to combine the two types i.e. prospective and retrospective...

Simon Wessely agrees with you about retrospective measures:

Retrospective assessment of pre exposure fatigue and/or life events is basically worthless.

 

[oceanblue]

I've just read this. Can't say I find it particularly exciting but will post some notes...

Thanks for all the excellent analysis, Dolphin. I'm many moons late on this but will post a few comments as I have time, mostly picking up on points you've made.

Female/male ratio looks suspiciously low for a CFS sample
Only 59% of this 1958 cohort were women, compared with 71% for the 1946 Birth Cohort, 78% for the 1970 Cohort, and 80% for the large Reyes prevalence study. Around 60% female is common to many illnesses but the 70%+ typically seen in CFS studies is more unusual (in other diseases). This raises further doubts about the accuracy of diagnosis in the 1958 cohort in particular.

'Imputation' relied on for missing data

Multiple imputation using regression addressed the issue of missing data on the analyses,[19] based on the assumption that data were missing at random[20]...

All variables reported in this paper were included in the imputation equations.21 Gender, socioeconomic position at 7 and 42, and qualifications at 33 that predict attrition were also included.20 Missing data on the independent variables ranged from <1 to 25% except for neglected appearance at 7 or 11 years (38%), maternal absence at 16 years (26%), school absence at 16 years (28%), sport at 16 years (28%) and physical activity in job at 33 years (30%)...

The odds ratios presented use imputed data, but analyses using the imputed and non-imputed data showed similar patterns of results.

I know almost nothing about imputation and it's validity but it's of particular concern for the cumulative adversity measure of 3+ items, as a large proportion of people in this group will have at least one imputed adversity. This Clark paper doesn't give details but the Goodwin 2011 sister paper (Table 1) shows that of those in the total sample with 2+ adversities, only 50% have complete data; the figure for 3+ must be even lower.

 

[oceanblue]

Evidence on childhood adversity inc. physical & sexual abuse is flaky

The abstract gives the impression that the "parental physical abuse" finding was a prospective one i.e. the data was collected when the individuals were children. However, it is actual based on a retrospective question, asked when the individuals were 45. This leaves open the possibility of what is called "recall bias".

Even in the results in Table 1 (i.e. only adjusted for gender), for the CFS/ME group (which I find the most interesting), there were only three significant odds ratios. Two of them are parental physical abuse (reported at age 45) and parental sexual abuse (reported at age 45). The third one, cumulative adversity, appears to combine the two types i.e. prospective and retrospective, judging by the abstract for reference 16:

Also, it's only increased for 3+ adversities, which only covers 6.3% of CFS/ME cases.

The respective percentages for parental physical abuse (reported at age 45) and parental sexual abuse (reported at age 45) are just 16.2% and 6.6% respectively so they can't explain most of the cases. They dont give a figure for the percentage with neither which could be anything from 77.2% to 83.8%.

Thanks for all those points. This is mainly a recap for my benefit:

1. The great strength of these birth cohort studies is that they collect the data prospectively without the risk of recall bias, unlike the retrospective measure of childhood sexual or physical abuse - which were collected retrospectively at age 45. And as Simon Wessely said: "Retrospective assessment of pre exposure fatigue and/or life events is basically worthless."

And as the authors say:

Prospectively measured childhood adversity was not a risk marker for CFS/ME

2. As you point out, the other measure that proved a predictor of CFS in table one, cumulative childhood adversity, includes the unreliable retrospective data Also, even retrospetively reported parental sexual abuse was not a statistically significant predictor of CFS in their multivariate model.

3. Finally, even these retrospective measures only apply to a small minority of CFS cases: 16.2% for physical abuse, 6.6% for sexual abuse and 6.3% for cumulative childhood adversity. These are total % CFS cases, what really matters is the excess cases for physical abuse etc compared with non-CFS cases. The figures for these are 10.2% for physical abuse, 4.0% for sexual abuse and 2.9% for 3+ cumulative childhood adversity; this is a side show at best.

re above post (#23) I can no longer edit
The point I make about using imputed data to calculate cumulative adversity is valid ie for 3+ adversities many cases will rely on imputed data not subjects questionnaire replies for at least one 'adversity'. However, the data I quote from the Goodwin data is for cumulative psychopathology (at age 16, 23 & 33) not for cumulative adversity, though the pattern is likely to be similar.

 

[Don Quichotte]

Don't know if you have seen this excellent "example" of a childhood trauma that explains everything.

http://www.youtube.com/watch?v=3RwToJB2bAE&feature=relmfu

(Seems like this guy realized that a good sense of humor, even if a bit cynical, is sometimes the only way to preserve your sanity).

 

[Valentijn]

Don't know if you have seen this excellent "example" of a childhood trauma that explains everything.

http://www.youtube.com/watch?v=3RwToJB2bAE&feature=relmfu

(Seems like this guy realized that a good sense of humor, even if a bit cynical, is sometimes the only way to preserve your sanity).

Funny stuff ... and it looks like he ultimately got an ME diagnosis, so he's even One Of Us

 

[Don Quichotte]

He is one of the many who have " unexplained medical symptoms" .

The name of this " none diagnosis" diagnosis (a diagnosis of exclusion, for which the main criterion is the ignorance of the physician) has changed many times, and really makes no difference.

It was called hysteria, conversion, neurasthenia, functional....

"what's in a name, that which we call a rose would any other name smell as sweet"

If you believe that someone could suddenly become paralyzed as an adult because he/she was abused in some way as a child, you can believe anything.

Not to say, that if you take this approach, it means that a person who had a true traumatic experience and managed to recover from it, is now doomed for life, because what ever happens to him/her in the future will be attributed to it.

Chloe Atkins story is an excellent example for that: The neglect of her parents, served to explain her illness, and for years she was denied proper medical care (to the extent that it endangered her life) because of that.

Ironically, quite often, the traumatic experiences that patients have under the hands of the more ignorant and arrogant members of the medical profession, serve to explain their lack of recovery from their illness, once it is properly diagnosed, after years of neglect. I would find this amusing if it wasn't sad.

I may be wrong, but I personally do not think that CFS is one disease. I think that one of the most neglected fields of medicine is the understanding of muscle function and energy metabolism in health and disease. The human body has been somewhat arbitrarily divided into different specialties, based on anatomical sites-the brain/nervous system (neurologists and psychiatrists who still seem to be confused about what belongs to whom), the heart (cardiologists), the endocrine glands (endocrinologists), the skeleton (orthopedics), the joints (rheumatologists), the blood vessels (vascular surgeons), the blood and bone marrow (hematologists), the kidneys (nephrologists). But, there is no field of medicine which is dedicated to muscles, even though it is a vital organ, just like the heart.
There are known diseases (endocrinological, neurological, infectious, autoimmune) which can be congenital or acquired, which lead to abnormal function of the muscles and energy metabolism, but I am sure there are as many that we do not yet know about.
The little we know is from sport's medicine.

Probably the muscles of a patient with CFS walking a short distance, respond (to some extent) like the muscles of a marathon runner at the end of a race. This means that you have reached your endurance limits doing what for others would be a normal every-day activity. This means that if you do a bit more (which for someone else would be unnoticeable, such as going up 5 instead of 3 steps) , you would be like a marathon runner that continues to run another 20 miles, which can lead to irreversible damage at some point.

There is one essential difference, though, between a marathon runner and a patient with CFS. the former can improve his endurance and VO2 max. by training and the latter can not. At least not in the same way.

Surprisingly, There are much more studies on marathon runners/sports than there are on patients who find it hard to walk a short distance. If you search VO2 max. you are going to find mostly sports related information and very few studies on patients. Even those are mostly on patients with high blood pressure or diabetes and not diseases that directly affect muscle function.

No doubt that a better understanding of the pathophysiology of abnormal endurance, can lead to better pharmacological and non-pharmacological treatments not only in CFS (what ever it is) but also in numerous other diseases and conditions, including ageing (which is a medical problem 100% of the population will eventually suffer from).

I don't know if you are aware of studies conducted by a drug company cytokinetics- http://www.cytokinetics.com/
the aim of which is to improve cardiac and skeletal muscle force and endurance, regardless of cause.

It is not surprising that their first studies were done on cardiac patients and published in the lancet, because all those patients are under the care of cardiologists and everyone knows what congestive heart failure is.
Studies on skeletal muscles are much harder to conduct because who would they work with-neurologists? who mostly have very little interest or understanding of muscle function. Psychiatrists, who have even less interest and understanding? Rheumatologists? Sport's physicians?

But, still they have done a trial on ALS, and have an ongoing trial in myasthenia.
But, those trials are conducted by neurologists who have very little understanding of muscle physiology.
It is not surprising that ALS patients (who do not have a primary muscle abnormality) have very limited benefit from this medication.

It may be different for patients with CFS, but it doesn't seem any one even considered it for this patient population.

 

[oceanblue]

Problems with using imputed data.

Edit: Perhaps I should first make the simple point that imputing data add a degree of uncertainty to the results, and given the CFS status itself is uncertain because it's self-reported that just adds to concerns about the findings.

I've just being looking into imputed data which features in this paper and is a big factor in the claim of multiple adversities where probably more than half of those in the 3 or more adversity group have at least one imputed rather than real adversity. I'd love someone who knows what they're doing to take a look at this, he says hopefully:

Nb it seems that imputation can't simply be dismissed as 'making up data', as far as I can tell and I really don't understand the details. Bit more on this below, but note that the authors of this paper say

The odds ratios presented use imputed data, but analyses using the imputed and non-imputed data showed similar patterns of results.

They don't mention if results from non-imputed data are still statistically significant, even if the pattern is similar. Compare this with an admittedly limited article on Wikipedia which comments very reasonably

The analysis should ideally take into account that there is a greater degree of uncertainty than if the imputed values had actually been observed

And this current study doesn't seem to do that ie ignores the greater uncertainty caused by imputation.

The best info I found on imputation was referenced by the authors themselves:
Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls (2009)

we review the reasons why missing data may lead to bias and loss of information in epidemiological and clinical research. We discuss the circumstances in which multiple imputation may help by reducing bias or increasing precision, as well as describing potential pitfalls in its application.

So as you can see it isn't straightforward. Nonetheless, the recommendations from the 2009 imputation paper don't appear to have been followed in this birth cohort study:

• If a large fraction of the data is imputed, compare observed and imputed values [not done for multiple adversities]
• Where possible, provide results from analyses restricted to complete cases, for comparison with results based on multiple imputation [not done: just commented on 'similar patterns' but without showing results]. If there are important differences between the results, suggest explanations, bearing in mind that analyses of complete cases may suffer more chance variation, and that under the missing at random assumption multiple imputation should correct biases that may arise in complete cases analyses
• Discuss whether the variables included in the imputation model make the missing at random assumption plausible [not done]
• It is also desirable to investigate the robustness of key inferences to possible departures from the missing at random assumption, by assuming a range of missing not at random mechanisms in sensitivity analyses. [not done]

another paper referenced in the birth cohort study also points to pitfalls with using imputation:

Hawkes (2006) in the Journal of the Royal Society of Statistics

We find that the best predictors of non-response at any sweep are generally variables that are measured at the previous sweep but, although non-response is systematic, much of the variation in it remains unexplained by our models.