Repurposing large health insurance claims data to estimate genetic and environmental contributions in 560 phenotypes (2019) Lakhani et al.

I think there may be different ways of expressing risk, which can confuse things. The one measure I understand is the risk of an identical twin of a patient getting an illness in their lifetime. For RA that is about25%. For ank spend it is nearer 60%. For lupus I think it is 30% but I forget.

 

So, to try and answer my own question, I've had a play with the data and I think what it shows is this (my knowledge of heritability statistics is pretty much nil, so please treat with caution and I'd welcome any comments/corrections):

Chronic fatigue syndrome
Heritability, h2 = 0.575, p=0.007
Shared Environment, c2 = -0.106, p=0.464 <--no significant association, 95% confidence intervals are above and below zero

Major depressive disorder
Heritability, h2 = 0.180, p=0.006
Shared Environment, c2 = 0.181, p<0.001

Anxiety disorder
Heritability, h2 = 0.349, p<0.001
Shared Environment, c2 =0.077, p=0.013

Looking at that, it appears to me that:

• Whilst relying on insurance data may be problematic due to coding errors (e.g. depression or anxiety being misdiagnosed/miscoded as CFS), there are substantial differences between the results for each condition in aggregate;
• CFS shows a strong heritable component and no statistically significant 'shared environment' component. i.e. if two siblings have CFS, it's not because of parental influence
• This would seem to contradict, for example, any paediatric CFS study that indicates a certain parental personality is a risk factor for a child developing CFS.
• MDD and anxiety disorders do show a shared environment component, in contrast to CFS. This (to me) provides yet more strong evidence - if any were needed - that CFS is not a mood disorder, that CFS is not just a form of depression or anxiety.
• (I'll admit I was a bit surprised that anxiety disorder had such a strong genetic component - however, unlike CFS, it does still have a shared environment component as well)

 

I did wonder if that graph - which if I'm reading it correctly suggests that published studies shower a stronger heretibility component to this insurance-based study - could be explained by the lower specificity/accuracy of insurance codings and also, perhaps, publication bias.

 

I wonder if this could be useful to Carol Monaghan for her debate next week. I know it is far from proven evidence, but it is another good indicator.

 

Revisiting this paper. Could someone help me to where the data on CFS are summarized? And is there information on how CFS diagnoses were made?

Thanks. Where can I find this individual data file?

On their app, I found the following graph (results with environmental factors) for CFS: http://apps.chiragjpgroup.org/catch/

 

http://apps.chiragjpgroup.org/catch/ under all individual results.

In one of my earlier posts there is also a raw data file formatted by me if I remember right.