Strangely, the NIH is actually funding a population based cohort (I think 200) [for the UK biobank] already and this does not seem to appear in the study.
As the NIH want to do a whole load of in-house testing inc MRI, TMS and physiological tests (using a metabolic chamber), I assume it wouldn't be practical to use patients/samples from the UK.
Sample size
The problem is, as Scarecrow, says, that it is very easy to miss something important with a 40 patient cohort - just for statistical reasons, considering that it is very likely that the important thing may only apply to 30% or so of your cases (maybe even only 10%) with a heterogeneous pathophysiology...
I would use a 40 patient cohort for a look see study costing $100,000 targeting one particular biomarker - where you have no Bonferoni problems, which makes it easier to avoid type 2 errors [false negatives]. For a high cost study looking at lots of variables I find it hard to see how you are going to get reliable stats out of 40. And the key thing is that you may well miss the crucial markers, so going on to study what showed up on bigger populations takes you further and further down a blind alley.
I agree this looks on the small size. I always bangs on about the dangers multiple comparisons and given they plan looking at over 2,000 proteins they would expect 100 false positives by chance alone using p<0.05 (100 biomarkers!). So they would in theory need to use a much lower p value, eg p<0.005, risking missing real positives.
But I wonder if there is a different way? Nath implies they have a different approach in mind:
Initially we will be storing them and what we’ll be doing is looking at cell-free fluid in the CSF and the serum for not just a small number of cytokines– actually 1500 lysates-analytes, ok? So… but we want to be very, very comprehensive, and I’ve developed a proteomics assay in my own lab which will look at about at least 2500 proteins.
So when we look at those, that composite, I think it will be very clear to us what cell types may be dysfunctional in these patients and how we can subgroup those individuals. And that will then allow us to go back and now say, ‘Well this looks like an NK cell function, let’s look at it. Or this looks like a B cell function,’ because there’s just innumerable amounts of very time-consuming, tedious assays for each cell type that you could potentially do, or interactions between cell types. So, instead of doing that at the get-go on everything you could possibly think (of), I think that’s a good screening tool, and then we can focus on the real aspects that we think are really dysfunctional.”
So basically they will look at the patterns in the overall data to identify which cell functions are off, then focus on the most promising area with fewer tests. I don't know how you would handle this statistically, but the basic idea makes sense to me. Does this make a difference from your perspective?
Even so, with heterogeneity, there's a real danger of having at least several different types of patients and I don't know if this kind of rich data is capable of detecting such subtypes. The point is they wouldnt' be looking for individual marker differences, but clustering of different markers. I don't know enough abou the stats as to how many markers/how many patients are needed to get robust results.
Post-infcetious fatigue "Versus" mecfs
I suppose the key question is, does PIF = ME. My gut says no..
I think it depends - certainly I've come across plenty of people with long-term mecfs that began as glandular fever. What seem to happen in various study is that as time goes by, fewer and fewer PIF cases recover:
Dubbo said:
The case rate for provisional post-infective fatigue syndrome was:
- 35% (87/250) at six weeks
- 27% (67/250) at three months
- 12% (29/250) at six months,
- and 9% (22/250) at 12 months.
- (from memory, it was about 6% at 24 months, different paper though)
So while just-6-months-fatigue cases may well be iffy, I doubt they will have many at all of those, and certainly once you get to 2 years you are probably looking at mecfs. Interestingly, Mady Hornig and Ian Lipkin had huge problems finding enough patients using <3 years ill for their study, which might explain why the NIH have set the threshold at 5 years. My guess is there will be very few cases in practice of under two years, but you would want that info and a sub analysis of the shortest-term cases.