Thyroid UK
82,632 members97,666 posts

TSH Testing - Fools Gold

I just replied in another thread saying that when I was first ill in 1976 my FT4 and FT3 were measured, but my TSH wasn't. I think this was St Thomas' Hospital in London, clearly the new TSH test hadn't quite taken hold there at that time.

It is widely stated that the usual reference range for TSH is the "normal" range. Normal has nothing to do with it. It isn't even how it is worked out.

This is how my local lab tell me the range is worked out. They take the average of the TSH from blood tests that are submitted, taken from people who do not yet have a diagnosis of thyroid problems. Bare in mind that most of those tests are taken from people who are unwell for some reason, and that they are not screened to ensure they are taken from people who do not have thyroid symptoms. The lab told me they are not allowed to take tests from healthy controls to determine the ranges.

So. They have these tests, taken from people who need a blood test. They test the TSH and determine the mean (average) value.

They then apply a mathematical trick. Not a medical test, not by questioning people to see at which level they start to feel hyper or hypo. No, a mathematical trick. They apply 2 standard deviations to ensure that a range is developed that covers 95% of the population from whom the tests were taken.

It is then randomly decided that those 95% cannot have thyroid problems and only the remaining 5% do. It is no coincidence that the incidence of thyroid disorder in most countries is 5%. Not because only 5% of people display thyroid symptoms, but because the artificially developed range only "allows" 5% of people to be treated.

(look at this too, ) Scroll down to "Reference Ranges".

Compare that with this study which found a TSH in healthy controls of 1.55 +/- 0.78

Of course the above applies to people with primary hypothyroidism. The TSH test is worse than useless for anyone with secondary/tertiary hypothyroidism or who has a conversion or uptake problem.

7 Replies

rosetrees, "The lab told me they are not allowed to take tests from healthy controls to determine the ranges." I'm pretty sure that violates standards but Diogenes would know better. They are supposed to establish the reference ranges from highly qualified controls.

The Australian guide is interesting. I don't see that realization often enough. PR

"There is considerable debate about the normal upper limit of the TSH reference range. The high background prevalence of autoimmune thyroid disease as well as the age, iodine status, smoking prevalence and ethnicity of the 'normal' population has raised the 'normal' upper limit. In people without these factors the upper limit is probably 2.5 mIU/L."


I don't think that using healthy controls is forbidden. Hospitals get their own ranges from as few as 20 subjects - lab staff might well be included. Statistically the rules are silly - I wouldn't derive a range on less than 250. The problem is who to include - all healthy, or people coming to the doctor but classified euthyroid, a mix of these and if so what mix. You have to use statistics to get a range to work with, but the ranges should not be used like a football goal with two rigid posts either side - ie if you score a goal you are euthyroid however close to the post you score it, and if you miss however finely you aren't. That is a silly way of proceeding. What statistical analysis actually points out is the steadily increasing probability of having dysfunction the more your TSH number draws away from the average and nears the statistical cutoff lines. When you cross the line the probability then shoots up quickly, but by statistical definition 2.5% of actually euthyroid subjects will fall outside the cutoff at each end. But are these actually euthyroid? This is the whole dilemma of how to make a TSH reference range, who do you use and how do you interpret? Finally why not do a 100% inclusion of the "normals" and use that as a range. The answer is that you'd get so many hypo people especially, put artificially in the euthyroid range to give a big chance of false negatives. The ranges should be set by what are known as ROC curves - ie setting the cutoff to exclude as many true hypos as possible, but including as many euthyroids as possible in the ref range. The problem is less at the hyper end, where the overlaps are less, but with patients on T4, the overlap at the hypo end is horrendous, even worse than in primary hypo states. All this really explains why TSH ranges differ so much between hospitals even with the same source of TSH test..

1 like

I can only tell you what the Hereford lab told me, that they are not allowed to use healthy controls to define the ranges. Possibly because it would cost too much. They didn't say they were forbidden (or is that the same thing, not sure).

I wonder how many other disorders have a curious incidence of 5% because the ranges are simply set that way?


All of them virtually. Its the way statistics works.


rosetrees, You said, "It is then randomly decided that those 95% cannot have thyroid problems and only the remaining 5% do." That is not what they are saying, the 5% are included in the out of range group which on the top end is anything beyond the reference range. It is not just limited to the 5%. Then you get into the subclinical range to 10. PR


I think I get what you are saying, but the range is designed to cover 95% of the population and the overwhelming majority of doctors are not interested in anyone whose tests fall within that range.

"Subclinical" is a fallacy. "Subclinical" by definition means that the patient has no symptoms. A subclinical disease is usually found on testing but the patient has no symptoms.

The definition of subclinical hypothyroidism is that the patient has "few or no symptoms or signs of thyroid dysfunction and thus by its very nature subclinical thyroid disease is a laboratory diagnosis. "

Clearly the overwhelming majority of people who are told they have subclinical hypothyroidism have anything but. They usually very symptomatic and therefore are hypo. They are not subclinical.


Subclinicals are a mixed bag. Some people will return to normal after a while re TSH etc, some remain in this kind of limbo for quite some time. And then they'll either eventually return to normal or go on to definite disease. But at least in Germany there is recognition that subclinical hypos and hypers do have tendencies for eventual illhealth above what is true for the normal person. And so there is a move to give low dose T4 to try to reduce the TSH back into the normal range. And to give carbimazole in low dose to normalize the subclinical hypers. One thing I wonder is whether, from time to time, all of us get short episodes of subclinical alterations, but because we're never tested (i.e. we don't go to the doctor) it will never be known except if by accident we do go for another reason and get thyroid function testing as a routine.


You may also like...