Roche Elecsys Thyroid Tests 2010

The following gives a detailed account/explanation of how Roche devised their lab ranges (which look very familiar...).

I'm assuming (possibly wrongly) that this is what has been adopted in the UK?

Unless I've misunderstood, there seems to be contradictory information given under the TSH section with regard to comparison with lab ranges in other parts of Europe. For instance, there is a statement:

"The so determined reference ranges correspond to those obtained in 2000 in a multicenter reference value study done in Spain with Elecsys TSH: 0.57 μIU/mL (0.48–0.68 μIU/mL) to 3.65 μIU/mL (3.41–3.90 μIU/mL) 95% central range (0.90 confidence interval of the 2.5% and 97.5% quantiles), n = 168."

However, they have determined their range to be 0.27 μIU/mL - 4.2 μIU/mL.

An upper limit of 4.2 is very different from 3.65.

Does anyone has the brainpower, time and energy (all of which I'm lacking at the moment) to have a more detailed look at the document and provide some cogent interpretation? It also covers FT3, FT4. T3, T4, TBC, anti-TPO, anti-Tg and Tg lab values and at 52 pages, it's quite a hefty document.


8 Replies

  • I didn't have to look any further than the numbers of subjects used to get the ranges. 168, I ask you! it's pathetically too small especially for a hormone whose values change so markedly over the day. Easy question: when were these measurements done, because they are so time-sensitive?. For example, how many were lab staff, how many patients? When I was in charge of developing ranges for tests like this, 500-1000 were thought of as a minimum to get good statistics. Perhaps with antibody tests, smaller numbers, but I wouldn't countenance less than 250. In any case, this is a small sample. The potential for significant error is very large. But this isn't an isolated case of doing the minimum to get a result. Lots of labs in the U K and elsewhere, when devising their own ranges, only use about 100  samples. It's statistically ignorant and potentially misleading. No wonder that the combination of this sort of behaviour and its application to determine "you're in the normal range thus OK" gives so many problems. With such small numbers you can believe in anything.

  • Many thanks Diogenes. The number of subjects quoted (168) actually refers to the Spain study, not the Roche. If you look at Table 1 on page 7, it gives the number of subjects - so, for example, adding up the number of adults aged between 20 and 70 in the TSH column, the total is 993.

    The study also looks at reference intervals in other European countries. According to Table 2 on page 15 the number of subjects tested in 2000 in Group N (Norway) was 70,000 so I would have thought that the outcome (i.e. an upper lab range of 3.6 for females and 3.4 for males) would be far more accurate. However, on p. 8, it states the combined population for the Group N study was only 604!

    Table 2 reveals that there is a differing range of values depending on gender, age, females taking contraceptives, females not taking contraceptives, females who are pregnant, females not pregnant, etc. So how any company can up with a one-size-fits-all reference range is beyond me.

    And yet they state:

    "Applying the given inclusion and exclusion criteria,[1] the ranges for the groups from various locations in different countries that were available for evaluation correspond basically to the ranges given in the package insert (e.g. GHD group[2]) or to the corresponding collectives of the blood donators from Leipzig (e.g. SE group[3] where only anti-TPO-negative samples were used)."

    [1] My understanding is that this is the criteria established by the National Academy of Clinical Biochemistry mentioned on page 6.

    [2] The GHD group is the sample from Heidelberg, though it doesn't provide the number of subjects tested, which is worrying.

    [3] The SE Group is Sweden, and comprised 457 subjects. Yet the upper range is 3.72! In biochemical terms I would have thought that 3.72 is significantly lower than 4.2?

    As you can see I've only looked at the TSH, but in only looking through the first 15 pages there seems to be all kinds of anomalies and contradictions. Obviously, I don't have the medical knowledge to make in-depth scientific judgements, but I have looked at it with a professional editor's eye and it is found wanting.

  • Thank you for the info. That's better but not fully and properly analysed. Re the application of the ref range. Many labs take the manufacturer's range as the basis for their diagnoses, but many still take only about 100 subjects from their own experience to get ranges. There still is a philosophy of the "uniqueness of my patients" in medicine. The rationale is that in different labs they have different mixes of patients thus different overall outcomes. There may be some truth in this, but the numbers often used to get lab-specific ranges are totally inadequate. 

  • In 2013 the National Academy of Clinical Biochemistry (AACC; cited in the Roche study) produced an article discussing the need to harmonize TSH levels ( There are some interesting comments which concur with your own research (apologies if this is already familiar to you):

    "Not only does the pituitary secrete TSH in a diurnal pattern, but many substances produced in the central nervous system, even in healthy euthyroid individuals, may enhance or suppress TSH production in addition to the feedback effect of thyroid hormone. Furthermore, although TSH levels rise and fall in response to changes in the concentration of free thyroxin (T4), individuals appear to have their own set-points, and factors such as race and age also contribute to variability in TSH levels. Alterations of the normal pituitary response are also common in patients with a variety of illnesses."


    "TSH Reference Interval

    For many years, most physicians have considered TSH levels >10 mIU/L evidence of thyroid failure, and levels of 5–10 mIU/L evidence of mild or subclinical hypothyroidism. During the past decade, however, there has been considerable debate about the correct upper limit of the reference interval for TSH.

    Although there is a consensus that the lower limit of the euthyroid reference interval for TSH should be 0.2–0.4 mIU/L, experts disagree about the appropriate upper limit. In 2002, researchers published an analysis of thyroid function test results from a large survey of individuals representative of the U.S. population (3). The study revealed that within a small standard error the mean TSH level in the general population is approximately 1.5 mIU/L. This finding prompted organizations to call for lowering the upper limit of the normal TSH reference range. The National Academy of Clinical Biochemistry recommended 4 mIU/L, while the American Association of Clinical Endocrinologists set the upper limit at 3 mIU/L, and other groups went as low as 2.5 mIU/L. Many clinicians resisted these new limits, because they worried that a significant number of patients would be unnecessarily labeled as having thyroid dysfunction, especially given the fact that there was no evidence that treatment of these individuals would provide any benefit.

    In 2007, researchers analyzed the data from the survey a second time to clarify the relationship of TSH and antibodies to thyroid peroxidase (TPO), a recognized marker of autoimmune thyroid disease (4). TSH levels correlated with anti-TPO positivity, and the investigators asserted that reference interval studies would support the lower upper limit if such individuals, who probably have occult autoimmune hypothyroidism, were excluded. While some groups also have challenged this position, there is growing consensus that one TSH reference interval does not fit all."


    These comments are taken out of context, so the whole article would have to be read.

    I'm not quite sure what is meant by the last paragraph as it seems they are proposing a two-tier testing system, though this is not borne out by the conclusions in the article, nor by the call for views by stakeholders in 2014: . It would be interesting to see what response they received.

  • The whole problem is that the thyroid diagnostic establishment is flailing about trying to justify the diagnostic certainty of using thyroid function parameters (TSH, FT4, rarely FT3) as a simple method of putting patients in a category. We now know this doesn't work. The variability of individuals and their almost limitless ability to make fools of rigid adherents to "ranges" simply challenges the unintelligent use of numbers. Numbers are valuable: they tell you where you are in the situation you are in, but they cannot be taken as absolutes. Thus, intelligent, personal diagnostic methods, that take into account the human race's ability for diversification. Not only numbers but presentation.

  • Thank you (and Diogenes) for highlighting and clarifying this - it's constantly perplexing to be denied medics' acknowledgement of symptoms when they adhere to blood results so clearly based upon lightweight evidence. Stiffens the resolve to go off-piste and take healthcare into own hands (with such able assistance from members).

  • 7 heard that there were only 12 people involved with this 'clinical trial' and lets fsce it, why use a pituartry hormone as the gold standard?

  • Precisely! But we are where we are and in cases such as mine then the more accurate the upper limit, the better. Having said this, I probably wouldn't get very far as my FT3 and FT4 levels are also in range (though in the bottom quadrant), so even if the TSH upper limit was lower and my TSH was then out of range I still wouldn't get any kind of diagnosis.

You may also like...