T Nation

Different Reference Ranges for Different Methods

Quite frequently it is discussed here why different labs/assays have different reference ranges.
Trying to provide some background why the comparison of lab values is pointless without considering the individual method specific reference range.

Lets take T3 as an example:

Basically two different assay types exist (also true for T for example): the immunoassay type (e.g. RIA, ECLIA) and the LC-MS based assay

LC-MS is considered the golden standard nowadays as it provides the highest specificity combined with a very good sensitivity. The variability of an LC-MS assay is typically somewhere around 10 to 15% (sometimes the 95% confidence interval is stated, ie 95% of the measurements are within this variability - thats an expression of the precission of the assay).

Lets look at the graph below. Its shows the repeat measurements of the same sample using many different assays (x axis). The mean of the reference method (MS) and the 95% confidence interval is indicated by the dotted line.
Assay C for example behaves highly comparable with the MS method, both in terms of accuracy (comparable means) and precision (comparable variability). Results are therefore highly comparable and a reference range established for the MS method might also be applicable for assay C.
Assay B, K or J for example show a very different response. They systematically overestimate the sample and a reference range determined by the MS method is definitly not applicable for these assays.
So how are reference ranges then established for assay K for example?
Two possibilities:

  • Either the lab of the supplier of the assay provides a reference range which is based on a large number of healthy patients - do directly established reference range using this method
  • Or the lab/supplier of the assay crossvalidates the assay to a reference method. Basically samples representing the lower and the higher ref range of the MS method are taken and these samples are measured multiple times with assay K. So the ref range for the MS method can be 0.5 to 2.5 which could translate do a ref range for assay K of 1.2 to 3.0 (numbers are just examples; even the slope of the assay - so how sensitive the assay responds to different concentrations can be different).

Hope that helps

Figure taken from

1 Like

The biggest issue is with free t direct measurement. You can try to explain to the guys here why free t measured is in very different scale than free t calculated.

It would be interesting to better understand how the different methods are actually physically completed by the lab person. Maybe that’s too over our head, I don’t know… but I think it would make people better understand it’s not a simple black and white measurement (like, say, measuring the weight of something). I hope that will help in understanding why different labs can have different ranges but not really be “different”.

But I could be wrong. I’ve never seen how the tests are performed.

Thoughts @johann77 ?

These are 2 interesting videos that give an idea of the ECLIA principle and the operation of the instrument


Yep. Above my pay grade.

Interesting publication comparing the results of 5 blood samples in 65 different labs.

Testosterone concentrations of five authentic human serum samples were analyzed by 65 clinical laboratories using 17 analytical systems. The target values ranged from 43.5 ng/dL to 534 ng/dL, simulating concentrations typical for children, adult males and females, and approximating cutoff values for clinical diagnosis of hypogonadism for men.

Eg Sample III T measured with the CDC LC/MS method was 302.6 ng/dl (Target value).
Results from the 65 labs varied between 203.8–422.0 ng/dL.

For Sample IV (target value 457.0 ng/dL) and Sample V (target value 534.0 ng/dL), participant laboratories ( n = 65) reported results ranging from 316.6 to 570.0 ng/dL and from 311.0 to 732.0 ng/dL, respectively.

A single lab used LC/MS technology.

The Roche Cobas System demonstrated the best accuracy (within +/- 10%) and good precision.

‚Additional effort is needed to improve accuracy/precision of measurements, especially at low concentrations.‘

Fig. 1. Analysis of bias (%) of each individual total testosterone result sorted by analytical systems


@johann77 how can we find out the age ranges of the people used to formulate the range for a given test? I was recently told that the total testosterone ranges were based on 18-25 year olds so someone in their 40s should really be in the lower-mid numbers within the range. I’ve searched but can’t find age ranges of subjects. Thanks.