Lots of sensitivities, or: How do analytical, diagnostical, and functional sensitivity differ?

Written by Dr. Janet Thode Posted in Method validation

Today it is once again about the clarification of terms. I recently came across a wide variety of sensitivities, which we will explain today in their respective contexts. Where do they come from and why do they cause confusion?

 

Calibration sensitivity and analytical sensitivity

Regarding analytical methods that are used for e.g. testing drugs, i.e., from the point of view of a chemist, the sensitivity of a quantitative analytical method describes how strongly the measurement signal changes as a function of the change in concentration of the analyte under investigation. It is thus reflected in the slope (m) of the calibration function. The larger the slope, the more sensitive the analytical method, since even small differences can be well distinguished. This is also referred to as "calibration sensitivity". However, since the calibration sensitivity does not indicate which concentration differences can be detected, the term analytical sensitivity takes this into account. Analytical sensitivity is the ratio of the slope of the calibration function to the standard deviation (SD) of the measured signal at a given concentration [1]. In other words, this describes the ability of an analytical method to distinguish between concentration-dependent measurement signals.

Thus, the analytical sensitivity has nothing to do with the limit of detection (LOD) or limit of quantification (LOQ), since the LOD can also be below the linear range (in which the slope of the calibration line is irrelevant) and the aspect of statistical certainty (à corresponding precision and trueness) required for the LOQ is also not given.

Analytical sensitivity is used, for example, in pH determinations, photometric determinations by means of the Lambert-Beer law and in analyses for determining water quality in accordance with ISO 11732:2005 [2].

 

Diagnostical sensitivity

And then there is the medical / diagnostic perspective related to the diagnostical sensitivity. This is a quality criterion of diagnostic tests and is defined by the Robert Koch Institute (RKI) as follows "Diagnostical sensitivity is the ability of the examination method to detect as many diseased persons as possible." [3]. In simplified terms, this is therefore the proportion of correctly positive (= true positive) results that come from persons who actually have the disease. This does not refer to analytical laboratory methods and their ability to detect particularly small amounts, but rather to a statistical statement that refers to true or false positive results. That is, how well can this test really diagnose the disease, can the results be trusted? Does a positive result mean that I am really sick? If the diagnostic sensitivity is e.g. only 50%, this means that half of actually existing diseases (like e.g. colorectal cancer) are overlooked and thus every second person is happily sent home, although he unfortunately has cancer...

In contrast, according to the RKI, "analytical sensitivity describes the detection strength of the laboratory method", which indeed can be somewhat misunderstood.

And if you now think that analytical and diagnostical sensitivity are not related to each other, there are cases where this is different, such as for screening tests that are used in the production of blood and stem cell preparations for donor testing and for which specifications are made by the Paul Ehrlich Institute (PEI). For this purpose, it must be ensured that the donor is free of certain diseases and that the test therefore a) does not deliver false negative results (à diagnostical sensitivity) and b) is sensitive enough to detect even the first signs of a disease, i.e. the first disease markers in the lowest concentration (à analytical sensitivity). According to the definition above of analytical sensitivity, I would rather speak of LOD in this context...

 

Functional sensitivity, LOB and the CLSI point of view

And finally, there is functional sensitivity. And what’s that?

To explain this, we’ve got to dig a little deeper. In the past, the term analytical sensitivity for diagnostic analytical methods was defined as "the lowest concentration that can be distinguished from background noise", reminiscent of LOD (and thus explaining the existing ambiguities and often still incorrect synonymous usage). For ELISAs, for example, the blank was measured several times, the SD was determined and the blank value + 2x SD then resulted in the analytical sensitivity (or the LOB, see below). This was then also stated by the manufacturer of diagnostic tests on the package insert. All well and good, but since this indication was not quite useful for clinical application, the concept of functional sensitivity was launched in the early nineties by a group of researchers working on thyrotropin (TSH) assays. According to them, functional sensitivity is "the lowest concentration at which an assay can report clinically useful results." And with "useful results" precision appeared, as a maximum coefficient of variation (CV or relative standard deviation) of 20% was established for this purpose [4]. As time went by, this concept was not exclusively applied to TSH assays but also to other (diagnostic) tests, and thus the definition of functional sensitivity as "the smallest concentration that can be measured with a CV of less than or equal to 20%" can still be found in current books on laboratory medicine [5]. In practice, functional sensitivity can be determined by analyzing the test material (e.g., patient sera) in different dilutions (the analyte of interest is thus present in different concentrations) several times, determining the CV of each dilution level, and then evaluating at which concentration the CV is still less than / equal to 20%. In terms of the basic idea, this already points into the direction of the LOQ, which is why functional sensitivity was often mistakenly equated with the LOQ, and unfortunately still is.

To make things even more complicated, in addition to analytical and functional sensitivity, some diagnostic test package inserts also include a LOB... The LOB stands for "Limit of Blank". It was introduced in 2004 as part of the EP17-A guideline published by the Clinical Laboratory and Standards Institute (CLSI) and is defined as "the highest concentration obtained in a series of results on samples that contain no analyte". This definition is the same as the "critical value" defined in ISO 11843-1, but the CLSI prefers the term LOB. Furthermore, the second version EP17-A2 of 2012 also distances itself from the terms analytical and functional sensitivity, as these were / are often incorrectly used for LOD and LOQ [6]. To learn how LOB and LOD relate to each other, click here.

 

In summary

Calibration sensitivity = slope of the calibration function

Analytical sensitivity = slope / standard deviationmeasurement signal; ≠ LOD !!!

Diagnostical sensitivity = true positive test results / (true positive test results + false negative test results); statistical statement, has nothing to do with an analytical method

Functional sensitivity = detectable minimum analyte concentration with a CV ≤ 20%; ≠ LOQ !!!

LOB = meanBlank + 1.65 x standard deviationBlank

 

References

[1] Skoog D.A., West D.W., Holler F.J., Crouch S.R. (9th ed. 2014) Fundamentals of Analytical Chemistry, Brooks / Cole, ISBN 978-0-495-55828-6

[2] B. Magnusson and U. Ornemark (eds.) Eurachem Guide: The Fitness for Purpose of Analytical Methods – A Laboratory Guide to Method Validation and Related Topics (2nd ed. 2014)

[3] Kommission „Methoden und Qualitätssicherung in der Umweltmedizin“ (2008) Leitlinien Diagnostische Validität, Bundesgesundheitsbl - Gesundheitsforsch -Gesundheitsschutz, 51:1353–1356

[4] Spencer CA. (1989) Thyroid profiling for the 1990's: free T4 estimate or sensitive TSH measurement, J Clin Immunoassay, 12:82-9.

[5] Müller M. (2020) Labormedizin 2020: in Frage und Antwort, irm-books, ISBN 978-3-751917100

[6] CLSI (2012) Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures; Approved Guideline – Second Edition. CLSI document EP17-A2. Wayne, PA: Clinical and Laboratory Standards Institute