Tests without reference ranges

RE: Tests without reference ranges

Beats — Mon, 25 Nov 2019 15:51:02 GMT

[quote user="Rob Davis"]Sorry to continue the pedantic line but... this is only correct if all 20 test results are completely independent of one another, and I don't think this would be the case for most commonly run panels....[/quote]

No apology needed. I hang my head in shame. Read "independent" where I have written "individual".

RE: Tests without reference ranges

Rob Davis — Mon, 25 Nov 2019 14:32:06 GMT

[quote user="Beats"]"If you do 20 individual tests on a group of animals, then approximately two-thirds will have at least one result outwith a 95% reference interval, while one-third will have all results within a 95% reference interval."[/quote]

Sorry to continue the pedantic line but... this is only correct if all 20 test results are completely independent of one another, and I don't think this would be the case for most commonly run panels....

RE: Tests without reference ranges

Beats — Fri, 22 Nov 2019 10:47:05 GMT

[quote user="David Mills"]Maybe I wasn't clear enough. Yes 95pc of "normal" animals will be within the reference range. But statistically, 5pc aren't, whether that's above or below the stated range. Therefore if you do 20 individual tests on a single animal, statistically, 1 of those values will be outside the reference range, either above or below.[/quote]

i think I still disagree, but happy to be labelled a pedant for doing so!

If we define a reference interval (as is commonly done) as the interval containing 95% of results for healthy animals for that test parameter, then there is a 5% chance (0.05 probability) that a random healthy animal selected from that population will have a test result outside of that reference interval. On this point I think there is no disagreement.

If we run 20 (independent) tests on that same patient, then there is a 95% chance (0.95 probability) for each test result that it will fall within the reference interval. Thus the chances of all 20 results falling within the reference interval would be 0.95^20. The chance of this not occurring would be 1-(0.95^20).

The statement: "Therefore if you do 20 individual tests on a single animal, statistically, 1 of those values will be outside the reference range, either above or below." would read to me [and here I am perhaps a pedant] that there is certainty, i.e. a probability of 1, that one [and to be most pedantic, perhaps precisely one, no more and no less] of those test results will be outwith the 95%-defined reference interval.

I would accept something vague like "...if you do 20 independent tests on a single animal, on average, at least one of those test results will be outside of the 95% reference interval" without quibble. And I appreciate that your use of the work statistically may be intended [perhaps justifiably] to mean on average, but I have a bee-in-my-bonnet about the word "statistically" being used in such a vague sense where its meaning could be misconstrued [I accept this is undoubtedly a pathological state from which I suffer!]

To be precise, if we run 20 independent tests on a single animal, then the probability of at least one of those tests being outside of a 95% reference interval is (to 2s.f.'s) 0.64, or 64% if we prefer. This is closer to a probability of 1 than a probability of 0, and above the 0.5 threshold to make it "more-likely-than-not" that at least one test result would be outwith the 95% reference interval, but is not the same as saying one of those test results will be outside the reference interval with certainty. Another way of saying this might be:

"If you do 20 individual tests on a group of animals, then approximately two-thirds will have at least one result outwith a 95% reference interval, while one-third will have all results within a 95% reference interval."

RE: Tests without reference ranges

David Mills — Fri, 22 Nov 2019 01:05:47 GMT

[quote user="robloxley"]

[quote user="Beats"]

[quote user="David Mills"]So in 20 measured variables 1 will be "out of range" statistically. [/quote]

I'm not sure I agree with that interpretation though!

[/quote]

If 95% of the population sit in the reference range, of the remaining 5% half are above and half are below the reference range, so if a result is above the reference range there's only 2.5% chance this is natural variation?

[/quote]

Maybe I wasn't clear enough. Yes 95pc of "normal" animals will be within the reference range. But statistically, 5pc aren't, whether that's above or below the stated range. Therefore if you do 20 individual tests on a single animal, statistically, 1 of those values will be outside the reference range, either above or below.

It amazes me that labs don't have reference ranges for immature animals. Similarly, that they don't develop specific urine and ear culture and sensitivity profiles - the current ones are simply useless, and I do wonder about the ethics of "selling" them.

RE: Tests without reference ranges

Evelyn Barbour-Hill — Thu, 21 Nov 2019 20:05:33 GMT

[quote user="Charlotte Marshall"]

Statistics are not my thing but I definately remember being told in a lecture by someone from idexx (At the harrogate ecc conference a few years back) that if I did an in house general health profile - so 12 tests at the time because this was before the catalyst machine, statistically there was good likelihood one would be outside the reference interval but be normal variation. I can't remember the exact likelihood but it was phrased as pretty much as you could expect it.

[/quote]

Ah, that rather confirms my view of blood results: if they confirm your diagnosis, they are reliable and confirm your diagnosis; if they don't they are obviously anomalous.

Only joking! (Well, half joking)

RE: Tests without reference ranges

Charlotte Marshall — Thu, 21 Nov 2019 19:57:38 GMT

RE: Tests without reference ranges

Rob Loxley — Thu, 21 Nov 2019 19:10:35 GMT

[quote user="Beats"]

[quote user="David Mills"]So in 20 measured variables 1 will be "out of range" statistically. [/quote]

I'm not sure I agree with that interpretation though!

[/quote]

RE: Tests without reference ranges

Beats — Thu, 21 Nov 2019 12:29:49 GMT

[quote user="David Mills"]Tests aren't inherently bad. It's their interpretation. [/quote]

Full agreement.

[quote user="David Mills"]So in 20 measured variables 1 will be "out of range" statistically. [/quote]

I'm not sure I agree with that interpretation though!

RE: Tests without reference ranges

David Mills — Thu, 21 Nov 2019 00:53:38 GMT

Hmmm reference ranges.

They're normally based statistically on around 100 lab animals.

The error acceptable is the 5pc of science.

So in 20 measured variables 1 will be "out of range" statistically.

Tests aren't inherently bad. It's their interpretation.

RE: Tests without reference ranges

Beats — Wed, 20 Nov 2019 22:59:12 GMT

[quote user="Evelyn Barbour-Hill"]Well, OK, if that's their approach. But it's axiomatic that different laboratories obtainly consistently different results for the same assay. It's not a matter of "knowing what's normal", it's a matter of needing that laboratory's reference range.[/quote]

Which is one of the reasons that the oft-cited IRIS guidelines are pointless from the perspective of a General Practitioner like myself...?

http://www.iris-kidney.com/pdf/IRIS_Staging_of_CKD_modified_2019.pdf

Back in my day, everyone talked about results in relation to a multiple of the upper reference limit (URL).

(and talk of a reference range was scorned, reference intervals being all the rage).

["First, some semantics; the reference limit is the upper and lower extreme of the reference interval, whereas the reference range refers to the difference between two values. If, for instance, we take the upper and lower reference limits for sodium as 135 and 145 mmol/L, respectively, the range is 10 mmol/L while the interval is 135–145 mmol/L"]

https://pmj.bmj.com/content/postgradmedj/94/1117/613.full.pdf

RE: Tests without reference ranges

Robert Lowe — Wed, 20 Nov 2019 19:56:22 GMT

Reminds of the vitamin E reference range. There didn’t use to be one but when supplementing in cockers with retinal pigment epithelial dystrophy it was useful to see the improvement and we could also compare it to an unaffected dog in the same household on the same diet.

Anyhow I had obviously seen and sampled a few dogs in a short period of time so the lab phoned me up and asked what I thought the reference range should be as they felt they couldn’t carry on sending out results without one. I rather plucked a figure out of the air which wasn’t too far out when someone finally put together some real data together.

RE: Tests without reference ranges

Evelyn Barbour-Hill — Wed, 20 Nov 2019 18:24:45 GMT

[quote user="Neil Wheadon"]

It's all about getting a feel for it, so if you've bled 3 and the results from the first 2 are 100 units whilst the 3rd is 6000 units, it might give an indication

I think I'm making sense??

[/quote]

Yes. But how do you know the first two weren't abnormal instead - or as well. Maybe that laboratory commonly obtains figures between 100 and 6000 for that assay?

So the laboratory is saying "Hey, we're just the technicians here, we'll measure the figures but it's up to you to know if they're normal or not"

Well, OK, if that's their approach. But it's axiomatic that different laboratories obtainly consistently different results for the same assay. It's not a matter of "knowing what's normal", it's a matter of needing that laboratory's reference range.

RE: Tests without reference ranges

Neil Wheadon — Wed, 20 Nov 2019 17:54:26 GMT

I would guess that if you haven't done enough of them on a particular species, you don't know what is normal? In the case it was a tortoise and species of tortoises vary enormously, so without bleeding 100 normal ones (though are they clinically well) you'll never know

It's all about getting a feel for it, so if you've bled 3 and the results from the first 2 are 100 units whilst the 3rd is 6000 units, it might give an indication

I think I'm making sense??

Neil