I gave up worrying about Vitamin D when I saw how much variability there is in labs that measure it. A CLIA-certified lab (i.e. what you need to do if you want to get insurance to pay) must only show that it delivers consistent results, not that its results are directly comparable to those of other labs. Here’s the variance that various labs show using the exact same control.
Note that this is a more serious problem than just “I can’t trust my lab results”. It also means you can’t compare results from one study to another unless you correct for the specific lab protocol that does the measuring. Bottom line: Vitamin D might be important but who knows because you can’t compare the results from studies that measure it.
Experimentally, NIST tested a bunch of tests 2021 and while they weren’t great (and many did not meet NIST’s target of ± 5% bias) they overall seemed to be OK-ish if your purpose is just to decide If your levels are low: (Fig. 3 and Fig. 4) https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=930349
I’m not sure what’s going on because this NIST test seems to include many of the same vendors shown in your figure. One possibility is that there’s a skill component to doing the assays? In the above paper, they were seemingly done by NIST. Maybe in your figure the assays were done by different people? It seems possible that this could lead to some extra error.
I gave up worrying about Vitamin D when I saw how much variability there is in labs that measure it. A CLIA-certified lab (i.e. what you need to do if you want to get insurance to pay) must only show that it delivers consistent results, not that its results are directly comparable to those of other labs. Here’s the variance that various labs show using the exact same control.
Note that this is a more serious problem than just “I can’t trust my lab results”. It also means you can’t compare results from one study to another unless you correct for the specific lab protocol that does the measuring. Bottom line: Vitamin D might be important but who knows because you can’t compare the results from studies that measure it.
Thanks. I’m having trouble verifying this, though. At a glance, CLIA-verification seems to require both precision (consistency) and accuracy: https://www.cms.gov/regulations-and-guidance/legislation/clia/downloads/6064bk.pdf
Experimentally, NIST tested a bunch of tests 2021 and while they weren’t great (and many did not meet NIST’s target of ± 5% bias) they overall seemed to be OK-ish if your purpose is just to decide If your levels are low: (Fig. 3 and Fig. 4) https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=930349
I’m not sure what’s going on because this NIST test seems to include many of the same vendors shown in your figure. One possibility is that there’s a skill component to doing the assays? In the above paper, they were seemingly done by NIST. Maybe in your figure the assays were done by different people? It seems possible that this could lead to some extra error.