Q and A: Is the 10:x rule still compulsory?

A recent question came to me via social media:

"Hello Sir, Good morning, How are you?
I wanted to know is still 10x rule is coumpulsary?[stet.] 

Even [when] we are using lab mean and lab SD.
Please share the details.

Thanks and Regards,
[anonymous]"


10:x is not compulsory by any regulation that I am aware of. Many regulations require you to perform QC, but the type of QC you must do is not specified. For example, CLIA says you must run QC once every 24 hours (for most tests). CLIA does not tell you what rules to use to do that.

If we move past compliance into quality itself (does the patient/clinician/test method _need_ the 10:x rule?), my short answer is a bit cryptic: 10:x is not necessary until it _is_ necessary. If you don't need a 10:x rule, you should not use it. If you do need a 10:x rule, however, then you should use it. Determining whether or not you need the 10:x is now the deeper question. 

Again, the shortest answer to that next question is, for methods of 3 Sigma and lower, the 10:x _is_ necessary. Those methods are not stable, and even small statistically significant errors are also clinically significant. For methods of 4 Sigma or higher, the 10:x is not needed for rejection. Some labs might still use it as a warning rule, but in general we discourage the use of warning rules.

So now the deeper question is, how do you determine the Sigma of a method? Well, for that, you need the analytical Sigma metric equation: (TE:a - |bias|)/ SD [this, all expressed in units. You can express it in % if your bias is small, then the SD is CV] 

There are deep divisions over whether or not the Westgard analytical Sigma metric equation is metrologically correct. The alternative is to determine measurement uncertainty, which usually means pretending that bias doesn't exist, and also abandoning any guidance to what QC is necessary. If you go down the path of mu, you are only in search of more uncertainty. For mu, there is no guidance on whether a 10:x is necessary or not. There is only more uncertainty.

Now, assuming you are going down the path of TE:a, the deeper question is, what is the TE:a for my method? Surprisingly, there is no global consensus for this. Instead, we have a mix of goals from all around the world, where you can find a big range of numbers from RCPA (Australia) to CLIA (USA) to Rilibaek (Germany), to biological variation-based goals (EFLM). Westgard provides consolidated tables of all these goals, as well as a recommendation for the most practical choice. We start with the CLIA 2025 goals as the base and then evaluate the options for tests not covered by CLIA.

So, that's a long, long, long answer. Probably more complex than you were hoping for. But QC for laboratory medicine is not "one size fits all." We don't mandate "10:x for all" or "10:x for none" - it's a qualified "10:x for some."

P.S. The use of 10:x practically mandates the use of your own mean and SD. If you use someone else's mean, you're probably going to encounter many 10:x violations (since you are not centered on your actual mean, but are just slightly off to one direction).