Q and A: Establishing a new mean and SD within the manufacturer's range

"I have a question about establishing a Levey-Jennings control chart. laboratories should collect a minimum of 20 measurements over at least 10 days to calculate the mean and SD for the chart. My question is: Is it required that these initial 20 data points must fall within the manufacturer's declared range (or assigned values) for the control material? (we use Randox or BioRad controls) Or is it acceptable to use the lab's own data to establish statistical limits even if some of the initial points fall outside the manufacturer's range?"

This is an excellent question, and a common scenario that all laboratories face.

The manufacturer's control range is meant to provide a starting point. It is typically mean to be a range that encompasses the mean of the control, which means that there could be data outside the range.

Labs often use this range in the beginning, and they do in fact exclude data outside the range, while they accumulate enough data to calculate their own actual mean and SD. By the way, the "10 days" that you mention is not a mandate. You could build your mean and SD over 20 days, try to get it done in just 5 days with multiple runs per day, etc. And then as more data is accrued, build up a cumulative mean and SD. 

If you have historical data that provides you with a working range and SD while you are building up a specific mean and SD, you are not obligated to use the manufacturer's range. If you are working with the same lot, it's safe to assume that the performance of different parts of the same lot should have similar performance. 

Nevertheless, I can see that if there are problems in the future, and the manufacturer learns you accepted data outside their provided range at one point, they will latch onto that discrepancy and declare any problem to be your fault, not theirs.

Another key question is how many points are outside the range. If "some of the initial points" fall outside the manufacturer's range, how many are out out of how many total? A few points are probably okay, but a lot of points are indicative of some significant difference in performance.