So EP21, the CLSI standard about total error, enters its fourth year in trying to advance from the “A” version to the “A2” version. Seems a pity since the changes to the A2 version are minor (all the calculation methods are identical). The biggest complaint about the new version was caused by an error that I made in overstating that pre- and post analytical error are part of the total error in EP21. I revised things to say that the opportunity for pre analytical error should not be excluded from EP21 experiments (rather than that EP21 would include all types of pre analytical error). But people still freak out at any mention of pre analytical error for EP21.
Any experiments to evaluate clinical laboratory assay error are just that – experiments. The protocols are not identical to the actual use of the assay but try to get as close as possible.
I contend that pre analytical error is not only part of the total error in EP21, it is also part of the bias in EP9 or part of the imprecision in EP5 and so on for other EP protocols.
Here is a pO2 example. For manufacturers (and perhaps some laboratories) for an EP9 pO2 bias experiment, the pO2 reference method is a certified tank of gas which circulates the gas in whole blood in a tonometer. A technician pulls some blood from the tonometer into a syringe to inject into the candidate instrument. Any air that is left in the syringe will alter the pO2 value and contribute to the result and likely be an error. Of course, an experienced technician will expel all air from the syringe and minimize the pre analytical error but not all technicians are experienced, and even experienced technicians make errors, and not all bubbles are easily seen. One could eliminate this source of pre analytical error by somehow getting an engineer to directly connect the tonometer to the candidate instrument using valves so that blood could be sampled directly into the instrument without the need for a syringe. Of course this is not done and should it be done, it would be bogus, because the pre analytical error that comes from the syringe is part of the assay error.
For another example, a whole blood potassium analyzer that is compared to a laboratory potassium result can potentially have (pre analytical) error due to hemolysis of the whole blood cells.
In another example, a point of care glucose meter that is compared to a laboratory glucose result can potentially have (pre analytical) error due to an improper finger stick for the point of care analyzer when compared to the venous sample drawn from the same patient for the laboratory analyzer.
For the pO2 and potassium examples, the pre analytical error comes with the experiment and represents what may occur with the actual use of the assay. It is with the glucose example that a caution is issued new to EP21 but equally valid for EP9, EP5, and other EP protocols – that it would be wrong to exclude the potential error from finger sticks by drawing venous blood for the point of care device.
Even for cases when a split serum sample is used, user error can occur that is independent from the analytical system. When I was head of conducting trials for a manufacturer, we noticed that customers in clinical trials almost never obtained results as good as our internal results. Although it is hard to pinpoint the source of the differences, the experience of company technicians – who ran the same assay for years – probably played a role. Thus, inexperienced operators did things – even on highly automated systems – to give worse results.
So the bottom line is that EP21 is merely mentioning that for most assays pre analytical error may be contained in the result and that if a situation arises such as the glucose example, don’t go out of your way to exclude pre analytical error by doing something that is outside of the routine use of the assay.
And finally, the “total” in total error is probably not the best term but I don’t know of a better one. This “total” error really means the “total” error from the experiment, which is of course a subset of the true total error which contains error from all reagent lots, all calibrations, and so on.