The Manufacturer Consumer Specification Dilemma

January 26, 2012

Below is a manufacturer specification for releasing a reagent for an assay.

It is an all or nothing approach, meaning that all reagents in the green are sold and have full value for the manufacturer, while all reagents out of spec (red line) are scrapped and have no value for the manufacturer. This is true for the two data points shown next to the arrow, one has full value, the other no value in spite of the fact that the data points have almost the same amount of error.

In the figure below is the consumer representation.

The blue line shows how patient harm is not an all or nothing approach but rather that patient harm increases with increasing error. For example, for a small glucose error in the hyperglycemic region, the insulin dose is slightly off. But a large glucose error (say a high hyperglycemic reading when truth is hypoglycemic) is life threatening.

This representation exists for many industries (say the fit of a car door, for example).

What’s the point?

In clinical chemistry, specifications from standard bodies (example glucose meter spec ISO 15197) use the manufacturer spec, not the consumer spec. But there are ways to express the consumer (laboratory) viewpoint, by using an error grid. For glucose meters, error grids exist but they are not part of ISO 15197.

CLSI EP21-A2 (Total Error) – rejected after 5 years

January 10, 2012

Reminds me of the Bluto clip “Seven years of college down the drain” So what happened?

A previous president of AACC explained his reject vote of EP21 …

“Pre-and post analytic error is not measurable in a reproducible manner. It will vary by location (ED vs. clinic, ICU private office etc.), personnel obtaining samples, reporting results.  It will vary depending on the LIS/interface of the institution, distance from the lab that the blood is drawn and then transported to, etc. The lab and manufacturers have very little control over this.  Bias and imprecision can be measured quite well and then total analytic error can be estimated.”

This person either doesn’t understand EP21 or pretends not to understand EP21.

EP21 does not attempt to assess error sources.
EP21 does not combine error sources such as implied in the last sentence of the commentator.
EP21 simply measures differences from the comparative assay as total error.

Pre-analytical error is part of what is measured in an EP21 experiment in certain cases. Some examples:

Air bubbles allowed in a syringe of arterial blood contaminate the pO2 measurement with room air.
A tacrolimus assay involves a manual pretreatment step that if performed incorrectly can affect results.

Moreover, EP21 is an experiment. No results are reported to clinicians. Patient acquisition, use of the LIS, and reporting of results will not be conducted according to the routine methods of the laboratory. So any errors that are part of the experiment and not representative of routine use should be excluded. Also, many pre-analytical errors that are important in routine use will be nullified by the fact that this is an experiment. Example: pre-analytical error of sample preparation is likely to be nullified if the experiment is a split sample design.

So why mention pre-analytical error at all?

Because there is the opportunity to introduce bias in the experiment by eliminating the opportunity for pre-analytical error to occur.

Example: compare split sample venous blood instead of a finger stick sample to venous blood for assays that routinely use finger stick and whose comparative assays use venous blood. This would be biased because one has excluded the possibility for error to occur from a finger stick.

So the commentator’s argument against mentioning pre-analytical error doesn’t seem like a valid reason to reject the document.

Air France 447 and clinical chemistry

January 5, 2012

I became aware of an article that discusses the crash of Air France 447. Since airline safety is always used as a reference for patient safety, one can ask what one can learn from this event. According to the article and comments to the Wachter blog, pilot error – specifically panic and decision paralysis in the face of stress were important factors (formally the Yerkes-Dodson law). Now, clinical chemists don’t face the acute stress that results from threats to personal survival – such as loss of an airspeed indicator at 37,000 feet. However, there is a less acute form of stress that affects clinical chemists – the pressure to release products. This can lead to products that harm patients. When bad decisions are made under the pressure of this stress, one can’t blame “management” (the source of the stress). It’s up to each individual to make the right decision.