EP21 and Sampling

EP21 continues to struggle – it’s going on 4 years which seems extreme for what was to be a simple revision. I have complained about obstructionists holding up documents but there are just too many smart people objecting to EP21 for me to chalk this issue up to obstructionists.

The issue has to do with pre analytical error and total error. As a glucose meter example, an EP21 evaluation would compare finger stick glucose meter results against a laboratory glucose method using a venous sample. Assume that the finger stick is performed by non laboratory personnel, which is what would routinely happen (in this hospital). In this case, some people comment that:

  1. The laboratory has no control over non laboratory personnel.
  2. The total error results may differ from this evaluation compared to one in a different hospital (due among other reasons to differences in training for the non laboratory personnel who perform the finger stick).
  3. The laboratory won’t know the performance of this ABC glucose meter’s analytical performance (i.e., separated from the error contribution from finger sticks).
  4. There may be variation in the error rate of the non laboratory personnel who perform the finger stick.

My response is that all of the above are true. EP21’s goal is to provide the distribution of differences between the candidate and comparative method that are representative of what will occur in routine use. To quote Cuthbert Daniel (one of my mentors) “The observations must be a fair representative, random) sample of the population about which inferences are desired.” To respond to 1-4:

  1. The laboratory has no control over non laboratory personnel. Dealing with interdepartmental interfaces may be difficult but this is no reason to subvert the goal of EP21.
  2. The total error results may differ from this evaluation compared to one in a different hospital (due among other reasons to differences in training for the non laboratory personnel who perform the finger stick). True, but what is important are the results for the hospital where the study was performed.
  3. The laboratory won’t know the performance of this ABC glucose meter’s analytical performance (i.e., separated from the error contribution from finger sticks). True, but this is not the goal of EP21.
  4. There may be variation in the error rate of the non laboratory personnel who perform the finger stick. The protocol should account for this variation by adequate sampling.

Quote is from Application of Statistics to Industrial Experimentation, Daniel C., Wiley, NY 1976 pp 5

Advertisements

2 Responses to EP21 and Sampling

  1. W Gernand says:

    Hello Jan,
    Total error (TE) is a figure calculated from a sample of N observations. I’m interested in calculation of confidence interval for TE. This is probably trivial question, but how to calculate standard error for TE?

  2. jkrouwer says:

    EP21A describes parametric and non parametric tolerance intervals. For a non parametric confidence interval, see Hahn GJ and Meeker WQ. Statistical intervals. A guide for practitioners. Wiley: New York, 1991, pp 103-105.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: