|
Article written by Thomas V. Coyner, General Manager, Analytical Products Group
Complete Article from Edition 1 APG eNewsletter
If you have ever failed a PT sample and could not understand why, it just might be the method EPA requires the PT providers to use to evaluate the samples. For most WP analytes and many WS analytes, the acceptance limits are based upon values calculated according to an EPA protocol. This method requires us to calculate an estimated Mean and an estimated Standard Deviation. Then calculate a limit as the Mean plus or minus three deviations for WP or the Mean plus or minus two deviations for WS. The Means and Deviations are calculated from regression constants supplied by the US EPA. How good the final limits are depends in large part on how good the regression constants are compared to the real world.
This all may seem complex but the simple analogy of a calibration curve might help explain the problem. When you do a calibration, you can judge whether the calibration is good by the R squared value given by most instruments or calculation programs. R squared values range from zero to one. Zero implies that the fitted line does not match the data while an R squared value close to one says the line fits the data very well. As an example, the NELAC Quality Standards suggest that the R squared for a calibration should be 0.95 or better for a good calibration. Most laboratories would never use a calibration with an R squared value of 0.9 or lower. A low R squared value says that the data simply does not match the line that was fitted.
Every regression equation has an R squared value much like a calibration. The US EPA supplies regressions are the R squared values for several WP and WS parameters and they are low. What this means to you is that although your performance on a PT sample may be just as good as most laboratories, you will fail because the US EPA equations predicted limits that are not consistent with the actual data being generated by the laboratories. The result is that for some analytes the US EPA equations result in very high laboratory failure rates. Some are higher than 40%. In fact, the R squared value for some analytes are as low as 0.2.
There are several reasons that the EPA-calculated equations are low but in most cases, the US EPA did not have sufficient data to develop appropriate equations. For most of the WP analytes, the US EPA used the data from twelve to thirteen old PT studies and many of these equations are fine. However, for six WP analytes and for ten WS analytes the Agency used data from fewer than five PT studies. Typically, 40% of laboratories fail Orthophosphate in the WS program because the equations are based on four studies. In contrast, the pass rate in the WP program for Orthophosphate is close to 90% because the equations used are much better.
Over the past two years, the EPA has received numerous complaints from PT providers and state certification officers but as of yet has failed to act. However, the NELAC PT Committee is attempting to work with the US EPA to develop new equations based upon data collected from NELAC PT Providers. APG is actively participating in this process.
|
