In Reply.—We are pleased with Mr Tholen's interest in our review article1 and appreciate his reference to the article by Lum et al2 demonstrating that individual laboratories whose calibration was verified and linear on the College of American Pathologists (CAP) Linearity (LN) surveys also had lower rates of proficiency testing failures. We are aware that the findings by Lum et al2 are further supported by a recent study performed by Kroll et al,3 which also shows that there is a strong link between successful performance on the LN surveys and acceptable performance on graded proficiency tests. However, we attempted to confine ourselves particularly to the issue of linearity.
Our statement, “This statistical approach has gained acceptance as the best statistical method to evaluate linearity of quantitative tests and has been adopted as an approved guideline (NCCLS EP6-A),” requires further clarification. Mr Tholen brought to our attention that this particular paragraph could be interpreted in more than one way. We agree with Mr Tholen's comment that the statistical methods used in the NCCLS EP6-A4 approved guideline and CAP LN surveys are different. Our intent was to indicate that both the CAP IRC and NCCLS EP6-A evaluation methods are based on the polynomial method (“this statistical method”) developed by Kroll and Emancipator.5 As described in the excellent article by Mr Tholen6 in Clinical Laboratory News, the EP6-A guideline was developed independently of the CAP protocol and was reviewed under its own scrutiny with comments from the scientific community. Because the CAP IRC protocol and EP6-A guidelines are based on the same polynomial method, we do not desire to state a preference of one evaluation method over the other; rather, the CAP IRC protocol was designed and modified to handle the special situation in which one is evaluating multiple surveys from many laboratories while attempting to minimize manual handling of the data.