Drug Information Journal - March 2009 - (Page 148) 148 MEDICAL INFORMATION Dijkman, Fraser, Treasure, Kapke Correlation (r) Between the Empirical Local Laboratory Reference Limit and the Actual Local Laboratory Reference Limit TABLE 3 Analyte Leucocyte count Leucocyte count Erythrocyte count Hemoglobin Platelet count Platelet count Serum calcium Serum calcium Serum creatinine Alanine aminotransferase Alkaline phosphatase Limit Lower Upper Lower Lower Lower Upper Lower Upper Upper Upper Upper r Value –0.01 –0.07 +0.19 +0.09 +0.08 –0.06 +0.38 +0.37 +0.12 +0.28 +0.14 F 0.2214 0.1127 0.3445 0.5966 0.0502 0.1576 0.0521 0.0323 0.0525 0.1316 0.2625 Number of Laboratories 180 180 62 175 123 179 84 44 99 154 156 Number of Observations 15,580 15,580 2,752 15,303 13,748 15,312 6,763 4,744 8,323 9,446 9,549 tion). If the choice of local reference limit is essentially arbitrary, then there should be no such trend and no significant correlation. Pearson’s correlation coefficient (r) was calculated for each analyte. Table 3 also records the fraction F of the whole data set cut off by the appropriate central laboratory reference limit and the number of laboratories and observations in the working data extract. A low r value signifies a poor correlation and hence justifies the use of a common reference interval to assess patient safety. The extent by which F varies characterizes the difference between the population considered here and a reference sample population. From Table 3, it can be seen that the population in the data set looks as would a reference sample population for serum calcium at one extreme (5% and 3% cutoff) but not for hemoglobin at the other extreme (60% below the LRL). It can also be seen from Table 3 that the correlation between empirical laboratory reference limit and actual laboratory reference limit is low (r less than 0.20) in most cases. There is no evidence of correlation for all the nonharmonized (that is, pooled) analytes and for the harmo- nized analyte serum creatinine. It is therefore reasonable to use a central reference interval rather than a local interval. With regard to the remaining analytes, that is, serum calcium, alanine aminotransferase, and alkaline phosphatase, present higher correlations (r more than 0.25) and alkaline phosphatase presents a low significance value (P less than .10). The conclusion that the laboratories are rationally responding to local (biological) variations in population should not be adopted lightly. Another potential source of betweenlaboratory variability for the three harmonized analytes with evidence of correlation is methodological differences. If an analyte is over- or underharmonized, then a spurious correlation will arise (indeed, for the present data set, there is less association between the serum calcium empirical and local reference limits if the data are used without applying calibration). Further work is planned to investigate the calibration process and its implications for reference intervals. Data on file at Covance show very little difference between central laboratory (Indianapolis, Geneva, Singapore, Sydney) variability in monthly averages of normals for the analytes
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.