Drug Information Journal - March 2009 - (Page 140) 140 MEDICAL INFORMATION Dijkman, Fraser, Treasure, Kapke reference interval from different local laboratory intervals (3). While this is a practical and feasible approach, it relies heavily on the relationships between the different local reference intervals in relation to their methods. The National Committee for Clinical Laboratory Standards has provided a guideline as to how reliable reference intervals can be defined (4). However, local clinical laboratories are generally limited in their ability to develop reliable reference intervals using such a strategy, due to constraints in the population size requirements and cost (5–7). In one approach to setting useful reference intervals, Fraser has proposed a hierarchy of methods to set these intervals, depending on the laboratory’s capability (8). In this study, we examine the hypothesis that local laboratories use rational criteria to establish their reference intervals by comparing the signaling of abnormality through local reference intervals with signaling of abnormality through a central laboratory reference interval,* applied across these laboratories. To compare certain data between local laboratories using different methods, a calibration process has been used to harmonize results among them. This calibration process is described in the literature for routine clinical biochemistry analytes (9,10). For routine hematology analytes, methods are assumed to be sufficiently comparable to allow pooling of the data. This assumption is based on results of external quality assessment schemes, such as the CAP and NEQAS surveys, and relevant studies (1 1). REFERENCE INTERVALS Reference intervals are usually derived from the results generated from reference samples collected from a sample population of reference individuals. The lower reference limit (LRL) and upper reference limit (URL) could, for example, be set to include 99% or 95% of that population within the reference interval. The analysis here is based on a clinical trial data set that, for many analytes, will not have the same characteristics as the usual reference population used by laboratories to generate their own reference intervals. The method to be described is empirical in the sense that it uses the data as seen in the data set, while relying on the rational basis through which the central laboratory reference interval was established. The central laboratory reference interval was derived from approximately 2,000 generally healthy clinical trial subjects. The subjects were from a variety of study protocols, representing a range of clinical indications. The central laboratory reference intervals represent the 99% nonparametric intervals for this population. The central laboratory has reviewed its global data through comparison of calculated average of normals and observes consistent data distributions across geography. For common safety analytes, the variation in average of normals between populations of different geography is negligible with respect to these analytes. CORRELATION BETWEEN HARMONIZATION FACTORS AND REFERENCE LIMITS If the reference intervals from local laboratories are related to their methods, there should be a correlation between the harmonization factor and the upper or lower reference limits. We have collected the (measured) harmonization factors for alkaline phosphatase from 20 laboratories. We related these to the reference limits for the age group 18–70 years. Since most laboratories have gender-independent reference intervals for alkaline phosphatase, we averaged the URL for laboratories with gender-specific intervals. The results are provided in Figure 1. Interestingly, Pearson’s correlation coefficient for all labs is 0.95, which suggests a strong correlation. However, from Figure 1, it seems obvious that four laboratories are using a distinctly different method (or methods) than the rest. If regression analysis is performed on the two subgroups, the correlation coefficients are 0.51 for the group with lower URLs and –0.57 for the four with *The reference interval used by Covance Central Laboratory Services, Indianapolis, Indiana.
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.