Bioquimico

Gary L.Horowitz

eJIFCC 2008, 19 (2)

Reference Intervals: Practical Aspects

REFERENCE INTERVALS: PRACTICAL ASPECTS
Gary L. Horowitz, MD
Beth Israel Deaconess Medical Center Boston, MA 02215 Arguably, one of the most important elements of a laboratory test is the reference interval, the values that help clinicians interpret their patients’ test results. Interestingly, though,laboratorians spend surprisingly little time formally addressing the issue of reference intervals. Typically, they adopt the intervals given them by manufacturers, neither establishing the intervals themselves nor even verifying the applicability of those intervals to their patients. In addition, for many analytes, conventional reference intervals, values typically representing the central 95% of healthyindividuals, have been superceded by decision limits. But, once again, individual laboratories rarely verify that their methods provide accurate values, a requirement if one is to use these decision limits. In this communication, I would like to review these three topics in some detail: 1) how laboratories can (and should) establish the accuracy of their values for those tests that have decisionlimits (e.g., cholesterol, glycated hemoglobin, neonatal bilirubin) 2) how laboratories can (and should) verify the applicability of reference intervals they adopt from other sources 3) how laboratories can establish reference intervals. Decision Limits If one were to perform a conventional reference interval study on 120 apparently healthy American adult men, one would find that the central 95% ofvalues would range from roughly 139 mg/dL to 273 mg/dL.[1] At one time, that was indeed the “reference interval” for total cholesterol. However, based on many studies, culminating in the publication of the NCEP ATP guidelines, [2] we learned that the “typical” cholesterol was not necessarily a healthy cholesterol. Indeed, we learned that apparently healthy individuals whose cholesterols were above200 mg/dL were at increased risk of coronary artery disease and that, by lowering their cholesterol levels, their risk could be lowered as well. Thus, the current reference interval, or better yet, decision limit, for cholesterol is 200 mg/dL. This is the value most laboratories use as their “upper limit of the reference interval”. But how does a laboratory know that its cholesterol values areaccurate (that is, that they match the values that a certified laboratory would get by the reference method on this same sample)? And that is what they must want – if their values are low by 10 mg/dL or high by 10 mg/dL (just 5%), a large percentage of their clinical samples will be miscategorized, since many samples have concentrations near the decision limit (50th percentile is 202 mg/dL).[1]

1Gary L.Horowitz

eJIFCC 2008, 19 (2)

Reference Intervals: Practical Aspects

Typically, laboratories assess their performance by daily internal quality control samples (for day-to-day precision) and by periodic external quality control (proficiency testing) samples (for comparability to other laboratories). The latter represent samples measured as patient samples but reported to acentral agency and compared, typically, to other laboratories using the same methods. Ideally, these results would be compared to “truth”, but, unfortunately, external quality control samples are rarely native human serum and, as a result, exhibit “matrix effects”. In brief, they act differently from human serum, and different methods may yield different values on these external quality control sampleseven when they yield comparable results on native human serum samples.[3] Thus, agencies grading external quality control samples typically resort to peer group grading to overcome this problem. As a result, for analytes like cholesterol, individual laboratories may know that their results agree with their peers, but that does not, in and of itself, insure that their results are accurate. The...