**Wendy Sutula** at Medstar asks about the PT reference interval calculation. Hi, George. Each time we convert to a new lot of PT reagent we run normals and calculate a geometric mean (geomean) to be used in the INR calculation. My question is this. When we are validating or re-validating the PT normal range, is the geometric mean used to calculate the “mean +/- 2SD” or do we use the arithmetic mean? If it is the arithmetic mean, could you please provide a bit of an explanation? I couldn’t find this information in any of the CLSI documents. Thanks so much.

Thank you for this question, Wendy. CLSI standard H47 A2 specifies the geomean for computing the mean normal prothrombin time in preparation for calculating the INR, and cautions against the arithmetic mean. For review purposes, the geomean is the n root of the product of n normal specimen PTs, a formula that diminishes the power of extreme data points. In contrast, reference intervals are routinely computed using the arithmetic mean, which is the sum of data points divided by n or n-1. The arithmetic mean is necessary for accurately computing variance and standard deviation, upon which we usually base our reference interval, +/– 2 SD (CLSI Document C28). So in fact, it is necessary to compute two different means.

By the way, although experts generally insist that the INR is confined only to monitoring stable vitamin K antagonist anticoagulation, we find that our physicians consistently refer to the INR for nearly all PT applications. For this reason, there is value in computing the reference interval in INR as well as the PT in seconds.

Hi, Heather, I discussed thisHi, Heather, I discussed this with Larry Brace, PhD while at the Thrombosis and Hemostasis Summit meeting last week and he suggests that the INR reference range be computed from the arithmetic mean directly. Do you see any instances in which the results of a patient PT in seconds and the results of the INR are not concordant, meaning, when one reports as abnormal and the other as normal?

This interesting. I haveThis is interesting. I have used our geometric mean to verify our PT reference range, and calculated the INR reference range from the geomean and ISI. Should the INR be calculated with an arithmetic mean?