An integral part of routine health checkups involves laboratory measurements on various analytes in the blood. It is then common to compare the value of two consecutive measurements sampled at different times from the same patient. A "significant" change requires an action (additional sample and/or clinical action). The current rule is to check whether the relative range of measurement is larger than a certain critical threshold. This rule should guarantee a specified confidence level (e.g., 95%), but its derivation was based on an approximation. We derive the exact distribution and show that it is related to the doubly noncentral F distribution. The currently used threshold values are compared with the exact ones, and some limited power calculations are presented to detect changes in the patient condition.