Dear All I have been asked to provide a method of determining the normal range of a marker (A) for patients on a certain drug based on a different marker (B). The effects of the drug mean that the normal range of A for people not on the drug is not a suitable normal range for patients on this drug. Marker B is too expensive and time consuming to measure on all patients but its normal range for the patients on the drug is known. I propose to do a simple regression of A on B then predict the normal range of A by substituting the two values for the normal range of B into the regression model. However, marker A is subject to a large coefficient of variation and it is vital that patients on the drug are not outside the normal range (so a measurement of marker A might appear to be within it's normal range for patients on the drug but the true value is outside). Given this, my question is, how can I adjust the normal range of marker A predicted from the regression to prevent this from happening i.e. to take account of its somewhat unreliable measurement? Thanks for any help Martin
_________________________________________________________________
http://clk.atdmt.com/UKM/go/msnnkmgl0010000007ukm/direct/01/
|