BQLs

From: peter.bonate@quintiles.com Subject: BQLs Date: Tue, 26 Jun 2001 09:01:13 -0500 I have a question for the group and that concerns the handling of BQLs. Some set them equal to missing, yet others set them equal to one-half the LOQ of the assay and then have some small residual variance term included in the error statement. This latter method has always bothered me because of the 1/2 LOQ issue. The value itself is completely arbitrary and wouldn't this set up the possibility of a mixture model for some pk parameters? What does the group think about instead of assigning one-half LOQ to the missing values, assigning a random value on the uniform interval 0(+) to LOQ. This would insure continuity. Then adding a dichotomous variable that is 0 or 1 to indicate whether the value is imputed and include this as a separate variance term to account for the additional uncertainty in the imputed concentration value. Any thoughts? Pete Bonate