Back to the Top
Hello all, I had a question where I am finding conflicting information from an internet search. I
wanted to see if the expertise of the group could give a definitive answer.
I am comparing drug uptake in xenograft models using two tumor lines. The study was done in n=6 with
one tumor type on the left flank and the second on the right. The level of detection was 10 ng/ml.
For tumor type 1, drug was detectable in all samples between 20-180 ng/ml. Tumor type 2 had drug
levels between >10-90 ng/ml. The two mice with the lowest concentration of drug in tumor type 1 on
one flank had undetectable drug levels in tumor type 2 on the opposite flank. If I drop the two
samples that are below the LLQ from tumor type 2 and then take the average of each group, the result
is minimally different which does not reflect the data from any of the samples where drug could be
quantitated in both tumor types. Is it more acceptable to exclude mice where drug could not be
detected in both tumors or to enter a value equal to the LLQ for the two unknowns and add a foot
note to the calculated average.
This is not being submitted to a regulatory agency.
Thanks in advance for your help
Michael D. Cameron, Ph.D.
Department of Molecular Therapeutics
The Scripps Research Institute
130 Scripps Way
Back to the Top
you have been careful to design a study with very well matched pairs. I think you should focus on
analyses which use the power of matched pairs. Which analysis you use depends on what you want. Do
you want proof of the difference? Or do you want a quantitative model of the difference? For proof,
there are paired non-parametric tests which get round the problem of the 2
tumors, allowing you to keep them all (though those two each other). For the simplest model (fixed ratio), I would see if the mean ratio for the 4 mice with
complete data would predict (based on the concentrations in type 1 tumors) that the concentration in
the other 2 mice would fall below the LOQ in type 2 tumors. If it does, I would say you are home and
dry, and you already have your best estimate. If it does not, then you probably cannot tell the
difference between a poor estimate (underestimate) of the fixed ratio and a more complex model with
a non-linear ratio. Since your ratio seems to be approximately 2, it could go either way.
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@lists.ucdenver.edu with "LOQ" as the subject
Copyright 1995-2014 David W. A. Bourne (firstname.lastname@example.org)