Bioequivalence and Bioavailability Forum

Hi Ohlbe,
I agree with everything above except one small point where I think Dr Rao's question needs clarification.

❝ ❝ c. In another rare case, if we have BLQ values in between (eg: 8 hrs.) and before 12 hours,

❝ ❝ 1. Can we use Status Codes and treat the timepoint as "Missing"?

❝

❝ BLQ is BLQ is BLQ, not missing. You do have a result.

Imagine his raw data stream looks like;



If Dr Rao put this data column through WinNonlin as it stands the "<1" or any other text values will in effect be treated as missing.

At 0.25 and 8 hours this may lead to a slight over estimation in AUClast but probably a good approximation, personally it would probably be my default for an analysis if I didn't have an SOP in place before I began my Study.

If you substituted all BLQ (or <1) for Zero you now will have an underestimation of AUClast, particularly between 5 and 12 hours as the curve is 'forced' through ZERO at 8 hours. This is sometimes termed an "embedded" BLQ since there are quantifiable values before and after.

The Status code option is a Tool in WNL that creates a second column where BLQ and other text values are replaced with numeric equivalents; this option can be used to make this substitution dependent on their position in the curve. (note I would retain the original column for tabulation and presentation in the report) e.g.

Before Tmax: 0 and 2.5 above, could set to "0"

After Tmax: <this is for embedded values> i.e here the 8 hours, I personally would set it to missing, so the point is effectively ignored, to avoid underestimating AUC.

First of consecutive after Tmax: after Tlast you can if you wish apply different rules for the FIRST e.g. 14 hours, and SUBSEQUENT points.

After First of Consecutive after Tmax: rule for SUBSEQUENT points.

If asked for a recommendation I would leave both of these to "0" or missing however some people use the "FIRST" function to set some 'real number' e.g half the LoQ i.e. 14 hours to "1". Their purpose here is trying to get a more representative AUC where perhaps the LoQ is high relevant to the observed conc. and this fudge factor makes the variablitly between subjects less at perhaps the first dosing level in an ascending FIM study.

One notable warning is if you do this substitution at FIRST with anything other than "0", Missing or other TEXT value, you will have added another point to your estimation of Lz - I would suggest excluding that point in the Slope selector of WNL if you do use this method.

Dr Rao - Lastly for this and any other confusion with WNL (or other tools) - why not read the manual whilst working through examples of different datasets/profiles to come to an informed decision of the method you are happiest with, and then document that as your SOP.