The PRDRUG data (CRF page 3) represent NSAIDS/selenium/vitamin E that were taken within 12 months prior to visit 1 but stopped prior to visit 1; the DRUG data represents con meds taken sometime beyond visit 1. Since Table 5 reports medications taken prior to the study then it would seem that PRDRUG should be included. I would suggest that the reporting be based on the preferred terms as per the drug coding via the DRUGCODE.CSV file. Regarding the other 5ARIs, the only one I know of is finasteride (Proscar, Propecia). For the next go-round on this we need clinical input on this issue. (Matt: Wed, 5 Oct 2005 10:09:35)

a) Table 5 - please ensure that only oral ketoconazole is included in the Table (topical ketoconazole is permitted). This will require use of the drug coding terms. Also, are you using the "Drug started Pretrial?" values (DGPRTR variable from the DRUG SAS dataset)? These values should correspond to the missing drug start dates and allow the needed categorization (prior to vs. during study).(Matt: Mon, 17 Oct 2005 09:46:36)

So if you are looking for, say, "Vitamin E", I would look through the DGTX values for vitamin E and note the corresponding CMDRGCOL value. Then select on the CMDRGCOL value to ensure that you get all the relevant entries regardless of spelling differences etc. It would be useful to provide a listing of all unique DGVTX values that map to the CMDRGCOL values of interest so that reviewers can ensure that the appropriate mapping has been done. The one tricky aspect of this is how to handle multivitamins (which may or may not contain Vitamin E). I need to look into this question a bit further and then get back to you on it.(Matt's e-mail: Wed, 19 Oct 2005 10:32:52)

I am assuming that you have merged together the DRUG and PRDRUG datasets with the DRUGCODE.CSV data, merging on the drug code value (ie the variable called "CMDRGCOL"). Note that in DRUGCODE.CSV multi-constituent meds have multiple records to indicate the different constituents. Hence you can end up with multiple records for a given con med record. In addition, a particular drug could be listed under 2 different levels of the "CMATC1" variable in DRUGCODE.CSV (for example, vitamin E or tocopherol maps to both "Alimentary Tract and Metabolism" and "Dermatologicals" ). So I would think that you would want the listing to include the ATC term ("CMATC1") and the preferred term ("PREF") from the DRUGCODE.CSV file, along with the verbatim terms from the DRUG and PRDRUG datasets. You can then summarize the data by CMATC1 and preferred term, being careful to count only unique occurrences of subjects for a given CMATC1 and preferred term since: a) subjects can have multiple entries of the same drug across time, and b) the same drug can be under multiple ATC terms as noted above (Tue, 21 Mar 2006 14:59:19).