Comparing the output of our in-house macro with OpenCDISC output

d Diane

on February 4, 2009

Hi guys, I've finally got test data written for all the Janus/WebSDM checks except for the following few. In general, it looks pretty good. Checks not programmed in our in-house tool, also not supported by OpenCDISC: * Needs MedDRA dictionary: IR4007 * Compares to "description file": IR4250, IR4259, IR4260, IR4261 * Only relevant when using Phase Forward's CTSD safety analysis tools: IR4255, IR4256, IR4257 * Checks for source data failing to load in WebSDM: IR4262 * Checks for using an unsupported Event Dictionary: IR4263 * Checks for value inclusion in a "study-specific codelist attached to a variable": IR4136 When I run both OpenCDISC and our tool, I get the following differences: * Checks firing for me, and not firing in OpenCDISC: CM: IR4134 (Missing CMDOSU when permissable variable CMDOSE provided) CM: IR4117 (CMENRF is expected when CMENDTC is missing) CM: IR4118 (CMSTRF is expected when CMSTDTC is missing) CO: IR4004 (COSEQ is not unique within USUBJID) *** EG: IR4113 (ECTEST length > 40) EX: IR4138 (Missing EXDOSU when expected variable EXDOSE provided) PC: IR4120 (PCEVLINT must be non-negative) PP: IR4122 (PPSTAT not done and PPREASND missing) SC: IR4113 (SCTEST length > 40) SE: IR4105 (If ETCD is 'UNPLAN' then SEUPDES can't be missing) RELREC: IR4001 (Should fire when RDOMAIN is missing) RELREC: IR4139 (Should fire when RDOMAIN is missing) SUPPEG: IR4001 (Should fire when RDOMAIN is missing) SUPPEG: IR4139 (Should fire when RDOMAIN is missing) SUPPXX: IR4258 (Should fire when USUBJID is missing, but might not be since SUPPXX is not a standard domain) *** IR4004 fired for a lot of records when it shouldn't have, but I believe it's due to the SEQs in my test data looking like "124809702001". My guess is that the tool is looking only at the first X number of digits, but that's only a guess. A suggestion for IR4125 and IR1428. These check for missing units when result is present. In order to avoid many false fires, I've programmed mine to only look for units when a value with a number in it appears in the result column. For example, I wouldn't expect a unit for a "NEGATIVE" result, but I would for ">5" or "122". I think that's it for now. Hope this is helpful. Let me know if you have any questions. Feel free to e-mail me if that method of communication is easier. You'd almost certainly get a more prompt response from me that way. Thanks, Diane

Forums: SDTM

t Tim

on February 4, 2009

RE: Comparing the output of our in-house macro with OpenCDISC ou

Hi Diane, Thanks for the feedback, we really appreciate it! The second alpha release (due out sometime in the next few days) will add additional rule types and fix some bugs, so hopefully that will allow for the validation of some of those missing rules. Additionally, I believe some of those definitions should be possible in the current implementation, but may not be included in the default rule-set. As a specific mention to the problem that you're having with rule IR4004, the issue is likely the length of your __SEQ variable. It looks like the values present in your data exceed the maximum value for the Java Integer type. I can change that to a larger data type though, so hopefully I'll be able to make a simple change to accommodate values of that size (and then it shouldn't produce any more false-negatives). Regards, Tim