False positives when validating define.xml with 2.1.1

t Thierry

on April 2, 2016

Hi,

When validating a define.xml, false positives have appeared:

DEFINE DD0031: 1 Error - Missing NCI Code for Codelist 'Null Flavors'
count: 1 | | Missing NCI Code for Codelist 'Null Flavors' (Error)

DEFINE DD0032: 2 Errors - Missing NCI Code for Term in Codelist 'Null Flavors'
count: 1 | CodedValue = NA | Missing NCI Code for Term in Codelist 'Null Flavors' (Error)
count: 1 | CodedValue = PINF | Missing NCI Code for Term in Codelist 'Null Flavors' (Error)

For these ones, null flavors are not part of the NCI CT, so the codelist and the terms cannot have NCI codes.

DEFINE DD0028: 10 Errors - Term/NCI Code mismatch in Codelist 'PK Units of Measure'
count: 2 | CodedValue, Code = h*ng/mL, C85624 | Term/NCI Code mismatch in Codelist 'PK Units of Measure' (Error)
count: 2 | CodedValue, Code = h*ng/mL/mg, C85627 | Term/NCI Code mismatch in Codelist 'PK Units of Measure' (Error)
count: 2 | CodedValue, Code = h, C25529 | Term/NCI Code mismatch in Codelist 'PK Units of Measure' (Error)
count: 2 | CodedValue, Code = ng/mL, C67306 | Term/NCI Code mismatch in Codelist 'PK Units of Measure' (Error)
count: 2 | CodedValue, Code = ng/mL/mg, C85747 | Term/NCI Code mismatch in Codelist 'PK Units of Measure' (Error)

DEFINE DD0028: 1 Error - Term/NCI Code mismatch in Codelist 'Trial Summary Parameter Test Name'
count: 1 | CodedValue, Code = Pharmacological Class of Invest. Therapy, C98768 | Term/NCI Code mismatch in Codelist 'Trial Summary Parameter Test Name' (Error)

DEFINE DD0033: 2 Errors - Unknown NCI Code value for Codelist 'PK Units of Measure'
count: 2 | NCICode = C85494 | Unknown NCI Code value for Codelist 'PK Units of Measure' (Error)

All these are false positives. My guess is that the bug comes from these codelists being defined in the odm.xml file with EnumeratedItems, while my define uses CodeListItems (decode = NCI preferred term).

Cheers
-- Thierry

Forums: Troubleshooting and Problems

s Sergiy

on April 20, 2016

Hi Thierry, Thank you for

Hi Thierry,

Thank you for reporting this bug!

We are already working on its fix in the next release.

The bug is due to new logic in validation algorithm around NCI codes. In previous version, NCI Code checks were executed if in a define.xml file a Codelits has a reference to standard CT (NCI Code for Codelist). However due to industry poor compliance with Define-XML standard we changed the check logic to be driven by our Standard’s metadata. For example, AEOUT variable should always follow standard CT (OUT). Therefore, P21 rules use Standard's metadata as a look-up source to assign CT (OUT) to AEOUT variable and generate checks for presence and correctness of NCI terms for Codelist and Codelist Terms. Such approach allows to identify errors in define.xml even if study metadata has incorrect or missing reference to standard CT (OUT) for AEOUT variable.

However, such check algorithm logic does not work if standard CT is used in custom domains or Value Level. Checks produces false-positive messages because they cannot find a reference to standard CT for custom domains or Value Level in Standard’s metadata.

There are three different cases of false-positive messages:

False-positive messages due to incorrect Standard’s metadata in P21 specs. For example, PCSTRESU variable has a reference to standard CT (UNIT) instead of (PKUNIT). While a correct reference is still under discussion, now P21 checks run validation for PCSTRESU Codelist assuming standard CT (UNIT). Another example is a missing assignment of standard CT to TSVALNF variable in P21 specs.
Non-clear, confusing messages. For example, DD0028 "Term/NCI Code mismatch in Codelist" messages are generated when actual Terms and NCI Code do match, but populated from non-applicable CT like (UNIT) used for VSSTRESU instead of (VSRESU). New checks with more explicit Messages are expected to handle such study metadata issues.
False-positive messages for custom domains or Value Level which are not specified in P21 Standard’s metadata.

Sorry for inconvenience.

Kind Regards,

Sergiy