2015-04-22 Lab Terminology Team LOINC/UCUM Discussion Meeting Notes

1 – Comparison of CDISC Lab Test terminology and LOINC

• Background:  History of the CDISC Lab Model.  Why was the decision made to not use LOINC initially?  CDISC Use of LOINC Codes.ppt

• CDISC having separate fields for each concept (analyte, specimen type, method, and unit) is a much better fit for Data Warehouses, Statistical Analysis, Informatics and Data Mining.

• Does the CDISC/FDA want method specific versus method less LOINC codes?

• Does the CDISC/FDA want time draw specific LOINC codes for true challenge test (e.g. post-dose glucose)?

• LOINC Units:  SI versus Conventional may be different codes (e.g. glucose mg/dL versus mmol/L)

• If we use LOINC codes (US based), why not SNOMED (EU, Canada, Australia based)?

  • What does Japan use?

• LOINC Code:  Possible factor for traceability. 

  • LOINC code supplied by the lab might make it easier to pool data from different sources.

    • BUT there are sponsors who need to pool data based on the separate fields for each concept.

    • SAS programmers cannot use LOINC easily.

  • Sponsors should not map data to LOINC

    • LOINC codes are available in electronic transmissions from large labs

    • BUT there are labs (local labs) that do not supply the LOINC code even when asked

• What is the business case for needing LOINC?

  • CDISC goal to get LOINC codes into SHARE.

    • What about everyone outside the US, is there a possibility that we will need to add other codes later like SNOMED?

• Additional comments after the meeting:

  • Covance:  As an FYI, Covance has offered LOINC code translations in its data transmissions for the last 5-6 years, but only two sponsors are taking advantage of this, and one is a startup who decided to base their lab dictionary on LOINC.

    • So basically, very few pharma companies have seen a need for LOINC codes, and most do not have a place in their resulting data bases for such a code. So don't expect the pharmas to spend a lot of time or energy on this.  Lilly is an exception to this, and Phil will try and get a contact there to explain their use case/storyboard reasons for wanting LOINC.

  • AstraZeneca:  We can use LOINC to supplement the lab information; we would not be able to use LOINC exclusively.  LOINC only covers a portion of the lab data in SHARE

  • Pfizer:  Question was raised on the value of LOINC code and its utility by the FDA.  No decision has been made by FDA. Current ‘research’ effort is proposed to explore its potential use.

  • Lilly:  We at Lilly will have to be very careful about how we utilize this tool- not all the detail will be there for every Lilly test code/ Lilly unit code combo & if the LOINC is not specific enough/ granular enough (in some cases) we should not try and force mapping/ or lumping/ or assigning a LOINC. 

    • SNOMED field may also be needed for EU trials (they do not use LOINC) and often local labs will not have LOINC or SNOMED.  

        

2 – There is a need to map LOINC codes to the current CDISC Lab Test metadata

• FDA Requirements:  How will the FDA be using the LOINC code?

• Presentation of the ALBUMIN CDISC/LOINC pilot mapping and findings     2015-04-22_LB_Test_Mapping.xlsx

• CDISC LAB TEST/LOINC mapping:  What is the best way to accomplish this task?

• CLARIFICATION:  the only reason that there were so many individual rows in the Albumin spreadsheet example is because of ease of loading into SHARE. 

• Proposal – only populate the following columns

  • Analyte, Specimen Type,  LOINC Property

    • Only populate the method when it is absolutely necessary or when it is particularly specified in LOINC

  • LOINC code is not needed for everything

• There are two needs for this mapping:  Therapeutic Areas and SHARE

• Therapeutic Areas (TA)

  • Requests received from the TA will have method only if the Subject Matter Experts think that it is needed.

  • Use abbreviated mapping version for the TAs

• SHARE

  • Use the same abbreviated mapping version as what is used for the TAs

  • BUT some sponsors feel that the best value SHARE content would contain all of the separate fields populated in a SDTM-compliant dataset such as location, method, and unit codes.

  • For information to be loaded into SHARE, NCI c-codes are necessary.

• LOINC codes can be added but why does it have to be the driving force on what lab metadata goes into SHARE?  SHARE will be used by everyone, not just the US.

• Instead of the Therapeutic Areas asking for subsets of specific lab tests, it would be better if the lab metadata already existed for a number of commonly used lab tests.  The Therapeutic Areas would then just have to pick and choose what they wanted.

• Additional comments after the meeting:

  • JNJ:  Why make a distinction between mass and substance concentration in LOINC?  The only time the unit type is important to call out, is when a test is measured in two non-interchangeable units…..as an example…bone specific alkaline phosphotase isoenzyme can be measured as an enzyme activity or mass concentration….these are not inter-convertible.

    • If there was no distinction on method, then it would make sense to have a unit type as a differentiator.

  • Covance:  We need to speed-up the SHARE definition process but LOINC will do so only if we learn to cluster certain CDISC concepts such a UOM codes into less granular concepts, such as the LOINC Property.

  • AstraZeneca:  From a sponsor’s point of view, the best value received from SHARE for lab tests would be to provide guidance for the most common location, most common specimen type, most common test method, and most common valid lab units.

    • Concerns over LOINC being US-based, not global

    • Rather than concentrating on tests performed for specific TA’s, we should view the bigger picture.  Perhaps the CDISC Lab team’s time would be better spent ensuring that the 200-300 most commonly performed tests across all TA’s are worked up in SHARE.  Consider enlisting a consultancy service to help since all the Lab Team members have day jobs.

3 – Comparison of CDISC Lab Unit terminology and UCUM

• Background:  History of the Unit codelist.  Why was the decision made to not use UCUM?

  • CDISC and UCUM: Why CDISC does not point to UCUM

    • document prepared by the CDISC Lab Terminology team in 2012     CDISC and UCUM_2012-12-20.docx

• This was not discussed at the meeting.

4 – Assigning the UCUM term to the current CDISC Unit terminology

• Is there an FDA requirement for UCUM?  How will the FDA be using the UCUM term?

• The CDISC UNIT/UCUM mapping task started in October 2014 (Package 19) with Erin Muhlbradt (NCI/EVS) and Jozef Aerts.  It was updated for P20 and is currently being updated for P21. 

  • Updated with each quarterly terminology release 

    • Currently stored in NCI/EVS and will also be accessible on the Controlled Terminology webpage with all of the other terminology mapping documents.     SDTM Terminology 2014-12-19_UCUM Mapping.xls

• Discuss mapping documents produced by Jozef Aerts and Diane Wold.     Diane_Wold_Units by Kind of Property.xlsx    

• Is there an FDA requirement for UCUM and how will the FDA be using the UCUM term? 

  • This was not discussed at the meeting

• The team was OK with the CDISC UNIT/UCUM mapping that is being done by NCI/EVS

• The LOINC Kind of Property, identified on Diane's mapping spreadsheet, will be a useful association to include, where appropriate, in order to further subset the existing unit of measure terminology

• As a first step, the available types of LOINC Kind of Properties on Diane's file is too long and needs to be abbreviated.  Decision made to start with a smaller set.