2016-10-26 Lab Terminology Team Discussion on Codelist Rules

Some team members are questioning the need for the two rules in red below.  If a majority consensus of the lab terminology team agrees that these rules should not exist then the lab terminology teams decision and comments will be taken to the other terminology teams for a vote.

Email comments have been entered below the rules.  Please read all of the comments and then vote in the table at the end of the discussion items.  If you have any additional comments, please use the comment section below.

The next row has all of the current rules for the lab terminology team.  There are also rules that apply to all terminology teams. Rules for all codelists_2016-04-20.docx

• CURRENT Rules for Laboratory Test Code/Name Codelists

• General
  • Lab tests where results can be expressed as qualitative, semi-quantitative or quantitative should all have the same test name and code.
  • Lab terminology includes lab tests for drugs of abuse, e.g., Methamphetamine.
  • Lab terminology does not include lab tests to perform therapeutic drug level monitoring, e.g., Digoxin.
  • If an instrument or lab commonly does the calculation and reports the observed and calculated value then the calculated results should have lab terminology, e.g., clearance and rates
    • Conversely, derived values, (i.e., delta, such as change from baseline) that require two or more separate collections or time points will not be included in lab terminology.
  • For erythrocyte maturation, we will use the RUBRI classification system.
  • For Allergen tests and Flow Cytometry tests, the lab test code will be the <C-Code> defined at the time of publication. 
  • Synonyms should have a one to one relationship with the test name, and also among other synonyms.
  • Within a codelist, if a term is in use as a CDISC submission value, that same term cannot also be a synonym for a different CDISC terminology concept.

• Definitions

  • The general format for absolute analyte definitions is: A measurement of the XXX in a biological specimen.

  • The general format for the differential analyte definitions is: A relative measurement (ratio or percentage) of XXX to XXX in a biological specimen.

  • For numbers less than 10, we should spell out the number.  The exception is any number less than 10 that is used in a scientific name or with a units of measure, dates, percentages, page numbers, or money.  If the number appears at the beginning of a sentence, then we should spell it out.

• Laboratory Test Names

  • Test names do not contain the specimen type. This information is populated in existing variable LBSPEC.

  • Test names do not contain units of measure. This information is populated in existing variables LBSTRESU, LBORRESU.

  • Test names do not contain methods of measurement. This information is populated in existing variable LBMETHOD.

  • Test names do not contain collection timing information. This information is populated within the various timing variables available such as LBTPT, LBELTM, etc.

  • Test names do not contain the fasting status of the specimen. This information is populated in existing variable LBFAST.

  •  For any analyte that has subtypes or subcomponents and may also be expressed as a total, the 'total' value will not contain "total" in the CDISC submission value or synonyms. Total will only be in the definition.

  •  For all differential test names, the numerator and denominator are spelled out as fully as possible (given the 40 character limitation) and separated by a forward slash.  Do not use the words ‘ratio’ or ‘percentage’, as that is made clear in the definition.

  •  For antibody test codes:  For IgG, IgM, IgA, IgE, IgD – Use all three characters if there is enough space, if we need to abbreviate use IG, IM, IA, IE, ID.

• Laboratory Test Codes

  •  Use the full analyte name as the test code for the term if the analyte name is 8 characters or less.  If there are known subtypes that are not yet published, create a test code less than 8 characters.Lab Test Code Naming Fragments:

• • Long Name                         Abbreviation                          Additional Instructions

  • acid                                    AC                                           suffix

  • active                                 AT                                              suffix

  • activity                               A                                                suffix

  • allergen                              AG                                           suffix

  • alpha                                  A or AL                                      

  • antibody                             AB                                              suffix

  • antigen                               AG                                              suffix

  • blasts                                 BLST                                          prefix

  • casts                                  CS                                              prefix

  • cells                                   CE                                              suffix

  • corrected                            CR                                              suffix

  • creatinine                           CREAT or CRT                          suffix

  • crystals                              CY                                              prefix

  • epithelial                             EPI

  • fragment                             FG                                              suffix

  • free                                      FR                                              suffix

  • <genus>                             use first letter

  • hyper                                  HPR                                         prefix

  • hypo                                   HPO                                         prefix

  • inactive                              IT                                                suffix

  • index                                  IDX                                             suffix

  • large                                  LG                                             

  • <notion of six>                   HEX

  • <species>                          use first two letters

  • rate                                     RT                                           suffix

  • receptor                              RC                                             suffix

  • toxin                                   TXN                                           suffix

• • If there is a roman numeral in the submission value and, following the above rules, the testcd value would exceed 8 characters then substitute a numeric digit for the roman numeral

  • For all differential test codes the absolute count is a short defined term and the ratio/percentage contains the same short mnemonic for the numerator followed by a second short mnemonic for the denominator. There is no forward slash in the testcd.

  • • abbreviations

      • eosinophils                               EOS

      • basophils                                  BASO

      • lymphoma cells                         LYMMCE

      • reticulocytes                             RETI

      • erythrocytes                             RBC

      • leukocytes                                LE

      • lymphocytes                             LYM (core test name), LY (denominator)

    • examples

      • eosinophils/leukocytes              EOSLE

      • reticulocytes/erythrocytes         RETIRBC

      • lymphoma cells/lymphocytes     LYMMCELY

      • lymphocytes/leukocytes              LYMLE

  • If there are multiple tests with the same base (core concept), the base text goes first and the qualifiers are suffixes, (e.g., Bilirubin, Direct Bilirubin, Indirect Bilirubin) BILI, BILDIR, BILIND.  This will then allow the test codes to sort together.

  • For any analyte that has subtypes or subcomponents and may also be expressed as a total, the components will have LBTESTCD named as the base analyte plus a suffix while the total value will be the base analyte code alone with no "total" in LBTEST - total will be in the definition, e.g., Bilirubin (BILI), Direct Bilirubin (BILDIR), Indirect Bilirubin (BILIND).

  • Do not use acronyms with multiple meanings for the test code.

• Rules for UNIT Codelist

  • A non-standard unit term, (e.g., cells or creatinine) that is contained in the test term will not be repeated in a unit of measure.

  •  Unit terms are mixed case and case sensitive and generally follow UCUM abbreviations where appropriate.

  •  The carat symbol is used to denote powers.

  •  The asterisk symbol is used to denote multiplication.

  •  The forward slash is used to denote division.

  •  Use the ‘left to right’ rule for order of operations. 

  •  Parentheses are only used when absolutely necessary, i.e., in cases where the order of operation does not follow the left-to-right rule and must be made explicit.

  •  No superscript capability; squared or cubed values will be the base unit code with 2 for squared or 3 for cubed, e.g., m2 (square meter) and m3 (cubic meter).

• Rules for Specimen Type Codelist

  • Submission value should be singular.
    

Email Comments from 26Dec2016 to 28Dec2016:

Erin: I don't agree but will go along with team consensus if I am in the minority. My thinking has changed on this over time; when this rule was added originally I didn't have a strong negative opinion against it. My reasons for dissenting are below, which were briefly discussed on Wednesday's lab team call (for those who were not present):

 1. I do agree that two distinct concepts within a single codelist cannot have the same submission value. This would break datasets since the novelty of a submission value ensures that reviewers are filtering or sorting on the exact same concept. The presence of a term in the synonym column that happens to be identical to another submission value does not break anything (at least in my mind). Synonyms are there to help in mapping. If there is a chance that two distinct lab tests can be known by the same abbreviation (e.g. ESR is perfectly valid for both Erythrocyte Sedimentation Rate and Estrogen Receptor), it would actually be a GOOD thing if the synonym was present. If a lab test of ESR was submitted by a local lab, the presence of ESR in two places would force the sponsor to make sure they are assigning the appropriate submission value for the test that was actually done. The removal of synonyms may in fact lead to errors in coding in this instance.

2. The creation of 40 character and 8 character TEST and TESTCD values is onerous. The downstream affect of this rule is that we must now work extra hard to ensure that whatever we assign as a TESTCD is not a valid synonym to something else. Additionally we will lose perfectly good /valid synonyms if we adhere to this rule. 

3. When assigning submission values for CDISC terms, some codelists have business rules that disallow the most appropriate term to be used. The LBTEST/CD codelists fall into this category since the 8 and 40 character limits are artificial and force the team to use the name of the analyte as the TEST/CD instead of the more appropriate concept of 'Analyte Measurement'. Submission values are aliases for a concept only; the presence of an artificially constructed TESTCD that happens to be the same as a synonym elsewhere doesn't actually matter from a semantic perspective.

I will happily go along with what the majority agrees to however I did want to make my dissent known. I don't believe that this actually 'breaks' anything in terms of terminology usage but this rule does add a burden to the team when assigning TEST and TESTCD values.

Audrey: I agree with Erin especially for the following “Synonyms are there to help in mapping. If there is a chance that two distinct lab tests can be known by the same abbreviation (e.g. ESR is perfectly valid for both Erythrocyte Sedimentation Rate and Estrogen Receptor), it would actually be a GOOD thing if the synonym was present. If a lab test of ESR was submitted by a local lab, the presence of ESR in two places would force the sponsor to make sure they are assigning the appropriate submission value for the test that was actually done. The removal of synonyms may in fact lead to errors in coding in this instance.”

Deb Rittenhouse: I was not at the meeting where this was discussed. My agreement was from the perspective of a user who may be naive and may select the wrong CD based on the first synonym they find. If the team feels there are reasonable times when more than one CD would have same Synonym, then this would need to be called out and not left for the user to come across by chance. Some users who are selecting the proper code are programmers and not experienced lab professionals.  I value your critique. What is missing from all code lists is comments to the user about how to use the values. The definition alone is not enough; it does not provide context. 

Audrey: We search on synonyms all the time based on synonyms that are in our current glossaries.  If the standard synonyms are not included in the synonym list it is very, very hard for the data mappers to do this job especially for labs.  I think ESR is a great example where our glossaries have the term ESR and if we were just to see this ESR as a synonym for " Estrogen Receptor" our data mappers would have most likely mapped it to " Estrogen Receptor", but it should really have been mapped to "Erythrocyte Sedimentation Rate".

If we were to see two entries for with the same synonym we would definitely ask more questions before mapping.  I think limiting the synonyms causes a much greater chance of mapping errors.

Joyce: I concur with Deb!  Having this documented would result in a more consistent conformance to our standard.

There are some of us who would like to automate the terminology process by interrogating the synonym list and selecting the best preferred term (test code, test name submission value) via software and not rely on human intervention.

In the interest of automation, the more consistently formatted guidance we can provide in  terminology the better.

Erin, I agree with your point number 1 assuming that what you mean by "submission value" is the column in the terminology spreadsheet which is used to designate the actual TEST or TEST CODE value and not the result.

Point 2 and 3 seem to be stating that the rule to limit test codes and test names to 8 and 40 characters respectively are causing challenges when assigning terminology values.

This issue is really the result of the way statistician at both pharma and the FDA pivot their analysis tables so that SDTM rows become actual columns.

Having a column of 8 characters is easily manageable when representing it in a report.

However, a column name that exceeds 40 characters is not easy to manage in a report.

Remember that the reason that our test codes across our SDTM Model is limited to 8 characters is to support analytical reporting that pivots those test codes into column names in a report. This was not done in an arbitrary manner but with this use case in mind.  If we wish to revisit this decision, all teams (especially ADaM) must be represented.

Nik: As a representative of a large Pharma that implements CDISC standards, we do not use controlled terminology off the shelf. We need to assign categories, and link to specimens and methods to create unique test concepts (define as a unique combination of LBTESTCD, LBCAT, LBSPEC and LBMETHOD). In doing this, we need to create our own sponsor dictionary…which we also use to stored sponsor defined test concepts. We actually manage our own list of synonyms also in our dictionary. We also need to manage unit and unit synonyms, conversion factors, standard units and link all of these to the test concepts. Do you know that labs have found so far, 89 different ways to represent the concept 10^3/uL as a raw unit…which we then theoretically need to represent as submission value 10^9/L (although we still do not comply and keep ORRESU mapped to the 10^3/uL (for traceability reason))….but I digress….one day I will climb down of that hobby horse.

I would say that CDISC a responsibility to provide terminology that is as easy to “unambiguously” consume as possible, and for that reason, I tend to agree that the best solution would be to either

(1)    Allow same synonyms for different tests in the same code list. ESR was one…..some others we have come across internally EGFR, TRAP.  

(2)    Or create some additional documentation describing “ambiguous” terms.

Or both

With more than 150 trials running concurrently, using a mixture of central and local labs, different DM CROs, internal and external data managers (the vast majority of which have less lab data experience than those on this team), I cannot hand on heart say that on some trials, mistakes are not being made. All we do as an organization, is try to provide training, documentation, programming checks and intelligent design to minimize the ambiguity and risk as much as we can.

Bernice (response to Erin's email comments): I do understand your concerns, Erin, but I feel that the rule is necessary to avoid confusion.  The end users are notifying me when they find this situation.

 Once the SAS transport file is changed or not used anymore then we will not have the 8 and 40 character limits.  Or we might someday use C-code for the test codes.  But for now I think that we should keep this rule.

Also, this rule applies to all terminology teams and all of the teams did agree to add it.  It also completed public review.

If the majority of people on the lab team believe that we should revisit this rule, I will then take it to all of the terminology teams for discussion and another vote.

VOTING - please answer yes or no in the table below,

additional comments can be written in the comments section.

Assume that Erythrocyte Sedimentation Rate and Estrogen Receptor do not yet exist.  Possible example where the acronyms are included in the synonym column and the testcd submission value does not equal the synonym.  But the testcd submission value does include the ESR as part of the testcd.  Any search on ESR in the synonym or submission value column would give you the information that you need.  The current lab abbreviations for rate is RT and for receptor is RC.

In this example, the need for the exact same acronym appearing in multiple concepts exists.  Also, the testcd submission value does not equal the synonym of a different concept.  I think that this scenario would solve everybody's concerns.  Do you agree?

2016-11-16:  This topic was discussed again in the lab terminology team meeting.  Here are the decisions:

• New Rules for lab terminology:
       1) For any controlled term, synonyms should have the exact same meaning as the submission value.  For any controlled term, synonyms should have the exact same meaning as the other synonyms for that particular term.
       2) Acronyms can be used as synonyms, and multiple concepts within the same codelist may have the same acronym as a synonym when it is appropriate.
       3) Within a codelist, if a term is in use as a CDISC submission value, that same term can also be a synonym for a different CDISC terminology concept.

The team agreed that they did not want to follow the existing rule below which is for all controlled terminology teams:

         4)  Within a codelist, if a term is in use as a CDISC submission value, that same term cannot also be a synonym for a different CDISC terminology concept.

The people who agreed to the above new rule # 3 and to remove #4  are:  Erin Muhlbradt, Phil Pochon, Anna Pron-Zwick, Sue Steen, Deb Rittenhouse, and Nik Pemble

The people who disagreed to the above new rule # 3 and who wanted to keep rule # 4 are:  Deb O'Neill and Bernice Yost

This information has been sent to Barrie Nelson for comments.