Wednesday, July 10, 2013

Is CDISC Controlled Terminology going the wrong way?

Today, I downloaded the latest (June 2013) CDISC controlled terminology (CDISC-CT) from the NCI website.
I am already communicating (well, it's more a one-way communication) with the CDISC-CT team for about a year now, as I do have strong concerns about the direction CDISC-CT is taking.

As you might know, CDISC-CT is just a list. It does not have a system (unlike LOINC), nor it is a taxonomy (like SNOMED-CT). It's just a list, with all imaginable problems.

Here are my major concerns:
- the CT does not show at all that there is a 1:1 relation between --TESTCD (test code) and --TEST (test name). There are separate lists for each of them, as if these were unrelated.
- CDISC-CT reinvents the wheel: some codelists like those for lab tests, units of measure, and microorganisms are unnecessary as there already exists well established CT used by healthcare. For lab tests there is LOINC, for units there is UCUM, and for microorganisms there is the "International Code of Nomenclature of Bacteria" (ICNB). All these have a system or are taxonomies, so why try to replace them with something basic like a simple list? Does CDISC-CT suffer under the "not invented here syndrome"?
Some within the CDISC-CT team claimed that their codelists are subsets of the mentioned ones. I checked, and it simply is not true! Others told me the CDISC-CT can easily be mapoed to the mentioned ones. I checked and it simply is not true! Due to the system (or better: lack of system) CDISC-CT is using, these lists will grow to almost infinity, and the length of the latest version (>7000 terms) strengthens my opinion on this. I heard of one sponsor who submitted over 500 "new term requests", only for lab tests.
So why not use something (LOINC, UCUM, ICNB) that is used by everyone in health care and that has proven its merits?

There are also some peculiarities in the CDISC-CT list that draw my attention. What to think of the following:
- the unit of pH is ... PH!
In my first year at high school (well, I do also have a PhD in chemistry) I learned that pH is a property that is dimensionless, i.e. it does not have a unit.

- bar: from the same teacher (my chemistry teacher was also my physics teacher), I learned that "bar" is a (even SI) unit for the property "pressure". According to CDISC-CT "bar" is defined by "A dosing measurement based on the bar unit". Do you understand what is meant here? I don't. Is this a circular definition (also not a definition). Is "bar" meant for something like "a bar of chocolate"? If so, is that a unit? Is a bar of chocolate the same amount in Belgium as it is in the US or Tokyo?
Under "unit" I understand that it is the same amount everywhere in the world (well, a "gram" is).

- beats/minute: this is a more subtle one. The definition in CDISC-CT is: "the number of heart beats measured per minute time". Makes sense, isn't it? No: it does NOT. First of all, the definition - is "beats/minute" limited to "heart beats"? I don't think so, I usually also use it for drum beats.
Secondly, the CDISC-CT mixes up the property (heart beats) with the units. A heart beat is measured as a frequeny, and some popular units for frequency are "/min" and "/sec".

- "bottle", "box", "can", "capsule", ... Is a "bottle" a "unit"? According to CDISC-CT, it is. I have a lot of bottles at home, and most can contain different amounts of liquids.
CDISC-CT is mixing up the concept of "object class" (in this case the parent class is "container") with the concept of "unit" (unity amount that is the same everywhere in the world).

I shared all these concerns with the CDISC-CT already more than half a year ago, but they just do not seem to care. The CDISC-CT-LAB subteam promised me to write up an answer to my suggestion not to define their own codelist, but to use the existing LOINC standard instead about 5 months ago, but I still haven't received it (though I send several reminders).

Do not misunderstand me - I am not against the development of CDISC controlled terminology! We really need CDISC-CT to be all "on the same wavelength" and especially to be able to compare values between different submissions of different sponsors.
But if we do, we need to do it correctly, and not try to reinvent the wheel. Nor should we forget our basics in physics, chemistry and especially in semantics, just for the sake of trying to controll everything.