This is a question that has continually arisen since organizations first started adopting SEND. If we just add all the SEND Controlled Terminology (CT) to our preclinical lexicons, then surely it makes SEND dataset generation automatic? It seems obvious, right? Well actually, this is a far more subjective topic than you’d imagine. So, in today’s blog I want to look at the pros and cons, things to consider, and offer my personal opinions on the topic.
The first point to consider concerns the structure of the libraries of the nonclinical data collection system. As these systems predate SEND, they don’t always have the same data structure to allow them to completely match up with the various SEND codelists. Think of the specimen qualifiers like Directionality and Laterality. Not all data collection systems split these out into separate libraries. The same may be true for the qualifiers for the finding like the Chronicity and Distribution. There are various other examples too. So, there are likely places where there isn’t an exact fit, but to be honest, these are only a few outliers, but important ones to consider.
Another point to consider is that the SEND CT is updated quarterly. Now in truth, relatively few terms are updated with each release, but there are changes that need to be assessed with each new CT version. Though the impact is usually minimal, there is the potential for significant frequent changes. Do we want to open ourselves up to the possibility of needing to make frequent changes to our data collection lexicons?
Another concern that I often hear on this topic is that many of the individuals who are collecting and/or reporting study data do not like the SEND terms. For example, some pathologists express strong opinions about not wanting to have to use the SEND terms. Also, with a few notable exceptions, the majority of SEND CT codelists exclusively use terms in uppercase. This doesn’t always suit data collection screens or reports.
Also, many CROs conduct plenty of studies that don’t make their way into SEND. This could be because they are studies for the chemical industry, or they are study types that are out of scope for SEND for some other reason. Such organizations may face obstacles when trying to introduce such a significant change for only a proportion of their studies.
However, the SEND datasets need to be compared back to the PDF tables in order to ensure that the electronic SEND records are an accurate representation of the study report. That’s obviously easier when they all use the same terms, in the same case.
So, there are advantages to aligning our nonclinical lexicons with the SEND CT terms, but there are potential pitfalls and things to consider too. Therefore, how much a particular organization is willing or able to align data collection with SEND CT will vary. Fortunately, SEND tools like Instem’s submit™ suite have extensive functionality to map terms in a data collection system to the SEND CT.
My personal recommendation would be to try to align the lexicons as closely as possible with SEND CT and accept that this can only ever go so far and then rely on the tools for the final translation to CT.
‘til next time
Marc
