Providing Review-Ready Results in SDTM

Findings results as originally collected may not be review ready for a variety of reasons, so SDTM has a variable, --ORRES, for the result as originally collected, and another variable, --STRESC, for the result in ready-for-review form.

Variability In Collection

In multi-center trials, different centers may report tests in different units. For example, a site in the US might report height in inches, while a site in Canada reports height in centimeters. Some sites might report a lab test like bilirubin concentration in mg/dL, while another reports it in umol/L. A reviewer needs all the results for a test in the same unit, so the preparer of an SDTM findings dataset chooses a consistent unit for each test and converts all results for that test to the chosen unit.

Tests with non-numeric results can also suffer from variability in reporting. For instance, at one site, the brand of dipstick used to test urine for white blood cells might report tests as “Negative”, “Trace”, “+”, “++”, or “+++”, while the brand used at another site might use the results “Negative”, “Trace”, “+”, “2+”, or “3+”. The preparer of the SDTM lab dataset will convert results to be consistent, perhaps using “2+” and “3+” instead of “++” and “+++”.

Scored QRS Instruments

Questionnaires and other instruments often contain questions with textual answers that have associated numeric values. For example, the instructions for answering a question might say to use a scale from 1 to 5 where 1 is “Never”, 2 is “Seldom”, 3 is “Sometimes”, 4 is “Often” and 5 is “Always.” A reviewer needs both the textual and numeric versions of the answer. The solution in SDTM QRS findings domains is to represent the textual answer in --ORRES and the numeric answer as a score in ---STRESC.

Post-collection Coding of Results

Some tests have such a broad range of results that the results are collected as free text. These tests are often open-ended tests that instruct the performer to record all abnormalities observed. Such results are often coded, using an external or sponsor-developed dictionary. In this scenario, the verbatim term is represented in --ORRES and the dictionary-derived term in --STRESC.

In the process of coding, the verbatim term may be edited to correct spelling or to remove words that are not relevant for coding. The edited verbatim term is represented in --MODIFY.

If the coding dictionary has a hierarchy, the top-level term corresponding to the dictionary-derived term in --STRESC is represented in --BODSYS.

Words that are not relevant for coding may still be of interest and can be represented in other variables. For example, the SEND Macroscopic Findings, Microscopic Findings, and Tumor Findings domains represent findings by a phathologist. The pathologist’s verbatim term is represented in –ORRES, often in the form of a text string that might include the name of the organ, a part of the organ, the severity, color, chronicity, distribution, staining, and other properties, depending on the domain. The value in –STRESC is the base process from controlled terminology, such as the SEND codelists NEOPLASM and NONNEO. A number of modifiers appearing in --ORRES (and not seen in --STRESC) can be represented in standard findings variables (e.g., --ANTREG, --LAT, --DIR, --PORTOT, --SEV, --CHRON, --DIST). Modifiers present in –ORRES that cannot be represented in standard variables are represented in the non-standard variable, --RESMOD.