Clinical data comes into electronic data capture systems in many different formats. The challenge for those working with the study data is to find the best way to map all that data to CDISC’s Study Data Tabulation Model (SDTM) format.
By Jeanna Radick, CDISC Expert, CSC
CDISC has done an excellent job with its implementation guides. The SDTM guide provides extensive information and is vital reading for anyone working with SDTM data. However, there are still gray areas and the only way to know how to maneuver through those trouble spots is through experience.
Every sponsor will have different expectations with regards to their study data, so it’s important to assess the circumstances, weigh them against past cases, including what the FDA has accepted in the past, and make a decision that best fits the current circumstances.
A common experience companies have is how to map unscheduled visits and unscheduled assessments Very often a patient will come for a scheduled visit, but during that visit the investigator conducts a particular lab test that hadn’t been scheduled. That will go into the raw data as unscheduled and it will appear as though the visit is unscheduled, so a decision needs to be made how best to treat that data.
During just such a situation with one company I was helping, the statisticians wanted to be able to identify those particular labs as unscheduled, but the sponsor wanted the visit to show as scheduled. So we needed to find another variable within the SDTM to define that this was an unscheduled assessment.
A very similar situation occurred with another company, but in this case the company wanted the unscheduled assessment to be defined as an unscheduled visit because the company wanted to represent the data as it was captured in the Electronic Data Capture tool. In this circumstance, it was a better fit to call it an unscheduled visit even though technically it was simply an unscheduled assessment.
The SDTM guide allows for interpretations with how to deal with such situations. There are some situations that the guide might not address, and other issues where the implementation guide provides several different approaches and leaves the decision to the sponsor. While this flexibility is what makes SDTM such a useful standard, it can also be challenging, and it’s particularly difficult for anyone who hasn’t had prior experience with mapping raw data to SDTM. Through many years of working with SDTM, you learn the best approach for a particular circumstance.
Errors and Warnings
Another challenge is that once a study is completed, you have to run a CDISC validation check. When data doesn’t map correctly to CDISC format, you’ll get an error or a warning and you either have to fix these or explain in the study data reviewer’s guide why you weren’t able to comply with CDISC. This is the reason why it is so important to implement standards at the start of a study, utilizing CDASH (Clinical Data Acquisition Standards Harmonization) standards so that you do not run into these CDISC non-compliance issues.
One situation I’ve been working on concerns the adverse event (AE) form. In the AE domain when the variable AESER = “Y”, one of several other seriousness criteria variables must be met, for example; the patient was hospitalized, the AE was life threatening, the AE resulted in significant disability, the patient died, or several other medically important events occurred. However, the particular study I’ve been working on failed to define these different variables, which prompted an error in CDISC validation. In one case, I noticed that death data had been collected so I updated the SDTM to create the variable AESDTH = “Y”. This addition helped remove a couple of the records relating to this error, but unfortunately it could not remove all the errors. Since the other data was not collected on the case report form (CRF) and there is no way to change the CRF after the fact, the next step was to explain in the reviewer’s guide why the report doesn’t comply with CDISC.
From previous experience, I know that the regulators have accepted this explanation, but whether they will continue to do so is uncertain. As of October 2016 for Japan’s PDMA and December 2016 for the U.S. FDA, SDTM will be mandatory. From that point, it’s possible the regulators will no longer accept these types of “exceptions.” That’s why it is vital that companies understand SDTM, the rules, and the types of errors and warnings that will come up, so they can address these issues when they collect their data, not after the fact.
With the mandatory requirement just months away, you want to make sure you are addressing your trial data and how you will map raw datasets to SDTM now. If you know what will be required and collect the data correctly from the outset, it will make the transition to the standards smoother.
Learn more at our live webinar, CDISC Clinical Data Standardization, on June 28.