PointCross has reviewed numerous SEND datasets prepared for test submissions to the FDA and has worked with the FDA on their KickStart program to ensure that received datasets are suitable for pharm/tox review. What we are observing is that creating a dataset that is technically compliant (i.e., one that passes the validation rules) is a low bar – it is necessary, but not sufficient to meet FDA review requirements.
In our experience, sponsors and CROs are having a much more difficult time ensuring that SEND datasets have no data quality issues that impact “fitness for review”. We have routinely observed cases where SEND datasets are inconsistent with the accompanying study report and have insufficient data for pharm/tox review despite the fact that these data often have been collected and tabulated in the study report. These issues are serious and can result in delays in the review and approval process.
These issues are serious and can result in delays in the review and approval process.
Technical Compliance Issues
Violations of the CDISC SEND standard or FDA Specific SEND Validation Rules in a SEND dataset prevent it from being loaded into the FDA’s NIMS. For example, a critical piece of missing information can disrupt data loading. We believe that simple formatting issues can be resolved with a little experience and self-run validations. Some examples of technical compliance issues beyond simple formatting that we have observed include:
- Syntax errors in datasets or Define.xml that prevent validation or loading into FDA’s NIMS
- Coded values that are not defined within the SEND dataset
Data Consistency, Quality and Sufficiency Issues
Being technically conformant to the SEND standards should not be the only concern for sponsors when creating SEND datasets. Frequently, we are seeing that the greatest challenge for companies is ensuring the SEND dataset is consistent and sufficient for FDA review.
These types of issues may be the result of relying on two separate processes to generate the study report and the SEND dataset. Because data typically originates from multiple LIMS systems with proprietary data models, sponsors or CROs creating SEND datasets are required to reconcile disparate terminologies, units, groupings, and coded data sources. As a result, there can be inconsistencies between the study report and SEND datasets.
If FDA reviewers see a signal of interest or calculate a group summary that does not match what is reported in the study report, these discrepancies will result in reviewers questioning the trustworthiness of the dataset. The review process may be disrupted until such data issues are resolved. In cases of extreme discrepancies, a submission could be put at risk