Leveraging the Value of CDISC SEND Data Sets for Cross-Study Analysis: Incidence of Microscopic Findings in Control Animals.

Implementation of the Clinical Data Interchange Standards Consortium (CDISC)'s Standard for Exchange of Nonclinical Data (SEND) by the United States Food and Drug Administration Center for Drug Evaluation and Research (US FDA CDER) has created large quantities of SEND data sets and a tremendous opportunity to apply large-scale data analytic approaches. To fully realize this opportunity, differences in SEND implementation that impair the ability to conduct cross-study analysis must be addressed. In this manuscript, a prototypical question regarding historical control data (see Table of Contents graphic) was used to identify areas for SEND harmonization and to develop algorithmic strategies for nonclinical cross-study analysis within a variety of databases. FDA CDER's repository of >1800 sponsor-submitted studies in SEND format was queried using the statistical programming language R to gain insight into how the CDISC SEND Implementation Guides are being applied across the industry. For each component needed to answer the question (defined as "query block"), the frequency of data population was determined and ranged from 6 to 99%. For fields populated <90% and/or that did not have Controlled Terminology, data extraction methods such as data transformation and script development were evaluated. Data extraction was successful for fields such as phase of study, negative controls, and histopathology using scripts. Calculations to assess accuracy of data extraction indicated a high confidence in most query block searches. Some fields such as vehicle name, animal supplier name, and test facility name are not amenable to accurate data extraction through script development alone and require additional harmonization to confidently extract data. Harmonization proposals are discussed in this manuscript. Implementation of these proposals will allow stakeholders to capitalize on the opportunity presented by SEND data sets to increase the efficiency and productivity of nonclinical drug development, allowing the most promising drug candidates to proceed through development.

SEND harmonization & cross-study analysis: A proposal to better harvest the value from SEND data.

The Standard for Exchange of Nonclinical Data (SEND) identifies an approach for representing nonclinical data in a structured format which has been widely adopted by the pharmaceutical industry as it is required for data submission to the United States Food & Drug Administration (US FDA). The SEND Implementation Guide (SENDIG) allows for considerable flexibility in how data is represented; interpretation of these guidelines has led to significant variability in the approach to SEND dataset creation. The purposes of this manuscript are to identify common variability in certain SEND domains and to describe how variability can be managed to enable valuable cross-study analysis use cases. The example of extracting a commonly used data point, animal age, is used to illustrate the complexity and variability of SEND datasets. Developing a solution framework to the variability problem that includes all stakeholders involved in the creation and use of SEND datasets may enable future, routine analysis of warehoused SEND data. Harmonizing the implementation and use of SEND is expected to benefit all involved stakeholders and to ultimately contribute to the goal of increased productivity in nonclinical research.

FDA Engages Collaborators to Address Nonclinical Data Challenges.

FDA and PhUSE cohosted a Computational Science Symposium (CSS) in 2012 that brought stakeholders from the pharmaceutical industry and government to work collaboratively to solve common needs and challenges. A nonclinical informatics workgroup was formed, dedicated to improving nonclinical assessments and regulatory science by identifying, collecting, and prioritizing key needs and challenges in the field and then establishing an innovative framework for addressing them in a collaborative manner. This paper discusses the process and outcomes of the nonclinical informatics workgroup during the CSS and describes an approach which crossed organizational barriers to optimize computational science for nonclinical assessment.