ACDM 2024

Bookmarking the aCRF –

automating the manual labour

by Ramon Jakobs (with Kimberley Santing) 

An integral part of any study submission to FDA is the annotated CRF. This annotated CRF must also be bookmarked. Bookmarks providing a concise overview of all forms and visits with hyperlinks to the associated page in the CRF for easier reference. Adding these bookmarks is often a manual, labour-intensive task that is not only tedious but also very prone to errors and inconsistencies.

The bookmarks are effectively a single list split into two parts: 1) form then visit, and 2) visit then form. Rather than in a list, these bookmarks can be represented in a two-dimensional matrix, or a spreadsheet. This matrix is the basis of our solution.

This presentation will present a solution which significantly reduces the time and effort required to bookmark a CRF. It revolves around preparing the bookmarks outside the CRF (i.e. outside the PDF) and in a spreadsheet which resembles the schedule of assessments. A simple program (SAS, R, VB) transforms this information into an XML file that can be imported into the CRF and that’s it – your aCRF is now ready for submission!

Mapping Engine:

An Approach to Standardising SAS Programming in Clinical Trials

by Kai Wanke (with Bas van Bakel)

One of the difficulties when working with clinical trial data is the variability of source data generated in different trials or by different vendors poses one of the major challenges faced when working with clinical trial data. Often this is the case even when the same EDC system is used throughout.

To ensure interoperability data is expected to be converted into standardised models before it is analysed and submitted to regulatory bodies. This process presents challenges since it can be time-consuming and may introduce a degree of variability, for example between data generated between different study teams.

To address these challenges, we have developed a data conversion framework. This framework is a suite of SAS macros that uses pseudo-code stored in a spreadsheet (called ‘mapping specifications’) to generate SAS programs that perform the conversion of raw data into specific models, such as SDTM or similar data models which may be company-specific. This framework is referred to as Mapping Engine.

As part of its specifications, Mapping Engine consumes metadata to ensure all its created datasets and variables adhere to the specifications of the selected model such as data types, labels, formats, et cetera.

One of the benefits of working with Mapping Engine is that its specifications are organised around individual source datasets and variables rather than complete trials, making it highly transferable and reusable between files, users, and trials. Moreover, this transferability allows the creation of pre-defined libraries of mapping specifications that are tailored around a specific vendor, therapeutic area, or compound, and describe how the data should be processed.

Using SAS code, it is possible to automatically match the raw data against the mapping libraries, automatically generating an initial version of the specifications that can be used to convert the raw data into the desired model. This results in the automated, metadata-driven conversion of clinical trial data for up to 90% of items in case report forms.

Mapping Engine allows to minimise the manual effort needed during data conversion, while at the same time enforcing programmatic standards and reducing variability between different studies.