Much ADaM about Nothing – a PROC Away in a Day

CDISC is rapidly becoming adopted as the data standard for clinical trials and submissions of clinical trial data to the FDA. Within CDISC, ADaM datasets are an integral part of clinical study analysis and require significant data derivation to fulfill the needs of TLF provision. Can the creation of ADaM datasets be automated? Yes! This paper proposes an Excel driven solution which greatly improves the efficiency and accuracy of ADaM dataset creation by defining ADaM dataset structure, complex variable derivation and data checks within Excel. Using a library of SAS macros, the definitions then drive the automated production of SAS scripts which produce analysis ready datasets that meet the ADaM specification and which can be validated in accordance with regulatory requirements. Additionally, data check reports are created for data management to speed up the data cleaning process. INTRODUCTION CDISC ADaM version 2.1 is now released on the CDSIC website. This document describes the fundamental principles of the ADaM analysis dataset and the design and purpose of submitted analysis datasets. The main principle of ADaM datasets is “analysis-ready” so that only minimal programming effort is needed to achieve the statistical results. More than one single statistical procedure may be required to calculate the statistical outputs, but the goal of ADaM is minimum programming effort and maximum concentration on the results. The ADaM Implementation Guide v.1.0, which is posted on the CDISC Website, describes in detail the standard structure of ADaM datasets, the variables they contain and the derivation methods used. Using metadata to define the structure and derivation methods of ADaM datasets greatly facilitates the automation of ADaM dataset creation directly from CDISC SDTM datasets. This paper outlines the fundamentals of our approach and describes a step by step ADaM dataset automation process.