Preparing ADaM: Why data traceability is crucial to the review process

When companies submit their clinical data to the regulatory authorities, reviewers not only want the raw data properly mapped and organized — as defined by the submission standard Study Data Tabulation Model (SDTM) — but they also want to be able to trace the raw data to where it originated.

This is where the standard known as ADaM (Analysis Data Model) steps in. The data standards body Clinical Data Interchange Standards Consortium (CDISC) introduced the ADaM standard specifically to assist with analysis of the data, as well as to enable data traceability.

When U.S. Food and Drug Administration (FDA) reviewers are assessing and analyzing the results of the datasets, ADaM allows them to trace the flow from the study data tabulation data to the analysis data. Along with these ADaM datasets, reviewers also need the metadata, which describes what is being provided. The metadata comes from define.XML and the analysis data comes from ADaM. Together, these standards give the reviewers assurance that the data they are seeing is genuine. Simply put, ADaM makes it easier for the reviewers to do their job.

To assist with the analysis of data, it needs to be formatted into tables, listings and figures. What makes ADaM so invaluable for analysis purposes is that it provides the means to quickly and simply create such tables, listings and figures and to conduct a thorough analysis of the data.

To carry out a statistical analysis of the study, data and metadata will be derived from many different parts of the study, such as data about the study subjects, efficacy data, safety or adverse event data, and so on.

The role of programmers

When creating the analysis dataset, programmers must include the subject-level analysis dataset (ADSL), which has one record per subject and contains core variables that are specified in the ADaM implementation guide, including demographics, key dates, stratification and subgrouping variables, and so on.

The other data structure included in the creation of the analysis dataset is the Basic Data Structure (BDS), which has one or more records per subject, per analysis parameter, per analysis time point. The dataset includes a central set of variables that describes the analysis parameter PARAM with related variables, defining type of data, baselines, visit timing, etc.  Sometimes the variables are derived from other variables — for example, if data is missing for a period — in which case that derivation is defined by the DTYPE parameter, for example, last observation carried forward (LOCF). Time-to-event (TTE) analysis domains generally used in therapeutic areas such as oncology are also BDS domains.

All pharmaceutical companies are required to submit ADaM datasets for analysis of information so that when the final report is submitted to the FDA, reviewers can quickly pull out information and conduct their analysis. It speeds the process for the reviewers, providing them with the assurance they need, and consequently hastens the review process, which is to the benefit of pharmaceutical companies seeking to bring their products to the market.

Nevertheless, creating the analysis dataset from SDTM to ADaM and carrying out the programming is not typically a core competency of pharmaceutical companies, and it seldom makes sense for them to keep CDISC experts and programmers on staff. This is where working with a partner with experience with CDISC and biometrics makes sense and can ensure that companies meet the ADaM requirements established by the FDA.

Han Zou is biometrics manager and Rao Mandava is CDISC lead in DXC’s Life Sciences group.

Speak Your Mind


This site uses Akismet to reduce spam. Learn how your comment data is processed.