Best Practices in ISS/ISE: Programming Perspective

Integrating data for submissions can be a laborious and difficult task, even with CDISC data, it can be a mammoth undertaking but when we add legacy data in the mix then it can make things far more difficult.

One of the ways of ensuring your approach to integration is effective is to have a good understanding of how the integration effort can be achieved. It is imperative to choose the right strategy for integrating your data from the start, thinking about the future projects you may have, or just focusing on the current effort.

On the 30th of July, we hosted a complimentary webinar titled “Best Practices in ISS/ISE: Programming Perspective” to share the most effective strategies for integrating data and discuss how to decipher what strategy to use.

The webinar was hosted by Caroline Gray, Director of Statistical Programming & FSP Partnerships of CROS NT. With more than a decade of experience in Life Sciences, having begun her career as a SAS programmer in a CRO. She is an industry recognised expert in CDISC (SDTM) and data analysis as well as a specialist in Six Sigma and Lean Methodology and Optimisation.

Continue reading to discover the questions from the webinar’s Q&A session answered by the speaker.

It can be tempting to skip the GAP Analysis due to the tight timelines associated with ISS/ISE. What are the risks associated with skipping the GAP Analysis?

You do not always have to do a GAP analysis especially if you are not creating SDTM datasets or you are integrating a small number of studies. The aim of the GAP analysis is to identify not only differences but also similarities across studies so the main risk with not performing the GAP analysis is that you go into your data integration blind and unaware of the state of the data. This can lead to you spending more time looking for similarities and differences during the harmonization process instead of having it available in one area for quick reference.

Should you create integrated datasets for data that are not directly involved in the ISS/ISE?

Focus on what you need but if you have time and plan to standardise your data centrally in the future then it may be an option, but it is not necessary to do the integration if not included in the ISS/ISE

When should the integration strategy be agreed with regulatory agencies?

This should be agreed as early as possible before any programming activities are stated, ideally at a pre-NDA meeting.

How long does a GAP Analysis take to complete?

This depends on how many studies you are going to harmonise but in general you can have two approaches to this: 1. Perform a GAP analysis on all available data which would take around 8 hours per study 2. Perform a GAP analysis on ISS/ISE-related data only which would take around 4 hours per study.

What are your thoughts and experience on the previous lack of industry-wide standards for integrated ADaM, and the proposed Integrated ADaM IG in development by CDISC?

There is a lack of guidance on ISS/ISE integration, however, this is getting better and PhUSE recently released a white paper on this topic which is an excellent resource. It would be great to have CDISC guidance on this topic also and I welcome an implementation guide that covers the difficulties and intricacies of integrating ADaM datasets.

Watch our free on-demand webinar “Best Practices in ISS/ISE: Programming Perspective”

Should all data be CDISC compliant?

Any trials completed after the FDA deadline will need to be in CDISC format. Discuss with the regulatory authorities to confirm or review the FDA recommendations for submissions.

What is the strategy need to follow to harmonise VISITS if integration is done at SDTM? For example Study 1 has Visit as “Cycle 1 Day1”, study 2 has Visit as Week 1.

Using intermediary integrated SDTM datasets may be useful in this instance as you can harmonize all the visits in one SDTM database. For visits, this can be difficult and a GAP analysis along with proc freq can help identify the differences between studies allowing you to implement some consistency across studies. You can also harmonise the visits within the integrated ADaM datasets modifying AVISIT and AVISITN but there is less transparency between SDTM and ADaM this way.

What is the biggest challenge when integrating data?

There are many challenges with integrating data but the main one and the one that will have the most impact on your teams is timelines. Generally, submissions have strict deadlines and ISS/ISE timelines can be very aggressive which may mean larger teams, working at risk i.e before a SAP is finalised or doing tasks in parallel.

What approach would you suggest if we have roll-over subjects in ISS (for ex: Subject 01 is participating in study 01 and the same subject is participating in study 05 with the different subject). What would be the best approach to follow?

USUBJID should be unique within the integrated datasets and throughout the process but it can happen especially with roll-over subjects. If the subject has rolled over from one project to another then we would expect just one USUBJID in the SDTM. We would also expect just one record in ADSL so you would choose which one is to remain and document this in the integrated analysis/study data reviewers guide. You could also discuss this and the approach that you would take with the regulatory authorities.

How do you ensure you can meet the strict timelines for an ISS/ISE?

You can use multiple strategies for this but in the majority of cases, you will need to scale up your team to meet the demanding programming effort. This will mean choosing a strong experienced lead programmer to coordinate activities, have regular communication and status updates with the team, and monitoring progress carefully. You can also assign sub-teams and datasets by type which will centralise the effort.

Which strategy do you think of when you have some of the data in pre-CDISC and some data in post-CDISC form?

Considering your source data for an ISS/ISE is in a mixed format, my suggestion would be to utilize an integrated ADaM approach directly from the source data that is if the source data does not need to be converted to SDTM (this will need to be verified with the regulatory authority).