FDA requirements state that an Integrated Summary of Safety (ISS) and Integrated Summary of Effectiveness (ISE) should be included in the Common Technical Document (CTD) for a regulatory submission.The word ‘summaries’, in this context, can be misleading as they are not summaries of clinical study reports of the individual studies. They are “but rather detailed integrated analyses of all relevant data from the clinical study reports that belong in Module 5”. Integrated summaries go far beyond being a regulatory requirement. They provide a valuable outlook towards the entire study.

Every study with the potential to include an integrated summary must give importance to it for these reasons:

  • Integrated summaries reiterate and validate the analytical conclusions derived in the individual studies.
  • They provide a chance to discover hidden trends and insights that could only be identified with a larger data pool.
  • Mistakes made in the individual study are highlighted when the data are pooled and analysed.

If you have any familiarity with the ISS process, you will be aware of the complexity and convoluted nature of its mechanisms. Our in-house team for Integration studies, named the Integration Summary Squadron (ISS, no pun intended), came out of a mammoth 10-month run of integrating a multitude of studies. We asked them to share some of their toughest challenges and how they conquered it. We’ve summarized some of the most important ones here:

1. Planning the execution

A significant amount of time needs to be spent in planning for the Integration in the early stages of clinical development. Planning the individual studies with Integration in mind will help identify critical endpoints of the component studies, helping build a comprehensive design for them. Focusing on the development of SDTM datasets will help create a hassle-free process for analysis datasets, and thereby for the tables, listings and graphs derived from them. Another crucial part is the need to upgrade the version of various studies to ensure their compliance to the current CDISC standard versions. Documents such as protocols, Statistical Analysis Plans (SAPs), annotated CRFs, coding dictionaries, define-xml need to be made available at the outset to avoid unnecessary delay.

2. Maintaining Consistency across studies

The most tedious part of an Integration process is reconciling the individual studies to be consistent before moving on to the integration portion. Often, studies are created with a lack of foresight on the potential implications they would have on the integration process. Standards may change drastically over the years, where entire domains could be removed, and new ones added. Also, in some cases, there are chances that different domains might have the same data. For example, our squadron faced an instance where MRI related information is collected in Study 001 in FA domain but in a custom domain called MR in Study 002.

The Squadron also highlighted some aspects that programmers tend to overlook:

  • Missing to re-populate the sequence numbers for Supplemental domains from the main domain
  • Ignoring variable level consistencies across similar studies. For example, in study 001 the result for a test is Y/N, while in study 002 the result is Yes/No.
  • Not Planning to combine the safety data in advance, leading to inconsistent coding of both AEs (MedDRA) and concomitant medications (usually WHO Drug dictionary).

3. Efficient Programming

It is easy to assume that while creating a submission ready ISS/ISE, all you need to do is stack the datasets of the component studies and apply the Statistical tests on the result, and voilà, your submission is ready.

But the reality is far from that. When you set all your datasets together, you will notice issues like

  • Multiple lengths getting assigned to the same variable
  • The datatype of the same variable being different in different studies
  • Same data points mapped with two different variables.

These programming intricacies may look trivial, but if predicted well in advance, can help save a ton of time. It increases the overall efficiency of the programming process allowing you to concentrate more on analyzing the data than transfiguring it. Additionally, the squadron recommends creating efficient individual programs since any integration process involves millions of data points, and saving a few minutes per program can translate into hours of saved run time. Review of programs as a Quality control procedure can be implemented for keeping the programs contemporaneous and help programmers stick to the bigger picture.

4. Extension Studies

One of the biggest challenges faced while integrating Extension studies (often open-label), involves implementing them into existing studies. Open-label extension studies are conducted right after a randomized trial and are used to assess long-term tolerability of the new drug. Reconciling the extension studies in the integration process is a monumental task that requires intuitive planning. Crucial decisions, like how the Demographics domain will represent subjects present in both studies, must be planned prior to the study. This should align with the analysis requirements of the integration.

5. Analysis Considerations

The focal point of any integration study is the analysis of its endpoints. Identifying data points that need to be analysed, and their source, plays a vital role. Combining the data points and analysing them is not straightforward and depends on the basis for data collection

The squadron quotes a few examples that they came across:

  • Studies usually do not have a homogeneous study population as the inclusion and exclusion criteria may differ. In such cases, the efficacy measures may be inconsistent and may mean that including certain studies in the ISS/ISE could be inappropriate.

Therefore, the rationale for including or excluding certain studies should be justified and documented prior to analysis.

  • Dose ranges will be different in different studies based on study design. Subject pooling strategy can be done in these situations, for dividing the population into groups that reflect the endpoints.

For example, Study A may dose patients for 3 weeks, compared to Study B, that doses patients for 6 weeks. A combined summary of the number of patients who reported an AE may not be appropriate, as one group of patients could be ‘at risk’ and followed up for a greater period. The statistician can help identify the appropriate methodology for addressing such issues. In this example, a solution may be to present AEs using a denominator that adjusts for the time at risk. Our Squadron can keep talking about their ISS/ISE conquests, but we’ve decided to give them a well-deserved break! Do you have any challenges that you’ve faced or a novel solution to the challenges mentioned above? Comment here and we’ll get our squadron to get in touch with you!