Skip to main content
Home / Resources / Blog / Diving into Best Practices for Pooling Clinical Trial Data

Diving into Best Practices for Pooling Clinical Trial Data

If you’ve worked with a client drug development team approaching submission for approval, it’s likely you’ve heard discussions like this:

Team member 1: “But, you can’t integrate the data from those studies because the treatment durations are different.”

Team member 2: “That doesn’t matter, we still have to pool the results into a single integrated summary for submission for approval of the product!”

Team member 1: “How can we do that when the designs of the studies aren’t the same?”

Team member 3: “Actually, two of the three Phase 3 studies have the same design, and one of the Phase 2 studies as well.”

Team member 4: “I’m confused, is there a difference between integrating and pooling?”

Team member 1: “Yes, but why would we ever integrate data from Phase 2 with Phase 3?”

Team member 5: “Do we even have time to pool the data?”

You also probably recognize some of the errors in the conversation above. Decisions on how to best pool or not pool clinical trial data for submission to regulatory agencies are not easy and require thought and debate within your team.

That’s when the advice of regulatory writing experts can make all the difference. Because when you’re drowning in data, it’s easy to end up doggie paddling in the shallow end instead of butterflying your way into the Olympic pool.  Your clinical data handling must be clear, efficient, and meet regulatory requirements. We help you identify solutions that will withstand rigorous reviews, both internally and by the regulatory authorities.

Companies are pushing the development of their products faster and faster.  In addition, a wide array of expedited development, review, and approval options is now offered by the FDA and other regulatory authorities.  In this environment, teams have less time to consider options, and more options to consider, when planning data strategies.  Unfortunately, this strategic planning occurs in a frantic rush as the studies are completing rather than when the protocols and statistical analysis plans for the studies are being generated.

Clinical trial data integration versus pooling

“Integration” and “pooling” of data are NOT synonymous.  As the 2015 FDA “Integrated Summary of Effectiveness” guidance specifies, integration summarizes in a single document all the information known, including both results from individual studies and published literature, on a particular drug. The purpose of the integrated analysis is to help the reviewer understanding the overall evidence for drug efficacy.

In contrast, pooling refers to combining data from multiple studies into a single dataset, so that analyses can be run on that new compiled dataset.  Therefore, in a submission, pooled data represent a subset of the integrated information to be presented.

Pros and cons of pooling

Pooled data are most useful for analyses and evaluations that benefit from a larger sample size, even if that sample is more heterogeneous (includes some differences in patient population, study design, treatment duration, etc):

  • Assessing variability in efficacy or safety effects in subgroups, such as older patients, women, or individuals with a particular baseline condition. Larger sample sizes improve the ability to identify statistical differences between subgroups.
  • Identifying rare adverse events (AEs). Larger sample sizes (eg, a pooled dataset of 3000 patients) increase the likelihood of seeing an event that happens in 1 of every 1000 patients relative to a dataset of 300 patients.
  • Modeling pharmacokinetics for a population, or testing the relationship between efficacy or safety and drug exposure. More samples, across a broader range of patients, increases the power of models to define the factors that influence systemic drug exposure and determine the relationship between exposure and specific efficacy or safety effects.

Pooled data are least useful when the degree of differences makes the results of the pooled data less meaningful or even meaningless. This applies to safety, but is especially true for efficacy pooling.  For example, no one would gain any useful information from pooling the efficacy data from two studies in different indications. Similarly, pooling safety data from a single-dose safety study with data from a 52-week safety study would likely dilute the overall AE incidence.

For efficacy, the FDA will always treat the individual study results as the primary support. They regard any pooled efficacy results as guidance on efficacy across the populations included in the pool. One exception to this is subgroups, for which pooled efficacy data take primacy over individual study results.

For pharmacokinetics, data are pooled for modeling purposes. In most cases, more data is better, regardless of the source, as long as all the pertinent dosing and patient information are available. Differences in dosing, study design, and patient population can be built into and tested in pharmacokinetic models.

Pooling considerations

Differences between studies can affect the validity of and ability to interpret pooled analyses. You should be proceed with caution when the studies differ with respect to:

  • Important demographic or disease characteristics (e.g., duration, severity, specific signs and symptoms, previous treatment, concomitant diseases and treatments, prognostic or predictive biomarkers)
  • Treatment practices, including methods of assessing effectiveness, specific test procedures (eg, measurements of biomarkers and clinical endpoints)
  • Study design features (eg, study duration, study size, doses studied, visit frequency).

Every discussion of pooling requires you to identify what, if anything, a particular pool of data is better at answering than the results of one or more of your individual studies separately. Your team will have to justify for pooling or not pooling data from particular studies or subsets of patients.  For example, pooling safety data from healthy volunteers with data from patients with the indicated disease or condition may be warranted in looking for possible rare adverse effects of the drug. In other cases, the data pooling would dilute the overall safety profile to be presented and mislead the reviewer.

Remember also, that the activity of pooling is largely a programming effort.  This impacts your pooling decisions in three main ways:

  1. The more similar the programming of each of the studies to be pooled, the easier the pooling will be (eg, pooling data from five recent studies from your own company will require significantly less work than pooling data from three studies done by three different companies over a 10-year period)
  2. It will likely take the programmers the same amount of time to pool a small subset of the data from 2 or more studies as to pool all the data from those studies (because of potential differences in variable naming and data structure)
  3. To save critical time to submission after your last study, you should plan for additional programming resources to complete the pooled output in parallel with completing the output for the final clinical study.

Nervous about plunging into the murky waters surrounding submission preparation?

I hope that this has helped clear up any confusion you may have had regarding some of the issues associated with getting your clinical trial data results ready for submission to a regulatory agency. Crafting a submission that presents your drug in the most clear, accurate, and informative light to regulatory authorities is both an art and a science. Read this case study to learn how we helped create an end-to-end solution for a biopharmaceutical company that helped them meet their aggressive timelines for submitting a New Drug Application (NDA).

About the author

Steve Sibley
By: Steve Sibley
With a career spanning more than 30 years in the pharmaceutical industry, Steve has extensive experience across regulatory writing, consulting, and project leadership roles. He has successfully supported projects from discovery through approval and life cycle management, including holding significant roles in more than 75 submissions and, in several cases, leading the entire submission team, overseeing all documentation from Modules 1 through 5, publishing, and transmission to Regulatory authority.