<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

The Data Management Perspective on the Interim Analysis

 Back view of businesswoman standing on crossroads and making choice.jpegAs a recognized expert in adaptive trials, Cytel has extensive experience designing and managing trials with interim analyses.  To ensure success in what are often complex studies, data management as well as statistical expertise is required.  Cytel data managers are well versed in the various nuances and demands of managing the successful delivery of an interim analysis from a data collection point of view. 

Success from the data management standpoint depends on three core elements- effective timeline management, thoughtful database design, and a proactive approach to data cleaning. In this blog, Patti Arsenault, our Global Head of Data Management shares her thoughts.

 Project plan.jpeg

Timeline Management

An interim database freeze should be considered as fundamental a milestone as a final database lock, and its timelines need to be managed with this in mind.  At the pre-planned Data Monitoring Committee (DMC) meeting, the  DMC will make decisions about the future course of the trial based on interim outputs provided by the Independent Statistical Center (ISC). In turn, these outputs depend on the data provided to the ISC by the data management team following the interim database freeze. The window between database freeze and the DMC meeting tends to be narrow, with little or no tolerance for delays. The data manager, therefore, needs to manage the timeline up to the database freeze extremely closely. 

 

Data Management on the Mechanism of Metal Gears..jpeg

Database Design

Nowadays, electronic data capture systems are available to facilitate the smooth running of the interim analysis.  When designing the database, there are a number of questions which you will need to consider and discuss with your CRO or in-house team.  Below, we outline some examples:

  • What datasets or variables (critical or not) will be used for interim analysis?
  • What external data are expected? How will these data be incorporated into the database and used in the interim analysis?

It’s important to note that lab or PK vendors may have data that are potentially unblinding, and so any blinded team members should not have access to these data. While this seems simple, unless planned for, the integrity of the trial could be affected.

  • How should the EDC behave during the data freeze?

For example, it can be useful to have a page Status of ‘Read Only’ available to allow the data manager to easily freeze/ unfreeze specific Pages or Subjects – this supports the cleaning process up to the interim lock.

  • How will user access to the EDC be managed?

It will need to be determined if an unblinded user is required as part of the data management team, and how their access will be managed.

  • Will enrollment continue or stop until the DMC has announced their findings?

Most EDC software now allows enrollment capping to be used at the study or the site level.  In a similar way, randomization capping can also be placed on an overall arm.

  • What is the enrollment rate expected to be?

If the trial is expected to be fast enrolling, then this will usually mean more forward planning is required.

Obviously, where any particular database module is required for use in the interim analysis such as medical coding or endpoint adjudication, then this needs to be programmed, fully tested and the users trained in advance of the interim data becoming available.

 

Database Cleaning

When planning for an interim analysis, the data management plan will reflect the required cleaning activities. Often, more frequent cleaning, coding, and SAE reconciliation will be planned at a general level, and specifically leading into the interim analysis itself. Depending on what is being reviewed at the interim analysis, data management activities could vary.  For instance, if only safety data is being examined, then this will be reflected in the data cleaning plan.

Use of the reporting functionality in the electronic data capture system can really come into its own in this setting, allowing the data management team to very efficiently keep on top of data entry and query management.

Recruitment reports can be generated to gauge how timely sites are with data entry. If there is any noticeable lag, the data manager will watch this closely. Where delays aren’t addressed, then the clinical monitoring team should be alerted to ensure that the timeline to interim database freeze is not impacted.

For query management, the same close management needs to be conducted.  It is part of the data manager’s role to ensure that the sites fully understand what is expected from each query, and the timeline expected for query closure.  It’s also important to define how many iterations of a query can be accommodated without risking the study timelines. Where a query is reissued and continues to be answered incorrectly, the data manager will raise this with the monitoring team.   While it is always important to ensure that stakeholders’ roles and expectations are being  adhered to, it becomes critical when timelines are tight, as with an interim analysis database freeze. If there are persistent issues with a site’s responsiveness to queries, then the data manager should consider that retraining might be necessary.

Unposed group of creative business people in an open concept office brainstorming their next project..jpeg

In all studies, having a cohesive team helps achieve the timelines and goals. Cytel Data Management works seamlessly with every Clinical Operations team to ensure quality is delivered on time.

 

Like this article?  Join our community of biopharma innovators and sign up for Cytel blog updates direct to your inbox. 

Subscribe

 Planning your clinical trial? Find out more about Cytel's data management services. 

Data Management

 

Related Blog Posts 

6 steps to timely database lock

Adaptive Designs: A Data Management Perspective

 

contact iconSubscribe back to top