In an ideal setting, clinical trials follow a double-blind design, where neither participants nor personnel involved in the conduct of the study are aware of treatment assignments during the course of the study. However, this gold standard is not always possible.

In this blog, we discuss randomized open-label studies and best practices to minimize potential bias when conducting the analyses.

Essential tenets of clinical trial design: Randomization and blinding

The two tenets essential to clinical trial design are randomization and blinding. For the latter, the gold standard for randomized clinical study design is the double-blind, whereby the participants, the clinical investigator, the sponsor, and all who handle and analyze the data while the study is ongoing are blinded to the randomized treatment assignment. This principle applies both to aggregated summaries by randomized treatment group as well as the full set of individual participant-level data.

Randomized open-label: When the double-blind design isn’t possible

Maintaining the blind is an important component to help demonstrate that the difference between treatment groups arises only from the randomized treatment and not from the expectations, behavior changes, or biased assessments by those involved in the conduct of the study.

However, the gold standard double-blind design is not always possible. This occurs when creating a dummy or sham comparator treatment or procedure is:

• impractical (e.g., comparator is “lifestyle change” versus experimental “intensive therapy”),
• unethical (a surgically implanted device),
• too costly (an acceptable placebo match is too expensive to manufacture),
• too burdensome (additional visits and interventions in a vulnerable population), or
• some combination of the above.

This design is called a randomized open-label study (or in case of the participant not knowing the treatment, randomized single-blind).

Protecting trial integrity

For many of these studies, the schedule of events differs between the two treatment groups, and the data collected will themselves reveal the treatment group assigned for each participant. As a result, this often makes it impossible to easily “blind” all the database elements for those who need to handle study data while the trial is ongoing (such as data managers, safety monitors, statisticians, analysts, and sponsor staff).

While open‑label designs may be accepted, regulators are explicit about the need to protect study integrity. A recent FDA poster emphasizes that access to accumulating subject‑level or group‑level data by treatment arm should be carefully restricted during trial conduct:

“It is critical to prevent access to accumulating subject-level and group-level study data that includes information on treatment assignment (either with the treatments identified or with codes such as “A” and “B”) outside of an external independent DMC and a supporting independent statistician(s) who prepares interim reports.

Knowledge of comparative summary-level interim outcome results by subjects, investigators, the sponsor, or the public can negatively impact trial conduct (e.g., recruitment, adherence, and retention) and impair ultimate interpretation of results.”

The challenge is how to manage access to treatment‑related data in combination with efficacy data in a way that minimizes the risk of operational bias while still enabling efficient, high‑quality analyses.

The Role of the Data Monitoring Committee/Data Safety Monitoring Board

Phase III randomized studies typically establish a Data Monitoring Committee (DMC)/ Data Safety Monitoring Board (DSMB) to safeguard participant safety and maintain the integrity and validity of the study.

DMCs/DSMBs will review the accumulating study data in an unblinded manner while the trial is ongoing to ensure participant safety and, when applicable, to support planned interim analyses for futility, efficacy, or sample size re-estimation. However, there will be the need beyond the DMC/DSMB to prepare and analyze study data, ideally in a manner that preserves blinding wherever possible during the course of the trial.

Best practices to minimize potential bias

To address this issue while conducting the statistical analyses of open-label studies, Cytel has provided several configurations to support our clients. Regardless of the exact methods, the following include our recommendations and best practices to minimize the potential for bias while providing transparency.

• Draft the SAP as early as started, preferably prior to first participant randomized.
• Ensure that the contributors to the SAP and any interim analysis approach methodologies (clinicians, statisticians, sponsor staff) do not have access to ongoing study data (full set of individual data or aggregate post-baseline summaries).
• Create a document (such as a Blinding Plan or Data Integrity Plan) that outlines:
  1. Access to specific types of data (individual efficacy data, aggregate efficacy data, individual post-baseline non-efficacy data, and aggregate post-baseline non-efficacy data)
  2. Approach to maintain the blind (e.g., use of dummy treatment, masked/dummy data as needed, separate team)
  3. List of entities, roles, and personnel with access to each data point
• Identify personnel responsible for creating blinded datasets and TLF outputs.

Final takeaways

The gold standard for randomized clinical study design is the double-blind, but when that isn’t feasible, early planning, clear governance, and fit-for-purpose operational models can ensure the potential for bias is minimized and that open-label randomized studies are conducted with transparency and without undue impact to timelines.

Interested in learning more?

Learn how Cytel works closely with sponsors to develop a streamlined workflow that preserves study integrity while remaining practical and deliverable:

Share Via