Breaking the blind

Steve Simon


Someone asked an interesting question on the IRBForum today. Under what circumstances should the blinding for a trial be broken when an unexpected Serious Adverse Event (SAE) occurs? The questioner did mention the International Conference on Harmonisation (ICH) guidelines which read as follows:

When the sponsor and investigator are blinded to individual patient treatment (as in a double-blind study), the occurrence of a serious event requires a decision on whether to open (break) the code for the specific patient. If the investigator breaks the blind, then it is assumed the sponsor will also know the assigned treatment for that patient. Although it is advantageous to retain the blind for all patients prior to final study analysis, when a serious adverse reaction is judged reportable on an expedited basis, it is recommended that the blind be broken only for that specific patient by the sponsor even if the investigator has not broken the blind. It is also recommended that, when possible and appropriate, the blind be maintained for those persons, such as biometrics personnel, responsible for analysis and interpretation of results at the study’s conclusion.

There are several disadvantages to maintaining the blind under the circumstances described which outweigh the advantages. By retaining the blind, placebo and comparator (usually a marketed product) cases are filed unnecessarily. When the blind is eventually opened, which may be many weeks or months after reporting to regulators, it must be ensured that company and regulatory data bases are revised. If the event is serious, new, and possibly related to the medicinal product, then if the Investigator’s Brochure is updated, notifying relevant parties of the new information in a blinded fashion is inappropriate and possibly misleading. Moreover, breaking the blind for a single patient usually has little or no significant implications for the conduct of the clinical investigation or on the analysis of the final clinical investigation data.

However, when a fatal or other “serious” outcome is the primary efficacy endpoint in a clinical investigation, the integrity of the clinical investigation may be compromised if the blind is broken. Under these and similar circumstances, it may be appropriate to reach agreement with regulatory authorities in advance concerning serious events that would be treated as disease-related and not subject to routine expedited reporting. Clinical Safety Data Management: Definitions And Standards For Expedited Reporting. E2A, page 9.

This is excellent advice. In particular, I support their comment about how breaking the blind for a single patient usually doesn’t have any significant implications on the scientific validity of the study. Sometimes researchers get hung up over running the “perfect” research study and let that unduly interfere with other important considerations.

Another consideration is that the research subjects themselves should have the option of breaking the blind. For example, a research subject is about to undergo an unexpected surgery and needs to know about the drug being administered in the clinical trial because one of the two drugs being studied thins out the blood and increases the risk of the operation.

You don’t want to break the blind for any old reason, but if the patient has a legitimate reason to know, we should respect that. Patients who volunteer for a research study cede a lot of their autonomy when they join. They let a random flip of the coin decide which treatment they get and in most cases, they are kept in the dark about that treatment they get until the study is over. There are good reasons why we ask patients to sacrifice this much autonomy, but we should recognize that it is indeed a sacrifice and we should respectfully return that autonomy early to the patient when there is a good medical justification.

In a perfect world, the rules under which the blinding might be broken should be specified in the research protocol.

A final consideration is that any data safety and monitoring board should have access to unblinded data, especially in trials where SAEs occur frequently.

You can find an earlier version of this page on my original website.