Your Own Protocol Deviations

Part of Principal Investigator (PI) Training: Eligibility and the Enrollment Decision.

About 4 minutes. Free, in your language, no account needed.

Why this matters

Deviations happen at every site that enrolls anybody. The distinguishing feature of a well-run site is not their absence; it is what happens in the seventy-two hours after one.

The guideline's expectation, in §2.5.3, is a four-part cycle: document, review, explain, prevent recurrence. Most sites do the first and skip the fourth, which is why the same deviation appears eleven times in a trial.

What counts

A deviation is any departure from the protocol as approved. That includes the small and tedious as well as the serious:

• A visit outside its window.
• An assessment not performed, performed in the wrong order, or performed with the wrong instrument.
• A dose given late, missed, or at the wrong amount.
• A prohibited concomitant medication started and not caught.
• Consent obtained on a superseded form version.
• An eligibility criterion not met — which is both a deviation and, usually, a serious one.

Sponsors will additionally classify some deviations as important or major, typically those affecting participant safety, participant rights, or the integrity of the primary data. Ask your sponsor for their definition at start-up; it varies, and it determines what has to be reported and how fast.

The four steps, done properly

Document. Contemporaneously, factually, in the source. What happened, when, to whom, discovered by whom and when. No interpretation, no defensiveness, no adjectives. A deviation log entry written the same day is worth more than a considered account written after the monitoring visit.

Review. By you. This is not a coordinator's task, because the question being asked is medical: did this affect the participant's safety, their rights, or the reliability of their data? A missed non-critical questionnaire and a missed safety laboratory test are the same category on a log and completely different clinically.

Explain. Root cause, not the immediate cause. "The visit was late" is what happened. "Our scheduling template calculated windows from the randomization date and the protocol counts from first dose" is why it happened. Only the second can be fixed. If your explanation has no mechanism in it, keep asking.

Prevent recurrence. A concrete change with an owner and a date — a corrected template, an added check, a re-training with a record, a second pair of eyes on a specific step. Then verify it worked, some weeks later. A preventive action nobody checked is a sentence, not a control.

Reporting outward

Three audiences, and they want different things.

• The sponsor — per the clinical trial agreement and their deviation procedure, usually within a defined number of days, faster for important deviations.
• The IRB or IEC — per their own requirements. Section 2.4 requires prompt communication of deviations implemented to eliminate an immediate hazard, of changes significantly affecting the conduct of the trial, and of adverse new information. Many committees ask for more than the minimum; follow theirs.
• Regulatory authorities — where applicable requirements demand it, typically through the sponsor.

When in doubt about whether to report, report. An over-reported deviation costs an email. An under-reported one becomes a finding about the reporting rather than about the original event.

Reading your own pattern

Once a quarter, read your whole deviation log at once rather than one entry at a time. You are looking for shape, not incidents.

Clustered around one participant suggests a person struggling with the schedule — possibly a withdrawal conversation. Clustered around one procedure suggests a training or process gap. Clustered around one team member suggests a supervision gap. Clustered around one time period — a holiday, a staff departure, a surge in enrollment — suggests a capacity problem, and capacity problems recur.

Rising deviations while enrollment rises is the clearest early signal that a site has taken on more than it can conduct properly. That is a §2.2.2 problem, and the honest response is to slow enrollment before somebody else slows it for you.

Key takeaways

• §2.5.3 requires four things: document, review, explain, prevent recurrence. Most sites omit the fourth.
• You review deviations personally, because the question — safety, rights, data reliability — is medical.
• Root cause means a mechanism. If the explanation contains no mechanism, nothing has been fixed.
• Read the log as a pattern each quarter; rising deviations alongside rising enrollment means the site is over capacity.