Protocol Deviation, Device Deficiency, and the Amendment Decision
A 6-site trial of a wearable cardiac monitor has a protocol-specified visit window of +/- 7 days for the 30-day follow-up. At one site, 11 of 22 participants had their 30-day visit between 10 and 14 days late because the site's only trained coordinator was on extended leave and no backup was trained. Separately, three devices returned from that site showed intermittent signal dropout that the sponsor's engineering team attributes to a firmware version that predates the version specified in the protocol. Two of those three participants had their late visits. The sponsor is preparing a substantial amendment to extend the follow-up window to +/- 14 days and to update the firmware specification. The IRB has a convened meeting in three weeks; the sponsor wants to implement the amendment immediately at all sites to 'stop the bleeding.'
1. Classify the issues and determine whether the amendment can be implemented before IRB approval.
Read the worked answer
Three distinct issues, with different rules:
Late visits (protocol deviation). Out-of-window visits are deviations. Eleven of 22 at one site is a pattern, not an anomaly, and the root cause (single trained coordinator, no backup) is a site capability failure. These are documented, assessed for impact on endpoint interpretability, and reported per the protocol and IRB policy. They do not by themselves authorize a protocol change.
Device deficiency (firmware). Devices running an out-of-specification firmware version is a device deficiency - a inadequacy in the device's labeling, instructions, or performance relative to the protocol. This is not a protocol deviation by the site; it is a sponsor-controlled problem requiring investigation of how nonconforming units reached the site, assessment of participant risk, and reporting per applicable device regulations.
The amendment. A change to eligibility, follow-up windows, or device specifications that affects risk or the rights and welfare of participants requires IRB approval before implementation, except where necessary to eliminate an apparent immediate hazard to participants. Extending a follow-up window is not an emergency measure; 'stop the bleeding' is an operational preference, not an immediate hazard. The amendment cannot be implemented at any site until the IRB approves it. If the firmware issue presents an immediate hazard, that is handled as an urgent safety measure with prompt reporting - a separate track from the amendment.
2. How should the monitor address the two participants whose late visits coincided with device signal dropout, and what does the overlap imply?
Read the worked answer
The overlap is the analytically important fact: it means the deviation and the device deficiency may be confounded. If the visit was late and the device was underperforming, the endpoint data for those participants may be uninterpretable for reasons that cannot be separated after the fact.
Monitor actions:
- Reconstruct the timeline for each participant: when was the device dispensed, what firmware, when did dropout begin, when was the visit actually conducted, and what data were collected in the interval.
- Determine whether the dropout affected the primary endpoint window. If the device was not capturing data during the protocol-specified window, the participant may need to be evaluated for exclusion from the per-protocol analysis, with the decision documented in advance of analysis (not after seeing results).
- Assess whether either participant experienced a safety consequence of the missed or degraded monitoring (e.g., an arrhythmia that went undetected). If so, that elevates the event and drives reporting.
- Query whether other sites received devices from the same nonconforming lot; the firmware finding should trigger a lot-level reconciliation, not just a review of the three returned units.
Implication for the amendment: the data support extending the window and correcting the firmware spec, but the amendment should be written with a pre-specified handling rule for out-of-window and device-deficiency-affected data, so the analysis plan does not have to be reverse-engineered from the deviations.