This is a ready-to-use register for capturing clinical protocol deviations in a structured, queryable form so they can be trended. Free-text deviation logs cannot be trended, and untrended deviations are a finding by themselves. Keep one per study. Replace every <<FILL: ...>> placeholder, and maintain it in a queryable system rather than prose. A field guide, a metrics section, and a filled specimen follow.
Why structure matters
A single missed visit is an event; the same missed visit at one site fifteen times is a broken process. You can only see the second if the data is structured: category, site, subject, dates, classification as discrete fields, not sentences. This register defines those fields and the metrics computed from them.
Part 1: Deviation register (one row per deviation)
| Field | Format | Required | Notes |
|---|---|---|---|
| Deviation ID | Text | Yes | Links to the deviation record |
| Site | Code | Yes | For per-site rates |
| Subject ID | ID only | Yes | No names |
| Category | Controlled list | Yes | e.g. inclusion/exclusion, out-of-window, missed assessment, IP/dosing, consent |
| Classification | Minor / Important | Yes | Drives the important-rate metric |
| Serious breach? | Y/N | Yes | EU determination |
| Event date / discovery date | Date / date | Yes | For time-to-detect |
| Report date | Date | Conditional | For time-to-report |
| Close date | Date | Yes (to close) | For time-to-close |
| Root cause code | Controlled list | Yes | Process / training / protocol / supply / subject |
| CAPA reference | Text | Conditional | Where raised |
Register table
| Deviation ID | Site | Subject ID | Category | Class | Serious? | Event / discovery | Report | Close | Root cause | CAPA |
|---|---|---|---|---|---|---|---|---|---|---|
<<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> |
Part 2: Trending metrics
Compute on a defined cadence (commonly monthly at study level, per-visit at site level). Use the right denominator so a high-enrolling site is not unfairly flagged.
| Metric | Definition | Threshold / action |
|---|---|---|
| Deviation rate per site | Deviations / subjects (or / visits) per site | <<FILL: e.g. > 1.5x study mean = review>> |
| Important-deviation rate | Important deviations / subjects, over time | <<FILL: rising trend = escalate>> |
| Rate by category | Deviations per category | <<FILL: top category = candidate amendment>> |
| Repeated identical deviations | Same category, same site, count | <<FILL: >= N = amendment signal>> |
| Time-to-report | Report date minus discovery date | <<FILL: > IRB clock = process issue>> |
| Time-to-close | Close date minus discovery date | <<FILL: target>> |
| Self-identification share | Site-found / total at site | <<FILL: low = site awareness issue>> |
Part 3: Review record
| Field | Entry |
|---|---|
| Review period | <<FILL>> |
| Outliers identified | <<FILL>> |
| Decisions (monitoring / retraining / amendment / CAPA) | <<FILL>> |
| Reviewer (name, date) | <<FILL>> |
Acceptance criteria
- Deviation data is captured in discrete, queryable fields, not free text.
- Metrics are computed on the defined cadence with the correct denominator.
- Site and category outliers are identified against a defined threshold.
- At least some reviews demonstrably led to action; a register where every review concludes “no action” is not a real review.
References
ICH E6 GCP (risk-based quality management; sponsor oversight of deviation patterns). FDA guidance on a risk-based approach to monitoring of clinical investigations. ICH E3 (important protocol deviations in the CSR).
Confirm current versions before issue.
Filled specimen
A trending review that found a real signal, illustrative.
Register (excerpt)
| Deviation ID | Site | Subject ID | Category | Class | Serious? | Event / discovery | Report | Close | Root cause | CAPA |
|---|---|---|---|---|---|---|---|---|---|---|
| PD-014-031 | 14 | S-014-007 | Out-of-window primary draw | Important | N | 03 Jun / 17 Jun | 23 Jun | 30 Jun | Process (no window check) | PD-CAPA-022 |
| PD-014-033 | 14 | S-014-011 | Out-of-window primary draw | Important | N | 10 Jun / 17 Jun | 23 Jun | 30 Jun | Process | PD-CAPA-022 |
| PD-014-035 | 14 | S-014-014 | Out-of-window primary draw | Important | N | 14 Jun / 17 Jun | 23 Jun | 30 Jun | Process | PD-CAPA-022 |
| PD-009-018 | 09 | S-009-004 | Out-of-window primary draw | Important | N | 12 Jun / 20 Jun | 27 Jun | 04 Jul | Process | PD-CAPA-022 |
Metric and decision
| Metric | Value | Decision |
|---|---|---|
| Repeated identical deviation (out-of-window primary draw) | 4 across 2 sites in one month (3 at site 14, 1 at site 9) | Not unique to one site, points to the scheduling tool, not just site 14 |
| Decision | Study-wide preventive action | Add an automated out-of-window alert to the visit-scheduling tool; verify the category rate drops to zero over two cycles |
Trending is what turned three separate “important” deviations into one systemic finding and one study-wide fix, instead of three site retrainings that would have left the same broken tool running everywhere. That is the difference between logging deviations and managing them.
Common inspection findings this register prevents
- Free-text deviation logs that cannot be queried, so trending never happens.
- A systemic cause fixed at one site while it recurs at the others.
- Repeated identical deviations that should have triggered a protocol amendment.
- Trend reviews that always conclude “no action,” signalling the review is theater.
How to adapt this register
- Set your controlled category and root-cause lists so the fields stay queryable.
- Calibrate the thresholds to your therapeutic area and enrollment; rare-disease trials carry different baselines.
- Feed the outputs into the monitoring plan so a flagged site drives targeted monitoring.