ICH Q9: written risk ≠ managed risk


Some plants have impeccable QRM in the archive and almost none on the shift. The process risk protocol was written at validation, signed, filed. Two years later a mixing deviation, an excipient supplier change, or a range tweak in the EBR follows the "usual" flow — without reopening the question the protocol had already answered.
In most cases this is not a "lack of QRM." It tends to be something else: written risk treated as managed risk. The text exists. Management, in the sense of ICH Q9(R1), has not yet reached the shop floor.
What Q9(R1) actually pushes
The current guideline — ICH Q9(R1), adopted at Step 4 in January 2023 and effective in Europe since 26 Jul 2023 — does not invent the idea of managing risk. It deepens what industry already knew and still does halfway: formality proportional to risk, risk-based decision-making (not only a "completed analysis"), and explicit care with subjectivity.
On the plant floor, that usually means three shifts in practice:
- Report size does not define management quality. A sixty-page FMEA on a low risk and a loose paragraph on a high risk are the opposite of what Q9(R1) describes as proportionate formality.
- Assessing is not controlling. Mapping failure modes without tying in what the system (or the procedure) does when the mode appears leaves the risk in the presentation only.
- Risk ages. Change, recurring deviation, new material, new equipment, new release path: the point at which risk must be reviewed matters as much as the point at which it was first assessed.
None of that asks anyone to "guarantee compliance" with a slide. It asks that the assessment stay in use where decisions happen.
Where this shows up day to day
That third shift — risk ages — is the most common pattern in operations.
The protocol already wrote that relative humidity at weighing is critical for attribute X; that mixer charge sequence matters; that the second person on the material-lot check is not a formality. In operations the deviation arrives like this: "humidity ran a bit high, but the batch passed"; "we swapped two charge steps because the silo was busy"; "the pharmacist confirmed later, on paper."
If deviation and change flows do not reopen the risk the protocol already named, the plant ends up running on two versions: the archive's and the shift's. Q9(R1) does not fix that alone. It makes clear that risk communication and review are part of the process — not an audit appendix.
The useful question, then, is not "do we have QRM?" It is: when something changes or deviates, does the matching risk come back for review — or only the deviation form?
ALCOA+ as a risk decision
Notice that in all three examples above, the problem runs through the data: the humidity value, the recorded sequence, the check done afterwards. That is why data integrity belongs here as a system and risk choice, not as a standalone compliance item.
If a critical process value depends on late entry, the risk is not only "keystroke error": it is weak contemporaneity, fragile attributability, and a second person who exists precisely because data provenance is poor.
In practice, when a parameter is judged critical enough to enter the FMEA, it usually makes sense to ask at the same time: where does this value come from, with what timestamp, and what does the system do if it is out of range? That is Q9 talking to record design — ALCOA+ as a consequence of the assessment, not a separate checklist.
Formality with focus
The same logic applies to how much process each risk calls for. Q9(R1) talks about proportionate formality. On the floor, too much formality without focus becomes a paper-only process; too little on a high risk becomes an inspection finding. The middle ground that tends to work is tying the level of effort to the level of patient and product-quality harm — and making that visible in the flow itself.
In practice, three patterns are common:
- High risk, loose control in the system. The document says "critical"; the EBR only raises a warning, and the shift closes and moves on. Assessment and execution diverge.
- Low risk, high bureaucracy. Every minimal change demands the same package. People work around it. The workaround disappears from the audit trail — and the real risk (the shortcut) was never in the FMEA.
- Well-written risk, no owner. The matrix exists; at change time nobody knows who reopens the assessment. It becomes an email queue.
The fix is not another template. It is owner, trigger, and evidence: who decides, what fires the review, and where that is recorded in a recoverable way.
Subjectivity: what Q9(R1) names openly
Even with owner and trigger defined, the risk score is still human judgement. One of the most useful contributions of the R1 revision is addressing that subjectivity directly. Dominant facilitator, "we have always done it this way," lead-time pressure: all of these bias severity scores and what counts as acceptable.
Judgement cannot be eliminated. It can be made visible and revisable: pre-agreed criteria, process data when they exist, and the habit of asking "what would change this score if the next batch failed?" In an electronic system that usually shows up as decision history in change control — not as a loose comment in the analysis spreadsheet.
Where the system of record helps
It is worth being careful here not to promise more than the guideline asks for. ICH Q9 does not require an MES. It requires management. What a well-designed system of record — for example an EBR with change and deviation tied to the batch — can do is reduce the space where risk exists only in the archive:
- prevent (or at least evidence) execution outside the range the assessment treated as critical;
- force limit, sequence, or material changes through a stateful flow — not silent email and reconfiguration;
- keep a trail of who changed what, when, and with what justification;
- link recurring deviation back to the same risk family instead of treating each occurrence in isolation.
That does not "implement Q9." It creates the conditions under which Q9 stops being only a PDF. The science of risk stays with people; the system closes the shortcut.
What not to confuse
- Having an FMEA ≠ managing risk. Without review in the change/deviation cycle, it is a snapshot of one moment.
- Tool ≠ method. FTA, HACCP, FMEA are means. Q9's question is whether the control decision follows the assessment.
- Zero risk is not the target. The target is known risk, controlled to what the product requires, and revisited when context changes.
- Inspection is not the only customer. If QRM only "wakes up" for inspection prep, it is not integrated into operations — and Q9(R1) is explicit about integration.
An honest sequence for the team
In practice, five checks help keep the assessment in use:
- The last three serious (or recurring) deviations show whether the matching protocol risk was reopened. When the answer is "no," the gap is not a template gap.
- The parameters the FMEA calls critical need to be compared with what the EBR/MES actually does: hard block, ignorable warning, or nothing.
- In change control, "risk reviewed? reference?" needs to be a required field; chronic "N/A" is a signal, not a time saver.
- Formality can be reduced where risk is low, deliberately, so capacity remains where risk is high. Proportionate formality is also effort allocation.
- Integrity of critical data needs to be part of the same risk package, not a parallel data integrity project.
Closing
ICH Q9(R1) does not ask for more risk paper. It asks that written risk keep governing decisions when the plant moves — change, deviation, new knowledge. While the assessment sits untouched in the archive and the shift handles things as best it can, with good intent, QRM is not managed: it is archived.
If in your plant the risk protocol is genuinely reopened when the deviation arrives — or only at the next validation campaign — that is usually the question that separates documentary compliance from management.
At T2, when we talk about electronic batch records and change control in the same system of record, the point is not "buying Q9." It is removing the shortcut that makes written risk lose its effect on the second shift.
Practical writing on GxP, MES, data integrity and shop-floor systems. A few times a month, no noise.


