Root Cause Analysis in GMP: Using 5-Why, Ishikawa, and FMEA Correctly

The investigation starts with two separate questions

At goods receipt in the fictitious RM-427 case, the seal on container 3 is found damaged. The observation is documented, and the material remains in quarantine for now. Two questions now need to be answered: How did the damage occur? And did the operational procedure capture the finding unambiguously and pass it on to the responsible review? A weakness in how the finding was passed on does not explain why the seal was damaged.

EU GMP Chapter 1, section 1.4(xiv), requires an appropriate level of root cause analysis for deviations, suspected product defects, and other problems. Where the true root cause cannot be determined, the most likely causes should be considered and addressed. Where human error is suspected, potential process, procedural, or system errors must not be overlooked. ^[1]

For RM-427, the first step is therefore a neutral statement of the initial finding: “Seal on container 3 found damaged during the documented incoming inspection.” It records the time and the observation without prematurely assigning responsibility to transport, the supplier, or an individual. The initial assessment of the impact on the material takes place in deviation management.

ICH Q10, section 3.2.2, links structured investigations with the CAPA system. Effort, formality, and documentation should be commensurate with the risk. For the investigation plan, this means the open question determines both the evidence needed and the method. ^[2]

The investigation plan makes data gaps visible

The team defines the material, time period, and affected process steps. For RM-427, these include the identity of container 3, the incoming inspection, available delivery and transport information, and SOP-WE-12 V3, which was effective at the time. For each investigative step, the question, the owner, and the expected evidence are recorded. Photos, finding descriptions, and interviews are reviewed for what they actually show.

ICH Q9(R1) links the choice of tool to a clearly defined problem or risk question and explicitly addresses data quality, uncertainty, and subjectivity. An information gap is therefore recorded as an open question. The absence of an image of the seal before transport neither refutes nor confirms transport damage. ^[3]

In the US CGMP context, 21 CFR 211.192 requires a thorough investigation of unexplained discrepancies and certain specification failures. The scope extends to other batches and products that may have been associated with the discrepancy; conclusions and follow-up must be recorded in writing. For RM-427, the investigation accordingly explains which further containers or deliveries are included in the review, and why. ^[4]

Batch traceability helps identify such relationships. Whether other material is actually affected is determined by the expert investigation.

The depth of review is guided by the potential impact on quality and by the open questions. For RM-427, for example, the investigation must establish which protective function the seal serves and whether other packaging layers are affected. A well-documented handover of the finding can make the investigation easier; on its own, it does not reduce the potential significance of a damaged protective function. The investigation plan keeps these two assessments apart.

Three tools serve different purposes

The methods can be combined. ICH Q9(R1) lists Ishikawa as a structuring aid and FMEA as a tool for considering potential failure modes and their effects. Which tool fits depends on the question and on the available process knowledge. ^[3]

The 5-Why method drills down into a specific question

With 5-Why, a problem is written down and then explored by repeatedly asking why. The original NHS Institute handbook describes this approach and treats five questions as a rule of thumb: fewer or more may be needed. This method description is not a pharmaceutical regulatory requirement. ^[5]

In the completed RM-427 example, the reviewed V3 wording and the execution records support the procedural finding. How the gap in the instruction arose remains open.

Stopping point: data missing. Steps 3 and 4 remain open follow-up questions. In the example, the evidence sufficiently supports the unclear instruction, but not the reason it arose. That is enough to justify clarifying the inspection step; the cause of the seal damage remains a separate matter.

The Ishikawa diagram opens up alternatives

An Ishikawa diagram, also called a fishbone diagram, organizes possible influences around a clearly stated problem. For RM-427, the areas considered might include material and packaging, handling and transport, work instructions, and observation and documentation. These categories are specific to the example and must be adapted to each investigation. ICH Q9(R1), Annex I.1, names the tool without prescribing any categories. ^[3]

The diagram organizes both lines of investigation. In the fictitious example, the wording review supports the procedural finding; the effective V3 snapshot refutes the hypothesis that a wrong SOP version was used. There is insufficient evidence for the packaging and transport hypotheses. “Open” does not mean “rejected”; transport packaging that looks unremarkable from the outside does not rule out every kind of transport stress.

The FMEA assesses potential failures in the inspection process

Failure mode and effects analysis, or FMEA, considers potential failure modes and their effects. ICH Q9(R1), Annex I.2, describes breaking down complex processes into manageable steps and using the results to prioritize risks and risk controls. I.3 covers the extension to FMECA. ^[3]

The FMEA looks ahead at potential future failures. Rating scales are defined and justified by the organization. The example row contains no risk priority number. The decision on whether the material can be used requires its own quality evidence.

For RM-427, the procedural finding stays separate from the seal damage

For the rest of the example, assume that the investigation confirms that the instruction on handling damaged seals was unclear. This procedural finding supports the action “Clarify the inspection step for damaged seals.” The physical origin of the seal damage is still not explained. The investigation report presents both results, each with its own level of certainty.

The CAPA takes over the substantiated procedural cause, the objective of the action, and the planned effectiveness check. Change control is initiated for the revised SOP. SOP-WE-12 V3 is preserved as a snapshot of the earlier execution; V4 is introduced in a controlled manner. The version change alters neither the original finding nor the basis for decisions on RM-427.

420+ keeps the investigation in its execution context

In the operational twin, the observation, material reference, executed inspection step, and follow-up task share the same process context. The SOP snapshot shows which instruction applied at the time of the inspection. This allows a later assessment to refer to the instruction that applied at the time.

The investigation connects findings and evidence with their expert conclusions. An audit trail makes changes to documented information traceable. A chronological sequence alone, however, does not establish a cause. The team continues to examine which explanation is supported by the available data.

Conformance checking can reveal differences between a described process and its documented process events. For RM-427, for example, it would be relevant whether the intended follow-up review was demonstrably triggered. Such a difference raises a question for the investigation. Whether it reflects an unclear instruction, a deviating execution, or a documentation gap must then be clarified on the basis of the relevant information.

Roles, independent second-person review, deadlines, and escalation are configurable defaults. The responsible experts determine what information is still needed and whether the report supports its conclusions. A further task can be created for an open question without prematurely recording a suspicion as an established cause.

Closure links what is known to the follow-up work

A closure that can be verified states the initial finding, scope, information evaluated, alternatives considered, and conclusions. It shows which cause is sufficiently supported and where uncertainty remains. EU GMP Chapter 1, section 1.4(xiv), also provides for addressing the most likely causes when the true root cause cannot be determined. ^[1]

For RM-427, the open mechanism behind the seal damage and the assumed procedural finding are assessed separately. The investigation hands over justified actions and supporting evidence for further processing. Whether the clarified inspection step works in day-to-day operations is then checked based on suitable practical experience. ICH Q10, section 3.2.2, explicitly links CAPA with the evaluation of its effectiveness. ^[2]

If later information calls an assumed cause into question, the assessment is updated in a traceable way. The original findings remain visible as the starting point. The action already derived is also reviewed to determine whether it still fits the identified problem or needs to be adjusted.

Frequently asked questions about root cause analysis in GMP

Are five “why” questions mandatory?

Five is a methodological rule of thumb. What matters is the question under investigation, robust answers, and a justified conclusion. Depending on the problem, fewer or more questions may be appropriate.

Is an Ishikawa diagram proof of a cause?

It organizes possible influences. Which explanation holds up follows from investigating the hypotheses and assessing the evidence.

Can an FMEA replace a deviation investigation?

It supports the assessment of potential failures and risks. The specific discrepancy, its causes, and its consequences must still be investigated on the basis of the actual case.

Is “human error” a sufficient conclusion?

This attribution needs a justification. Potential errors in the process, procedure, or system must be carefully considered as well.