How do you perform an FMEA in pharma and calculate the risk priority number (RPN)?

1What FMEA is.

Failure modes and effects analysis establishes how an item or process might fail to perform its function, so that any treatment needed can be identified. It identifies the failure modes and their effects, may also identify their causes, and lets the failure modes be prioritised. The international standard for the method is IEC 60812, whose third edition was published on 10 August 2018.¹

ICH Q9(R1) lists FMEA among the tools of quality risk management. It evaluates potential failure modes for processes and their likely effect on outcomes or product performance, relies on product and process understanding, and “methodically breaks down the analysis of complex processes into manageable steps”. It can be used to prioritise risks and monitor the effectiveness of risk controls, and applied to equipment, facilities and manufacturing operations.² WHO adds that it assumes a comprehensive understanding of the process, with critical process parameters defined before the assessment starts.³

2How to perform an FMEA, step by step.

1. Define the risk question and the scope. State the problem and assumptions, gather the background data, name a leader and set the deliverables.²
2. Assemble the team. Quality risk management is usually done by interdisciplinary teams, with experts from areas such as the quality unit, development, engineering, production and regulatory affairs, plus people who know the risk process.²
3. Map the process. A flow diagram covering all operations and controls in the process can help, including the steps before and after the one assessed.³
4. List failure modes, effects and causes. For each step, record how it could fail, what the effect on the product or process would be, and what could cause it.¹²
5. Score each failure mode. Rate severity, occurrence and detection on scales defined in advance.³
6. Calculate the RPN and rank. Multiply the three scores and rank the failure modes by the result.³²
7. Reduce the risk. For risks above the acceptable level, act to eliminate, contain, reduce or control the failure, by lowering severity or probability or by improving detection.²
8. Rescore and review. Risk reduction can introduce new risks, so revisit the assessment after controls are in place, and keep reviewing it as new knowledge arrives.²

3How to calculate the RPN.

WHO’s quality risk management guideline defines the risk priority number as a numeric assessment of risk for a process or process step in an FMEA: “Each failure mode gets a numeric score that quantifies likelihood of occurrence, likelihood of detection and severity of impact. The product of these three scores is the RPN for that failure mode.”³

RPN = severity × occurrence × detection

For a higher RPN to mean a higher priority, the detection scale has to run the opposite way to the other two: a failure that is hard to detect gets a high score. On scales a team has defined from 1 to 10, a failure mode scored 8 for severity, 3 for occurrence and 4 for detection has an RPN of 8 × 3 × 4 = 96.

4Setting scales and thresholds.

Neither ICH Q9(R1) nor WHO’s guideline sets the scales or the RPN at which action is required. WHO’s guideline says the scoring system and trigger points for mitigating action “are subjective so the rationale for score categorization should be defined in as much detail as possible”, and that the mitigating action is as important as the score.³ For its probability and impact matrix, WHO gives examples of five levels, scored 1 to 5, or weighted scores such as 1, 3, 5, 7 and 10.³ ICH Q9(R1) asks for qualitative descriptors such as high, medium and low to be defined in as much detail as possible, and for risk scores to rest on evidence, science and knowledge.² The acceptable level of risk is decided case by case.²

5Where the RPN misleads.

A product of three scores hides how it was reached. On the same scales, a failure mode rated 10, 2 and 3 and another rated 3, 5 and 4 both give an RPN of 60, though the first has a far more serious effect. The 2018 edition of IEC 60812 added alternative means of calculating risk priority numbers and a criticality matrix method.¹

Subjectivity is the other weakness. ICH Q9(R1) says it enters through poorly defined risk questions and “poorly designed risk scoring scales”, and asks for it to be managed by addressing bias and assumptions, using the tools properly and making full use of relevant data.² WHO tells manufacturers that inspectors will check whether risk assessments underrate likelihood or overrate detection so that patient risk is underestimated, and that the evidence behind statements should be robust to challenge.³ Detection still counts: ICH Q9(R1) notes that detection controls can reduce the probability of harm, even where a method has no separate detectability factor.²

6FMEA, FMECA and other tools.

An FMEA becomes a failure mode, effects and criticality analysis (FMECA) when it also examines the degree of severity, the probability of occurrence and the detectability of each failure mode; the product or process specifications should be established first, and the output is a relative risk score for each failure mode, used to rank them.² IEC 60812 calls the analysis FMECA when criticality is ranked using at least the severity of consequences.¹ Other tools in ICH Q9(R1) answer different questions: fault tree analysis works back from an assumed failure to its causes, HACCP sets critical control points, and risk ranking and filtering compares diverse risks.²

7How it runs in Seal.

Seal’s quality risk management blueprint keeps a process FMEA as its own template, holding failure modes, causes and controls. It can reference the same functions, controls, requirements and source evidence as other risk files while keeping its own terminology and assessment criteria. The categories, definitions and criteria are configured for your context, with their version and the basis for each estimate recorded, so a reviewer can see what an assigned score means. Each control links to its verification, and a failed test, complaint or deviation challenges the assessment it concerns.

When the analysis points to a stronger control, change control carries it with the verification the assessment will rely on. In process characterisation, prior knowledge and risk questions connect to the factors and quality attributes worth studying.