Regulations and compliance · Practical guide

ICH Q9(R1): reducing subjectivity in risk assessment

Two experienced colleagues assess the same change differently. One trusts an existing control, while the other questions its effectiveness. Averaging their scores conceals the disagreement. ICH Q9(R1) addresses subjectivity in quality risk management. QA, IT and process owners can make assumptions explicit, identify gaps in knowledge and apply criteria others can follow.

Table of contents

At a glance

How can you reduce subjectivity in an ICH Q9(R1) risk assessment?

ICH Q9(R1) does not seek to remove professional judgement. It explains how to recognise, manage and reduce subjectivity. Define the risk question, use sufficient knowledge, state assumptions and make rating criteria understandable. Match effort and formality to the importance, uncertainty and complexity of the decision. A deadline or shortage of resources does not justify an artificially low rating or inadequate formality.

  • Attribute differing assessments to different assumptions.
  • Scoring scales should include understandable criteria rather than mere numbers.
  • New insights may alter risk assessment and acceptance decisions.

Minimize subjectivity without implying a professional opinion.

ICH Q9(R1) considers quality risks in the pharmaceutical life cycle. Scientific findings and the protection of the patient form central points of reference. Chapter 5.3 explicitly deals with subjectivity: Different perceptions, assumptions and assessment methods can influence decisions. This influence cannot be completely eliminated; however, it should be recognized and reduced.

The version of the EMA used herein incorporates Corr.2 dated 23 January 2025. The document history indicates a typographical correction, not a new technical revision Q9(R2). The following example relates to an assumed computerized material process within the human medicinal product GMP environment. It does not constitute a universal evaluation guidance for every business or safety decision.

Sources: [1] [2]

Formulate the question in such a way that it becomes answerable

The question "Is the system safe?" is too broad for a specific assessment. Instead, describe a potential failure mode, its conditions, and the relevant consequence. For example, the question is: Could a faulty unit assignment following a change to an interface lead to the use of an incorrect material quantity during the manufacturing process? This identifies the process, the change, and the potential impact on quality.

Distinguish known facts from assumptions. Known elements include the modified field, the intended unit, and the current data pathway. The question of whether a claimed plausibility check actually detects the error remains unresolved. Q9(R1) Chapter 4.2 and 4.3 emphasizes the importance of risk assessment, available information, and acknowledged uncertainty. A broad table header does not substitute for this preparatory work.

Sources: [1]

Associate evaluation criteria with specific evidence

A number may appear objective, even if its meaning is interpreted differently within the team. Therefore, prior to classification, agree on which observations support a judgment. For qualitative categories, descriptions should be as clear as possible. An ordinal risk score does not automatically represent a measured probability; multiplication of weak assumptions does not yield strong data.

The following completed matrix serves as an internal reference tool. It does not include any ICH-defined scale. Its purpose is to distinguish statements regarding control and knowledge gaps prior to making a decision.

Criteria for assessing a change of units
Assessment questionSupportable source informationUnspecified assumptionAppropriate next test result
What could go wrong? A field in a interface changes its unit. All involved systems interpret the value as equal. Compare field definitions against actual data transfers.
Which consequence is relevant? The value affects the supplied material quantity. A user would promptly identify any incorrect value. Investigate the workflow and clearly identifiable information in detail.
What control is in place? A plausibility check is described. It exactly covers the new defect scenario. Run a representative test that should pass and one that should fail.
How reliable is the classification? Existing tests cover the old field status. The previous evidence remains transferable without change. Assess the impact of changes and coverage of existing tests.
What does a decision permit? Intended use and potential consequences are known. The schedule replaces missing evidence. Explicitly define open verification and authorized decision-making.

Sources: [1]

Two assessments based on the same facts resolve.

Example: A transfer interface is intended to transfer 5.0 kg. An incorrect unit reference could result in 5.0 g instead. Mathematically, 5.0 kg corresponds exactly to 5,000 g; 5.0 g would represent one-thousandth of the intended quantity. These values apply solely to this example and do not constitute product specifications.

Employee A assesses the risk as low, assuming an effective plausibility check is in place. Employee B assigns a higher risk level, as the available tests only reflect the current field status. The appropriate next step is not to average the two assessments. Instead, the differing assumptions must first be evaluated. The resulting assessment may then support a justified re-evaluation; however, this does not trigger automatic release.

Example: completed decision pathway in case of conflicting evaluation
StepEstablished factWorking decisionAvailable evidence
Starting point 5.0 kg intended; incorrect reference could deliver 5.0 g. Consider specific unit-level errors as a scenario. Value, unit, data path, and potential consequences documented
Evaluation A Control is considered effective. Keep the assumption open until the evidence confirms it. Stated function and associated evidence
Evaluation B The test validation applies only to the previous version. Clarify the scope of the altered state. Assigned to configuration and test setup
Targeted check Correct and incorrect units defined as separate test cases. Define the expected behavior prior to execution. Test protocol with required criteria; no claimed result yet
The following decision Results and remaining uncertainties are evaluated. Evaluate control, additional actions, or justified acceptance. Decision with conditions and responsible party
Later updated finding Another transmission pathway is identified. Reassess the validity of previous assumptions. Updated risk assessment instead of unchanged old risk score

Sources: [1] [3]

Align formality with significance and uncertainty

Chapter 5.1 describes formality as a continuum. Key factors include uncertainty, the significance of the decision, and complexity. A brief assessment may be appropriate if the situation is understood and sufficient evidence is available. In the presence of significant knowledge gaps, a more detailed evaluation may be required, even if the affected system is small.

Resource limitations should not be used as justification for reduced formality, as specified in Q9(R1). This does not imply a requirement to provide the most comprehensive documentation package. Rather, the chosen approach must be professionally justified. Evaluate whether the additional information actually improves the decision, rather than simply filling out more evaluation fields.

Sources: [1]

Address disputes and decision authority in a traceable manner

A professional disagreement may indicate a lack of knowledge or differing perceptions regarding the extent of harm. Therefore, document the essence of the discrepancy. An experienced moderator can assist in separating assumptions and establishing concrete evidence. Q9(R1) describes interdisciplinary teams as a common, though not universally required, approach; it does not imply a fixed role assignment for every situation.

The decision must take into account the remaining uncertainties and conditions. A low score does not justify a practice that would be inappropriate under applicable requirements. Similarly, an accepted residual uncertainty should not be considered resolved merely because the assessment is later transferred into a report. Document clearly who is authorized to make the decision and what triggers a re-evaluation.

Sources: [1]

Checklist to Prevent Unjustified Classifications

This dedicated checklist supports the preparation and facilitation of a specific risk assessment. It neither mandates an additional standalone risk analysis nor requires a universal second signature. Existing procedures may already include the necessary steps.

  • The risk question identifies failure modes, conditions, and potential quality impacts.
  • Known facts are separated from assumptions.
  • Uncertainty and lack of knowledge remain visible.
  • Appropriate scaling criteria are established for this case.
  • Existing controls are evaluated based on appropriate evidence.
  • Different assessments are clarified in content, not merely averaged.
  • Formality aligns with significance, uncertainty, and complexity.
  • The timing and budget pressure do not change the classification quietly.
  • Decision, conditions, and rationale for the re-evaluation are documented.

Sources: [1]

Assessing in operation maintains learnability

Chapter 4.6 provides for the review of risk management outcomes based on new insights and experiences. Planned changes and unexpected events may affect the original decision. The review frequency is risk-based; a one-size-fits-all annual deadline for each review cannot be derived from this.

For SMEs, a brief review of recurring assumptions is warranted: Which controls are frequently considered to be in place? Where is reliable evidence lacking? Which assessments have had to be revised following new information? Leverage these insights to enhance criteria and knowledge. The objective is to support better decision-making for the specific process, not to achieve a uniformly consistent color across every risk matrix.

Sources: [1]

Project experience and collaboration

Resolve differences in risk ratings by examining their assumptions

If team members assess the same change differently, we first record each view and its assumptions. Together, we trace potential failure consequences and check the evidence behind claimed controls. A more precise risk question may suffice. In other cases, essential information must be obtained first. We facilitate the comparison and document the reasoning for the decision. An average score or a tight deadline cannot resolve the underlying uncertainty.

Frequently asked questions about Subjectivity under ICH Q9

Does ICH Q9(R1) require fully objective risk assessments?

The guideline acknowledges that subjectivity cannot be entirely eliminated. It aims to recognize, control, and minimize subjectivity. Scientific understanding, appropriate criteria, and transparent assumptions are essential. Therefore, a form filled with numbers still does not constitute objective or adequately justified evaluation.

Sources: [1]

What is the difference between uncertainty and high risk?

Uncertainty describes a lack of knowledge regarding hazards, adverse effects, or their associated risks. It is not automatically equivalent to a high risk. However, it influences the robustness of a risk assessment and determines the level of formalities or additional evidence required. Therefore, risk assessment and knowledge boundaries should be kept distinctly visible.

Sources: [1]

May we simply average different risk scores?

An arithmetic mean may mask different underlying assumptions. First, clarify whether all stakeholders evaluate the same error, the same consequence, and the same control effect. The proposed approach is derived from ICH Q9's guidance on handling assumptions and subjectivity; it is not a universally mandated prohibition against calculation.

Sources: [1]

Does Q9(R1) specify a fixed risk priority number?

No. The guideline describes different tools and qualitative or quantitative approaches. It does not establish a universal risk priority score or a general acceptance criterion applicable to all processes. A used point system requires appropriate definitions and must not overstate its credibility.

Sources: [1]

Is a smaller document package acceptable for SMEs?

The formality should align with the significance, uncertainty, and complexity involved. A streamlined approach may be appropriate if it adequately addresses the specific case. However, limited resources alone do not justify reduced formality. Professional appropriateness is paramount, not the size of the organization or a preferred page count.

Sources: [1]

Must every risk analysis be conducted by a cross-functional team?

Q9(R1) describes such teams as a typical, but not absolutely mandatory, approach. Appropriate knowledge of the subject matter and risk management is required. The scope, involvement, and decision-making authority should be defined such that the specific question is adequately addressed; a universal role list does not provide a prescriptive framework.

Sources: [1]

Can a low score justify a regulatory non-compliant practice?

No. Q9(R1) explicitly states that Quality Risk Management is not intended to justify otherwise inappropriate practices under regulations or guidance. Therefore, a cost-benefit analysis does not replace the evaluation of applicable requirements and their fulfillment.

Sources: [1]

Must every change trigger a completely new risk assessment?

The impact on existing knowledge and current decisions must be assessed. Q9(R1) also addresses decisions within existing processes and does not require a comprehensive new analysis for every change. However, new information may invalidate assumptions and necessitate a targeted reassessment.

Sources: [1]

Should each risk assessment be reviewed annually?

Q9(R1) does not specify a universal time frame for each evaluation. Chapter 4.6 links reviews to new knowledge, events, and risk levels. Establish appropriate triggers and a justified frequency. A fixed internal organizational deadline should be distinguished from any purported general ICH guidance.

Sources: [1]

How do we identify an improved assessment in the unit example?

At the traceable link between defect signature, existing control, appropriate evidence, and decision. The calculation 5.0 kg = 5,000 g only clarifies the magnitude. Whether the interface detects a defect and what level of usage is acceptable must be assessed based on the actual system; a lower risk score alone does not indicate improvement.

Sources: [1] [3]

Sources and review date

Source date: 7 October 2026. The EMA Step-5 version of ICH Q9(R1), Corr.2 dated 23 January 2025 is applied; the history indicates a typographical correction. Annex 11: Revision January 2011, as current in the EU GMP Index. Matrices and expected results are internal work aids and do not represent completed product testing.

  1. EMA: ICH Q9(R1) – Quality Risk Management, Corr.2EMA version dated 23 January 2025, Corr.2. This is Q9(R1), applicable in the EU since 26 July 2023. See sections 1–6, particularly 4.3 and 5.1–5.3.
  2. European Commission: EudraLex Volume 4Status as of 07.10.2026; Human medicinal products and veterinary law separated.
  3. EU GMP Annex 11 – Computerised SystemsRevision January 2011; listed in the human medicinal product section. Principle, §§1–17.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Jointly develop the specific evidence process

We support your team in sharpening specific risk questions, providing justifications for assessments based on available evidence, and deriving appropriate testing and decision documentation from those evaluations.

  • GAMP 5 consulting →

    Understand how GAMP 5, EU GMP Annex 11 and 21 CFR Part 11 apply to your systems. We translate the relevant requirements into proportionate validation work and evidence your team can defend in an audit.

Send a project enquiry Customer testimonials →