Short answer
ISO 14971:2019 replaced the 2007 edition and is the current state of the art for medical device risk management; in Europe it is EN ISO 14971:2019, and amendment A11:2021 relates it to the MDR. For a software risk file the main changes are explicit risk acceptability criteria defined before analysis, a defined method for evaluating overall residual risk, an explicit benefit-risk conclusion, a documented link from post-market data back to the risk file, terminology aligned with the 2019 definitions, and coverage of security-related risks. IEC/TR 80002-1 still helps with software examples but was written for the 2007 edition, so use it together with ISO 14971:2019 and ISO/TR 24971:2020.
Why a 2007-era risk file struggles in EU MDR review
ISO 14971:2019 (third edition) was published in December 2019 and replaced ISO 14971:2007. In Europe it is published as EN ISO 14971:2019; amendment A11:2021 added the European annexes that relate the standard to the requirements of the MDR and IVDR. Only standards whose references are published in the Official Journal give a presumption of conformity with the MDR (Article 8), so check the Commission's list of harmonised standards for the current citation. The 2007 edition is no longer the current state of the art. This matters for the whole technical file — see also our IEC 62304 guide, which must align with your risk file.
In our documentation audits, we still regularly see legacy risk management files built on the ISO 14971:2007 structure and terminology, and these typically generate findings when an NB reviews them against the current edition and the MDR. The good news: updating a 2007 risk file to 2019 is entirely manageable if you know what needs to change — a technical documentation gap analysis will show which parts of your file are affected.
The five key changes in ISO 14971:2019
1. Benefit and benefit-risk: new definitions, plus the MDR's stricter expectation
The 2019 edition defines "benefit" and renames the 2007 "risk/benefit analysis" to "benefit-risk analysis". Within the standard, a benefit-risk analysis is used when a residual risk — or the overall residual risk — is not acceptable under the criteria in your risk management plan: you then have to show that the benefits of the intended use outweigh that residual risk.
The EU MDR goes further. Annex I requires that all known and foreseeable risks are minimised and acceptable when weighed against the evaluated benefits, and that the overall residual risk is acceptable. In practical terms, your risk management report should include a section that states what clinical benefits the device provides, what residual risks remain after all controls, and an explicit conclusion on the benefit-risk balance for the intended use and patient population. This conclusion must be substantiated — cross-referencing your clinical evaluation is the typical approach.
2. Risk acceptability criteria — including for the overall residual risk
Risk acceptability criteria were already required in the 2007 edition. The 2019 edition makes the framework more explicit: top management defines a policy for establishing the criteria, taking into account applicable regulations and standards, the generally acknowledged state of the art and known stakeholder concerns; and the risk management plan must also define a method to evaluate the overall residual risk and the criteria for its acceptability.
Practically, this means your risk management plan must define: the probability and severity scales you will use, the risk matrix or criteria that define what constitutes acceptable and unacceptable risk, how you will accept risks when the probability of harm cannot be estimated, and how you will evaluate the overall residual risk. These criteria must be defined before the analysis — you cannot set your criteria after seeing your risk analysis results. In the EU, note that the MDR requires risks to be reduced as far as possible without adversely affecting the benefit-risk ratio (Annex I, section 2), so an ALARP-style policy that lets economic considerations stop risk reduction does not fit MDR expectations.
3. Expanded production and post-production requirements
The 2019 version expands the requirements for feeding post-market data back into risk management. Your risk management process must have a defined mechanism for collecting and reviewing information from production and post-production — complaint data, vigilance reports, literature surveillance, post-market performance data — and evaluating this information for its implications on risk estimates and risk controls.
This requirement directly connects your risk management file to your post-market surveillance process. Your QMS must have a documented procedure that links PMS data review to risk management file updates.
4. Overall residual risk evaluation
An evaluation of the overall residual risk already existed in the 2007 edition, but the 2019 edition tightens it: the method and acceptability criteria must be defined in the risk management plan, the evaluation must consider the contributions of all individual residual risks together, and significant residual risks must be disclosed to users in the accompanying documentation. This is different from evaluating each risk individually and must be documented separately.
5. Risk management review and report
The risk management report is the culminating document. Under the 2019 edition, before commercial release you carry out a risk management review confirming that the risk management plan has been appropriately implemented, that the overall residual risk is acceptable, and that appropriate methods are in place to collect and review production and post-production information; the results are documented in the risk management report. For the EU, add the overall benefit-risk conclusion and the cross-reference to your clinical evaluation described above.
Software-specific risk considerations under ISO 14971:2019
For SaMD, ISO 14971 risk management must address software-specific hazards that are not present in hardware devices. The technical report IEC/TR 80002-1:2009 (Guidance on the application of ISO 14971 to medical device software) provides software-specific guidance; it was written for the 2007 edition, so read it together with ISO 14971:2019 and ISO/TR 24971:2020. Software-specific hazards include:
- Algorithmic errors: Errors in the software's logic or processing that produce incorrect outputs — incorrect diagnoses, false positives or negatives, inaccurate calculations. For AI software, this includes model prediction errors, distribution shift, and edge case failures.
- Interface hazards: Risks arising from software-hardware interfaces (for embedded software), software-user interfaces (use errors driven by poor UX), and software-software interfaces (integration errors with other clinical systems, APIs, databases).
- SOUP hazards: Risks introduced by third-party software components — library failures, known vulnerabilities in open-source dependencies, version incompatibilities.
- Cybersecurity risks: Vulnerabilities that could compromise the software's integrity, leading to incorrect clinical outputs or denial of service. ISO 14971:2019 explicitly lists data and systems security among the risks its process applies to, and MDR Annex I requires software to be developed taking into account risk management, including information security. See our cybersecurity guide — MDCG 2019-16 provides more specific guidance on this for EU MDR.
- Change-related risks: Risks introduced by software updates — including unintended regression effects on previously validated functionality and performance degradation in production environments.