Medical software development standards are the shared rulebook for how software that touches patient care should be planned, built, tested, released, and maintained. They span the software life cycle (IEC 62304), the quality system surrounding it (ISO 13485), and risk management (ISO 14971). Regulators such as the FDA and EU notified bodies expect your records to demonstrate that you followed them. This guide walks through each standard in simple terms, shows how US and EU rules build on them, and explains how to apply them without swamping your team in paperwork.
Key Takeaways
- Medical software development standards center on IEC 62304, with ISO 13485 and ISO 14971 covering quality and risk around it.
- What the software is intended to do, not the technology behind it, determines whether it is regulated as a medical device.
- The A, B, or C safety class reflects the harm the software might contribute to, and it determines how much evidence you must produce.
- The FDA’s QMSR has been in effect since February 2, 2026, and under EU MDR Rule 11 most clinical decision software lands in class IIa or higher.
- Cybersecurity, third-party code, and traceability should be handled from the first day, not squeezed in before release.
- AI-enabled software requires a plan for model updates, while privacy, interoperability, and post-market duties apply alongside the safety standards.
- IEC 62304 is under revision, so keep your processes adaptable.
What Counts as Medical Software?
Regulators focus on what your software is supposed to do, not on how it was built. When the intended purpose is to diagnose, prevent, monitor, or treat a disease or condition, the software is likely a medical device, no matter whether it runs on a phone, in the cloud, or inside a pump.
Two terms appear again and again. Software in a medical device (SiMD) operates or controls physical hardware, an infusion pump for instance. Software as a medical device (SaMD) performs the medical function by itself, typically on general-purpose hardware. A symptom checker that proposes a probable diagnosis counts as SaMD, whereas an app that merely sends appointment reminders does not, since it makes no medical decision.
State the intended purpose with care and use the same wording in every document. Classification, testing requirements, and the application of medical software development standards to your product all depend on its intended purpose. Solid medical device software development begins with that single sentence.Understanding the regulatory requirements early is particularly important for medical software development standards, as it helps teams identify whether their product falls under medical device regulations before moving into development.
Core Medical Software Development Standards at a Glance
In practice, medical software development standards act as a small family that works together. The table below shows how the main medical device software standards fit together.
| Standard | What it covers | Why it matters to your team |
| IEC 62304 | The software life cycle, from planning through maintenance | Forms the backbone of your development process |
| ISO 13485 | Quality management system for medical device organizations | Handles document control, design reviews, and supplier oversight |
| ISO 14971 | Risk management through the whole product life cycle | Drives safety classification and risk controls |
| IEC 62366-1 | Usability engineering | Cuts down use errors that could harm users |
| IEC 82304-1 | Safety of health software products | Applies to standalone software sold without dedicated hardware |
| IEC 81001-5-1 | Security activities throughout the software life cycle | Fits security work into an IEC 62304 process |
| IEC 60601-1 | Safety of medical electrical equipment | Relevant when your software runs inside that equipment |
Current editions: IEC 62304:2006+A1:2015, ISO 13485:2016, ISO 14971:2019, IEC 62366-1:2015+A1:2020, IEC 82304-1:2016, and IEC 81001-5-1:2021.
IEC 62304 in Plain Language
Among the essential medical software development standards, IEC 62304 is one of the first that development teams need to understand. It lays out the processes needed over the software life cycle: development, maintenance, risk management, configuration management, and problem resolution. Within development you plan, gather requirements, design the architecture and its details, implement and verify units, integrate, test the whole system, and release.
Two things tend to surprise people. The standard does not require a waterfall approach, and it does not set document formats. What it wants is proof that the required tasks happened, and your own tools can store that proof provided it is controlled and traceable.For teams working on medical software development standards, following IEC 62304 from the beginning can make it easier to manage software requirements, testing, documentation, and future updates without creating unnecessary compliance challenges.
Software Safety Classes
The class indicates how rigorous your process must be. It is based on the most serious harm the software could contribute to, on the assumption that the software may fail.
- Class A: there is no possibility of injury or damage to health.
- Class B: non-serious injury is possible.
- Class C: death or serious injury is possible.
Risk controls that sit outside the software can affect the class, but the clinical benefit of the product cannot. Higher classes demand more work: unit verification applies to classes B and C, for example, and class C also expects detailed design documentation.
Third-Party Code and SOUP
Software of unknown provenance (SOUP) covers any component that was not built for medical use, such as an operating system, a database, or an open-source library. Every such component calls for defined requirements, a risk review, and a plan for handling updates. Keeping a living SOUP register that lists version, purpose, and known issues makes audits far smoother.
A Note on Edition 2
Like other medical device software standards, IEC 62304 is being revised. Draft material suggests a broader scope that reaches health software, along with two rigor levels instead of three classes. Sources disagree on timing and the text may still shift, so continue working to the current edition and check the IEC website before planning around the new one.
ISO 13485 and ISO 14971: The Frame Around Your Code
ISO 13485 is the quality management system standard for organizations that make medical devices. For software teams it appears as document and record control, design reviews, supplier management, and corrective and preventive action. It also expects you to validate software tools that influence quality, such as your build pipeline or issue tracker, to a degree that matches the risk.
ISO 14971 deals with risk management, and IEC 62304 depends on it, so software risk work does not stand apart. The system-level analysis identifies hazardous situations, and the software side examines how failures might contribute to them. IEC/TR 80002-1 provides guidance on applying ISO 14971 to software.Medical software development standards requires a structured approach to quality management and risk assessment. Implementing ISO 13485 and ISO 14971 early helps smaller teams establish reliable processes as their products evolve.
Usability, Security, and Health Software Standards
Besides life cycle and risk management, several other medical software development standards help address usability, security, and product safety.Poor usability harms patients as well, so IEC 62366-1 expects you to study how people actually use the product and to design out use errors. This is especially important for dosing screens and alarms.
IEC 82304-1 is aimed at health software products that run on general computing platforms and ship without dedicated hardware. IEC 81001-5-1 addresses security activities across the life cycle and was designed to work next to IEC 62304. Expect threat modeling, secure coding, vulnerability management, and update processes.
Verification and Validation: Not the Same Thing
People often use these two words interchangeably. Verification asks whether you built the product correctly, in other words whether it satisfies its specified requirements; it includes reviews and unit, integration, and system tests. Validation asks whether you built the correct product, meaning that it serves real users in the intended setting.
IEC 62304 is mostly concerned with verification within the software life cycle. Validation generally belongs to your quality system and design controls, and it often relies on usability evaluation and, when needed, clinical evidence. Plan for both from the beginning, since validation results can send you back to the requirements.
How the FDA and EU Use Medical Software Development Standards
United States
The FDA’s Quality Management System Regulation (QMSR) came into effect on February 2, 2026. It amends 21 CFR Part 820 and brings in ISO 13485:2016 by reference, though FDA-specific requirements continue to apply.
For the software content of a submission, the FDA’s guidance on premarket submissions for device software functions, issued in June 2023, replaced the 2005 version. It sets two documentation levels, basic and enhanced, chosen according to the risk of the software function. The FDA recognizes IEC 62304 as a consensus standard, and the guidance follows the same life cycle logic.
Connected products face an additional layer. Section 524B of the FD&C Act, in effect since March 29, 2023, requires cyber devices to include a postmarket vulnerability plan, evidence of secure design, and a software bill of materials in a machine-readable format such as SPDX or CycloneDX. The FDA reissued its cybersecurity guidance on February 3, 2026.
European Union
Under the EU MDR, software is classified using Rule 11 in Annex VIII. MDCG 2019-11, revised in June 2025, explains how to decide whether software is a medical device and which class it falls into. Software that provides information for diagnostic or therapeutic decisions is generally class IIa or higher. Software that monitors physiological processes is class IIa, or IIb when the vital parameters could change in a way that creates immediate danger. Class I is a small leftover group. From class IIa upward, a notified body takes part.
Industry bodies point out that the European Commission proposed a revised Rule 11 in December 2025, so keep watching it. If your product uses AI, see the AI section below, which adds extra layers in both regions.
Other Markets
Other countries, including the UK, Canada, Australia, and Japan, operate their own systems. Most rely on ISO 13485 and IMDRF ideas, yet classification, registration, and local representative rules differ. Before launching abroad, review each regulator’s current software rules rather than assuming a CE or FDA route will carry over.
SaMD Risk Categories: The IMDRF Framework
Many regulators begin from a shared starting point when judging SaMD risk: a framework by the International Medical Device Regulators Forum (IMDRF). It poses two questions. How serious is the health situation the software addresses (non-serious, serious, or critical)? And how strongly does its output influence the healthcare decision, that is, does it inform, drive, or directly support treatment and diagnosis?
Putting the two answers together yields four categories, I to IV, where IV has the greatest impact. Software that supports diagnosing a critical condition falls into category IV, while software that merely informs the management of a non-serious condition is category I.
The framework is a conceptual tool, not a legal class. The FDA and the EU each convert risk into their own device classes under their own rules, and MDCG 2019-11 indicates roughly how IMDRF categories map to MDR classes. Only the MDR classification is binding, but working through the IMDRF grid early gives you a sound basis for your classification and helps you plan clinical evidence.
AI and Machine Learning Software
AI-enabled software brings a twist, because the model can change after release. In the US, the FDA published final guidance on predetermined change control plans (PCCPs) on December 4, 2024. A PCCP allows a manufacturer to describe planned model updates, the testing behind them, and the impact assessment in advance, so authorized changes do not each require a fresh submission. The guidance applies to AI-enabled device software functions in general, not just machine learning.
In the EU, AI that is a medical device, or a safety component of one, may be treated as high-risk under the AI Act. After the Digital Omnibus, reports indicate that high-risk obligations for AI in regulated products such as medical devices apply from August 2, 2028. Use that as a planning marker and verify it against official EU sources, because other AI Act obligations follow different dates.
For your life cycle this means recording how training, tuning, and test data are managed, planning checks for bias, and handling every model update as a controlled change. ISO 14971 stays the risk foundation, and AAMI TIR34971:2023 gives guidance on applying it to machine learning.
Privacy, Data Protection, and Interoperability
Safety standards say nothing about privacy, so privacy law applies in parallel. In the US, HIPAA covers covered entities and their business associates, so a software vendor that handles protected health information for a provider will usually need a business associate agreement. In the EU, the GDPR classes health data as a special category, with tighter conditions for processing it.
ISO/IEC 27001 is a popular framework for an information security management system. It supports IEC 81001-5-1 without replacing it, since one governs the organization and the other governs the product life cycle.
Interoperability needs the same attention. Connections to hospital systems, devices, and cloud services belong to your architecture under IEC 62304, and each is a point where data can go wrong. Exchange formats such as HL7 FHIR are widespread in healthcare, so decide which ones you support, what you assume about incoming data, and how failures are handled and tested.
Can Agile Teams Follow Medical Software Development Standards?
Yes. IEC 62304 does not prescribe a life cycle model, so Scrum, Kanban, or a mix can work as long as you map your activities to the standard and justify the approach. AAMI TIR45 provides practical guidance for agile in this setting.
The difference lies in your definition of done. A story is complete only when requirements are reviewed, risks are considered, tests are linked, and the evidence is under version control. Your ticket system, repository, and CI pipeline can hold most of it, keeping medical device software development both quick and auditable.Agile methods can be especially useful in medical software development standards because they allow teams to build and improve products in smaller stages while maintaining the documentation, testing, and traceability required for compliance.
After Release: Post-Market Surveillance, Maintenance, and Legacy Software
Release is a milestone rather than the finish line. In the EU, every device needs a post-market surveillance plan. Devices in class IIa and above also generate periodic safety update reports, and serious incidents must be reported to the authorities. In the US, complaint handling, medical device reporting, and corrective action stay on your list of obligations. What matters is a feedback loop: field data should return to risk management, change control, and your next release.
For software, prepare for the usual triggers: newly found vulnerabilities in third-party code, operating system changes, and anomalies reported by users. Your change control should specify when a change is a simple fix and when it demands a new risk review, extra testing, or even a new regulatory submission.
Legacy software is a special case. When a product is already on the market but cannot show that it followed the current IEC 62304 process, the standard offers a path: assess the risks, identify the gaps, and add the missing evidence. Take care here, because the FDA expects documentation to be created prospectively, so plan the clean-up together with your regulatory team.

Common Mistakes to Avoid
Most problems with medical software development standards stem from timing rather than talent. Watch out for these:
- Treating risk management as a task for just before release. It works only if it starts with the first requirement.
- Writing documentation retroactively. The FDA’s guidance says software documentation should be established prospectively.
- Reading the safety class as proof that the product is safe. It sets how rigorous the process is, and nothing beyond that.
- Leaving SOUP unmanaged, including its updates and known vulnerabilities.
- Relying on a vague intended purpose, which makes classification hard to defend.
- Forgetting maintenance. Problem resolution, change control, and security updates carry on after release. Avoiding these mistakes helps medical software development standards maintain compliance and reduce development risks.
A Practical Five-Step Path to Compliance
Here is a simple way to turn medical software development standards into a working plan.
- State the intended purpose and confirm whether your software is a medical device in each target market.
- Classify it under regional rules, then assign the IEC 62304 safety class.
- Set up your quality system and software development plan, covering traceability, change control, and SOUP handling.
- Build, verify, integrate, and test, with risk and security work running in parallel.
- Release with a full record, then monitor, maintain, and update. A clear compliance strategy is essential for medical software development standards , particularly when planning product classification, regulatory submissions, risk management, and long-term maintenance.
Frequently Asked Questions
What are the main medical software development standards?
The core set consists of IEC 62304 for the software life cycle, ISO 13485 for quality management, and ISO 14971 for risk management. Most teams add IEC 62366-1 for usability, IEC 81001-5-1 for security, and IEC 82304-1 for standalone health software. Regional rules such as the FDA’s QMSR and the EU MDR sit on top.
Is IEC 62304 mandatory?
The standard is voluntary, as are most medical device software standards. Still, regulators and notified bodies expect you either to follow it or to justify an alternative. The FDA recognizes it as a consensus standard, which can make a software review go more smoothly.
What are the three software safety classes in IEC 62304?
Class A means no injury or damage to health is possible, class B means non-serious injury is possible, and class C means death or serious injury is possible. A higher class requires more verification and more design documentation.
Does IEC 62304 apply to mobile apps and SaMD?
Yes, whenever the software has a medical purpose. Medical device software standards care about intent, and the platform makes no difference. A diagnostic app on a phone, a cloud service, and embedded firmware all count as medical device software when their intended purpose is medical.
What is the difference between ISO 13485 and IEC 62304?
ISO 13485 is the quality management system for the entire organization. IEC 62304 defines the software life cycle processes that operate inside that system. You need both.
Bottom Line
Medical software development standards may look like a wall of acronyms, yet they follow simple logic: know your intended purpose, understand the risk, build with evidence, and keep watching after release. Begin with IEC 62304, ISO 13485, and ISO 14971, add the usability and security standards, and check the FDA and EU rules for your market. Treat medical device software development as an ongoing habit of building evidence, not a last hurdle.
Standards and guidance keep changing, so confirm current editions with official sources before finalizing your plan. This article offers general information, not regulatory or legal advice.




