FDA publications issued in late October 2026 directly affect how manufacturers plan, justify, and present medical device evidence. Final classification orders set out clinical performance expectations for three categories of diagnostic software. A revised draft guidance addresses drug-delivery performance across the product lifecycle, including post-change assessment. A separate draft guidance describes the proposed electronic submission process for investigational device exemption applications.
For manufacturers and regulatory professionals, the immediate priority is to identify evidence gaps before study designs and submission packages are finalized. The publications do not have the same regulatory status. The classification orders establish binding US requirements for the specified device types, whereas the two guidance documents remain in draft form
1 Clinical Evidence Requirements for Diagnostic Software
On 30 October 2026, FDA published final orders codifying Class II classifications and special controls for skin-cancer adjunctive diagnostic software, fibrotic-lung-disease referral software, and over-the-counter sleep-apnea risk assessment. The underlying classifications date to 2024. The current development is their codification, rather than a new clinical finding or first authorization. [1–3]
For skin-cancer software under 21 CFR 878.1830, the intended role is a second read for physicians without training in the clinical diagnosis and management of skin cancer. Clinical data must demonstrate superior accuracy with device assistance compared with unaided users. Testing must represent relevant skin phototypes, risk factors, anatomical areas, and mimic lesions. Standalone output testing must demonstrate at least 90% sensitivity for lesions with high metastatic potential, unless a clinical consideration justifies a lower value; specificity also requires clinical justification. [1]
For evidence planning, algorithm accuracy alone may not address the central regulatory question: whether use of the device improves the intended physician’s characterization of suspicious lesions. Manufacturers developing similar adjunctive products should confirm that their evidence reflects both the intended users and the actual assisted-use workflow. A reader study limited to specialists may not support a claim directed to a different user population.
For fibrotic-lung-disease referral software under 21 CFR 892.2085, FDA requires a justified reference standard and an appropriate clinical control. Diagnostic-accuracy measures must be prespecified. Test datasets must contain sufficient cases in important cohorts to characterize subgroup estimates and confidence intervals, including clinically relevant demographics, confounders, early disease, and image-acquisition characteristics. The intended output supports referral within the diagnostic assessment process. [2]
This makes subgroup planning a protocol-level consideration. A large overall dataset may still be insufficient to characterize performance in a clinically important intended-use subgroup. Before data collection begins, teams should define the subgroup conclusions required, establish the necessary case mix, and confirm with a statistician that the planned sample will provide adequate precision.
For over-the-counter sleep-apnea risk assessment under 21 CFR 868.2378, clinical testing must include a representative population with symptomatic and asymptomatic participants. Requirements include evaluating relevant subgroups, comparing performance with a justified clinical comparator such as polysomnography, and assessing consistency across expected data sources and quality. For machine-learning algorithms, clinical validation must use a dataset separate from training. The device provides a risk notification and is not a standalone diagnostic. [3]
Manufacturers should determine whether the development dataset represents the population expected to use the product. Evidence generated in an enriched specialist population requires a clear applicability assessment before it can support a consumer-facing claim. The regulatory submission should explain recruitment, exclusions, disease spectrum, and how the tested population relates to the intended users.
2 Essential Drug Delivery Performance Characteristics
FDA’s revised draft, Essential Drug Delivery Performance Characteristics for Devices Intended to Deliver Drugs and Biological Products, is dated October 2026, Revision 1. Its availability notice was published on 30 October. It replaces the July 2024 draft on essential drug-delivery outputs, clarifies testing and control recommendations, and aligns the document with the Quality Management System Regulation. It remains nonbinding and not for implementation. Comments are due on 28 January 2027. [4]
The draft recommends clinically meaningful specifications for essential drug-delivery device performance characteristics. Supporting justification may include literature, anthropometric data, or simulated bench testing. Verification should account for relevant storage, shipping, and use conditions. The document also addresses performance at expiry and assessment of design or manufacturing changes, including whether existing verification and validation remain applicable. [5]
Manufacturers should use this framework to connect engineering acceptance criteria to the intended patient population and delivery function. Each proposed specification should be supported by a documented rationale explaining why the range is appropriate for the intended use. Regulatory and engineering teams should align on that rationale before relying on a passing test as support for a clinical claim.
The draft also provides a clear basis for assessing evidence reuse after a change. The assessment should identify the affected delivery characteristic, compare the tested and proposed configurations, and explain which existing results remain applicable. If additional testing is proposed, the plan should state the uncertainty that testing is intended to resolve. If no new testing is proposed, the rationale should be equally clear. This is an evidence-planning principle, not an indication that every modification requires a new clinical study.
SOTA and literature-review teams should record the specific purpose served by each source: defining a clinically acceptable specification, characterizing intended users, or supporting the relevance of a test condition. Relevance to a device category does not, by itself, establish that a publication supports a particular acceptance criterion.
3 Electronic Submissions for IDE Applications
FDA issued Electronic Submission Template for Investigational Device Exemption Applications on 28 October 2026. The draft describes IDE eSTAR content and the proposed electronic-submission framework. It covers original applications and related submissions, with specified exemptions. FDA intends to allow at least one year after final guidance before requiring covered electronic submissions. eSTAR use is currently voluntary. Comments are due on 28 December 2026. [6]
The template organizes material including prior investigations, product descriptions, investigational plans, and consent documentation. FDA’s draft is intended to improve submission consistency and review efficiency. It does not establish a new universal clinical-performance threshold. [6]
Sponsors should begin by mapping existing documents to the proposed structure and assigning ownership for each contribution. This review may identify version discrepancies, incomplete descriptions of prior testing, or inconsistencies between the investigational plan and the device description. Addressing these issues during study design can reduce avoidable rework when the application is assembled.
A structured submission still requires a coherent scientific rationale. Teams should retain the reasoning that links prior evidence, residual uncertainties, and the proposed investigation. Completing the required fields does not, on its own, demonstrate that the study addresses the appropriate clinical question.
Implications for Evidence Programs
Manufacturers pursuing both US and European pathways should review an active product against its intended use and claims. For each claim, identify the supporting evidence, record the evaluated device configuration, and state any limits on applicability. The requirements of each jurisdiction should then be assessed independently.
For a CER, this review can distinguish standalone performance evidence from evidence generated with the intended user. For an IVD PER, scientific validity, analytical performance, and clinical performance should be considered separately when assessing gaps or changes. These FDA publications do not modify EU MDR or IVDR obligations. Applying their study-design principles within a European programme is a methodological decision that requires its own justification.
The outcome should be a focused evidence plan that identifies the reference standard, study population, clinically defensible endpoints, and the basis for determining which prior results remain applicable. Maintaining these links as the product evolves gives reviewers a clearer basis for assessing both the evidence and the claims it is intended to support.
References
[1] FDA. Classification of the Software-Aided Adjunctive Diagnostic Device for Use by Physicians on Lesions Suspicious for Skin Cancer. Final order. 30 October 2026. 21 CFR 878.1830. FR document 2026-22179. Official source
[2] FDA. Classification of the Radiology Software for Referral of Findings Related to Fibrotic Lung Disease. Final order. 30 October 2026. 21 CFR 892.2085. FR document 2026-22180. Official source
[3] FDA. Classification of the Over-the-Counter Device To Assess Risk of Sleep Apnea. Final order. 30 October 2026. 21 CFR 868.2378. FR document 2026-22182. Official source
[4] FDA. Essential Drug Delivery Performance Characteristics for Devices Intended To Deliver Drugs and Biological Products. Revised draft guidance availability notice. 30 October 2026. Docket FDA-2024-D-2560. FR document 2026-22203. Official source
[5] FDA. Essential Drug Delivery Performance Characteristics for Devices Intended to Deliver Drugs and Biological Products. October 2026. Revision 1. Draft guidance. Sections V and VII. Official source
[6] FDA. Electronic Submission Template for Investigational Device Exemption Applications. 28 October 2026. Draft guidance. GUI00007080. Docket FDA-2026-D-11188. Official source
