FDA Review of Multiple Function Device Products: Policy Framework and 510(k) and De Novo Precedents

How FDA reviews multiple function device products: policy framework and 510(k) and De Novo precedents

Many devices now combine several measurement channels, software features, or hardware functions in one product. A bedside monitor can measure a dozen physiological parameters. A consumer watch can host a cleared notification feature next to hundreds of non-device apps. A PCR instrument can run cleared assays and research workflows. FDA uses two different approaches here. When every function needs premarket review, as with a multiparameter monitor or a glucose/ketone meter, each function must be supported in the submission. When a product mixes reviewed functions with "other functions" (non-device, 510(k)-exempt, or under enforcement discretion), FDA reviews only the function under review and looks at the other functions only for their impact on it. This article covers both approaches and the precedents that show how FDA has applied them.

Key takeaways

• FDA's controlling policy is the final guidance Multiple Function Device Products: Policy and Considerations (July 29, 2020) ⁵⁸². FDA reviews the device function(s) for which clearance or approval is sought, and its clearance or approval applies only to those reviewed functions ⁶²⁶.
• An "other function" is assessed only to the extent it could adversely affect the safety or effectiveness of the device function under review, or when the sponsor wants a "labeled positive impact" considered ⁶²⁶⁷.
• FDA applies these principles to all multiple function device products, including software, hardware, and mixed-function products. The statute itself addresses only software ²⁵⁶.
• Recent 510(k) and De Novo decision summaries for Apple, Samsung, ResMed, Centervue, Metrovision, and BD products include an explicit multiple-function statement ^{10911111512028927827999241}.
• When all functions need review, as with multiparameter patient monitors and multi-analyte meters, sponsors usually pair a primary predicate with parameter-specific or analyte-specific predicates or reference devices ^153169122.

Statutory basis: FD&C Act section 520(o)(2)

Section 520(o)(2) covers a product that contains at least one software function excluded under section 520(o)(1) (or otherwise outside the section 201(h) device definition) and at least one device function. For these products, FDA "shall not regulate" the non-device software function as a device. FDA may, however, assess how the non-device function affects the device function(s) ¹. FDA uses a general-purpose computing platform as its example: it does not regulate the platform as a device, but it may assess the platform's impact on a mobile medical application running on it ⁶.

FDA identifies the 2020 Multiple Function Device Products guidance as its interpretation of section 520(o)(2). The separate guidance Changes to Existing Medical Software Policies Resulting from Section 3060 of the 21st Century Cures Act (September 27, 2019) covers section 520(o)(1)(A) to (D) ⁸. FDA's final Clinical Decision Support Software guidance (January 29, 2026) says the same approach applies when a product contains both Non-Device CDS and device software functions ⁴.

Core definitions

The 2020 guidance sets out the vocabulary that later decision summaries use ⁵²:

The guidance does not decide which functions are devices or which fall under enforcement discretion ⁵. Sponsors still have to make those classification calls using the relevant software and wellness policies.

Scope of FDA review

FDA does not review non-device functions just because they are packaged with a device function. It also does not review enforcement-discretion or 510(k)-exempt device functions for that reason alone ⁶⁵. The guidance gives this example: if clearance is sought for an analysis function but not for a trend function, FDA reviews the analysis function and looks at the trend function only to the extent it could harm analysis. The clearance covers only the analysis function ⁶².

FDA also says its public decision summaries will state that a product may contain functions not subject to premarket review, and that FDA assessed those functions only for adverse impact or as a labeled positive impact ⁶³¹³. That commitment explains the standard wording in the precedents discussed below.

The impact assessment

Two-step inquiry

FDA frames the assessment as two questions ⁵⁸:

1. Does the other function affect the safety or effectiveness of the device function-under-review?
2. If it does, could that effect increase risk or degrade performance (a negative impact)?

Manufacturers should look at relationships such as shared computational resources, data dependencies, shared code, shared memory or storage, a shared user interface or output screen, effects on processing time or memory, shared programming pointers, and whether the device function needs privileges to avoid delays or interruptions. A relationship alone does not prove an impact ⁵⁹⁶⁰. Examples FDA flags include an other function that supplies input to a critical calculation or writes configuration data to shared storage ⁵⁹.

Negative impact

Increased risk can arise when the other function creates a new hazardous situation or a new cause of an existing one, increases the severity of harm, serves as a risk control, or affects a risk control for the device function ⁶⁴. On the effectiveness side, FDA focuses on performance, including responsiveness, usability, and efficiency. Where an adverse effect is possible, verification and validation should characterize it ⁶⁴. One example of a negative impact is an other function that slows computation time ⁶².

Positive impact and labeled positive impact

A positive impact is a benefit to the device function when the other function works as intended, with the condition that failure of the other function does not harm the device function. Improved processing speed is FDA's example ⁶⁷⁶². FDA generally does not want every positive impact documented in the submission. If the sponsor wants a positive impact reflected in labeling (a "labeled positive impact"), the supporting documentation must be submitted ⁶²⁶⁷.

Cybersecurity

The cybersecurity assessment should assume that other functions could be used, maliciously or unintentionally, to affect the device function. For example, an attack on an other function could make the device function unavailable ⁶⁴. FDA notes that modular or separated architectures may be needed so that a threat to or from an other function does not reach the device function ²³.

Quality system documentation versus submission content

All impact assessments (no impact, negative, and positive) must be done and documented under the quality system as part of design validation. FDA may review them during inspection ⁶²⁶⁷¹⁴. What goes into the premarket submission depends on the result ¹⁴⁶⁷:

Recommended submission content

The Content of Premarket Submissions for Device Software Functions guidance (June 14, 2023) builds these recommendations into software documentation. The architecture diagram should clearly separate device functions under review from other functions ⁵². The software risk assessment should address adverse effects or labeled positive effects of other functions ⁵³. Requirements and the Software Design Specification should describe the relationships with other functions ⁴⁷⁵². The Off-The-Shelf Software Use in Medical Devices guidance (August 11, 2023) also points sponsors of multiple function products to the 2020 policy ⁴⁸.

Changes to other functions

A change to an other function should be assessed for whether it could significantly affect the safety or effectiveness of the reviewed device function. Depending on that assessment, a new premarket submission may be needed ¹³.

Adjacent frameworks

Molecular diagnostic instruments with combined functions (final, November 12, 2014). This guidance came before the 2020 policy and follows the same logic for instruments that run cleared assays and research-use functions. FDA reviews non-cleared functions only to assess the risks they pose to cleared functions. It does not evaluate their performance or how they are marketed ¹⁷⁸. Submissions should show non-interference, include a hazard analysis of the functions coexisting, and describe design controls ¹⁸²¹⁸⁰. For new instruments, software should clearly separate cleared and other functions. FDA suggests a dual-boot design as one method ¹⁷⁸. Cleared functions need separate labeling, and reports for non-cleared use may carry a "Not approved/cleared by FDA" statement ¹⁸²¹⁸⁰.

Bundling multiple devices or indications (final, June 22, 2007). When every function or indication needs review, the bundling guidance governs whether they can share one submission. Bundling works when the supporting data are similar, mainly one review division is involved, and the devices or indications are similar ²⁰⁵. Multiple related indications often fit in one 510(k). A new, uncleared use, or indications spanning specialties reviewed by different divisions, should be submitted separately ²⁰⁸. Each device or indication in a bundle must independently meet premarket requirements ²⁰⁶.

510(k) and De Novo precedents

Device software functions on general-purpose consumer platforms

The DPCF summary shows that the impact assessment can be backed by bench testing against the reviewed function's performance specification. The HTNF and SANF summaries show that a sponsor with an established platform can rely on a consistent, repeated platform impact assessment across features ^111289120.

Devices with 510(k)-exempt or supporting other functions

• K243504, Centervue MAIA (Mar. 17, 2025). The reviewed function was confocal retinal imaging. Automatic perimetry (product code HPT, 510(k)-exempt) was treated as an other function and assessed only for adverse impact or labeled positive impact ²⁷⁹. This is a direct hardware example of the "exempt device function" category.
• K253553, ResMed Orion (June 4, 2026). The reviewed function was remote configuration of therapy settings for compatible PAP devices. Therapy-data display, patient-record management, reporting, user-access controls, and interoperability features were other functions that do not adversely affect the reviewed function ²⁷⁸.
• K212936, Metrovision MonCvONE (Nov. 21, 2022) and K201814, BD FACSLyric flow cytometer system (Sept. 28, 2020) both carry the standard statement that the product has functions subject to and not subject to premarket review, and that FDA assessed the latter only for adverse impact (and, in K212936, labeled positive impact) ⁹⁹²⁴¹.

IVD instruments with cleared and research-use modes

• K123955, Life Technologies QuantStudio Dx (Mar. 8, 2013). IVD and RUO modes are mutually exclusive and cannot run at the same time. IVD use requires login to controlled IVD software and certified test definitions. Internal verification showed that co-installing RUO software did not affect IVD functionality ²⁶⁴.
• K091409, Roche LightCycler MRSA Advanced Test on LightCycler 2.0 (July 6, 2010). FDA applied the combination RUO/IVD instrumentation policy. The sponsor provided software documentation and an IVD-only user manual ²⁶⁵.

Multiparameter monitors: predicate strategy when every function is reviewed

In multiparameter patient monitors, all measurement functions are reviewed. The regulatory question is how to support each parameter.

• K250854, Comen N-series monitors (Dec. 17, 2025) used BeneVision N Series (K213799) as the predicate. Reference devices supported specialized functions: Masimo Radical-7 and Root for pulse CO-oximetry, gas/CO2 analyzers, BISx, a nerve stimulator, and cardiac output devices ¹⁵³. It is a clear example of a primary monitor predicate supplemented with parameter-specific references.
• K180903, Iradimed 3880 MRI Patient Monitoring System (Mar. 29, 2019) added neonatal indications for SpO2, capnography, and anesthetic-agent monitoring. It relied on the prior 3880 clearance plus the 3160 MRI monitor and Masimo Rad-8 and Root as comparators ¹⁶⁹.
• K200015, Mindray ePM Series (June 8, 2020) used the prior ePM clearance (K191769) as primary predicate and a Spacelabs multi-parameter module (K103142) as a reference device ¹⁵⁸.
• K233354, Athena GTX WVSM Pro (June 26, 2024) claimed a subset of the ZOLL Propaq M (K202375) indications. It omitted invasive pressure, automated 12-lead interpretation, and a TBI dashboard ¹⁷⁵¹⁷⁶.
• K223498, Masimo Radius VSM (June 1, 2023) reported function-specific NIBP performance to ISO 81060-2 (mean difference of 5 mmHg or less, standard deviation of 8 mmHg or less) ¹⁶⁴.

Multi-analyte meters: analyte-by-analyte support

• K130094, Abbott FreeStyle Precision Pro (June 27, 2013). The β-ketone strip was already cleared on the meter, so ketone performance relied on K080960. New glucose performance data, including precision, were submitted because the glucose strip had not been cleared on that meter ¹²⁵¹⁴⁵.
• K122688, Nova Max Mini (Apr. 4, 2013) used separate predicates for each analyte's strips: glucose strips identical to those in K112638 and ketone strips identical to those in K091547 ¹²⁷¹⁴².
• K201037, Taidoc FORA ADVANCED GD40 Glucose, β-Ketone and Cholesterol (June 16, 2022) carried glucose and ketone forward from K161738 and added cholesterol with its own predicate (K163406) ^122124126139.
• K173345, Taidoc TD-4140 (May 4, 2018) added β-ketone to a previously cleared glucose system (K162382) and presented ketone-specific repeatability and intermediate precision testing ¹²¹.

Practical implications for submission planning

1. Define the function under review precisely. Keep the indications for use limited to the reviewed function unless you intend to claim a labeled positive impact ⁶⁷. The precedents show FDA's clearance statement follows this scope ⁶²¹⁰⁹.
2. Classify every other function first. Decide whether each one is non-device, exempt, or under enforcement discretion before filing, because the 2020 guidance does not make that call for you ⁵. The Movano and Centervue summaries show how sponsors state the basis for exclusion ⁸⁹²⁷⁹.
3. Document the impact assessment in the design history file even when it is not submitted. FDA can request it during inspection ⁶²¹⁴.
4. Build separation into the architecture. Logical separation, code/data partitioning, and mutually exclusive modes simplify review. The QuantStudio Dx precedent shows the approach in an IVD instrument ²³²⁶⁴.
5. Tie platform-impact claims to the device function's own performance specification. Bench evidence that the reviewed function still meets its tolerances with other functions active, as in the DPCF summary, is stronger than a bare risk conclusion ¹¹¹.
6. When every function needs review, map predicates to functions. Use a primary predicate for the overall device type and parameter- or analyte-specific predicates or reference devices where the primary predicate does not cover a function ¹⁵³¹²². Consider the 2007 bundling criteria before combining disparate indications ²⁰⁵²⁰⁸.
7. Plan for change control on other functions. Changes to other functions still need an impact assessment and may trigger a new submission ¹³.