(165 days)
Not Found
Yes
The document explicitly states that the device includes a rhythm classification algorithm that leverages machine learning techniques, specifically mentioning a convolutional neural network based architecture.
No.
The device is intended to provide a retrospective estimate of AFib burden and does not provide individual irregular rhythm notifications or replace traditional methods of diagnosis, treatment, or therapy.
No
The text explicitly states, "It is not intended to provide individual irregular rhythm notifications or to replace traditional methods of diagnosis, treatment, or MFib," indicating that its purpose is not diagnostic.
Yes
The device is explicitly described as an "over-the-counter ("OTC") software-only mobile medical application" in both the Intended Use and Intended User sections. While it relies on data from the Apple Watch hardware, the device itself is the software component that analyzes this data.
Based on the provided information, this device is not an In Vitro Diagnostic (IVD).
Here's why:
- IVDs analyze samples taken from the human body. The definition of an IVD typically involves the examination of specimens such as blood, urine, tissue, etc., in vitro (outside the body).
- This device analyzes data collected from the body. The Atrial Fibrillation History Feature analyzes pulse rate data collected by the Apple Watch's PPG sensor on the user's wrist. This is a non-invasive method of data collection from the body, not the analysis of a sample taken from the body.
- The intended use is for monitoring and providing information to the user, not for diagnostic testing of a sample. The feature provides an estimate of AFib burden and tracks trends, but explicitly states it is "not intended to provide individual irregular rhythm notifications or to replace traditional methods of diagnosis, treatment, or MFib."
Therefore, while it is a medical device that uses data from the body, it does not fit the definition of an In Vitro Diagnostic.
No
The letter does not explicitly state that the FDA has reviewed, approved, or cleared a Predetermined Change Control Plan (PCCP) for this specific device.
Intended Use / Indications for Use
The Atrial Fibrillation (AFib) History Feature is an over-the-counter ("OTC") software-only mobile medical application intended for users 22 years of age and over who have a diagnosis of atrial fibrillation (AFib). The feature opportunistically analyzes pulse rate data to identify episodes of irregular heart rhythms suggestive of AFib and provides the user with a retrospective estimate of AFib burden (a measure of the spent in AFib during past Apple Watch wear).
The feature also tracks and trends estimated AFib burden over time, and includes lifestyle data visualizations to enable users to understand the impact of certain aspects of their AFib. It is not intended to provide individual irregular rhythm notifications or to replace traditional methods of diagnosis, treatment, or MFib.
The feature is intended for use with the Apple Watch and the Health app on iPhone.
Product codes
QDB
Device Description
The Atrial Fibrillation History Feature (AFib History Feature) is comprised of a pair of mobile medical apps - one on Apple Watch and the other on the iPhone.
The AFib History Feature is intended to analyze pulse rate data collected by the Apple Watch PPG sensor on Apple Watch Series 4. Series 5. and SE to identify episodes of irregular heart rhythms consistent with AFib and provides the user with a retrospective estimate of AFib burden (a measure of the amount of time spent in AFib during past Apple Watch wear).
The AFib History Feature uses PPG pulse rhythm data from compatible Apple Watches. Apple Watch uses green LED lights paired with light-sensitive photodiodes to detect relative changes in the amount of blood flowing through a user's wrist at any given moment. When the heart beats it sends a pressure wave down the vasculature, causing a momentary increase in blood volume when it passes by the sensor. By monitoring these changes in blood flow, the sensor detects individual pulses when they reach the peripherv and thereby measure beat-to-beat intervals.
The AFib History Feature iPhone App is part of the Health App. which allows users to store, manage, and share health and fitness data, and comes pre-installed on every iPhone.
The AFib History Feature provides users visualizations of AFib burden estimate data alongside clinically relevant lifestyle data and presents estimates of AFib burden in three different ways. These visualizations empower users to observe and understand the impact of lifestyle on their AFib burden, and to better understand their condition generally.
- Weekly Estimate an estimate of the amount of time a user was in Atrial Fibrillation over the past calendar week during watch wear, presented to the user as a percentage.
- Day of Week Estimate an estimate of the amount of time a user was in Atrial Fibrillation on each day of the week over the previous 42 days during watch wear, presented to the user as a percentage. That is, all Mondays over the past 42 days, all Tuesdays over the past 42 days.
- Time of Day Estimate an estimate of the amount of time a user was in Atrial Fibrillation on 4-hour segments of the day over the previous 42 days during watch wear, presented to the user as a percentage. That is, all 12 am - 4 am segments over the past 42 days, all 4 am - 8 am segments over the past 42 days.
The AFib History Feature is intended to serve as an extension of the predicate Irregular Rhythm Notification feature, but has been optimized for users with a diagnosis of Afib.
Mentions image processing
Not Found
Mentions AI, DNN, or ML
The AFib History Feature includes a rhythm classification algorithm that leverages machine learning techniques to differentiate between AFib and non-AFib rhythms. The classifier algorithm is the same that is used in the predicate device, but has been optimized for use in the AFib History Feature's indicated use population, where there is an a priori expectation of AFib.
The rhythm classification algorithm uses a convolutional neural network based architecture and was trained extensively using data collected in a number of development studies.
The AFib History Feature and the Irregular rhythm Notification Feature (IRNF 2.0) use the same tachogram classification algorithm that leverages machine learning techniques to differentiate between AFib and non-AFib rhythms. The classification algorithm analyzes pulse rhythm samples collected by Apple Watch and uses a convolutional network based architecture.
Input Imaging Modality
Photoplethysmograph (PPG) pulse rhythm data
Anatomical Site
wrist
Indicated Patient Age Range
users 22 years of age and over
Intended User / Care Setting
over-the-counter ("OTC") software-only mobile medical application
Description of the training set, sample size, data source, and annotation protocol
The rhythm classification algorithm uses a convolutional neural network based architecture and was trained extensively using data collected in a number of development studies. In total, the studies included over 2500 subjects and collected over 3 million pulse rate recordings on a variety of rhythms including: atrial fibrillation, normal sinus rhythm, sinus arrhythmia, and other ectopic beats (PVCs, PACs).
The studies used to train the convolutional network recruited demographically diverse populations with broad representation of age, sex, BMI, race, and skin tones.
For the purpose of developing the algorithm, the data was split into four sets with matching distributions of rhythms and demographics: Training, Validation, Test, and Sequestration sets. The model was trained on the Training set, with the Validation set used for early stopping and threshold selection.
Description of the test set, sample size, data source, and annotation protocol
For the purpose of developing the algorithm, the data was split into four sets with matching distributions of rhythms and demographics: Training, Validation, Test, and Sequestration sets. The model was trained on the Training set, with the Validation set used for early stopping and threshold selection. The model was then evaluated on the Testing set at regular intervals during model development. When development was complete the model was locked, and then evaluated on the Sequestration set as a last test to ensure it had not been over-fit to the development data.
Summary of Performance Studies (study type, sample size, AUC, MRMC, standalone performance, key results)
Development Studies Performance:
Study Type: Rhythm Classification Algorithm Development Studies
Sample Size: Over 2500 subjects
Key Metrics:
AFib History Feature: Sensitivity 97%, Specificity 99.0%
IRNF 2.0 (predicate): Sensitivity 79.6%, Specificity 99.9%
Key Results: The classifier algorithm's performance on the development studies dataset when used in both the AFib History Feature & the Irregular Rhythm Notification Feature is characterized by the receiver operating characteristic (ROC) curve & table 3.
Clinical Performance Testing:
Study Type: Clinical Study
Sample Size: 413 participants
Data Source: Enrolled subjects wore an Apple Watch and a reference electrocardiogram (ECG) patch concurrently for up to 13 days.
Objective: To assess the accuracy of the weekly AFib burden estimate generated by the feature compared to a weekly AFib burden reference measurement through a Bland-Altman Limits of Agreement (LoA) approach.
Key Results:
Of the 413 enrolled subjects, 280 contributed data to the primary endpoint analysis.
Lower and upper Bland-Altman limits were -11.4% and 12.8%, respectively.
The average difference between the feature's weekly burden estimate and reference weekly burden was 0.67%.
92.9% (260/280) of subjects had paired weekly AFib burden differences within ±5%.
95.7% (268/280) of subjects' weekly AFib burden estimates were within +/- 10%.
The classification algorithm’s performance in the clinical validation study:
AFib History Feature: Sensitivity 92.6%, Specificity 98.8%
IRNF 2.0 (predicate): Sensitivity 85.5%, Specificity 99.6%
These results demonstrate that the AFib History Feature is effective in generating accurate AFib burden estimates.
Key Metrics (Sensitivity, Specificity, PPV, NPV, etc.)
Rhythm Classification Algorithm - Development Studies Performance:
AFib History Feature: Sensitivity 97%, Specificity 99.0%
IRNF 2.0 (predicate): Sensitivity 79.6%, Specificity 99.9%
Classification Algorithm - Clinical Validation Study Performance:
AFib History Feature: Sensitivity 92.6%, Specificity 98.8%
IRNF 2.0 (predicate): Sensitivity 85.5%, Specificity 99.6%
Clinical Performance:
The average difference between the feature's weekly burden estimate and reference weekly burden was 0.67%.
92.9% (260/280) of subjects had paired weekly AFib burden differences within ±5%.
95.7% (268/280) of subjects' weekly AFib burden estimates were within +/- 10%.
Predicate Device(s)
Reference Device(s)
Not Found
Predetermined Change Control Plan (PCCP) - All Relevant Information
Not Found
§ 870.2790 Photoplethysmograph analysis software for over-the-counter use.
(a)
Identification. A photoplethysmograph analysis software device for over-the-counter use analyzes photoplethysmograph data and provides information for identifying irregular heart rhythms. This device is not intended to provide a diagnosis.(b)
Classification. Class II (special controls). The special controls for this device are:(1) Clinical performance testing must demonstrate the performance characteristics of the detection algorithm under anticipated conditions of use.
(2) Software verification, validation, and hazard analysis must be performed. Documentation must include a characterization of the technical specifications of the software, including the detection algorithm and its inputs and outputs.
(3) Non-clinical performance testing must demonstrate the ability of the device to detect adequate photoplethysmograph signal quality.
(4) Human factors and usability testing must demonstrate the following:
(i) The user can correctly use the device based solely on reading the device labeling; and
(ii) The user can correctly interpret the device output and understand when to seek medical care.
(5) Labeling must include:
(i) Hardware platform and operating system requirements;
(ii) Situations in which the device may not operate at an expected performance level;
(iii) A summary of the clinical performance testing conducted with the device;
(iv) A description of what the device measures and outputs to the user; and
(v) Guidance on interpretation of any results.
0
Image /page/0/Picture/0 description: The image contains the logo of the U.S. Food & Drug Administration (FDA). On the left is the Department of Health & Human Services logo. To the right of that is the FDA logo, which is a blue square with the letters "FDA" in white. To the right of the blue square is the text "U.S. FOOD & DRUG ADMINISTRATION" in blue.
June 3, 2022
Apple Inc. Luke Olson Regulatory Affairs 1 Apple Park Way Cupertino, California 95014
Re: K213971
Trade/Device Name: Atrial Fibrillation History Feature Regulation Number: 21 CFR 870.2790 Regulation Name: Photoplethysmograph analysis software for over-the-counter use Regulatory Class: Class II Product Code: QDB Dated: Mav 2, 2022 Received: May 3, 2022
Dear Luke Olson:
We have reviewed your Section 510(k) premarket notification of intent to market the device referenced above and have determined the device is substantially equivalent (for the indications for use stated in the enclosure) to legally marketed predicate devices marketed in interstate commerce prior to May 28, 1976, the enactment date of the Medical Device Amendments, or to devices that have been reclassified in accordance with the provisions of the Federal Food, Drug, and Cosmetic Act (Act) that do not require approval of a premarket approval application (PMA). You may, therefore, market the device, subject to the general controls provisions of the Act. Although this letter refers to your product as a device, please be aware that some cleared products may instead be combination products. The 510(k) Premarket Notification Database located at https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm identifies combination product submissions. The general controls provisions of the Act include requirements for annual registration, listing of devices, good manufacturing practice, labeling, and prohibitions against misbranding and adulteration. Please note: CDRH does not evaluate information related to contract liability warranties. We remind you, however, that device labeling must be truthful and not misleading.
If your device is classified (see above) into either class II (Special Controls) or class III (PMA), it may be subject to additional controls. Existing major regulations affecting your device can be found in the Code of Federal Regulations, Title 21, Parts 800 to 898. In addition, FDA may publish further announcements concerning your device in the Federal Register.
Please be advised that FDA's issuance of a substantial equivalence determination does not mean that FDA has made a determination that your device complies with other requirements of the Act or any Federal statutes and regulations administered by other Federal agencies. You must comply with all the Act's
1
requirements, including, but not limited to: registration and listing (21 CFR Part 807); labeling (21 CFR Part 801); medical device reporting of medical device-related adverse events) (21 CFR 803) for devices or postmarketing safety reporting (21 CFR 4, Subpart B) for combination products (see https://www.fda.gov/combination-products/guidance-regulatory-information/postmarketing-safety-reportingcombination-products); good manufacturing practice requirements as set forth in the quality systems (OS) regulation (21 CFR Part 820) for devices or current good manufacturing practices (21 CFR 4, Subpart A) for combination products; and, if applicable, the electronic product radiation control provisions (Sections 531-542 of the Act); 21 CFR 1000-1050.
Also, please note the regulation entitled, "Misbranding by reference to premarket notification" (21 CFR Part 807.97). For questions regarding the reporting of adverse events under the MDR regulation (21 CFR Part 803), please go to https://www.fda.gov/medical-device-safety/medical-device-reportingmdr-how-report-medical-device-problems.
For comprehensive regulatory information about mediation-emitting products, including information about labeling regulations, please see Device Advice (https://www.fda.gov/medicaldevices/device-advice-comprehensive-regulatory-assistance) and CDRH Learn (https://www.fda.gov/training-and-continuing-education/cdrh-learn). Additionally, you may contact the Division of Industry and Consumer Education (DICE) to ask a question about a specific regulatory topic. See the DICE website (https://www.fda.gov/medical-device-advice-comprehensive-regulatoryassistance/contact-us-division-industry-and-consumer-education-dice) for more information or contact DICE by email (DICE@fda.hhs.gov) or phone (1-800-638-2041 or 301-796-7100).
Sincerely,
Jennifer Shih Kozen Assistant Director Division of Cardiac Electrophysiology, Diagnostics and Monitoring Devices Office of Cardiovascular Devices Office of Product Evaluation and Quality Center for Devices and Radiological Health
Enclosure
2
Indications for Use
510(k) Number (if known) K213971
Device Name Atrial Fibrillation History Feature
Indications for Use (Describe)
The Atrial Fibrillation (AFib) History Feature is an over-the-counter ("OTC") software-only mobile medical application intended for users 22 years of age and over who have a diagnosis of atrial fibrillation (AFib). The feature opportunistically analyzes pulse rate data to identify episodes of irregular heart rhythms suggestive of AFib and provides the user with a retrospective estimate of AFib burden (a measure of the spent in AFib during past Apple Watch wear).
The feature also tracks and trends estimated AFib burden over time, and includes lifestyle data visualizations to enable users to understand the impact of certain aspects of their AFib. It is not intended to provide individual irregular rhythm notifications or to replace traditional methods of diagnosis, treatment, or MFib.
The feature is intended for use with the Apple Watch and the Health app on iPhone.
| Type of Use (Select one or both, as applicable) | |
|---|---|
| Prescription Use (Part 21 CFR 801 Subpart D) | Over-The-Counter Use (21 CFR 801 Subpart C) |
CONTINUE ON A SEPARATE PAGE IF NEEDED.
This section applies only to requirements of the Paperwork Reduction Act of 1995.
DO NOT SEND YOUR COMPLETED FORM TO THE PRA STAFF EMAIL ADDRESS BELOW.
The burden time for this collection of information is estimated to average 79 hours per response, including the time to review instructions, search existing data sources, gather and maintain the data needed and complete and review the collection of information. Send comments regarding this burden estimate or any other aspect of this information collection, including suggestions for reducing this burden, to:
Department of Health and Human Services Food and Drug Administration Office of Chief Information Officer Paperwork Reduction Act (PRA) Staff PRAStaff(@fda.hhs.gov
"An agency may not conduct or sponsor, and a person is not required to respond to, a collection of information unless it displays a currently valid OMB number."
3
510(k) Summary
This summary of 510(k) safety and effectiveness information is submitted in accordance with the requirements of 21 CFR §807.92:
5.1 Submitter
| Applicant | Apple Inc.
One Apple Park Way
Cupertino, CA 95014 |
|----------------------------|--------------------------------------------------------------------------------------------|
| Primary
Correspondent | Luke Olson
Regulatory Affairs
Phone: (408) 609-2001
Email: luke_olson@apple.com |
| Secondary
Correspondent | Dachan Kwon
Regulatory Affairs
Phone: (669) 268-5659
Email: dachan_kwon@apple.com |
| Date Prepared | May 28, 2022 |
5.2 Device Names and Classifications
Subject Device:
| Name of Device | Atrial Fibrillation History Feature |
|---|---|
| Classification Name | Photoplethysmograph Analysis Software For Over-The-Counter Use, |
| 21 CFR 870.2790 | |
| Regulatory Class | Class II |
| Product Code | QDB |
| 510(k) Review | |
| Pannel | Cardiovascular |
Predicate Device:
| Predicate
| Manufacturer | Apple Inc. |
|---|---|
| Predicate Trade | |
| Name | Irregular Rhythm Notification Feature |
| Predicate 510(k) | K212516 |
4
5.3 Device Description
The Atrial Fibrillation History Feature (AFib History Feature) is comprised of a pair of mobile medical apps - one on Apple Watch and the other on the iPhone.
The AFib History Feature is intended to analyze pulse rate data collected by the Apple Watch PPG sensor on Apple Watch Series 4. Series 5. and SE to identify episodes of irregular heart rhythms consistent with AFib and provides the user with a retrospective estimate of AFib burden (a measure of the amount of time spent in AFib during past Apple Watch wear).
The AFib History Feature uses PPG pulse rhythm data from compatible Apple Watches. Apple Watch uses green LED lights paired with light-sensitive photodiodes to detect relative changes in the amount of blood flowing through a user's wrist at any given moment. When the heart beats it sends a pressure wave down the vasculature, causing a momentary increase in blood volume when it passes by the sensor. By monitoring these changes in blood flow, the sensor detects individual pulses when they reach the peripherv and thereby measure beat-to-beat intervals.
The AFib History Feature iPhone App is part of the Health App. which allows users to store, manage, and share health and fitness data, and comes pre-installed on every iPhone.
The AFib History Feature provides users visualizations of AFib burden estimate data alongside clinically relevant lifestyle data and presents estimates of AFib burden in three different ways. These visualizations empower users to observe and understand the impact of lifestyle on their AFib burden, and to better understand their condition generally.
- · Weekly Estimate an estimate of the amount of time a user was in Atrial Fibrillation over the past calendar week during watch wear, presented to the user as a percentage.
- Day of Week Estimate an estimate of the amount of time a user was in Atrial Fibrillation on each day of the week over the previous 42 days during watch wear, presented to the user as a percentage. That is, all Mondays over the past 42 days, all Tuesdays over the past 42 days.
- Time of Day Estimate an estimate of the amount of time a user was in Atrial Fibrillation on 4-hour segments of the day over the previous 42 days during watch wear, presented to the user as a percentage. That is, all 12 am - 4 am segments over the past 42 days, all 4 am - 8 am segments over the past 42 days.
The AFib History Feature is intended to serve as an extension of the predicate Irregular Rhythm Notification feature, but has been optimized for users with a diagnosis of Afib.
5
5.4 Indications for Use
The Atrial Fibrillation (AFib) History Feature is an over-the-counter ("OTC") software-only mobile medical application intended for users 22 years of age and over who have a diagnosis of atrial fibrillation (AFib). The feature opportunistically analyzes pulse rate data to identify episodes of irreqular heart rhythms suggestive of AFib and provides the user with a retrospective estimate of AFib burden (a measure of the amount of time spent in AFib during past Apple Watch wear).
The feature also tracks and trends estimated AFib burden over time, and includes lifestyle data visualizations to enable users to understand the impact of certain aspects of their lifestyle on their AFib. It is not intended to provide individual irregular rhythm notifications or to replace traditional methods of diagnosis, treatment, or monitoring of AFib.
The feature is intended for use with the Apple Watch and the Health app on iPhone.
5.5 Comparison with the Predicate Device
| Item | Subject Device | Predicate Device |
|---|---|---|
| AFib History Feature | IRNF 2.0 App | |
| Manufacturer | Apple Inc. | Apple Inc. |
| Submission | ||
| Reference | K213971 | K212516 |
| Intended Use | Photoplethysmograph analysis | |
| software for over-the-counter use. A | ||
| photoplethysmograph analysis | ||
| software device for over-the-counter | ||
| use analyzes photoplethysmograph | ||
| data and provides information for | ||
| identifying irregular heart rhythms. | ||
| This device is not intended to provide | ||
| a diagnosis. | Photoplethysmograph analysis | |
| software for over-the-counter use. A | ||
| photoplethysmograph analysis | ||
| software device for over-the-counter | ||
| use analyzes photoplethysmograph | ||
| data and provides information for | ||
| identifying irregular heart rhythms. | ||
| This device is not intended to provide | ||
| a diagnosis. | ||
| Subject Device | Predicate Device | |
| Item | AFib History Feature | IRNF 2.0 App |
| Indications for | ||
| Use | The Atrial Fibrillation (AFib) History | |
| Feature is an over-the-counter | ||
| ("OTC") software-only mobile medical | ||
| application intended for users 22 | ||
| years of age and over who have a | ||
| diagnosis of atrial fibrillation (AFib). | ||
| The feature opportunistically analyzes | ||
| pulse rate data to identify episodes of | ||
| irregular heart rhythms suggestive of | ||
| AFib and provides the user with a | ||
| retrospective estimate of AFib burden | ||
| (a measure of the amount of time | ||
| spent in AFib during past Apple Watch | ||
| wear). |
The feature also tracks and trends
estimated AFib burden over time, and
includes lifestyle data visualizations to
enable users to understand the
impact of certain aspects of their
lifestyle on their AFib. It is not
intended to provide individual irregular
rhythm notifications or to replace
traditional methods of diagnosis,
treatment, or monitoring of AFib.
The feature is intended for use with
the Apple Watch and the Health app
on iPhone. | The Irregular Rhythm Notification
Feature is a software-only mobile
medical application that is intended to
be used with the Apple Watch. The
feature analyzes pulse rate data to
identify episodes of irregular heart
rhythms suggestive of atrial fibrillation
(AFib) and provides a notification to
the user. The feature is intended for
over-the-counter (OTC) use. It is not
intended to provide a notification on
every episode of irregular rhythm
suggestive of AFib and the absence of
a notification is not intended to
indicate no disease process is
present; rather the feature is intended
to opportunistically surface a
notification of possible AFib when
sufficient data are available for
analysis. These data are only captured
when the user is still. Along with the
user's risk factors the feature can be
used to supplement the decision for
AFib screening. The feature is not
intended to replace traditional
methods of diagnosis or treatment.
The feature has not been tested for
and is not intended for use in people
under 22 years of age. It is also not
intended for use in individuals
previously diagnosed with AFib |
| Principle of
Operation | The AFib History Feature acquires
platform sensor data from Apple
Watch. After acquisition, the Afib
History Feature algorithms analyze
pulse rate data to identify episodes of
irregular heart rhythms suggestive of
atrial fibrillation (AFib) and aggregates
those episodes to provide the user
with an estimate of atrial fibrillation
burden during watch wear. | The IRN 2.0 app acquires platform
sensor data from Apple Watch. After
acquisition, the IRN app algorithms
analyze pulse rate data to identify
episodes of irregular heart rhythms
suggestive of atrial fibrillation (AFib)
and provides notification to the user. |
| | Subject Device | Predicate Device |
| Item | AFib History Feature | IRNF 2.0 App |
| Clinical
Performance | See below for a discussion of clinical
performance testing supporting the
AFib History Feature. | Apple conducted a clinical validation
study to assess the performance of
IRNF 2.0 app relative to that of the
IRNF 1.0 on a common sensor dataset.
IRNF 2.0 person-level sensitivity
(88.6%) and specificity (99.3%) were
both demonstrated to be non-inferior
to those of the IRNF 1.0. |
| Compatibility
with Intended
Platforms | iOS version 16.0 or later
watchOS version 9.0 or later | iOS version 15.5 or later
watchOS version 8.5 or later |
| | Apple Watch Series 4, 5, SE
iPhone 6s and later | Apple Watch Series 3, 4, 5, SE
iPhone 6s and later |
Table 1. AFib History Feature Comparison with the Predicate
6
7
5.6 Performance Testing
The AFib History Feature was verified and validated according to Apple's internal design control processes and in accordance with the special controls for Photoplethysmograph Analysis software for over-the-counter use (21 CFR 870.2790). The testing demonstrated that the device performed according to its specifications and that the technological and performance criteria are comparable to the predicate device.
The AFib History Feature includes a rhythm classification algorithm that leverages machine learning techniques to differentiate between AFib and non-AFib rhythms. The classifier algorithm is the same that is used in the predicate device, but has been optimized for use in the AFib History Feature's indicated use population, where there is an a priori expectation of AFib.
The rhythm classification algorithm uses a convolutional neural network based architecture and was trained extensively using data collected in a number of development studies. In total, the studies included over 2500 subjects and collected over 3 million pulse rate recordings on a variety of rhythms including: atrial fibrillation, normal sinus rhythm, sinus arrhythmia, and other ectopic beats (PVCs, PACs).
8
The studies used to train the convolutional network recruited demographically diverse populations with broad representation of age, sex, BMI, race, and skin tones. Table 2 below summarizes approximate development study demographic characteristics:
| Age Group (years) | |
|---|---|
| =55 to =65 | 35.1% |
| Sex | |
| Male | 49.6% |
| Female | 50.4% |
| BMI (kg/m²) | |
| =18.5 to =25.0 to =30.0 | 32.9% |
| Race | |
| White | 71.5% |
| Black or African American | 18.0% |
| Other | 10.5% |
Table 2. Development Study Subject Demographics
For the purpose of developing the algorithm, the data was split into four sets with matching distributions of rhythms and demographics: Training, Validation, Test, and Sequestration sets. The model was trained on the Training set, with the Validation set used for early stopping and threshold selection. The model was then evaluated on the Testing set at regular intervals during model development. When development was complete the model was locked, and then evaluated on the Sequestration set as a last test to ensure it had not been over-fit to the development data.
The classifier algorithm's performance on the development studies dataset when used in both the AFib History Feature & the Irreqular Rhythm Notification Feature is characterized by the receiver operating characteristic (ROC) curve & table 3 below.
9
Image /page/9/Figure/2 description: The image is a plot of sensitivity versus 100 - specificity. The y-axis is labeled "Sensitivity (%)" and ranges from 70 to 100. The x-axis is labeled "100 - Specificity (%)" and ranges from 0.0 to 2.0. The plot shows a green curve that starts at approximately (0.1, 79) and increases to approximately (1.1, 97). The plot also shows two points labeled "IRNF 2.0" and "AFib History".
Table 3. Rhythm Classification Algorithm - Development Studies Performance
| Sensitivity | Specificity | |
|---|---|---|
| AFib History Feature | 97% | 99.0% |
| IRNF 2.0 (predicate) | 79.6% | 99.9% |
10
5.7 Clinical Performance
The performance of the AFib History Feature was extensively tested in a clinical study of 413 participants ages 22 and older with a mix of AFib diagnoses (paroxysmal & permanent). Enrolled subjects wore an Apple Watch and a reference electrocardiogram (ECG) patch concurrently for up to 13 days. Study demographic characteristics are summarized in the table below:
| N=413 | |
|---|---|
| Age Group (years) | |
| =55 to =65 | 255 (61.7%) |
| Sex | |
| Male | 219 (53.0%) |
| Female | 194 (47.0%) |
| Ethnicity | |
| Hispanic or Latino | 19 (4.6%) |
| Non-Hispanic or Latino | 394 (95.4%) |
| Race | |
| White | 371 (89.8%) |
| Black or African American | 31 (7.5%) |
| Other | 11 (2.7%) |
Table 3. IRNF 2.0 Clinical Study Subject Demographics
The objective of the study was to assess the accuracy of the weekly AFib burden estimate generated by the feature compared to a weekly AFib burden reference measurement. To do so, Apple employed a Bland-Altman Limits of Agreement (LoA) approach. A LoA approach is a way of assessing agreement accuracy between two measurement methods.
11
The objective of the study was to assess the accuracy of the weekly AFib burden estimate generated by the feature compared to a weekly AFib burden reference measurement. To do so, Apple employed a Bland-Altman Limits of Agreement (LoA) approach. A LoA approach is a way of assessing agreement accuracy between two measurement methods.
Of the 413 enrolled subjects. 280 contributed data to the primary endpoint analysis to determine if the level of agreement between the reference ECG AFib burden and the feature's AFib burden estimate was acceptable. Based on the results of the study, the lower and upper Bland-Altman limits (i.e., two standard deviations from the mean difference) were -11.4% and 12.8%, respectively.
The average difference between the feature's weekly burden estimate and reference weekly burden was 0.67%. 92.9% (260/280) of subjects had paired weekly AFib burden differences within ±5%; 95.7% (268/280) of subjects' weekly AFib burden estimates were within +/- 10%.
The AFib History Feature and the Irregular rhythm Notification Feature (IRNF 2.0) use the same tachogram classification algorithm that leverages machine learning techniques to differentiate between AFib and non-AFib rhythms. The classification algorithm analyzes pulse rhythm samples collected by Apple Watch and uses a convolutional network based architecture. For use in the AFib History Feature the algorithm's operating point was adjusted to prioritize sensitivity. Table 4 below outlines the sensitivity and specificity of the classification algorithm for the AFib History Feature and IRNF 2.0 in the clinical validation study.
| Sensitivity | Specificity | |
|---|---|---|
| AFib History Feature | 92.6% | 98.8% |
| IRNF 2.0 (predicate) | 85.5% | 99.6% |
Table 4. Classification Algorithm - Clinical Validation Study Performance
These results demonstrate that the AFib History Feature is effective in generating accurate AFib burden estimates.
5.8 Human Factors Testing
The AFib History Feature was found to be safe and effective as compared to the predicate for the intended users, uses, and use environments. This conclusion is supported by iterative human factors analyses and evaluations on the Feature, resulting design modifications, and the ultimate analysis of the summative/validation testing results.
12
5.9 Conclusion
The AFib History Feature is substantially equivalent to IRNF 2.0 as they are identical with respect to intended use and there are no differences in technological or performance characteristics that raise new questions of safety and effectiveness.