The global prevalence of atrial fibrillation (AF) is over 60 million and increases steadily with age. This estimate is widely expected to increase and almost double by 2050, reaching over 100 million. AF is a major risk factor for stroke, accounting for up to 15-20% of all cases, and is a substantial economic burden on individuals, health systems and society.
Numerous studies have shown that increased screening leads to increased AF diagnosis (STROKESTOP, LOOP, mSToPS, SAFER), and continuous forms of screening lead to a higher detection rate than intermittent monitoring (implantable loop recorders > patch monitors > intermittent 12-lead ECGs). These trials showed that, while different methods of screening can increase the detection of AF in various populations, the impact of screening on stroke prevention and other clinical outcomes remains unclear. The effect of false positives and the cost of screening to the public health system are potential negatives. In fact, the USPSTF does not recommend screening in asymptomatic individuals as evidence is insufficient to assess the balance of benefits and harms.
Advances in wearable consumer technology have made it possible to monitor heart rate and rhythm on smart watches and other devices that can detect irregular pulses using optical sensors and algorithms. These devices can alert users to possible AF and encourage them to seek further medical attention. Many of these technologies and algorithms have even received FDA approval for ‘over-the-counter’ medical technology labeling. In some cases, physicians are already using these devices with their patients for screening purposes, and very often, patients are presenting to clinics with data from devices they have decided to use on their own. As a result, it is important to understand the data and technology behind these devices. In our last post we discussed the internal data behind the Apple Watch that was used for FDA approval, but there are many peer reviewed clinical trials covering multiple devices that are important to be aware of, which will be the subject in this post.
PPG vs ECG
The two popular technologies used for consumer AF monitoring are photoplethysmography (PPG) and single lead recordable ECGs. There is real world data available for both, with different strengths and weaknesses. In general, PPG signals can be continuously monitoring in the background whenever the watch is being worn, which should yield more frequent detection (similar to loop recorders), though in reality the algorithms used by these devices have opportunistic windows of screening – they are not constantly monitoring the pulse signal. All of the algorithms are slightly different in terms of monitoring frequency and what constitutes a positive result that will trigger an alert to the user.
There are four large clinical trials evaluating consumer PPG technology for AF screening in the general public: (i) The Apple Heart Study, (ii) the FitBit Heart Study, (iii) the Huawei Heart Study, and (iv) the WATCH AF Trial (Samsung Watch). The PPG detection algorithms are slightly different for each, which I encourage you to review (we presented the Apple details in our previous post). In general, the studies were large (>200,000 participants, with the exception of WATCH AF) and young (mean age of <60 yrs). The primary outcome reported for these trials was the positive predictive value (PPV), or the probability that a patient with a positive notification actually has the outcome of interest, i.e. AF. The PPV was highest for the FitBit algorithm (98%), though it was over 80% in all cases. Importantly, as noted above, these studies were completed in relatively young and healthy individuals. PPV increases as prevelance increases, so algorithm performance would be expected to be higher in older, less healthy populations with greater risk of AF. While PPV was high, sensitivity varied to as low as 67% in some studies (when reported), so it is very important to remember that while false positives were low, a lack of irregular rhythm notification does not rule out AF when there is high clinical suspicion and PPG should not be used as the only screening method for symptomatic individuals.
| Study | PPV |
| Apple Heart Study | 84% |
| FitBit Heart Study | 98% |
| Huawei Heart Study | 98% |
| WATCH AF Study | 92% |
Many smart watches and other consumer tech devices (KardiaMobile) also allow users to record an ECG signal (usually a single lead). This technology can be used for stand-alone, intermittent screening as users can record a 30 second ECG signal at random times throughout the day, or can be combined with PPG technology to increase the specificity and PPV of irregular rhythm notifications by confirming or ruling out AF as the culprit rhythm. There are numerous trials available in the medical literature evaluating this technology. By comparison, these trials have been smaller and completed in more clinically diverse populations.
The greatest level of clinical evidence for ‘over-the-counter’ ECG monitors belong to the KardiaMobile and Apple Watch platforms. In the studies available for review, the majority show accuracy and PPV levels of >90%, but some are as low as 61% (though this was completed in the hospital setting for patients recovering from cardiac surgery). Sensitivity and specificity levels also vary by study, but are generally over 70%. One large, clinically relevant study evaluated these two devices in a head-to-head comparison. The aptly named SMART WARS Study compared the Apple Watch to the Kardia Band in an outpatient, elderly (>65 years) population. In a monitored setting, patients generated back-to-back 30 second ECGs from the Apple Watch and the Kardia Band while simultaneously recording a 12-lead ECG to serve as the gold standard. Results showed the Apple Watch and Kardia Band generated diagnostic results in 70% and 80% of cases, respectively. The majority of the non-diagnostic results were due to devices labeling rhythms as unclassified, although cardiologists overreading the ECG strips were able to determine a diagnosis on all occasions (i.e. not due to severe artifact). The resultant diagnostic level accuracy and PPV of the two tech platforms was 93% and 100% (Apple) and 94% and 84% (KardiaMobile).
| SMART WARS Study | |||
| Apple Watch | Kardia Band | ||
| Accuracy: 93% | PPV: 100% | Accuracy: 94% | PPV: 84% |
Summary
Consumer facing medical technology is rapidly expanding and improving from year-to-year. It offers many promising benefits, including the ability to screen and identify asymptomatic AF in the general population. Early identification and treatment of AF has the potential to reduce stroke risk and lower the economic burden on the healthcare system. These potential benefits are currently unproven and it is not yet clear how to manage asymptomatic, device detected AF. There are also a number of important limitations to keep in mind: (i) there is no standardization among smart watch algorithms for AF detection which can affect reliability and accuracy, (ii) users with the biggest potential gain (>65 years) have been under-reprented in large trials and are less likely to use the technology, (iii) false positives (this rate appears low in the available evidence) can lead to over-investigation and increased anxiety in low risk populations, and (iv) there is no current mechanism in place for providers and hospitals to seek reimbursement for time spent assessing the large amount of data being generated by these devices.
Please subscribe to our Newsletter for email notifications containing new posts as soon as they are published:
Cardio Cypher 