Smartphone-based screening for atrial fibrillation: Findings from the eBRAVE-AF study

Approximately 20-30% of all strokes are associated with atrial fibrillation (AF) (Grond et al., 2013; Kishore et al., 2014). However, because AF often occurs sporadically and without any symptoms, it can be difficult to detect (Jaakkola et al., 2016). However, early identification of AF is extremely important as this is the only way to initiate therapy and reduce the risk of stroke. In this context, the relevant medical guidelines recommend - in the event of a positive test result as part of the AF diagnosis - an assessment of the individual risk of stroke using the so-called CHA2DS2-VASc risk score, which is determined by awarding points for fulfilling certain risk factors (e.g. Reaching defined age thresholds or the presence of other risk-increasing diseases) is determined (Hindricks et al., 2021). Below a score of 2, simply adjusting your lifestyle may be sufficient, while with a score ≥ 2, so-called oral anticoagulation (OAC), ie administration of medication to inhibit blood clotting (also known as “blood thinners”), is regularly indicated. In the case of an indicated OAC, this can reduce the risk of stroke by around 70% (Hart et al., 2007; López-López et al., 2017) - which is also economically important considering stroke-associated direct medical costs of an average of around 43,200 euros (Birkemeyer et al ., 2020).

Smartphone-based screening
vs. standard diagnostics
>> Conventional AF diagnostics include measuring cardiac activity using electrodes and displaying them in curves (so-called electrocardiogram (ECG)). A functioning heart produces a regular image on the ECG, while the corresponding image in the case of an existing cardiac arrhythmia is visibly irregular. However, regular ECG measurements have limited sensitivity in relation to AF, as the irregularities described often do not occur permanently. Even a long-term ECG, where the recording period is usually 24 hours,1 is in many cases too short to detect an AF episode. Relevant studies have shown that the AF detection rate only comes close to an optimum when the measurement period is extended to up to 14 days (Turakhia et al., 2013; Quer et al., 2020). In addition, common ECG measurement methods are too unwieldy and time-consuming for large-scale screening and are therefore not suitable for identifying asymptomatic cases on a large scale. Rather, more manageable, cost-effective and at the same time reliable screening methods are required so that the cardiac arrhythmia in such cases does not only become apparent as a result of a stroke (Jaakkola et al., 2016). One such option is the two-stage AF screening from Preventicus GmbH. The central components - as shown in Figure 1 - are a clinically validated screening app (Preventicus Heartbeats), an affiliated telemedicine center (Ulm Heart Clinic) and a 14-day ECG , which is recorded with an external ECG loop recorder (CardioMem CM 100XT, GETEMED) in the event of an abnormal app measurement.
Preventicus Heartbeats is a certified Class IIa medical device that can distinguish between normal and irregular heart rhythms with a reported sensitivity and specificity of 89.9% and 99.1% (based on a one-minute measurement) (Brasier et al., 2019). To do this, the user positions the smartphone camera on any fingertip. Using the LED flash and the smartphone camera, Preventicus Heartbeats - based on the photoplethysmography (PPG) process, which commercially available pulse oximeters also use - records the heart rhythm, analyzes it and summarizes the results in an ECG-comparable report. All you need to use it is a standard smartphone. The purchase of any accessories is not necessary, which increases the suitability for large-scale and regular screening.
Any abnormal results are automatically sent to the connected telemedicine center for confirmation. If the abnormalities are confirmed, a 14-day long-term ECG is recorded to confirm the diagnosis, which goes far beyond the usual 24-hour recording in standard care. The external ECG loop recorder used for this purpose can automatically detect, record and save various cardiac arrhythmias. This so-called event triggering is a necessary condition for a 14-day ECG, since the automated pre-selection of potentially relevant measurement phases makes it possible to analyze such a long measurement period - despite limited time resources in cardiological care.
In statutory health insurance (GKV), the program is offered as part of the RhythmLife contract in accordance with Section 140a SGB V (Brandt et al., 2021). After the smartphone-based screening, the external ECG loop recorder is used in a telemedicine version. This means that all events are transmitted to a telemedicine center for technical preliminary assessment. Participating cardiologists can access all data via a special web application and use it to make a diagnosis.
The eBRAVE-AF study
In the eBRAVE-AF study (eHealth-based Bavarian Alternative Detection of Atrial Fibrillation), the effectiveness of the described smartphone-based AF screening program compared to regular care was scientifically examined for the first time (Freyer et al., 2021; Rizas et al., 2022). The following study report provides information about the central key points and findings of this randomized controlled study (RCT).

Design and implementation
The primary research interest was the question of whether AF can be detected in a significantly higher number of “hidden” cases by the screening program shown in Figure 1 than in standard care. The focus here was on high-risk patients who are eligible for subsequent OAC due to an increased CHA2DS2-VASc score.
This resulted in the following inclusion criteria:
• Age between 50 and 90 years,
• CHA2DS2-VASc score ≥ 1 (men)
or ≥ 2 (women),
• no existing VHF and
• no ongoing OAK.

The study participants were recruited across Germany in 2020 from among insured people from the cooperating Bavarian Insurance Chamber. The number of people in the target group who were invited, participated and dropped out during the course of the study is shown in the flow chart shown in Figure 2. A total of 67,488 insured people - based on the case data from the Bavarian Insurance Chamber - met the inclusion criteria and were invited to take part in the study. Of these, 5,587 (8.3%) downloaded the specially developed study coordination app, filled out the initial questionnaire stored there and electronically declared their consent to participate in the study. The majority (5,551 people) continued to meet the defined inclusion criteria - even taking into account the information provided in the initial questionnaire - and were included in the study. Ultimately, 4,453 participants completed the study completely, while 100 participants were diagnosed with AF with indicated OAC during the course of the study, i.e. H. the primary endpoint was reached and 633 participants dropped out for other reasons (e.g. withdrawal of informed consent or death).
The study concept envisaged a parallel group comparison (intervention group = smartphone-based screening (Preventicus) vs. control group = standard care), whereby the standard
care was modeled in such a way that control group
Penpatients – basically left to their own devices – did not receive any diagnostic services within the study. Group allocation was randomized and unblinded because the nature of the intervention was not amenable to blinding. After six months in the control and intervention groups, a six-month cross-over phase followed, i.e. i.e., the group assignment was swapped. In this way, the innovative screening program did not have to be withheld from any of the study participants. In addition, the cross-over design enabled both inter- and intragroup comparisons, which - in addition to randomization - secured the study results against possible biases resulting from structural differences between the comparison groups. If AF was diagnosed during the study, participants were referred to their local treating physician. The respective doctors were not involved in the study and made all subsequent treatment decisions, such as the initiation of OAC, at their own discretion. The research data was collected during the study using app-based questionnaires and telephone calls and was supplemented at the end of the study by case data from the Bavarian Insurance Chamber (diagnoses made according to ICD-10-CM and medications dispensed according to ATC).
The effect of screening on the primary endpoint was examined using a logistic regression analysis to estimate the influence of an independent variable (here: type of AF screening) on the time until the occurrence of a dependent event (here: detection of AF with indicated OAC). . The effect size of the smartphone-based screening on the primary endpoint was quantified using the so-called odds ratio (OR). The OR here represents the calculated rate for the occurrence of the event “detection of AF with indicated OAC” in the ratio of the intervention group to the control group. It can take values between 0 and ∞, with values > 1 representing a greater chance of detection in the intervention group.
The effects of screening on AF diagnoses in general (i.e. also those without indicated OAC), OAC prescriptions, strokes, thromboembolism and major bleeding were examined as secondary endpoints. The analyzes were carried out separately for both study phases and according to the intention-to-treat principle, with patients with known AF or ongoing OAC being excluded. In all analyses, differences were considered statistically significant if the two-tailed P value was less than 0.05.

Results and interpretation
The number of participants achieved is shown in the flowchart shown in Figure 2. Of the 5,551 individuals included, 31% were female. The median age was 65 years and the median CHA2DS2-VASc score was 3. The acceptance of the smartphone-based screening program was high, which was reflected in comparatively good adherence and a low dropout rate:
• Over 50% of the participants in the intervention group carried out at least one smartphone measurement per week over the entire six months in phase 1.
• 618 participants (11.1%) dropped out of the study prematurely, of which only 421 (7.6%) while they were in the intervention group.
These are encouraging values for a digital study without direct contact with subjects. In this context, it should be emphasized that the tested care innovation was also well received by the older participants, with the most PPG measurements tending to be carried out in the age group of 65–71 (P < 0.05). There were also no differences between older and younger participants in terms of the quality of the measurements or their correct implementation. There were no significant differences between the groups with regard to the characteristics “age”, “gender”, “BMI” or with regard to various existing illnesses and treatment references. A distorting influence on the analysis results caused by the potential disruptive factors mentioned could therefore be largely ruled out. Cumulative event rates for the primary endpoint are shown in Figure 3.
Within the first 6 months (Phase 1), the primary endpoint was reached by 38 participants (1.33% out of 2,860) who were assigned to the smartphone-based screening, while in the regular care control group only 17 participants (0.63% out of 2,691) were reached. met the primary endpoint. The corresponding OR for the primary endpoint was 2.12 (95% CI: 1.19-3.76) and was significant (P = 0.010) (see Figure 3a). Of the remaining participants from phase 1, 4,752 accepted the invitation to take part in the subsequent six-month phase 2, which involved a change in group membership (cross-over). Here too, smartphone-based screening demonstrated its superiority over standard care. Treatment-relevant AF was diagnosed in 33 participants in the intervention group (1.38% out of 2,387), while standard care only led to 12 corresponding diagnoses (0.51% out of 2,365). The OR in the second phase was 2.75 (95% CI: 1.42–5.34) and was also significant (P = 0.003) (see Fig. 3b).
In this context, it should be taken into account that the ORs are not comparable between the phases. The Phase 2 participants differ in one key feature in that they are already characterized by a negative screening result from Phase 1. Overall, the chance of detecting an AF requiring treatment but previously hidden was more than twice as great in both study phases if participants took part in the smartphone-based screening program.
Regarding the secondary endpoints, there were no significant differences between the groups in clinical endpoints such as the occurrence of stroke or relevant bleeding. However, in both phases of digital screening, AF was diagnosed more often and oral blood thinners were prescribed more often (each significant, i.e. P < 0.05).
There are related studies on digital AFib screening using smartwatches or comparable tracking devices that are worn on the wrist - e.g. B. the Apple, Fitbit or Huawei heart study (Guo et al., 2019; Perez et al., 2019; Lubitz et al., 2021). However, the eBRAVE-AF study is the first of its kind in which the screening program examined was initially purely smartphone-based, i.e. H. without special hardware and in which the effectiveness of the screening program was systematically compared with standard care. For the first time, it was demonstrated in an RCT, which is considered the gold standard of clinical study designs, that the detection rate of AF requiring treatment can be increased two to three times by using such a low-threshold screening strategy. Similar to previous digital screening studies, recruitment followed a direct-to-participant approach. However, instead of recruiting the participants from a pool of owners of corresponding smartwatches, in this case policyholders from a private health insurance company were selected based on specific risk criteria. As a result, the relevant risk group was addressed much more specifically as part of the eBRAVE-AF study. This resulted in not only diagnostic but also therapeutic effectiveness, as AF detection demonstrably led to OAC prescriptions. This in turn suggests a positive effect on stroke prevention, as the risk of stroke in risk groups can be significantly reduced by OAC (Hart et al., 2007; López-López et al., 2017). Especially since AF is associated with a variety of other diseases, the development or progression of which early AF detection can potentially help prevent.
Finally, some limitations – both with regard to the study conduct and with regard to the screening program examined – should be reported. The fact that the study was carried out exclusively among those insured by private health insurance limits its representativeness because in Germany almost 90% are cared for under statutory health insurance (selection bias). Due to the design of the GKV as social insurance based on the principle of solidarity, its insured structure differs from that of the PKV, which requires an individual risk assessment at the start of insurance. Furthermore, participation in the study – including in the control group – could have aroused attention to the topic in the first place. As a result, it cannot be ruled out that participation in the study as such had a distorting influence on the use of AF-related diagnostic services in standard care. If necessary, the study conditions would have put the effect shown here into perspective by artificially intensifying the diagnostic process in the control group and the screening would have been even more effective under real conditions.
As with most screenings, there is a risk of unnecessary testing and costs resulting from false-positive measurements. Due to the vast majority of “healthy” people, false-positive measurements are simply more significant. In the eBRAVE-AF study, AF was subsequently not confirmed in two of three participants with abnormal PPG measurements. Some of the unconfirmed cases could be explained by the fact that arrhythmias other than AF led to the abnormal PPG measurement or that the patient subsequently did not use any confirmatory ECG diagnostics at all or that AF occurred while the ECG device was being worn has not occurred. On the other hand, due to the preventive nature, there are various avoidable medical costs. A differentiated health economic model calculation has already come to the conclusion that smartphone-based AF screening is also advantageous from an economic point of view - especially because avoidable stroke costs overcompensate for the costs of screening (Birkemeyer et al., 2020).
Conclusion
The results of the eBRAVE-AF study show that, using a screening strategy based on common smartphones, the detection rate of AF requiring treatment in the 50+ age group can be more than doubled compared to standard care. The high prevalence of smartphones in the population makes corresponding screenings in the target group easily accessible and scalable. In this context, the study results also indicate a high level of acceptance of the procedure among older users (65+). As a result, in many of the AF cases identified in this way, treatment with blood-thinning medication can be initiated in order to avoid potential secondary diseases - especially strokes. However, future studies are needed to examine the effects of digital AF screening on clinical endpoints such as stroke prevention. <<

background
With a lifetime risk of about one in four, atrial fibrillation (AF) is one of the most common cardiac arrhythmias (Lloyd-Jones et al.; 2004). In those affected, the heart beats persistently irregularly, which leads to a lack of blood supply to the body. As a result, dangerous blood clots can form. These can in turn trigger a so-called thromboembolism - that is, a blockage of blood vessels - and, among other things, lead to an ischemic stroke - the second most common cause of death worldwide (Lozano et al., 2012) (Kirchhof et al., 2016). In Europe alone, over 430,000 people died in 2013 as a result of this circulatory disorder in the brain, which is also known as a cerebral infarction - almost 60,000 of them in Germany (Eurostat, 2016). In addition, stroke is a leading cause of acquired disability in adulthood (Murray et al., 2012).