Mass Screening for Untreated Atrial Fibrillation: The STROKESTOP Study

The study design was described in a previous publication.¹¹ Individuals born in 1936 to 1937 living in Stockholm County (n=23 888) or in the Halland region (n=4880) at the end of 2011 were identified by their unique civic registration numbers.

A computerized 1:1 randomization was performed in the 75- to 76-year-old population with stratification for sex, year of birth, and region. If the individual was randomized to the screening arm, an invitation to participate in an AF screening program was sent by mail. Nonresponders received 1 reminder in Halland and 2 reminders in Stockholm. Individuals who died before or during the invitation process were identified and excluded (Figure 1). Information on nationality at birth was obtained from the Swedish Central Bureau of Statistics.

Participants were informed orally and in writing at a screening center and signed informed consent forms before entering the study. Medical history, including AF, stroke/TIA, heart failure, hypertension, diabetes mellitus, myocardial infarction, or vascular disease, was obtained, as well as whether participants were on OAC or antiplatelet therapy. In Stockholm, participants also self-assessed height and weight.

Participants without a previous history of AF who were in sinus rhythm on the first visit were instructed in the use of a handheld ECG recorder for intermittent ECG recordings over 2 weeks. An index ECG for detection of permanent arrhythmia was obtained. A 1-lead ECG recorder from Zenicor (www.zenicor.com) with an integrated mobile transmitter that sends 30-second ECG strip data to a database was used. Participants placed their thumbs on the device twice daily and whenever they noticed palpitations. The device has been shown to have higher sensitivity for detection of AF than conventional 24-hour Holter recordings.^12–14 In cases of inconclusive ECG tracings, participants were offered additional ECG recordings according to the investigating cardiologist’s judgment.

AF was defined as at least one 30-second recording with irregular rhythm without p waves¹⁵ or a minimum of 2 similar episodes lasting 10 to 29 seconds during 2 weeks of intermittent recording. Research nurses, whose ECG skills were verified by random controls, manually assessed all ECG recordings. All abnormal ECGs were referred to the investigating cardiologist. If there was uncertainty about the presence of AF, the ECGs were adjudicated by a consensus group. In patients in whom other significant arrhythmias were detected, referral was made as appropriate.

All individuals with new AF and AF patients without OAC treatment were offered structured follow-up by a cardiologist to ensure adequate treatment, following current European guidelines.⁷

Continuous variables are reported as mean±SD. For continuous variables, the Student t test was used. For proportions, the Fisher exact test or χ² tests were used. Ordinal data were analyzed with the Mann-Whitney U or Kruskal-Wallis test. Two-tailed tests were applied. A multivariable analysis was performed with logistic regression. The discriminative ability of the model was estimated as c statistics. A value of P<0.05 was regarded as significant. These analyses were performed with IBM SPSS statistics version 22 software (IBM SPSS Statistics, IBM Corp, Somers, NY) and Open-Epi (Open Source Epidemiological Statistics for Public Health) version 3.01.

The study complies with the Declaration of Helsinki, and the protocol was approved by the regional ethics committee (DNR 2011-1363-31/3). Informed consent was obtained from all participants in the screening program (http://www.clinicaltrials.gov; identifier, NCT01593553).

Screening started in March 2012 and concluded in June 2014. In total, 14 387 inhabitants were invited to take part in screening. Before the invitation process was completed, 1056 individuals died. The remaining individuals’ response was 54% (7173 participants), which increased with additional invitations (Figure I in the online-only Data Supplement).

The participants registered 189 715 ECG recordings, with an average of 26.4 per subject. Only 1% made <15 ECG recordings. Because of difficulties in diagnosing AF from the recordings, 3.5% of participants were referred for 24-hour monitoring. Two patients withdrew their consent to participate. Referral for further workup was made in 10 patients as a result of high-grade atrioventricular block or sick sinus syndrome. Only 2% of participants (n=144) were of non-European descent.

Characteristics for participants are shown in Table 1.

New AF was detected in 218 patients (3.0%; 95% confidence interval [CI], 2.7–3.5). A previous diagnosis of AF was present in 666 patients (9.3%; 95% CI, 8.6–10.0), the majority of whom were men (n=407, 61.6%). Among those with known AF, 517 of 666 (77.6%) were using OACs at the index visit. Thus, 149 patients with known AF were not using OACs, constituting 2.1% (95% CI, 1.8–2.4) of the screened population. Hence the prevalence of untreated AF was 5.1% (95% CI, 4.6–5.7).

In participants who received a new diagnosis of AF, the mean number of registrations with AF was 4.5 (95% CI, 3.4–5.6). In 40 individuals, the diagnosis was made from 1 single pathological registration during intermittent screening (Figure 2). In 12 individuals, AF was diagnosed from ≥2 episodes ranging from 10 to 29 seconds.

Most participants with new AF were diagnosed during the first days of their 2-week ECG registration period (Figure 3). Only 37 cases were diagnosed from the ECG at the index visit. Intermittent monitoring diagnosed 4 times as many individuals with new AF compared with the index ECG (Figure II in the online-only Data Supplement).

Atrial flutter was diagnosed in 8 patients, and they were included in the AF group.

The total number of AF cases in the screened population was 884 (12.3%). Intermittent screening revealed 33% more cases of AF than previously known.

Men with known AF were treated with OACs more often (80.8% versus 72.4%; P<0.05). In summary, 5.1% (95% CI, 4.6–5.7) of the participants in screening had AF and were without OAC protection. Men (n=367) constituted 53.7% of the untreated population.

In participants with new AF, 93% accepted starting OAC treatment. The main reason for not initiating OAC treatment in participants with new AF was patient preference (see Figure 4).

In participants with known but untreated AF, anticoagulant treatment was initiated in 70 of 149 (47%). The main reason for not initiating OAC treatment was patient preference. Contraindications to OAC treatment were found in 14% of individuals in this group. Of 128 participants without contraindications, OAC treatment was started in 70 (55%).

Initiation of OAC treatment was made in 3.7% (95% CI, 3.3–4.2) of the screened population. The choice of which OAC to be prescribed was made according to the patients’ preferences. New OACs were initiated in 73%.

Participants with known vascular disease were more likely to be diagnosed with AF, and women were less likely to receive a new diagnosis of AF (Table 1 and Figure III in the online-only Data Supplement).

Participants with new AF were in general significantly taller and heavier with a higher body mass index compared with participants without AF. Mean CHA₂DS₂-VASc score did not differ significantly between the newly diagnosed AF group and participants free of AF. However, there was a significant (P<0.001) association between increasing CHA₂DS₂-VASc score and prevalence of AF (Figure 5). Participants with known AF had higher CHA₂DS₂-VASc scores (mean, 3.94; median, 4) than participants with newly detected AF (mean, 3.47; median, 3; P<0.001) and participants without AF (mean, 3.40; median, 3; P<0.001).

A multivariable analysis showed that the strongest predictor for AF (new or known) in the screened population was congestive heart failure, followed by previous stroke/TIA, diabetes mellitus, height (odd ratio per 1-cm increase), and weight (odds ratio per 1-kg increase; see Table 2).

Previous studies^16,17 have used risk scores to predict the development of AF. We calculated a modified score on the basis of the work of the Cohorts for Heart and Aging Research in Genomic Epidemiology–Atrial Fibrillation (CHARGE-AF)¹⁶ Consortium, using height, weight, history of diabetes mellitus, hypertension, vascular disease, and congestive heart failure, which showed a modest capacity to predict in which patients AF was most likely (c statistic, 0.692; 95% CI, 0.670–0.717).

To study whether any group would not benefit from screening, a multivariable analysis was performed to compare risk factors for participants in whom AF was detected compared with the group with no detection of AF (Table I in the online-only Data Supplement). Female sex, lower weight, and absence of vascular disease were significantly associated with no detection of AF. In women with a body mass index <25 kg/m², screening yielded only 1.3% new AF.

Participation in the screening program was higher in rural Halland than in urban Stockholm (64% versus 52%; P<0.001), and AF was a more commonly found in Halland (4.0% versus 2.8%; P=0.02). In Halland, individuals with known AF were more likely to be on OACs (87% versus 75%; P=0.001; Table II in the online-only Data Supplement).

Compared with established screening programs in Sweden, participation was lower.^18–20 However, participation in epidemiological studies has shown a declining trend,²¹ and compared with participation in another large research study in Sweden, our participation was higher.²² Factors that could explain a lower participation include the fact that the general public could perceive that partaking in a research screening program over 14 days is more cumbersome than participating in a single-visit established screening program. In addition, the age selected for our study was higher than in other programs; higher age is associated with more disability, which could affect participation.²³ The invitation was written in Swedish, which could exclude non–Swedish-speaking participants.

Previous screening studies for AF have focused on strategies using screening for AF at a single time point, which is likely to detect only permanent arrhythmia and might miss most patients with paroxysmal arrhythmia. Intermittent ECG recording has shown a higher detection rate of AF compared with 24-hour Holter monitoring.¹³ In a meta-analysis of screening programs, using a single ECG recording found new AF in 1.4% of subjects >65 years of age.²⁴ The current recommendation in the European Society of Cardiology guidelines for opportunistic screening for AF is based on a study comparing opportunistic ECG screening using pulse palpation in which the investigators found 1.64% new AF with systematic screening using 12-lead ECGs and 1.62% new AF was found in a population with a mean age of 75 years.^7,23 Our study found 0.5% new AF via the initial ECG recording (Figure II in the online-only Data Supplement), but with intermittent systematic screening, detection of new AF increased 4-fold to 3% of the screened population, which is almost double the prevalence of previous studies in which single-time-point ECG screening was performed.^23,24 This might indicate that AF is well recognized and treated in Sweden and that most individuals with a more permanent arrhythmia have already been diagnosed. The use of an intermittent ECG recorder is more likely to detect patients earlier before AF becomes permanent.¹⁵

Before intermittent ECG screening, the screened population reported a 9.3% prevalence of AF. This is a higher prevalence of AF compared with studies of populations of similar age groups in North America,^25,26 the UK,²⁷ and Greece²⁸ and slightly lower than Icelandic and Dutch prevalence studies.^27,29,30 Systematic screening for AF in populations of a similar ethnicity in which the prevalence of known AF is lower would likely yield a higher rate of newly discovered AF, which would further increase the cost-effectiveness of systematic AF screening.

The majority of AF prevalence studies have been performed in Western European countries and in North America. Studies from low- and middle-income countries indicate a lower AF prevalence,³¹ but as the population ages and the prevalence of risk factors for AF increases, AF prevalence is likely to increase in these areas,³² and systematic AF screening might become beneficial.

Our multivariable analysis showed that the strongest predictor for AF (new or known) in the screened population was heart failure, previous stroke/TIA, and diabetes mellitus. Our results are consistent with results from the Framingham study in which congestive heart failure was also shown to be one of the most important risk factors for AF with a 4.5- and 5.9-fold increased risk in men and women, respectively, whereas diabetes mellitus conferred a risk of 1.4 and 1.6, respectively.³³ A history of stroke/TIA was also a predictor for AF. According to current European stroke guidelines,³⁴ these patients should already have been screened for AF with 24-hour Holter monitoring unless another cause for stroke was apparent. This short monitoring period, which differs from the more extended screening for AF in the American Heart Association guidelines,³⁵ presumably leads to underdetection of AF.³⁶

Individuals with vascular disease were diagnosed with new AF to a greater extent, despite the fact that they most likely had already been subjected to cardiovascular workup (increasing the probability that AF should have been diagnosed). This could signify that individuals with vascular disease are more prone to developing AF³⁷ but also shows what a difficult end point AF is because of its asymptomatic and intermittent nature. Participants who weighed more, were taller, and had a higher body mass index were more likely to receive a diagnosis with AF, which is in accordance with previous studies.^16,38

In women with a body mass index <25 kg/m², only 1.3% new AF was found. This might indicate that in women AF screening might be initiated at an older age because the prevalence of AF is lower in women compared with men in the same age category.⁹ However, women have a higher risk of stroke compared with men,³⁹ so continuing screening at the same age as men might still be pertinent because the individuals found are at a higher risk.

Participants with known AF had more comorbidity. Increasing CHA₂DS₂-VASc scores correlate with an increased presence of AF (Figure 5).

In a pilot study, individuals 75 years of age with at least 1 additional risk factor for stroke underwent screening for silent AF with intermittent ECG registration.⁴⁰ Previously untreated AF was found in 7% of the participants. In the pilot study, patients with newly diagnosed AF had a CHA₂DS₂-VASc mean score of 3.85 and a median score of 4; in the present study, the respective scores were 3.47 and 3. This difference (P=0.03) in comorbidity may explain some of the difference in AF prevalence.⁴⁰

In patients with new AF, acceptance of treatment with OACs despite a lack of symptoms was high. A possible reason is that a cardiologist with special interest in AF did the workup. More than 70% of participants chose treatment with novel OACs.

Compared with previous studies,^9,41,42 a greater proportion of participants with known AF were on OAC treatment than expected. The invitation stated that patients with AF would be referred for OAC treatment; hence, patients with a negative attitude toward OACs might have chosen not to participate. Individuals attending screening might be more health aware and therefore on correct treatment. Initiation of OAC treatment in participants with known AF was lower compared with those with new AF. The same cardiologists initiated treatment for both groups, so patient information was similar. There could be several reasons for lower initiation. Individuals with known AF could have tried and discontinued OAC treatment, making them less willing to try again. There could be difficulties in realizing the risks of untreated AF, especially in patients who avoided long-term treatment without suffering consequences or received erroneous information in the past that aspirin would suffice. In this study, almost 20% of individuals with known AF were treated with aspirin despite poor effects on stroke prevention in individuals with AF.^8,43 In an AF population less well treated with OAC, it is probable that systematic screening could yield further increases in OAC treatment among participants.

In Halland, screening uptake was greater, which could be explained in part by decentralized organization of screening compared with the centralized organization in Stockholm. Differences in screening uptake within each area might also be correlated to socioeconomic status.⁴⁴

Proportionally more cases of new AF were found in Halland compared with Stockholm (4.0% versus 2.8%). Screening uptake is usually poorer among individuals who are more prone to have the condition being investigated.⁴⁵ The regional difference in participation might explain the difference in detection of new AF if AF prevalence is higher in the group not attending in Stockholm.

The Swedish central government agency of Dental and Pharmaceutical Benefits, TLV, has the role of determining whether a pharmaceutical product shall be subsidized by the state and has recently published an extensive analysis on the cost-effectiveness of screening for AF with intermittent ECG recordings.⁴⁶ It predicts that the STROKESTOP screening program would incur a cost of €4164 per quality-adjusted life-year when a lifelong perspective is used, which is regarded as a low cost per quality-adjusted life-year. It could hence be implied that, in a 75- to 76-year-old population of similar ethnicity, screening for AF will be cost-effective.

The World Health Organization lists 10 conditions (Table III in the online-only Data Supplement) that should be fulfilled to justify mass screening.⁴⁷ AF meets all of these criteria. In screening for malignancies, conditions that might not affect the patients’ life expectancy or symptoms might be found. In contrast, in screening for AF, there is an instantaneous indication for initiating OAC treatment in individuals at risk of thromboembolic events.

Several other screening programs have already been implemented and have been accepted by society.^18–20 The diagnostic yield of these programs is lower than that of screening for AF.

Because the registration period was limited to 14 days and the participants were monitored only <1% of the time, there was probably underdetection of AF. A patient who has AF detected in this brief period of monitoring is presumably one with a high burden of AF. It is plausible that continuous monitoring with an implantable device would increase the yield of AF, but the upfront costs would be greater, and the procedure probably would be less acceptable to the participants.^3,48

Atrial flutter can remain undiagnosed because its detection can be difficult with this method.¹² Because atrial flutter is an arrhythmia more commonly observed in blacks compared with other racial groups, this might be more of a concern in black populations.⁴⁹

These limitations should be weighed against the ease of use and high degree of compliance.

Most clinicians unfailingly agree to initiate OAC treatment in individuals with asymptomatic AF and increased CHA₂DS₂-VASc score if AF is found en passant on a regular health check, and a recent study of stroke risk in incidentally discovered AF supports this.⁵⁰ However, the relationship between AF burden and stroke risk is not fully established. In 28.6% of individuals with AF detected by intermittent ECGs, the diagnosis was made from 1 episode of AF, and 12 individuals presented with ≥2 episodes of AF <30 seconds, likely representing a low AF burden. In studies using continuous monitoring such as the Asymptomatic Atrial Fibrillation and Stroke Evaluation in Pacemaker Patients and the Atrial Fibrillation Reduction Atrial Pacing Trial (ASSERT), no clear temporal relationship between duration of AF and stroke has been seen.^3,51 In a recent study of patients with implanted cardiac devices, a threshold of ≥5 minutes during a median monitoring period of 24 months was statistically significantly associated with the occurrence of ischemic stroke.⁵² However, AF of a duration <5 minutes was not studied. Compared with continuous monitoring, the finding of AF on intermittent monitoring is likely to signify a high AF burden because the monitoring represents a short temporal time frame.

Self-reporting of risk factors could lead to partially erroneous reporting of comorbidity, but in our pilot study, patient information on (self-reported) comorbidity was accurate in 99% of cases.⁴⁰

Studies comparing AF prevalence across different racial groups have shown heterogeneity with a higher prevalence in whites compared with blacks, Hispanics, and Asians.^32,49,53,54 This difference remains even if AF prevalence is studied in pacemaker studies, in which differences in healthcare consumption patterns are of no importance.⁵⁴ Even though AF was less common in blacks compared with whites in a study of an American biracial population, stroke remained a vast problem, with higher incidence rates than in the white population,⁵⁵ raising the possibility that the importance of AF as a risk factor for stroke might vary between ethnic groups.

Our data describe a population in which 98% of participants were of European descent. Hence, systematic screening for AF in a population with different racial composition might yield a lower prevalence.

The questions of whether screening for AF and initiating OAC treatment will reduce the risk of ischemic stroke and to what extent individuals are compliant with treatment remains to be investigated. We plan to follow up on our participants in 5 years using data from the national health registries and the national prescription registries to observe whether intermittent screening for AF and initiating OAC treatment will reduce the risk of stroke compared with the nonscreened population.

Mass screening in 75- to 76-year-old individuals with intermittent ECG recordings yields a considerable proportion of individuals with untreated AF who can be started on OAC treatment.