Subclinical Atherosclerosis Is Associated With Incident Atrial Fibrillation

A Systematic Review and Meta-analysis

Kit Engedal Kristensen; Cille Cederholm Knage; Liv Havgaard Nyhegn; Bart A. Mulder; Michiel Rienstra; Isabelle C. Van Gelder; Axel Brandes


Europace. 2020;22(7):991-1000. 

In This Article


Search Outcome

The initial search conducted in October 2018 resulted in 466 articles. The full search and selection process is displayed in a PRISMA flow diagram (Figure 1). Of the 466 articles 97 were duplicates, which left 369 articles for screening at the title and abstract level. We excluded 348 articles because they did not meet the inclusion criteria. Three additional articles were identified through chain search, resulting in 24 articles for full-text screening. Data from the Rotterdam study were presented as a single cohort and in a pooled analysis, where we chose to include the single cohort study by Heeringa et al.[11,12] Two studies based on the Multi-Ethnic Study of Atherosclerosis (MESA) cohort were included, since one used cIMT and the other used CACS as measurement for subclinical atherosclerosis.[12,13] Authors of nine articles were contacted for further information. None of these correspondences led to further data material. The updated literature search conducted in January 2019 resulted in additional 21 articles, where none were eligible for inclusion.

Figure 1.

PRISMA flow diagram of the meta-analysis.

After full-text assessment, seven cohort studies were included in the final analyses, of which five measured cIMT and two measured CACS.[11–16]

Study Designs and Patient Characteristics

An overview of the study designs, individual characteristics, and baseline measurements is shown in Table 1. Overall, the included studies had relatively similar study designs and patient characteristics, but some differences should be noted.

Five of the seven studies were population-based prospective cohort studies.[11–14] Vinter et al. reported a retrospective registry-based cohort study in a population, which was notably different from the other studies. The population in this study consisted of individuals, which were suspected, but proven free of CAD.[16] The study based on the Campania-Salute Network by Losi et al. differed as well, as this was a retrospective cohort study in a hypertensive population.[15]

There were also differences in, how prevalent AF and clinically manifest atherosclerosis were determined in the studies to exclude these cases (Table 1). In the Atherosclerosis Risk in Communities (ARIC), MESA, and Rotterdam Studies, AF on the baseline electrocardiogram (ECG) was defined as prevalent AF.[11–13] In the three other studies, the definition of prevalent AF was based on registry-recorded diagnoses of AF before enrolment.[14–16] Clinically manifest atherosclerosis was determined from patient questionnaires and subsequent review of medical records in the ARIC, MESA, and Rotterdam Studies,[11–13] while the three other studies used registry-recorded diagnoses.[14–16] Interestingly, the ARIC study cohort included few patients with a history of myocardial infraction and heart failure.[12]

Moreover, the seven studies also differed in the data sources for their outcome incident AF. Five of them used a combination of ECG documentation and hospital diagnoses of AF,[11–13,15] of which four utilized additional data sources, e.g., death certificates and medical files from general practitioners.[11–13] The two other studies by Adamsson Eryd et al. and Vinter et al. only relied on hospital diagnoses from administrative registries.[14,16] Most studies did also not distinguish between AF and atrial flutter (Table 1).[11–14,16]

All studies included both men and women, while six of them also adjusted or stratified for sex in their analyses (Table 2 and Table 3). The mean age was between 52.0 and 68.5 years, with a mean of all study participants of 58.3 years. The length of follow-up in the included studies was between a median of 2.9 years and 17.8 years (Table 1).

Carotid Intima-media Thickness and Incident Atrial Fibrillation

The results of the studies investigating the association between cIMT and incident AF are shown in Table 2. The HRs and 95% confidence intervals (CI) are the most comprehensively adjusted. In four of the five studies, the cIMT groups were stratified in quartiles or quintiles, and the HRs for incident AF during the follow-up period were calculated comparing participants in the highest with the lowest quartile or quintile.[11,12,14]

All five studies showed a significant association between increased cIMT and incident AF. Hazard ratios (95% CI) were 1.29 (1.08–1.54), 1.94 (1.15–3.28), 1.52 (1.08–2.16), 1.90 (1.20–3.00), and 1.51 (1.27–1.79), respectively, with an overall HR of 1.43 (95% CI 1.27–1.59) (Table 2 and Figure 2).

Figure 2.

Association between subclinical atherosclerosis, defined by increased carotid intima-media thickness, and incident atrial fibrillation. Weights are from fixed-effects analysis. CI, confidence interval.

The HRs stratified for sex did not show consistent results. Chen et al. (ARIC study) and Heeringa et al. found a statistically significant association between cIMT and risk of AF in women, but not in men.[11,12] On the contrary, Adamsson Eryd et al.[14] only found a statistically significant HR in men.

It is important to note that the five studies used different measurement methods of cIMT and different definitions of carotid plaque. In the Rotterdam Study, Heeringa et al. only used qualitative criteria to define the presence of carotid plaques,[11] while Adamsson Eryd et al. and the ARIC Study used quantitative and qualitative criteria to define the presence of carotid plaques.[12,14] The two other studies only used (semi-)quantitative criteria, i.e., a cut-off value for IMT (Table 1).[12,15]

Coronary Artery Calcium Score and Incident Atrial Fibrillation

Table 3 displays the results of the two studies investigating the association between CACS and incident AF. The HRs and 95% CI shown are the most comprehensively adjusted. The results were stratified by CACS groups and the HRs were calculated for each group using CACS = 0 as reference.[13,16]

The two studies showed a significant association between increased CACS and incident AF in all groups except for the group CACS 1–99 in the study by Vinter et al.[16] Both studies showed an increase in HR with increasing CACS, except for a CACS >1000 in the study by Vinter et al.

The two studies analysed CACS as a continuous variable to examine the risk of AF when the CACS doubled. These analyses also showed a significant increase in the risk for incident AF with increased CACS with a Log2 (CACS + 1) HR of 1.1 (95% CI 1.05–1.13) and 1.05 (95% CI 1.03–1.08), respectively. The results remained unchanged in both studies when stratified or adjusted for sex. The overall HR was 1.07 (95% CI 1.02–1.12) (Table 3 and Figure 3). The test for heterogeneity showed an I 2 of 76.8% (P = 0.038), which represent a substantial and considerable heterogeneity.

Figure 3.

Association between subclinical atherosclerosis, defined by increased coronary artery calcium score, and incident atrial fibrillation. Weights are from random-effects analysis. CI, confidence interval.

Both studies used the Agatston score method for the measurement of CACS.[17]

Validity and Study Quality Assessment

The quality assessment of the seven studies according to the NOS is summarized in Table 4. All studies had sufficient exposure ascertainment. The studies by Adamsson Eryd et al.,[14] Losi et al.,[15] and Vinter et al.[16] did not explicitly evaluate, whether AF was present at baseline. However, Adamsson Eryd et al.[14] did not use AF, but hospitalization for AF as an outcome, and demonstrated in this way that their specific outcome of interest was not present at the beginning of the study.