36
Electrocardiographic PR Prolongation and Atrial Fibrillation Risk: A Meta-Analysis of Prospective Cohort Studies MIN CHENG, Ph.D., XIANGFENG LU, Ph.D., JIANFENG HUANG, Ph.D., SHU ZHANG, Ph.D., and DONGFENG GU, Ph.D. From the State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
PR Interval and Atrial Fibrillation. Background: Electrocardiographic PR interval prolongation is considered a benign condition, but recent studies have challenged the notion by demonstrating that prolonged PR interval is associated with an increased risk of atrial fibrillation (AF). Objective: The purpose of this study was to perform a meta-analysis of prospective cohort studies to evaluate the evidence supporting an association of prolonged PR interval with AF incidence. Methods: We searched the MEDLINE and EMBASE database (from inception to May 2014) supplemented by manual searches of references of relevant retrieved articles. Prospective cohort studies were included with hazard ratio (HR) of prolonged PR interval for incident AF. Results: The search strategy yielded 6 cohort studies meeting eligibility criteria. A total of 328,932 participants were included, with 14,191 participants suffering from AF during follow-up. Pooled HRs of prolonged PR interval for incident AF was 1.30 (95% CI: 1.13 to 1.49) using random-effect model (I2 = 30%). There was a significant difference of combined HRs between studies with and without adjustment for taking of AV nodal blocking agents in subgroup analysis. Sensitivity analysis supported the robustness of the results. Conclusions: Prolonged PR interval is not a totally benign condition but an independent risk factor for AF incidence. The mechanisms underlying the association of prolonged PR interval with AF incidence need further research. (J Cardiovasc Electrophysiol, Vol. 26, pp. 36-41, January 2015) atrial fibrillation, AV node, epidemiology, Framingham study, PR interval Introduction Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, occurring in 1–2% of the general population, and its prevalence is estimated to be at least doubled in the next 50 years. The mechanisms causing and sustaining AF are multifactorial, and AF can be complex and difficult for clinicians to manage. AF confers a 5-fold risk of stroke and is associated with increased mortality.1-3 AF is also expensive, adding approximately $8,700 per year (estimate from 2004 to 2006) for a patient with AF compared to a patient without AF. PR interval is the time required for an electrical impulse to travel from the atrial myocardium adjacent to the sinus node through the atrioventricular node (AVN) to the Purkinje fibers. Electrocardiographically, prolonged PR interval, or first-degree atrioventricular (AV) block, is defined by PR interval >200 milliseconds. In the majority of cases, it is due to delayed conduction in the AV node, although conduction This study was supported by the National Natural Foundation. No disclosures. Address for correspondence: Dongfeng Gu, Ph.D., State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beili Shi road 167, Beijing, People’s Republic of China. Fax: +01-08-839-6147; E-mail: [email protected] Manuscript received 5 July 2014; Revised manuscript received 30 July 2014; Accepted for publication 19 August 2014. doi: 10.1111/jce.12539
delay in the atrium, the His-Purkinje system or multiple sites may contribute to the prolonged PR interval.4 In the absence of structural heart disease or other conduction abnormalities, prolonged PR interval has been traditionally considered a benign phenomenon.5,6 The association of prolonged PR interval with AF has been investigated in some studies but results were inconclusive. Though the Framingham Heart Study found an increase in AF incidence associated with prolonged PR interval,7 the Finnish Social Insurance Institution’s Coronary Heart Disease Study failed to show any association between them.8 We therefore set out to conduct a meta-analysis to evaluate the evidence supporting an association between prolonged PR interval and AF incidence. Methods We followed the Meta-analysis of Observational Studies in Epidemiology protocol throughout the design, implementation, analysis, and reporting for this study9 (see Supplemental Appendix S1). Literature Search Strategy We performed a literature search of MedLine and EMBASE using the key words “PR interval,” “atrioventricular block,” “arrhythmia,” and “atrial fibrillation” from inception to May 2014. We reviewed all articles with an abstract suggesting relevance. We checked the reference lists of relevant articles to find other eligible trials. Prospective cohort studies were included. Supplemental Appendix S2 provides a detailed description of the search strategies.
Cheng et al.
PR Interval and Atrial Fibrillation
37
Figure 1. Flow diagram of the study selection process.
Data Extraction and Quality Evaluation The following details were recorded for each study: author, year of publication, cohort study name, the geographic location of study, study period, participants’ sex, age range at baseline, health at baseline, and outcome (defined as incidence of AF/flutter). For each study, we obtained information on the most adjusted hazard ratio (HR) for prolonged PR interval with corresponding 95% confidence intervals (CIs) or P value. Data abstraction was conducted independently by 2 investigators, with disagreements resolved by a third investigator. The Newcastle-Ottawa Quality Assessment Scale was used to assess bias in included studies. Statistical Analysis In this meta-analysis, the adjusted HR with 95% CIs was considered as the effect size for all studies. A randomeffects meta-analysis was performed to estimate pooled HRs weighted by the inverse variance method considering different measurement methods and reference range of PR interval. The heterogeneity among studies was estimated using the I2 statistic. Heterogeneity was confirmed with a significance level of P < 0.10. Meta analysis with fixed-effects model was conducted for sensitivity analysis. Publication bias was assessed using the Begg-adjusted rank correlation test. All statistical analysis was performed using Stata software (Stata 12.0; Stata Corp, College Station, TX, USA). All reported probabilities were 2-sided with P < 0.05 considered statistically significant. Results Figure 1 shows the results of literature research and selection. There were 2,799 citations retrieved on initial research. After removal of 219 replicates, 2,553 citations were excluded at the title or abstract level. Full-text assessment of 27 articles resulted in 6 prospective cohort studies entering into final meta-analysis. Description of Studies Included Table 1 shows the information extracted from the included studies. These studies included 328,932 participants in total. During a mean follow-up ranging from 5.7 to 30 years, 14,191
participants suffered from AF. Three studies were conducted in the United States7,10,11 ; the remaining studies were conducted in Australia,12 Finland,8 or Denmark.13 Participants’ mean age ranged from 44 to 74 years at baseline. Female percentage ranged from 48% in the Aro study8 to 56% in the Knuiman study.12 Taking of β-blockers or calcium antagonists on the day of ECG recording was 27% in the Nielsen study.13 Prescription of chronotropic medication was 4.4% in the Aro study.8 Selected medication including amiodarone, cardiac glycosides, calcium channel blockers and β-blockers was used in 41.7% participants in the Magnani study.10 The number of persons on AV node blocking agents was not mentioned in remaining studies. The reporting quality of the included studies was globally acceptable (see Supplemental Appendix S2). There was a concern of selection bias in the Nielsen study, which enrolled people with an electrocardiography (ECG) recorded at Copenhagen General Practitioners’ Laboratory. AF incidence in the population was higher compared with the general population.13 There was also a concern of potential selection bias in the Magnani study, which enrolled people aged 70–79 years old.10 AF incidence was documented using ECG in all studies. Method of PR interval measurements was different across included studies. Lead II was used for PR measurement in the Cheng7 and Magnani studies.10 The distance between the earliest detection of atrial and ventricular depolarization in any lead was used for PR measurement in the Nielsen study.13 Bipolar limb lead in which PR interval was longest was used in the Aro study.8 The mean P-wave duration plus the mean PR-segment duration in the 12-lead ECG was used in the Soliman study.11 The method of PR measurement was not specified in the Knuiman study.12 Effect size was estimated for prolonged PR interval (PR interval more than 200 milliseconds) with comparison with a “normal” reference PR interval that was defined in individual studies differently. PR interval less than 200 milliseconds was defined as “normal” reference in the Cheng,7 Aro,8 Soliman11 and Magnani10 (PR interval < 80 milliseconds excluded) studies. The “normal” range was defined as PR interval less than 200 milliseconds and more than 120 milliseconds in the Knuiman study.12 In the Nielsen study,13 PR interval ranged from 150 to 161 milliseconds was
10,758
Finland, 30 years 55%
48%
56%
55%
54
43
52
54
73
46
Mean Age (years)
Limb lead in which PR is longest The mean P-wave duration plus the mean PR-segment duration
The distance between the earliest atrial and ventricular depolarization NS
Lead II
Lead II
PR Measurement Method
1.59 (0.77–3.30)
1.03 (0.74–1.45)
1.29 (0.68–2.44)
1.26 (1.17–1.35)
1.26 (0.99–1.61
2.06 (1.36–3.12)
HR (95% CI)
Sex, age, height, HTN treatment, BMI Sex, age, chronotropic medication, cardiovascular disease, BMI, HF Sex, age, ethnicity, HTN, SBP, diabetes, blood lipids, smoking, BMI
PR ࣘ 200 milliseconds PR ࣘ 200 milliseconds
Sex, age, cardiovascular disease status, HR, BMI, HTN, smoking, diabetes, total/HDL cholesterol, atrial premature beats, valve disease, ECG LVH Sex, age, cardiovascular disease, selected medications (amiodarone, cardiac glycosides, calcium channel blockers, β-blockers), BMI, HF, SBP, DBP, site, smoking, total/HDL, ECG LVH, hypertension treatment Sex, age, HTN, HF, MI, valvular heart disease, treatment with AV nodal-blocking medication, diabetes, hyperthyroidism, HR, QT interval, ECG LVH
Variables Adjusting for Effect Estimates
Normal PR
150 milliseconds < PR ࣘ 161 milliseconds
80 milliseconds < PR ࣘ 200 milliseconds
PR ࣘ200 milliseconds
Defined Reference PR
HR = heart rate; BMI = body mass index; HDL = high-density lipoprotein; HTN = hypertension; ECG = electrocardiography; LVH = left ventricular hypertrophy; MI = myocardial infarction; HF = heart failure; SBP = systolic blood pressure; NS = not stated.
15,429
4,267
Australia, 15 years
Soliman, 2009, ARIC Study United States, 6.97 years
Knuiman, 2013, the Busselton Health Study Aro, 2014, CHD Study
288,181
52%
54%
Sex, Female (%)
Journal of Cardiovascular Electrophysiology
Nielsen, 2013, Copenhagen Denmark, 5.7 ECG Study years
2,722
Magnani, 2013, Health ABC United States, 10 years
Total Number (n) 7,575
Location and Duration
United States, 12 years
Cheng, 2009, Framingham Heart Study
Author, Publication year, Cohort
TABLE 1 Characteristics of Cohort Studies of Prolonged PR Interval and Atrial Fibrillation Included in the Meta-Analysis with Multivariable-Adjusted Estimates
38 Vol. 26, No. 1, January 2015
Cheng et al.
PR Interval and Atrial Fibrillation
39
Figure 2. Combined HR of prolonged PR interval for AF incidence using random-effects model. HR = hazard ratio; AF = atrial fibrillation.
Figure 3. Pooled HR in subgroup analysis according to the adjustment for AV block agents. HR = hazard ratio.
used as reference. Confounding factors adjusted were different across studies. Age, sex, and hypertension (treatment) or blood pressure were adjusted for effect size estimates in all studies, but taking of AV nodal-blocking medication was adjusted only in the Magnani,10 Aro,8 and Nielsen13 studies. Heart rate was adjusted in the Nielsen13 and Cheng7 studies. BMI was not adjusted in the Nielsen study.13 ECG LVH was adjusted in the Nielsen,13 Magnani,10 and Cheng7 studies. PR Prolongation and Risk of AF In primary analysis, the pooled HR for AF incidence of prolonged PR interval was 1.30 (95% CI: 1.13 to 1.49; P = 0.000) using random-effects model. Heterogeneity was
moderate between studies (I2 = 30% and P = 0.21) (see Fig. 2). Pooled HRs were different in subgroup analysis according to confounding factors adjusted (P = 0.039). Pooled HR from studies adjusting for taking of AV nodal blocking agents was 1.25 (95% CI, 1.17–1.34; I2 = 0%) for AF incidence, while pooled HR was 1.75 (95% CI: 1.28 to 2.40, I2 = 0%) from studies without taking of AV nodal blocking agents adjustment (see Fig. 3). Sensitivity Analysis and Publication Bias Sensitivity analyses using fixed-effects model supported the robustness of the primary results. The Begg-adjusted rank
40
Journal of Cardiovascular Electrophysiology
Vol. 26, No. 1, January 2015
correlation test provided no evidence of substantial publication bias in the primary meta-analysis (P > 0.05). Because we included a small number of studies, formal assessment of publication bias might not be appropriate. Discussion Results of this meta-analysis demonstrate that electrocardiographic PR interval prolongation is a risk factor for AF incidence independent of age, sex, and hypertension. Although the notion that electrocardiographic PR interval prolongation is a benign condition has been challenged by the Framingham Heart Study in which prolonged PR interval was associated with more than twofold increases in AF incidence,7 results from a recent large cohort study are still supportive. In the Finnish Social Insurance Institution’s Coronary Heart Disease Study, prolonged PR interval was normalized in almost 30% of the subjects during 30 years follow-up.8 There are no definite explanations for the difference of results across studies. It is speculated that varying underlying substrate for PR interval prolongation may account for the difference. For example, enhanced vagal tone is assumed to be a major contributor in the young, and the prolonged PR interval may normalize with aging. In contrast, structural remodeling such as atrial fibrosis may play a more important role in the development of PR prolongation in the old, and it is well known that AF risk is proportional to atrial structural remodeling. In this case, prolonged PR interval may just be an indirect marker of atrial structural remodeling, which is a substrate for AF. This speculation was supported by the results of Soliman’s study, which demonstrated that associations of PR interval with adverse outcomes were dependent on the level of P-wave duration contribution to the overall length of PR interval, a contribution that varied across populations.14 In this meta-analysis with strict inclusion criteria, we found that PR prolongation was significantly associated with increased risk of AF. Furthermore, the moderate heterogeneity between included studies was mainly due to the differential adjustment of confounding factors. Pooled HRs were significantly different between studies with and without the taking of AV block agents adjustment. Since AV nodal blocking agents are frequently used in patients with high AF risk, it is not surprising to find out that the taking of AV blocking agents is a significant confounding factor for the effect estimates of prolonged PR interval for AF risk. There are several potential explanations for the observed association of prolonged PR interval with AF incidence. First, prolonged PR interval may share genetic background with AF. The heritability estimate of PR interval ranged from 34% to 78%.15-17 Genome wide association studies have found that chromosome loci from voltage gated sodium channel and cardiac developmental genes are associated with PR interval and many also associated with AF.18-24 In addition, AF associated genetic variant rs 220073 in chromosome 4q25 has been found associated with PR interval, further supporting the notion of a shared genetic background for both PR prolongation and AF incidence.25,26 Second, PR interval duration is strongly influenced by the autonomic nervous system, and PR prolongation may be a marker of abnormal cardiac autonomic abnormality.27,28 It is well known that both parasympathetic and sympathetic tone, or more likely interplay between the two, play an important role in the genesis
of AF.29-31 Third, PR prolongation is a marker of atrial electrical and structural remodeling and reflective of increased atrial conduction time that may directly increase the risk of AF. AV dyssynchrony may also play an important role in the association between PR prolongation and risk of AF. It is, however, still unclear what is the most important contributor to the observed association of prolonged PR with AF, and further research needed. The results of this meta-analysis are meaningful for the battle against AF. First, with inclusion of more than 300,000 participants, this meta analysis robustly supports the predictive role of prolonged PR interval for AF incidence. Second, research on prolonged PR interval will improve our understanding of AF pathophysiology. Third, 30% increases in AF risk associated with prolonged PR interval suggest a potential management benefit of PR prolongation for AF prevention. At present, there are few effective approaches to directly normalize prolonged PR interval except for cardiac pacing, and reversal of atrial remodeling that may contribute to P-wave duration might be a biologically plausible means of reducing AF. However, it is worth mentioning that though normalization of prolonged PR interval may reduce AF risk, there is no evidence to support the assumption that the shortening of normal PR interval (defined as PR interval
Electrocardiographic PR Prolongation and Atrial Fibrillation Risk: A Meta-Analysis of Prospective Cohort Studies MIN CHENG, Ph.D., XIANGFENG LU, Ph.D., JIANFENG HUANG, Ph.D., SHU ZHANG, Ph.D., and DONGFENG GU, Ph.D. From the State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
PR Interval and Atrial Fibrillation. Background: Electrocardiographic PR interval prolongation is considered a benign condition, but recent studies have challenged the notion by demonstrating that prolonged PR interval is associated with an increased risk of atrial fibrillation (AF). Objective: The purpose of this study was to perform a meta-analysis of prospective cohort studies to evaluate the evidence supporting an association of prolonged PR interval with AF incidence. Methods: We searched the MEDLINE and EMBASE database (from inception to May 2014) supplemented by manual searches of references of relevant retrieved articles. Prospective cohort studies were included with hazard ratio (HR) of prolonged PR interval for incident AF. Results: The search strategy yielded 6 cohort studies meeting eligibility criteria. A total of 328,932 participants were included, with 14,191 participants suffering from AF during follow-up. Pooled HRs of prolonged PR interval for incident AF was 1.30 (95% CI: 1.13 to 1.49) using random-effect model (I2 = 30%). There was a significant difference of combined HRs between studies with and without adjustment for taking of AV nodal blocking agents in subgroup analysis. Sensitivity analysis supported the robustness of the results. Conclusions: Prolonged PR interval is not a totally benign condition but an independent risk factor for AF incidence. The mechanisms underlying the association of prolonged PR interval with AF incidence need further research. (J Cardiovasc Electrophysiol, Vol. 26, pp. 36-41, January 2015) atrial fibrillation, AV node, epidemiology, Framingham study, PR interval Introduction Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, occurring in 1–2% of the general population, and its prevalence is estimated to be at least doubled in the next 50 years. The mechanisms causing and sustaining AF are multifactorial, and AF can be complex and difficult for clinicians to manage. AF confers a 5-fold risk of stroke and is associated with increased mortality.1-3 AF is also expensive, adding approximately $8,700 per year (estimate from 2004 to 2006) for a patient with AF compared to a patient without AF. PR interval is the time required for an electrical impulse to travel from the atrial myocardium adjacent to the sinus node through the atrioventricular node (AVN) to the Purkinje fibers. Electrocardiographically, prolonged PR interval, or first-degree atrioventricular (AV) block, is defined by PR interval >200 milliseconds. In the majority of cases, it is due to delayed conduction in the AV node, although conduction This study was supported by the National Natural Foundation. No disclosures. Address for correspondence: Dongfeng Gu, Ph.D., State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beili Shi road 167, Beijing, People’s Republic of China. Fax: +01-08-839-6147; E-mail: [email protected] Manuscript received 5 July 2014; Revised manuscript received 30 July 2014; Accepted for publication 19 August 2014. doi: 10.1111/jce.12539
delay in the atrium, the His-Purkinje system or multiple sites may contribute to the prolonged PR interval.4 In the absence of structural heart disease or other conduction abnormalities, prolonged PR interval has been traditionally considered a benign phenomenon.5,6 The association of prolonged PR interval with AF has been investigated in some studies but results were inconclusive. Though the Framingham Heart Study found an increase in AF incidence associated with prolonged PR interval,7 the Finnish Social Insurance Institution’s Coronary Heart Disease Study failed to show any association between them.8 We therefore set out to conduct a meta-analysis to evaluate the evidence supporting an association between prolonged PR interval and AF incidence. Methods We followed the Meta-analysis of Observational Studies in Epidemiology protocol throughout the design, implementation, analysis, and reporting for this study9 (see Supplemental Appendix S1). Literature Search Strategy We performed a literature search of MedLine and EMBASE using the key words “PR interval,” “atrioventricular block,” “arrhythmia,” and “atrial fibrillation” from inception to May 2014. We reviewed all articles with an abstract suggesting relevance. We checked the reference lists of relevant articles to find other eligible trials. Prospective cohort studies were included. Supplemental Appendix S2 provides a detailed description of the search strategies.
Cheng et al.
PR Interval and Atrial Fibrillation
37
Figure 1. Flow diagram of the study selection process.
Data Extraction and Quality Evaluation The following details were recorded for each study: author, year of publication, cohort study name, the geographic location of study, study period, participants’ sex, age range at baseline, health at baseline, and outcome (defined as incidence of AF/flutter). For each study, we obtained information on the most adjusted hazard ratio (HR) for prolonged PR interval with corresponding 95% confidence intervals (CIs) or P value. Data abstraction was conducted independently by 2 investigators, with disagreements resolved by a third investigator. The Newcastle-Ottawa Quality Assessment Scale was used to assess bias in included studies. Statistical Analysis In this meta-analysis, the adjusted HR with 95% CIs was considered as the effect size for all studies. A randomeffects meta-analysis was performed to estimate pooled HRs weighted by the inverse variance method considering different measurement methods and reference range of PR interval. The heterogeneity among studies was estimated using the I2 statistic. Heterogeneity was confirmed with a significance level of P < 0.10. Meta analysis with fixed-effects model was conducted for sensitivity analysis. Publication bias was assessed using the Begg-adjusted rank correlation test. All statistical analysis was performed using Stata software (Stata 12.0; Stata Corp, College Station, TX, USA). All reported probabilities were 2-sided with P < 0.05 considered statistically significant. Results Figure 1 shows the results of literature research and selection. There were 2,799 citations retrieved on initial research. After removal of 219 replicates, 2,553 citations were excluded at the title or abstract level. Full-text assessment of 27 articles resulted in 6 prospective cohort studies entering into final meta-analysis. Description of Studies Included Table 1 shows the information extracted from the included studies. These studies included 328,932 participants in total. During a mean follow-up ranging from 5.7 to 30 years, 14,191
participants suffered from AF. Three studies were conducted in the United States7,10,11 ; the remaining studies were conducted in Australia,12 Finland,8 or Denmark.13 Participants’ mean age ranged from 44 to 74 years at baseline. Female percentage ranged from 48% in the Aro study8 to 56% in the Knuiman study.12 Taking of β-blockers or calcium antagonists on the day of ECG recording was 27% in the Nielsen study.13 Prescription of chronotropic medication was 4.4% in the Aro study.8 Selected medication including amiodarone, cardiac glycosides, calcium channel blockers and β-blockers was used in 41.7% participants in the Magnani study.10 The number of persons on AV node blocking agents was not mentioned in remaining studies. The reporting quality of the included studies was globally acceptable (see Supplemental Appendix S2). There was a concern of selection bias in the Nielsen study, which enrolled people with an electrocardiography (ECG) recorded at Copenhagen General Practitioners’ Laboratory. AF incidence in the population was higher compared with the general population.13 There was also a concern of potential selection bias in the Magnani study, which enrolled people aged 70–79 years old.10 AF incidence was documented using ECG in all studies. Method of PR interval measurements was different across included studies. Lead II was used for PR measurement in the Cheng7 and Magnani studies.10 The distance between the earliest detection of atrial and ventricular depolarization in any lead was used for PR measurement in the Nielsen study.13 Bipolar limb lead in which PR interval was longest was used in the Aro study.8 The mean P-wave duration plus the mean PR-segment duration in the 12-lead ECG was used in the Soliman study.11 The method of PR measurement was not specified in the Knuiman study.12 Effect size was estimated for prolonged PR interval (PR interval more than 200 milliseconds) with comparison with a “normal” reference PR interval that was defined in individual studies differently. PR interval less than 200 milliseconds was defined as “normal” reference in the Cheng,7 Aro,8 Soliman11 and Magnani10 (PR interval < 80 milliseconds excluded) studies. The “normal” range was defined as PR interval less than 200 milliseconds and more than 120 milliseconds in the Knuiman study.12 In the Nielsen study,13 PR interval ranged from 150 to 161 milliseconds was
10,758
Finland, 30 years 55%
48%
56%
55%
54
43
52
54
73
46
Mean Age (years)
Limb lead in which PR is longest The mean P-wave duration plus the mean PR-segment duration
The distance between the earliest atrial and ventricular depolarization NS
Lead II
Lead II
PR Measurement Method
1.59 (0.77–3.30)
1.03 (0.74–1.45)
1.29 (0.68–2.44)
1.26 (1.17–1.35)
1.26 (0.99–1.61
2.06 (1.36–3.12)
HR (95% CI)
Sex, age, height, HTN treatment, BMI Sex, age, chronotropic medication, cardiovascular disease, BMI, HF Sex, age, ethnicity, HTN, SBP, diabetes, blood lipids, smoking, BMI
PR ࣘ 200 milliseconds PR ࣘ 200 milliseconds
Sex, age, cardiovascular disease status, HR, BMI, HTN, smoking, diabetes, total/HDL cholesterol, atrial premature beats, valve disease, ECG LVH Sex, age, cardiovascular disease, selected medications (amiodarone, cardiac glycosides, calcium channel blockers, β-blockers), BMI, HF, SBP, DBP, site, smoking, total/HDL, ECG LVH, hypertension treatment Sex, age, HTN, HF, MI, valvular heart disease, treatment with AV nodal-blocking medication, diabetes, hyperthyroidism, HR, QT interval, ECG LVH
Variables Adjusting for Effect Estimates
Normal PR
150 milliseconds < PR ࣘ 161 milliseconds
80 milliseconds < PR ࣘ 200 milliseconds
PR ࣘ200 milliseconds
Defined Reference PR
HR = heart rate; BMI = body mass index; HDL = high-density lipoprotein; HTN = hypertension; ECG = electrocardiography; LVH = left ventricular hypertrophy; MI = myocardial infarction; HF = heart failure; SBP = systolic blood pressure; NS = not stated.
15,429
4,267
Australia, 15 years
Soliman, 2009, ARIC Study United States, 6.97 years
Knuiman, 2013, the Busselton Health Study Aro, 2014, CHD Study
288,181
52%
54%
Sex, Female (%)
Journal of Cardiovascular Electrophysiology
Nielsen, 2013, Copenhagen Denmark, 5.7 ECG Study years
2,722
Magnani, 2013, Health ABC United States, 10 years
Total Number (n) 7,575
Location and Duration
United States, 12 years
Cheng, 2009, Framingham Heart Study
Author, Publication year, Cohort
TABLE 1 Characteristics of Cohort Studies of Prolonged PR Interval and Atrial Fibrillation Included in the Meta-Analysis with Multivariable-Adjusted Estimates
38 Vol. 26, No. 1, January 2015
Cheng et al.
PR Interval and Atrial Fibrillation
39
Figure 2. Combined HR of prolonged PR interval for AF incidence using random-effects model. HR = hazard ratio; AF = atrial fibrillation.
Figure 3. Pooled HR in subgroup analysis according to the adjustment for AV block agents. HR = hazard ratio.
used as reference. Confounding factors adjusted were different across studies. Age, sex, and hypertension (treatment) or blood pressure were adjusted for effect size estimates in all studies, but taking of AV nodal-blocking medication was adjusted only in the Magnani,10 Aro,8 and Nielsen13 studies. Heart rate was adjusted in the Nielsen13 and Cheng7 studies. BMI was not adjusted in the Nielsen study.13 ECG LVH was adjusted in the Nielsen,13 Magnani,10 and Cheng7 studies. PR Prolongation and Risk of AF In primary analysis, the pooled HR for AF incidence of prolonged PR interval was 1.30 (95% CI: 1.13 to 1.49; P = 0.000) using random-effects model. Heterogeneity was
moderate between studies (I2 = 30% and P = 0.21) (see Fig. 2). Pooled HRs were different in subgroup analysis according to confounding factors adjusted (P = 0.039). Pooled HR from studies adjusting for taking of AV nodal blocking agents was 1.25 (95% CI, 1.17–1.34; I2 = 0%) for AF incidence, while pooled HR was 1.75 (95% CI: 1.28 to 2.40, I2 = 0%) from studies without taking of AV nodal blocking agents adjustment (see Fig. 3). Sensitivity Analysis and Publication Bias Sensitivity analyses using fixed-effects model supported the robustness of the primary results. The Begg-adjusted rank
40
Journal of Cardiovascular Electrophysiology
Vol. 26, No. 1, January 2015
correlation test provided no evidence of substantial publication bias in the primary meta-analysis (P > 0.05). Because we included a small number of studies, formal assessment of publication bias might not be appropriate. Discussion Results of this meta-analysis demonstrate that electrocardiographic PR interval prolongation is a risk factor for AF incidence independent of age, sex, and hypertension. Although the notion that electrocardiographic PR interval prolongation is a benign condition has been challenged by the Framingham Heart Study in which prolonged PR interval was associated with more than twofold increases in AF incidence,7 results from a recent large cohort study are still supportive. In the Finnish Social Insurance Institution’s Coronary Heart Disease Study, prolonged PR interval was normalized in almost 30% of the subjects during 30 years follow-up.8 There are no definite explanations for the difference of results across studies. It is speculated that varying underlying substrate for PR interval prolongation may account for the difference. For example, enhanced vagal tone is assumed to be a major contributor in the young, and the prolonged PR interval may normalize with aging. In contrast, structural remodeling such as atrial fibrosis may play a more important role in the development of PR prolongation in the old, and it is well known that AF risk is proportional to atrial structural remodeling. In this case, prolonged PR interval may just be an indirect marker of atrial structural remodeling, which is a substrate for AF. This speculation was supported by the results of Soliman’s study, which demonstrated that associations of PR interval with adverse outcomes were dependent on the level of P-wave duration contribution to the overall length of PR interval, a contribution that varied across populations.14 In this meta-analysis with strict inclusion criteria, we found that PR prolongation was significantly associated with increased risk of AF. Furthermore, the moderate heterogeneity between included studies was mainly due to the differential adjustment of confounding factors. Pooled HRs were significantly different between studies with and without the taking of AV block agents adjustment. Since AV nodal blocking agents are frequently used in patients with high AF risk, it is not surprising to find out that the taking of AV blocking agents is a significant confounding factor for the effect estimates of prolonged PR interval for AF risk. There are several potential explanations for the observed association of prolonged PR interval with AF incidence. First, prolonged PR interval may share genetic background with AF. The heritability estimate of PR interval ranged from 34% to 78%.15-17 Genome wide association studies have found that chromosome loci from voltage gated sodium channel and cardiac developmental genes are associated with PR interval and many also associated with AF.18-24 In addition, AF associated genetic variant rs 220073 in chromosome 4q25 has been found associated with PR interval, further supporting the notion of a shared genetic background for both PR prolongation and AF incidence.25,26 Second, PR interval duration is strongly influenced by the autonomic nervous system, and PR prolongation may be a marker of abnormal cardiac autonomic abnormality.27,28 It is well known that both parasympathetic and sympathetic tone, or more likely interplay between the two, play an important role in the genesis
of AF.29-31 Third, PR prolongation is a marker of atrial electrical and structural remodeling and reflective of increased atrial conduction time that may directly increase the risk of AF. AV dyssynchrony may also play an important role in the association between PR prolongation and risk of AF. It is, however, still unclear what is the most important contributor to the observed association of prolonged PR with AF, and further research needed. The results of this meta-analysis are meaningful for the battle against AF. First, with inclusion of more than 300,000 participants, this meta analysis robustly supports the predictive role of prolonged PR interval for AF incidence. Second, research on prolonged PR interval will improve our understanding of AF pathophysiology. Third, 30% increases in AF risk associated with prolonged PR interval suggest a potential management benefit of PR prolongation for AF prevention. At present, there are few effective approaches to directly normalize prolonged PR interval except for cardiac pacing, and reversal of atrial remodeling that may contribute to P-wave duration might be a biologically plausible means of reducing AF. However, it is worth mentioning that though normalization of prolonged PR interval may reduce AF risk, there is no evidence to support the assumption that the shortening of normal PR interval (defined as PR interval
