Several studies have reported that inflammatory markers are associated with atrial fibrillation (AF). The white blood cell (WBC) count is a widely available and broadly used marker of systemic inflammation. We sought to investigate the association between an increased WBC count and incident AF and whether this association is mediated by smoking, myocardial infarction, and heart failure. We examined the participants in the Framingham Heart Study original cohort. Cox proportional hazard regression analysis was used to examine the relation between the WBC count and incident AF during a 5-year follow-up period. We adjusted for standard AF risk factors, smoking, previous myocardial infarction, and interim myocardial infarction and heart failure before the incident AF. Our sample consisted of 936 participants (mean age 76 ± 6 years and 61% women). The median WBC count was 6.4 × 10 9 /L (25th to 75th percentile 5.6 × 10 9 /L to 7.8 × 10 9 /L). During a median 5-year follow-up period, 82 participants (9%) developed new-onset AF. After adjusting for standard risk factors for AF, an increased WBC count was significantly associated with incident AF, with a hazard ratio per SD (0.26 × 10 9 /L) increase of 2.22 (95% confidence interval 1.10 to 4.48; p = 0.03). We found no substantive differences adjusting for smoking, previous myocardial infarction, interim myocardial infarction, or heart failure. In conclusion, in our community-based sample, an increased WBC count was associated with incident AF during 5 years of follow-up. Our findings provide additional evidence for the relation between systemic inflammation and AF.
The most widely available marker of systemic inflammation is the white blood cell (WBC) count. To date, the relations between the WBC count and atrial fibrillation (AF) have not been extensively explored. An increased WBC count during the perioperative period was predictive of postoperative AF, and a small observational study observed that an increased WBC count was associated with recurrence of AF after pulmonary vein isolation. However, whether an increased WBC count is associated with AF in the general population is unclear. We hypothesized that increased WBC count is related to incident AF in the community. We further postulated that the relation between the WBC count and incident AF might be attenuated by adjustment for factors associated with inflammation that might serve as potential mediators of an association between the WBC count and AF, including body mass index, smoking, previous myocardial infarction, and interim myocardial infarction and heart failure before incident AF.
We studied the participants from the Framingham Heart Study original cohort. The participants were initially recruited in 1948. For the present analysis, we used data from 1,401 participants attending the 20th biennial examination cycle (1986 to 1990). The participants were excluded for the following reasons: prevalent AF (n = 120), missing WBC count (n = 315), and missing covariate data (body mass index, n = 24; PR interval, n = 3; hypertension treatment, n = 2; and cardiac murmur, n = 1). The characteristics of the included versus excluded participants attending the 20th examination cycle are reported in Supplementary Table 1 . The participants were monitored until the first AF event with a maximum follow-up duration of 5 years. In secondary analyses, the follow-up duration was increased to 10 years. The Boston University Medical Center institutional review board approved the protocol of the Framingham Heart Study, and all participants provided written informed consent.
The body mass index was calculated as the weight in kilograms divided by the height in meters squared. The ascertainment of antihypertensive treatment was by self-report. Current smoking was defined as the self-reported regular use of cigarettes in the year preceding the examination. A clinically significant precordial murmur was diagnosed when a physician auscultated a systolic murmur of at least grade 3 of 6 intensity or any diastolic murmur. The subjects were diagnosed with heart failure using standard major and minor clinical criteria.
Nonfasting blood samples were obtained the morning of the Framingham Heart Study clinic visit and frozen at −80°C. The plasma WBC count was measured using the Serono-Baker System 9000 Hematology Analyzer (Serono-Baker Diagnostics, Allentown, Pennsylvania). The normal range of WBC count using this analyzer was 3.8 to 12.4 × 10 9 /L.
AF was considered present if either atrial flutter or AF was present on an electrocardiogram, Holter monitor, or outside clinical records. The data were obtained at Framingham Heart Study clinic visits, on outside clinician visits or during hospital admissions, and a Framingham Study cardiologist reviewed all the records.
In our primary analysis, we tested the hypothesis that a greater WBC count is related to incident AF using Cox proportional hazards regression analysis. We examined robust variance estimators to account for relatedness among Framingham Heart Study participants. We examined the appropriateness of the proportionality assumption by including time-dependent risk factors (i.e., interactions between risk factors and log[time]). The time-dependent risk factors were not statistically significant, supporting the assumption of proportionality of hazards. Because the WBC count was positively skewed, the WBC count was natural logarithmically transformed (log e ). We used 1 SD of the log e WBC count to quantify the effect size. We examined the association between the WBC count and incident AF in age- and gender-adjusted models and in multivariable models adjusted for established AF risk factors, such as used in a previously published Framingham Heart Study AF risk score. The variables in the AF risk score included the age, gender, body mass index, systolic blood pressure, treatment of hypertension, PR interval, clinically significant cardiac murmur, and heart failure. In the secondary analyses, we also adjusted for smoking, previous myocardial infarction, interim myocardial infarction, and interim heart failure. We performed additional models using the method of Fine and Gray to adjust the risk estimates for the competing risk of death. All analyses were performed using SAS software, version 9.2 (SAS Institute, Cary, North Carolina), and R software, version 2.11.1 (R Foundation for Statistical Computing), and a 2-tailed p value of <0.05 was considered statistically significant.
Our final sample consisted of 936 participants. Their mean age was 76 ± 6 years, and 61% were women. Myocardial infarction was present in 10% of participants, and 10% were current smokers. The median WBC count was 6.4 × 10 9 /L (25th to 75th percentile 5.6 × 10 9 /L to 7.8 × 10 9 /L). Sample characteristics are listed in Table 1 . The SD of the log e WBC count was 0.26 × 10 9 /L.
|Variable||Participants (n = 936)|
|Age (years)||76 ± 6|
|Body mass index (kg/m 2 )||26.6 ± 4.6|
|Systolic blood pressure (mm Hg)||147 ± 22|
|Antihypertensive therapy||506 (54%)|
|PR interval duration (ms)||170 ± 33|
|Precordial murmur ⁎||67 (7%)|
|Heart failure||28 (3%)|
|Myocardial infarction||90 (10%)|
|White blood cell count (10 9 /L)|
|Interquartile range †||5.6–7.8|
|Log e white blood cell count||1.88 ± 0.26|
During a median 5-year follow-up period, 82 participants (9%) developed incident AF. After adjusting for standard risk factors for AF, the log e WBC count was significantly associated with incident AF, with a hazard ratio of 2.22 per SD increase (95% confidence interval 1.10 to 4.48; p = 0.03; Table 2 ). In Figure 1 , the cumulative incidence of AF by tertiles of the WBC count distribution is shown.
|Variable||Log e WBC Count|
|Hazard Ratio ⁎ (95% Confidence Interval)||p Value|
|Age and gender adjusted||2.78 (1.36–5.66)||0.005|
|Standard atrial fibrillation risk factors †||2.22 (1.10–4.48)||0.03|
|Standard atrial fibrillation risk factors plus smoking||2.33 (1.14–4.75)||0.02|
|Standard atrial fibrillation risk factors plus previous myocardial infarction||2.21 (1.10–4.45)||0.03|
|Standard atrial fibrillation risk factors plus interim myocardial infarction ‡||2.16 (1.07, 4.35)||0.03|
|Standard atrial fibrillation risk factors plus interim heart failure ‡||2.16 (1.07, 4.39)||0.03|
|Standard atrial fibrillation risk factors plus interim myocardial infarction plus heart failure ‡||2.16 (1.07, 4.35)||0.03|
To test whether smoking, previous myocardial infarction, and myocardial infarction and heart failure occurring during follow-up before incident AF were mediatory factors by which the relation between WBC count and AF was (partly) explained, we also adjusted for these factors ( Table 2 ). However, neither the baseline covariates (smoking and previous myocardial infarction) nor the interim covariates (myocardial infarction and heart failure during follow-up) substantively attenuated the association between the WBC count and incident AF. No significant interactions between the WBC count and smoking (p = 0.34), heart failure (p = 0.34), or myocardial infarction (p = 0.34) were observed.
During a 5-year follow-up period, 140 participants (15%) died without having incident AF. To address a potential bias, we computed additional models accounting for the potential competing risk of death. We found comparable risk estimates using this method; the results of the competing risk models are listed in Supplementary Table 2 .
In the secondary analyses, we expanded the follow-up duration beyond the prespecified 5 years of follow-up to investigate whether the found WBC-AF relation remained present for longer follow-up durations. In Supplementary Table 2 , we report the results of the analyses accounting for the competing risk of death during 10 years of follow-up. The observed hazards ratio during 10 years of follow-up was similar to the hazard ratio observed during 5 years of follow-up; however, the WBC-AF relation was not statistically significant.