|Home | About | Journals | Submit | Contact Us | Français|
The influence of systolic and diastolic blood pressure (BP) on incident atrial fibrillation (AF) is not well studied among initially healthy, middle-aged women.
34221 women participating in the Women’s Health Study were prospectively followed for incident AF. The risk of AF across categories of systolic and diastolic BP was compared using Cox proportional-hazards models. During 12.4 years of follow-up, 644 incident AF events occurred. Using BP measurements at baseline, the long-term risk of AF was significantly increased across categories of systolic BP and diastolic BP. Multivariable adjusted hazard ratios (HR) (95% confidence interval (CI)) for systolic BP categories (<120, 120–129, 130–139, 140–159 and ≥160 mmHg) were 1.0, 1.00 (0.78–1.28), 1.28 (1.00–1.63), 1.56 (1.22–2.01) and 2.74 (1.77–4.22) (p for trend <0.0001). Adjusted HRs (95% CI) across baseline diastolic BP categories (<65, 65–74, 75–84, 85–89, 90–94, ≥95 mmHg) were 1.0, 1.17 (0.81–1.69), 1.18 (0.84–1.65), 1.53 (1.05–2.23), 1.35 (0.82–2.22) and 2.15 (1.21–3.84) (p for trend 0.004). When BP changes over time were accounted for in updated models, multivariable adjusted HRs (95% CI) were 1.0, 1.14 (0.89– 1.46), 1.37 (1.07–1.76), 1.71 (1.33–2.21) and 2.21 (1.45–3.36) (p for trend <0.0001) for systolic BP categories, and 1.0, 1.12 (0.82–1.52), 1.13 (0.83–1.52), 1.30 (0.89–1.88), 1.50 (1.01–1.88) and 1.54 (0.75–3.14) (p for trend 0.026) for diastolic BP categories.
In this large cohort of initially healthy women, BP was strongly associated with incident AF, and systolic BP was a better predictor than diastolic BP. Systolic BP levels within the non-hypertensive range were independently associated with incident AF even after taking into account BP changes over time.
Atrial fibrillation (AF) is the most common cardiac arrhythmia and its prevalence in the general population is increasing rapidly.1–3 The importance of AF as a public health problem is further underscored through its association with stroke, heart failure, death, cognitive dysfunction and a reduced quality of life.4–8 Because treatment of established AF has limited long-term success rates and significant risks,9, 10 characterizing treatable risk factors for AF has substantial clinical relevance.
Several studies have shown that arterial hypertension is an independent risk factor for incident AF,11–13 but optimal blood pressure (BP) levels for AF prevention have not been established. In addition, few studies have assessed more in detail the association between different BP components or the change in these components over time and subsequent risk of incident AF. Given the non-linear relationship between age and diastolic BP but not systolic BP,14 alterations in systolic or diastolic BP may also differentially affect the risk of incident AF. Along these lines, a recent study suggested that pulse pressure may be more predictive for AF than either systolic or diastolic BP alone.15 However, these relationships may differ in younger individuals or women, where the prevalence, outcome and underlying comorbidities associated with AF differ.4, 5, 11, 16
To address these issues, we assessed the relationship between systolic and diastolic BP levels and the risk of incident AF in a large cohort of middle-aged women who were free of cardiovascular disease at baseline.
All study subjects were participants of the Women’s Health Study, a completed randomized trial evaluating the risks and benefits of low dose aspirin and vitamin E in the primary prevention of cardiovascular disease and cancer. Details of the study design have been described previously.17–19
Briefly, beginning in 1993, 39876 female health professionals in the United States who were 45 years or older and free of cardiovascular disease, cancer or other major illnesses were randomized to receive 100 mg aspirin every other day, 600 IU vitamin E every other day, both agents or placebo. The trial initially had a beta carotene arm that was terminated early.20 Randomized treatment ended on March 31, 2004, and women were invited to participate in continued observational follow-up, which for the current study was truncated on October 31, 2006. Of the original cohort, 4320 opted out of the observational follow-up. These women were excluded from this analysis because their AF could not be reliably confirmed. However, very similar results were obtained when we repeated our analyses using self-reported AF events among all women as the main outcome variable (data not shown).
We also excluded 813 women with a history of AF at baseline and 465 women with missing information on baseline systolic or diastolic BP. The final study population for the present analysis consisted of 34221 women with a median (interquartile range) follow-up of 12.4 (11.8–12.8) years. Written informed consent was obtained from all participants. The study was approved by the institutional review board of Brigham and Women’s Hospital, Boston, and was monitored by an external data and safety monitoring board.
Information on baseline variables was collected using mailed questionnaires. Follow-up questionnaires asking participants about study outcomes and other information were sent every six months during the first year and every 12 months thereafter. Systolic and diastolic BP were self-reported at randomization, and then again at 12, 48, 120 and 132 months of follow-up. At randomization and at 12 months, women were asked to categorize their BP levels into nine different categories for systolic and seven different categories for diastolic BP. Beginning at 48 months, women were asked to directly report their current systolic and diastolic BP. Such self-reported BP measurements among female health professionals have proven to be highly accurate in prior studies.21–23
Other covariates of interest that were assessed at study entry and at various points of follow-up included age, smoking, diabetes, history of hypercholesterolemia (self-reported cholesterol of at least 240 mg/dl (6.22 mmol/l)), body mass index (weight in kilograms divided by the square of height in meters), exercise, alcohol consumption and highest education level achieved.
Women were asked to report diagnoses of incident AF at baseline, 48 months, and then annually thereafter. Beginning on September 19, 2006, women enrolled in the continued observational follow-up who reported an incident AF event on at least one yearly questionnaire were sent an additional questionnaire to confirm the episode and collect additional information. They were also asked for permission to review their medical records, particularly available electrocardiograms, rhythm strips, 24-hour electrocardiograms and information on cardiac structure and function. For all deceased participants who reported AF during the trial and extended follow-up period, we contacted family members to obtain consent and additional relevant information. An endpoint committee of physicians reviewed medical records for reported events according to predefined criteria. An incident AF event was confirmed if there was electrocardiographic evidence of AF or if a medical report clearly indicated a personal history of AF. The earliest date in the medical records when documentation was believed to have occurred was set as the date of onset of AF. Only confirmed events are included in the present report.
The primary analysis examined the relationship between self-reported BP at baseline and the subsequent development of AF over the course of the study. Continuous BP values were assigned to each participant by taking the midpoint value for their reported BP category. Participants with systolic BP <110 or ≥180 mmHg were assigned the values 95 or 190 mmHg, respectively; those with diastolic BP <65 or ≥105 mmHg were assigned the values 55 or 110 mmHg, respectively. Because of the high correlation between baseline and one year BP in this cohort,22 the average of the two measurements was used as baseline BP, in order to reduce BP variability and potential reporting errors. Women were then classified in the following pre-specified clinically relevant baseline BP categories: <120 mmHg, 120–129 mmHg, 130–139 mmHg, 140–159 mmHg and ≥160 mmHg for systolic BP, and <65 mmHg, 65–74 mmHg, 75–84 mmHg, 85–89 mmHg, 90–94 mmHg and ≥95 mmHg for diastolic BP.
Subsequently, Cox proportional-hazards models were constructed to calculate hazard ratios and 95 percent confidence intervals across BP categories. Multivariable models were adjusted for age, body mass index, history of diabetes, smoking, history of hypercholesterolemia, exercise, alcohol consumption, education and randomized treatment assignments (aspirin, vitamin E, beta carotene). In a first step, we performed separate analyses for systolic and diastolic BP. We then assessed the joint contribution of systolic and diastolic BP by including both BP variables in the same models. The –2 log likelihood was used to compare the fit of nested BP models.
Since the rounding of BP into categories may differentially influence the strength of the association between systolic or diastolic BP and incident AF, we performed a sensitivity analysis examining the relationship between the directly reported BP values obtained at 48 months and subsequent risk of AF. Also, since the majority of participants indicated a baseline diastolic BP between 75 and 84 mmHg (Table 1), we performed an additional secondary analysis that subdivided the range between 75–84 mmHg into two equally spaced categories at 48 months (80–84 mmHg and 85–89 mmHg).
To incorporate the effect of BP changes over time on the risk of incident AF, we utilized a modified Kaplan-Meier method that estimated cumulative probabilities of incident AF according to the most recent BP level using time-updated BP categories. 24 We then constructed multivariable adjusted Cox models where BP categories were similarly updated at 12, 48, 120 and 132 months of follow-up, and the most recent BP measurement prior to the event was used to estimate risk. For example, the 12-month BP measurement was used for the 12 month through 48-month follow-up period, the 48-month measurement was used for the 48-month to 120-month follow-up period, and so on. Similarly, the effect of antihypertensive treatment over the course of the study was assessed by including an indicator for antihypertensive treatment as a time-varying covariate in these models. We obtained a rough estimate of the potential benefit of lowering systolic BP <120 mmHg among women with higher BP levels by using the following formula: (1-(1/HR)) × 100.
An association between BP and incident AF may be caused by intercurrent cardiovascular events, given the strong relationship between BP and cardiovascular disease.22, 23 We therefore refitted all Cox models after censoring women at the date of their first cardiovascular event. A cardiovascular event was defined as myocardial infarction, stroke or coronary revascularization.19 Finally, we also performed analyses excluding 4621 (13.5%) women who took antihypertensive treatment at baseline.
In all regression models, categorical variables were entered using binary indicator variables, and tests for linear trend were performed using integer scores across categories. We tested for deviation from linearity by including a quadratic term in the models. Effect modification was assessed using multiplicative interaction terms. The proportional hazards assumption was examined for all models by including a BP category by logarithm of time interaction into the model.25 No violation of this assumption was detected. All analyzes were carried out using SAS version 9 (SAS Institute Inc, Cary, NC). A two-tailed p value <0.05 was considered to indicate statistical significance. The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
During a median follow-up of 12.4 years, 644 out of 34221 women had at least one confirmed episode of incident AF. Baseline characteristics of the study population are shown in Table 1. Mean age was 55 ± 7 years, 16.8% of participants had a systolic BP ≥140 mmHg and 4.8% had a diastolic BP ≥90 mmHg.
There was a highly significant increase in risk of incident AF across increasing categories of systolic and diastolic BP. While multivariable adjustment somewhat attenuated these trends, both BP components remained strongly associated with incident AF (Table 2). Even women with systolic BP between 130–139 mmHg or diastolic BP between 85–89 mmHg, i.e. below the current threshold for diagnosing and treating hypertension,26, 27 had a significantly increased risk of incident AF (hazard ratio (95% confidence interval) 1.28 (1.00– 1.63), p=0.05, and 1.53 (1.05–2.23), p=0.03, respectively). When both BP components were entered in a joint multivariable model, we found a persistent risk gradient across systolic but not diastolic BP categories. Accordingly, adding diastolic BP to the multivariable systolic BP model did not improve model fit (likelihood ratio chi square 2.23, p=0.82).
When we assessed baseline systolic and diastolic BP as continuous variables, we found an adjusted hazard ratio (95% confidence interval) of 1.16 (1.09–1.23), p<0.0001 per 10-mmHg increase of systolic BP and 1.17 (1.05–1.29), p=0.003 per 10-mmHg increase of diastolic BP. In a combined model systolic but not diastolic BP remained a significant predictor of incident AF (hazard ratio (95% confidence interval) per 10 mmHg increase 1.17 (1.08–1.27), p<0.0001 for systolic BP; and 0.98 (0.86–1.12), p=0.74 for diastolic BP).
When we examined the continuous relationship between directly reported BP measurements ascertained at 48 months and subsequent risk of AF, systolic but not diastolic BP was associated with AF (hazard ratio (95% confidence interval) per 10-mmHg increase 1.12 (1.05–1.19), p=0.001 for systolic, and 1.00 (0.90–1.12), p=0.94, for diastolic BP). In a combined model, systolic BP remained significantly related to incident AF (hazard ratio (95% confidence interval) 1.16 (1.08–1.25, p=0.0001), and an inverse association between diastolic BP and incident AF began to emerge (hazard ratio (95% confidence interval) 0.88 (0.78– 1.00), p=0.05).
Importantly, women with high normal systolic BP (130–139 mmHg) at 48 months continued to be at increased risk of subsequent AF (hazard ratio (95% confidence interval) 1.50 (1.12–2.02), p=0.006). At 48 months, there was an increased risk even among women with systolic BP between 120 and 129 mmHg (hazard ratio (95% confidence interval) 1.35 (1.01–1.82), p=0.04). When we subdivided diastolic BP at 48 months into one additional category (<65, 65 to 74, 75 to 80, 80–84, 85 to 89, 90 to 94 and >94 mmHg), we obtained the following adjusted hazard ratios (95% confidence intervals): 1.0, 0.97 (0.69–1.35), 0.65 (0.43–1.00), 0.92 (0.66–1.28), 0.91 (0.59–1.40), 1.01 (0.65–1.57), 1.38 (0.62–3.07). When diastolic BP categories were entered into the systolic BP model, a U-shaped relationship between diastolic BP and AF emerged (hazard ratio (95% confidence interval) 1.0, 0.84 (0.60–1.18), 0.52 (0.33–0.80), 0.67 (0.46–0.96), 0.62 (0.39–0.99), 0.67 (0.41–1.08), 0.86 (0.37–2.00)), associated with a borderline significant improvement in model fit (likelihood ratio 12.58, 6df, p=0.05). The addition of a quadratic diastolic BP term to this model provided a significant result (p=0.031), again suggesting a non-linear relationship.
In models where BP was updated over time, systolic BP remained a strong predictor of AF and women with systolic BP between 130 and 139 mmHg continued to have significantly elevated risks of AF as compared to women with a systolic BP <120 mmHg (hazard ratio (95% confidence interval) 1.37 (1.07–1.76), p=0.01) (Table 2, Figure 1). These analyses also suggest that lowering systolic BP <120 mmHg in women with systolic BP between 120–129 mmHg, 130–139 mmHg, 140–159 mmHg and ≥160 mmHg may be associated with reductions of incident AF of 12%, 27%, 42% and 55%, respectively.
Diastolic BP was also associated with incident AF in the updated model (p for linear trend 0.026). In an updated combined model, systolic BP remained strongly and positively associated with incident AF, also among women with BP values between 130 and 139 mmHg (hazard ratio (95% confidence interval) 1.43 (1.09–1.87), p=0.009), but the relationship between diastolic BP and AF became non-significant.
Findings were similar when we took into account the effect of antihypertensive treatment during follow-up. In these models, multivariable adjusted hazard ratios (95% confidence intervals) were 1.0, 1.10 (0.86–1.41), 1.28 (0.99–1.64), 1.54 (1.18–2.00) and 1.93 (1.26–2.97) for increasing systolic BP categories, and 1.0, 1.10 (0.81–1.51), 1.08 (0.79–1.46), 1.18 (0.81–1.71), 1.32 (0.88–1.96) and 1.32 (0.64–1.70) for increasing diastolic BP categories. Treatment by systolic or diastolic BP interaction terms were not statistically significant (p=0.80 and p=0.31, respectively).
In total, 47 women had a cardiovascular event prior to the development of new-onset AF. Censoring these women at the date of the cardiovascular event did attenuate but not offset the association between incident AF and baseline BP (Table 2), and linear trends across BP categories remained significant for systolic BP (p for linear trend=0.0001) and diastolic BP (p for linear trend=0.02). Results also did not significantly change if we excluded the 13.5% of women who took antihypertensive treatment at baseline (data not shown). Analysis for effect modification revealed that the risk of AF according to both systolic and diastolic BP was similar for women <65 years and those ≥65 years (p for interaction 0.32 and 0.39 for systolic and diastolic BP, respectively), and similar for those who took antihypertensive treatment at baseline and those who did not (p for interaction 0.24 and 0.75 for systolic and diastolic BP, respectively).
The present study demonstrates that BP is a strong and independent predictor of incident AF in initially healthy, middle-aged women. This study also found that BP values below the current threshold for the diagnosis of arterial hypertension are significantly associated with the risk of incident AF, extending prior studies that have consistently shown a strong and independent relationship between hypertension and incident AF.11–13 Women with high normal systolic (130 to 139 mmHg27) or diastolic (85 to 89 mmHg27) BP at baseline had a 28% and 53% increased risk of incident AF compared to women with systolic BP <120 mmHg or diastolic BP <65 mmHg, respectively. Thus, in our study, even slightly elevated BP levels at baseline imposed some degree of increased risk. These results parallel findings from prior studies that demonstrated continuous relationships between BP and other cardiovascular disease outcomes.22, 23 However, further studies in individuals with low BP levels are needed to confirm the absence of a BP threshold below which the risk of incident AF is not increased.
To our knowledge, this is one of the first studies examining the influence of BP changes over time and subsequent risk of AF. Not surprisingly, we found that women with a systolic BP ≥140 mmHg during follow-up had a significantly increased risk of incident AF compared to those with systolic BP values <140 mmHg. Of note, even women with systolic BP values between 130–139 mmHg during follow-up had a significantly increased risk of subsequent AF. Although no firm recommendations can be made based on these observational findings, they nevertheless suggest that tight BP control may help to reduce the burden of AF among women.
We believe that our study also has other potential implications. First, guidelines for the management of arterial hypertension do not currently classify AF as a risk factor, target organ damage or associated clinical condition.26,27 However, the tight association between BP, incident AF and subsequent cardiovascular events4–8 suggests that future hypertension guidelines may assign a more important role to AF for cardiovascular risk stratification in patients with hypertension. Second, current guidelines do not assign a lower BP treatment target for patients suffering from AF.26, 27 Given the strong, continuous relationship between systolic BP and incident AF found in this study, individuals with AF may also benefit from a lower BP treatment threshold of <140/90 mmHg. Our time-updated analyses suggest that substantial risk reductions may be obtained if systolic BP levels were maintained at or lowered to levels <120 mmHg (Table 2). However, optimal BP targets should ideally be evaluated in a randomized trial.
Prior work indicated that pulse pressure is an important risk factor for AF in a middle-aged to elderly community sample. In this prior study, diastolic BP provided significant additional information when added to a model containing systolic BP in this population.15 Accordingly, the investigators suggested that aortic stiffness might be an important factor in the pathogenesis of AF. In the current study among middle aged women, we found that systolic BP was a stronger predictor of incident AF than diastolic BP. However, additional analyses using directly reported continuous BP measures at 48 months suggested that after taking into account systolic BP, a U-shaped relationship between diastolic BP and incident AF emerged. These models indicated that women with a diastolic BP <65 mmHg had the highest risk of developing AF, providing some evidence that elevated pulse pressure and aortic stiffness may also play a role in this population.
U-shaped relationships for diastolic BP have been described for other cardiovascular disease outcomes, but mainly in the elderly. For example, in elderly individuals with coronary disease there was a J-shaped association between diastolic BP and cardiovascular events.28 Similar relationships have been reported among apparently healthy elderly individuals.29 By contrast, among middle-aged civil servants participating in the Whitehall study, both systolic and diastolic BP were significantly related to the incidence of coronary heart disease, but only the association with systolic BP remained significant in a combined BP model. 30 Further studies are needed to gain more insights in the relationships between different BP components and risk of AF in different population groups.
There are several other potential mechanisms in addition to arterial stiffness that could underlie the relationship between BP and incident AF. For example, elevated systolic BP may be associated with increases in left atrial fibrosis,31, 32 which in turn is related to prevalent AF.33 Left ventricular hypertrophy and increased left atrial size are other important mediators of the relationship between BP and incident AF.34, 35 However, it is important to emphasize that prior studies showed independent BP effects even after adjustment for echocardiographic variables.15
Strengths of the present study include its prospective design, sample size, and long-term follow-up with a large number of confirmed events. Potential study limitations also require discussion. First, the study population consisted of predominantly white, middle-aged female health professionals, and our findings may not be generalizable to men or other populations of women. Second, BP was self-reported. However, the prognostic value of self-reported BP in cohort studies involving US health professionals is similar to directly measured BP values in participants of other cohort studies.23 Furthermore, the validity of this approach has been examined in the Nurses’ Health Study, where 99% of the women who reported high BP had their diagnosis confirmed by medical record review.21 Moreover, in the Women’s Health Study, self-reported BP, total cholesterol and body mass index are strong predictors of cardiovascular risk, with relative risks consistent in magnitude with those observed in other major studies.22, 36–38 However, it is possible that the rounding of baseline BP into categories may have differentially affected the relationship between systolic or diastolic BP and incident AF. Our sensitivity analyses at 48 months suggest that in middle-aged women, there might be a U-shaped association between diastolic BP and incident AF. Further studies are needed to confirm these preliminary findings.
In middle-aged, initially healthy women, BP is a powerful predictor of incident AF, and systolic BP is a better predictor than diastolic BP. Our study also revealed that even participants with BP values currently considered as normal have an increased risk of developing AF. Taken together, our findings indicate that tight BP control may help to reduce the growing burden of AF in the community.
David Conen was supported by the grant PASMA 118586/1 from the Swiss National Science Foundation. The Women’s Health Study was supported by grants HL-043851, HL-080467 and CA-047988 from the National Heart, Lung and Blood Institute and the National Cancer Institute.
Clinical Trial Registration: http://clinicaltrials.gov/ct2/show/NCT00000479 (NCT00000479).
Conflict of Interest Disclosures