Impact of alcohol habits and smoking on the risk of new-onset atrial fibrillation in hypertensive patients with ECG left ventricular hypertrophy: The LIFE Study

Abstract

Background. The incidence of new-onset atrial fibrillation (AF) is increased by uncontrolled hypertension, and antihypertensive treatment reduces new-onset AF. However, it is unclear whether alcohol intake and smoking influence the risk of new-onset AF during antihypertensive treatment. Methods. In the Losartan Intervention For Endpoint reduction in Hypertension (LIFE) study, a double-blinded, randomized, parallel-group study, 9193 hypertensive patients with electrocardiogram (ECG)-documented left ventricular hypertrophy (LVH), randomized to once-daily losartan- or atenolol-based antihypertensive therapy were followed for a mean of 4.8 years. At baseline, 8831 patients (54% women, mean age 67 years, mean blood pressure 174/98 mmHg after placebo run-in) had neither a history of AF nor AF on ECG, and they were thus at risk of developing this condition during the study. Results. New-onset AF occurred in 353 (4%) patients. Univariate Cox analyses showed that intake of alcohol > 10 units/week compared with less or no alcohol intake predicted new-onset AF (Hazard ratio, HR = 1.60 [95% CI 1.02–2.51], p = 0.043). Multivariate Cox regression analysis showed that intake of alcohol > 10 units/week predicted new-onset AF (p = 0.010) independently of most other univariate predictors, except when also baseline serum cholesterol, serum potassium and urinary albumin/creatinine ratio were included in the model (HR = 1.60 [95% CI 0.94–2.72], p = 0.081). Impact of smoking was not significant in Cox univariate or multivariate analyses, and there were no significant interactions between high alcohol intake and either smoking or gender on the risk of getting AF. Conclusions. Up to 10 drinks of alcohol per week appears to be safe with respect to the risk for AF in hypertensive patients with LVH. Our data suggest that alcohol intake above this level may be marginally deleterious, while no effect of smoking on risk of AF was detected in hypertensive patients with LVH.

Key Words::

• arrhythmia
• Cornell voltage–duration product
• ethanol
• Sokolow–Lyon

Introduction

Several studies report a U-shaped association between alcohol intake and mortality, with higher risk of death for abstainers and those reporting an alcohol intake above 18 standard drinks (12 g alcohol per drink) per week (1). In the “holiday heart syndrome”, acute alcohol intake is also associated with cardiac arrhythmias (2). Subjects with long-term moderate alcohol consumption and those who do not drink have a similar risk of atrial fibrillation (AF) (3,4). However, the Framingham study and other studies in men suggested a threshold effect with significantly increased risk of AF among drinkers of more than 21–35 standard drinks per week (3,5,6). A study in women reported a lower threshold for AF; women with an intake of 18 or more standard drinks per week had an increased risk of new-onset AF (7). In addition, a study of elderly people found no increased risk of AF with alcohol intake; conversely, ex-drinkers, who quit drinking during the 9 years of follow-up, had an increased risk of new-onset AF compared with long-term abstainers (4). Recently a meta-analysis indicated that the risk of AF increased with increasing levels of alcohol consumption (8).

The incidence of new-onset AF is increased in hypertensive people and even more if left ventricular hypertrophy (LVH) has developed (9,10). Thus, the hypertensive patients with electrocardiographic (ECG) LVH enrolled in Losartan Intervention For Endpoint reduction in Hypertension (LIFE) were prone to develop more AF than normal subjects, and hence in them, provoking or protecting lifestyle factors might be more distinctly revealed. In addition, and important in this respect, alcohol consumption in the LIFE study population was associated with a graded lower risk of myocardial infarction, whereas the risk of stroke increased sharply with high intake (11). However, it remains unclear how alcohol intake influences the risk of new-onset AF during antihypertensive treatment, a relation previously not reported in patients with LVH. Thus, we aimed to investigate the influence of alcohol intake, and simultaneously smoking habits, on the risk of new-onset AF during antihypertensive treatment. We hypothesized that high alcohol intake predicts new-onset AF in hypertensive patients with ECG-LVH.

Methods

Design and participants

The LIFE study was a double-blinded, randomized, parallel-group design in 9193 essential hypertensive patients with ECG-LVH who were treated with once-daily losartan- or atenolol-based antihypertensive therapy to compare the effect of the two treatments on cardiovascular morbidity and mortality. To be included, patients had to have a mean trough sitting diastolic blood pressure of 95–115 mmHg and/or a mean sitting systolic blood pressure of 160–200 mmHg after 1–2 weeks on single-blind placebo treatment. The age range for inclusion was 55–80 years. Alcohol abuse, suspected alcohol abuse or psychiatric disorders were exclusion criteria at investigators’ discretion (12). The study took place in the Nordic countries (70% of participants), the UK and the USA during 1995–2001. The rationale, objectives, design and methods have been described previously (12), as have the baseline characteristics (13). The blinded treatment period continued until 1040 patients had experienced a primary endpoint, defined as cardiovascular death, non-fatal myocardial infarction or non-fatal cerebral stroke.

New-onset AF

New-onset AF was identified from annual in-study ECGs that underwent Minnesota coding at a single ECG core center (12). At baseline, 8831 patients had neither a history of AF nor AF on ECG (n = 342), or were missing baseline ECG-data (n = 20), and were thus at risk of developing new-onset AF during the study. The mean age at inclusion was 66.8 years, 54% were female and the mean baseline blood pressure was 174.3/97.9 mmHg. The patients were followed for 4.8 ± 0.9 years.

Reported alcohol intake

Information on weekly smoking habits and alcohol intake was collected at baseline by the study investigators actively asking the participants (12). Alcohol consumption was quantified as number of national standard drinks and categorized as none, 1–4, 5–7, 8–10 and > 10 standard drinks of alcohol per week. Smoking habits were categorized as never smoker, ex-smoker and active smoker, also with the average number of cigarettes per day (14).

Statistical methods

Analysis was done using SPSS ver. 17.0 (Chicago, IL). We used Cox proportional hazard analysis to test the effect of alcohol intake on risk of new-onset AF. The group with the highest baseline alcohol intake (> 10 alcohol drinks/week) was compared with the composite group of less or no alcohol intake-categories. Multivariate models were created based on previously described predictors of AF in the LIFE study and included treatment allocation and independent baseline and in-treatment clinical variables such as heart rate, blood pressure and Cornell product (15,16). Finally smoking, which is closely related to alcohol consumption, was included in multivariate analyses. We tested for interaction between alcohol intake and the effects of smoking, gender and treatment allocation; p-values < 0.05 were considered significant.

Results

New-onset AF was confirmed by ECG in 353 patients, whereas 8378 patients never developed AF on ECG. The group with new-onset AF was significantly older, had higher baseline systolic blood pressure, Cornell voltage–duration product, Framingham risk score, urine albumin/creatinine ratio and frequency of Caucasian ethnicity, in addition to lower baseline diastolic blood pressure, serum potassium and frequency of treatment allocation to losartan compared with the patients who never developed AF (Table I) (15,16).

Table I. Baseline characteristics in patients who did and did not develop atrial fibrillation.

	All patients
	New-onset AF (n = 353)	Sinus rhythm (n = 8478)	p-value
Male gender, n (%)	178 (50)	3844 (45)	0.060
Age (years)	71 ± 6	67 ± 7	< 0.001
Caucasian ethnicity, n (%)	343 (98)	7818 (94)	0.001
BMI (kg/m^2)	28.3 ± 5.0	28.0 ± 4.8	0.235
Alcohol intake > 10 drinks/week, n (%)	20 (6)	331 (4)	0.097
Current smoker, n (%)	48 (14)	1398 (17)	0.155
No exercise, n (%)	86 (24)	1845 (22)	0.247
Systolic blood pressure (mmHg)	177 ± 14	174 ± 14	< 0.001
Diastolic blood pressure (mmHg)	97 ± 8	98 ± 9	0.031
Heart rate (beats/min)	73 ± 11	74 ± 11	0.121
Sokolow–Lyon voltage (mV)	30.6 ± 11.3	29.9 ± 10.3	0.276
Cornell product (mV × ms)	2986 ± 962	2808 ± 1018	0.001
Potassium (mmol/l)	4.1 ± 0.4	4.2 ± 0.4	0.029
Log UACR	0.3 ± 0.7	0.2 ± 0.7	0.001
Total cholesterol (mmol/l)	6.0 ± 1.1	6.1 ± 1.12	0.134
HDL-cholesterol (mmol/l)	1.5 ± 0.5	1.5 ± 0.4	0.503
Diabetes, n (%)	47 (13)	1059 (13)	0.647
History of coronary disease, n (%)	54 (15)	1106 (13)	0.220
Framingham risk score, %	24 ± 9	22 ± 9	< 0.001
Treatment allocation to losartan, n (%)	143 (41)	4296 (51)	< 0.001

AF, atrial fibrillation; BMI, body mass index; Log UACR, logarithm of urine albumin/creatinine ratio; HDL, high-density lipoprotein.

When dichotomizing high alcohol intake (> 10 drinks/week) versus lower or no alcohol intake, high intake of alcohol predicted new-onset AF in univariate Cox regression analysis (HR = 1.60 [95% CI 1.02–2.51], p = 0.043) (Table II). As previously described (15,16), additional univariate predictors of incident AF in the LIFE study population were higher heart rate and Cornell voltage–duration product over time, baseline systolic blood pressure, diastolic blood pressure, Cornell voltage–duration product, serum total cholesterol, urine albumin/creatinine ratio, age, male gender, Caucasian ethnicity and Framingham risk score (15,16). Body mass index, diabetes and history of coronary heart disease did not predict new-onset AF whereas allocation to losartan predicted less AF.

Table II. Univariate potential predictors of new-onset atrial fibrillation.

	All
Variable	HR (95% CI)	p-value
Male gender	1.26 (1.03–1.56)	0.028
Age (per year)	1.09 (1.07–1.11)	< 0.001
Caucasian race, n (%)	3.34 (1.49–7.49)	0.003
Body mass index (per kg/m^2)	1.01 (0.99–1.03)	0.272
Alcohol intake > 10 drinks/week^a	1.60 (1.02–2.51)	0.043
Current smoking	1.23 (0.91–1.67)	0.117
No exercise	1.17 (0.92–1.49)	0.207
History of coronary disease	1.21 (0.91–1.62)	0.197
Framingham risk score (per %)	1.02 (1.01–1.03)	< 0.001
Lower in-treatment systolic blood pressure (per 10 mmHg)	0.94 (0.89–1.00)	0.063
Baseline systolic blood pressure (per 10 mmHg)	1.16 (1.08–1.25)	< 0.001
Lower in-treatment diastolic blood pressure (per 10 mmHg)	0.93 (0.83–1.04)	0.190
Lower baseline diastolic blood pressure (per 10 mmHg)	0.87 (0.77–0.97)	0.013
In-treatment heart rate (beats/min)	1.03 (1.02–1.03)	< 0.001
Lower baseline heart rate (per beat/min)	0.99 (0.98–1.00)	0.134
In-treatment Cornell product (per mV × ms/100)	1.01 (1.01–1.02)	0.002
Baseline Cornell product (per mV × ms/100)	1.01 (1.00–1.02)	0.004
Lower baseline total cholesterol (per mmol/l)	0.90 (0.81–0.99)	0.036
Baseline high-density lipoprotein cholesterol (per mmol/l)	1.06 (0.83–1.36)	0.651
Lower baseline potassium (per mmol/l)	0.74 (0.55–1.01)	0.054
Log baseline UACR (per 1 log unit)	1.35 (1.16–1.59)	< 0.001
Diabetes	1.08 (0.79–1.46)	0.642
Treatment with losartan	0.67 (0.54–0.83)	< 0.001

^aCompared with less or no alcohol per week. HR, hazard ratio; CI, confidence interval; Log UACR, logarithm of urine albumin/creatinine ratio.

In multivariate Cox regression analysis, intake of alcohol > 10 drinks/week independently predicted new-onset AF when adjusting for age, gender, treatment allocation, smoking, in-treatment Cornell product, and baseline and in-treatment systolic blood pressure and heart rate (HR = 1.83 [95% CI 1.16–2.91], p = 0.010, Table III). Neither the Framingham risk score nor Caucasian ethnicity interfered with the effect of high alcohol intake on risk of new-onset AF in the multivariate model. However, addition of baseline serum total cholesterol, serum potassium or log urine albumin/creatinine ratio to the multivariate model weakened the effect of high alcohol consumption on risk of new-onset AF (HR = 1.60 [95% CI 0.94–2.72], p = 0.081, Table IV). Including concomitant treatment with diuretic (maximal dose of hydrochlorothiazide) did not influence the risk of new-onset AF in any analysis.

Table III. Predictors of new-onset atrial fibrillation including effect of high alcohol intake at baseline in multivariate Cox regression analysis.

Variable	Hazard ratio (95% CI)	p-value
In-treatment heart rate (beats/min)	1.04 (1.03–1.04)	< 0.001
Age (years)	1.09 (1.07–1.11)	< 0.001
Treatment allocation to losartan	0.55 (0.44–0.68)	< 0.001
Baseline heart rate (per beat/min)	1.02 (1.01–1.03)	< 0.001
In-treatment systolic blood pressure (per 10 mmHg)	1.12 (1.05–1.20)	< 0.001
Male gender	1.46 (1.17–1.82)	0.001
Baseline systolic blood pressure (per 10 mmHg)	1.12 (1.04–1.22)	0.004
Baseline alcohol intake > 10 drinks/week^a	1.83 (1.16–2.91)	0.010
In-treatment Cornell product (per mV × ms/100)	1.01 (1.00–1.02)	0.048
Baseline smoking	1.19 (0.87–1.63)	0.278

^aCompared with less or no alcohol per week. CI, Confidence interval. The variables are arranged by predictive power.

Table IV. Predictors of new-onset atrial fibrillation including effect of high alcohol intake, total cholesterol, potassium and UACR at baseline in multivariate Cox regression analysis.

Variable	Hazard ratio (95% CI)	p-value
In-treatment heart rate (beats/min)	1.04 (1.03–1.05)	< 0.001
Age (years)	1.09 (1.07–1.11)	< 0.001
Treatment allocation to losartan	0.56 (0.45–0.71)	< 0.001
In-treatment systolic blood pressure (per 10 mmHg)	1.13 (1.06–1.21)	< 0.001
Baseline heart rate (per beat/min)	1.02 (1.01–1.03)	< 0.001
Male gender	1.42 (1.10–1.81)	0.006
Baseline systolic blood pressure (per 10 mmHg)	1.13 (1.03–1.23)	0.007
Baseline potassium (per mmol/l)	1.52 (1.12–2.07)	0.007
Baseline alcohol intake > 10 drinks/week^a	1.60 (0.94–2.72)	0.081
Log baseline UACR (per 1 log unit)	1.14 (0.96–1.35)	0.138
In-treatment Cornell product (per mV × ms/100)	1.01 (1.00–1.02)	0.183
Baseline total cholesterol (per mmol/l)	1.07 (0.96–1.18)	0.227
Baseline smoking	1.22 (0.87–1.72)	0.255

^aCompared with less or no alcohol per week. CI, Confidence interval. Log UACR, logarithm of urine albumin/creatinine ratio. The variables are arranged by predictive power.

New-onset AF occurred in 5.7% of all patients with alcohol intake > 10 drinks/week (n = 20 of total n = 351). This was mainly a finding in men (7%, n = 19 of total n = 287), which was significant both in univariate Cox regression analysis (HR = 1.68 [95% CI 1.04–2.70], p = 0.033) and in the multivariate adjusted Cox regression model, which did include total cholesterol, potassium and urine albumin/creatinine ratio (HR = 1.81 [95% CI 1.04–3.16], p = 0.036). For women, the cell number in the highest alcohol category was too small to make meaningful analysis, and the overall alcohol–gender interaction for new-onset AF was not statistically significant.

There was no significant interaction between high alcohol intake and treatment allocation. Still, high alcohol intake was a significant predictor of new-onset AF in the atenolol arm but not in the losartan arm with fewer instances of new AF in univariate and multivariate analyses (data not shown). There was no significant difference in risk of new-onset AF in smokers compared with non-smokers, across smoking categories including ex-smokers or never smokers or in multivariate Cox analyses. There was no significant interaction between high alcohol intake and smoking (data not shown).

Discussion

This study showed that intake of alcohol > 10 units/week predicted new-onset AF in univariate and in multivariate Cox regression analyses in hypertensive male patients with ECG LVH, but not in the total study population of hypertensive patients of both genders. There was no interaction between gender and high alcohol intake > 10 units/week; still men had a higher proportion with alcohol > 10 units/week. Smoking did not predict AF, and there was no significant interaction between high alcohol intake and smoking.

Studies on alcohol intake and new-onset AF in general populations have reported an increased risk of AF at a much higher threshold values, above about 20 drinks per week, than in our hypertensive population with LVH (5,6). The fact that hypertensive patients with LVH already are at higher risk of developing AF may possibly explain that we found a lower alcohol threshold for risk of new-onset AF.

Our findings in AF contrast with the apparent protective alcohol effect on the risk of myocardial infarction, which fell gradually with higher intake (11). On the other hand, there was a sharp rise in the risk of stroke with more alcohol (11). This suggests that the increase in stroke risk attributable to alcohol is not mediated through atherosclerosis, but rather to thrombo-embolism secondary to AF.

Increased adrenergic response including a higher heart rate has been suggested as a mechanism linking alcohol consumption with the risk of new-onset AF. Increased 24-hour heart rate was positively correlated to alcohol intake in normal men (17). Increased levels of circulating plasma catecholamines and increased heart rates were found in alcoholic men during withdrawal (18). A study of men with a history of AF episodes related to binge-drinking compared with controls showed increased lymphocytic β-adrenoreceptor density during alcohol intake and increased low-frequency/high-frequency ratio in spectral analysis of heart rate variability, suggesting increased sympathetic activity (19). A more recent study reported a similar gradient for low-frequency/high-frequency ratio in addition to increased heart rate through the range non-drinkers-moderate-heavy drinkers (20).

Increased heart rate is associated with onset of AF (21), and an increase in heart rate during antihypertensive therapy was associated with a greater likelihood of new-onset AF in the LIFE study (16). From this, we might expect that a beta-blocker should provide some protection against catecholamine-related AF. However, in the LIFE study, losartan prevented AF more effectively than atenolol (15), and meta-analyses of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers showed blockade of the renin–angiotensin system to be superior to prevent recurrent AF (22). In the present analysis, there was no significant interaction between high alcohol intake and treatment allocation. Concomitant treatment with diuretic could possibly reduce the risk of new-onset AF through reduction in intra- and extra-vascular volume load; however, even though we had a detailed record of hydrochlorothiazide as add-on medication and its maximal dose, we found no association of hydrochlorothiazide with new AF in any analysis.

The present study is a retrospective analysis of data from the LIFE study, which was neither designed nor powered to investigate the exact relation between alcohol intake and new-onset AF. The category > 10 drinks of alcohol per week comprises the range from high to excessive drinking (though alcohol abuse was an exclusion criterion and nobody with excessive drinking habits should have been included). Thus, the present data do not permit the definition of an exact consumption threshold. Furthermore, analysis is limited by the low number of women in the higher drinking categories. The untoward effect of alcohol on new AF in this population was also marginalized after multiple entering and deleting of univariate variables, and was of borderline significance only (p = 0.081) when baseline cholesterol, potassium and urine albumin/creatinine ratio were included into the model. Our data can, however, justify the conclusion that up to 10 drinks of alcohol per week appears to be safe with respect to the risk for AF in hypertensive patients with LVH.

The alcohol content in a standard British drink is less than in Scandinavian countries and in the USA (23). Because of response categories, we did not succeed in re-calculating the alcohol intake to an equal standard drink size. We might thus have overestimated the alcohol intake in the British participants, who constituted 9% of the study population. Only baseline alcohol intake was registered, and the drinking pattern was not assessed, so no measure of binge drinking was available. Furthermore, we could not differentiate the ex-drinkers from the never drinkers within the non-alcohol category, which may have concealed an increased risk of AF in ex-drinkers. However, the information on alcohol intake that was collected was quite powerful in predicting components of the primary endpoint in the LIFE study (11) and as such indicated a high degree of trial precision and statistical power for various endpoints in this large population. Data is still limited to hypertensive patients above 55 years with LVH without alcohol abuse at outset.

Finally, the incidence of new-onset AF may have been somewhat underestimated, since we have limited the diagnosis of new-onset AF to the cases confirmed with ECGs interpreted according to the Minnesota Code classification system, whereas AF episodes reported as adverse events were not included in the analysis. However, these limitations taken into account, our data suggest that an alcohol intake of 10 or more drinks per week was associated with increased adjusted 5-year risk of new-onset AF in hypertensive male patients with LVH, suggesting a considerably lower threshold for cardiac toxic and pro-arrhythmic effect of alcohol than seen in studies of the general population.

Disclosures

This work was supported by Merck & Co., Inc. Drs Gjesdal, Olsen, Ibsen, Devereux, Okin, Kjeldsen, Dahlöf and Wachtell have received grant support from Merck & Co, Inc., the sponsor of the LIFE study.