The Estimated Risk of Atrial Fibrillation Related to Alcohol Consumption

Michael A. Rosenberg, MD 1, Kenneth J. Mukamal, MD, MPH 2

1Cardiovascular Institute, Beth Israel Deaconess Medical Center, Boston, MA, USA.2Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, Boston, MA, USA.

Abstract

The risk of acute heavy alcohol intake on the development of atrial fibrillation (AF), aka ?holiday heart syndrome?, has been well-described. However, whether chronic alcohol intake is also associated with increased risk of AF, or might even be protective as has been observed with other cardiac conditions, is more uncertain. A number of studies, from basic science to large cohort studies have been performed to analyze the association between alcohol and AF. Basic-level studies have found that alcohol causes changes in tissue electrophysiology, ion channels, and circulating hormones, which might promote development and maintenance of AF. Clinical studies have generally shown groups with the highest regular intake of alcohol to be at increased risk, with no association with more moderate use. However, these studies have not always accounted for other AF risk factors, been inconsistent in the assessment and validation of the quantity of alcohol consumed across populations, and been unable to completely separate drinking patterns from overall health of participants. As a result, solid conclusions about a threshold level for ?safe? chronic alcohol intake cannot be made with regard to AF risk, but it appears to be safe within currently recommended limits of 1 drink daily for women and 2 for men. In this review, we discuss these findings, limitations, and conclusions.


Correspondence to: Kenneth Mukamal, MD, MPH, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 1309 Beacon Street, 2nd Floor, Brookline, MA, 02446.

Introduction

In 1978, Philip Ettinger and colleagues described in the American Heart Journal a condition in which patients would present with cardiac arrhythmias after episodes of heavy alcohol consumption.1 They called the condition ‘holiday heart syndrome,’ since it was noted that arrhythmia presentations seemed to occur on weekends or during the holidays, times when people tended to increase their alcohol consumption. Importantly, all of the cases in Ettinger’s seminal paper occurred in people who had a long history of heavy alcohol intake: daily alcohol consumption in the patients varied from six ounces of martini daily to 12-15 beers daily to daily whiskey in “quart quantities.”1 In a paper in The Lancet in 1984, Thornton and colleagues pointed out this limitation of Ettinger’s study, before going on to describe cases of four patients without a history of long-term heavy alcohol use who presented with atrial fibrillation (AF).2 These four cases, noted to be ‘light’ drinkers as they consumed less than 30 grams of alcohol per day, presented in AF while still apparently intoxicated (the authors estimate consumption of about 138 grams of alcohol),2 with reversion to sinus rhythm the following morning after presumed metabolic elimination of the ingested alcohol. Thornton’s interpretation, supported by a variety of metabolic and electrocardiographic changes noted at the time,1,3,4 was that it was the heavy acute intake of alcohol that was the precipitant for acute AF, and this hypothesis has since been supported by a number of studies that have described the risk of binge drinking on development of AF.2,5-8

Other cardiac conditions have also been associated with binge drinking, including myocardial infarction,9 stroke10 and heart failure,11 well-illustrating the potential of heavy alcohol intake to act as a direct cardiac toxin. However, more recently, observational studies have suggested alcohol might have a protective effect if taken in low to moderate amounts, with reduction in the risk of coronary heart disease, stroke, congestive heart failure,12-16 as well as all-cause mortality.17 For example, one six-month randomized trial examined regular wine intake among diabetic survivors of acute myocardial infarction and found a benefit of modest drinking on myocardial infarction.18 These findings have raised the question of whether a similar paradoxical effect might be observed between alcohol and AF, with moderate alcohol intake having a protective effect and only binge drinking increasing risk. Since Ettinger, a number of studies have attempted to answer these questions. In this review, we will examine the evidence for an association between alcohol intake and AF, as well as the methodological limitations surrounding studies of the association.

Mechanistic Associations

Mechanistically, a number of hypotheses have been put forward with regard to how alcohol might increase the risk of AF (Figure 1). In their initial study, Ettinger and colleagues noted a modest increase in the PR, QRS, and QT intervals on the electrocardiograms of alcoholics presenting in AF,1 findings which have since been replicated in several other studies in humans.19,20 Specifically, abnormalities in the P wave, which represents atrial depolarization, have been noteworthy, including a study by Uyarel et al., who performed a randomized crossover study in otherwise healthy humans, and found that consumption of moderate-dose ethanol (0.97+/-0.12 g/kg body weight over 1 hour) was associated with a significant lengthening of both the P wave duration, as well as an increase in P wave dispersion.21 Similarly, Steinbigler et al. found that P-wave signal average studies identified individuals who might be at risk for alcohol-induced AF,22 indicating that the changes observed in the P wave with alcohol intake might be a precursor to the development of AF.

Figure 1. Possible mechanisms of alcohol-induced atrial fibrillation



At a more basic level, acute ethanol infusion in animal models had been demonstrated as early as 1962 to shorten the atrial refractory period,4 an effect that has since been observed in human atria23 as well as in rabbit pulmonary vein isolates,24 although it should be noted that no changes were observed in the rate of spontaneous firing or triggering in the pulmonary veins, as might have been expected in AF. Computer models have shown that shortening of the atrial refractory period allows for multiple reentry wavelets, and thus alcohol-induced shortening of the atrial refractory period would create a pro-arrhythmic substrate accommodating the development and maintenance of AF.25-27 Changes in several ion channels have been implicated in this process, including decreased activity of inward sodium (INa) and calcium (ICa,L) currents and increased activity of outward potassium currents (including IK1 and IKur).25 Several experiments have shown that treatment with ethanol causes depression of sodium28,29 and calcium signaling,28,30,31 as well as increased potassium channel function,32 all of which would in theory shorten the atrial refractory period, although the reproducibility of some results has been variable.33

Cardiomyopathy as a result of excess alcohol intake is well-known, and many studies have identified depressed contractility with ethanol infusion in the ventricles34 as well as the atria.35 In rats consuming alcohol for 2 months,36 atrial toxicity developed in the form of decreased response to inotropic agents and decreased calcium sensitivity,36 supporting the notion that alcohol may cause direct effects on the health of atrial tissue leading to AF. However, despite these findings, replication of inducible AF in animal models has been variable. While Anadon and colleagues infused pigs with high doses of alcohol (up to 1,230 mg/kg of ethanol) and found that they could induce AF as well as atrial flutter more easily and for more sustained periods than at baseline,37 Fenelon et al. were unable to detect any changes in atrial refractoriness, conduction velocity, or arrhythmia inducibility in dogs infused with moderate amounts of ethanol.34 These results suggest that perhaps high doses of alcohol are needed to induce AF, although further studies are needed.

Other mechanisms by which heavy alcohol consumption has been noted to increase the risk of AF include the hyperadrenergic state of drinking and withdrawal,38 impairment of vagal heart rate control and autonomic function,38,39 oxidative stress,40 and an abnormal electrolyte profile.38 As AF patients have also been noted to have increased oxidative stress,41 this possible mechanism may have interesting treatment implications, as in one study investigators were able to reduce post-operative AF (another form of AF that has been linked to oxidative stress41) by treatment with the antioxidant ascorbate.42 Another recent interesting mechanistic link was described by Djousse et al., who found a positive correlation between alcohol intake and plasma atrial natriuretic peptide (ANP)43. Others have also described an increase in plasma ANP after alcohol injection in rats,44 and elevated ANP has been well-described in AF.45,46 As ANP has also been noted to shorten atrial refractoriness,47 it may also be implicated directly in the arrhythmogenesis of alcohol-induced AF, although this specific mechanism, as well as the role of serum ANP levels in predicting development of alcohol-induced AF, has not been studied.

Clinical Studies

One of the earliest studies examining the relationship between alcohol intake and AF was performed in 1985 by Rich and colleagues.8 This study was a case-control review of the medical records of 40 individuals presenting to the hospital with ‘idiopathic’ acute AF, with 64 age- and gender-matched controls. Among their findings, was the observation that not only did patients presenting with AF have nearly twice the likelihood of having a history of heavy alcohol use, but that they also were more likely than other patients without AF to have withdrawal symptoms. In addition to the likely sampling bias—62% of the AF cases had a history of heavy alcohol use—this study had a number of shortcomings, which will be reviewed in detail later.

Other observational studies, both case-control and cohort studies, on the association of alcohol and AF have been performed, and have revealed some informative, although mixed, results (Table 1). A number of case-control studies have been performed using cases of both chronic73 and paroxysmal AF,74,75 as well as right atrial flutter.23 In two studies of AF with significant findings,73,74 only the highest consumption groups—over 21 units per week74 and over 42 units per week73—were associated with an increased risk of AF (OR 1.7 and 2.4 respectively). A third study, by Mattioli and colleagues did not find any significant association with alcohol and AF.75 There was also no significant association in the AF arm of the study by Marcus and colleagues,23 although this study did find an association of right atrial flutter and alcohol intake in a subgroup of participants under 60 years of age who consumed over 1 drink per day.

In addition to the common limitations of case control studies, these were primarily registry or medical record reviews, with the largest having 1,035 cases.73 In few of the studies was the focus of the analysis specifically on alcohol and AF, and thus there was limited validation of alcohol consumption. All of these studies were limited by a small number of participants in the highest consuming group (range from 5.9% of 525 cases74 to 11% of 116 cases75), and as a result even those with significant results had wide confidence intervals. For example, in the largest case-control study, the General Practice Research Database study, the 95% confidence interval was 1.4 – 4.1 (ARR 2.4) for the group with the highest alcohol intake (over 42 units/week).73 A key limitation of these case-control studies was that there was no information about whether the significance of the groups with the highest alcohol intake was driven by individuals with extremely high levels of chronic intake or whether there was a threshold effect at the chosen cut-point.

The role of alcohol on incident AF has been examined in several large cohort studies, including the Cardiovascular Health Study,72,76 the Framingham Heart Study,77 the Copenhagen City Heart Study,61 the Women’s Health Study,62 and the Danish Diet, Cancer, and Health Study.78 Similar to the case-control studies, the general trend among cohort studies was that only the group or quintile with the highest consumption had a significant association with AF.

An important recognition among investigators of cohort studies was the need to stratify by gender, as women in general tended to consume less alcohol than men. This issue is exemplified in the Women’s Health Study, which analyzed data from a cohort of women health professionals in the United States and found that women who consumed more than 2 drinks per day were at an increased risk of AF compared with nondrinking women.62 Other studies that stratified by gender, including the Framingham Study,77 the Danish study,78 and the Copenhagen City Heart Study,61 all failed to find a significant association between alcohol and AF in women, despite non-significant trends toward higher risk in the group with the highest intake. One explanation for this phenomenon has been that fewer women were available in each study that drank in the highest consumption group;62 however, other studies with non-significant results have had large numbers of women in the highest consumption group. For example, in the Danish study, which divided participants into quintiles, the highest quintile with 5,084 participants consumed 38.8 g per day of alcohol, yet lacked even a trend toward increased risk of AF with an unadjusted HR of 1.01 (CI 0.63 – 1.61), while in men a significantly increased risk of 45% was observed with consumption in the highest quintile (68.7 g/day).

Despite the general trend of increased risk in the highest consuming group, extrapolation of results across studies is difficult. For one, the actual amount consumed varies substantially across studies. The highest group in the Framingham study consumed greater than 36.0 g/day and had an increased relative risk of 1.34 (CI 1.01 – 1.78),77 while in the Danish study the highest quintile consumed 68.7 g/day on average, with an increased risk of 1.45 (CI 1.05 – 2.01). Interestingly, quintile 4 of the Danish study, which consumed 38.1 – 68.7 g/day did not have a significantly increased risk of AF (RR 1.24, CI 0.88 – 1.73), highlighting this important limitation in comparisons across studies. In studies that examined consumption in drinks per week, there was also a large variation present in terms of actual amounts of alcohol between studies. In the Copenhagen City Heart Study, the highest group of men consumed over 35 drinks per week, and had an increased risk of AF of 1.62 (CI 1.16 – 2.27) in the age-adjusted model.61 On the other hand, in the Cardiovascular Health Study, there was no increased risk seen in the highest consumption group; however, the highest consumption group only consumed greater than 14 drinks per week,72 well below that of the Copenhagen City Heart Study. One likely reason for this discrepancy was the substantial difference in age between the cohorts; nonetheless, such a wide range of cut-points for the highest level of alcohol consumption across studies makes generalization difficult in terms of finding a threshold risk level.

In addition to the variability in quantities, there are also differences in the type of alcohol consumed, i.e. beer, wine, or spirits, across populations, and some have speculated that substances in certain alcoholic beverages, such as resveratrol in red wine, might have anti-oxidant properties that prevent AF.79 Mukamal and colleagues looked at this issues specifically in the Copenhagen City Heart Study, and did not detect a difference based on the type of alcoholic beverage consumed,61 although other studies that have looked at this issue have noted a difference in the drink preferences of consumption groups within the same study, such as in the study by Frost and colleagues, who found that while heavy drinking men (who were also the only group at higher risk for AF) drank mostly beer, lighter drinking men and all women tended to drink mostly wine,78 and had no increased risk.

In addition to the above studies, meta-analyses have also been performed80,81 exploring, among other issues, the question of what is the safe limit of alcohol consumption with regard to AF risk. The most recent, by Kodama and colleagues, examined pooled estimates from fourteen eligible studies that included both case-control studies and cohort studies. The overall conclusion from their analysis was that consumption of over 4 drinks per day was significantly associated with a 74% increased risk of AF.81 Further analysis using linear and spline regression models found that there was a very modest dose-response relationship between alcohol intake and AF with an increase in relative risk of 1.08 (CI 1.05 – 1.10) for every increase of 10 grams of alcohol consumed per day. Although the graph of the cubic spline suggested a possible threshold of 33.1 g/day, which is similar to that seen in individual cohort studies,61 there was no significant difference between the spline model and the linear one (p = 0.77). The authors’ interpretation of their analysis was that the risk of AF is increased at any level of alcohol consumption.81 However, a more interesting finding from the meta-analysis was the stronger effect in case-control than cohort studies, which was nearly statistically significant (p=0.06). As case control studies tend to assess recent consumption better than cohort studies because the reporting in cohort studies may occur years or even decades before AF develops, this finding suggests that the effect might be due to acute, rather than chronic, alcohol consumption, and thus supports the hypothesis that the highest drinking groups were merely identifying individuals more likely to binge drink.

The large variation in the design and results of the individual studies of alcohol consumption and AF suggests that combining studies in a meta-analysis, while useful in principle, may not be entirely feasible. The authors of this most recent meta-analysis acknowledged these limitations and others in their discussion, in addition to finding statistical evidence of publication bias (via an Egger’s test p = 0.03).81

Table 1. Clinical studies of alcohol and atrial fibrillation
STUDIES Number of AF cases Alcohol consumption of highest risk group Adjusted Risk (CI) versus nondrinkers Adjusted covariates
Case-control studies
UK General Practice Research Database73, 2002 1,035 > 42 units/week OR 2.4 (CI 1.4 – 4.1) Age, sex, cardiovascular morbidity
Mattioli et al.75, 2005 116 >50 mL/day Non-significant Age, sex, life stress, espresso coffee consumption, body mass index, income and smoking
UK General Practice Research Database,74 2005 525 >21 units/week OR 1.7 (CI 1.1 – 2.6) Age, sex, BMI, cardiovascular morbidity, smoking
Marcus et al.23, 2008 195 > 2 drinks/day *Non-significant Age, sex, race, hypertension, congestive heart failure, coronary artery disease, BMI
Cohort studies
Framingham Heart Study,77 2004 1,055 > 36 g/day RR 1.34 (CI 1.01 – 1.78) Sex-stratified, adjusted for systolic blood pressure, age at baseline examination, education, and cumulative history of myocardial infarction, congestive heart failure, diabetes mellitus, left ventricular hypertrophy, and valvular heart disease
Danish Diet, Cancer, and HealthStudy,78 2004 556 Mean 68.7 ± 22.8 g/week in highest quintile HR 1.46 (CI 1.05 – 2.04) Sex-stratified, adjusted for age, body height, body mass index, smoking, systolic blood pressure, treatment for hypertension, total serum cholesterol, and level of education
Copenhagen City Heart Study,61 2005 1071 >35 drinks/week HR 1.45 (CI 1.02 – 2.04) Sex-stratified; adjusted for age, smoking, education, cohabitation, family history of cardiovascular disease, diabetes, income, physical activity, BMI, FEV1, and height
Cardiovascular Health study,72 2007 1232 >14 drinks/week †Non-significant Age, sex, race, income, height, waist circumference, physical activity, use of psychoactive medication, diabetes, hypertension, coronary heart disease, congestive heart failure, and total cholesterol level
Women’s Health Study,62 2008 653 > 2 drinks/day HR 1.60 (CI 1.13 – 2.25) (Women only) Age, systolic blood pressure, history of hypertension, body mass index, smoking, history of diabetes, history of hypercholesterolemia, and randomized treatment assignment

* Significant effect in subgroup of AFL under 60 years of age (see text) † Non-drinkers limited to lifelong abstainers, former drinkers excluded (see text)

Limitations of Clinical Studies

In addition to the specific limitations mentioned above, some general limitations exist in studies attempting to estimate the risk of AF related to alcohol consumption. The first limitation applies to any epidemiological analysis of AF, which is the effect of confounding by the multiple risk factors that have been associated with increased risk. Such factors as advanced age, hypertension,48-50 gender,51-53 diabetes,54 myocardial infarction,54 congestive heart failure,54 body mass index,55 kidney function,56 bradycardia,57 forced-expiratory volume at 1 second (FEV-1),58 C-reactive protein59 and N-terminal-pro-Brain Natriuretic Peptide (NT-proBNP) level,60 all increase the risk of AF, and as a result, any non-randomized control study with a goal of examining a potential new mediator of AF may be limited by potential confounding unless it is large enough to adjust for these multiple factors. The study by Rich et al. only included 104 total participants, which allowed adjustment for only age, gender, and cardiac function.8 In contrast, the recent analyses from the Copenhagen City Heart Study61 and the Women’s Health Study62 examined data from 16,415 and 34,715 participants, respectively, and thus were able to adjust for significantly more covariates, as well as exclude participants with risk factors, such as self-reported coronary heart disease, stroke, use of cardiac medication or antihypertensives.61 The importance of having a sufficient sample size also concerns the limitation of having enough members of the ‘high risk’ group, as in the Women’s Health Study only 3.9% of participants consumed alcohol in the highest amount (> 2 drinks/day),62 a finding that the investigators suggested may have accounted for the prior negative findings of smaller prior studies in women.

Because of confounding, it is ultimately impossible to ascribe causation in observational studies. Moderate alcohol intake as a marker for healthy lifestyle has been described in other studies,67,68 and abstinence has been shown to reflect underlying illness,69 prior alcoholism,70 or lower socioeconomic class.71 To this point, Mukamal and colleagues found that former drinkers had a 25% increased risk of AF compared with lifelong abstainers, in addition to a 27% increase in mortality.72 In that same study, the investigators noted from the Kaplan-Meier curves that heavy drinkers had an initial benefit that attenuated over time, postulated to reflect a ‘healthy user’ phenomenon in which only healthy older adults are able and willing to consume alcohol.72 In some occasions, multivariate analysis can be performed to adjust for some of these factors, although in many cases it is difficult as these characteristics are often not known or easily quantified.

A second general limitation of studies assessing the association of alcohol and AF is difficulty in the quantification of alcohol intake by participants. Investigators have been left to questionnaires, interviews, and medical record review to assess the quantity of alcohol intake, with variable degrees of validity. Initial efforts, such as those of Rich et al.8 as well as several other early studies,53,63,64 used a dichotomous approach based on intake of alcohol above or below a certain amount,8,63 or the presence or absence of alcohol abuse,53,64 to analyze the effect of alcohol. As a result, these early studies were limited by the inability to analyze or describe a dose-response effect of alcohol, or to explore potential thresholds for increased risk, as later studies that included multiple incremental categories were able to do.61,62

However, even with the use of more sophisticated and thorough validation methods, investigators using current methods are still limited in their ability to assess the day-to-day alcohol intake and blood alcohol levels in participants. This leaves open the important question of whether the risk of alcohol is due to chronic intake at a constant level, or fluctuating above a given threshold on a given number of occasions—a critical issue if one is to make assessments about a ‘safety threshold’ for alcohol consumption. As mentioned, binge drinkers are well-known to have an increased risk for AF,2,65 but many studies did not obtain enough information to determine if the highest consuming individuals were also more likely to have periods of excess consumption or binging, and whether it was this behavior rather than the chronic consumption that increased risk.

Conclusions

Atrial fibrillation is the most common sustained cardiac arrhythmia, and its prevalence in the population is increasing.82 Treatment of AF, regardless of the etiology, is a heavy burden on the healthcare system, with annual U.S. costs for AF management $6.65 billion total, with $2.93 billion for hospitalizations alone.83 Moderate alcohol intake has been associated with a lower risk of cardiovascular disease, as well as overall mortality, and in terms of AF, most studies to date suggest that alcohol intake at moderate levels (2 drinks/day for men and 1 drink/day for women) is not associated with an increased risk. As far as the threshold level of alcohol consumption at which risk for AF increases, the data are less clear. Several barriers exist in the study of the effects of alcohol and AF, including study size limitations to adjust for the multiple other risk factors for AF, limitations in methods for accurate determination of alcohol consumption patterns across differing populations and demographics, and ultimately limitations in our understanding of the specific arrhythmogenic mechanisms for alcohol-induced AF. Despite these limitations, clinical and basic science studies seem to suggest that individuals who consume large amounts of alcohol are at increased risk of AF. The key question is whether these are individuals who are more likely to have ‘binge’ drinking episodes, which have been well-described to increase the risk of AF, or whether there is a risk from chronic alcohol use, whose safety threshold remains to be determined.

Atrioventricular nodal ablation and pacemaker implantation is another strategy that may be safe in select patients. However, close attention to the type of pacemaker (biventricular versus single versus dual chamber), the programming, and the patient’s functional status is likely needed to select those who will benefit most. Furthermore, there is limited data comparing the relative efficacy of this approach, in which patients may remain in atrial fibrillation but with better controlled ventricular rates, with that of pulmonary vein isolation / trigger-guided ablation, in which maintenance of sinus rhythm is the principal goal.

To date, published data support that ablative strategies are relatively safe and efficacious in elderly patients. Future randomized prospective trials are needed to further assess the safety and efficacy of pulmonary vein isolation in elderly patients. Elderly subsets of the CABANA trial may offer some insights into the relative efficacy of pulmonary vein isolation when compared with antiarrhythmic drugs in managing patients with atrial fibrillation. Further studies are needed to better evaluate the relative benefits and risks of different approaches aimed at rate versus rhythm control in the elderly population and specifically, the role and timing of ablation strategies.

References

  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]