Both AS and MR were related to AF, as well as to the composite endpoint (valvular surgery or death). In the univariate analysis, patients with significant AS had 273% higher risk of AF and 290% higher risk of the composite endpoint than those without significant AS. For patients with significant MR, the risk of AF and the composite endpoint was increased by 152% and 107%, respectively, compared to patients without significant MR.
We also performed multivariate analysis with adjustment for age, sex, other valvular abnormalities, LVEF and LA diameter. Notably, AS was independently related to both endpoints, whereas MR was not independently related to either endpoint in this analysis. Age, AS, LVEF, LA and “AS+MR” as a four-group variable were independently related to both endpoints in the multivariable analysis. Sex and TR were independently related only to the composite endpoint, while there was still a significant association between AR and AF. MR was not related to either endpoint.
Our results suggest pathophysiological difference between AS and MR in AF development. It had been shown that MR causes structural changes in LA including dilatation, myofibril hypertrophy and fibrosis . The atrial refractory period shortens as the atrium dilates and the vulnerability to AF increases [12, 13]. LA size is a well-known powerful independent predictor of AF [14–16]. Subjects with significant MR usually have dilated LA due to volume overload. Indeed, in the present study they had on average 6.3 mm larger LA size than subjects without significant MR.
What do we know about LA size in patients with AS? Dalsgaard et al. have shown that the degree of valvular AS is an independent predictor of LA enlargement . LV pressure overload causes compensatory concentric LV hypertrophy , which leads to both LA pressure overload and enlargement . In our study individuals with significant AS only had 1.5 mm larger LA diameter than subjects without significant AS. This indicates that AF in subjects with significant AS not only precipitated by increased LA size, but rather by pressure overload with subsequent LA structural remodeling.
There is a strong and extensive evidence of atrial fibrosis being a key player in the development and persistence of AF [19, 20]. Li and colleagues have shown that atrial fibrosis causes localized regions of conduction slowing, increasing conduction heterogeneity and thereby providing an AF substrate . Very recently it has been demonstrated that mast cells can infiltrate the atrium of pressure-overloaded mice and contribute to the pathogenesis of atrial fibrosis and AF susceptibility .
AS and MR are often seen together in clinical practice. MR has been reported to be present in up to 70% of patients with symptomatic severe AS [23, 24]. The present study clearly shows that, among patients with combined AS and MR, the risk for valvular surgery or death was mostly influenced by the AS. However, AS and MR were nearly equally important for the risk of AF alone. Subjects with combined AS and MR seem to have the highest risk for both AF and the composite endpoint.
The presence of AF in patients with aortic valve disease is a poor prognostic sign, associated with considerable increase in morbidity and mortality. Bergeron et al.  reported a 75% one-year mortality following the onset of AF in patients with AS that exceeds 50% one-year mortality following the onset of heart failure. According to the latest American  and European  guidelines for aortic valve replacement, patients with asymptomatic severe AS and preserved LV systolic function  should be managed by “watchful waiting” with close clinical follow-up. While these asymptomatic patients are followed-up, the pathophysiological changes in the LV lead to the structural, functional and electrophysiological abnormalities in the LA, culminating in the development of AF . In patients with severe AS, the onset of AF may cause immediate clinical deterioration  and dramatically worsened prognosis . According to current guidelines, with onset of symptoms, surgery may be considered in these patients. However, these patients are likely to have enlarged and structurally abnormal LA. AF and its hazards are difficult to control once established and, therefore, a preventive approach to the problem may be preferable . Possibly, aortic valve replacement might be justified in some of these patients even at an asymptomatic stage. This approach can perhaps be more easily accepted with the development of new, less invasive methods. This should be further investigated.
This study has potential limitations, besides those caused by its retrospective nature. The major limitations include the absence of some potentially important clinical patient data (e.g. history of hypertension, ischemic heart disease, hyperthyroidism or detailed medication list). We had no access to this information at the time of the baseline examination 1996, because the hospital electronic patient record system was established a few years later. Therefore, adjustment for these factors was not possible.
Due to the retrospective nature of the study, the method for AF detection in our study was based on the presence of AF on clinically motivated ECG which can obviously underestimate the true number of patients with paroxysmal form of arrhythmia and could have affected results if dedicated AF screening would had been used.
Finally, differences between echocardiography guidelines in force today and practice that existed at the time of baseline examination 1996 pose another limitation that needs to be considered while interpreting our findings. AS severity at baseline examination was assessed using mainly peak transaortic valve pressure gradient, which can be affected by other heamodynamic parameters such as the changes in cardiac output or preload state. Information on AS jet velocity, mean transaortic gradient or valve area, which are the fundametnals of AS grading today  was not available for all subjects and reevaluation of source historical image data not possible. In regard to LVEF, the practice at the time was to rely on visual estimation of LVEF, which may be considered not as robust as quantification techniques used today . It has, however, been suggested that visual estimation of LVEF is basically as accurate as LVEF estimation by Simpson's biplane method of disc when compared to a gold standard method . Also, it would have made the data more robust if only one observer had performed grading of echocardiography, findings, which however was not possible due to the nature of the investigation.