OUP user menu

Primary prevention of atrial fibrillation by statins: still unresolved?

Harry J.G.M. Crijns
DOI: http://dx.doi.org/10.1093/eurheartj/ehp362 2302-2303 First published online: 30 August 2009
This commentary refers to ‘Effects of rosuvastatin on atrial fibrillation occurrence: ancillary results of the GISSI-HF trial’, by A.P. Maggioni et al., on page 2327

Atrial fibrillation (AF) frequently complicates heart failure (HF), and vice versa, and—when found together—for clinicians they form an insufferable odd couple. Once AF appears, HF can worsen and stroke rate increases. Therefore, prediction and prevention of AF may be desirable goals. Secondary prevention using conventional rhythm control measures is, however, difficult to achieve and has not shown survival benefit.1,2 In high-risk patients primary prevention of AF using non-antiarrhythmic drugs such as angiotension-converting enzyme (ACE) inhibitors, angiotensin receptor blockers, and statins may be useful because of their secondary antiarrhythmic effects.3,4 The ancillary study of GISSI-HF by Maggioni et al. found that compared with placebo rosuvastatin reduces the incidence of AF in patients with HF.5 This substudy is very interesting and the first to report the effects of a statin compared with placebo on the incidence of AF in a stable HF population. Unfortunately, the antiarrhythmic effect is not large and the question remains of whether suppressing AF in HF patients is beneficial or only kills the messenger.

This GISSI-HF substudy mainly concerned primary prevention since patients with AF on the pre-randomization ECG were excluded and eventually only ∼15% of study patients had had paroxysmal AF before inclusion. However, one must bear in mind that many patients with HF may have asymptomatic or otherwise unrecognized AF episodes, and these were not checked before randomization. Even so, the question is whether it was reasonable to exclude patients with previous AF. In addition, a related question is whether the endpoint should have been incidence of new AF. In stead, development of or progression to non-self-terminating (i.e. persistent) AF would have been a more reasonable endpoint for a statin trial. Alternatively, the burden of AF (e.g. by looking at the number of episodes over time) might have been taken. This relates to the fact that the final common and most plausible mechanism by which statins act is prevention of structural atrial remodelling, which in turn is connected to persistence (or progression) and recurrences (or burden) of AF, respectively. Obviously, with such endpoints patients with previous paroxysmal AF would also have qualified for the study. In fact, such a study is still ‘doable’ in the GISSI-HF database since the investigators may consider defining all patients with AF on minimally two of their last follow-up ECGs as showing progression. For burden, simply the number of ECGs with AF may be counted. Obviously such an analysis is methodologically not ideal and may suffer from small numbers. It may be worthwhile, especially in view of the recent finding that AF burden rather than one new-onset episode of AF seems very relevant for predicting stroke and possibly also other major cardiovascular events.6 In addition, progression to persistent AF may serve as a clinically relevant endpoint in patients who are otherwise stable with their AF.7

The authors suggest that the Kaplan–Meier curves show a steadily diverging course. However, most of the effect seems to occur acutely since the curves separate most markedly at the very first visit after randomization. This suggests an acute antiarrhythmic effect rather than a gradual antiremodelling effect. Which acute effect is uncertain. The authors mention the anti-inflammatory effect of statins but cannot substantiate this by incident AF-related C-reactive protein levels (not available). It is questionable if such short-term mechanisms of AF prevention by statins can be found in studies such as GISSI-HF since the numbers of patients developing AF in the short term are extremely low. Notwithstanding the above, the divergence of the curves seen near the end of follow-up seems to support the authors' notion that a longer follow-up might have shown statistically significant effects. Once more the mechanism is unclear. Chronic inhibition of the renin–angiotensin system has been proposed but represents a mechanism which cannot be tested in GISSI-HF since almost all patients had inhibitors of the system, blurring any possible statin effect. Also suppression of adrenergic activity has been proposed, which may be studied separately in the GISSI-HF database. Statin effects are most plausible through antiatherogenic actions leading to less ischaemia-related ventricular dysfunction and concurrent atrial stretch. However, the latter notions are not supported by the data since the incidence of major adverse events including myocardial infarction was similar among both treatment arms, even not in the subgroup (42% of study patients) with an ischaemic aetiology of HF.8 Obviously, this hypothesis will be difficult to test in the future since the time is gone when patients with coronary disease can be selected for studies such as GISSI-HF. This makes the GISSI-HF a valuable database for future post hoc analyses.

Across all pre-defined subgroups the effect of rosuvastatin was similar. However, the subgroups did not take atrial characteristics into account. The investigators used age, ejection fraction, New York Heart Association (NYHA) class, aetiology of HF, diabetes, renal function, history of AF, and cholesterol levels into account. To predict future events or response to rhythm control in patients at risk of (recurrence of) AF these clinical characteristics are frequently used, but they are only indirectly related to the ‘scene of calamity’. The ‘scene of calamity’ is the atrial wall— the place where AF resides—whereas the patient's age and other characteristics are only indirectly related to it.9 Unfortunately, GISSI-HF did not broadly record electrocardiographic or echocardiographic parameters such as the PR-interval reflecting P-wave duration, atrial volumes, or total atrial activation time.10,11 This might have enhanced the pinpointing of mechanisms by which statins can reduce AF. In addition, it might have helped to identify patients who would benefit from statin treatment, not only in terms of AF prevention but also with respect to atrial remodelling and developing stroke or recurrent AF-related HF. A recent policy conference at the European Heart House states that future studies aiming to predict incident or recurrent AF, and studies trying to identify factors associated with adverse cardiovascular events related to AF, should take specific ‘scene of calamity’ parameters into account more often.7

The definition of AF occurrence during follow-up was quite remarkable. There is general agreement that 30 s of AF recorded on an ECG suffices.7 Apart from the mere presence on the ECG at the scheduled follow-up visits, the GISSI-HF definition also included ‘AF as a cause of worsening HF or of hospital admission or AF as an event occurring during a hospital admission’. However complex this definition is, it simply boils down to recording an ECG and finding AF on it or not. Using their definition, the investigators mixed up establishing the presence of AF with recording the clinical consequences of AF. This does not seem logical and it illustrates how clinicians and investigators look at AF as an arrhythmia directly causing adverse events such as stroke and HF rather than as a simple marker for a more adverse prognosis. This image of AF continuously feeds the notion that abolishing AF does away with its risks, which—as yet—has never been shown to be the case.1 Both for clinicians and for trialists it is important to note that once AF is recorded, the stroke and HF risks need to be assessed irrespective of the setting in which AF occurred. Thereafter therapy needs to be adjusted, which in patients with the combination of AF and HF frequently is not done.12 At the end of the day, the GISSI-HF definition of AF did seem to be very useful since definition-based frequency distributions of AF occurrences are not presented. In addition, considering the jumps in the AF-free survival curves at each planned visit, one may assume that most incidences of AF were found at those visits, these patients being asymptomatic in terms of their AF and not suffering from HF symptoms or HF progression at the time of their new-onset AF.

The present report from GISSI-HF provides valuable information on the potential antiarrhythmic effects of a non-antiarrhythmic drug in HF patients. The CORONA study may shed further light on this issue, but for now GISSI-HF showed that uncoupling the insufferable odd couple HF and AF with statins is not easy.

Conflict of interest: none declared.


  • The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.

  • doi:10.1093/eurheartj/ehp357


View Abstract