Prognostic implications of valvular heart disease in patients with non-valvular atrial fibrillation

Atrial fibrillation (AF) and valvular heart disease (VHD) are frequently encountered in clinical practice, and often coexist, especially in the elderly population [1‐3]. Both conditions are associated with increased mortality and morbidity [4‐6]. Recent guidelines suggest careful evaluation of patients with AF and VHD due to the puzzling nature of their coexistence [7‐9].

Post-hoc sub-analyses of the existing randomized controlled trials on oral anticoagulation have demonstrated the prognostic significance of VHD across a plethora of outcomes among patients with AF [10‐15]. Valve lesions have been compared on the basis of their association with specific outcomes, revealing contradictory results between studies [11, 13, 16]. However, the predictive performance of VHD in AF outside highly-selected trial cohorts remains debatable, since outcomes have been mainly analyzed under the scope of comparisons between type and dosage of oral anticoagulants [192021].

The main objectives of this study were to evaluate the association of sVHD and individual valve lesions with clinical outcomes, and to detect specific patient subgroups where the prognostic value of sVHD is particularly evident.

In this post-hoc analysis of the MISOAC-AF clinical trial, comprising a relatively multimorbid patient population with non-valvular AF, several findings were noted: (1) moderate or severe VHD (sVHD) accounted for almost 6 out of 10 patients with non-valvular AF, while patients with sVHD and non-valvular AF received significantly more VKAs compared with NOAC, (2) sVHD was an independent predictor of mortality and HF-related hospitalization; (3) the prognostic value of sVHD was particularly evident in patients aged < 80 years old and in those without history of heart failure; and (4) moderate/severe AS and TR were both associated with all-cause and cardiovascular mortality, while moderate/severe AS and MR were associated with HF-hospitalization.

Five post-hoc analyses of large randomized clinical trials (RCTs) on oral anticoagulation (ENGAGE AF-TIMI [2017], ORBIT-AF [2017], RE-LY [2016], ARISTOTLE [2015], ROCKET-AF [2014]) have analyzed the prevalence and prognostic value of VHD in patients with AF [10‐14]. To place these pivotal trials into perspective, the prevalence of sVHD (moderate/severe VHD) was 13% (n = 2824) in ENGAGE AF-TIMI, 27.7% (n = 2705) in ORBIT-AF, 21.8% (n = 3950) in RE-LY, 26.4% (n = 4808) in ARISTOTLE and 14.1% (n = 2003) in ROCKET-AF. In our study, 58.5% (n = 575) had sVHD, which is remarkably higher compared with the aforementioned studies. This discrepancy has three main etiologies; (1) the higher rate of comorbidities of our patient population, (2) the strict eligibility criteria of the RCTs compared with the MISOAC-AF trial, and (3) the varying definitions of sVHD across studies. Indeed, our study included only AF inpatients who were hospitalized for cardiac reasons [28], which is undoubtedly a relatively multimorbid group of patients. This is additionally confirmed by the higher age (median of 76 [14]) and CHA2DS2-VASc score (mean of 4.3) of our patients, compared to other AF registries. Furthermore, the less strict inclusion/exclusion criteria of our study [22], compared with other large trials, extended the range of included patients. Thus, our study could be a better reflection of “real-world” patients with AF, and our results could be indicative of the true prevalence of VHD among patients with non-valvular AF. Lastly, our study also included patients with TR, which was highly prevalent in our population (36.8%). This could partially explain the big difference in sVHD prevalence between our and other studies such as the ROCKET-AF and ENGAGE-AF, which did not include TR in their definition of sVHD.

In the ENGAGE AF-TIMI 48 study, the presence of VHD significantly increased the risk for all-cause (aHR 1.40) and cardiovascular death (aHR 1.47), major bleeding (aHR 1.21), major adverse cardiac events (aHR 1.29), as well as composite endpoints of stroke/SEE or death (aHR 1.30) [10]. However, the outcomes of stroke/SEE were similar in AF patients with and without VHD [10]. In the ORBIT-AF study, VHD was significantly associated with all-cause death (aHR 1.23), while stroke and major bleeding were not related with VHD status after adjustment for covariates. In the RE-LY and ROCKET-AF studies, major bleeding was the only outcome related with VHD status (aHR 1.32 in both studies). Contrarily, in the ARISTOTLE study, patients with VHD had higher rates of stroke/SSE (aHR 1.34) and death (aHR 1.48), but similar rates of major bleeding.

The aforementioned results reveal a lack of homogeneity across studies regarding the prognostic effect of VHD on clinical outcomes. A recent meta-analysis that analyzed patients enrolled in these studies reported that patients with VHD had higher risks of mortality and major bleeding but not stroke or SEE [21]. In our study, sVHD was significantly associated with all-cause and cardiovascular mortality, which is in agreement with results from the ENGAGE AF-TIMI, ORBIT AF and ARISTOTLE studies. No association was noted between VHD status and risk of stroke/SEE in our study, results that were consistent with four out of five studies. However, the analysis of the risk of major bleeding in our study produced conflicting results with those of other studies, since it appeared to be similar between VHD groups, and even numerically reduced in patients with VHD. The reasons for this specific finding are unknown, and not explained by the higher baseline bleeding risk of the VHD subgroup. Interestingly, the risk of HF-hospitalization, which is an outcome that was not examined in previous studies, appeared to be 2.5 times greater in patients with sVHD in our study. This observation has several implications regarding increase in health care costs, and warrants further investigation in future studies.

Of note, the prognostic effect of sVHD was particularly evident in patients < 80 years old and in those without history of heart failure. This could be partially explained by the fact that older patients or patients with HF have competing risks for worse outcomes due to multimorbidity; hence, sVHD is being heavily competed by other clinical factors (frailty, cancer, etc.). As a result, the presence of sVHD has an increased impact on clinical outcomes and becomes more evident/profound in “younger, HF-free” patients. Even though results were adjusted for covariates, it is likely that some residual confounding still exists. To our knowledge, this is the first subgroup analysis in patients with VHD and AF. Thus, careful evaluation of the presence and severity of VHD in these specific subgroups and modifications in the management of these patients is encouraged to ultimately improve prognosis.

The association of individual valve lesions with clinical outcomes has been inadequately investigated in patients with AF. Few studies have reported on the prognostic value of AS [11, 18, 29]. A sub-analysis of the ROCKET-AF trial [29] and a Danish nationwide study [18] compared AS with AR and MR, showing its prognostic superiority for all-cause death, stroke/SEE and major bleeding. In a post-hoc analysis of the ORBIT-AF trial, AS was associated with higher risk of death, but not stroke or major bleeding [11]. Data from our study suggest that AS has a graded and independent association with increased risk of all-cause death, cardiovascular death and several net clinical outcomes. Patients with moderate/severe AS had 1-year mortality risk of 33.9%, which is considerably higher compared with cohorts of patients with [18] or without AF [30]. Stroke/SEE and major bleeding events did not differ based on VHD status in our study. However, the risk of HF-hospitalization was more than two-fold in patients with AS, which has not been reported in previous studies. AR did not have any association with clinical outcomes neither in our study, nor in other registries [11, 18, 29]. MR appeared to be the most prevalent valve lesion across AF studies [21], even though it has not been showed conflicting results when analyzed. Specifically, recent results associate MR with higher risks of all-cause and cardiovascular death [31] but not thromboembolic events [18, 31], which is contradictory to earlier studies that reported the protective effect of MR against stroke [32, 33]. In our registry, MR was indeed the most prevalent valve lesion, while patients with moderate/severe MR were at higher risk of HF-hospitalization but not death, stroke/SEE or major bleeding. The prognostic effect of TR has not been analyzed in patients with AF, since large AF trials did not include TR in their definition of VHD due to its unlikely impact on thromboembolic risk. Results from our study suggest that moderate/severe TR was an independent predictor of all-cause and cardiovascular death, beside AS, indicating the need for careful evaluation of patients with AF and TR.

Two particularly obvious, yet very interesting, observations should be underlined that warrant a careful evaluation of patients with valvular heart disease. Firstly, the results of this study suggest that the generally accepted term “non-valvular AF” is a misnomer, since more than half of our patient population with non-valvular AF had significant VHD, even after exclusion of those with mitral stenosis or mechanical heart valves. Secondly, the so-called “non-valvular AF” could also be largely misleading and confusing in daily clinical practice, as demonstrated by the fact that significantly more patients with sVHD that could benefit from the use of NOACs [9, 34] were treated with VKAs in our cohort, in contrast to patients without sVHD in whom NOACs were preferred. Currently, there is no evidence that the presence of VHD, other than moderate-to-severe MS and mechanical prosthetic heart valves, should modify the choice of oral anticoagulation [35]. However, as indicated in our study the current labeling and description of these agents as being indicated for non-valvular AF may be leading to undertreatment of patients with VHD. These observations have also been stressed out in a paucity of studies [13, 14, 19, 35], which ultimately led to a suggestion for abandonment of the terms “valvular/non-valvular AF” in the recent 2020 AF guidelines [9].

This study has a number of strengths. It is the first one that examined the risk of HF-hospitalization according to VHD status. The study is also unique because of the subgroup analyses, and more importantly the identification of specific subgroups of clinical interest in which the prognostic effect of VHD becomes more evident. Moreover, the study explored the association of individual valve lesions with several outcomes, which has not been adequately investigated in previous studies. Finally, it could be argued that the patients of this study represented more accurately a “real-world” AF population, since the MISOAC-AF trial avoided the strict eligibility criteria of other large RCTs.

As for the study’s limitations, this post-hoc analysis in patients with and without VHD was not prespecified. The number of cases for some events were too small, which may have weakened the validity of Cox regression analysis. The medium sized sample resulted in smaller subgroups; thus, interpretation should be considered in this context. However, the high event rates and the large proportion of patients with VHD likely provided adequate statistical power to detect heterogeneity across subgroup analyses. There was no information on the cause of valve lesions. Patients had substantial differences in baseline characteristics based on VHD status, and even though multivariable adjustment was performed, some residual confounding likely still exists. Comparisons of risks of events in patients treated with either VKA or NOAC cannot be made since this study was not designed for this.

In this contemporary registry of patients with non-valvular AF that were hospitalized in the Cardiology ward, almost 60% had moderate/severe VHD. VHD was associated with mortality and HF-hospitalization and its prognostic effect was particularly evident in patients aged < 80 years old and in those without history of heart failure. Moderate/severe AS had the greatest prognostic value among valve lesions. Increased use of VKAs was noted in patients with VHD, which implied that the term “non-valvular AF” might be confusing and misleading in daily clinical practice. Thus, re-evaluation of this potentially obsolete term is warranted.