Objectives Although guidelines support aortic valve replacement (AVR) in patients with severe aortic regurgitation (AR) and left ventricular ejection fraction (LVEF) <50%, severe left ventricular dysfunction (LVEF <35%) is thought to confer high surgical risk. We sought to determine if a survival benefit exists with AVR compared with medical management in this high-risk, relatively rare population.
Methods A large institutional echocardiography database was queried to identify patients with severe AR and LVEF <35%. Manual chart review was performed. Due to small sample size and population heterogeneity, corrected group prognosis method was applied, which calculates the adjusted survival curve for each individual using fitted Cox proportional hazard model. Average survival adjusted for comorbidities and age was then calculated using the weighted average of the individual survival curves.
Results Initially, 2 54 614 echocardiograms were considered, representing 1 45 785 unique patients, of which 40 patients met inclusion criteria. Of those, 18 (45.0%) underwent AVR and 22 (55.0%) were managed medically. Absolute mortality was 27.8% in the AVR group and 91.2% in the medical management group. After multivariate adjustment, end-stage renal disease (HR=17.633, p=0.0335) and peripheral arterial disease (HR=6.050, p=0.0180) were associated with higher mortality. AVR was associated with lower mortality (HR=0.143, p=0.0490). Mean follow-up time of the study cohort was 6.58 years, and mean survival for patients undergoing AVR was 6.31 years.
Conclusions Even after adjustment for clinical characteristics and patient age, AVR is associated with higher survival for patients with low LVEF and severe AR. Although treatment selection bias cannot be completely eliminated by this analysis, these results provide some evidence that surgery may be associated with prolonged survival in this high-risk patient group.
- valve disease surgery
- heart failure with reduced ejection fraction
- aortic regurgitation
Statistics from Altmetric.com
- valve disease surgery
- heart failure with reduced ejection fraction
- aortic regurgitation
The management of patients with both severe aortic regurgitation (AR) and severely depressed left ventricular ejection fraction (LVEF) presents a therapeutic dilemma.1 According to the 2014 guidelines published by the American Heart Association and the American College of Cardiology, aortic valve replacement (AVR) is a class I indication for patients with severe symptomatic AR regardless of left ventricular systolic function, as well as asymptomatic patients with chronic, severe AR and a LVEF <50%.2 The European Society of Cardiology and European Association for Cardiothoracic Surgery 2017 guidelines indicate that in asymptomatic patients with severe AR and impaired LVEF, LV end-diastolic diameter of >70 mm, or left ventricular end-systolic diameter >50 mm have demonstrated worse outcomes and should also be offered surgical intervention.3 4 Nevertheless, clinicians might not offer surgery to these patients given the perception that depressed LVEF may be a poor prognostic factor following AVR.5 6
Despite that, patients with severe AR and low LVEF appear to have similar operative outcomes compared with patients with higher LVEF.7 These results should build some enthusiasm for proceeding with AVR in these patients. However, a rigorous assessment of whether patients benefit from surgery must compare patients who receive surgery with those who do not receive surgery. As in many decisions to proceed with surgery or interventions, data exploring this comparison are lacking due to a paucity of data on outcomes for patients who do not receive intervention. This information is critical to managing these patients. For example, in some cases, surgical ineligibility—a subjective assessment made by a surgeon—is a marker of poor prognosis.8 In other cases, patients have similar outcomes with less extensive or even no procedure being performed.9 10 Information about patients who do not receive surgery is typically not available from surgical registries or case series.7 In this specific clinical situation, relative outcomes data are even more limited given the relative rarity of the combination of severe AR and low LVEF.
In the context of this critical need to understand if patients with severe AR and low LVEF benefit from surgical intervention, we used a large institutional database of echocardiograms to detect both patients who received and those who did not receive surgery. In particular, we compared survival in patients who received isolated AVR with survival in those patients who were managed without surgery.
We first identified all patients, both operative and non-operative, who had severe AR and LVEF <35% between January 2001 and March 2016 documented on echocardiography from a large institutional echocardiography database. AR severity was determined by an attending cardiologist. Manual chart review by a cardiac surgeon (AGF) with clinical input from a cardiologist (JHW) was performed on all patients to ascertain whether or not patients received isolated AVR without concomitant replacements or repair of additional valves or coronary artery bypass grafting, and to identify key demographic, clinical, operative details and cause of death if applicable. Given that the study cohort had low LVEF, we conducted chart reviews to ascertain if patients were treated with medical therapy consisting of beta-blockers, ACE inhibitors or angiotensin II receptor blockers (ARB), mineralocorticoid receptor antagonists, implantable cardioverter-defibrillators (ICDs) or cardiac resynchronization therapy (CRT). A cardiologist (JHW) conducted chart reviews to add information from the echo reports that were not yet described as structured data (end-diastolic and end-systolic dimensions, body surface area and left atrial dimension). We also evaluated based on clinical history and echocardiographic data for each patient, whether the severe AR likely caused the low LVEF, as opposed to another cause. These data from chart reviews were merged with structured echocardiographic data. The resulting dataset was then linked to a demographic research database, the Research Patient Data Registry (RPDR) to obtain additional information including date of death. In RPDR, date of death is extracted from the Social Security Death Index. Patients aged <18 years, those with active endocarditis, significant concomitant non-aortic valvular pathology or those who had undergone any additional surgical procedure other than isolated AVR were excluded. When aortic root replacement was indicated, we included the patient as long as the primary indication for surgery was severe AR. These patients were then divided into two groups: those who underwent AVR and those who did not undergo surgical intervention.
In addition, to assess the possible role of frailty as a confounding factor between operative and non-operative patients, we reviewed medical records to detect if frailty played a role in the decision about whether to operate. Furthermore, although all echocardiograms had been evaluated by attending cardiologists, we wanted to confirm and report specific echocardiographic parameters related to AR. Therefore, an echocardiographer (MMW) rereviewed all available echocardiogram images to confirm and report jet vena contracta, jet width as a percentage of the left ventricular outflow tract width, pressure half-time of the AR jet and velocity at end-diastole of flow reversal in the upper descending aorta.
Outcomes and covariates
The primary outcome of interest was the amount of time between the diagnosis of severe AR and death. Patients still alive at the time of the analysis were censored in the analysis.
To assess other covariates associated with the primary outcome, we considered tobacco use, chronic obstructive pulmonary disorder, diabetes, preoperative creatinine, end-stage renal disease (ESRD), hypertension, peripheral vascular disease (PVD), prior cerebrovascular accident, coronary artery disease (CAD), atrial fibrillation (AF), congestive heart failure, presence of left main coronary artery disease and New York Heart Association (NYHA) class. With respect to ESRD, this was defined as patients found to have chronic kidney disease stage IV with or without dialysis dependence as extracted from manual chart review. PVD was defined as clinically significant peripheral arterial disease as documented in the electronic medical record. We also included echocardiographic data, including LVEF and degree of mitral regurgitation.
Data are expressed as mean±SD for continuous variables, and absolute value and percentage for categorical variables, unless otherwise stated. Differences between two groups were assessed by using Χ2 test for categorical variables and student’s t-test for continuous variables, unless cell frequencies were <5, in which case we used Fisher’s exact test for categorical variables. We used forward selection to identify the variables that met significance at a p<0.15 cut-off, and then included the retained variables in survival analysis. We forced age into the final model, given known association with mortality. Survival was analysed by using the Kaplan-Meier and Cox proportional hazards methods. Due to the small sample size, we adjusted survival rates using corrected group prognosis method, which calculates the survival curve for each individual using a fitted Cox model.11 12 The average survival is then calculated as a weighted average of the individual survival curve. Results are reported as HRs with 95% CIs and p values. Finally, we tested the proportional hazards assumption in the model with a supremum test and assessed discrimination with the C-index. Statistical analysis was performed by using SAS V.9.4. Figures were generated using SAS and R software.
The need for patient consent was waived due to the retrospective, observational nature of the study.
Initially, 2 54 614 echocardiograms were considered, representing 1 45 785 unique patients. Of the 2 54 614 echocardiograms, 884 echocardiograms (0.3%) had severe AR. Of those 884 echocardiograms, 43/884 (4.9%) had LVEF <35%. One echocardiogram was excluded because of a pseudoaneurysm of the aortic root as the primary indication for surgery, and was excluded per our exclusion criteria. Two other echocardiograms were excluded because they did not represent unique patients. After the exclusion criteria were applied, 40 echocardiograms representing 40 unique patients remained for the analysis. Of the 40 included patients, 18 (45.0%) underwent surgery and 22 (55.0%) were managed medically.
Characteristics of surgical and non-surgical patients
Baseline patient characteristics are summarised in table 1. The medically managed patients were older than the surgical group (72 vs 59, p=0.018) and were more likely to carry a diagnosis of hypertension (20 vs 8, p=0.0014). The surgical patients had higher creatinine than the group that was medically managed (1.6 vs 1.2 mg/dL, p=0.068). The proportion of patients with CAD was no different in the surgical and medical management group (6% vs 9%, p=1.000). Additionally, NYHA class III or IV, diabetes and additional significant valvular disease were not statistically different between the surgical and medically managed groups. Specific therapies for heart failure, including beta-blockers, ACE inhibitors/ARBs, mineralocorticoid receptor antagonists such as eplerenone or spironolactone, ICD therapy and CRT therapy were not statistically significant between the surgical and medically managed groups. In 16/40 patients (40%), the cause of the low LVEF was deemed related to the severe AR, with no difference between surgical and medically managed groups (p=0.90). In 4/40 patients (10%), frailty was considered in the decision about whether to operate, with no statistically significant difference between surgical and medically managed groups (p=0.61).
Among the 18 surgical patients, 15 patients had traditional isolated AVR. Among the three other operative patients, one had a composite aortic root replacement, one had an AVR with ascending aortic graft and one had a Ross procedure.
Validation of severity of aortic regurgitation
Of the 40 patients, 25/40 (62.5%) had echocardiogram images available to review. The mean vena contracta width was 8.2 mm (SD 2.1), the mean percentage of the left ventricular outflow tract was 72.7% (SD 8.8%), the mean pressure half-time was 263 ms (SD 118 ms) and the mean velocity of flow reversal at end-diastole was 25.5 cm/s (SD 8.2). Of the 18/25 patients who had abdominal aortic flow well visualised, 17 (94.4%) had holodiastolic flow reversal in the abdominal aorta. All of the 25 patients had severe AR as classified by meeting criteria for at least two of the four parameters adjudicated above. Of the 25, 6 patients (24.0%) also had some reported degree of aortic stenosis.
Survival for surgical and non-surgical patients
Mean follow-up for all patients was 6.58 years (SD 4.51). The mean survival for patients undergoing AVR was 6.3 years (SD 5.3 years), while the mean survival for patients with no surgery was 0.5 years (SD 1.0 year).
At the time of the end of follow-up, three patients in the surgical group had died. Among those, one died of non-cardiac causes, one died of a ruptured abdominal aortic aneurysm and one died of unknown causes. At the time of the end of follow-up, nine patients in the no surgery group had died. Of those, four died of heart failure, one died of sudden cardiac death and four died of unknown causes. At the end of the last observed event, unadjusted survival rate in the group with no surgery was 8.8%, with a death rate of 91.2%. When adjusting for ESRD, AF, age >65 years and PVD, the survival rate in the group with no surgery increased to 28.7%, with a death rate of 71.3%. Unadjusted survival rate in the group with AVR was 71.5%, with a death rate of 28.5%. When adjusting for ESRD, AF, age >65 years and PVD, the survival rate in the group with AVR increased to 72.2%, with a death rate of 27.8% (table 2).
In the final model, ESRD (HR=17.633, p=0.0335) and peripheral arterial disease (HR=6.050, p=0.0180) were associated with higher risk of long-term mortality. Age 65 or older was not significantly associated with higher risk of long-term mortality (HR=1.789, p=0.5108). AF was also not significantly associated with higher risk of long-term mortality (HR=3.385, p=0.0887). Relative to no surgery, surgery was associated with lower mortality, adjusting for ESRD, AF, age >65 years and peripheral artery disease (HR=0.143, p=0.0490, table 3). The proportional hazards assumption was not violated in the model by the supremum test (p>0.05 for all variables) and the C-index of the model was 0.64, consistent with moderate discrimination. Figure 1 shows unadjusted and adjusted Kaplan-Meier survival curves. Figure 2 shows a forest plot of the HRs.
Despite the recommendations for AVR in patients with severe AR and low LVEF, the surgical risk has been considered by some to be prohibitively high.13 Here we have demonstrated that in a real-world setting, nearly half of patients do not receive surgery. Furthermore, we have demonstrated that the patients turned down for surgery were older, with more peripheral arterial disease and renal failure. However, we also have demonstrated that even after adjustment for these risk factors, the decision to proceed with surgery is associated with a nearly 90% reduction in mortality relative to no surgery. The mean survival for the patients who did not receive surgery was <1 year. All the no surgery patients who later died (for whom a cause of death was available), died of cardiac causes. Although these results are observational and retrospective, and therefore hypothesis-generating, we believe that these results should encourage AVR in these clinical circumstances.13
These findings are critical to informing the operative decision, since this is a rare clinical circumstance without much data. In fact, our results show fewer than 1 out of 20 echocardiograms with severe AR also had low LVEF. Since AR itself is relatively rare, only 40 patients qualified for the analysis out of 1 45 785 patients in the database. Perhaps in the context of this rarity, there is substantial divergence between the guidelines and actual practice. For example, among patients who receive AVR for AR, only 2.7% have an ejection fraction <30%.14 The fact that few AVRs are performed among patients with low LVEF may reflect older analyses demonstrating poor postoperative survival. Similar to our findings, others have reported a 3-year survival of only 64% when the ejection fraction was <50%.5
More recent analyses have also demonstrated that performance of AVR for patients with very low LVEF and severe AR may be a reasonable option. For example, Kaneko et al compared patients who received AVR for severe AR with different levels of ventricular function.7 That analysis demonstrated similar postoperative survival for different levels of ventricular dysfunction. We believe that these results should build enthusiasm for operating on these patients. However, this type of analysis cannot compare the relative benefit of surgery in specific patients with severe AR and impaired LVEF. In particular, cardiac surgeons may have had a higher threshold to operate on sicker patients with impaired LVEF, raising the postoperative survival of that particular group.
For that reason, comparative evidence between surgery and no surgery reduce the selection bias common to analyses of patients who received surgery. In fact, Kamath et al has previously demonstrated a mortality benefit of AVR for these patients compared with patients who did not receive surgery.15 In that analysis, 1-year survival was 88% for patients who received surgery compared with 65% for patients who did not receive surgery. However, in that analysis, 16/53 (30.1%) of the patients had a concomitant mitral valve replacement or repair, and the primary valvular lesion affecting ventricular function may not have been the aortic valve. Furthermore, 17/53 (32.1%) patients had concomitant coronary artery bypass grafting, which can improve ventricular function postoperatively. By demonstrating an association between AVR and long-term survival in a cohort that includes isolated AVR only, our results here are compatible with and extend these results.
In that context, we believe that these results suggest that even when AR is likely the cause of depressed ventricular function, the decision to proceed with AVR is associated with mortality benefit. While any decision to operate on high-risk patients must involve shared decision-making regarding relative risks and benefits, our analysis has shown an improvement over time in survival with surgery for these patients. In this contemporary, real-world cohort, surgery was associated with a greater than fivefold increase in length of survival relative to no surgery.
Our study should be interpreted in the setting of important limitations. First, given that this is a rare clinical scenario, our sample size is small and therefore, we were not able to ensure that our results are generalisable to the overall population, due to the chance of type I statistical error. Further studies with larger sample sizes are needed to investigate the association between choice of surgery and mortality. This study is a retrospective cohort study and as such may contain selection bias. We attempted to minimise selection bias on survival with corrected group prognosis method with comorbidities as covariates, but unmeasured confounders are still possible including clinical judgement and frailty.16 Given the retrospective nature of our study, we were not able to include the frailty index of the patients as it was not routinely charted in the electronic medical record. We are reassured, however, that only a tenth of patients in the analysis had chart-reviewed frailty considered in the operative decision and there were not statistically significant differences between groups with respect to frailty playing a role in the decision about whether to operate. Finally, it should be noted that as a single-centre study, the external validity of our results is uncertain. In particular, as a high-volume Heart Valve Centre of Excellence, outcomes in this rare condition from our hospital may be different than outcomes at other hospitals.17
Even after adjustment for clinical characteristics and patient age, surgical AVR is associated with higher survival for patients with low LVEF and severe AR. Although treatment selection bias cannot be completely eliminated with this retrospective cohort design, these results provide some evidence that surgery may be associated with prolonged survival even in this high-risk patient group. Although our findings demonstrate a statistically significant association between survival and surgery in this population, the data must be interpreted with caution, given the small sample size and the retrospective nature of the study.
What is already known on this subject?
Severe symptomatic aortic regurgitation is a class I indication for aortic valve replacement (AVR), yet the efficacy of AVR in high-risk patients with low left ventricular ejection fraction (LVEF) is not well established.
What does this study add?
For patients with severe aortic regurgitation and low LVEF, absolute mortality is 27.8% for surgical patients and 91.2% for non-surgical patients with a mean follow-up of 6.58 years. AVR was associated with lower mortality (HR=0.143, p=0.0490) relative to no surgery. The mean survival for patients undergoing AVR was 6.3 years (SD 5.3 years), while the mean survival for patients with no surgery was 0.5 years (SD 1.0 year).
How might this impact on clinical practice?
These results provide some evidence that AVR is associated with increased survival in this high-risk, but relatively rare clinical condition. The data must be interpreted with caution, given the small sample size and the retrospective nature of the study.
Contributors Conception and design of study, AGF, JHW. Acquisition of data, AGF, VB, MMW, JHW. Analysis/interpretation of data, AGF, VB, JHW, MMW. Drafting the manuscript, AGF, VB, EL, MP, MMW, GT, SM, TMS, JHW. Revising the manuscript critically for important intellectual content, AGF, VB, JHW. Approval of the version of the manuscript to be published, AGF, VB, EL, MP, MMW, GT, SM, TMS, JHW.
Funding JHW is supported by a career development award from the National Institutes of Health through Harvard Catalyst (KL2 TR001100).
Competing interests None declared.
Ethics approval Partners Healthcare IRB.
Provenance and peer review Not commissioned; externally peer reviewed.
Correction notice Since this paper was first published online the image labelled figure 1 has now been updated to figure 2. The image which was published as figure 2 is now figure 1. Neither of the figure legends have been altered.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.