Author + information
- Received May 26, 2006
- Revision received October 25, 2006
- Accepted November 1, 2006
- Published online April 3, 2007.
- Sunil K. Bhudia, MD⁎,
- Patrick M. McCarthy, MD⁎,⁎ (, )
- Ganesh S. Kumpati, MD⁎,
- Joe Helou, MD⁎,
- Katherine J. Hoercher, RN⁎,
- Jeevanantham Rajeswaran, MSc† and
- Eugene H. Blackstone, MD⁎,†
- ↵⁎Reprint requests and correspondence to:
Dr. Patrick M. McCarthy, Co-Director of the Bluhm Cardiovascular Institute, Chief of Cardiothoracic Surgery Division, and Professor of Surgery at the Feinberg School of Medicine Northwestern University, Division of Cardiothoracic Surgery, 201 East Huron Street, Suite 11-140, Chicago, Illinois 60611-29968.
Objectives Among patients undergoing aortic valve surgery for chronic aortic regurgitation (AR), we sought to: 1) compare survival among those with and without severe left ventricular dysfunction (LVD); 2) identify risk factors for death, including LVD and date of operation; and 3) estimate contemporary risk for cardiomyopathic patients.
Background Patients with chronic AR and severe LVD have been considered high risk for aortic valve surgery, with limited prognosis. Transplantation is considered for some.
Methods From 1972 to 1999, 724 patients underwent surgery for chronic AR; 88 (12%) had severe LVD. They were propensity matched to patients with nonsevere LVD to compare hospital mortality, interaction of operative date with severity of LVD, and late survival. Propensity score-adjusted multivariable analysis was performed for all 724 patients to identify risk factors for death.
Results Survival was lower (p = 0.04) among patients with severe LVD than among matched patients with nonsevere LVD (30-day, 1-, 5-, and 25-year survival estimates were 91% vs. 96%, 81% vs. 92%, 68% vs. 81%, and 5% vs. 12%, respectively). However, survival of patients with severe LVD improved dramatically across the study time frame (p = 0.0004): hospital mortality decreased from 50% in 1975 to 0% after 1985, and time-related survival in patients with severe LVD operated on since 1985 became equivalent to that of matched patients with nonsevere LVD (p = 0.96).
Conclusions Neutralizing risk of severe LVD has improved early and late survival such that aortic valve surgery for chronic AR and cardiomyopathy is no longer a high-risk procedure for which transplantation is the best option.
Guidelines for managing aortic regurgitation (AR) in patients with normal left ventricular (LV) function generally are clear (1,2). However, for patients with AR-induced cardiomyopathy, manifested at times by massive ventricular dilation from volume overload, management guidelines reflect conflicting evidence, and operative results have been difficult to predict (1). Heart transplantation has sometimes been suggested (3).
Previous studies have identified greater severity of LV dysfunction (LVD) as a risk factor for poor long-term outcome among patients treated both surgically and medically (4–6). However, recent reports of mitral valve surgery in patients with both severe regurgitation and severe LVD (also historically considered a high-risk procedure) have demonstrated good early and late survival, relief of symptoms, and improved LV geometry (7,8). In light of these results, we reexamined valve surgery for AR in patients with severe LVD. Specifically, among patients undergoing aortic valve surgery for chronic AR, we sought to: 1) compare survival among those with versus without severe LVD; 2) identify risk factors for death, including LVD and operative date; and 3) estimate the contemporary risk of aortic valve surgery for the cardiomyopathic patient.
Patients and Methods
From January 1972 to January 1999, 724 patients underwent primary isolated aortic valve surgery (replacement, n = 722; repair, n = 2) for chronic AR, with or without concomitant repair of the aorta, at Cleveland Clinic. Preoperative, operative, and postoperative variables and preoperative echocardiographic variables were retrieved from the prospective Cardiovascular Information Registry and Echocardiography Database, respectively. Both were approved for use in research by the institutional review board.
The study population comprises patients with moderately severe or severe AR of at least 3 months’ duration, including ones with chronic aortic dissection (all had diagnosis of dissection 3 months or more before surgery), coexisting ascending aortic aneurysm, or connective tissue diseases (Marfan syndrome and cystic medial necrosis).
Patients not having primary, isolated aortic valve surgery or aortic valve surgery with concomitant repair of the aorta were excluded. Also excluded were patients who had mixed aortic stenosis and regurgitation (because the fundamental pathophysiology of valves with stenosis is different than for pure regurgitation), previous chest radiation, unknown duration of aortic dissection, or endocarditis within 3 months of operation (defined as blood-culture positive or systemic sepsis with the clinical syndrome of infective endocarditis, culture-positive valvar vegetations, or antibiotic treatment for diagnosed or suspected endocarditis).
Of the 724 patients, 88 (12%) had a LV ejection fraction of 30% or less on preoperative left ventriculogram or echocardiogram; they constituted the severe LVD group. The remaining patients (n = 636) had LV ejection fraction greater than 30% and were designated the nonsevere LVD group. A subgroup of patients from 1986 onward had preoperative echocardiographic measurements of structure and function, and these revealed the expected larger LV diastolic (7.5 ± 0.7 cm vs. 6.4 ± 1.0 cm, p < 0.0001) and systolic (5.9 ± 0.79 cm vs. 4.2 ± 0.89 cm, p < 0.0001) dimensions of the severe LVD group, reduced fractional shortening (0.22 ± 0.07 vs. 0.34 ± 0.08, p < 0.0001), but similar wall thickness (1.2 ± 0.23 cm vs. 1.2 ± 0.24 cm, p = 0.5) (Appendix A).
Characteristics of these groups and their operation are contrasted in Table 1.In patients with severe LVD, mean age was 56 ± 12 years, and 80 (91%) were men. The most common etiologies of AR were degeneration (59%) and anuloaortic ectasia without aneurysm (28%). In contrast, patients with nonsevere LVD were younger (mean age 50 ± 15 years), and a smaller proportion were men (67%), but the most common etiologies of AR were similar: degeneration in 49% and anuloaortic ectasia in 30%. Both groups had similar prevalence of aorta repairs.
Before 1985, the proportions of mechanical (n = 165 of 352 replacements, 47%) and biological (n = 187, 53%) prostheses implanted were nearly equal. Since 1985, biological prostheses have been implanted somewhat more frequently (n = 211 of 370 replacements, 57%); however, 80 of these were allografts, which were not used before 1985.
Patients were followed routinely every 2 years after surgery and cross-sectionally in 2001 to determine functional status, vital status, and whether they had undergone cardiac reoperation. For patients who were untraced, the Social Security Death Index was used to estimate vital status (9,10).
Follow-up extended reliably to 25 years, with 5,785 patient-years of information available for analysis. Among survivors, median follow-up was 6 years (mean 8.3 ± 6.5 years), with 75% followed for more than 3 years, 25% more for 13 years, and 10% more than 19 years.
Data analysis addressed the following questions: 1) What is the difference in time-related survival after aortic valve surgery between patients with versus without severe LVD, but who are otherwise comparable? 2) What are the risk factors for mortality? Specifically, after adjusting for other risk factors, did LVD and date of surgery affect early versus long-term mortality differently? 3) What is the impact of LVD on mortality in the current era?
Characteristics of patients with severe LVD differed in many ways from those of patients with nonsevere disease (Table 1). Therefore, to make a fair survival comparison given the impossibility of randomizing patients to having severe LVD or not, we used propensity matching (11,12). In brief, multivariable logistic regression analysis was used to identify factors associated with severe LVD (Table 2,Appendix B), and to these were added variables representing groups of demographic, symptomatic, and comorbid factors that might be related to unrecorded selection factors (saturated model). The c statistic was 0.84. A propensity score was calculated for each patient by solving the saturated model for the probability of being in the severe LVD group. Propensity-matched pairs were then selected using a greedy matching strategy. This yielded 77 well-matched pairs of patients (Appendix C).
Nonparametric survival estimates were obtained using the Kaplan-Meier method (13). A parametric method was used to resolve the number of phases of instantaneous risk of death (hazard function) and to estimate shaping parameters of each (14). Survival comparison was performed in the hazard function domain for propensity-matched patients.
Risk factors for mortality
A multivariable analysis of all 724 patients was performed simultaneously for each hazard phase, with an indicator for severe LVD forced into each phase (Appendix B). Bootstrap bagging using 250 resampled data sets was then used to identify other risk factors appearing in at least 50% of the models at p ≤ 0.05 (15). Because of the limited number of events, the propensity score for each patient was then also forced into the model to further adjust for known differences between patients with and without severe LVD (16).
Severe LVD in the current era
The influence of a risk factor, such as severe LVD, on survival can be neutralized by general improvements in surgical and medical care of patients across time or by direct modulation of its influence. We quantified the latter by examining the interaction of severe LVD with date of surgery. Such an analysis permits estimation of patient prognosis in the current era by solving the resulting multivariable equation for the patient’s risk factors, but substituting a contemporary date of operation.
To confirm and amplify understanding of the influence of date of operation, we also examined hospital mortality and long-term survival in propensity-matched pairs operated on before and after January 1, 1985 (35 patients with severe LVD and 37 with nonsevere LVD underwent surgery after that date).
Simple descriptive statistics were used to summarize the data. Categorical data are summarized as frequencies and percentages, with comparisons made using chi-square and the Fisher exact tests. Continuous variables are summarized as mean ± standard deviation, with comparisons made using the Wilcoxon rank-sum nonparametric test. Analyses were performed using SAS statistical software (version 9.1, SAS Institute, Cary, North Carolina). Uncertainty is expressed by confidence limits (CLs) equivalent to ±1 standard error (68%).
Survival comparison across entire study time frame
Survival after aortic valve surgery across the entire study time frame was worse among propensity-matched patients with severe LVD than otherwise-similar patients with nonsevere LVD (81% vs. 92% at 1 year, 68% vs. 81% at 5 years, 46% vs. 62% at 10 years, 26% vs. 41% at 15 years, 12% vs. 24% at 20 years, and 5% vs. 12% at 25 years, respectively) (p [log rank] = 0.04) (Fig. 1A).Difference in survival was considerably greater in unadjusted comparisons (Appendix Figure 1). The hazard ratio comparing patients with severe LVD versus nonsevere LVD was greater than unity both early and late after surgery; although the ratio was >2 early after surgery, its confidence limits suggest an increased proportional hazard across years of follow-up (Fig. 1B).
Risk factors for mortality
Among all 724 patients, severe LVD was a risk factor for early and late mortality and remained so after adjusting for propensity score (Table 3).Early date of surgery within the series was a risk factor for early mortality, and risk of death decreased for both severe (p = 0.0004) and nonsevere (p < 0.001) LVD patients over the study time frame. The decline in risk was similar (p = 0.3) for both groups.
Severe LVD in the current era
Hospital mortality among propensity-matched patients undergoing surgery before 1985 was 17% in the severe LVD group (7 of 42; CL 11% to 25%) and 3% in the nonsevere LVD group (1 of 40; CL 0.3% to 8%; p = 0.03). From 1985 onward, hospital mortality was 0% for both groups (0 of 35 severe LVD patients, 0 of 37 nonsevere LVD patients, both CL 0% to 5%; p = 1) (Fig. 2).
In large part because of the opportunity for more substantial decline of early risk in the severe LVD group, 10-year survival in propensity-matched patients who underwent surgery in 1973 was 20%, increased to 49% by 1985, and reached 60% by the mid-1980s. Thus, by the mid-1980s, survival in this group became equivalent (widely overlapping confidence limits) to that of non-severe LVD patients (Fig. 3).
Long-term survival in patients who underwent surgery in 1985 and after
Among propensity-matched patients operated on in 1985 and beyond, survival at 1, 5, and 10 years after surgery was 92%, 79%, and 51% for severe LVD patients and 96%, 83%, and 55% for nonsevere LVD patients, respectively (p > 0.9) (Fig. 4).
Previous studies have shown that long-term results of aortic valve replacement are poor in patients with preoperative severe LVD (17,18). We also found that long-term survival across the entire study time frame was worse in those with severe LVD.
However, because our study extended from 1972 to 1999, we were also able to investigate the influence that date of operation (with coincident improvements in medical and surgical management and valvar prostheses) had on early and late survival. Long-term survival of patients with severe LVD improved progressively and by about 1985 was comparable with that of patients with nonsevere LVD; hospital mortality decreased to a comparable value. Early and late survival was already good and did not change appreciably over the experience in patients with better LV function.
Furthermore, severe LV dysfunction was associated with extremely enlarged ventricles and large volumes, with mean LV diastolic dimension of 7.5 cm and end-systolic dimension of 5.9 cm, as noted in Appendix A. Tables 1 and 2present other markers for high-risk patients. Clinicians should be aware that patients with these characteristics can recover after surgery.
Implications for therapy
Guidelines for managing patients with AR and normal LV function generally are clear, but for patients with severe LVD, they reflect conflicting evidence because the results of surgery have historically been poor (19). It is recommended that patients with AR and in New York Heart Association functional class III or IV with preserved LV function, those in functional class II with progressive LV dilation or angina, and asymptomatic patients with mild-to-moderate LVD should undergo aortic valve surgery. Surgery also should be performed in patients with moderate-to-severe AR undergoing a concomitant cardiac surgery. However, patients with severe LVD and symptomatic patients with advanced LVD have presented difficult management issues (19).
Unfortunately, patients with AR may not present until they have advanced heart failure with severe LVD and ventricular dilation. The natural history of such patients (and even of patients with AR and preserved LV function) is limited (20). For some, heart transplantation has been considered because of perceived high perioperative mortality and poor late survival after aortic valve surgery (3). However, we have shown in this study that aortic valve surgery in patients with chronic AR and cardiomyopathy is no longer a high-risk procedure for which transplantation is a better option. Neutralization of the effect of severe LVD was accomplished by improved early results of aortic valve surgery in general, not by specific improvements for these patients alone.
There are many reasons that patients fare better after surgery in the current era. Medical management of LV dysfunction and heart failure improved substantially since the earliest time frame of this study, with use of angiotensin-converting enzyme inhibitors and beta blockers becoming routine. Intraoperatively, myocardial protection became more sophisticated for patients with LV dysfunction, with antegrade and retrograde cold-blood cardioplegia becoming routine. Intraoperative transesophageal echocardiography now allows the surgeon to reduce complications, such as atheroembolism from aortic calcification, and aids in optimizing inotrope management of LV function and removal of intraventricular air. Compared with the Starr-Edwards high-gradient prosthesis used in the early days, replacement heart valves are considerably improved, with low-gradient bioprostheses and mechanical valves readily available. Perioperative management now includes use of newer inotropes such as the phosphodiesterase inhibitor milrinone. Finally, postoperative optimal medical therapy and device options, such as biventricular synchronous pacing and implantable cardioverter-defibrillators, are now available to improve mid-term survival.
This was a clinical study with data collected from our prospective clinical registries. It thus reflects the inherent bias of such studies, including referral biases and patient selection bias. Because LV dysfunction is not randomizable, we used propensity adjustment in multivariable modeling and matching to reduce uncontrollable natural selection bias. Nonsignificance of the propensity score in early hazard indicates that usual clinical variables sufficed for this, but not for late hazard, where it provided important bias reduction over and above that provided by individual clinical variables.
Duration of preoperative AR could not be determined and conceivably differed between groups; perhaps the severe LVD group had more long-standing AR. This was also a single-institution study whose results may not be generalizable. Unfortunately, we have been unable to identify which of many factors contributed to lowering of mortality, but we believe they include those discussed above.
Yet, it is conceivable that the progressive decline in early mortality, particularly from the 1970s through the mid-1980s, could be an artifact of temporal changes in data quality, definitions, or methods of assessment of LV function (echocardiogram vs. ventriculogram). We believe this is not the case. First, although there is imperfect correlation of echocardiography and ventriculography, most of the decline in mortality occurred in the era of routine ventriculography. Second, there has been an ongoing, active quality assurance and improvement process for the primary registry used in this study. Although specific temporal changes in the database, generally corresponding to discrete changes in American College of Cardiology/Society of Thoracic Surgeons definitions, have occurred, we have not observed temporal drifts.
In addition, the most dramatic decline in mortality has been that which occurrs in-hospital. Although hospital mortality is a hard, completely documented end point, it does not encompass the entire period of early surgical risk. This phenomenon has been long recognized by surgeons, as detailed in all 3 editions of Cardiac Surgery(21). It stimulated Blackstone and colleagues to develop their temporal decomposition methods for time-related analysis in the early 1980s, which is a nonproportional hazards model that can identify risk factors that dominate early after surgery, and possibly different ones that modulate late risk (15,22). The use of this model leads to a more medically meaningful analysis of postintervention events than does a proportional-hazards model. Thus, hospital mortality, which is surely influenced by ever-shorter length of hospital stay and underestimates early risk even more in serious conditions than in straightforward cardiac surgery, is surely not an ideal end point (23). For this reason, the primary analysis for this study has used all deaths from surgery through end of follow-up and the temporal decomposition method to identify associations of severe LVD and date of operation with mortality in each phase of hazard.
Although severe LVD has been a risk factor for reduced early and late survival after aortic valve surgery for AR, it has been neutralized in this series since 1985. Patients should still be referred for surgery earlier in their clinical course, or per American Heart Association/American College of Cardiology guidelines (19). However, for the patient who presents late with AR and severe LVD, surgery is the preferred treatment and can be performed with acceptable risk and late survival.
For Appendices A, B, and C, please see the online version of this article.
- Abbreviations and Acronyms
- aortic valve regurgitation
- confidence limits
- left ventricular
- left ventricular dysfunction
- Received May 26, 2006.
- Revision received October 25, 2006.
- Accepted November 1, 2006.
- American College of Cardiology Foundation
- Dujardin K.S.,
- Enriquez-Sarano M.,
- Schaff H.V.,
- Bailey K.R.,
- Seward J.B.,
- Tajik A.J.
- Bishay E.S.,
- McCarthy P.M.,
- Cosgrove D.M.,
- et al.
- Boyle C.A.,
- Decoufle P.
- Newman T.B.,
- Brown A.N.
- Rosenbaum P.R.,
- Rubin D.B.
- Drake C.,
- Fisher L.
- Klodas E.,
- Enriquez-Sarano M.,
- Tajik A.J.,
- Mullany C.J.,
- Bailey K.R.,
- Seward J.B.
- Bonow R.O.,
- Carabello B.A.,
- Kanu C.,
- et al.
- Kouchoukos N.T.,
- Blackstone E.H.,
- Doty D.B.,
- Hanley F.L.,
- Karp R.B.
- Kouchoukos N.T.,
- Blackstone E.H.,
- Doty D.B.,
- Hanley F.L.,
- Karp R.B.
- Osswald B.R.,
- Blackstone E.H.,
- Tochtermann U.,
- Thomas G.,
- Vahl C.F.,
- Hagl S.