Mortality and Morbidity of Aortic Regurgitation in Clinical Practice
A Long-Term Follow-Up Study
Background—The outcome of aortic regurgitation conservatively followed in clinical practice is poorly defined.
Methods and Results—Long-term outcome of 246 patients with severe or moderately severe aortic regurgitation diagnosed by color Doppler echocardiography was analyzed. With conservative management, mortality rate was higher than expected (at 10 years, 34±5%, P<0.001) and morbidity was high (10-year rates of 47±6% for heart failure and 62±4% for aortic valve surgery). At 10 years, 75±3% of patients had died or had surgery and 83±3% had had cardiovascular events. In multivariate analysis, predictors of survival were age (P<0.001), functional class (P<0.001), comorbidity index (P=0.033), atrial fibrillation (P=0.002), and left ventricular end-systolic diameter corrected for body surface area (P=0.025). Ejection fraction was also an independent predictor of overall survival, including postoperative follow-up of surgically treated patients (P<0.001). High risk during conservative treatment, with mortality rate in excess of that expected, was noted among patients with severe, even transient, symptoms (24.6% yearly, P<0.001) but also in those with mild (class II) symptoms (6.3% yearly, P=0.02) and in asymptomatic patients with left ventricular ejection fraction <55% (5.8% yearly, P=0.03) or with end-systolic diameter normalized to body surface area ≥25 mm/m2 (7.8% yearly, P=0.004). Surgery performed during follow-up was independently associated with reduced cardiovascular mortality (adjusted hazard ratio, 0.54; P=0.048).
Conclusions—Patients diagnosed with severe aortic regurgitation in clinical practice incur excess mortality and high morbidity, underscoring the serious prognosis of the disease. Surgery, which reduces cardiac mortality rates, should be considered promptly in high-risk patients.
In patients with valvular heart disease, clinical decision making is based largely on the risk of mortality and morbidity incurred by patients conservatively treated,1 and in subsets of patients with excess mortality rates, an aggressive approach is often considered justified.
In patients with severe aortic regurgitation (AR), survival with conservative treatment is poorly defined.2 Previous studies of the natural history of AR reported variable mortality rates, ranging at 5 years between 2% and 62%.3 4 Such disparate results are due to selection biases, small populations, poorly defined degrees of AR, and the confounding effect of associated diseases5 and are difficult to reconcile and use for clinical decision making. Furthermore, the cohorts with low mortality rates included patients much younger3 6 7 than those treated in clinical practice,8 9 10 and the risk of excess mortality in comparison with expected survival is unknown.
Analysis of determinants of survival with conservative treatment is essential to define high-risk groups that require aggressive treatment. Although symptoms caused by AR improve after surgery8 11 and are widely accepted indications for surgery,12 13 14 their independent impact on outcome is unclear. Previous studies suggested that asymptomatic patients have a good outcome.3 7 However, uncertainty persists about the fate of patients with minimal symptoms15 and about subsets of asymptomatic patients who may be at high risk. Furthermore, the impact of left ventricular (LV) function at diagnosis on survival under conservative management is unclear. Therefore, the concept of surgical correction of AR indicated solely for alterations of LV function, although appealing,14 has been challenged12 13 because it is supported only by postoperative outcome results in mostly old series11 16 that recently have been contradicted.12 17 18
Because of their inherent selection criteria, the series on natural history of AR were either small4 19 20 21 or observed few deaths,3 7 22 lacked power, could not define predictors of survival, and provided limited information on morbidity in patients conservatively treated after the diagnosis of AR.3 4 6 19 20 21 22 These uncertainties are sources of doubt about which subsets of patients may benefit most from an aggressive surgical approach8 or from treatment with vasodilators.23 24
These limitations can be addressed with Doppler echocardiography, which allows comprehensive assessment of LV function25 and degree of AR26 and offers the unique opportunity to define a large cohort of patients diagnosed with severe or moderately severe AR in routine clinical practice.
The study patients were those in whom moderately severe (grade III/IV) or severe (grade IV/IV) AR was first diagnosed with color flow Doppler echocardiography26 between 1985 and 1994 at the Mayo Clinic. Exclusion criteria were aortic dissection, previous valve surgery, and associated mitral valve or congenital heart disease. Patients were not excluded on the basis of treatment received. Symptoms at baseline were those occurring within 1 month before diagnosis. The patients were not followed exclusively at our institution and received care by their attending physician. Information on events after diagnosis was obtained for all patients between February and July 1996. The events and cause of death were established by review of medical, coroner, and autopsy records and death certificates. Associated comorbid conditions at baseline were summated as a comorbidity index.5
All measurements and gradations were collected in routine clinical practice prospectively and transferred unaltered electronically. The degree of AR was graded semiquantitatively by color flow Doppler on a scale of 1+ to 4+.26 Cardiac diameters were measured by 2-dimensional echocardiography–guided M-mode and indexed to body surface area. LV end-systolic wall stress27 and ejection fraction (EF)25 28 were calculated.
Continuous variables were expressed as mean±1 SD and survival as observed±SE. After echocardiographic diagnosis, the rates of events were estimated by the Kaplan-Meier method and linearized yearly rates. For analysis of outcome with medical treatment, patients operated on were censored at surgery, but the entire follow-up was used to analyze the effect of surgery on outcome. The 1-sample log-rank test was used to compare survival with expected survival of the age- and sex-matched 1990 US white population. Comparison of survival between subgroups of patients was based on the k-sample log-rank test. Baseline predictors of outcome were identified by proportional hazards regression analysis. To determine the effect on survival of events occurring after diagnosis (surgery, congestive heart failure), a time-dependent proportional hazards analysis was performed within a multivariate model that included the baseline predictors of survival.
The impact of potential referral biases (geographical origin and comorbidity) was tested in multivariate analysis. The analysis was also repeated in patients who were not rapidly (within 1 month) referred to surgery or were initially in functional class I or II. A probability value <0.05 was considered statistically significant.
The clinical characteristics of the 246 patients who met the inclusion criteria are listed in Table 1⇓. Among the 122 patients without dyspnea (class I), 9 had angina or syncope; therefore, 113 were strictly asymptomatic. The presumed cause of AR was degenerative in 99 patients, congenital in 69, aortic root enlargement (with or without aneurysm) in 46, rheumatic in 15, aortitis in 8, endocarditis in 7, and miscellaneous in 2. The AR was grade 3+ in 160 (65%) and grade 4+ in 86 (35%) and was confirmed in 61 (97%) of 63 patients undergoing aortography. Medications received for at least 6 months or the duration of medical follow-up included diuretics in 89 patients, angiotensin-converting enzyme inhibitors in 71, calcium channel blockers in 41, hydralazine in 4, β-blockers in 34, digoxin in 103, and none of these in 73.
Survival in Conservatively Managed Patients
The mean follow-up was 7±3 years (3.7±3.8 years for conservative management) (Table 2⇓). Of the 246 patients, 43 died during medical follow-up, and long-term survival was worse than expected survival (P<0.001) (Figure 1⇓). The cause of death was cardiovascular in 33 patients (intractable heart failure in 20, sudden death in 6, cerebral thromboembolism or hemorrhage in 3, and aortic dissection in 4) and noncardiac in 10. Cardiac deaths at 5 and 10 years were 18%±3% and 27%±5%, respectively.
In multivariate analysis (Table 3⇓), the baseline variables independently predictive of survival with medical treatment were age, New York Heart Association (NYHA) class, comorbidity, LV systolic dimension corrected for body surface area (LVS/BSA), and atrial fibrillation.
Symptoms were major predictors of survival (Table 3⇑ and Figure 2⇓). Survival was significantly lower in symptomatic than in asymptomatic patients (P<0.001). Among patients in class III or IV at baseline, 46 (77%) of 60 ultimately underwent surgery, but 14 remained medically treated because of refusal of surgery (4 patients) or high risk (1 patient) and because functional improvement occurred with treatment in all patients. Despite this improvement, survival was dismal (yearly mortality rate 24.6%), worse than expected survival (P<0.001), and worse than survival of patients in class I or II (P<0.001). Patients with mild dyspnea (class II) had survival worse than that of patients in class I (P=0.04) and displayed excess mortality rates compared with expected survival (6.3% yearly, P=0.02). In these patients, 10 (83%) of 12 deaths were cardiac. The 10-year mortality tended to exceed expected mortality, but the difference did not reach statistical significance in patients without dyspnea (class I) (25%±5%, P=0.38) or in strictly asymptomatic patients (24%±5%, P=0.37).
LV measurements at baseline were also predictive of survival. EF and LVS/BSA were the only independent predictors of survival. Excess mortality rate compared with expected rate was observed with EF <50% (at 10 years, 74%±10%, P<0.001) and with EF of 50% to 55% (at 10 years, 35%±12%, P=0.039). Patients with EF ≥55% had a lower mortality rate (P<0.001), not different from expected (at 10 years, 22%±5%, P=0.73). Survival with conservative management stratified according to LVS/BSA (with a threshold of 25 mm/m2) is presented in Figure 3⇓. However, in multivariate analysis, LVS/BSA was the strongest predictor of survival with conservative management (Table 3⇑, model 1), whereas EF was the strongest predictor of overall survival, including postsurgical survival (Table 3⇑, model 3). Asymptomatic patients with EF <55% or with LVS/BSA ≥25 mm/m2 had excess mortality rates compared with those without these LV characteristics (P=0.022 and P<0.001, respectively) and compared with expected survival (P=0.03 and P=0.004, respectively) (Table 2⇑).
Morbidity in Medically Treated Patients
The morbidity in medically treated patients is noted in Table 2⇑ and Figure 4⇓. Predictors of heart failure are noted Table 3⇑. Heart failure was associated with high subsequent mortality (adjusted hazard ratio 15.4 [95% CI 7.5 to 31.8]; P<0.001). Vascular complications occurred in 13 patients (thromboembolism in 9 and aortic dissection in 4) after diagnosis of AR (1.5% yearly) (Figure 4⇓).
Aortic Valve Surgery
Surgery was performed in 132 patients (110 at our institution and 22 elsewhere). Indications for surgery were determined by attending physicians and were symptom class III or IV in 84 patients (dyspnea in 79, angina pectoris in 5), asymptomatic LV dysfunction in 17 or enlargement in 8, aortic aneurysm in 14, infective endocarditis in 5, and physician’s preference in 4. Coronary angiography was performed in 95 patients and coronary bypass grafting in 24. Operative mortality was 3.8% (5/132). The cumulative likelihood (Figure 4⇑) of aortic valve surgery and combined end points are noted in Table 2⇑.
Analysis of the effect of surgery on survival showed that in a time-dependent multivariate analysis, after adjustments for age, EF, and symptoms at baseline, surgery during follow-up was associated with a trend toward reduction of overall mortality (adjusted hazard ratio 0.69; P=0.20) and a significant reduction of cardiovascular mortality (adjusted hazard ratio 0.54 [95% CI 0.23 to 0.99]; P=0.048) (Table 3⇑, model 4).
Effect of Referral Patterns on Outcome
The local or distant referral had no independent effect on outcome (P=0.72). The comorbidity was low, indicating that the choice of medical rather than surgical treatment was motivated primarily by cardiac status and not by excessive comorbidity.
Among the 187 patients who were not rapidly referred for surgery (within 1 month after diagnosis), the 10-year incidences of surgery and of surgery or death were 51%±5% and 67%±4%, respectively. Among the 186 patients initially in class I or II, the 10-year incidences of heart failure, surgery, and surgery or death were 43%±6%, 55%±5%, and 68%±4%, respectively, showing that even in these patients high complication rates are present.
In this study, after the diagnosis of severe AR, patients followed conservatively in clinical practice incur (1) excess mortality compared with expected, directly related to cardiac disease; (2) excess mortality in patients with severe symptoms even transient, but also in patients with mild, class II symptoms; (3) excess mortality rate even in patients asymptomatic with EF <55% or LVS/BSA ≥25 mm/m2; and (4) high cardiovascular morbidity; but (5) surgical treatment during follow-up is associated significantly with reduced cardiovascular mortality rate.
Death Associated With Severe AR
The natural history of severe AR is poorly defined, with widely disparate reported estimates of long-term survival. Survival rates at 10 years after diagnosis were 96%,3 80%,6 76%,20 62%,29 and 50%.21 Other series reported survival at 8 years as low as 32%4 or, at 5 years, between 66%30 and <20%.19 This wide range of reported survival is difficult to reconcile to provide guidelines for routine practice. This considerable variability in mortality rates may be due to small populations,4 19 20 21 29 30 variable20 or ill-defined6 19 21 29 degrees of AR, and associated comorbidity.5 However, selection bias appears to be an essential reason for this variability. In observational studies,4 19 29 particularly those including patients after catheterization,4 high mortality rates were observed. Conversely, in studies enrolling patients volunteering for prospective follow-up, patients were much younger3 7 15 22 31 than those treated in clinical practice,8 9 10 and barely any cardiac or noncardiac death was observed. This extremely low mortality rate of cohort studies,3 7 15 22 31 much lower than expected mortality rate in the general population, is probably due to the “healthy volunteer effect.” Such limitations warrant analysis of the risk of death incurred by patients who in routine clinical practice are diagnosed with severe AR. Furthermore, recommendations for surgery3 8 13 are not based on natural history studies, in which predictors of survival could not be defined because of small populations4 19 20 21 29 30 or very low death rates.3 22
In the present study, patients diagnosed with severe AR in routine clinical practice incurred, while conservatively managed, excess mortality rates compared with expected, underlining the severity of this condition. This excess mortality rate was not randomly distributed, and independent predictors of survival allow prognostic stratification.
Symptoms were major determinants of survival. Dyspnea class III-IV is an accepted indication for surgery3 13 because of symptomatic improvement after surgery.8 11 Such symptoms, even if transient and relieved by medical treatment, without surgical intervention are associated with a dismal outcome.20 32 Therefore, clinicians should not be deceived by functional improvements and delay the operation. The outcome with mild, class II symptoms has not been clearly defined3 6 15 20 22 but is characterized by excess mortality rates of cardiovascular causes. Therefore, these patients should be considered at high risk and promptly be offered surgical correction of AR, which provides an excellent long-term outcome.18 Even strictly asymptomatic patients display excess mortality rates in subsets defined by LV size and function.
In AR followed conservatively, LV size has been associated with progression of symptoms,3 15 22 but the impact of LV characteristics on survival is uncertain. Their role in the surgical indications has been based on postoperative outcome. Conflicting results showing significant10 11 16 or no significant12 13 17 33 relation to postoperative survival have been reported, and the concept of surgical indications based purely on LV alterations14 has been challenged.12 13 In the present study, LVS/BSA and EF were independent determinants of survival. Of note, the uncorrected LVS was a weaker predictor of outcome than the corrected value.33 34 Although both LVS/BSA and EF are powerful predictors of outcome, the former appears stronger with persistent volume overload under conservative management (Table 3⇑, model 1). Conversely, EF is stronger when postsurgical survival is included (Table 3⇑, model 3), with a large proportion of patients with volume overload relieved by valve replacement. Patients with LVS/BSA ≥25 mm/m2 or EF <55%, even if asymptomatic, have a follow-up mortality rate higher than expected and higher than without these LV abnormalities. These patients should be considered at high risk and evaluated for prompt surgical intervention.
Another important result of the present study is the benefit of surgery performed during follow-up, with a trend toward reduction of total mortality rates and significant reduction of cardiac mortality rates. Short of results from a randomized trial, yet to be conducted, this effect strongly supports surgical indications in subgroups demonstrating excess mortality rates with conservative management.
Follow-Up Events Associated With Severe AR
Cardiac morbidity was also high in the present study, with a 10-year incidence of congestive heart failure of 47%±6% overall and 43%±6% in patients initially in NYHA class I or II. This progression of symptoms is independently determined by LVS/BSA at baseline,3 22 which is another reason to include this variable in risk stratification. Ten years after the diagnosis, low incidences of endocarditis and new atrial fibrillation are observed, but 15% of patients incur vascular complications, 75% either die or undergo surgical treatment,20 and 83% have a fatal or nonfatal cardiovascular event. High morbidity in addition to excess mortality should alert clinicians to the serious prognosis of severe AR, and surgery should promptly be considered on the basis of the patient’s condition at the time of diagnosis.
Methodology of the Study
A natural history study of AR, with medical and surgical treatments withheld, is not conceivable. Conversely, the design of this study allows determination of mortality and morbidity incurred while surgery is deferred. Deferral was motivated not by poor operative risk but mostly by paucity of symptoms, making these results relevant to routine practice.
Vasodilators have been shown to favorably affect LV remodeling24 and to delay surgery23 but have not been shown to improve survival of patients with severe AR.1 In the present study, there was no association between medications received and survival. Without a randomized trial, the effect of surgery on survival is uncertain, but it appears, adjusting for baseline characteristics, to decrease cardiovascular mortality rates.
The LV diameters and EF were measured by echocardiography, were used as originally calculated, and were strong predictors of outcome.10 The grading of AR by color flow Doppler has also been validated26 ; the diagnosis of severe AR was confirmed in 97% of patients undergoing aortography, and all the patients operated on had severe lesions. Therefore, these methods do not represent a limitation of the present study.
The study was not a cohort prospectively followed up but rather a study in clinical practice of patients diagnosed with severe AR. The disadvantage of this design is that progression of LV remodeling cannot be analyzed. The advantages of this design are that the healthy participant bias is avoided and the mortality and morbidity observed are applicable to routine practice. Adjustment for geographic origin, for coexisting conditions, and for direct referrals for surgery did not affect the results, and referral bias is of low likelihood.
Conclusions and Clinical Implications
Severe AR diagnosed in clinical practice is a serious condition complicated both by excess mortality and high morbidity. Within 10 years after diagnosis, 75% of the patients had either died or required aortic valve replacement and 83% had a cardiac event.
Patients at high risk under conservative management include those with (1) severe symptoms, even transient and improved by treatment; (2) mild, class II symptoms; (3) EF <55% or LVS/BSA ≥25 mm/m2 with or without symptoms; or (4) atrial fibrillation. Because surgery during follow-up is associated significantly with reduced cardiac mortality rates, these patients should promptly be considered for surgical correction of AR.
Conversely, strictly asymptomatic patients with an EF ≥55% and LVS/BSA <25 mm/m2 incurred a 10-year mortality rate between 14% and 17%, not significantly different from expected (P>0.80) and form the only subgroup at relatively low risk. The identification among these patients of subsets at higher risk, based on predictors of other events such as sudden death or aortic dissection, will require larger populations in future studies. Currently, in these patients, it appears reasonable to defer surgery and consider vasodilators, which may retard LV remodeling24 and surgery.23 Therefore, the present study is in agreement with the elective approach to surgery3 but defines new criteria for patients at high risk requiring surgery, based on excess mortality rates under conservative treatment.
Dr Dujardin was supported by a fellowship from the Belgian American Educational Foundation. The authors thank Julie Wilson for excellent help with data analysis.
- Received September 10, 1998.
- Revision received December 16, 1998.
- Accepted December 30, 1998.
- Copyright © 1999 by American Heart Association
Bonow RO, Lakatos E, Maron BJ, Epstein SE. Serial long-term assessment of the natural history of asymptomatic patients with chronic aortic regurgitation and normal left ventricular systolic function. Circulation. 1991;84:1625–1635.
Spagnuolo M, Kloth H, Taranta A, Doyle E, Pasternack B. Natural history of rheumatic aortic regurgitation: criteria predictive of death, congestive heart failure, and angina in young patients. Circulation. 1971;44:368–380.
Turina J, Turina M, Rothlin M, Krayenbuehl HP. Improved late survival in patients with chronic aortic regurgitation by earlier operation. Circulation. 1984;70(suppl I):I-147–I-152.
Daniel WG, Hood WP Jr, Siart A, Hausmann D, Nellessen U, Oelert H, Lichtlen PR. Chronic aortic regurgitation: reassessment of the prognostic value of preoperative left ventricular end-systolic dimension and fractional shortening. Circulation. 1985;71:669–680.
Schwarz F, Ehrmann J, Olschewski P, Scheurlen H, Saggau W, Kubler W. Patients with significant aortic incompetence should not be operated on until they are symptomatic. Z Kardiol. 1986;75(suppl 2):133–136.
Bonow RO. Asymptomatic aortic regurgitation: indications for operation. J Card Surg. 1994;9(suppl 2):170–173.
Borer JS, Hochreiter C, Herrold EM, Supino P, Aschermann M, Wencker D, Devereux RB, Roman MJ, Szulc M, Kligfield P, Isom OW. Prediction of indications for valve replacement among asymptomatic or minimally symptomatic patients with chronic aortic regurgitation and normal left ventricular performance. Circulation. 1998;97:525–534.
Henry WL, Bonow RO, Borer JS, Ware JH, Kent KM, Redwood DR, McIntosh CL, Morrow AG, Epstein SE. Observations on the optimum time for operative intervention for aortic regurgitation, I: evaluation of the results of aortic valve replacement in symptomatic patients. Circulation. 1980;61:471–483.
Fioretti P, Roelandt J, Bos RJ, Meltzer RS, van Hoogenhuijze D, Serruys PW, Nauta J, Hugenholtz PG. Echocardiography in chronic aortic insufficiency: is valve replacement too late when left ventricular end-systolic dimension reaches 55 mm? Circulation. 1983;67:216–221.
Turina J, Hess O, Sepulcri F, Krayenbuehl HP. Spontaneous course of aortic valve disease. Eur Heart J. 1987;8:471–483.
Siemienczuk D, Greenberg B, Morris C, Massie B, Wilson RA, Topic N, Bristow JD, Cheitlin M. Chronic aortic insufficiency: factors associated with progression to aortic valve replacement. Ann Intern Med. 1989;110:587–592.
Reichek N, Wilson J, St. John Sutton M, Plappert TA, Goldberg S, Hirshfeld JW. Noninvasive determination of left ventricular end-systolic stress: validation of the method and initial application. Circulation. 1982;65:99–108.
Bland EF, Wheeler EO. Severe aortic regurgitation in young people: a long-term prospective with reference to prognosis and prosthesis. N Engl J Med. 1957;256:667–672.
Smith HJ, Neutze JM, Roche AH, Agnew TM, Barratt-Boyes BG. The natural history of rheumatic aortic regurgitation and the indications for surgery. Br Heart J. 1976;38:147–154.
Klodas E, Enriquez-Sarano M, Tajik AJ, Mullany CJ, Bailey KR, Seward JB. Surgery for aortic regurgitation in women: contrasting indications and outcomes compared with men. Circulation. 1996;94:2472–2478.