T 1980s, the survival benefits of both coronary artery

Transcription

1 Determinants of Survival and Recovery of Left Ventricular Function After Aortic Valve Replacement James J. Morris, MD, Hartzell V. Schaff, MD, Charles J. Mullany, MD, Amita Rastogi, MD, Christopher G. A. McGregor, MD, Richard C. Daly, MD, Robert L. Frye, MD, and Thomas A. Orszulak, MD Division of Thoracic and Cardiovascular Surgery and Division of Cardiovascular Diseases and Internal Medicine, Mayo Clinic and Mayo Foundation, Rochester, Minnesota To determine factors that influence survival and recovery of ventricular function in patients undergoing aortic valve replacement in the current surgical era, baseline risk factors related to outcome were analyzed in 1,012 consecutive patients undergoing aortic valve replacement between 1983 and Forty-two percent of patients underwent concomitant coronary bypass. Observed survival probabilities (expressed as 30-day/5-year) were 0.97/0.81 overall, 0.99/0.89 for patients aged less than 70 years, and 0.95/0.74 for patients aged 70 years or greater. Advanced age (p < O.OOOl), decreased ejection fraction (p < O.OOOl), extent of coronary disease (p < 0.006), smaller prosthetic valve (p < 0.03), and advanced New York Heart Association class (p < 0.04) were incremental risk factors for mortality. In patients with preoperative ventricular dysfunction (ejection fraction s0.45), ejection fraction measured 1.4 years after aortic valve replacement improved in 72% and the mean increment in ejection fraction was (95% confidence interval, to 0.195). The increment in ejection fraction was greater in female patients than in male patients (p < 0.02) and greater in patients without than with coronary disease (p < 0.02). Female sex (p < 0.02) and lesser extent of coronary disease (p < 0.05) were independent predictors of change in ejection fraction. In all patients, early improvement in ejection fraction conveyed an independent subsequent survival benefit (p < ). The results of aortic valve replacement in the current era are excellent, and the majority of patients with ventricular dysfunction demonstrate significant improvement. Early improvement in ejection fraction, influenced by coexistent coronary artery disease and sex-associated factors, importantly affects subsequent survival. ( 1993;56:22-30) hrough the decade of the 1970s and into the early T 1980s, the survival benefits of both coronary artery bypass grafting (CABG) and aortic valve replacement (AVR) significantly improved over time [ Important refinements in anesthetic and surgical techniques resulting in improved myocardial preservation [l, 4-71 and more complete myocardial revascularization [5, 7-91 introduced in the 1970s became uniformly practiced in the 1980s. At the same time, however, the success of surgical therapy has led to an expansion of the population considered for operation and altered baseline characteristics of the patient group undergoing AVR. As a result, earlier analyses of predictors of outcome may not truly reflect the spectrum of patients undergoing AVR in the current era nor was the overall survival benefit of AVR as great then as now. Also in recent years, increasing use of noninvasive methods, particularly echocardiography, has provided a means for the serial assessment of ventricular perfor- Presented at the Twenty-ninth Annual Meeting of The Society of Thoracic Surgeons, San Antonio, TX, Jan 25-27, Address reprint requests to Dr Morris, Mayo Clinic, 200 First St SW, Rochester, MN mance in large numbers of patients after AVR. We therefore undertook a retrospective study to determine factors that influence survival and recovery of left ventricular function in patients undergoing AVR in the current surgical era. Material and Methods Patient Population Baseline risk factors and subsequent outcome were retrospectively analyzed in 1,012 consecutive patients undergoing operation for AVR with and without concomitant CABG at the Mayo Clinic, Rochester, MN, between January 1983 and July Clinical, angiographic, and operative data were prospectively acquired by using a computerized data base. Patients less than age 30 years, those with prior cardiac operations, those with active endocarditis, and those undergoing concomitant cardiac procedures other than CABG were excluded from this analysis. Preoperative coronary angiography was performed in 920 of 1,012 patients (91%): in 638 of 678 patients (94%) aged 55 years or greater, in 87 of 123 patients (71%) aged less than 55 years, and in 30 of 48 patients (63%) aged less than 45 years. Preoperative left ventricular ejection frac by The Society of Thoracic Surgeons /93/$6.00

2 1993;56:2230 MORRIS ET AL 23 tion was determined by angiography, echocardiography, or both in 96% of patients an average of 46 days (95% confidence interval [CI], 23 to 70 days) before operation. Ejection fraction determinations by the two methods generally were in close agreement, and the preoperative ejection fraction measured by echocardiography was preferentially used for analysis except in the 21% of patients with only angiographically determined ejection fraction. The designation of valve disease type (stenosis, insufficiency, or combined stenosis and insufficiency) was in all patients determined by the operating surgeon incorporating preoperative clinical data, noninvasive and invasive hemodynamic data, and intraoperative findings. Specific operative techniques varied by surgeon, but the general approach was similar. Standard techniques included cardiopulmonary bypass with moderate systemic hypothermia, with cold potassium crystalloid or blood cardioplegia used for myocardial protection in all patients. Major coronary vessels or branches with 50% or greater stenosis supplying presumed viable myocardium were routinely bypassed unless, in the surgeon s opinion, the vessel was small and noncritical, as in the case of a nondominant right coronary artery. Internal mammary artery grafts were used with increasing frequency after Bioprosthetic valves were placed in 65% of patients, mechanical valves in 30%, and homografts in 5%. Postoperative left ventricular ejection fraction was determined by echocardiography performed at the Mayo Clinic, Rochester, MN, in 664 patients (66%) an average of 1.4 years (CI, 1.2 to 1.5 years) after operation. The indication for postoperative echocardiography varied but most frequently echocardiography was performed routinely on an annual basis. In the case of more than one postoperative echocardiogram for a single patient, the first echocardiogram obtained more than 2 months postoperatively was selected for analysis. Echocardiographic data obtained in the presence of major detectable prosthetic valve dysfunction or after a subsequent cardiac surgical procedure were excluded from analysis. Postoperative change in ejection fraction was calculated in 646 patients who had both preoperative and postoperative ejection fraction determinations. Patient follow-up information was obtained by mailed questionnaire, telephone interview by trained personnel, and review of the medical record during a 6-month interval between January 1992 and July Survival data include all deaths. Mean duration of follow-up was 4.2 years, and follow-up was complete in 988 of 1,012 patients (98%). Data Analysis Baseline patient variables used in this analysis and tested for association with risk were as follows: Preoperative variables Age Sex Body surface area New York Heart Association class Status of operation Elective Urgent Emergency Conduction defect None Right bundle-branch block Left bundle-branch block Second- or third-degree heart block Type of valvular heart disease Stenosis Insufficiency Combined stenosis and insufficiency Left ventricular ejection fraction Left ventricular end-diastolic pressure Aortic valve gradient Extent of coronary artery disease (250% stenosis) None 1-vessel disease 2-vessel disease 3-vessel disease Left main disease (250% stenosis) Operative variables Year of operation Prosthetic valve type Mechanical Bioprosthetic Homograft Prosthetic valve size Completeness of revascularization (calculated as number of major coronary vessels bypassed divided by number of major vessels diseased [250% stenosis]; = 0 if no vessels diseased) These variables are similar to risk factors previously reported [4, 7, 9-11]. Differences between patient groups in baseline clinical, angiographic, echocardiographic, operative, and follow-up variables were tested for statistical significance by t or 2 tests or by analysis of variance as appropriate. Multivariate analysis of all risk factors relating to outcome was performed for the overall population by using a stepwise Cox proportional hazards model [12] for overall survival and a regression analysis for postoperative change in ejection fraction. Variables were included in multivariate analyses if univariate significance was less than or equal to Unadjusted overall patient group survival probabilities relative to baseline characteristics were compared by a log-rank test. All values are mean 2 standard deviation except as otherwise noted. A p value less than 0.05 was considered significant. Results Of 1,012 consecutive patients undergoing AVR, 77% had aortic stenosis (AS), 11% had aortic insufficiency (AI), and 12% had mixed AS and AI. The group with AS had the greatest mean age, the greatest proportion of female patients, and the highest frequency of coronary artery disease (all p < 0.01) (Table 1). Mean preoperative left ventricular ejection fraction was the lowest in patients with A1 (p < 0.01), but the proportion of patients with a

3 24 MORRISETAL 1993;56:22-30 Table 1. Comparison of Baseline Patient Characteristics by Valve Disease Type Variable AS A1 ASIA1 p Value No. of patients Preoperative Mean age (y) 70 f * f 12 <0.001 Female sex 35% 25 % 23% <0.008 Coronary artery disease 54% 40% 36% <o Mean number of vessels diseased NS Mean ejection fraction 0.55 f f f 0.14 c0.004 Ejection fraction % 36% 27% NS Operative Concomitant CABG 55% 35 % 30% <0.002 Prosthetic valve type 27% 44% 33 % (0.007 (% mechanical) Median prosthetic valve size (mm) A1 = aortic insufficiency; AS = aortic stenosis; CABG = coronary artery bypass grafting; "3 = not significant. preoperative ejection fraction of 0.45 or less did not differ (p = not significant [NS]) among valve disease type subgroups. Overall, 42% of patients underwent concomitant CABG. The frequency of CABG was significantly higher in AS patients than in A1 or AS/AI patients, but the frequency of CABG closely approximated the frequency of coronary artery disease in each valve disease type subgroup (see Table 1). Overall, CABG was performed in 83% of patients with 50% or greater stenosis of one or more coronary arteries. During the study interval, the overall frequency of concomitant CABG each year did not change (p = NS), and the mean age of patients undergoing AVR each year also did not change (p = NS) (overall mean age, 68.2? 11.5 years), reflecting the relative stability of patient characteristics over time. Survival Analysis Overall 30-day operative mortality was 3% for the entire group of patients and 1% in patients aged less than 70 years. Thirty-day survival probabilities were similar ac- Table 2. Early and Late Unadjusted Survival Probabilities Group All patients Age <70 years 270 years Operation AVR only AVR + CABG Disease AS A1 ASIA1 30-Day 5-Year Survival Survival p Value < < cording to valve disease type, but the 5-year survival probability of 0.91 in patients with ASIA1 was significantly better (p < ) than the survival probability of 0.80 in patients with either AS or A1 (Table 2). Univariate analysis of risk factors for overall mortality identified 13 preoperative and operative variables as determinants of outcome (Table 3). Multivariate analysis identified five independent predictors of mortality (Table 4). Advanced age was the most important predictor of mortality, followed by impairment of preoperative left ventricular ejection fraction. Other significant risk factors were greater extent of coronary artery disease, smaller prosthetic valve size, and advanced preoperative New York Heart Association functional class. Unadjusted survival curves relative to each of the identified multivariate risk factors for mortality are presented in Figures 1 through 5. Table 3. Univariate Analysis of Risk Factors for Mortality in All Patients Variable 2 p Value Age Extent coronary artery disease NYHA class Ejection fraction Prosthetic valve size Left main disease Incomplete revascularization Bioprosthetic valve type Smaller body surface area Extent of conduction defect Valve disease type (AS or AI) Female sex Aortic valve gradient A1 = aortic insufficiency; AS = aortic stenosis; AVR = aortic valve replacement; CABG = coronary artery bypass grafting. A1 = aortic insufficiency; Heart Association. AS = aortic stenosis; NYHA = New York

4 ~ ; MORRISETAL 25 Table 4. Multivariate Analysis of Risk Factors for Mortality in All Patients Variable 2 p Value Age 18 Ejection fraction 16 Extent of coronary artery disease Prosthetic valve size NYHA class NYHA = New York Heart Association. Change in Ejection Fraction In the entire group of 646 patients in whom postoperative change in ejection fraction was determined, the overall mean change in ejection fraction measured an average 1.4 years after AVR was (CI, to 0.017; median, 0; and range, to 0.42). In the 167 patients with preoperative ejection fraction of 0.45 or less (mean, C 0.073), postoperative ejection fraction increased in 72% of patients, and the mean increment in ejection fraction was (CI, to 0.195). In the 28% of patients without improvement in postoperative ejection fraction, the mean decrement in ejection fraction was (CI, to ). Overall, for the entire group of patients with preoperative ejection fraction of 0.45 or less, the mean change in ejection fraction was (CI, to 0.126). Improvement in ejection fraction occurred in 87 of 120 (73%) of patients with AS, 18 of 28 (64%) patients with AI, and 15 of 19 (79%) patients with AS/AI, and the magnitude of the increment in ejection fraction did not differ (p = NS) according to valve disease type: (CI, to 0.193) in AS patients, (CI, to 0.200) in A1 patients, and (CI, to 0.288) in AS/AI patients. The observed improvement in ejection fraction among patients with preoperative left ventricular dysfunction was significantly greater in female patients as a group than in male patients (p < 0.02) (Table 5). Ejection fraction increased in 28 of 37 female patients (76%) and 92 of 130 male patients (71%) (p = NS), but the mean increment in ejection fraction was (CI, to 0.261) in female EF -SO nr n.299-2x-45 n <30 n-122 "." Fig 2. Unadjusted survival probability relative to preoperative ejection fraction (EF). patients and only (CI, to 0.183) in male patients (p < 0.02). There was also a trend toward a greater improvement in ejection fraction in patients with lesser extent of coronary artery disease as a group (p < 0.07) (see Table 5). Ejection fraction increased in 58 of 72 patients (81%) without coronary disease compared with 62 of 94 patients (66%) with coronary disease (p = NS), but the mean increment in ejection fraction was (CI, to 0.173) and (0.048 to 0.108) (p < 0.02), respectively. By multivariate regression analysis, both female sex (p < 0.02) and lesser extent of coronary artery disease (p < 0.05) were demonstrated to be independent predictors of improvement in postoperative ejection fraction among patients with preoperative ejection fraction of 0.45 or less. Change in Ejection Fraction and Survival For the entire group, subsequent survival was significantly better in patients with an early improvement in ejection fraction than in patients without an improvement in ejection fraction (Fig 6). In patients with a preoperative ejection fraction of 0.45 or less, 5-year survival probability was 0.82 in patients with an increment in ejection fraction greater than or equal to 0.10, 0.61 in patients with an increment in ejection fraction less than 0.10 or a decre- 1.o -.-, -<40 n= n= n=128 n= n=373 ->80 n= I I I I I I I I Fig I. Unadjusted survival probability relative to age. I CAD -0VD n=492-1 VD n=167-2vd n=152-3vd n= Fig 3. Unadjusted survival probability relative to extent of coronary artery disease (CAD). (VD = vessels diseased.)

5 26 MORRIS ET AL 1993; mm n= mm n= mm n mm n _. >25mm n= ' I I I I I I I Fig 4. Unadjusted survival probability relative to prosthetic valve size. ment less than -0.10, and 0.54 in patients with a decrement in ejection fraction more than (p < 0.01). For the entire patient population, reentering postoperative change in ejection fraction into the Cox multivariate analysis of risk factors for mortality demonstrated that early improvement in ejection fraction conveyed a significant independent survival benefit (p < ) (Table 6). Comment Recently reported multivariate analyses [6, 7, 10, 13-15] of patients undergoing AVR with and without concomitant CABG have noted operative mortality rates of 3% to 7% with 5-year survival probabilities of 0.70 to However, most patients in these series underwent operation in the 1970s and early 1980s. Since that time, there have been significant alterations in baseline characteristics of the patient population undergoing AVR. In more recent years, a greater proportion of patients have coronary artery disease and are older [15]. At our institution, the percentage of patients aged 70 years or greater undergoing AVR and AVR plus CABG progressively increased from 13% and 23% in 1967 through 1976 to 30% and 42% in 1982 through 1983 [lo] to 43% and 63% in 1983 through 1991, respectively. Despite this generally perceived worsening of baseline characteristics of the population undergoing AVR compared with populations in earlier eras, we demonstrated the results of AVR remain excellent. Overall 30-day survival probability was 0.97 and 5-year survival probability was As others [15] have maintained, in current surgical practice operative mortality has a significant impact on outcome only in patients 70 years of age or greater. The risk of operation is extremely low (1%) in patients aged less than 70 years undergoing either AVR or AVR plus CABG. The continuous variables of age and preoperative left ventricular ejection fraction were the most important predictors of mortality in this series. Of note, although there was a progressive decline in overall survival by decade of age, in patients greater than age 80 years, 5-year survival was still 62%. The extent of coronary artery disease and preoperative functional class were also important multivariate determinants of long-term survival, affirming prior observations [7, 10, 11, 15, 161. Earlier year of operation, previously identified as a multivariate risk factor for survival in the decades of the 1960s and 1970s [lo, 11, 16, 171, was not shown to be an important risk factor in the current series. Refinements in anesthetic and surgical techniques, particularly improved myocardial preservation [l, 4, 51 and more complete revascularization [5, 7, 151, introduced during the 1970s and early 1980s were uniformly practiced during the study period. Aortic insufficiency, another variable earlier identified [8, 161 as a risk factor also was no longer a predictor of outcome, likely for similar reasons and perhaps because of earlier referral of patients with A1 for AVR. Concomitant CABG also did not independently contribute to mortality. Interestingly, larger prosthetic valve size appeared to convey a significant incremental survival benefit independent of prosthetic valve type, aortic valve disease type, patient age, sex, and body surface area. As shown in Figure 4, the survival differences according to valve size appeared to be attributable to differences in both early and late mortality. Several factors may account for this observation. Unrecognized variables or interaction between confounding variables not incorporated into the multivariate model may have affected the analysis of certain risk factors. The use of a smaller prosthetic valve may be an indicator of factors that contribute to risk in an additive manner such as a combination of critical senescent aortic stenosis, a small calcified aortic root, marked ventricular hypertrophy with narrowing of the left ventricular outflow tract, a small body surface area, and a preexistent sedentary or debilitated activity level. Smaller valve size may also be an indicator of patient-prosthetic valve mismatch and a significant postoperative left ventricularaortic pressure gradient. However, for several reasons, this is not likely the case. In general, use of too small a prosthetic valve was avoided in this series by frequent use of annular enlarging techniques as previously described [18]. Overall, 104 patients (10%) underwent a concomitant annular enlargement procedure to accommodate a larger prosthesis. Furthermore, valve size as a continuous vari- - Class 3 n.502 I - Class 4 n=189 I o.2 t I t I I I I 0.0 I Fig 5. Unadjusted survival probability relative to preoperative New York Heart Association (NYHA) class.

6 1993; MORRIS ET AL 27 Table 5. Change in Ejection Fraction After Aortic Valve Replacement No. of Group Patients Mean Change 95% CI p Value All patients to Preop ejection fraction to > to Preop ejection fraction Disease AS to A to ASIA to Sex Male to Female to No. of vessels diseased to to to to A1 = aortic insufficiency; AS = aortic stenosis; CI = confidence interval NS <0.02 <0.07 able was observed to convey a survival benefit. Survival was better in patients with valve size 22 to 23 mm compared with patients with valve size less than or equal to 21 mm, but also survival was better in patients with valve size greater than 25 mm compared with valve size 24 to 25 mm. This also suggests that differences in outcome were not attributable to postoperative pressure gradients. All aortic prostheses larger than 21 mm generally provide satisfactory performance in the majority of adult patients [ll]. There also was no observed correlation between prosthetic valve size and postoperative change in ejection fraction, again indicating survival differences were likely not due to postoperative valve gradients. The exact reasons for this apparent association are unclear, and further investigation is necessary to confirm this observation. - C t P< A EF >I0 n=146 AEF-lOtolO ~416 - A EF +lo n= Fig 6. Unadjusted survival probability relative to postoperative change in ejection fraction (AEF). An additional purpose of the present study was to analyze factors that influence recovery of ventricular performance across the entire spectrum of patients undergoing AVR in the recent surgical era to draw conclusions applicable to clinical decision-making in current practice. A number of prior studies have characterized the effects of AVR on ventricular function [17, 19-26], but most were relatively small series of patients predominantly with AI. The immediate effects of valve replacement on left ventricular performance are principally decreases in preload and afterload [25]. Later effects include ventricular adaptation and remodeling with regression of hypertrophy and left ventricular mass [19, 201. Both end-diastolic and end-systolic volumes decrease [19, 21, 22, 251, apd there is an increase in the end-systolic pressure-volume ratio implying an improvement in contractile performance [25]. Postoperative ejection fraction has been reported to be maintained or increased after AVR in patients with normal preoperative ejection fraction, and ejection fraction tends to increase in patients with preoperative impairment of ejection fraction [19-23, 25, 261. However, not all patients with left ventricular dysfunction improve after AVR, and persistent left ventricular dysfunction adversely influences subsequent survival [17, 271. Preoperative impairment of ejection fraction, prior myocardial infarction, and preoperative aortic valve gradient have been shown to be associated with postoperative impairment of ejection fraction [24]. Our study confirms several of these observations. Ejection fraction was maintained after AVR in patients without preoperative ventricular dysfunction and increased after AVR in patients with preoperative ventricular dysfunction. Overall, 72% of patients with preoperative ven-

7 28 MORRIS ET AL 1993;56:2230 Table 6. Multivariate Analysis of Risk Factors for Mortality lncludinn Postoperative Channe in Ejection Fraction Variable J p Value Preoperative ejection fraction 31 Postoperative change in ejection haction 21 Age Extent of coronary artery disease tricular dysfunction demonstrated a substantial improvement in postoperative ejection fraction at a mean 1.4 years after AVR, and the mean increment in ejection fraction was Of the 28% of patients who did not demonstrate an improvement, the mean decrement in ejection fraction was By multivariate analysis, preoperative ejection fraction was the only variable that exceeded the importance of postoperative change in ejection fraction as a predictor of survival. Preoperative duration of left ventricular dysfunction has previously been identified as a determinant of reversibility of dysfunction after AVR for A1 [22]. That variable was not included in our analysis. Duration of preexistent ventricular dysfunction is difficult to determine retrospectively, and in current practice patients are not commonly followed prospectively with progressive deterioration of ventricular performance before referral for AVR. Lesser extent of coronary disease and female sex were each identified as independent predictors of ventricular functional recovery after AVR. Patients with preoperative left ventricular dysfunction and more extensive coronary disease likely have a greater degree of irreversibly damaged myocardium or scar before AVR. As expected, these patients demonstrate less improvement in postoperative ejection fraction. The extent of coronary disease also influences the risk of perioperative myocardial infarction with AVR [6], which may further influence postoperative recovery of function. Recent studies [ have demonstrated sexassociated differences in left ventricular adaptation to aortic stenosis. With similar degrees of symptoms and left ventricular outflow tract obstruction, female patients more frequently demonstrate preservation of left ventricular fractional shortening, pressure generation, ejection time, mean pulmonary artery pressure, and cardiac index compared with male patients [28]. For a given left ventricular systolic load, female patients with AS also have an increased relative left ventricular wall thickness and an exaggerated hypertrophy response accounting for increased fractional shortening [29]. Abnormal myocardial collagen fiber architecture contributing to ventricular dysfunction is also less commonly observed in female patients than in male patients with AS [30]. The present study, demonstrating an independent effect of female sex on recovery of depressed left ventricular ejection fraction in a large population after AVR, suggests there are also significant sex-associated differences in regression of left ventricular adaptation to chronic pressure and volume overload. Possible hormonal and cellular mechanisms accounting for these observed differences are as yet uncharacterized. Female sex was observed to be a univariate risk factor for mortality. Yet, female sex was also independently associated with improvement in postoperative ejection fraction, which was shown to convey a subsequent survival benefit. Testing for differences in the magnitude of the survival benefit attributable to improvement in ejection fraction according to sex, Cox model analysis demonstrated no significant interaction (p > 0.7) between the effects of sex and improvement in ejection fraction on survival. Therefore, the survival benefit attributable to improvement in ventricular function was similar in male and female patients. One recognized limitation of the present study is that postoperative assessment of left ventricular function was obtained in only 664 of 1,012 patients (66%) after AVR. An important concern with any retrospective observational analysis is the influence of patient selection bias, and in this case, selection bias may have influenced the completeness of follow-up ventricular functional data. Obviously, mortality precluded determination of postoperative ejection fraction, but other variables may have also influenced whether or not postoperative ventricular function was assessed. Comparison of baseline characteristics between groups of patients with and without determination of postoperative ejection fraction indicated no differences (all p = NS) in the proportion of female patients, preoperative New York Heart Association functional class, or valve disease type. However, among patients without postoperative ejection fraction determinations, a greater proportion of patients were greater than age 70 years (62% versus 50%; p < 0.001), had coronary artery disease (57% versus 47%; p < 0.01), had a preoperative ejection fraction of 0.45 or less (37% versus 26%; p < 0.001), and underwent AVR with CABG (48% versus 39%; p < 0.01). In patients with a postoperative determination of ejection fraction, mortality was 3.4% at 1.4 years after AVR, the average time of postoperative echocardiography, versus 10% (p < 0.01) in patients without a postoperative determination of ejection fraction. Although significant, differences in baseline characteristics were not large in magnitude and early mortality would account for, at the most, only a minimal portion of unobtained postoperative ventricular functional data. For these reasons, and because the absolute number of patients with postoperative functional data is large, conclusions drawn regarding change in ejection fraction after AVR seem to be valid. In summary, despite the advanced age and more frequent coronary artery disease in the patient population, the early and intermediate-term results of AVR in the current surgical era remain excellent. The majority of patients with preoperative left ventricular dysfunction demonstrate a significant early improvement in ejection fraction after AVR. Early recovery of ventricular function, influenced by coexistent coronary artery disease and sex-associated factors, importantly affects subsequent survival.

8 1993:56:22-30 MORRISETAL 29 The significant contributions of G. K. Danielson, MD, J. R. Pluth, MD, F. J. Puga, MD, and the members of the Division of Cardiology at the Mayo Clinic in the care of these patients are gratefully acknowledged. References 1. Cohn LH, Allred EN, DiSesa VJ, Sawtelle K, Shemin RJ, Collins JJ Jr. Early and late risk of aortic valve replacement. J Thorac Cardiovasc Surg 1984;88: Cosgrove DM, Loop FD, Lytle BW, et al. Determinants of 10 year survival after primary myocardial revascularization. Ann Surg 1986;202: Pryor DB, Harrell FE, Rankin JS, et al. The changing survival benefits of coronary revascularization over time. Circulation 1987;76(Suppl 5): Lytle BW, Cosgrove DM, Taylor PC, et al. Primary isolated aortic valve replacement. J Thorac Cardiovasc Surg 1989;97: Richardson JV, Kouchoukos NT, Wright JO 111, Karp RB. Combined aortic valve replacement and myocardial revascularization: results in 220 patients. Circulation 1979;59: Christakis GT, Weisel RD, Fremes SE, et al. Can the results of contemporary aortic valve replacement be improved? J Thorac Cardiovasc Surg 1986;92: Czer LSC, Gray RJ, Steward ME, De Robertis M, Chaux A, Matloff JM. Reduction in sudden late death by concomitant revascularization with aortic valve replacement. J Thorac Cardiovasc Surg 1988;95: Lytle BW, Cosgrove DM, Loop FD, et al. Replacement of aortic valve combined with myocardial revascularization: determinants of early and late risk for 500 patients, Circulation 1983;68: Lytle BW, Cosgrove DM, Gill CC. Aortic valve replacement combined with myocardial revascularization. J Thorac Cardiovasc Surg 1988;95: Mullany CJ, Elveback LR, Frye RL, et al. Coronary artery disease and its management: influence on survival in patients undergoing aortic valve replacement. J Am Coll Cardiol 1987;10: Kirklin JW, Barrett-Boyes BG. Aortic valve disease. In: Kirklin JW, Barrett-Boyes BG, eds. Cardiac surgery. New York: Churchill-Livingstone, 1993: Cox DR. Regression models and life tables. J Res Stat SOC [B] 1972;34: Magovern JA, Pennock JL, Campbell DB, et al. Aortic valve replacement and combined aortic valve replacement and coronary artery bypass grafting: predicting high risk groups. J Am Coll Cardiol 1987;9:3% Craver JM, Weintraub WS, Jones EL, Guyton RA, Hatcher CR. Predictors of mortality, complications, and length of stay in aortic valve replacement for aortic stenosis. Circulation 1988;78(Suppl 1): Lytle BW. Impact of coronary artery disease on valvular heart surgery. Cardiol Clin 1991;9: Scott WC, Miller DC, Haverick A, et al. Determinants of operative mortality for patients undergoing aortic valve replacement. J Thorac Cardiovasc Surg 1985; Bonow RO, Picone AL, McIntosh CL, et al. Survival and functional results after valve replacement for aortic regurgitation from 1976 to 1983: impact of preoperative left ventricular function. Circulation 1985;72:124& Piehler JM, Danielson GK, Pluth JR, et al. Enlargement of the aortic root or anulus with autogenous pericardial patch during aortic valve replacement. Long-term follow-up. J Thorac Cardiovasc Surg 1983;86: Kennedy JW, Doces J, Stewart DK. Left ventricular function before and following aortic valve replacement. Circulation 1977;56: Pantely G, Morton M, Rahimtoola SH. Effects of successful, uncomplicated valve replacement on ventricular hypertrophy, volume, and performance in aortic stenosis and in aortic incompetence. J Thorac Cardiovasc Surg 1978;75: Smith N, McAnulty JH, Rdlhimtoola SH. Severe aortic stenosis and impaired left ventricular function and clinical heart failure: results of valve replacement. Circulation 1978;58: Bonow RO, Rosing DR, Maron MJ, et al. Reversal of left ventricular dysfunction after aortic valve replacement for chronic aortic regurgitation: influence of duration of preoperative left ventricular dysfunction. Circulation 1984; Carabello BA, Usher BW, Hendrix GH, Assey ME, Crawford FA, Leman RB. Predictors of outcome for aortic valve replacement in patients with aortic regurgitation and left ventricular dysfunction: a change in the measuring stick. J Am Coll Cardiol 1987;10: Hwang MH, Hammermeister KE, Oprian C, et al. Preoperative identification of patients likely to have left ventricular function after aortic valve replacement. Circulation 1989; SO(Supp1 1): Harpole DH, Jones RH. Serial assessment of ventricular performance after valve replacement for aortic stenosis. J Thorac Cardiovasc Surg 1990;99: Borer JS, Herrold EM, Hochreiter C, et al. Natural history of left ventricular performance at rest and during exercise after aortic valve replacement for aortic regurgitation. Circulation 1991;84(Suppl 3): Cunha CLP, Giuliani ER, Fuster V, Seward JB, Brandenburg RO, McGoon DC. Preoperative M-mode echocardiography as a predictor of surgical results in chronic aortic insufficiency. J Thorac Cardiovasc Surg 1980;79: Carroll JD, Carroll El', Feldman T, et al. Sex-associated differences in left ventricular function in aortic stenosis of the elderly. Circulation 1992;86: Aurigemma GP, Silver KH, McLaughlin M, Orsinelli D, Sweeney AM, Gaasch WH. Gender influences the pattern of left ventricular hypertrophy in elderly patients with aortic stenosis. Circulation 1992;86(Suppl 1): Villari B, Hess OM, Campbell SE, Krayenbuehl HP. Sexdependent response of the left ventricle to aortic valve stenosis: influence of the collagen network. Circulation 1992; 86(SuppI 1):538. DISCUSSION DR HENRY M. SPOTNITZ (New York, NY): Doctors Morris, Schaff, and colleagues have presented a study characterized by very low surgical mortality and excellent long-term patient survival. Their results also demonstrate interesting positive effects of larger prosthesis size and female sex on survival. I wish to thank the authors for faxing a copy of their recently revised study last week. I have some comments and questions. The strength of this analysis is in demonstrating a favorable effect on patient survival of both improvement in postoperative ejection fraction and higher levels of preoperative ejection fraction. Weaknesses include omission of indications for operation and preoperative symptoms. Overall, 42% of patients in the study underwent coronary artery bypass. Also, 42 of 167 with low ejection fractions had triple-vessel disease. This leads to the question of whether angina, syncope, heart failure, or no symptoms were present in these patients and whether symptoms were

9 30 MORRIS ET AL 1993;56:2230 related to outcome. Similarly, reduced afterload after aortic valve replacement may not improve ejection fraction if the cause of low ejection fraction is a thick scar related to infarction. Was an attempt made to distinguish patients with asymmetric preoperative wall motion or fixed defects on thallium scan? In relation to valve size, was root enlargement itself tested as an independent variable affecting survival? Many previous studies have attempted to predict effects of valve replacement on ejection fraction and to relate changes in intraoperative ejection fraction to long-term results. Aortic stenosis and insufficiency are considered different pathophysiologically in such studies, although no outcome differences were apparent in the present study. Our data show a statistically significant decrease in intraoperative ejection fraction in aortic insufficiency but an increase in aortic stenosis. Wall stress decreases somewhat in patients with aortic insufficiency after valve replacement, but similar data for aortic stenosis, show a marked reduction in afterload. If the expected effect of operation on ejection fraction could be reliably predicted, the present study indicates that survivorship could also be predicted. Analysis of reasons for outcome better or worse than predicted might improve surgical methods and results. The importance of the present study is not so much in identifying mechanisms as in documenting both the current state of the art in aortic valve operations and the relation of changes in ejection fraction to survival. For this Dr Morris and associates should be heartily congratulated. DR RICHARD P. ANDERSON (Seattle, WA): I am intrigued with this report, and one of the findings that particularly caught my attention was the relationship between valve size and longterm estimates of survival. However, there was one finding that was a little paradoxical. Perhaps you could explain it. I understood you to say that the effect of valve size on survival was independent of body size. If that is the case, I wonder whether that observation is not more a problem with the type of regression analysis than it is with reality, because there should probably be a correlation, at least in the small valve sizes, between body size and valve size. If you are doing a multiple linear regression you probably would have some kind of a problem with multiple colinearity if you do not combine those two. Perhaps you could address that. DR MORRIS: I would like to thank Dr Spotnitz and Dr Anderson for their comments and insightful critique. Symptoms of angina, heart failure, and syncope were not included in the multivariate analysis; including them might add additional information. Likewise, information regarding the presence or absence of transmural scar or other evidence of nonviability of myocardium, which also might be important additional information, was not included in the analysis. We did not specifically test root enlargement or annular enlarging procedures as a variable in the multivariate model. We have examined smaller subsets of patients to specifically assess the effect of annular enlarging procedures, and while we do not have sufficient data to address the issue of survival, we have seen that there is an apparent significant improvement in regression of left ventricular mass that can be attributable to root enlargement procedures. Perhaps with longer duration of follow-up there might be a beneficial effect on survival also. With regard to the issue of valve size being a predictor of outcome, the data did demonstrate that larger valve size conveyed a significant incremental survival benefit independent of prosthetic valve type and, as I mentioned, valve disease type, sex, and body surface area. Body surface area was a univariate predictor of mortality, but entered into the Cox multivariate analysis it did not emerge as a multivariate predictor of mortality. Likewise, indexing valve size to body surface area was not a significant predictor of mortality. Several factors might account for our observation of the association between valve size and subsequent mortality. First, this is a large series of patients who underwent AVR in the current surgcal era with uniformly practiced methods of optimal myocardial revascularization and myocardial protection. This was not necessarily true in earlier series that did not show this association and, as a result, those earlier analyses of predictors of outcome may not apply to patients undergoing AVR in current practice. Second, there may be unrecognized variables or interaction between confounding variables that might affect the analysis of certain risk factors. Small valve size may actually be an indicator of a number of factors that contribute to risk in an additive fashion, such as a combination of critical senescent aortic stenosis, a small calcified aortic root, marked left ventricular hypertrophy with narrowing of the left ventricular outflow tract, small body surface area along with a perceived preexistent sedentary or debilitated lifestyle. Third, smaller valve size may actually be a marker for a significant postoperative left ventricular aortic pressure gradient. For a variety of reasons we think that that is not likely the case. In general the use of too small a valve was avoided in this series by the frequent use of annular enlarging techniques. Ten percent of patients had a concomitant annular enlarging procedure to accommodate a larger valve. Also, valve size as a continuous variable was shown to convey a survival benefit. Not only was survival better with a 21-mm or 23-mm valve compared with a 19-mm valve, but survival was also better with a 25-mm valve compared with a 23-mm valve, again suggesting that survival differences were not attributable to differences in gradients. And, finally we did not observe a correlation between a smaller prosthetic valve size and change in postoperative ejection fraction, again suggesting that the survival differences could not be attributable to differences in gradients. The exact reasons for this apparent association are as yet not entirely clear. We are not advocating a routine practice of upsizing of prosthetic valves, which might likely be accompanied by an additional set of problems. Further investigation is necessary to confirm and clarify this observation.