VALVULAR HEAR? DISEASE
Determinants of One-year Outcome from Balloon Aortic Valvuloplasty Charles J. Davidson, MD, J. Kevin Harrison, MD, Karen S. Pieper, MS, Michael Harding, MD, James B. Hermiller, MD, Katherine Kisslo, RDMS, Cynthia Pierce, RN, and Thomas M. Bashore, MD
Balloon aortic valvuloplasty (BAV) has been a therapeutic alternative treatment for severe symptomatic aortic stenosis. Previous studies have been unable to predict l-year outcome because of limited acute and follow-up clinical, invasive and echocardiographic data. The purpose of this study was to predict long-term outcome based on comprehensive data obtained at the time of valvuloplasty and at 3 and 6 months aiter the procedure. Of 170 consecutive patients undergoing BAV, 108 (mean age 78 years) were at least 1 year from their procedure. Prospective clinical, micromanometer hemodynamic, digital ventriculographic and echocardiographic/ Doppler data were collected at baseline and immediately after the procedure. Echocardiographi6 data were also obtained at 3 and 6 months. With use of Cox model analysis, major events (defined as cardiac death [n = 301, aortic valve replacement [n = 211 qr repeat BAV [n = 131) were predicted by advanced age, baseline heart failure class, and baseline echocardiographic-determined diastolic left ventricular diameter. Only baseline left ventricular ejection fraction proved to be a significant predictor of cardiac death (p = 0.002) in a multivariate model. Absolute values after BAV (stroke work, first derivative of left ventricular pressure, valve area, end-systolic volume, Fick cardiac output, transvalvular gradient) and acute changes measured by catheterization or echocardiography did not provide additional predictive information over that of postprocedure ejection fraction. Similarly, echocardiographic valve area and transvalvular gra-
dient at 3 months added no further prognostic data. With an ejection fraction 145% (n = 63), cardiac survival at 1 year was SO%, irrespective of age, sex, congestive heati failure class or severity of coronary artery disease. Thus, prognosis after BAV can be determined by noninvasive clinical and echocardiographic data obtained before cardiac catheterization. Despite advanced age and concomitant medial problems, patients undergoing BAV with baseline ejection fractions 245% have an excellent cardiac survival to 1 year. (Am J Cardiol 1991;68:75-60)
alloon aortic valvuloplasty (BAV) has been an alternative treatment for severe aortic steno&L-~ Although restenosisand recurrent symptoms occur often at a variable time after the procedure, some patients experience sustained improvement.l-l 2 Previousstudieshave describedshort-term predictors of outcome, based on either invasive or noninvasive parameters.3-7However, analysis of factors predictive of long-term outcome have yet to be delineated. In an elderly and debilitated population of patients, BAV often representsthe only treatment option other than medical therapy. Therefore, it becomesimportant lo clarify its role in the treatment of symptomatic aortic stenosis. Controversy exists as to what factors, if any, constitute a successfulprocedure (e.g., final aortic valve area, and changesin aortic or echocardiographicvalve areas 1 to 3 days after BAV).3-7 Previous studies have limited long-term data and have typically failed to analyze both invasive, hemodynamic and serial echocardiographic data and their relation to patient outcome. Prospectivedetermination of clinical, hemodynamic or From Duke University Medical Center, Department of Medicine, Divi- echocardiographic variables predictive of 1-year outsionof Cardiology, Durham, North Carolina. This study wassupported come would help clarify the patient groups that are in part by Grant 5T32 HL07 101from the National Institutes of Health, National Research Service Award, Bethesda,Maryland. Manuscript likely to benefit from BAV. This study identifies the received January 15, 1991; revised manuscript received March 13, clinical, acute hemodynamic and echocardiographic 1991,and acceptedMarch 17. variables that predict l-year outcome after BAV, and Address for reprints: Charles J. Davidson, MD, Duke University determines whether 3- or &month echocardiographic Medical Center, Department of Medicine, Division of Cardiology, Box 31195,Durham, North Carolina 27710. data provide additional prognostic information.
B
BALLOON AORTIC VALVULOPLASTY
75
METHODS Of 170 consecutivepatients undergoing BAV at Duke University Medical Center, 108 had undergone valvuloplasty at least 1 year previously and thus were eligible for 1 year follow-up. BAV was performed in patients that were deemed high risk for aortic valve replacement after all patients were evaluated by cardiology and cardiothoracic surgery. The median age was 78 years (range 53 to 91). Forty-four percent were men. Before BAV, all patients had severe symptomatic aortic stenosis and were in either New York Heart Association Functional classIII or IV congestive heart failure (n = 86) or Canadian Functional classIII or IV angina (n = 49), or both. Basedon digital subtraction supravalvular aortography and 2-dimensional echocardiography, the cause of aortic stenosis was assumed to be senile degenerative calcilic aortic stenosisin 96 and bicuspid aortic stenosisin 12. Significant coronary artery diseasewas present in 59 (55%). Fourteen had significant 3-vessel disease. Significant diseasewas usually defined as 175% luminal diameter narrowing. Valvuloplasty procedure: BAV was performed after informed consent from a previously approved institutional review board protocol. A 20-mm balloon was used as the initial balloon in all patients. The retrograde technique using a single balloon (Mansfield, Inc., Mansfield, Massachusetts)of 20 or 23 mm was used in 92 and 6 patients, respectively. Double balloons of 20 and 10 mm were used in 10 patients. Left ventricular and aortic pressureswere simultaneously measured immediately before and after BAV with dual-sensor, high-fidelity micromanometer catheters (Millar, Inc., Houston, Texas). R-wave gated masked-mode digital subtraction left ventriculography and supravalvular aortography were obtained before and after the intervention in 106 of 108 patients. Left ventricular volume was calculated with the area-length formula and a regression equation based on 20 previously calibrated human heart casts of known volume. Aortic regurgitation was graded 0 to 4+ with a standard angiographic grading scale. Angiographic imageswere acquired with an ADAC 4100-C digital radiography system interfaced with a GE- MPX L/U x-ray unit. Images were recorded in a 9-inch image intensifier mode using a 256 X 256 X 8 pixel matrix at 30 frames per second. Oxygen consumption was measuredwith a metabolic cart and Fick cardiac output was obtained. The effective aortic valve area was calculated using the Gorlin formula. khocardiographic data: Echocardiographic data were obtained with a Hewlett-Packard 7702OAphasedPatient population:
76
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
array sector scanner immediately before and 1 to 3 days after valvuloplasty in all patients. Fifty-eight patients underwent a 3-month echocardiographic study, and 53 patients a 6-month study. Left ventricular enddiastolic dimension was measuredfrom the parasternal view using M-mode or 2-dimensional imaging according to the recommendationsof the American Society of Echocardiography.13,14Subaortic diameter was measured immediately below the aortic cusps in the longaxis parasternal view. The subaortic area was calculated from diameter measurementsduring 3 to 5 beats in systole. Subaortic velocity was measured by pulsed Doppler and displayed by fast-Fourier transform spectral analysis. Signals were obtained from the cardiac apex with a cursor positionedin the outflow tract at the same level as the diameter measurement. Transducer angulation was optimized to obtain maximal velocity. Peak systolic velocity was calculated for 5 cardiac cycles in patients with sinus rhythm and 10 cycles in patients with atria1 fibrillation. Continuous-wave Doppler was performed with a 1.9-MHz transducer, and velocities were optimized. Peak instantaneousand mean gradients were calculated from resident software. Aortic valve area was calculated with the continuity equation.r5 Paiient evaluation: Patient status was evaluated by physician interview and examination at baselineand at 3 months, 6 months and 1 year after BAV. End points were achieved when the patient died, underwent aortic valve replacement, underwent repeat BAV, or survived to 1 year after the procedure. If death occurred, telephone interview with both the primary care physician and family were obtained in order to classify the cause as either cardiac or noncardiac. Statistics: All continuous results were expressedas median (twenty-ftith, seventy-fnth percentile). Categorical variables were expressed as percentages.The Kaplan-Meier methodi was used to calculate median survival rates. All patients were followed for 1 year unlessthey died or had a major event, at which date they were withdrawn and analyzed as alive (i.e., event-free). To evaluate the changes in values from before to after valvuloplasty, paired t tests were used after the assumption of normality was checked and accepted.A Wilcoxon rank-sum test was used to compare baseline echocardiographic measurementsin patients with and without a 3-month echocardiogram. Becauseof the large number of variables collected in relation to the relatively small number of events, a subsetof potential prognostic variables from each of the categories (clinical, hemodynamic 1, hemodynamic 2 and echocardiographic) was identified. The Cox proportional-hazards regressionmodel*7*18 was used to ex-
JULY 1, 1991
amine the univariable relation of each with the outcomesof cardiovascular death and total cardiac events (cardiac death, aortic valve replacement or repeat valvuloplasty). When the relation between the variable of interest and the logarithm of the hazard ratio was assumedto be linear, spline functionst9-*l were usedto describethe true shape. These graphically displayed functions were then used to ascertain the appropriate transformations required to meet the linearity assumption. Further variable reduction was required, thus, the stepwisetechnique was used within each of the 4 categories for each outcome. The resultant 4 sets of variables were then used in a final model, again using stepwise techniques to describe best the joint relation of these with each of the 2 outcomes. The added prognostic value of 3- and 6-month echocardiograms was evaluated by adding them as timedependent covariates into Cox proportional-hazards models, which included the baseline echocardiographic data and the baseline resultant variable(s) from stepwise regression as fixed covariates. RESULTS Patient
outcome: The demographics of the study group are listed in Table I. The median time for followup was 252 days (range 105 to 471 days). All patients were potentially eligible for 1 year of clinical follow-up. There were 35 deaths, 30 of which were cardiovascular. Death occurred at a median of 125 days (48, 248 days). Specific causesof cardiovascular death were progressivecongestiveheart failure in 17, sudden death in 7, stroke in 2 and undetermined cardiovascular causes in 4. Other major events included 21 patients who underwent aortic valve replacement, and 13 requiring repeat BAV owing to recurrent symptomsand restenosis. The probability of event-free survival was 38% at 1 year and the probability of cardiac survival was 65% at 1 year (Figure 1). Of the 73 patients who were alive at 1 year, 28 had overall improved symptoms, 11 had symptoms that returned to baseline, and 34 were considered to have overall worse symptoms than before BAV. Congestive heart failure was present at baseline in 86 patients, and was improved by at least 1 functional classin 24 of 27 patients who had not incurred a major event at 1 year. Twenty-six cardiac deaths occurred, 17 had an aortic valve replacement, 12 underwent repeat BAV and 4 noncardiac deaths occurred. Canadian functional class III or IV angina was present in 49 patients before BAV. Twelve of 13 patients evaluated at 1 year were improved by at least 1 functional class. Fifteen incurred a cardiac death, 12
TABLE I Distribution of Baseline Clinical and Catheterization Characteristics Characteristics
Number
Age (years) Men/women Number of coronary arteries narrowed > 50% in diameter 1 2 3 Ejection fraction (%I Ejection fraction
Determinants of One-year Outcome from Balloon Aortic Valvuloplasty Charles J. Davidson, MD, J. Kevin Harrison, MD, Karen S. Pieper, MS, Michael Harding, MD, James B. Hermiller, MD, Katherine Kisslo, RDMS, Cynthia Pierce, RN, and Thomas M. Bashore, MD
Balloon aortic valvuloplasty (BAV) has been a therapeutic alternative treatment for severe symptomatic aortic stenosis. Previous studies have been unable to predict l-year outcome because of limited acute and follow-up clinical, invasive and echocardiographic data. The purpose of this study was to predict long-term outcome based on comprehensive data obtained at the time of valvuloplasty and at 3 and 6 months aiter the procedure. Of 170 consecutive patients undergoing BAV, 108 (mean age 78 years) were at least 1 year from their procedure. Prospective clinical, micromanometer hemodynamic, digital ventriculographic and echocardiographic/ Doppler data were collected at baseline and immediately after the procedure. Echocardiographi6 data were also obtained at 3 and 6 months. With use of Cox model analysis, major events (defined as cardiac death [n = 301, aortic valve replacement [n = 211 qr repeat BAV [n = 131) were predicted by advanced age, baseline heart failure class, and baseline echocardiographic-determined diastolic left ventricular diameter. Only baseline left ventricular ejection fraction proved to be a significant predictor of cardiac death (p = 0.002) in a multivariate model. Absolute values after BAV (stroke work, first derivative of left ventricular pressure, valve area, end-systolic volume, Fick cardiac output, transvalvular gradient) and acute changes measured by catheterization or echocardiography did not provide additional predictive information over that of postprocedure ejection fraction. Similarly, echocardiographic valve area and transvalvular gra-
dient at 3 months added no further prognostic data. With an ejection fraction 145% (n = 63), cardiac survival at 1 year was SO%, irrespective of age, sex, congestive heati failure class or severity of coronary artery disease. Thus, prognosis after BAV can be determined by noninvasive clinical and echocardiographic data obtained before cardiac catheterization. Despite advanced age and concomitant medial problems, patients undergoing BAV with baseline ejection fractions 245% have an excellent cardiac survival to 1 year. (Am J Cardiol 1991;68:75-60)
alloon aortic valvuloplasty (BAV) has been an alternative treatment for severe aortic steno&L-~ Although restenosisand recurrent symptoms occur often at a variable time after the procedure, some patients experience sustained improvement.l-l 2 Previousstudieshave describedshort-term predictors of outcome, based on either invasive or noninvasive parameters.3-7However, analysis of factors predictive of long-term outcome have yet to be delineated. In an elderly and debilitated population of patients, BAV often representsthe only treatment option other than medical therapy. Therefore, it becomesimportant lo clarify its role in the treatment of symptomatic aortic stenosis. Controversy exists as to what factors, if any, constitute a successfulprocedure (e.g., final aortic valve area, and changesin aortic or echocardiographicvalve areas 1 to 3 days after BAV).3-7 Previous studies have limited long-term data and have typically failed to analyze both invasive, hemodynamic and serial echocardiographic data and their relation to patient outcome. Prospectivedetermination of clinical, hemodynamic or From Duke University Medical Center, Department of Medicine, Divi- echocardiographic variables predictive of 1-year outsionof Cardiology, Durham, North Carolina. This study wassupported come would help clarify the patient groups that are in part by Grant 5T32 HL07 101from the National Institutes of Health, National Research Service Award, Bethesda,Maryland. Manuscript likely to benefit from BAV. This study identifies the received January 15, 1991; revised manuscript received March 13, clinical, acute hemodynamic and echocardiographic 1991,and acceptedMarch 17. variables that predict l-year outcome after BAV, and Address for reprints: Charles J. Davidson, MD, Duke University determines whether 3- or &month echocardiographic Medical Center, Department of Medicine, Division of Cardiology, Box 31195,Durham, North Carolina 27710. data provide additional prognostic information.
B
BALLOON AORTIC VALVULOPLASTY
75
METHODS Of 170 consecutivepatients undergoing BAV at Duke University Medical Center, 108 had undergone valvuloplasty at least 1 year previously and thus were eligible for 1 year follow-up. BAV was performed in patients that were deemed high risk for aortic valve replacement after all patients were evaluated by cardiology and cardiothoracic surgery. The median age was 78 years (range 53 to 91). Forty-four percent were men. Before BAV, all patients had severe symptomatic aortic stenosis and were in either New York Heart Association Functional classIII or IV congestive heart failure (n = 86) or Canadian Functional classIII or IV angina (n = 49), or both. Basedon digital subtraction supravalvular aortography and 2-dimensional echocardiography, the cause of aortic stenosis was assumed to be senile degenerative calcilic aortic stenosisin 96 and bicuspid aortic stenosisin 12. Significant coronary artery diseasewas present in 59 (55%). Fourteen had significant 3-vessel disease. Significant diseasewas usually defined as 175% luminal diameter narrowing. Valvuloplasty procedure: BAV was performed after informed consent from a previously approved institutional review board protocol. A 20-mm balloon was used as the initial balloon in all patients. The retrograde technique using a single balloon (Mansfield, Inc., Mansfield, Massachusetts)of 20 or 23 mm was used in 92 and 6 patients, respectively. Double balloons of 20 and 10 mm were used in 10 patients. Left ventricular and aortic pressureswere simultaneously measured immediately before and after BAV with dual-sensor, high-fidelity micromanometer catheters (Millar, Inc., Houston, Texas). R-wave gated masked-mode digital subtraction left ventriculography and supravalvular aortography were obtained before and after the intervention in 106 of 108 patients. Left ventricular volume was calculated with the area-length formula and a regression equation based on 20 previously calibrated human heart casts of known volume. Aortic regurgitation was graded 0 to 4+ with a standard angiographic grading scale. Angiographic imageswere acquired with an ADAC 4100-C digital radiography system interfaced with a GE- MPX L/U x-ray unit. Images were recorded in a 9-inch image intensifier mode using a 256 X 256 X 8 pixel matrix at 30 frames per second. Oxygen consumption was measuredwith a metabolic cart and Fick cardiac output was obtained. The effective aortic valve area was calculated using the Gorlin formula. khocardiographic data: Echocardiographic data were obtained with a Hewlett-Packard 7702OAphasedPatient population:
76
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
array sector scanner immediately before and 1 to 3 days after valvuloplasty in all patients. Fifty-eight patients underwent a 3-month echocardiographic study, and 53 patients a 6-month study. Left ventricular enddiastolic dimension was measuredfrom the parasternal view using M-mode or 2-dimensional imaging according to the recommendationsof the American Society of Echocardiography.13,14Subaortic diameter was measured immediately below the aortic cusps in the longaxis parasternal view. The subaortic area was calculated from diameter measurementsduring 3 to 5 beats in systole. Subaortic velocity was measured by pulsed Doppler and displayed by fast-Fourier transform spectral analysis. Signals were obtained from the cardiac apex with a cursor positionedin the outflow tract at the same level as the diameter measurement. Transducer angulation was optimized to obtain maximal velocity. Peak systolic velocity was calculated for 5 cardiac cycles in patients with sinus rhythm and 10 cycles in patients with atria1 fibrillation. Continuous-wave Doppler was performed with a 1.9-MHz transducer, and velocities were optimized. Peak instantaneousand mean gradients were calculated from resident software. Aortic valve area was calculated with the continuity equation.r5 Paiient evaluation: Patient status was evaluated by physician interview and examination at baselineand at 3 months, 6 months and 1 year after BAV. End points were achieved when the patient died, underwent aortic valve replacement, underwent repeat BAV, or survived to 1 year after the procedure. If death occurred, telephone interview with both the primary care physician and family were obtained in order to classify the cause as either cardiac or noncardiac. Statistics: All continuous results were expressedas median (twenty-ftith, seventy-fnth percentile). Categorical variables were expressed as percentages.The Kaplan-Meier methodi was used to calculate median survival rates. All patients were followed for 1 year unlessthey died or had a major event, at which date they were withdrawn and analyzed as alive (i.e., event-free). To evaluate the changes in values from before to after valvuloplasty, paired t tests were used after the assumption of normality was checked and accepted.A Wilcoxon rank-sum test was used to compare baseline echocardiographic measurementsin patients with and without a 3-month echocardiogram. Becauseof the large number of variables collected in relation to the relatively small number of events, a subsetof potential prognostic variables from each of the categories (clinical, hemodynamic 1, hemodynamic 2 and echocardiographic) was identified. The Cox proportional-hazards regressionmodel*7*18 was used to ex-
JULY 1, 1991
amine the univariable relation of each with the outcomesof cardiovascular death and total cardiac events (cardiac death, aortic valve replacement or repeat valvuloplasty). When the relation between the variable of interest and the logarithm of the hazard ratio was assumedto be linear, spline functionst9-*l were usedto describethe true shape. These graphically displayed functions were then used to ascertain the appropriate transformations required to meet the linearity assumption. Further variable reduction was required, thus, the stepwisetechnique was used within each of the 4 categories for each outcome. The resultant 4 sets of variables were then used in a final model, again using stepwise techniques to describe best the joint relation of these with each of the 2 outcomes. The added prognostic value of 3- and 6-month echocardiograms was evaluated by adding them as timedependent covariates into Cox proportional-hazards models, which included the baseline echocardiographic data and the baseline resultant variable(s) from stepwise regression as fixed covariates. RESULTS Patient
outcome: The demographics of the study group are listed in Table I. The median time for followup was 252 days (range 105 to 471 days). All patients were potentially eligible for 1 year of clinical follow-up. There were 35 deaths, 30 of which were cardiovascular. Death occurred at a median of 125 days (48, 248 days). Specific causesof cardiovascular death were progressivecongestiveheart failure in 17, sudden death in 7, stroke in 2 and undetermined cardiovascular causes in 4. Other major events included 21 patients who underwent aortic valve replacement, and 13 requiring repeat BAV owing to recurrent symptomsand restenosis. The probability of event-free survival was 38% at 1 year and the probability of cardiac survival was 65% at 1 year (Figure 1). Of the 73 patients who were alive at 1 year, 28 had overall improved symptoms, 11 had symptoms that returned to baseline, and 34 were considered to have overall worse symptoms than before BAV. Congestive heart failure was present at baseline in 86 patients, and was improved by at least 1 functional classin 24 of 27 patients who had not incurred a major event at 1 year. Twenty-six cardiac deaths occurred, 17 had an aortic valve replacement, 12 underwent repeat BAV and 4 noncardiac deaths occurred. Canadian functional class III or IV angina was present in 49 patients before BAV. Twelve of 13 patients evaluated at 1 year were improved by at least 1 functional class. Fifteen incurred a cardiac death, 12
TABLE I Distribution of Baseline Clinical and Catheterization Characteristics Characteristics
Number
Age (years) Men/women Number of coronary arteries narrowed > 50% in diameter 1 2 3 Ejection fraction (%I Ejection fraction
