Author + information
- Received April 22, 1999
- Revision received September 20, 1999
- Accepted November 3, 1999
- Published online March 1, 2000.
- Per Kvidal, MD∗,* ( )(, )
- Prof.Reinhold Bergström, PhD‡,
- Lars-Gunnar Hörte, BA, PM§ and
- Elisabeth Ståhle, MD, PhD†
- ↵*Reprint requests and correspondence: Dr. Per Kvidal, Department of Cardiology, University Hospital, S-751 85, Uppsala, Sweden
We sought to evaluate the effects of a number of factors that can potentially determine the optimal time for aortic valve replacement (AVR) and the observed and relative survival after the operation.
Aortic valve replacement is performed in patients within a wide age span, but the proportion of elderly patients is increasing. In survival analyses, adjustment for the effects of age is therefore essential. Analysis of relative survival provides additional information on excess or disease-specific mortality and its risk factors.
Survival was analyzed in 2,359 patients (1,442 without and 917 with concomitant coronary artery bypass graft surgery) undergoing their first AVR. By relating observed survival to that expected among the general Swedish population stratified by age, gender and five-year calendar period, the relative survival and disease-specific survival were estimated.
Early mortality after AVR (death within 30 days) was 5.6%. Relative survival rates (excluding early deaths) after 5, 10 and 15 years were 94.6%, 84.7% and 74.9%, respectively. There was an excess risk of dying during the entire follow-up period. Advanced New York Heart Association functional class, preoperative atrial fibrillation and pure aortic regurgitation were independent risk factors for observed and relative survival. Patients in the oldest age group showed decreased observed survival but excellent relative survival.
Old age was not a risk factor for excess mortality after AVR, whereas atrial fibrillation decreased relative survival substantially.
Aortic valve replacement (AVR) is performed in patients within a wide age range. During recent years the average age of the entire patient population has increased as a result of a higher proportion of elderly patients (1).
In the present study we investigated the impact of risk factors on relative survival after AVR. Calculation of relative survival is one way of measuring excess mortality among patients who have had heart valve replacement in relation to the mortality in the general population. Thus, relative survival and the effects on it by factors related to different aspects of the status of native aortic valve disease at the time of operation provide additional information on patient outcome after heart valve surgery that may be useful when considering the optimal time for surgical repair in the individual patient.
In this study relative survival was estimated in all patients who underwent AVR in Uppsala, Sweden, during a 15-year period. To obtain the relative survival in our study group of 2,359 patients, we compared the survival rate in these patients with that in the total Swedish population, adjusted for gender, age and five-year calendar period.
From January 1980 through December 1995, 2,359 patients underwent primary AVR, with (n = 917) or without (n = 1,442) concomitant coronary artery bypass graft surgery (CABG) at the Department of Thoracic and Cardiovascular Surgery of the University Hospital, Uppsala, Sweden.
There were 1,499 men (64%; mean age 63.2 years, range 18 to 88) and 860 women (36%; mean age 67.4 years, range 21 to 86). Preoperative coronary angiography was performed in all patients ≥50 years old and in all patients with angina or in whom coronary artery disease was suspected on a clinical basis.
All operations were carried out using a standard technique for cardiopulmonary bypass and moderate hypothermia (25 to 32°C).
Our current policy is to recommend a bioprosthesis to patients ≥70 years old. However, the type and trademark of the prosthesis were left to the discretion of each surgeon. Of the 2,359 patients, 1,776 received a mechanical valve (St. Jude, n = 575; Björk-Shiley, n = 625; Duromedics, n = 199; Carbomedic, n = 299; composite graft, n = 20; and other unspecified, n = 58) and 583 patients received a bioprosthesis (Carpentier-Edwards, n = 213; Sorin, n = 196; St. Jude, n = 32; and miscellaneous types, n = 142). Seven patients underwent concomitant mitral annuloplasty.
Patients with a mechanical valve were prescribed life-long treatment, whereas most patients with a bioprosthesis were recommended three months of anticoagulation, unless atrial fibrillation (AF) or other conditions indicated a high risk of thromboembolism. The therapeutic level of prothrombin complex (factors II, VII and X) ranged from 10% to 25%, corresponding to an international normalized ratio (INR) of 2.1 to 4.3.
Data collection, follow-up and outcome events
All clinical data were recorded prospectively and stored in a computer. New York Heart Association (NYHA) functional classification (2)of congestive heart failure was made on the basis of the preoperative interview with the patient. Patients who had slight discomfort in their normal activities but were able to walk a mile at their own speed and could climb stairs slowly without undue discomfort were allocated to functional class IIIA. Patients who could manage only the lightest activity without discomfort, those who were able to walk only short distances without resting and those who had difficulty climbing stairs were allocated to functional class IIIB.
A unique 10-digit national registration number is allocated to every Swedish citizen. In January 1996, all patients were followed up with respect to survival by computerized linkage to two national registers—namely, the Swedish Cause of Death Register and a continuously updated population register. By use of these combined registers, all patients could be assigned a date of death or identified as being alive on December 31, 1995. The mean follow-up period was 71.5 months.
Early mortality (death from any cause within 30 days postoperatively)
For identification of factors related to early outcome, logistic regression analysis was performed (3). The odds ratio (OR) from this analysis was used as a measure of the relative risk.
Long-term survival (death from any cause after 30 days postoperatively)
The observed survival rate for all causes of death was calculated by the actuarial (life-table) method.
The following variables were entered into the Cox analyses of observed survival: demographic variables (age at operation, gender, year of surgery), disease history (previous myocardial infarction), symptoms and clinical status (dyspnea, left heart failure, NYHA functional class, preoperative heart rhythm [sinus rhythm, AF or other]), associated conditions (hypertension, diabetes, other serious diseases [e.g., malignancies]), preoperative catheterization data (presence of significant coronary artery disease [i.e., ≥50% stenosis in at least one coronary artery], type of lesion [pure aortic regurgitation, aortic stenosis or combined lesion]) and characteristics of the surgical procedure (concomitant CABG, mechanical vs. biologic prosthesis).
Age as a variable was used both as a linear continuous variable and in categorized form. There was little difference between the explanatory power of the two alternatives (log rank statistics: 124.17 with 1 dffor age in continuous form and 117.46 with 3 dffor variables categorized into four 10-year intervals as shown in Table 1). In the age-adjusted models, the difference between the estimates of the effects of the other variables was small enough to be of no consequence. Thus, age can be used in either categorized or continuous form, depending on which alternative is more suitable in a certain situation.
The relative survival rate was computed as the ratio of the observed to the expected survival rate (4,5). The expected survival rates were calculated from life-tables compiled from the total population of Sweden stratified by gender, five-year age group and five-year calendar period (6). The ratio O/E—where O = observed number of deaths and E = expected number of deaths in a cohort in the general population similar to the study cohort of patients with respect to gender, age and intervention period—was also calculated. The reference cohort in the general population was updated annually, with correction for withdrawals. Univariate and multivariate analyses of relative survival were performed for the factors that gave independent information on observed survival in the multivariate Cox model. The multivariate analyses were based on a multiplicative model, where the ratio O/E (or corresponding death risk) was assumed to depend on the explanatory variables of age, NYHA functional class, concomitant CABG, AF, aortic regurgitation and follow-up year. The model was estimated by the maximal likelihood method on the assumption that the observed number of deaths had a Poisson distribution (7). Multivariate analysis of relative survival was based on grouped data, with the explanatory variables categorized as in Table 1. In addition, follow-up time was included, with 15 categories (years 1 through 15).
One hundred and thirty-two patients (5.6%) died within the first postoperative month. In patients undergoing AVR, the following factors influenced early mortality independently: advanced NYHA functional class (IIIB: OR 1.63, 95% confidence interval [CI] 1.10 to 2.39; class IV: OR 4.26, 95% CI 2.35 to 7.73), preoperative AF (OR 1.74, 95% CI 1.07 to 2.80) and older age (≥70 years: OR 1.45, 95% CI 1.01 to 2.09).
All analyses of long-term survival were based on the 2,227 patients who were alive after one month. Observed and relative survival is depicted in Figure 1. The observed survival rates after 5, 10 and 15 years were 82.9%, 63.4% and 46.2%, respectively. The corresponding figures for relative survival were 94.6%, 84.7% and 74.9%.
The observed death risk during the first year postoperatively was 0.044 (Fig. 2). The risk was at a minimum during the third and fourth years and then increased. During the last few years of follow-up, there were too few events for accurate estimation.
The expected death risk was 0.023 during the first year and gradually increased with increasing follow-up time (Fig. 2). During the tenth year it was 0.034. This meant that in the study cohort there was an excess risk of dying during each of the first 12 years of follow-up.
There was a very strong association between age and the other explanatory variables (NYHA functional class, aortic regurgitation, concomitant CABG, and AF) (Table 1). During the study period, the proportion of elderly patients increased (17% of the patients undergoing AVR during the period 1980 through 1985 were ≥71 years old, compared with 34% during the period 1986 through 1990 and 47% during the period 1991 through 1995). The proportion of patients who underwent concomitant CABG also increased, from 24% during the earlier time period to 44% during the period 1991 through 1995. The distribution of the other explanatory variables was virtually unchanged during the study period.
The results of univariate analyses by the Cox proportional hazards model are given in Table 2. After adjustment for age, the previous insignificant effect of aortic regurgitation became significant (Table 2).
The full multivariate model produced results that were very close to those obtained with adjustment for age only. All five variables included in the model became strongly significant (Table 2).
Analyses were performed on two different follow-up periods—0 to 5 years and >5 years (Table 3)—and the effects of the risk factors were allowed to change between the intervals. The effects of aortic regurgitation, concomitant CABG and AF still persisted after five years of follow-up, and in some cases even strengthened. The effect of NYHA functional class decreased markedly, with relative hazards of 1.44 and 1.56 after more than five years, compared with 2.03 and 2.75 during the first five years.
Figures 3 through 7⇓⇓⇓⇓⇓show the relative survival in relation to different categories of the explanatory variables. Patients in NYHA functional class II (Fig. 3)had a relative survival >100% for basically the entire follow-up period (i.e., survival was better than of the corresponding group in the general population cohort). Patients in functional class IIIB or IV had an excess mortality rate of 20% and 32% 10 and 15 years, respectively, after AVR.
Age was the exceptional variable, for which the relative survival analyses yielded quite different results from those based on observed survival (Fig. 4, Table 1). The oldest age group had a relative survival >100% for a large part of the follow-up period.
Observed and expected death risks
When considering basic descriptive data on the observed and expected numbers of deaths (Table 4), we found that in the youngest age group the O/E ratio was 4.5 (31 observed deaths compared with 6.8 expected ones) whereas in the oldest age group there were almost the same numbers of observed (n = 212) and expected deaths (n = 208.2), with an O/E ratio of 1.02.
The full multivariate model is shown in Table 5. The age effect was strongly negative, with a relative risk of 0.21 for the oldest age group compared with the youngest age group. For the other explanatory variables, the difference between analyses based on observed survival, with adjustment for age and relative survival, was limited. The relative hazard (RH) associated with concomitant CABG was now so small that it was not significant at the 5% level (p = 0.095, RH 1.15, 95% CI 0.98 to 1.36).
This study was a prospective study of a substantial number of consecutive patients (n = 2,359) who underwent AVR. The numbers of patients within the different subgroups were adequate for valid generalization of the results to patients with the characteristics in question. With a mean follow-up time of 6 years, corresponding to 14,046 patient-years, the present study has given us one of the most comprehensive analyses of survival after AVR.
In this study we used as our control the “expected” survival of the Swedish population. The survival rates of the total patient group did not return to the normal expected survival of their age- and gender-matched cohort in the general population. The study results underline the palliative rather than curative potential of heart valve replacement in most patient groups (1,8–11).
Factors that influence survival after AVR
The presence of AF preoperatively reduced the observed and relative survival substantially in patients with an aortic lesion. Among patients who underwent AVR and had preoperative AF, 40% were alive after 10 years, corresponding to an excess mortality rate of 42%. After 15 years, only 18% of the patients were alive, corresponding to an excess mortality rate of 67%. Atrial fibrillation is a known independent risk factor for deaths after AVR (12). This might be attributed to more advanced myocardial dysfunction, to an associated increased risk of lethal thromboembolism or to the occurrence of major hemorrhages in patients with biologic valve substitutes who otherwise would not be anticoagulated. However, AF may also imply a compromised hemodynamic situation and disturbed heart rate regulation during both rest and physical stress (13). This may counteract the recovery of the left ventricular morphology as well as geometry after the operation, thereby accelerating the development of heart failure (14,15).
This study also confirms previous findings of increased excess mortality among patients with aortic regurgitation (10–12). These patients are younger than patients with isolated aortic stenosis or mixed valve disease. Both the etiology and clinical characteristics in this type of lesion differ from those in pure stenosis or the combined lesion.
Patients with concomitant CABG showed decreased observed long-term survival. However, these patients were older at the time of operation. Consequently, age adjustment reduced the effect on observed survival, and there was no significant difference in relative survival. However, we saw a trend toward increased disease-specific mortality after ∼8 to 10 years, probably indicating vein graft degeneration or worsening coronary artery disease (16). The number of patients operated on with concomitant CABG increased during recent years. Hence, with prolonged follow-up, concomitant ischemic heart disease may prove to be a relative risk factor in this study group.
Patients in preoperative NYHA functional class IIIB or IV had a significantly higher operative mortality and worse postoperative observed and relative survival in comparison with those in class I or II. In contrast, patients with minimal or no symptoms (class I or II) had low operative mortality and excellent long-term survival not different from the expected survival. However, AVR should not be considered curative in these patients. The optimal timing of such replacement is a difficult decision that cannot be based solely on relative survival analyses. To be able to recommend the operation to each individual patient at the optimal point in the natural course of aortic valve disease, a number of considerations must be taken into account. Despite excellent relative survival, the valve prosthesis introduces the patient to a new disease process in which complications include thromboembolism, anticoagulant-related bleeding, infective endocarditis and structural deterioration of the valve.
Our multivariate analyses were based on an essentially multiplicative Poisson model, which means that the excess mortality is obtained by modeling of the ratio of the observed to the expected risk of death. The regression model most closely connected with relative survival, as used in the descriptive analyses, is the mixed additive-multiplicative model that was previously used by our group (17–20). Here the excess (disease-specific) mortality is modeled additively as the difference between the observed death hazard and the hazard expected in the comparison group from the general population. In cases where the excess mortality defined in this way is very small, the mixed additive-multiplicative model becomes doubtful. A more detailed discussion of the models can be found in the Statistical Appendix.
Several factors must be taken into account when the relative survival function (4,21–23)is used to measure the outcome. First, in calculations of the expected survival in the study group, it is assumed that the survival in the general population is unaffected by deaths related to the disease being studied. Because deaths from heart valve disease are uncommon among younger people in the general population, the relative survival rates calculated in this study should reflect the disease-specific death risk, at least in the younger age groups. However, with increasing age, the prevalence rates of heart valve disease (24)and coronary artery disease in the general population increase, and these constitute a common cause of death. Second, patients accepted for heart valve surgery are selected with respect to their better overall medical condition concerning noncardiac diseases at the time of operation. This is probably more pronounced in the older age groups and will partly explain their high relative survival rate. Furthermore, the present study is focused on the prognosis in patients surviving the immediate postoperative period. Exclusion of deaths within the first month also biases the results in favor of operation. If deaths within the first 30 days postoperatively in, for example, the highest age group were considered (n = 56), the number of deaths in this category during the 15 years of follow-up would be 268, as compared with the 208.2 expected in that age group. In the youngest age group, the observed number of deaths would increase by another 30% if operative deaths were considered.
It is a limitation of our study that left ventricular function, a major predictor of long-term survival after AVR (1,25), was not considered. However, data on NYHA functional class and preoperative AF were available in all patients and may at least theoretically reflect different aspects of left ventricular performance.
This study emphasizes the impact of age on all of the risk factors we studied. The patients showed an excess mortality during the entire follow-up period. Three characteristics on study inclusion—namely, advanced NYHA functional class, AF and pure aortic regurgitation—were independent risk factors, whereas older age was not. Studies of relative survival provide information supplementary to the estimation of the benefits in different age groups.
In an analysis based on relative survival, a formal model must be chosen. It is not self-evident which model should be used, and several alternatives are available. The relative survival (RS) in a group with a death risk of P1is obviously RS = (1 − P1/(1 − P0), if the corresponding risk in the general population is P0. The relative survival is approximately equal to 1 − (P1− P0) for low P0and P1. Thus, the basically additive model is closely connected with the concept of relative survival, which supports the use of this type of model when the basic descriptive results are presented in the form of relative survival.
However, in cases where the excess (disease-specific) mortality is very small, the mixed additive-multiplicative model becomes doubtful. Therefore, in the present study, the multiplicative model based on the Poisson distribution was used for multivariate analysis of relative survival. This model is formulated basically in terms of the ratio P1/P0and compares the observed number of deaths (or the observed death risk) with the expected number of deaths (or the general death risk). The results of this analysis can sometimes look completely different from those based on relative survival, even though the qualitative conclusions are often the same. The use of the multiplicative model is the reason for the discrepancy between the age effect on relative survival depicted in Figure 4and the relative hazards obtained in the multivariate analysis.
☆ Financial support was provided by the Swedish National Association against Heart and Lung Disease and the Uppsala County Association Against Heart and Lung Disease.
- atrial fibrillation
- aortic valve replacement
- coronary artery bypass graft surgery
- confidence interval
- New York Heart Association
- odds ratio
- relative hazard
- Received April 22, 1999.
- Revision received September 20, 1999.
- Accepted November 3, 1999.
- American College of Cardiology
- Lund O
- Wade O.L,
- Bishop J.M
- Kleinbaum D.G,
- Kupper L.L,
- Morgenstern H
- Lawless J.F
- Blackstone E.H
- Hakulinen T,
- Abeywickrama K.H
- Lindblom D,
- Lindblom U,
- Quist J,
- Lundström H
- Abdelnoor M,
- Nitter-Hauge S,
- Trettli S
- Verheul H.A,
- van den Brink R.B.A,
- Bouma B.J,
- et al.
- Botker H.E,
- Toft P,
- Klitgaard N.A,
- Simonsen E.E
- Ståhle E,
- Bergström R,
- Holmberg L,
- et al.
- Ståhle E,
- Kvidal P,
- Nyström S.O,
- Bergström R
- Cutler S.J,
- Axtell L.M
- Lindroos M,
- Kupari M,
- Heikkila Tilvis R