Author + information
- Received March 10, 1997
- Revision received June 10, 1997
- Accepted June 21, 1997
- Published online October 1, 1997.
- David R. Holmes Jr., MD, FACCA,
- Kevin E. Kip, MSPHB,* (, )
- Sheryl F. Kelsey, PhDB,
- Katherine M. Detre, MD, DrPH, FACCB,
- Allan D. Rosen, MSB,1,
- for the NHLBI PTCA Registry Investigators
- ↵*Mr. Kevin E. Kip, University of Pittsburgh, Graduate School of Public Health, Epidemiology Data Coordinating Center, 127 Parran Hall, 130 DeSoto Street, Pittsburgh, Pennsylvania 15261.
Objectives. We examined cause of death in relation to age, length of follow-up and other baseline characteristics in patients in the 1985–1986 National Heart, Lung, and Blood Institute Percutaneous Transluminal Coronary Angioplasty (NHLBI PTCA) Registry.
Background. The manner in which cardiac versus noncardiac mortality of patients with coronary revascularization varies in relation to patient and study characteristics has not been well documented.
Methods. Cause of death determined from a review of 5 years of annual follow-up forms and death certificates was analyzed in 2,127 patients who had coronary angioplasty (mean age 57.6 years) without acute myocardial infarction.
Results. Within 5 years of the initial procedure, there were 205 deaths (9.6%), with 52.7% attributed to cardiac causes. Patients with a low baseline ejection fraction, history of hypertension, previous bypass surgery, previous myocardial infarction, inoperable or high surgical risk or multivessel disease had significantly higher 5-year cardiac mortality. Patients with a history of diabetes, congestive heart failure or severe concomitant noncardiac disease had higher rates of both cardiac and noncardiac mortality. As length of follow-up increased, older patients died of noncardiac causes more often than cardiac causes. Age ≥65 years was a strong independent predictor of 5-year noncardiac mortality (p < 0.001), but not cardiac mortality (p = 0.08).
Conclusions. All-cause mortality rates may be high in elderly revascularized patients, yet cardiac mortality may be less than that expected because of a high risk of noncardiac death. Although all-cause mortality is a more reliable end point than cause-specific mortality, both cardiac and all-cause mortality should be considered in coronary intervention studies involving older patients and long-term follow-up.
Over the past two decades, long-term mortality has been used as an end point after coronary revascularization [1, 2]. In some reports [3, 4], cardiac and noncardiac causes of death have been analyzed together, whereas in others they have been separated. The relation between age, length of follow-up and cause of mortality (cardiac vs. noncardiac) has not been well documented. This is important because mortality from noncardiac causes may, theoretically, obscure interpretation of the long-term effects of coronary revascularization, particularly in nonrandomized studies. Indeed, the results from several longitudinal (>1 year) coronary intervention studies (Appendix A, Table A1) suggest that the cardiac/noncardiac mortality rate ratio tends to decrease as patient age within the study cohort or length of study follow-up, or both, increase. Accordingly, we examined cause of death over a 5-year follow-up period in relation to age, time since first percutaneous transluminal coronary angioplasty (PTCA) procedure and other patient characteristics in the 1985–1986 National Heart, Lung, and Blood Institute (NHLBI) PTCA Registry.
1.1 Patient group
From 16 participating NHLBI PTCA Registry sites, 2,127 consecutive patients who underwent first-time PTCA outside the setting of an acute myocardial infarction (MI) (within the last 10 days) were included, as previously described [2, 5–7]. All subjects gave written informed consent for routine annual follow-up after the initial procedure. The study protocol was approved by the Institutional Review Board for Biomedical Research at the University of Pittsburgh, Coordinating Center for the Registry. The mean age of the cohort was 57.6 years (27% were ≥65 years old), 25.6% were women, 51.5% had multivessel disease and 97.0% were symptomatic before or at the time of the initial procedure, or both.
1.2 Patient follow-up and ascertainment of cause of death
Five-year follow-up data were available on 2,064 (97%) of the baseline cohort of 2,127 patients. Of the remaining 63 patients with incomplete follow-up, 40 (1.9%) were lost to follow-up and 23 (1.1%) refused follow-up. In conjunction with routine annual contact made with participating site coordinators since 1985–1986, periodic searches of the National Death Index were conducted to identify previously unknown deaths among patients with incomplete follow-up data. Death certificates were obtained for all but five patients known to have died within 5 years of their index procedure. Cause of death was ascertained by physician review (D.R.H.) of patient follow-up forms and death certificates. Each patient was classified into one of the following mutually exclusive categories: alive, periprocedural death, death due to congestive heart failure (CHF), sudden cardiac death, death due to MI or noncardiac cause of death. Deaths classified as periprocedural included patients with complications during their index procedure or as a result of repeat revascularization either by PTCA or coronary artery bypass graft surgery (CABG). For the five deceased patients without death certificates, cause of death was determined by a review of patient follow-up forms and baseline medical history (e.g., history of CHF, number of diseased vessels). Three of these five patients were considered to have died of cardiac causes. Parallel analyses were performed with and without these five patients, and the results were virtually identical.
1.3 Statistical analysis
Cardiac and noncardiac mortality over 5 years of follow-up was estimated by use of life-table analysis . This included estimates within baseline subgroups. Deaths due to noncardiac causes were censored in determining cardiac mortality, and vice versa. Similarly, in the estimation of individual cardiac causes of death, deaths from all other causes were censored. To illustrate cause-specific censoring (e.g., to determine cardiac mortality) for patients who did not die from a cardiac cause, follow-up time was determined by the maximal number of days from study entry to last follow-up or death of a noncardiac cause. Although death is a one-time event, estimating mortality rates in this manner theoretically permits a cause-specific death after a patient is censored for another cause-specific death . Because theoretic cause-specific deaths are not mutually exclusive, the sum of mortality rates for individual causes will always be greater or equal to the mortality rate of the combined causes (i.e., cardiac mortality plus noncardiac mortality greater than or equal to all-cause mortality) . Cox regression analysis was used to assess the relative risk (RR) of older age (≥65 years) and other baseline factors on cardiac, noncardiac and all-cause mortality at different follow-up intervals. A cumulative set of 25 baseline (predilation) variables that were collected from the baseline form (and were primarily cardiac in nature) were initially screened by univariable analysis (at p < 0.10). The remaining variables were subsequently assessed (at p < 0.10) using stepwise selection. Baseline ejection fraction was excluded from stepwise models because of substantial missing data and because attempts to impute values for the missing data (using other baseline variables) proved to be unreliable. Estimates of RR were declared significant at p < 0.05.
2.1 Cause-specific mortality and relation to follow-up
There were 205 deaths (9.6%) in the cohort within 5 years of the initial procedure, with slightly more than half of the deaths (52.7%) being attributed to cardiac causes (Table 1). The leading cardiac causes of death were sudden death and MI. Twenty-two deaths (1.0%) occurred within the first 30 days of follow-up, 17 of which were from cardiac causes (Fig. 1). Hence, roughly 10% of the total deaths occurred within the first 30 days. During the first 2 years after the initial procedure, patients died increasingly more often from cardiac causes than from noncardiac causes. In contrast, during years 3 through 5, the divergence between cardiac and noncardiac mortality decreased. Roughly 80% of the periprocedural deaths occurred within the first year after the initial procedure (Fig. 2), although, on average, 42% of the total cohort underwent repeat PTCA or CABG, or both, at some time during the 5-year follow-up. Among the 18 patients who died of a periprocedural cause, the median time to first repeat revascularization was 36.5 days, and the median time to death was 139 days. The occurrence of CHF, MI and sudden cardiac death increased steadily over the study period, with nearly 2% of all patients having died of sudden cardiac death by 5 years.
2.2 Subgroup cause-specific mortality
Table 2displays the 5-year incidence of cardiac and noncardiac mortality in relation to baseline characteristics. Patients with low ejection fraction, history of hypertension, previous CABG, previous MI, inoperable or high surgical risk or triple-vessel disease had significantly higher 5-year cardiac mortality. These prognostic factors were not evident for noncardiac mortality. In contrast, patients ≥65 years old or with a history of diabetes, CHF or severe concomitant noncardiac disease had significantly higher cardiac and noncardiac mortality.
2.3 Predictors of cardiac and all-cause mortality
During the first year after the initial procedure, history of CHF, inoperable or high surgical risk status and older patient age were strong independent predictors of both cardiac and all-cause mortality (Table 3). The consistency of these prognostic factors was influenced by the preponderance of deaths that occurred from cardiac causes (36 of 58) during the first year. In years 2 through 5, 72 (49.0%) of 147 deaths were attributed to cardiac causes. Severe concomitant noncardiac disease at baseline and history of hypertension emerged as significant predictors of cardiac and all-cause mortality. In addition, variant angina and previous CABG conferred increased risk of cardiac-only mortality during this interval. Older patient age (≥65 years) was predictive of all-cause mortality during years 2 through 5, but not cardiac-only mortality.
2.4 Cardiac and noncardiac mortality in relation to patient age and follow-up
Consistent with the results in Table 3a pronounced shift from cardiac to noncardiac mortality was evident in older patients (Fig. 3), with patients ≥65 years old initially more likely to die of cardiac causes, yet by 5 years, more likely to have died of noncardiac causes (10.1% vs. 8.2%). Table 4further illustrates the independent effect of patient age on mortality. Older age conferred a more than threefold risk of both cardiac and noncardiac mortality in the first year after the initial procedure. The excess risk associated with older age remained evident for noncardiac mortality in later years (years 2 through 5) of follow-up (RR 2.92, p < 0.001), but was not evident for cardiac-only mortality (RR 1.06, p = 0.83). Overall, older age was a strong independent predictor of 5-year noncardiac mortality (RR 3.30, p < 0.001), yet was not significantly related to cardiac mortality.
The results of this study show that within 5 years of an initial PTCA procedure, patients in the 1985–1986 NHLBI PTCA Registry were about equally likely to die of noncardiac causes of death as of cardiac causes. Patients with classic cardiac risk factors such as previous CABG had higher rates and risk of 5-year cardiac mortality, yet some established cardiac risk factors (history of CHF and diabetes) were also indicative of elevated rates of noncardiac mortality. From these findings, it is clear that patients undergoing revascularization for symptomatic coronary artery disease often possess a variety of comorbid conditions, both cardiac and noncardiac in nature, and these conditions may yield high rates of both cardiac and noncardiac mortality.
3.1 Competing risks associated with age and follow-up
As suggested by previous reports of cardiac and noncardiac mortality in patients who have had CABG or PTCA (Appendix A), older age and increased length of follow-up were predictive of a shift from cardiac to noncardiac causes of death in patients in the Registry. Because age is a risk factor for most chronic diseases, the probability of patients of advanced age with cardiovascular disease dying of a noncardiac cause is higher than in that of younger patients. Moreover, as the time of study follow-up increases, the relative influence of periprocedural mortality on the cardiac/noncardiac mortality rate ratio diminishes. Hence, the competing risk of death from noncardiac causes in patients undergoing coronary revascularization is related to both age at study entry and length of study follow-up.
3.2 Methodologic considerations
Our results have several important methodologic implications. First, in studies of patients undergoing coronary revascularization, both all-cause mortality and cardiac-only mortality should be considered as study outcomes if the cohort under study is relatively advanced in age and to be followed long-term. This seems particularly relevant to today’s interventional cardiology environment, where patients with advanced age and advanced coronary artery disease routinely undergo percutaneous and surgical procedures. Cardiac mortality is an important end point for evaluating the specific impact of coronary revascularization. However, all-cause mortality is also necessary because of the known unreliability of cause of death ascertainment from death certificate data [12–15]. Moreover, the cost and manpower associated with cause of death adjudication within a study can be substantial. Therefore, all-cause mortality should never be replaced by cardiac-only mortality, but rather the two end points should be complementary. In addition, nonfatal clinical outcomes (i.e., repeat revascularization) and quality-of-life measures (i.e., functional status) are important and useful as outcomes in studies involving elderly patients and long-term follow-up.
Second, when all-cause mortality is used as an end point in coronary revascularization studies, sufficient consideration should be given (in both the planning and analysis phases of the study) to the expected noncardiac disease mortality to be experienced by the cohort during the follow-up period. High noncardiac mortality may lead to inappropriate study conclusions. For example, a previously suggested survival advantage of a particular revascularization strategy in a cohort of younger patients might appear less profound in a cohort of older-aged patients. In short, if noncardiac mortality is high and is similar across revascularization strategies, the relative advantage of a particular therapy on all-cause mortality will be diluted. In contrast, direct assessment of cardiac mortality is also obscured by the competing risk of a noncardiac death. Cardiac mortality rates in an elderly population may be less than expected because of a high risk for noncardiac death.
Finally, given the shift from cardiac to noncardiac causes of death as age and length of follow-up increase, analyses of coronary revascularization cases may be most informative if early periprocedural mortality is described separately from long-term mortality and the cohort is subdivided into discrete follow-up intervals (i.e., first year, years 2 through 5).
3.3 Study limitations
Despite a careful review of all available documentation, cause of death ascertainment was based largely on death certificate data and was determined by a single physician. As previously noted, cause of death ascertainment from death certificates may be unreliable [12–15], and misclassification of cause of death could have occurred. Assuming nondifferential cause of death misclassification, the probable net effect would be biased toward the null. This rationale would also apply to the 63 patients (3% of study cohort) with incomplete follow-up in whom misclassification may have occurred. The life-table estimates for cardiac and noncardiac mortality in this report also assume statistical independence of the times to cardiac and noncardiac death—an assumption that cannot be tested.
In addition, all of the Registry patients were treated in 1985–1986, with 5-year follow-up extending into 1991. Technologic advances in the field of interventional cardiology over the last decade, including widespread use of stents and increased use of platelet glycoprotein IIb/IIIa inhibitors, may be providing improved long-term survival after more recent coronary revascularization. Therefore, it is possible that the cardiac/noncardiac mortality rate ratio would be different in a cohort of similar patients treated today and followed long-term. This possibility further underscores the importance of investigating all-cause and cause-specific mortality in coronary intervention studies involving elderly patients and long-term follow-up.
In this follow-up study of 2,127 consecutive patients undergoing initial PTCA in the 1985–1986 NHLBI PTCA Registry, at 5 years after the index procedure, 9.6% had died. Cardiac causes were identified in 52.7%, with the leading cause secondary to sudden cardiac death or MI. The majority of cardiac deaths occurred within the initial 2 years; after this, the majority of deaths were related to noncardiac causes. On multivariate analysis, age ≥65 years was an independent predictor of 5-year mortality but not cardiac mortality. Although all-cause mortality is a more reliable end point than cause-specific mortality, both cardiac and all-cause mortality should be considered in coronary intervention studies involving older patients and long-term follow-up.
Appendix A Appendix
- coronary artery bypass graft surgery
- congestive heart failure
- myocardial infarction
- National Heart, Lung, and Blood Institute
- percutaneous transluminal coronary angioplasty
- relative risk
- Received March 10, 1997.
- Revision received June 10, 1997.
- Accepted June 21, 1997.
- The American College of Cardiology
- King SB III.,
- Schlumpf M
- Detre K,
- Yeh W,
- Kelsey S,
- et al.
- Holmes DR Jr.,
- Holubkov R,
- Vlietstra RE,
- et al.
- Detre KM,
- Holubkov R,
- Kelsey SF,
- et al.
- Mood AM,
- Graybill FA,
- Boes DC
- ↵Cox DR. Regression models and life tables [with discussion]. J R Stat Soc B 1972;74:187–20.