The risk scores developed for the prediction of an adverse outcome in patients after ST-segment elevation myocardial infarction (STEMI) have mostly addressed patients treated with thrombolysis and evaluated solely all-cause mortality as the primary end point. Primary percutaneous coronary intervention in patients with STEMI has improved the outcome significantly and might have changed the relative contribution of different risk factors. Our patient population included 1,484 consecutive patients admitted with STEMI who had undergone primary percutaneous coronary intervention. The clinical, angiographic, and echocardiographic data obtained during hospitalization were used to derive a risk score for the prediction of short-term (30-day) and long-term (1- and 4-year) cardiovascular mortality and hospitalization for heart failure. During a median follow-up of 30 months, 87 patients (6%) died from cardiovascular mortality or were hospitalized for heart failure. Multivariate Cox regression analyses identified age ≥70 years, Killip class ≥2, diabetes, left anterior descending coronary artery as the culprit vessel, 3-vessel disease, peak cardiac troponin T level ≥3.5 μg/L, left ventricular ejection fraction ≤40%, and heart rate at discharge ≥70 beats/min as relevant factors for the construction of the risk score. The discriminatory power of the model as assessed using the areas under the receiver operating characteristic curves was good (0.84, 0.83, and 0.81 at 30 days and 1 and 4 years, respectively), and the patients could be allocated to low-, intermediate-, or high-risk categories with an event rate of 1%, 6%, and 24%, respectively. In conclusion, the current risk model demonstrates for the first time that 8 parameters readily available during the hospitalization of patients with STEMI treated with primary percutaneous coronary intervention can accurately stratify patients at long-term follow-up (≤4 years after the index infarction) into low-, intermediate-, and high-risk categories.
Several risk scores have been proposed for predicting long-term survival after myocardial infarction. However, most of them were developed from patient cohorts treated with thrombolysis. In the Western countries, patients with ST-segment elevation acute myocardial infarction (STEMI) should preferably be treated with primary percutaneous coronary intervention (PCI). Primary PCI in patients with STEMI results in limited infarct size and preserved left ventricular systolic function. As previously shown, the infarct size and left ventricular ejection fraction are powerful determinants of long-term survival in these populations and form a part of established risk scores. However, the wide use of primary PCI might have changed the relative contribution of these parameters to the prediction of long-term outcome. Data concerning which risk factors are most important in this contemporary population of patients for the prediction of cardiovascular mortality and heart failure hospitalization during long-term follow-up are not available. In addition, risk models focusing on cardiovascular mortality and the development of heart failure have not been explored. These outcomes might be more relevant end points in this population than all-cause mortality. Therefore, the aim of the present evaluation was to derive a risk score for the prediction of short- and long-term cardiovascular mortality and hospitalization of heart failure in patients with STEMI treated with primary PCI using clinical, angiographic, and echocardiographic parameters available during the hospitalization for the index infarction.
Methods
Since February 2004, the clinical, angiographic, and echocardiographic data from consecutive patients who were admitted with STEMI to the Leiden University Medical Center were prospectively collected in the departmental cardiology information system (EPD-Vision) and retrospectively analyzed. All patients underwent primary PCI and were treated according to the institutional protocol for patients admitted with STEMI (MISSION!). This protocol is based on the most recent American College of Cardiology/American Heart Association/European Society of Cardiology guidelines and includes a prehospital, inhospital, and outpatient clinical framework designed to optimize the care of these patients. Evidence-based medical therapy is initiated early during hospitalization. In addition, the left ventricular ejection fraction is assessed with 2-dimensional echocardiography within 48 hours of admission to refine the risk stratification and clinical treatment of the patients.
For the present evaluation, the clinical, angiographic, and echocardiographic data from consecutive patients admitted with STEMI and who were not in cardiogenic shock at admission were analyzed. From the various clinical, angiographic, and echocardiographic variables that were routinely collected, a practical risk score was created to accurately predict cardiovascular mortality and hospitalization for heart failure at short-term (30-day) and long-term (1- and 4-years) in this contemporary population of patients with STEMI.
Coronary angiography was performed in all patients in the setting of primary PCI. During angiography, the coronary artery in which the culprit lesion was located, the number of diseased vessels (defined as ≥50% diameter stenosis), the time of first balloon dilation, and the final Thrombolysis in Myocardial Infarction flow grade were noted. Thereafter, patients were transferred to the coronary care unit, and 2-dimensional echocardiography was performed within 48 hours of admission. The left ventricular ejection fraction was calculated from the end-systolic and end-diastolic volumes measured at the apical 4- and 2-chamber views using the biplane Simpson method. All measurements were performed by 2 experienced observers. The inter- and intravariability for the echocardiographic measurements were good, as previously published.
All patients were scheduled for visits at the outpatient clinic at 1, 3, 6, and 12 months after treatment according to the protocol. Data on the occurrence of adverse events after discharge were collected by reviewing the medical records, retrieval of survival status through the municipal civil registries, and telephone interviews. The primary end point was defined as a composite of cardiovascular mortality and hospitalization for heart failure. All medical records were reviewed independently by 2 observers, and the primary cause of death was recorded. All deaths were classified as cardiac unless unequivocally proven noncardiac. Hospitalization for heart failure was defined as hospitalization for either new-onset or worsening heart failure. In addition, both cardiovascular mortality and hospitalization for heart failure were assessed as individual end points. Patients without data for the last 6 months were considered lost to clinical follow-up. The data from these patients were included up to the last date of follow-up.
Continuous data are presented as the mean ± SD or median with the 25th and 75th percentiles, as appropriate. Categorical data are presented as frequencies and percentages. Differences in the baseline characteristics between patients who reached the composite end point versus patients who remained event free were evaluated using the unpaired Student t test and chi-square test. Continuous variables that were not normally distributed were compared using the Wilcoxon rank-sum test.
Event rates for cardiovascular mortality and hospitalization for heart failure were analyzed using the Kaplan-Meier method. Differences in event rates were assessed using the log-rank test. In the presence of missing data, the single imputation procedure was applied. In studies with a small number of missing variables (<10% for any parameter), a single imputation has been shown to perform equally to multiple imputation techniques.
To obtain a risk score, composed of robust, reproducible, and nonclinician-driven parameters, the use of medication was not used in the analysis. All variables were entered as categorical variables according to cutoff values previously defined in published studies ( Table 1 ). Age was categorized as ≥70 years or <70 years. Three-vessel disease was defined as ≥50% stenosis in 3 major epicardial branches. The symptom onset to balloon time was categorized as ≥4 hours and <4 hours. The peak cardiac troponin T level was categorized as ≥3.5 μg/L or <3.5 μg/L. The glucose level was categorized as ≥8 mmol/L or <8 mmol/L. The renal clearance was estimated using the formula of Cockcroft-Gault and categorized as abnormal (≤60 ml/min) or normal (>60 ml/min). The left ventricular ejection fraction was categorized as ≤40% and >40%. The heart rate was categorized as ≥70 beats/min or <70 beats/min, and the systolic blood pressure was categorized as ≤100 mm Hg or >100 mm Hg.
| Variable | All Patients (n = 1,484) | Primary End Point (n = 87) | Event Free (n = 1,397) | p Value |
|---|---|---|---|---|
| Age (years) | 61 ± 12 | 65 ± 14 | 61 ± 12 | 0.002 |
| Age ≥70 years | 366 (25%) | 38 (44%) | 328 (24%) | <0.001 |
| Women | 356 (24%) | 20 (23%) | 336 (24%) | 0.82 |
| Killip class | <0.001 | |||
| 1 | 1,430 | 73 | 1,357 | |
| 2 | 40 | 8 | 32 | |
| 3 | 14 | 6 | 8 | |
| Killip class≥2 | 54 (4%) | 14 (16%) | 40 (3%) | <0.001 |
| Current smoker | 702 (47%) | 43 (49%) | 659 (47%) | 0.68 |
| Diabetes mellitus | 175 (12%) | 26 (30%) | 149 (11%) | <0.001 |
| Family history of coronary artery disease | 613 (41%) | 32 (37%) | 581 (42%) | 0.38 |
| Hypercholesterolemia ⁎ | 290 (20%) | 13 (15%) | 277 (20%) | 0.27 |
| Hypertension † | 517 (35%) | 30 (35%) | 487 (35%) | 0.94 |
| Previous myocardial infarction | 129 (9%) | 10 (12%) | 119 (9%) | 0.34 |
| Left anterior descending as culprit artery | 679 (46%) | 55 (63%) | 624 (45%) | 0.001 |
| Number of narrowed coronary arteries | 0.001 | |||
| 1 | 698 | 30 | 668 | |
| 2 | 504 | 27 | 477 | |
| 3 | 282 | 30 | 252 | |
| Three-vessel disease | 282 (19%) | 30 (35%) | 252 (18%) | <0.001 |
| Symptom onset to balloon time (min) | 0.05 | |||
| Median | 174 | 178 | 174 | |
| 25th, 75th percentile | 128, 255 | 144, 291 | 126, 254 | |
| Symptom onset to balloon time ≥240 min | 423 (29%) | 26 (30%) | 397 (28%) | 0.77 |
| Final Thrombolysis in Myocardial Infarction flow | 0.91 | |||
| 0 | 7 | 0 | 7 | |
| 1 | 22 | 1 | 21 | |
| 2 | 79 | 5 | 74 | |
| 3 | 1,376 | 81 | 1,295 | |
| Thrombolysis in Myocardial Infarction flow ≤2 | 108 (7%) | 6 (7%) | 102 (7%) | 0.89 |
| Peak creatine phosphokinase (U/L) | <0.001 | |||
| Median | 1,488 | 3,430 | 1,417 | |
| 25th, 75th percentile | 647, 2,921 | 1689, 5,530 | 619, 2,676 | |
| Peak cardiac troponin T (μg/L) | <0.001 | |||
| Median | 3.8 | 9.2 | 3.6 | |
| 25th, 75th percentile | 1.4, 7.7 | 3.8, 14.5 | 1.4, 7.3 | |
| Peak cardiac troponin T level ≥3.5 μg/L | 772 (52%) | 67 (77%) | 705 (51%) | <0.001 |
| Glucose (mmol/L) | 8.5 ± 3.0 | 9.8 ± 4.2 | 8.4 ± 2.9 | 0.003 |
| Glucose ≥8 mmol/L | 705 (48%) | 51 (59%) | 654 (47%) | 0.03 |
| Estimated glomerular filtration rate (ml/min/1.73 m 2 ) | 98 ± 33 | 89 ± 39 | 99 ± 33 | 0.008 |
| Estimated glomerular filtration rate ≤60 ml/min/1.73 m 2 | 172 (12%) | 67 (77%) | 20 (23%) | 0.001 |
| Left ventricular ejection fraction (%) | 47 ± 9 | 41 ± 10 | 48 ± 9 | <0.001 |
| Left ventricular ejection fraction ≤40% | 315 (21%) | 38 (44%) | 277 (20%) | <0.001 |
| Heart rate at discharge (beats/min) | 70 ± 12 | 77 ± 16 | 70 ± 12 | <0.001 |
| Heart rate ≥70 beats/min | 730 (49%) | 57 (66%) | 673 (48%) | 0.002 |
| Systolic blood pressure at discharge (mm Hg) | 115 ± 16 | 111 ± 17 | 115 ± 16 | 0.02 |
| Systolic blood pressure ≤100 mm Hg | 270 (18%) | 23 (26%) | 247 (18%) | 0.04 |
| Diastolic blood pressure at discharge (mm Hg) | 70 ± 22 | 67 ± 12 | 70 ± 23 | 0.20 |
| Medication at discharge | ||||
| Angiotensin-converting enzyme inhibitors/angiotensin receptor blockers | 1,434 (98%) | 86 (100%) | 1,397 (100%) | 1.00 |
| Antiplatelets | 1,484 (100%) | 82 (95%) | 1,365 (98%) | 0.17 |
| β Blockers | 1,390 (95%) | 83 (97%) | 1,386 (99%) | 0.01 |
| Statins | 1,456 (99%) | 76 (88%) | 1,327 (95%) | 0.008 |
⁎ Total cholesterol ≥190 mg/dl or previous pharmacologic treatment.
† Blood pressure ≥140/90 mm Hg or previous pharmacologic treatment.
Thereafter, univariate Cox regression analysis was performed with the composite end point of cardiovascular mortality and hospitalization for heart failure. All parameters with p <0.10 were included in a multivariate Cox regression model. Using backward stepwise elimination, the least significant parameter was discarded from the model until all parameters had reached a value of p <0.25. Subsequently, each remaining significant variable in the model was assigned a weighted score proportional to the regression coefficient. For this purpose, the base regression coefficient was assigned the value of 1 point, and all variables were given the associated score, according to their multiplication of this base regression coefficient, with rounding to the nearest whole number. Thus, 0.46 was used as the base regression coefficient and was assigned the value of 1 point. The ability of the risk score to discriminate between patients who did and did not reach the composite end point was estimated by the area under the curve of the receiver operator characteristic curve at short-term (30-day) and long-term (1- and 4-year) follow-up. The developed risk score based on the whole study cohort was further evaluated by drawing 1,000 bootstrap samples, with replacement, to estimate the extent to which the predictive accuracy of the model was overoptimistic. The mean C-index and corresponding SEM was reported. In addition, the discriminative capacity of the derived risk score was evaluated for the individual end points cardiovascular mortality and hospitalization for heart failure at 30 days and 1 and 4 years. Finally, after determination of the individual risk score per patient, cutoff values were determined to divide the population into low-, intermediate-, and high-risk categories. These cutoff values were chosen to optimize the discriminative effect of the model without making the different groups too small. p Values <0.05 were considered significant, and analyses were performed with the Statistical Package for Social Sciences, version 16.0 (SPSS, Chicago, Illinois).
Results
A total of 1,523 consecutive patients admitted with STEMI and undergoing primary PCI were evaluated in the present study. During hospitalization for the index infarction, 39 patients (2%) died and were excluded from additional analysis. The final study population included 1,484 patients. The characteristics of all included patients are listed in Table 1 . The mean patient age was 61 ± 12 years, and 76% of the patients were men; 4% of the patients presented with Killip class ≥2, and 12% had diabetes. The left anterior descending coronary artery was the culprit vessel in 46% of the patients, and the peak creatine phosphokinase level and peak cardiac troponin T level was 1,488 U/L (range 647 to 2,921) and 3.8 μg/L (range 1.4 to 7.7), respectively. Baseline echocardiography performed within 48 hours of admission revealed a mean left ventricular ejection fraction of 47 ± 9%. At discharge, the mean heart rate was 70 ± 12 beats/min. In addition, the use of evidence-based medical therapy at discharge was high; 98% of the patients were treated with angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, 95% with β blockers, and 99% with statins.
Clinical follow-up was completed in 1,389 patients (94%), and the median follow-up duration was 30 months (25th and 75th percentile 13 and 48). During this period, 87 patients (6%) reached the composite end point. Of these 87 patients, 52 (4%) died from cardiovascular mortality and 46 (3%) were hospitalized for new-onset or worsening heart failure. A total of 78 patients (5%) died during the follow-up period, and only 67% of the deaths were defined as having a cardiovascular cause. In the present population, the noncardiovascular deaths were mostly from malignancy.
Univariate and subsequent multivariate Cox regression analyses identified 8 variables for the construction of the risk score: age ≥70 years, Killip class ≥2, diabetes, left anterior descending coronary artery as the culprit vessel, 3-vessel disease, peak cardiac troponin T level ≥3.5 μg/L, left ventricular ejection fraction ≤40%, and heart rate at discharge ≥70 beats/min ( Table 2 ). The regression coefficient of heart rate at discharge of 0.46 was used as the base regression coefficient. For each variable, a weighted risk score was assigned according to the corresponding regression coefficient ( Table 2 ). Then, a risk score was calculated for each patient by summing the points for each risk factor present. The area under the receiver operator characteristic curve for the risk score and the composite end point at 30 days and 1 and 4 years was 0.77, 0.81, and 0.79, respectively, indicating good discriminatory power for the model. The mean C-index of the risk score as obtained in the 1,000 bootstrap samples was fairly similar at 0.78 ± 0.04 SEM, 0.82 ± 0.03, and 0.79 ± 0.79 for the composite end point at 30 days and 1 and 4 years, respectively.
