- Email: [email protected]
A Novel Percutaneous Coronary Intervention Risk Score to Predict One-Year Mortality
A Novel Percutaneous Coronary Intervention Risk Score to Predict One-Year Mortality Gabriel Maluenda, MD, Cedric Delhaye, MD, Michael A. Gaglia, Jr., MD, MSc, Itsik Ben-Dor, MD, Manuel A. Gonzalez, MD, MPH, Nicholas N. Hanna, MD, Sara D. Collins, MD, Kohei Wakabayashi, MD, Rebecca Torguson, MPH, Zhenyi Xue, MS, Lowell F. Satler, MD, Augusto D. Pichard, MD, and Ron Waksman, MD* Clinical and angiographic risk factors associated with adverse outcomes after percutaneous coronary intervention (PCI) have been included in previous validated risk scores. Complications after PCI are known to increase mortality and morbidity but have not been included in any model. Records of 6,932 consecutive patients who underwent PCI from 2000 to 2005 were reviewed. Patients presenting with cardiogenic shock were excluded. Logistic regression and bootstrap methods were used to build an integer risk score for estimating risk of death at 1 year after PCI using baseline, angiographic, and procedural characteristics and postprocedural complications. This risk score was validated in a set of consecutive patients who underwent PCI from 2006 to 2007. The following 8 variables were significantly correlated with outcome: older age, history of diabetes mellitus, chronic renal failure, heart failure, left main coronary artery disease, lower baseline hematocrit, greater hematocrit decrease after PCI, and Thrombolysis In Myocardial Infarction grade <3 flow after PCI. In the validation population (n ⴝ 973), average receiver operating characteristic curve area was 0.836. In conclusion, we developed and validated a simple integer risk score, including postprocedural variables that closely predict long-term mortality after PCI. This model emphasizes the significant impact of complications occurring after PCI on long-term outcomes. © 2010 Elsevier Inc. All rights reserved. (Am J Cardiol 2010;106:641– 645) Several statistical models have been developed and validated to predict adverse outcomes and complications after percutaneous coronary intervention (PCI).1–11 To simplify the process of risk prediction, risk scores provide useful information from corresponding logistic regression models to clinicians. All these models include baseline clinical, angiographic, and procedural characteristics in patients presenting with acute myocardial infarction (MI)1,3– 6,11 and in unselected populations2,7,9,10 undergoing PCI. Postprocedural complications such as slow coronary flow after PCI,12–14 bleeding,15,16 and renal failure17 have been reported to negatively affect outcomes, most significantly by increasing mortality. We recently reported the value of hematocrit before and after PCI as a strong predictor for long-term outcome.18 However, none of the previously validated risk scores has included postprocedural complications to assess the negative impact of such adverse events. The goals of the present study were to (1) identify preprocedural, clinical, angiographic, and procedural characteristics and postprocedural risk factors associated with longterm mortality in a consecutive series of patients undergoing contemporary PCI; (2) construct a simple risk score for prediction of long-term mortality after PCI using pre- and postprocedural risk factors; and (3) internally validate this
Department of Internal Medicine, Division of Cardiology, Washington Hospital Center, Washington, DC. Manuscript received March 5, 2010; revised manuscript received and accepted April 7, 2010. *Corresponding author: Tel: 202-877-2812; fax: 202-877-2715. E-mail address: [email protected] (R. Waksman). 0002-9149/10/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.amjcard.2010.04.011
risk score with an independent cohort of patients undergoing PCI. Methods An ongoing registry of catheter-based coronary procedures is maintained at our institution. For data analysis and risk score construction, we included consecutive patients who underwent PCI and completed 1-year follow-up from January 1, 2000 to December 31, 2005. Patients in cardiogenic shock before PCI were excluded from the study, because it is already known this population has a very poor prognosis. Internal validation was performed using patients who underwent PCI from January 1, 2006 to December 31, 2007. This study was approved by the institutional review board at Washington Hospital Center and MedStar Research Institute (Washington, DC). In all cases, the interventional strategy was at the discretion of the responsible physician. Periprocedural anticoagulation was ensured using bivalirudin or unfractionated heparin to achieve an activated clotting time ⬎250 seconds in all patients. Glycoprotein IIb/IIIa inhibitor administration was also at the operator’s discretion. All patients received aspirin 325 mg and continued this regimen indefinitely. All patients received a clopidogrel loading dose of 300 to 600 mg, followed by a maintenance dose of 75 mg indefinitely. Independent research personnel blinded to the objectives of the study conducted clinical follow-up by telephone contact or office visits. In case of hospitalization, data were obtained by systematic review of source documents. All clinwww.ajconline.org
642
The American Journal of Cardiology (www.ajconline.org)
Table 1 Clinical, angiographic/procedural, and postprocedural characteristics of 6,932 patients undergoing percutaneous coronary intervention from 2000 to 2005 Clinical (n ⫽ 6,932) Age (years), mean ⫾ SD 65.1 ⫾ 11.8 Men 4,544 (65.6%) Diabetes mellitus 2,387 (34.8%) Previous myocardial infarction 2,398 (36.8%) Previous percutaneous coronary intervention 1,988 (29.9%) Previous coronary artery bypass grafting 1,668 (24.3%) Heart failure 1,017 (15.7%) Chronic renal failure 872 (12.7%) Stable angina pectoris 2,700 (39.2%) Unstable angina pectoris 3,433 (49.8%) Acute myocardial infarction 755 (11.0%) Hematocrit (%), mean ⫾ SD 39.9 ⫾ 11.9 Angiographic (n ⫽ 12,411 lesions) Left main coronary artery disease 215 (1.7%) Type C–B target lesion 10,847 (92.7%) Left ventricular ejection fraction (%), mean ⫾ SD 48 ⫾ 14 Stenting 10,522 (84.8%) Drug-eluting stent 7,372 (59.4%) Postprocedural (n ⫽ 6,932) Number of narrowed coronary arteries, mean ⫾ SD 1.98 ⫾ 0.88 Multilesion percutaneous coronary intervention 3,452 (47.5%) Thrombolysis In Myocardial Infarction grade ⬍3 flow 264 (3.8%) Hematocrit decrease (%), mean ⫾ SD 2.53 ⫾ 3.5 Creatinine increase ⬎1 mg/dl 162 (2.6%)
ical events were adjudicated by independent physicians who were not involved in the procedures. The outcome of interest was death at 1 year. Death was defined as all-cause mortality. MI was defined as the association of ⱖ1 clinical and ⱖ1 biological criteria: acute onset chest pain and/or typical changes on electrocardiogram (ST or T-wave changes or new left bundle branch block) and an increase of troponin ⬎99th percentile of the upper reference limit.19 Chronic renal insufficiency was defined as the presence of previously documented renal failure and/or a baseline serum creatinine ⬎2.0 mg/dl. Heart failure was defined as objective evidence of fluid retention due to cardiac causes before admission. Angiographic lesions were designated as A, B or C using the American College of Cardiology/ American Heart Association classification.20 Angiographic flow immediately after PCI was classified using Thrombolysis In Myocardial Infarction (TIMI) criteria.14 Hematocrit decrease was defined as the baseline value minus the nadir hematocrit level after PCI. Indications for PCI were classified as stable angina pectoris, unstable angina pectoris (defined as pain at rest or prolonged episodes of pain associated with objective evidence of ischemia), and acute MI. Continuous variables are presented as mean ⫾ SD, and discrete variables are summarized as frequencies and percentages. Logistic regression analysis was used to estimate odds ratios and associated p values for the primary end point. Clinically relevant baseline angiographic, procedural, and postprocedural variables were included in the initial model: age, gender, history of diabetes mellitus, chronic renal failure, heart failure, previous MI, hematocrit at baseline, presentation with acute MI, left main coronary artery disease (ⱖ50% stenosis), multilesion PCI (⬎1 lesion
Table 2 Association of baseline, angiographic/procedural, and postprocedural characteristics with one-year mortality
Age (years)* ⬍50 ⱖ50–⬍60 ⱖ60–⬍70 ⱖ70–⬍80 ⱖ80 Men Diabetes mellitus Previous myocardial infarction Heart failure Chronic renal failure Presentation with acute myocardial infarction Hematocrit at baseline (%)* ⱖ40 ⱖ35–⬍40 ⱖ30–⬍35 ⬍30 Left main coronary artery disease Type C target lesion Multilesion percutaneous coronary intervention Thrombolysis In Myocardial Infarction grade ⬍3 flow Hematocrit decrease (%)* ⬍5 ⱖ5–⬍10 ⱖ10–⬍15 ⱖ15 Creatinine increase ⬎1 mg/dl
Odds Ratio
Confidence Interval
p Value
— 0.76 1.46 3.21 4.35 0.74 2.44 1.85 5.28 4.87 1.44
— 0.4–1.4 0.9–2.5 1.9–5.3 2.6–7.4 0.59–0.92 1.95–3.05 1.47–1.87 4.20–6.63 3.85–6.17 1.05–1.98
— 0.3 0.2 ⬍0.0001 ⬍0.0001 0.008 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.023
— 1.70 4.08 5.27 2.89 1.36 1.20
— 1.3–2.3 3.0–5.5 3.4–8.1 1.88–4.45 1.07–1.75 0.9–1.6
— 0.0002 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.013 0.16
7.62–14.1
⬍0.0001
— 1.3–2.1 2.6–5.9 3.4–11.4 4.86–10.4
— 0.0002 ⬍0.0001 ⬍0.0001 ⬍0.0001
10.4
— 1.61 3.91 6.19 7.12
* Odds rates estimated point using the first category group as a reference for comparison.
treated), American College of Cardiology/American Heart Association type C lesion treated, presence of TIMI grade ⬍3 flow immediately after PCI, hematocrit decrease after PCI as a continuous variable, and creatinine increase after PCI ⱖ1.0 mg/dl from baseline value. The bootstrap method was then used to avoid overfitting the data.21 One thousand bootstrap samples were used. Backward selection with a p value ⬍0.05 for statistical significance was used to remove variables in each sample. Variables selected ⱖ800 times (80%) in the overall sample were included in the final model. To construct a simple risk score, variables identified through the multivariable model were assigned an integer coefficient. Integers were chosen to be approximately proportional to the estimated continuous coefficient from the logistic model. Each risk factor’s corresponding coefficient was added to obtain a final risk score from 0 to 26. For the final score, 5 strata were defined (very low, 0 to 4; low, 5 to 8; moderate, 9 to 10; high, 11 to 14; and very high, ⱖ15) based on predicted event rates of each score. Observed and expected numbers of events were calculated within each group. Model adequacy of the scoring system was then evaluated with the Hosmer-Lemeshow goodness-of-fit test.22 For the validation set of procedures the estimated probability of death at 1-year follow-up was calculated
Coronary Artery Disease/Long-Term Mortality Risk Assessment After PCI
643
Table 3 Adjusted predictors of death at one-year follow-up from index percutaneous coronary intervention Variable Thrombolysis In Myocardial Infarction grade ⬍3 flow14 Heart failure Left main coronary artery disease Chronic renal failure Diabetes mellitus Hematocrit decrease Hematocrit at baseline Age, number of decades after 40 years Intercept
Integer Score
Model Coefficient
Odds Estimated
95% Confidence Interval
p Value
7 4 3 3 2 1 1 1 N/A
2.035 1.155 0.783 0.730 0.527 0.382 0.374 0.296 ⫺5.276
7.65 3.17 2.19 2.07 1.69 1.46 1.45 1.34 N/A
5.3–10.9 2.4–4.1 1.3–3.5 1.6–2.7 1.3–2.2 1.2–1.7 1.3–1.7 1.2–1.5 N/A
⬍0.0001 ⬍0.0001 0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 N/A
Figure 1. Estimated rate of death at 1-year follow-up for the integer score system. Percentages at risk are shown for each of the 5 strata: ⱕ2%, very low risk; ⬎2% to 5%, low risk; ⬎5% to 10%, moderate risk; ⬎10% to 25%, high risk; ⬎25%, very high risk.
using the integer risk score. The discriminatory capacity of the model was assessed using an area under the receiver operating characteristic (ROC) curve, and goodness-of-fit was tested with the Hosmer-Lemeshow statistic (p ⬎0.05 considered to indicate lack of deviation between the model and the observed event rates). All statistical analysis was performed using SAS 9.1 (SAS Institute, Cary, North Carolina). Results A total of 6,932 patients who underwent PCI from January 1, 2000 to December 31, 2005 were included in the study. At 1-year follow-up there were 383 deaths (5.6%). Mean age was 65.1 ⫾ 11.8 years, 65.6% were men, and unstable angina pectoris was the most common indication for PCI. This set of patients represented a high-risk population, with a high prevalence of co-morbidities, as listed in Table 1. This population also had a high prevalence of high-risk angiographic findings; most patients had multivessel disease and complex coronary lesions requiring multilesion PCI. Stenting was used in most cases, with drug-eluting stents used in 60% of procedures. Univariable associations between covariables initially included in the bootstrapping sample and death at 1 year are presented in Table 2.
Five clinical, 1 angiographic, and 2 postprocedural variables were selected in ⱖ80% of bootstrapping samples. Variables selected from the original model included presence of TIMI grade ⬍3 flow after PCI, history of heart failure, left main coronary artery disease, chronic renal failure, diabetes mellitus, hematocrit decrease after PCI and hematocrit at baseline, and age (Table 3). The data showed a lack of deviation from the model, as indicated by the Hosmer-Lemeshow test result (p ⫽ 0.43). Mean area under the ROC curve of the bootstrap samples was 0.818, indicating a good discriminatory capacity between patients who developed death at 1-year follow-up and those who did not. Based on the integer risk score, 39.5% of procedures were classified as very low risk, 32.7% as low risk, 14.7% as moderate risk, 9.6% as high risk, and 3.6% as very high risk. Observed (and expected) rates of death in these strata were 1.0% (ⱕ2%) for very low risk, 3.3% (⬎2% to 5%) for low risk, 7.1% (⬎5% to 10%) for moderate risk, 15.9% (⬎10% to 25%) for high risk, and 43.0% (⬎25%) for very high risk (Figure 1). The validation set included 973 patients who underwent PCI from January 1, 2006 to December 31, 2007. In this population, there were 100 deaths (10.3%) at 1 year. The model predicted 81.5% of mortal events. In this validation
644
The American Journal of Cardiology (www.ajconline.org)
Figure 2. Observed and predicted rates of mortality in the validation set by integer score displayed as risk score in the validation set (x axis), mortality rate at 1-year follow-up (y axis), predicted mortality rate at 1-year derived from the risk score (solid line), and observed rate of 1-year mortality in the validation set (bars).
population the area under the ROC curve was 0.836, and the Hosmer-Lemeshow statistic p value was 0.573, demonstrating that the model fit the data well and correlated appropriately with patient risk (Figure 2). Of 738 presenting with stable/unstable angina pectoris, there were 85 deaths at 1 year, and the model predicted 79.3% of those events (area under the ROC curve 0.816, Hosmer-Lemeshow statistic, p ⫽ 0.933). In addition, in patients presenting with acute MI (n ⫽ 181), there were 24 deaths at 1 year, and the model predicted 88.9% of those (area under the ROC curve 0.903, and Hosmer-Lemeshow statistic, p ⫽ 0.625). Discussion We presented a simple integer risk scoring model that includes preprocedural, procedural, and postprocedural variables, which closely predicts long-term mortality in unselected patients undergoing contemporary PCI. This model, which was validated internally, highlights the significant impact of complications occurring after PCI on long-term outcomes. To our knowledge, this is the first risk model for long-term mortality to include postprocedural complications. In regard to baseline and procedural risk factors, our results agree with previously validated models.1–11 Similar to these risk models, we identified 5 baseline clinical characteristics (older age, presence of heart failure, chronic renal failure, diabetes mellitus, and lower baseline hematocrit) and 1 angiographic variable (presence of left main coronary artery disease) as the most significant predictors before PCI for long-term mortality. In addition, we identified 2 postprocedural variables (TIMI grade ⬍3 flow after PCI and lower hematocrit after PCI) as important predictors after PCI for long-term mortality. Extensive clinical evidence supports the predictive value of slow coronary flow12–14 and development of anemia or bleeding after PCI15,18,23,24 as known markers of poor prognosis. This study emphasizes the significant impact of postprocedural complications on
long-term risk of death and infers the underlying importance of measures to minimize bleeding and to achieve normal coronary flow after PCI. The presented risk score model has an excellent predictive capacity for 1-year mortality (c-statistics 0.84). Its performance remains accurate for patients presenting without acute MI (c-statistics 0.816) and even better for those presenting with acute MI (c-statistics 0.903). In general, previously reported risk score models for unselected patients undergoing PCI have predicted the risk of in-hospital complications after the procedure.2,7,9,10 No previous risk models, however, have included postprocedural complications and have not predicted the long-term risk of mortality after PCI. We believe that inclusion of postprocedural complications allows us to account for the significance of those events on long-term prognosis, and the addition of these components may improve the performance of risk models on long-term outcome prediction. Somewhat unexpectedly, presentation with acute MI was not associated with the primary outcome. Exclusion of patients presenting with cardiogenic shock may have minimized the role of acute MI in the model. In addition, inclusion of postprocedural variables, which are also complications frequently related to presentation with acute MI, could also account for this issue. In addition, procedurerelated renal failure was not part of the final model. The close relation of this complication and previous chronic renal failure may explain the lack of association between this variable and long-term mortality rate. The primary limitation of the present study is that our data are from a single center, and therefore the applicability to other populations remains unknown. Exclusion of patients presenting with cardiogenic shock and how this exclusion might result in omission of other variables in the current score are unclear. Furthermore, operator volume and expertise were not included in the analysis. External validation in a different dataset is therefore required.
Coronary Artery Disease/Long-Term Mortality Risk Assessment After PCI 1. Morrow DA, Antman EM, Charlesworth A, Cairns R, Murphy SA, de Lemos JA, Giugliano RP, McCabe CH, Braunwald E. TIMI risk score for ST-elevation myocardial infarction: A convenient, bedside, clinical score for risk assessment at presentation: An intravenous nPA for treatment of infarcting myocardium early II trial substudy. Circulation 2000;102:2031–2037. 2. Resnic FS, Ohno-Machado L, Selwyn A, Simon DI, Popma JJ. Simplified risk score models accurately predict the risk of major inhospital complications following percutaneous coronary intervention. Am J Cardiol 2001;88:5–9. 3. Addala S, Grines CL, Dixon SR, Stone GW, Boura JA, Ochoa AB, Pellizzon G, O’Neill WW, Kahn JK. Predicting mortality in patients with ST-elevation myocardial infarction treated with primary percutaneous coronary intervention (PAMI risk score). Am J Cardiol 2004; 93:629 – 632. 4. Granger CB, Goldberg RJ, Dabbous O, Pieper KS, Eagle KA, Cannon CP, Van De Werf F, Avezum A, Goodman SG, Flather MD, Fox KA. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med 2003;163:2345–2353. 5. Eagle KA, Lim MJ, Dabbous OH, Pieper KS, Goldberg RJ, Van de Werf F, Goodman SG, Granger CB, Steg PG, Gore JM, Budaj A, Avezum A, Flather MD, Fox KA. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA 2004;291: 2727–2733. 6. Halkin A, Singh M, Nikolsky E, Grines CL, Tcheng JE, Garcia E, Cox DA, Turco M, Stuckey TD, Na Y, Lansky AJ, Gersh BJ, O’Neill WW, Mehran R, Stone GW. Prediction of mortality after primary percutaneous coronary intervention for acute myocardial infarction: the CADILLAC risk score. J Am Coll Cardiol 2005;45:1397–1405. 7. Wu C, Hannan EL, Walford G, Ambrose JA, Holmes DR Jr, King SB III, Clark LT, Katz S, Sharma S, Jones RH. A risk score to predict in-hospital mortality for percutaneous coronary interventions. J Am Coll Cardiol 2006;47:654 – 660. 8. Singh M, Rihal CS, Selzer F, Kip KE, Detre K, Holmes DR. Validation of Mayo Clinic risk adjustment model for in-hospital complications after percutaneous coronary interventions, using the National Heart, Lung, and Blood Institute dynamic registry. J Am Coll Cardiol 2003;42:1722–1728. 9. Singh M, Lennon RJ, Holmes DR Jr, Bell MR, Rihal CS. Correlates of procedural complications and a simple integer risk score for percutaneous coronary intervention. J Am Coll Cardiol 2002;40:387–393. 10. Madan P, Elayda MA, Lee VV, Wilson JM. Predicting major adverse cardiac events after percutaneous coronary intervention: the Texas Heart Institute risk score. Am Heart J 2008;155:1068 –1074. 11. Negassa A, Monrad ES, Srinivas VS. A simple prognostic classification model for postprocedural complications after percutaneous coronary intervention for acute myocardial infarction (from the New York State percutaneous coronary intervention database). Am J Cardiol 2009;103:937–942. 12. Simes RJ, Topol EJ, Holmes DR Jr, White HD, Rutsch WR, Vahanian A, Simoons ML, Morris D, Betriu A, Califf RM, Ross AM; for the
13.
14.
15.
16.
17. 18.
19. 20.
21. 22. 23.
24.
645
GUSTO-I Investigators. Link between the angiographic substudy and mortality outcomes in a large randomized trial of myocardial reperfusion. Importance of early and complete infarct artery reperfusion. Circulation 1995;91:1923–1928. French JK, Hyde TA, Patel H, Amos DJ, McLaughlin SC, Webber BJ, White HD. Survival 12 years after randomization to streptokinase: the influence of thrombolysis in myocardial infarction flow at three to four weeks. J Am Coll Cardiol 1999;34:62– 69. Hamada S, Nishiue T, Nakamura S, Sugiura T, Kamihata H, Miyoshi H, Imuro Y, Iwasaka T. TIMI frame count immediately after primary coronary angioplasty as a predictor of functional recovery in patients with TIMI 3 reperfused acute myocardial infarction. J Am Coll Cardiol 2001;38:666 – 671. Ndrepepa G, Berger PB, Mehilli J, Seyfarth M, Neumann FJ, Schomig A, Kastrati A. Periprocedural bleeding and 1-year outcome after percutaneous coronary interventions: appropriateness of including bleeding as a component of a quadruple end point. J Am Coll Cardiol 2008;51:690 – 697. Kinnaird TD, Stabile E, Mintz GS, Lee CW, Canos DA, Gevorkian N, Pinnow EE, Kent KM, Pichard AD, Satler LF, Weissman NJ, Lindsay J, Fuchs S. Incidence, predictors, and prognostic implications of bleeding and blood transfusion following percutaneous coronary interventions. Am J Cardiol 2003;92:930 –935. Gupta R, Gurm HS, Bhatt DL, Chew DP, Ellis SG. Renal failure after percutaneous coronary intervention is associated with high mortality. Catheter Cardiovasc Interv 2005;64:442– 448. Maluenda G, Lemesle G, Collins SD, Ben-Dor I, Syed AI, Torguson R, Kaneshige K, Xue Z, Pakala R, Suddath WO, Satler LF, Kent KM, Lindsay J, Pichard AD, Waksman R. The clinical significance of hematocrit values before and after percutaneous coronary intervention. Am Heart J 2009;158:1024 –1030. Thygesen K, Alpert JS, White HD. Universal definition of myocardial infarction. J Am Coll Cardiol 2007;50:2173–2195. Ryan TJ, Bauman WB, Kennedy JW, Kereiakes DJ, King III SB, McCallister BD, Smith Jr. SC, Ullyot DJ (Committee Members). Guidelines for percutaneous transluminal coronary angioplasty. A report of the American College of Cardiology/American Heart Association Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures (Committee on Percutaneous Transluminal Coronary Angioplasty). J Am Coll Cardiol 1993;22:2033–2054. Davison AC. Bootstrap Methods and Their Applications. Cambridge, UK: Cambridge University Press, 1997. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15:361–387. Kim P, Dixon S, Eisenbrey AB, O’Malley B, Boura J, O’Neill W. Impact of acute blood loss anemia and red blood cell transfusion on mortality after percutaneous coronary intervention. Clin Cardiol 2007; 30(suppl 2):II35–II43. Sattur S, Harjai KJ, Narula A, Devarakonda S, Orshaw P, Yaeger K. The influence of anemia after percutaneous coronary intervention on clinical outcomes. Clin Cardiol 2009;32:373–379.
Department of Internal Medicine, Division of Cardiology, Washington Hospital Center, Washington, DC. Manuscript received March 5, 2010; revised manuscript received and accepted April 7, 2010. *Corresponding author: Tel: 202-877-2812; fax: 202-877-2715. E-mail address: [email protected] (R. Waksman). 0002-9149/10/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.amjcard.2010.04.011
risk score with an independent cohort of patients undergoing PCI. Methods An ongoing registry of catheter-based coronary procedures is maintained at our institution. For data analysis and risk score construction, we included consecutive patients who underwent PCI and completed 1-year follow-up from January 1, 2000 to December 31, 2005. Patients in cardiogenic shock before PCI were excluded from the study, because it is already known this population has a very poor prognosis. Internal validation was performed using patients who underwent PCI from January 1, 2006 to December 31, 2007. This study was approved by the institutional review board at Washington Hospital Center and MedStar Research Institute (Washington, DC). In all cases, the interventional strategy was at the discretion of the responsible physician. Periprocedural anticoagulation was ensured using bivalirudin or unfractionated heparin to achieve an activated clotting time ⬎250 seconds in all patients. Glycoprotein IIb/IIIa inhibitor administration was also at the operator’s discretion. All patients received aspirin 325 mg and continued this regimen indefinitely. All patients received a clopidogrel loading dose of 300 to 600 mg, followed by a maintenance dose of 75 mg indefinitely. Independent research personnel blinded to the objectives of the study conducted clinical follow-up by telephone contact or office visits. In case of hospitalization, data were obtained by systematic review of source documents. All clinwww.ajconline.org
642
The American Journal of Cardiology (www.ajconline.org)
Table 1 Clinical, angiographic/procedural, and postprocedural characteristics of 6,932 patients undergoing percutaneous coronary intervention from 2000 to 2005 Clinical (n ⫽ 6,932) Age (years), mean ⫾ SD 65.1 ⫾ 11.8 Men 4,544 (65.6%) Diabetes mellitus 2,387 (34.8%) Previous myocardial infarction 2,398 (36.8%) Previous percutaneous coronary intervention 1,988 (29.9%) Previous coronary artery bypass grafting 1,668 (24.3%) Heart failure 1,017 (15.7%) Chronic renal failure 872 (12.7%) Stable angina pectoris 2,700 (39.2%) Unstable angina pectoris 3,433 (49.8%) Acute myocardial infarction 755 (11.0%) Hematocrit (%), mean ⫾ SD 39.9 ⫾ 11.9 Angiographic (n ⫽ 12,411 lesions) Left main coronary artery disease 215 (1.7%) Type C–B target lesion 10,847 (92.7%) Left ventricular ejection fraction (%), mean ⫾ SD 48 ⫾ 14 Stenting 10,522 (84.8%) Drug-eluting stent 7,372 (59.4%) Postprocedural (n ⫽ 6,932) Number of narrowed coronary arteries, mean ⫾ SD 1.98 ⫾ 0.88 Multilesion percutaneous coronary intervention 3,452 (47.5%) Thrombolysis In Myocardial Infarction grade ⬍3 flow 264 (3.8%) Hematocrit decrease (%), mean ⫾ SD 2.53 ⫾ 3.5 Creatinine increase ⬎1 mg/dl 162 (2.6%)
ical events were adjudicated by independent physicians who were not involved in the procedures. The outcome of interest was death at 1 year. Death was defined as all-cause mortality. MI was defined as the association of ⱖ1 clinical and ⱖ1 biological criteria: acute onset chest pain and/or typical changes on electrocardiogram (ST or T-wave changes or new left bundle branch block) and an increase of troponin ⬎99th percentile of the upper reference limit.19 Chronic renal insufficiency was defined as the presence of previously documented renal failure and/or a baseline serum creatinine ⬎2.0 mg/dl. Heart failure was defined as objective evidence of fluid retention due to cardiac causes before admission. Angiographic lesions were designated as A, B or C using the American College of Cardiology/ American Heart Association classification.20 Angiographic flow immediately after PCI was classified using Thrombolysis In Myocardial Infarction (TIMI) criteria.14 Hematocrit decrease was defined as the baseline value minus the nadir hematocrit level after PCI. Indications for PCI were classified as stable angina pectoris, unstable angina pectoris (defined as pain at rest or prolonged episodes of pain associated with objective evidence of ischemia), and acute MI. Continuous variables are presented as mean ⫾ SD, and discrete variables are summarized as frequencies and percentages. Logistic regression analysis was used to estimate odds ratios and associated p values for the primary end point. Clinically relevant baseline angiographic, procedural, and postprocedural variables were included in the initial model: age, gender, history of diabetes mellitus, chronic renal failure, heart failure, previous MI, hematocrit at baseline, presentation with acute MI, left main coronary artery disease (ⱖ50% stenosis), multilesion PCI (⬎1 lesion
Table 2 Association of baseline, angiographic/procedural, and postprocedural characteristics with one-year mortality
Age (years)* ⬍50 ⱖ50–⬍60 ⱖ60–⬍70 ⱖ70–⬍80 ⱖ80 Men Diabetes mellitus Previous myocardial infarction Heart failure Chronic renal failure Presentation with acute myocardial infarction Hematocrit at baseline (%)* ⱖ40 ⱖ35–⬍40 ⱖ30–⬍35 ⬍30 Left main coronary artery disease Type C target lesion Multilesion percutaneous coronary intervention Thrombolysis In Myocardial Infarction grade ⬍3 flow Hematocrit decrease (%)* ⬍5 ⱖ5–⬍10 ⱖ10–⬍15 ⱖ15 Creatinine increase ⬎1 mg/dl
Odds Ratio
Confidence Interval
p Value
— 0.76 1.46 3.21 4.35 0.74 2.44 1.85 5.28 4.87 1.44
— 0.4–1.4 0.9–2.5 1.9–5.3 2.6–7.4 0.59–0.92 1.95–3.05 1.47–1.87 4.20–6.63 3.85–6.17 1.05–1.98
— 0.3 0.2 ⬍0.0001 ⬍0.0001 0.008 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.023
— 1.70 4.08 5.27 2.89 1.36 1.20
— 1.3–2.3 3.0–5.5 3.4–8.1 1.88–4.45 1.07–1.75 0.9–1.6
— 0.0002 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.013 0.16
7.62–14.1
⬍0.0001
— 1.3–2.1 2.6–5.9 3.4–11.4 4.86–10.4
— 0.0002 ⬍0.0001 ⬍0.0001 ⬍0.0001
10.4
— 1.61 3.91 6.19 7.12
* Odds rates estimated point using the first category group as a reference for comparison.
treated), American College of Cardiology/American Heart Association type C lesion treated, presence of TIMI grade ⬍3 flow immediately after PCI, hematocrit decrease after PCI as a continuous variable, and creatinine increase after PCI ⱖ1.0 mg/dl from baseline value. The bootstrap method was then used to avoid overfitting the data.21 One thousand bootstrap samples were used. Backward selection with a p value ⬍0.05 for statistical significance was used to remove variables in each sample. Variables selected ⱖ800 times (80%) in the overall sample were included in the final model. To construct a simple risk score, variables identified through the multivariable model were assigned an integer coefficient. Integers were chosen to be approximately proportional to the estimated continuous coefficient from the logistic model. Each risk factor’s corresponding coefficient was added to obtain a final risk score from 0 to 26. For the final score, 5 strata were defined (very low, 0 to 4; low, 5 to 8; moderate, 9 to 10; high, 11 to 14; and very high, ⱖ15) based on predicted event rates of each score. Observed and expected numbers of events were calculated within each group. Model adequacy of the scoring system was then evaluated with the Hosmer-Lemeshow goodness-of-fit test.22 For the validation set of procedures the estimated probability of death at 1-year follow-up was calculated
Coronary Artery Disease/Long-Term Mortality Risk Assessment After PCI
643
Table 3 Adjusted predictors of death at one-year follow-up from index percutaneous coronary intervention Variable Thrombolysis In Myocardial Infarction grade ⬍3 flow14 Heart failure Left main coronary artery disease Chronic renal failure Diabetes mellitus Hematocrit decrease Hematocrit at baseline Age, number of decades after 40 years Intercept
Integer Score
Model Coefficient
Odds Estimated
95% Confidence Interval
p Value
7 4 3 3 2 1 1 1 N/A
2.035 1.155 0.783 0.730 0.527 0.382 0.374 0.296 ⫺5.276
7.65 3.17 2.19 2.07 1.69 1.46 1.45 1.34 N/A
5.3–10.9 2.4–4.1 1.3–3.5 1.6–2.7 1.3–2.2 1.2–1.7 1.3–1.7 1.2–1.5 N/A
⬍0.0001 ⬍0.0001 0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 N/A
Figure 1. Estimated rate of death at 1-year follow-up for the integer score system. Percentages at risk are shown for each of the 5 strata: ⱕ2%, very low risk; ⬎2% to 5%, low risk; ⬎5% to 10%, moderate risk; ⬎10% to 25%, high risk; ⬎25%, very high risk.
using the integer risk score. The discriminatory capacity of the model was assessed using an area under the receiver operating characteristic (ROC) curve, and goodness-of-fit was tested with the Hosmer-Lemeshow statistic (p ⬎0.05 considered to indicate lack of deviation between the model and the observed event rates). All statistical analysis was performed using SAS 9.1 (SAS Institute, Cary, North Carolina). Results A total of 6,932 patients who underwent PCI from January 1, 2000 to December 31, 2005 were included in the study. At 1-year follow-up there were 383 deaths (5.6%). Mean age was 65.1 ⫾ 11.8 years, 65.6% were men, and unstable angina pectoris was the most common indication for PCI. This set of patients represented a high-risk population, with a high prevalence of co-morbidities, as listed in Table 1. This population also had a high prevalence of high-risk angiographic findings; most patients had multivessel disease and complex coronary lesions requiring multilesion PCI. Stenting was used in most cases, with drug-eluting stents used in 60% of procedures. Univariable associations between covariables initially included in the bootstrapping sample and death at 1 year are presented in Table 2.
Five clinical, 1 angiographic, and 2 postprocedural variables were selected in ⱖ80% of bootstrapping samples. Variables selected from the original model included presence of TIMI grade ⬍3 flow after PCI, history of heart failure, left main coronary artery disease, chronic renal failure, diabetes mellitus, hematocrit decrease after PCI and hematocrit at baseline, and age (Table 3). The data showed a lack of deviation from the model, as indicated by the Hosmer-Lemeshow test result (p ⫽ 0.43). Mean area under the ROC curve of the bootstrap samples was 0.818, indicating a good discriminatory capacity between patients who developed death at 1-year follow-up and those who did not. Based on the integer risk score, 39.5% of procedures were classified as very low risk, 32.7% as low risk, 14.7% as moderate risk, 9.6% as high risk, and 3.6% as very high risk. Observed (and expected) rates of death in these strata were 1.0% (ⱕ2%) for very low risk, 3.3% (⬎2% to 5%) for low risk, 7.1% (⬎5% to 10%) for moderate risk, 15.9% (⬎10% to 25%) for high risk, and 43.0% (⬎25%) for very high risk (Figure 1). The validation set included 973 patients who underwent PCI from January 1, 2006 to December 31, 2007. In this population, there were 100 deaths (10.3%) at 1 year. The model predicted 81.5% of mortal events. In this validation
644
The American Journal of Cardiology (www.ajconline.org)
Figure 2. Observed and predicted rates of mortality in the validation set by integer score displayed as risk score in the validation set (x axis), mortality rate at 1-year follow-up (y axis), predicted mortality rate at 1-year derived from the risk score (solid line), and observed rate of 1-year mortality in the validation set (bars).
population the area under the ROC curve was 0.836, and the Hosmer-Lemeshow statistic p value was 0.573, demonstrating that the model fit the data well and correlated appropriately with patient risk (Figure 2). Of 738 presenting with stable/unstable angina pectoris, there were 85 deaths at 1 year, and the model predicted 79.3% of those events (area under the ROC curve 0.816, Hosmer-Lemeshow statistic, p ⫽ 0.933). In addition, in patients presenting with acute MI (n ⫽ 181), there were 24 deaths at 1 year, and the model predicted 88.9% of those (area under the ROC curve 0.903, and Hosmer-Lemeshow statistic, p ⫽ 0.625). Discussion We presented a simple integer risk scoring model that includes preprocedural, procedural, and postprocedural variables, which closely predicts long-term mortality in unselected patients undergoing contemporary PCI. This model, which was validated internally, highlights the significant impact of complications occurring after PCI on long-term outcomes. To our knowledge, this is the first risk model for long-term mortality to include postprocedural complications. In regard to baseline and procedural risk factors, our results agree with previously validated models.1–11 Similar to these risk models, we identified 5 baseline clinical characteristics (older age, presence of heart failure, chronic renal failure, diabetes mellitus, and lower baseline hematocrit) and 1 angiographic variable (presence of left main coronary artery disease) as the most significant predictors before PCI for long-term mortality. In addition, we identified 2 postprocedural variables (TIMI grade ⬍3 flow after PCI and lower hematocrit after PCI) as important predictors after PCI for long-term mortality. Extensive clinical evidence supports the predictive value of slow coronary flow12–14 and development of anemia or bleeding after PCI15,18,23,24 as known markers of poor prognosis. This study emphasizes the significant impact of postprocedural complications on
long-term risk of death and infers the underlying importance of measures to minimize bleeding and to achieve normal coronary flow after PCI. The presented risk score model has an excellent predictive capacity for 1-year mortality (c-statistics 0.84). Its performance remains accurate for patients presenting without acute MI (c-statistics 0.816) and even better for those presenting with acute MI (c-statistics 0.903). In general, previously reported risk score models for unselected patients undergoing PCI have predicted the risk of in-hospital complications after the procedure.2,7,9,10 No previous risk models, however, have included postprocedural complications and have not predicted the long-term risk of mortality after PCI. We believe that inclusion of postprocedural complications allows us to account for the significance of those events on long-term prognosis, and the addition of these components may improve the performance of risk models on long-term outcome prediction. Somewhat unexpectedly, presentation with acute MI was not associated with the primary outcome. Exclusion of patients presenting with cardiogenic shock may have minimized the role of acute MI in the model. In addition, inclusion of postprocedural variables, which are also complications frequently related to presentation with acute MI, could also account for this issue. In addition, procedurerelated renal failure was not part of the final model. The close relation of this complication and previous chronic renal failure may explain the lack of association between this variable and long-term mortality rate. The primary limitation of the present study is that our data are from a single center, and therefore the applicability to other populations remains unknown. Exclusion of patients presenting with cardiogenic shock and how this exclusion might result in omission of other variables in the current score are unclear. Furthermore, operator volume and expertise were not included in the analysis. External validation in a different dataset is therefore required.
Coronary Artery Disease/Long-Term Mortality Risk Assessment After PCI 1. Morrow DA, Antman EM, Charlesworth A, Cairns R, Murphy SA, de Lemos JA, Giugliano RP, McCabe CH, Braunwald E. TIMI risk score for ST-elevation myocardial infarction: A convenient, bedside, clinical score for risk assessment at presentation: An intravenous nPA for treatment of infarcting myocardium early II trial substudy. Circulation 2000;102:2031–2037. 2. Resnic FS, Ohno-Machado L, Selwyn A, Simon DI, Popma JJ. Simplified risk score models accurately predict the risk of major inhospital complications following percutaneous coronary intervention. Am J Cardiol 2001;88:5–9. 3. Addala S, Grines CL, Dixon SR, Stone GW, Boura JA, Ochoa AB, Pellizzon G, O’Neill WW, Kahn JK. Predicting mortality in patients with ST-elevation myocardial infarction treated with primary percutaneous coronary intervention (PAMI risk score). Am J Cardiol 2004; 93:629 – 632. 4. Granger CB, Goldberg RJ, Dabbous O, Pieper KS, Eagle KA, Cannon CP, Van De Werf F, Avezum A, Goodman SG, Flather MD, Fox KA. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med 2003;163:2345–2353. 5. Eagle KA, Lim MJ, Dabbous OH, Pieper KS, Goldberg RJ, Van de Werf F, Goodman SG, Granger CB, Steg PG, Gore JM, Budaj A, Avezum A, Flather MD, Fox KA. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA 2004;291: 2727–2733. 6. Halkin A, Singh M, Nikolsky E, Grines CL, Tcheng JE, Garcia E, Cox DA, Turco M, Stuckey TD, Na Y, Lansky AJ, Gersh BJ, O’Neill WW, Mehran R, Stone GW. Prediction of mortality after primary percutaneous coronary intervention for acute myocardial infarction: the CADILLAC risk score. J Am Coll Cardiol 2005;45:1397–1405. 7. Wu C, Hannan EL, Walford G, Ambrose JA, Holmes DR Jr, King SB III, Clark LT, Katz S, Sharma S, Jones RH. A risk score to predict in-hospital mortality for percutaneous coronary interventions. J Am Coll Cardiol 2006;47:654 – 660. 8. Singh M, Rihal CS, Selzer F, Kip KE, Detre K, Holmes DR. Validation of Mayo Clinic risk adjustment model for in-hospital complications after percutaneous coronary interventions, using the National Heart, Lung, and Blood Institute dynamic registry. J Am Coll Cardiol 2003;42:1722–1728. 9. Singh M, Lennon RJ, Holmes DR Jr, Bell MR, Rihal CS. Correlates of procedural complications and a simple integer risk score for percutaneous coronary intervention. J Am Coll Cardiol 2002;40:387–393. 10. Madan P, Elayda MA, Lee VV, Wilson JM. Predicting major adverse cardiac events after percutaneous coronary intervention: the Texas Heart Institute risk score. Am Heart J 2008;155:1068 –1074. 11. Negassa A, Monrad ES, Srinivas VS. A simple prognostic classification model for postprocedural complications after percutaneous coronary intervention for acute myocardial infarction (from the New York State percutaneous coronary intervention database). Am J Cardiol 2009;103:937–942. 12. Simes RJ, Topol EJ, Holmes DR Jr, White HD, Rutsch WR, Vahanian A, Simoons ML, Morris D, Betriu A, Califf RM, Ross AM; for the
13.
14.
15.
16.
17. 18.
19. 20.
21. 22. 23.
24.
645
GUSTO-I Investigators. Link between the angiographic substudy and mortality outcomes in a large randomized trial of myocardial reperfusion. Importance of early and complete infarct artery reperfusion. Circulation 1995;91:1923–1928. French JK, Hyde TA, Patel H, Amos DJ, McLaughlin SC, Webber BJ, White HD. Survival 12 years after randomization to streptokinase: the influence of thrombolysis in myocardial infarction flow at three to four weeks. J Am Coll Cardiol 1999;34:62– 69. Hamada S, Nishiue T, Nakamura S, Sugiura T, Kamihata H, Miyoshi H, Imuro Y, Iwasaka T. TIMI frame count immediately after primary coronary angioplasty as a predictor of functional recovery in patients with TIMI 3 reperfused acute myocardial infarction. J Am Coll Cardiol 2001;38:666 – 671. Ndrepepa G, Berger PB, Mehilli J, Seyfarth M, Neumann FJ, Schomig A, Kastrati A. Periprocedural bleeding and 1-year outcome after percutaneous coronary interventions: appropriateness of including bleeding as a component of a quadruple end point. J Am Coll Cardiol 2008;51:690 – 697. Kinnaird TD, Stabile E, Mintz GS, Lee CW, Canos DA, Gevorkian N, Pinnow EE, Kent KM, Pichard AD, Satler LF, Weissman NJ, Lindsay J, Fuchs S. Incidence, predictors, and prognostic implications of bleeding and blood transfusion following percutaneous coronary interventions. Am J Cardiol 2003;92:930 –935. Gupta R, Gurm HS, Bhatt DL, Chew DP, Ellis SG. Renal failure after percutaneous coronary intervention is associated with high mortality. Catheter Cardiovasc Interv 2005;64:442– 448. Maluenda G, Lemesle G, Collins SD, Ben-Dor I, Syed AI, Torguson R, Kaneshige K, Xue Z, Pakala R, Suddath WO, Satler LF, Kent KM, Lindsay J, Pichard AD, Waksman R. The clinical significance of hematocrit values before and after percutaneous coronary intervention. Am Heart J 2009;158:1024 –1030. Thygesen K, Alpert JS, White HD. Universal definition of myocardial infarction. J Am Coll Cardiol 2007;50:2173–2195. Ryan TJ, Bauman WB, Kennedy JW, Kereiakes DJ, King III SB, McCallister BD, Smith Jr. SC, Ullyot DJ (Committee Members). Guidelines for percutaneous transluminal coronary angioplasty. A report of the American College of Cardiology/American Heart Association Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures (Committee on Percutaneous Transluminal Coronary Angioplasty). J Am Coll Cardiol 1993;22:2033–2054. Davison AC. Bootstrap Methods and Their Applications. Cambridge, UK: Cambridge University Press, 1997. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15:361–387. Kim P, Dixon S, Eisenbrey AB, O’Malley B, Boura J, O’Neill W. Impact of acute blood loss anemia and red blood cell transfusion on mortality after percutaneous coronary intervention. Clin Cardiol 2007; 30(suppl 2):II35–II43. Sattur S, Harjai KJ, Narula A, Devarakonda S, Orshaw P, Yaeger K. The influence of anemia after percutaneous coronary intervention on clinical outcomes. Clin Cardiol 2009;32:373–379.