Relation Between Lesion Characteristics and Risk With Percutaneous Intervention in the Stent and Glycoprotein IIb/IIIa Era
An Analysis of Results From 10 907 Lesions and Proposal for New Classification Scheme
Background—The currently used American College of Cardiology/American Heart Association lesion classification scheme dates from an era when balloon angioplasty was the only percutaneous treatment available and major complications occurred in ≈7% of patients. Major advances in treatment options would suggest that this scheme may be outmoded, but the schemes that have been suggested to update lesion classification have not been widely accepted.
Methods and Results—Four thousand one hundred eighty-one consecutive patients (6676 lesions) formed a training set and 2146 patients (4231 lesions) formed a validation set treated from 1995 to 1997 at a single center used by 3 hospital groups. Twenty-seven pretreatment candidate variables were analyzed with the use of stepwise proportional logistic regression, and 9 (nonchronic total occlusion with TIMI flow 0, degenerated vein graft, vein graft age >10 years, lesion length ≥10 mm, severe calcium, lesion irregularity, large filling defect, angulated ≥45 degrees plus calcium, and eccentricity) were independently correlated (P<0.05) with ranked adverse outcome (death, Q-wave or creatine kinase ≥3× normal myocardial infarction, or emergency coronary artery bypass grafting≫creatine kinase 2 to 3× myocardial infarction≫possibly related to non–Q-wave myocardial infarction≫no complication). A scheme based on these findings and the old American College of Cardiology/American Heart Association scheme were found to have c-statistics in the validation set of 0.672 and 0.620 (P=0.010 vs old scheme), respectively.
Conclusions—Appreciation of these contemporary risk factors for complications of coronary intervention may assist in patient selection and in risk adjustment for comparison of outcomes between providers.
The American College of Cardiology/American Heart Association (ACC/AHA) lesion classification scheme was proposed in 19861 and modified in 1990.2 It is still widely used to assess risk of patients and lesions undergoing percutaneous intervention and serves as a risk adjustment parameter in “scorecarding” individual operators and hospitals. Dramatic changes have occurred in the approach to percutaneous intervention since that time, allowing treatment of more complex lesions3 4 5 and lower overall risk.6 7 8 It would be reasonable to suspect that the ACC/AHA scheme would need to be revised, but several schemes that have been suggested have not been widely adapted.9 10 11
Our purpose was to develop a new predictive model for adverse outcomes with the use of contemporary data and to compare its performance in a validation sample against that of the current modified ACC/AHA scheme.2
Data from 6327 consecutively treated patients undergoing coronary intervention from January 1, 1995, until December 31, 1997, treated by physician operators from 1 of 3 cardiology groups (full-time Cleveland Clinic staff, Ohio Permanente Medical Group staff, part-time associate staff) were analyzed. Baseline clinical data were entered prospectively into an established database, and in-hospital clinical outcomes were evaluated by a cadre of trained personnel independent from the physician operators. ECGs were routinely obtained before and immediately after the procedure, the morning after the procedure, and in the event of suspected ischemia. Creatine kinase (CK) levels were routinely assayed 6 to 8 hours after, the morning after the procedure, and in the event of a suspected myocardial infarction (MI).
Variables and Definitions
Baseline Clinical Variables
Acute MI (<24 hours), age, cardiogenic shock, current smoking, diabetes mellitus (insulin-dependent and non–insulin-dependent), gender, hypercholesteremia (total cholesterol ≥240 mg%, LDL cholesterol ≥130 mg%, or HDL cholesterol <35 mg%), hypertension, New York Heart Association congestive heart failure class, prior remote MI (>2 weeks), recent MI (1 to 14 days), saphenous vein graft age (years), and surgically inoperable, unstable angina. Except for vein graft age (analogous to “chronic” total occlusion in the ACC/AHA scheme in that it relates a time duration with a morphological or locating variable), these variables were used for descriptive purposes only.
See Table 1⇓ for variables assessed and their definitions. Angiographic variables were analyzed by 1 of 2 independent observers blinded to clinical outcome in the training set analysis and by 1 of 12 clinical interventionists overread by the Cleveland Clinic Core Angiographic Laboratory in the validation set analysis.
Treatment variables were abciximab and primary treatment device [balloon, directional atherectomy, laser, Rotablator, stent, and transluminal extraction catheter (TEC)].
Outcome variables were cardiac death, emergency bypass surgery (with acute ischemia or for the prevention of acute ischemia within 24 hours of the procedure), and MI [Q-wave and non–Q-wave/subcategorized by CK level (ECG changes or enzyme elevations consequent to an acute MI at the patient’s presentation are excluded)].
Continuous data are presented as mean±1 SD if normally distributed and as median and interquartile range if skewed. Categorical data are presented as a percentage.
Cases for the training set analysis were composed of all patients treated during 1995 to 1996 with any ischemic complication and a randomly selected cohort of patients without complications (3:1 patients with no complications:patients with complications). This cohort was termed the enriched training set. This approach allowed us to study characteristics not routinely evaluated without having to review thousands of cineangiograms. However, if characteristics were relatively infrequent in this sample (n <50) and recorded routinely in our database, the entire 1995 to 1996 data set was used to ascertain their relation to clinical outcome. Secondary analyses of the population without the most predictive variables was performed to improve identification of intermediate-risk lesion morphologies. Cases for the validation set analysis were composed of all patients treated during 1997.
Adverse outcomes were characterized as most severe: cardiac death, Q-wave MI, non–Q-wave MI with CK ≥5× upper limit of normal and positive cardiac isoenzymes, or emergency bypass surgery; severe: non–Q-wave MI with CK elevation 3 to 5× upper limit of normal in the presence of abnormal cardiac isoenzymes; moderate: non–Q-wave MI with CK elevation 2 to 3× upper limit of normal with positive cardiac isoenzymes; uncertain: the presence of non–Q-wave MI, possibly associated with the lesion evaluated (this typically occurred when multiple lesions were treated and there was no obvious angiographic complication); and no complication associated with lesion treatment. Because of the small number of lesions associated with the severe category in the test population, the most severe and severe categories were subsequently collapsed for the purposes of all analyses subsequent to our preliminary investigation.
Stepwise proportional odds logistic regression analysis12 was performed to identify independent correlates of adverse ischemic outcomes in the training sample. This, in contradistinction to standard multiple logistic regression, allows for simultaneous fitting of a model to a ranked ordinal outcome.
On the basis of their relation to graded adverse outcome, variables were classified as strongly or moderately correlated. With the use of the training sample, a scheme to optimize the true-positive, false-negative relation was developed. The presence of any strong correlate qualified the lesion as highest risk (class IV). The presence of ≥3 moderate characteristics qualified the lesion as high risk (class III), the presence of 1 to 2 moderate characteristics qualified the lesion as moderate risk (class II), and absence of any risk factor led to the assessment of low risk (class I).
The predictive capability of the new scheme and the current ACC/AHA scheme were then tested in the 1997 validation population. Comparison of their predictive capacities were performed with the use of the methods of Hanley and McNeil.13 Subgroup analyses were also performed in the 1997 stent-plus-abciximab–treated population.
All statistical analyses were performed with SYSTAT (SYSTAT version 7.0, SPSS Inc) or SAS/STAT (SAS Institute Inc, version 6.1).
Baseline patient demographics for the training, validation, and total population studied are provided in Table 2⇓. Six thousand three hundred twenty-seven patients and 10 907 lesions in total were analyzed. Patients were typically male and middle-aged. Sixty-two percent had unstable angina. Twenty-nine percent had prior bypass surgery. Devices and drug utilization frequency for the 1995 to 1996 population were as follows: stents, 40.7% (planned in 36.9% and bailout in 3.8%); abciximab, 26.2% (planned in 24.0% and bailout in 2.2%); Rotablator, 18.9%; directional coronary atherectomy, 0.9%; excimer laser coronary angioplasty, 0.2%; and TEC, 0.2%. The figures for the 1997 population were as follows: stent, 64.2% (62.6% planned and 1.6% bailout); abciximab, 41.1% (38.4% planned, 2.7% bailout); Rotablator, 17.7%; directional coronary atherectomy, 0.6%; excimer laser coronary angioplasty, 0.5%; and TEC, 0.1%. No significant differences were observed between the training and validation samples except for the use of stents and abciximab.
For the training and validation cohorts together, technical success achieved at ≥1 target sites was achieved in 95.8% of patients, and major complications (death, Q-wave MI, or emergency bypass surgery) occurred in 1.9% of patients (Table 3⇓). There was no significant difference between the training and validation sets for any major outcome.
Training Sample Evaluation
In the enriched training sample, there were 85 lesions associated with “most severe” complications, 11 lesions associated with “severe” complications, 93 lesions associated with “moderately severe” complications, and 57 lesions associated with “possible” complications (a total of 246 adverse event–related lesions). This number increased to 257 in the full evaluation consequent to our inability to retrieve cineangiograms on 7 patients (11 lesions) associated with complications during the analysis of the enriched training sample.
As can be seen in Table 4⇓, 10 variables were identified as independent correlates of graded adverse outcome. Detailed clinical outcomes for the 2 most important correlates in the entire 1995 to 1997 cohort were for nonchronic total occlusion: death, 4.5%; Q-wave infarction, 0.1%; emergency bypass surgery, 2.8%; non–Q-wave infarction, 5.2% (CK >5× 0.6%, CK 3 to 5× 3.9%, CK 2 to 3× 0.7%); for degenerated saphenous vein grafts: death, 1.1%; Q-wave infarction, 2.8%; emergency bypass surgery, 0.4%; non–Q-wave infarction, 12.7% (CK >5× 1.5%, CK 3 to 5× 8.8%, CK 2 to 3× 2.4%). The proposed schema derived from this analysis is presented in Figure 1⇓ (the 2 lesion length variables were combined). The c-statistic evaluating the predictive value of the new model for complications in the training set was 0.701.
Validation Sample Evaluation
Performance of the new model in the 1997 validation sample is shown in Table 5⇓ and Figure 2⇓. The calibration coefficient (r) for the relation between risk in the training and validation samples for each risk group (Figure 2⇓) is 0.971. The c-statistic for the new model was 0.672 (the c-statistic for the ACC/AHA scheme was 0.620, P=0.010). The new model had its predictive accuracy maintained in the prespecified stent+abciximab subset (c-statistic=0.663), whereas by the ACC/AHA scheme, the c-statistic was only 0.589.
The ACC/AHA lesion risk classification schema has been criticized because of its subjectivity and because it emanates from an era when stents and glycoprotein IIb/IIIa inhibitors in particular were not available. Nonetheless, in the hands of core angiographic laboratories and in those practitioners who are both objective and well versed in its meaning, this system has been demonstrated repeatedly to have predictive power in the assessment of risk for populations of patients.14 15 16
Our comprehensive reevaluation of the relation between lesion morphology and short-term outcome with percutaneous intervention was undertaken with several goals and considerations. First, we wanted to develop and validate a better predictive instrument suitable for the current era of intervention. We expected that changes in treatment and outcome might attenuate the correlation of some previously important lesion characteristics with outcome. We intended to make use of inferences from observations with the use of intracoronary ultrasound and angioscopy17 18 19 to postulate new or modified angiographic characteristics to evaluate. We recognized, however, that prediction of risk generally becomes more difficult as overall risk diminishes (as it has in the decade since the modified ACC/AHA scheme was developed). Second, we hoped to more objectively characterize certain angiographic risk factors to lessen the subjectivity of the risk assessment. Third, we wanted for the first time to have a system that reflected the gradation of clinically meaningful adverse events.20 Fourth, we hoped that the new instrument would be relatively simple and easy to implement.
The proposed classification scheme is, in fact, more predictive than the prior modified ACC/AHA lesion scheme, although perhaps not by as much as one might have suspected beforehand. Importantly, it was validated in a patient population reflecting current utilization practice (stenting in 64%, glycoprotein IIb/IIIa inhibitors in 41%), and maintained its predictive accuracy in the patient populations treated with stents and ReoPro (c-statistic=0.66). In addition, it is somewhat simpler than the previous system analyzing 9 specific elements instead of 16. Further, we were able to more specifically define the nature of certain characteristics evaluated. Previous definitions that were particularly troublesome due to their subjectivity included that for degenerated vein grafts, calcification, and proximal tortuosity.
Also important, particularly as it pertains to the assessment of individual patient risk, is the absence of importance of certain characteristics that were previously thought to be associated with heightened risk. These include lesion angulation per se, bifurcation location, ostial location, proximal tortuosity, and small thrombus. It is likely that improvements in device profile and trackability, as well as the capacity for stenting to treat dissections that were particularly prone to develop with balloon angioplasty in certain circumstances, account for these differences.
In addition, the identification of specific lesion characteristics that continue to be associated with complications, even with current therapies, provides objective data from which to challenge industry to improve percutaneous treatment devices and adjunct pharmacology. The importance of aged and degenerated vein graft lesion morphology in particular should heighten interest in the development of treatments that may minimize the risk of complications associated with treatment of such lesions, for example, emboli entrapment devices and covered stents. The association of nonchronic total occlusion with adverse outcome is partly due to its correlation with the setting of acute MI and the subsequent risk of cardiac death; hence it may be less modifiable. Nonetheless, if patients with acute MI were excluded from analysis, this variable maintains considerable independent prognostic power (OR=3.60, P=0.003 [compare with Table 4⇑]).
There are 2 limitations in particular that should be noted in relation to this analysis. First, although the model was prospectively validated in a population representative of currently treated patients, one must be somewhat cautious about overgeneralizing the results. The results were derived from outcomes of 18 generally experienced interventionists operating in an environment when senior operator assistance was readily available. Annual interventional patient volume of these operators ranged from 42 to 323 over this 3-year period. In addition, the assessment of lesion morphology was performed by angiographers who were trained and highly experienced. Second, although from a statistical standpoint this model provides considerable predictive power, it is far from perfect. Certainly, clinical factors such as patient age, presentation with acute MI, or poor left ventricular function, as well as operator characteristics,21 22 23 even in the era of stent utilization, would be expected to affect outcome and therefore diminish the sole predictive capability of lesion morphology. In the end, any measure of the relation between lesion characteristics and outcome in this setting, although somewhat helpful in the choice of potential therapies for a given patient, will always have greater power in assessing the risk of large populations.
In conclusion, utilization of this contemporary risk assessment algorithm by interventionists working under similar circumstances should improve their assessment of patient risk, and, in particular, extend availability of percutaneous intervention to some patients who otherwise would have been turned down for treatment because of perceived high risk.
The authors would like to acknowledge the expert assistance of Patti Durnwald in the preparation of the manuscript.
- Received April 21, 1999.
- Revision received June 15, 1999.
- Accepted July 13, 1999.
- Copyright © 1999 by American Heart Association
Ryan TJ, Faxon DP, Gunnar RM, Kennedy JW, King SB, Loop FD, Peterson KL, Reeves TJ, Williams DO, Winters WL, Fisch C, DeSanctis RW, Dodge HT, Reeves TJ Weinberg SL. 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 (subcommittee on percutaneous transluminal coronary angioplasty). J Am Coll Cardiol. 1988;12:529–545.
Ellis SG, Vandormael MG, Cowley MJ, DiSciascio G, Deligonul U, Topol EJ, Bulle TM. Coronary morphologic and clinical determinants of procedural outcome with angioplasty for multivessel coronary disease: implications for patient selection. Circulation. 1990;82:1193–1202.
Baim DS, Kent KM, King SB, Safian RD, Cowley MJ, Holmes DR, Roubin GS, Gallup D, Steenkiste AR, Detre K, for the NACI Investigators. Evaluating new devices: acute (in-hospital) results from the New Approaches to Coronary Intervention Registry. Circulation. 1994;89:471–481.
Reifart N, Vandormael M, Krajcar M, Göhring S, Preusler W, Schwarz F, Störger H, Hofmann M, Klöpper J, Müller S Haase J. Randomized comparison of angioplasty of complex coronary lesions at a single center: excimer laser, rotational atherectomy, and balloon angioplasty comparison (ERBAC) Study. Circulation. 1997;96:91–98.
Park S-J, Park S-W, Hong M-K, Cheong S-S, Lee CW, Kim J-J, Hong MK, Mintz GS Leon MB. Stenting of unprotected left main coronary artery stenoses: immediate and late outcomes. J Am Coll Cardiol. 1997;31:37–42.
Haude M, Erbel R, Hoepp HW, Heublein B, Sigmund M, Meyer J, and the STENT-BY Study Group. STENT-BY Study: a prospective randomized trial comparing immediate stenting versus conservative treatment strategies in abrupt vessel closure or symptomatic dissections during coronary balloon angioplasty. Eur Heart J. 1996;17:172. Abstract.
Kimmel SE, Berlin JA, Strom BL, Laskey WK, for the Registry Committee of the Society for Cardiac Angiography and Interventions. Development and validation of a simplified predictive index for major complications in contemporary percutaneous transluminal coronary angioplasty practice. J Am Coll Cardiol. 1995;26:931–938.
Special topics. Polytomous logistic regression: introduction to the model and estimation of the parameters. In: Hosmer DW, Lemeshow S, eds. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989:216–245.
Hannan EL, Arani DT, Johnson LW, Kemp HG, Lukacik G. Percutaneous transluminal coronary angioplasty in New York State: risk factors and outcomes. JAMA. 1993;268:3092–3097.
Ellis SG, Omoigui N, Bittl JA, Lincoff M, Wolfe MW, Howell G, Topol EJ. Analysis and comparison of operator-specific outcomes in interventional cardiology: from a multicenter database of 4860 quality-controlled procedures. Circulation. 1996;93:431–439.
Fitzgerald PJ, Ports TA, Yock PG. Contribution of localized calcium deposits to dissection after angioplasty: an observational study using intravascular ultrasound. Circulation. 1992;86:64–70.
Mintz GS, Popma JJ, Pichard AD, Kent KM, Satler LF, Chuang YC, Ditrano CJ, Leon MB. Patterns of calcification in coronary artery disease: a statistical analysis of intravascular ultrasound and coronary angiography in 1155 lesions. Circulation. 1995;91:1959–1965.
White CJ, Ramee SR, Collins TJ, Escobar AE, Karsan A, Shaw D, Jain SP, Bass TA, Heuser RR, Teirstein PS, Bonan R, Walter PD, Smalling RW. Coronary thrombi increase PTCA risk: angioscopy as a clinical tool. Circulation. 1996;93:253–258.
Califf RM, Abdelmeguid AE, Kuntz RE, Popma JJ, Davidson CJ, Cohen EA, Kleiman NS, Mahaffey KW, Topol EJ, Pepine CJ, Lipicky RJ, Granger CB, Harrington RA, Tardiff BE, Crenshaw BS, Bauman RP, Zuckerman BD, Chaitman BR, Bittl JA, Ohman EM. Myonecrosis after revascularization procedures. J Am Coll Cardiol. 1998;31:241–251.
Hannan EL, Racz M, Ryan TJ, McCallister BD, Johnson LW, Arani DT, Guerci AD, Sosa J, Topol EJ. Coronary angioplasty volume-outcome relationships for hospitals and cardiologists. JAMA. 1997;279:892–898.
Ellis SG, Weintraub W, Holmes D, Shaw R, Block PC, King SB. Relation of operator volume and experience to procedural outcome of percutaneous coronary revascularization at hospitals with high interventional volumes. Circulation. 1997;95:2479–2484.