Author + information
- Received June 5, 1998
- Revision received October 19, 1998
- Accepted January 5, 1999
- Published online April 1, 1999.
- Francois Ledru, MDa,* (, )
- Pierre Théroux, MD†,
- Jacques Lespérance, MD‡,
- Jean Laurier, MSc‡,
- Pierre Ducimetière, PhD∗,
- Jean-Léon Guermonprez, MDa,
- Benoit Diébold, MD, PhDa and
- Didier Blanchard, MDa
- ↵*Reprint requests and correspondence: Dr. Francois Ledru, Département de Cardiologie, Hopital Broussais, 96 rue Didot, 75014 Paris, France
We sought to identify the angiographic predictors of a future infarction, to study their interaction with time to infarction, patient risk factors and medications, and to evaluate their clinical utility for risk stratification.
Identification of coronary lesions at risk of acute occlusion remains challenging. Stenosis severity is poorly predictive but other stenosis descriptors might be better predictors.
Eighty-four patients with an acute myocardial infarction and a coronary angiogram performed within the preceding 36 months (baseline angiogram), and after infarction were selected. All coronary stenoses (from 10% to 95% lumen diameter reduction) at baseline angiogram were analyzed by computer-assisted quantification. Each of the 84 lesions responsible for the infarction (culprit) was compared with the nonculprit stenoses (controls) in the same patient.
Culprit lesions were more symmetrical (symmetry index +15%; p < 0.001), had steeper outflow angles (maximal angle +4°; p < 0.001), were more severe (percent stenosis +5%; p = 0.001) and longer (+1.5 mm, p = 0.01) than controls. The symmetry index and the outflow angles were the two independent predictors of infarction at three-year follow-up. Stenosis severity predicted only infarctions occurring within 1 year after angiography. In moderately severe stenoses (40% to 70% stenosis), stratification using the symmetry index and outflow angles accurately predicted lesions remaining free of occlusion and infarction at three-year follow-up.
Better characterization of stenosis geometry might help to understand the pathophysiologic mechanisms triggering coronary occlusion and to stratify patients for improved care.
Histopathologic and angiographic studies have clearly documented that acute myocardial infarction (AMI) results from the formation of an intraluminal occlusive thrombus on a disrupted atherosclerotic plaque (1). Ruptured lesions, compared with stable plaques, contain more soft lipids and have a thinner cap (1). The currently available investigation
methods, however, do not allow characterization of the serial changes that mark the transition from a stable to an unstable plaque in humans and from a disrupted plaque to an occlusive thrombus. Patient management is therefore mainly dictated by evaluation of the extent and severity of coronary artery disease (CAD) (2,3). The onset of AMI, however, is not predictable and occlusion often occurs on a nonsevere stenosis (4–7). Several morphologic features of stenoses, such as irregular contours or ulcerations, stenosis length, inflow and outflow angles and the presence of a division branch within the stenosis (6,7)have been associated with risk of occlusion. The present study attempted to identify the most powerful angiographic predictors of a future AMI in patients with known coronary anatomy, to study their interaction with time to infarction, patient risk factors and medications and to evaluate their clinical utility for risk stratification.
Patients and methods
Patients were identified from the database of the Montreal Heart Institute, 1980–1993, using as selection criteria: 1) an AMI defined by ST segment elevation at admission and confirmed by cardiac enzymes elevation and/or new Q-waves on the resting 12-lead electrocardiogram (ECG); 2) unequivocal identification of the culprit coronary lesion on an angiogram obtained within the 4 weeks after the AMI; 3) an angiogram performed within the previous 36 months; and 4) no coronary event or intervention between the first angiogram and the qualifying infarction.
One hundred four patients met these criteria. Twenty patients were excluded by predefined criteria of reinfarction at the site of a previous AMI or a new occlusion proximal to a previous occlusion (17 patients), diffuse multifocal atherosclerosis precluding valid measurement of the reference diameter (1 patient) and technically sub-optimal angiograms (2 patients). The demographic, historical, clinical and biological data at the time of the first angiogram were collected. Diabetes mellitus was defined as a fasting glucose above 6.8 mmol/liter or the prescription of hypoglycaemic agents. Hypercholesterolemia was defined as total cholesterol above 5.2 mmol/liter combined with LDL cholesterol above 3.3 mmol/L or the prescription of a cholesterol-lowering agent. Systemic hypertension was defined as the intake of antihypertensive drugs. The medication used at the time of the AMI was also recorded. The indication of angiography after AMI was recurrent angina in 41 patients (49%), part of a research protocol on thrombolysis in 17 (20%), physician’s choice in 16 patients (19%) and a positive treadmill test in 10 (12%).
The coronary angiograms recorded on 35-mm cine films were performed as usual, following a standard protocol including administration of sub-lingual or intravenous nitroglycerin. For the purpose of this study, all angiograms were re-analyzed by a cardiovascular radiologist and an experienced cardiologist. The infarct-related coronary lesion (IRL) was identified on the post-AMI angiogram by the presence of a thrombolysis in myocardial infarction (TIMI) 0, 1 or 2 flow grade, or of an intra-luminal defect strongly suggestive of a thrombus and concordant with the location of the infarct by the 12-lead ECG and by left ventricular angiography. Eight patients had a TIMI 3 flow grade in the infarct-related coronary artery, but a single obstructive lesion that had clearly caused the AMI.
All visible lesions, including wall irregularities, were analyzed on the pre-AMI angiogram blindly to the IRL location. Multiple lesions within one coronary artery segment were considered distinct whenever separated by a visually smooth arterial wall. Four hundred thirty-five lesions were identified in the 84 patients. Sixty of them were excluded: 36 complete occlusions, 10 located proximal to a complete occlusion, 8 unsuitable for quantitative coronary analysis (QCA) and 6 lesions part of diffuse aneurysmal dilatation of the artery. All remaining 375 lesions were submitted to quantitative coronary analysis (QCA). Once all qualitative and quantitative data had been entered into the database, the two angiograms of each patient were projected side-by-side to correctly identify the IRL. Thus, the 375 lesions present at baseline were divided into two groups: the 84 lesions corresponding to the IRL location were defined as the culprit lesions, and the remaining 291 nonculprit lesions served as controls.
The qualitative analysis included ostial, proximal, mid or distal site of the lesion, presence of a bifurcation within or in the immediate vicinity of a stenosis and presence or absence of calcifications. The first three segments of the right coronary artery were considered as proximal, since this artery does not usually provide any major division branch.
QCA was performed on all stenoses greater than 10% lumen diameter reduction (by visual assessment), excluding vessels less than 1.0 mm diameter and segments poorly perfused with TIMI 1 or 2 grade flow. QCA was applied to the entire length of each analyzed segment. Measurements were made using the Coronary Measurement System (CMS; Medis, Leiden, The Netherlands) (8)in the projection showing the most severe diameter reduction, at the time of best filling and during diastole whenever possible, after 2.5 times image magnification. The minimal lumen diameter (MLD), interpolated reference diameter, percent lumen diameter stenosis (PS) and lesion length were measured from the diameter function of the CMS automated edge detection algorithm, as previously described (8). In ostial lesions, the user-defined reference diameter is more reliable and was used. The digitized images were calibrated by reference to the catheter size.
The symmetry index was calculated as the percent ratio of the plaque areas of the two borders of the stenosis, the plaque area being the area between the estimated vessel wall and the luminal stenosis contours within the obstruction boundaries (Fig. 1). A ratio of 0 indicated extreme eccentricity and 100% perfect symmetry (8). The inlet (convergent) and the outlet (divergent) angles of the diameter function were computed by three different methods using the Math Lab software (the Math Works Inc., Natick, Massachusetts) to obtain inflow and outflow angles, respectively, as detailed in Figure 1. The inflow angles were not assessed in ostial lesions.
The variability of the quantitative measurements obtained by the CMS was assessed in a random sample of 25 coronary lesions as the intra-class correlation between two independent readings of the same cine frame by the same observer (intra-observer variability) and by two observers (inter-observer variability). The intra- and inter-observer intra-class correlation were 0.96 and 0.94, respectively, for the MLD, 0.92 and 0.92 for the PS, 0.92 and 0.90 for the stenosis length, 0.84 and 0.66 for the symmetry index, 0.95 and 0.93 for the average inflow angle, 0.91 and 0.94 for the average outflow angle, 0.79 and 0.87 for the maximal inflow angle and 0.75 and 0.79 for the maximal outflow angle. All correlations were highly significant.
The statistical analyses were performed using the S.A.S. software (Cary, North Carolina). Continuous variables were expressed as means ± SD, and discrete data as frequency distribution. p values <0.05 were considered statistically significant. For each quantitative stenosis descriptor, we defined at first three sub-groups using terciles and calculated the prevalence of culprit lesions in each of these sub-groups. The likelihood of a linear trend across the three sub-groups was tested by the Chi-square method for trends. Univariate and multivariate comparisons of continuous variables were performed by conditional logistic regression (PHREG procedure), allowing matched comparisons of culprit lesions with a variable number of control lesions. In these analyses, each patient was his own control. In the multivariate conditional logistic regressions, the dependent variable was the stenosis outcome (culprit or control) and the covariates were percent lumen diameter stenosis (PS), length, symmetry index, inflow and outflow angles.
To explore whether the time to the infarction could modify the predictive value of the geometric characteristics of stenoses, linear regression analyses were performed plotting the time interval from initial angiography to occurrence of AMI (time to infarction) against each characteristic of the culprit and control lesions, respectively. This covariance analysis was done using general estimating equations (GENMOD procedure) in order to take into account the excess of control lesions per patient (eg, 291 control lesions in 84 patients). Eventually, interactions between patient characteristic and angiographic parameters were tested by the likelihood ratio method.
The time interval (mean ± SD) between the first coronary angiography and the qualifying AMI was 15 ± 10 months (range 8 days to 36 months), and between the AMI and the post-AMI coronary angiography was 8 ± 6 days (0 to 24 days). The clinical features of the population at the time of the initial angiogram are displayed in Table 1. Only 31% of patients were taking aspirin at the time of the AMI, and 5% a lipid-lowering drug despite the high prevalence of hypercholesterolemia. In most patients, the ST segment elevation evolved to a Q wave AMI.
Coronary lesions were located in the right coronary artery, the left anterior descending artery and the circumflex artery in 36%, 35% and 29%, respectively, were ostial in 8% and involved a bifurcation in 29%, with no statistically significant difference between culprit and control lesions. However, culprit lesions tended to be more proximal than control lesions (56% vs 45%; p = 0.04) and were located in larger coronary arteries (reference diameter 2.9 ± 0.6 mm vs 2.7 ± 0.7 mm; p = 0.04).
Prognostic impact of stenosis geometry
Using terciles (Fig. 2), we observed an increasing frequency of culprit lesions with increasing symmetry index, stenosis length, outflow angles and PS. Frequencies of culprit lesions were unrelated to the magnitude of the MLD or the inflow angles.
Culprit and control lesions differed with regards to their angiographic characteristics, as shown in Table 2. The most striking difference between the culprit and control lesions was a higher symmetry index in culprit lesions. The other statistically significant differences sorted in decreasing order of importance were steeper outflow angles (either geometric, average or maximal), more severe obstructions and longer stenoses.
Since weak but significant correlations between PS, length, symmetry index and inflow and outflow angles were found, a multivariate approach was used to identify the best predictors of acute occlusion. The logistic analysis, conditioned to the comparison between the culprit lesion and the control lesions within each patient, retained the symmetry index and the maximal outflow angle as independent predictors of occlusion and AMI throughout the 36-month follow-up (Table 3). Consideration of average or geometric inflow and outflow angles yielded results similar to maximal angle.
Interaction with time to AMI
Figure 3shows that the shorter the time to AMI, the greater were the symmetry index, the PS and the outflow angle of the culprit stenoses. The slopes of the linear regression in the culprit and control groups were significantly different for the PS and for the maximal outflow angle. On the other hand, time did not influence the predictive value of the symmetry index. No influence of time to AMI was observed for absolute diameters, inflow angles and stenosis length. The findings with average and geometric angles were similar to those obtained with the maximal angles.
Conditional multivariate analyses were obtained for infarctions occurring early (≤12 months, n = 38) or late (>13 months, n = 46) after the initial angiogram. Patients with early and late infarctions had similar age, gender, risk factors, clinical presentation, mean number of diseased coronary segments and drug intake. In early infarctions, the symmetry index (p = 0.005), the outflow angle (p = 0.03) and the PS (p = 0.04) were independent predictors, whereas late infarctions could only be predicted by the symmetry index (p = 0.03).
Interaction with patient-related characteristics
No consistent and significant interaction existed between the angiographic pattern of the culprit stenoses compared with controls and patient-related characteristics such as age, sex, risk factors and medications, except aspirin.
Data on aspirin use were missing in nine patients. Aspirin users (n = 23) and non-users (n = 52) were similar with regards to age, gender, risk factors, clinical presentation, medication used, extent of CAD and characteristics of control lesions. As shown in Table 4, culprit stenoses in aspirin users versus control lesions were significantly more symmetrical, more severe and tended to be steeper. In non-users, culprit stenoses compared with control lesions were also more symmetrical and steeper, but had similar PS. The interaction between PS and aspirin use was significant (p = 0.05). The multivariate analyses retained the symmetry index as the independent predictor of AMI in aspirin users (p = 0.02), and the symmetry index (p = 0.005) and the maximal outflow angle (p = 0.03) as predictors in non-users.
Stenosis geometry and risk profile
To seek whether these findings could help define stenoses of intermediate severity (SIS) (ranging from 40% to 70% diameter reduction) at higher risk (lower risk) of acute occlusion, we defined three groups of lesions using the combination of the symmetry index and the maximal outflow angle, provided that these variables were the two independent predictors of occlusion by multivariate analysis (Table 3). The threshold value of these variables was arbitrarily defined as the lower limit of the tercile that yielded a prevalence of culprit lesions above 25% in univariate analysis, namely 68% and 27° for the symmetry index and the maximal outflow angle, respectively (Fig. 3). The low-risk criteria were a symmetry index and a maximal outflow angle below 68% and 27°, respectively, and the high-risk criteria were a symmetry index and a maximal outflow angle above 68% and 27°, respectively. The remaining lesions were defined as being at medium-risk. There were 165 SIS. Figure 4shows that 49% of the SIS having the high-risk criteria were culprit lesions at 3-year follow-up. Sensitivity, specificity, positive and negative predictive values were 33%, 85%, 49% and 76%, respectively, for the high-risk criteria. On the other hand, the low-risk criteria identified control and stable SIS with a sensitivity, specificity, positive and negative predictive values of 44%, 83%, 87% and 38%, respectively.
This study has characterized the angiographic features of the coronary artery stenoses associated with a high risk of AMI within the following 3 years. Stenosis symmetry, which was defined as the percent ratio of the plaque areas of the two borders of the stenosis, was associated with an increased risk of occlusion at any time during follow-up, whereas stenosis severity and outflow abruptness were predictors of early infarctions. It has also demonstrated that these features can be modulated by aspirin use.
Stenosis geometry favoring occlusion
In this study, the angiographic symmetry of the coronary lesions was a strong and independent predictor of occlusion, contrasting with the generally accepted concept that stenosis eccentricity is a marker of unstable atheromatous plaques (9). However, the visually assessed type II eccentric lesion, initially described in patients with unstable angina, was not predictive of AMI in another study (4). An autopsy study of patients who died from CAD also observed that culprit lesions were typically concentric as opposed to nonculprit plaques (10). Potential mechanisms relating plaque symmetry with rupture and acute thrombosis could be mechanical, such as inhomogeneity in the distribution of the circumferential wall shear stress (11,12), or rheologic, such as promotion of platelet deposition and aggregation in areas of secondary flow (13)initiating formation of fibrin rich occlusive thrombi (14), as opposed to platelet aggregate formation at the apex of asymmetrical stenoses, with fragmentation by high shear rates and distal embolization as described in unstable angina (14–16). Alternatively, plaque structure and composition may influence angiographic morphology. The high lipid content together with the thin and collagen-depleted fibrous cap of a vulnerable plaque (1,17)may favor angiographic symmetry as the plaque gets softer and less resistant to the plastic deformation induced by hemodynamic shear forces. Therefore, the symmetry index could indirectly indicate plaque softness and composition. This hypothesis could explain the relative time-independence of its prognostic value as well as the absence of aspirin interaction.
The critical role of the outflow abruptness of the plaque is re-emphasized in this study (6). Steep outflow angles predicted thrombotic occlusion independently of other geometric parameters, including stenosis severity. Abrupt stenosis expansion decreases pressure recovery and increases flow separation and vortices generation distal to the apex (18), which may favor wall vibration and plaque fissuration. Low shear vortices and secondary flows may further promote platelet aggregation and thrombus growth (19), although the absence of a protective effect of aspirin may suggest that this latter mechanism is less critical. The inlet angles had no prognostic importance in this study, as opposed to previous findings (6), but in agreement with experimental data showing little influence of this angle on blood flow disturbances and pressure recovery (18). Platelet-rich thrombi were previously shown to form distal to the stenosis apex but not in the stenosis inlet, where shear stress is the highest (20).
The association between stenosis severity and coronary occlusion is complex, but important for clinical decision making (2,3). High-grade stenoses occlude more frequently (1,7,21), but often silently without AMI (7,22). Conversely, stenoses of moderate severity occlude less often (23), but lead more frequently to AMI (1,4,5,7). Accordingly, stenoses that led to AMI in our study, although more severe than control lesions, were not of high-grade severity.
Influence of time
The design of this study based on a single transversal look at coronary anatomy did not allow the study of the stenosis progression rate as a predictor of AMI (22,24). Its size, however, has allowed characterization of interactions between geometric features and time to AMI. The closer the onset of the AMI, the greater the outflow angles and the PS of the culprit lesions as compared with the control lesions. These two variables were indeed predictive of early coronary occlusion, within 12 months after angiography, suggesting that factors affecting rheology may have a critical importance in triggering plaque rupture or thrombosis formation. On the other hand, the symmetry index was predictive of coronary occlusion at any time during follow-up, suggesting that this index may be primarily related to plaque structure, composition and stability.
Influence of patient-related factors
Lesions that occlude in patients taking aspirin were more severe, suggesting that lesion obstruction promotes thrombus formation. On the other hand, aspirin did not influence the prognostic value of the symmetry index, suggesting that platelet activation is not critical in destabilization of symmetrical stenoses.
Limitations of the study
The retrospective design of this study may have introduced limitations such as biases in the population selection or incomplete collection of clinical data. The most important bias could have been selection of patients with prior coronary heart disease since all patients had had an angiogram before the onset of the qualifying AMI. This precludes any extrapolation of our findings to patients with an inaugural AMI. One third of our patients developed a non-Q wave AMI, although all patients had ST segment elevation at hospital admission. Patients with a Q wave or a non-Q wave AMI had similar clinical and angiographic characteristics. The findings could have been different in patients presenting with ST segment depression or T wave inversion. To minimize “patients effects” as much as possible, a statistical approach that matched a case (the culprit lesion) with a various number of control cases (the nonculprit lesions in the same patient) was used, focusing on “lesion effects,” each patient being his own control.
Coronary angiography is not the best tool to quantitate stenosis dimensions. Intravascular ultrasounds allow more a precise description of selected stenoses and provides insights in plaque composition (25). Since they require intracoronary manipulation and allow the study of a limited number of arterial segments, this approach appears inadequate and potentially unsafe to describe multiple lesions in a large population of patients. Quantitative analysis of multiple angiographic projections could have theoretically allowed three-dimensional reconstruction of the vascular lumen (26), but currently available softwares impose specific projections, precluding an extensive description of a relatively large number of stenoses in a given patient. Similarly, information provided by videodensitometric measurements have been disappointing so far (27). It was believed that the known limitations of monoplane angiography were more likely to introduce random noise, which could have obscured the differences between culprit and control lesions, rather than measurement biases. Besides, the statistical analyses have provided consistent and corroborating results, beyond the variability of the methods. Eventually, the limited impact of our findings to accurately predict future unstable coronary lesions might hopefully benefit from the more precise vessel wall contour detection obtained with last-generation QCA softwares (28).
Clinical implications and conclusion
Identification of coronary lesions at high risk of occlusion and AMI remains challenging. Making a therapeutic decision about stenoses of intermediate severity (ranging between 40% and 70% diameter stenosis) is especially difficult since these lesions may not be severe enough to cause ischemia in daily life, although some may result in acute occlusion and AMI. Correct risk identification would have a great clinical impact in orienting the need for preventive targeted treatment, as well as in avoiding useless revascularization procedures. We thus sought to evaluate whether the combination of the two independent angiographic predictors of occlusion could help discriminate those SIS at highest risk or lowest risk of occlusion. Beyond the fact that all patients developed an AMI, which could bias the predictiveness of any “high-risk” criteria, we found that this model did not accurately identify the culprit lesions; the positive predictive value was only around 50% in the high-risk sub-group. On the other hand, the combination of a symmetry index below 68% and a maximal outflow angle below 27° appeared to be helpful to discriminate 40% to 70% of stenoses that would remain free of occlusion and infarction over a 3-year follow-up period, with 83% specificity and 87% positive predictive value. With regards to prevention of infarction, the clinical use of these two stenosis descriptors might prove to be useful in avoiding useless revascularization of lesions less than 70% stenosis.
Provided that these findings are confirmed prospectively in a less selected population of patients, using even more accurate edge-detection softwares alone or in the purpose of orienting intravascular ultrasound stenosis assessment, study of stenosis geometry might not only contribute to the understanding of the pathophysiology of plaque rupture and coronary thrombosis but might also be useful and cost-effective to stratify risk and to orient conservative therapeutic strategies.
We thank Francois Harel and André Couturier for their expert statistical assistance.
☆ Dr. François Ledru was partly supported by a grant from the Assistance Publique-Hôpitaux de Paris, Paris, France.
- acute myocardial infarction
- coronary artery disease
- Coronary Measurement System
- infarct-related lesion
- percent lumen diameter stenosis
- quantitative coronary analysis
- stenoses of intermediate severity
- Thrombolysis In Myocardial Infarction
- Received June 5, 1998.
- Revision received October 19, 1998.
- Accepted January 5, 1999.
- American College of Cardiology
- Falk E,
- Shah P.K,
- Fuster V
- Ryan T.J,
- Bauman W.B,
- Kennedy J.W,
- et al.
- Little W.C,
- Constantinescu M,
- Applegate R.J,
- et al.
- Taeymans Y,
- Théroux P,
- Lespérance J,
- Waters D
- Lespérance J,
- Théroux P,
- Hudon G,
- Waters D
- Reiber J.H.C,
- van der Zwet P.M.J,
- von Land C.D,
- et al.
- Ambrose J.A,
- Winters S.L,
- Arora R.R,
- et al.
- Cheng G.C,
- Loree H.M,
- Kamm R.D,
- et al.
- Schoephoerster R.T,
- Oynes F,
- Nunez G,
- et al.
- Lassila R,
- Badimon J.J,
- Vallabhajosula S,
- Badimon L
- Mailhac A,
- Badimon J.J,
- Fallon J.T,
- et al.
- Davies M.J,
- Richardson P.D,
- Woolf N,
- et al.
- Baumgartner H,
- Schima H,
- Tulzer G,
- Kühn P
- Strony J,
- Beaudoin A,
- Brands D,
- Adelman B
- Alderman E.L,
- Corley S.D,
- Fisher L.D,
- et al.
- Lichtlen P.R,
- Nikutta P,
- Jost S,
- et al.
- Dacanay S,
- Kennedy H.L,
- Uretz E,
- et al.
- Waters D,
- Craven T,
- Lespérance J
- Nissen S.E,
- Gurley J.C,
- Grines C.L,
- et al.
- Rosenfield K,
- Losordo D,
- Ramaswamy K,
- et al.
- Strauss B.H,
- Escaned J,
- Foley D.P,
- et al.
- van der Zwet P.M,
- Reiber J.H