Author + information
- Received March 7, 1997
- Revision received April 21, 1998
- Accepted May 11, 1998
- Published online September 1, 1998.
- Conor F Lundergan, MDa,*,
- Jonathan S Reiner, MD, FACCa,
- William F McCarthy, PhDa,
- Karin S Coyne, RN, MPHa,
- Robert M Califf, MD, FACC∗,
- Allan M Ross, MD, FACCa,
- for the GUSTO-I Angiographic Investigators
- ↵*Address for correspondence: Dr. Conor F. Lundergan, Cardiovascular Research Institute & Division of Cardiology, The George Washington University Medical Center, 2150 Pennsylvania Avenue, NW, Washington, DC 20037
Objectives. The purpose of this study was to determine patient characteristics that are a priori predictors of early infarct related artery patency following thrombolytic therapy, and to provide a paradigm which may identify patients who would be most likely to achieve restoration of normal (TIMI 3) coronary flow in response to thrombolytic therapy.
Background. Restoration of infarct-related artery perfusion in acute myocardial infarction is necessary for preservation of ventricular function and mortality reduction. Clinical variables that are a priori predictors of early patency with currently available thrombolytic regimens have not been fully characterized.
Methods. The probability of early infarct-related artery patency (TIMI 3 flow) was determined by multivariable logistic regression. We determined a reduced (parsimonious) model for predicting early (90 min) infarct-related artery patency (TIMI grade 3) based on data from 1,030 patients in the GUSTO-I Angiographic study.
Results. Predictors of 90 min TIMI 3 flow are use of an accelerated t-PA regimen (vs. streptokinase containing regimens) (χ2= 39.1; p ≤ 0.0001), infarct related artery (RCA/Lcx vs. LAD) (χ2= 12.7; p = 0.0004), body weight (χ2= 10.3; p = 0.001) and history of smoking (χ2= 7.4; p = 0.007). Time from symptom onset to treatment was not significant (p = 0.71).
Conclusions. The efficacy of currently available thrombolytic regimens is chiefly dependent on choice of thrombolytic regimen, body weight, infarct-related coronary artery and smoking history. Clinical variables alone correctly predict a priori TIMI 3 flow in the infarct-related artery 64% of the time. Patients with body weights greater than 85 kg are at a significant disadvantage with regard to achieving successful thrombolysis compared to those with lesser body weights.
Mortality reduction in acute myocardial infarction is dependent upon the efficacy of thrombolytic regimens with regard to reestablishing normal infarct-related artery flow (1). Successful reperfusion of initially occluded infarct-related coronary arteries is the result of a complex interplay among clinical, hemodynamic, mechanical and biochemical factors (2). Clinical variables that determine the efficacy of thrombolytic therapy, however, have been poorly described. The Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries (GUSTO-I) Angiographic Study offered the unique opportunity to determine clinical determinants of early (90 min) infarct-related artery patency. The purpose of this study is to describe those patient or treatment variables of greatest importance in determining a priori early infarct-related artery patency following thrombolytic therapy. We also sought to develop a predictive model based on clinical variables that could possibly be used as a paradigm for determining the likelihood of thrombolytic therapy success.
The GUSTO-I Angiographic study has been described in detail previously (3). Briefly, patients were enrolled in GUSTO-I if they experienced chest pain <6 hs in duration and had electrocardiographic evidence of acute myocardial infarction (ST segment elevation). Patients were randomized to one of four treatment regimens which included 1) streptokinase with subcutaneous heparin; 2) streptokinase with intravenous heparin; 3) accelerated dose t-PA with intravenous heparin; and 4) combination therapy with t-PA and streptokinase with intravenous heparin. Patients were also randomized to one of four time periods for angiography (3). This study includes coronary angiographic data from those patients randomized to 90 min postthrombolytic therapy angiography with cine films adequate for determining infarct-related artery flow grade.
Core angiographic laboratory cineangiographic analysis
All cine films were interpreted by an experienced angiographer (CFL, JSR, AMR) who had no knowledge of treatment allocation or angiographic time randomization as described in the Thrombolysis in Myocardial Infarction (TIMI) trial (4). Coronary segments identified as the infarct-related segment were considered proximal if they occurred prior to or immediately following the first septal perforator in the left anterior descending coronary artery, prior to the second marginal branch of the right coronary artery or prior to the second marginal branch of the left circumflex artery. All other infarct-related segments were considered distal sites.
Exploratory data analyses were performed to determine the relationship between individual clinical variables and patency (TIMI 3 flow) in 1,030 patients randomized to 90 min angiography with complete data sets. Analyses of patency determinants were done only for those patients with native coronary arteries as the infarct-related vessel. Boxplots and contingency tables were utilized to conceptually compare the distributions of continuous and discrete variables with respect to patency. The Kolmogorov-Smirnov test was used to compare the empirical distribution of continuous variables (5). ANOVA and the median test were used to compare the means and medians of continuous variables respectively (5). Contingency tables and the Fisher’s exact test were used to compare categorical data (5). Single variable and multivariable logistic regression models were developed to determine which clinical variables were predictors of 90 min postthrombolytic therapy patency (6). Both ordinal (proportional odds model) and binary outcomes were considered. The goal of multivariable modelling was to create a final reduced (parsimonious) model that would predict early patency as well as a full model. The c-index was used to quantify the predictive discrimination of a specific model (7). The c-index is equivalent to the area under the receiver operating characteristic curve. A c-index of 0.5 would indicate that a given model predicted patency no better than a coin flip (8). The relative weight of each independent variable in the final reduced model was determined by comparison of the standardized estimate coefficient of each variable (9). The full model contained all possible predictors of infarct-related artery patency, including appropriate interaction terms. These variables were chosen based on their contribution to 30-day postinfarction mortality as described by Lee et al. and based on data from the 41,021 patient GUSTO-I trial (7). They included age, gender, diabetes, history of hypertension, smoking status (never smoked, current smoker, history of smoking), time from symptom onset to treatment, height (cm), body weight (kg), infarct sight (anterior, other), history of hypercholesterolemia, history of coronary artery bypass surgery, history of previous myocardial infarction, history of angina, study entry pulse, systolic and diastolic blood pressure, study entry Killip class, and age and Killip class interaction, treatment regimen and body weight interaction. Killip class at study entry was dropped from the final model due to the extreme skewness of the distribution of this data set with only 1% of patients entered into the study being classified as Killip class III or IV. We did not examine ECG related variables, such as ST segment changes, because the purpose of this study was to identify patient-related variables that may identify a priori (that is, prior to thrombolytic drug administration) the likelihood of successful thrombolysis. Unadjusted and adjusted odds ratios for 90-min patency (TIMI 3 flow) were determined from single and multivariable logistic regression models.
The GUSTO-I Angiographic Study enrolled 1,077 patients who were randomized to 90-min postthrombolytic therapy angiography who had adequate coronary angiographic images for determination of TIMI flow grade. Tables 1 and 2⇓⇓display 90-min infarct related artery patency status according to various patient characteristics and treatment regimens. Patients who achieved normal (TIMI 3) coronary flow 90 min posttherapy were more likely to have the right or left circumflex coronary artery as the infarct related vessel (Table 1A). As shown in Table 1B, 90-min TIMI 3 flow was noted in greater proportion of those patients with either proximal right coronary artery or proximal left circumflex coronary artery as the presumed site of the culprit lesion compared to angiographically determined distal locations of either of these vessels or a proximal left anterior descending coronary artery location. Patients with early TIMI 3 flow had significantly lower body weights and were more likely to have a history of smoking compared with those patients with less than normal flow. Patients with 90-min TIMI 3 flow were also more likely to have received the accelerated t-PA regimen rather than a streptokinase regimen. Of note, the distribution of time from symptom onset to treatment were similar in patients who achieved normal coronary flow and those with less than normal flow. Univariate analysis failed to detect any difference in those patients with TIMI 3 flow compared to those with less than TIMI 3 flow with regard to age, gender or traditional cardiovascular risk factors such as diabetes, hypertension or lipid status. Hemodynamic parameters at the time of study entry such as heart rate, systolic or diastolic blood pressure were not clinically different between these two groups.
|LAD||124/395 (31%)||6/20 (30%)||0.002∗|
|LCx||56/122 (50%)||2/10 (20%)||0.038†|
|RCA||160/375 (43%)||37/108 (34%)||0.008‡|
↵legend Values represent proportion of infarct-related artery location with 90 min post thrombolytic therapy TIMI 3 flow.
↵∗ = p Value for comparison of proximal LAD vs all other categories.
↵† = p Value for comparison of proximal LCx vs. all other categories.
↵‡ = p Value for comparison of proximal RCA vs. all other categories.
|B.legend||TIMI 3||TIMI 2||TIMI 0,1||p Value|
|SK(IV + SQ)||36.8||45.2||64.1|
|t-PA + SK||26.7||31.5||21.5|
Predictors of 90-min TIMI 3 flow are shown in Table 3. Use of an accelerated t-PA regimen, infarct related artery (right and circumflex coronary artery vs left anterior descending coronary artery), body weight and history of smoking were each significant determinants of 90-min infarct-related artery flow status. The relative weight of each of these variables with regards their contribution to the probability of 90-min TIMI 3 flow is 2.2:1.3:1.2:1.0, respectively. Substitution of infarct related artery specific segment of presumed occlusion (i.e., proximal vs. distal) for infarct-related artery as a candidate variable for testing by multivariable analysis resulted in the inclusion of proximal right and left circumflex infarction related coronary segments (vs. all other sites) in the model (χ2= 16.0; p = 0.0001). Time from symptom onset to treatment was not significant as a predictor of early TIMI 3 flow (χ2= 0.13; p = 0.71). Time to treatment was only of marginal significance (p = 0.03) when patency outcome was analyzed as a dichotomous variable (TIMI 2,3 vs TIMI 0,1) (odds ratio 0.89; 95% confidence limits 0.80–0.99); Of note, age (p = 0.08), gender (p = 0.16), diabetes (p = 0.08), and hypercholesterolemia (p = 0.46) were not significant predictors of 90-min infarct-related artery TIMI 3 flow. The c-index for this model is 0.64. This reduced model contains >90% of the information found in a model containing all tested clinical variables.
Figure 1displays the probability of 90-min TIMI 3 flow as a function of body weight adjusted for treatment regimen and infarct related artery in patients who never smoked. There is a 1.6 fold (95% confidence limits 1.2–2.1) greater likelihood of achieving 90-min posttherapy TIMI 3 flow if the infarct-related artery is either the right or circumflex coronary vessel compared with the left anterior descending coronary artery regardless of the treatment regimen. This effect is relatively constant over the range of patient body weights included in this study.
Figure 2shows there is a 2.5 times greater likelihood of achieving 90-min TIMI 3 flow with the use of an accelerated t-PA regimen compared with streptokinase regimens. This effect is also constant over the range of patient body weights and infarct-related arteries. Figure 2also displays the adjusted odds ratios for 90-min TIMI 3 flow according to infarct related artery, patient smoking history and a range of body weights. The body weights chosen represent the 10%, 50%, 75%, 90% and 95% percentiles of patient body weights in this study. For any 1 kg difference in body weight, the odds ratio for achieving 90-min TIMI 3 flow is 1.02 with the greater likelihood in favor of the lower body weight. For any 17, 28, 38, or 43 kg difference in body weight, the respective odds ratios are 1.3, 1.6, 1.9 and 2.0, respectively, in favor of the lower body weight.
The GUSTO-I trial and GUSTO-I Angiographic Study provided powerful evidence supporting the hypothesis that infarct-related artery patency is of prime importance with regards to preservation of left ventricular function and mortality reduction in the setting of acute myocardial infarction (1,3,10). The GUSTO-I Angiographic Study, due to its large size, also provided for the first time a unique opportunity to examine the primary clinical predictors of early (90 min) and normal (TIMI 3) reperfusion following thrombolytic therapy. The primary predictors of 90-min infarct-related artery patency are: 1) use of an accelerated t-PA regimen; 2) body weight; 3) the infarct-related artery; and 4) a history of smoking.
Patency, thrombolytic therapy regimen and body weight
A substantial reduction in the likelihood of achieving TIMI 3 flow occurs for successive increments in body weight when t-PA is the thrombolytic agent of choice and delivered in an accelerated manner as described by Neuhaus et al. (11). In patients weighing ≥85 kg, the probability of achieving TIMI 3 flow in an occluded left anterior descending coronary artery is less than 50% (47% in patients with a history of smoking; 38% in those who never smoked) using the t-PA regimen described in GUSTO-I. For patients weighing ≥105 kilograms the probability of 90-min TIMI 3 flow with t-PA falls to 39% in those with a smoking history and 30% in those patients who never smoked. Of note, 23% of the patients enrolled in the GUSTO-I Angiographic Study had body weights greater than 85 kg. These data are very similar to the TEAM-3 results of 24-h infarct-related artery patency using t-PA delivered over 3 h (12). Use of a streptokinase regimen resulted in even greater reduction in the likelihood of early patency between the lowest body weight quantile and the highest quantile. Streptokinase containing regimens are associated with a 45%–48% reduced likelihood of TIMI 3 flow 90 min postthrombolytic therapy over the entire range of patient body weights examined compared with the accelerated t-PA regimen in those patients with acute anterior myocardial infarctions.
These data suggest that the current dosage recommendations for t-PA and streptokinase may be less than optimal for achieving a maximal effect on early patency in patients weighing more than 80–85 kg.
Patency and infarct-related artery
There are conflicting reports concerning the effect of infarct related artery location on early patency in response to thrombolytic therapy. The TIMI I investigators reported a greater reperfusion rate (TIMI 2 + TIMI 3) in the left anterior descending coronary artery compared to either the right or circumflex coronaries in patients who received double chain t-PA (4). They were unable to detect an effect on infarct segment location (proximal vs. distal) on the likelihood of reperfusion (4). The TAMI I investigators reported a statistically insignificant trend toward a higher patency rate (TIMI 2 plus TIMI 3) 90 min after initiation of thrombolytic therapy with t-PA when the left anterior descending coronary artery was the infarct-related artery (13). They reported an insignificant trend toward less successful reperfusion with proximal infarct related segments compared with distal segments (13). Multivariable modeling for patency determinants was not done in either the TIMI I nor TAMI I studies (4,13). Anderson et al., reporting for the TEAM-2 Investigators, found a significantly greater reperfusion rate (TIMI 2 + TIMI 3) in the combined right and circumflex coronary arteries compared to the left anterior descending with anistreplase but no difference in vessel-specific patency rates with streptokinase (14). Pacouret et al. and Leizorovicz et al. failed to find any significance of infarct-related artery on patency (TIMI 2 + TIMI 3) rates with either anistreplase or streptokinase even after adjustment for clinical variables (15,16). Our database represents the largest single trial patient population with complete demographic and angiographic data sets thus far utilized to detail important predictors of early infarct-related artery patency. It appears by multivariable analysis that the likelihood of achieving TIMI 3 flow 90-min postthrombolytic therapy in the right or left circumflex coronary artery is approximately 1.3 to 2.1 fold greater than the likelihood of achieving normal reperfusion in the left anterior descending coronary vessel. The location of the infarct segment within the infarct-related artery, that is, proximal or distal, also appears to be of importance and retains significance following adjustment for multiple clinical variables. The reason for the greater likelihood of achieving early normal flow in the right or circumflex vessels and in more proximal infarct segments of these vessels is most likely multifactorial and related, for example, to thrombus burden (2,4), plaque burden and possibly collateral flow in the infarct region (17). We are unable to determine the effect of collateral flow on early postthrombolytic therapy infarct-related artery patency because our database does not include pretreatment assessment of collateral flow status.
Patency and smoking
A detailed analysis of the paradoxical beneficial effects of smoking with regard to mortality in patients receiving thrombolytic therapy for acute myocardial infarction has been reported using the GUSTO-I Angiographic Study database (18). A significant difference was noted in early (90-min) TIMI 3 patency rates between patients with a history of smoking and those who never smoked in favor of those with a smoking history (41% vs. 33%, p = 0.02). The TEAM-2 investigators also reported that current smokers have a significantly greater chance of achieving TIMI 3 flow in infarct related arteries 90–240 min following thrombolysis in patients treated with either streptokinase or anistreplase (19). The beneficial effect of smoking remained significant in that study after adjustment for baseline clinical and angiographic variables and was independent of the infarct-related artery (19). Our current analysis, based on a much larger data set, supports the TEAM-2 findings with regard to early patency and patient smoking status. The physiologic mechanism responsible for this phenomena remains to be explained but may be related to a greater initial thrombus burden and less plaque burden in smokers (18).
Patency and time to treatment
Of note, time to treatment was not an important predictor of the likelihood of successful (TIMI 3) infarct-related artery reperfusion following adjustment for multiple clinical variables. The likelihood of successful reperfusion has been noted to decline with increasing times from symptom onset to thrombolytic therapy with streptokinase and anistreplase in several previous studies (4,16,20–22). Recently, the RAPID II investigators, upon univariate subgroup analysis, reported a significant effect of time to treatment on the occurrence of TIMI 3 flow in the infarct-related artery with both double bolus reteplase and an “accelerated” infusion of t-PA (23). The TIMI I investigators, in a much smaller sample size compared to GUSTO-I, reported that streptokinase exhibited diminished efficacy with regard to reperfusion, measured as TIMI 2 or TIMI 3 flow, in the infarct-related artery in patients treated between 4–6 h after symptom onset compared to those treated within 4 h (44% vs 27%) (4). They noted no such time dependent difference with regard to patency when t-PA was used as the thrombolytic agent (4). The TIMI I (4)investigators did not report separate 90 min postthrombolysis TIMI 2 and TIMI 3 patency rates for comparison with our data. However, it is of interest that repeat analysis of our data using the endpoint of 90 min postthrombolysis patency defined as a dichotomous variable (TIMI 2,3 vs TIMI 0,1) did demonstrate that time from symptom onset to treatment was of marginal significance with regard to patency (p = 0.03). Time to treatment may assume more importance with regard to the likelihood of achieving either TIMI 2 or 3 infarct-related artery flow when time from symptom onset to treatment exceeds 6 h as was the case in TIMI I (4)and RAPID II (23). However, we are unable to address this issue with our database as GUSTO-I limited enrollment to those patients with symptom onset ≤6 h prior to treatment. It is generally recognized that TIMI 2 flow represents an inadequate degree of reperfusion postthrombolysis (1,24). However, a significant proportion of patients with TIMI 2 flow 90 min postthrombolytic therapy develop TIMI 3 flow in the infarct-related vessel during the first week postinfarction and subsequently demonstrate improvement in systolic ventricular function (25). Therefore, shorter times to treatment, within a 6-h window, may be of some benefit, although probably marginal, with regard to restoration of infarct-artery patency. However, time to treatment should not be considered of special importance with regard to the likelihood of achieving fully normal (TIMI 3) reperfusion postthrombolysis if thrombolysis is attempted within 6 h of symptom onset.
Implications of results
The quest for greater reperfusion success with alternative dosing regimens of existing and investigational thrombolytic agents must be balanced against the risks of increasing the incidence of stroke or cerebral hemorrhage. Although the use of an accelerated dosing regimen of t-PA is associated with an increased risk of stroke (relative to streptokinase) (10), detailed data regarding the dose/unit body weight/infusion rate vs. stroke are not available. Based on aggregate data from several small studies it would appear that early infarct-related artery patency is dependent on the rate of infusion of t-PA, whereas the likelihood of intracranial hemorrhage is more related to total thrombolytic dose and the level of concomitant anticoagulation (26,27). On the basis of the present study observations it would seem clear that further investigation is needed regarding alternative lytic dosing in patients with body weights greater than 85 kg using the thrombolytic agents described in this study.
The principle limitation of this study is the sensitivity and specificity of our multivariable model for determining the probability of 90-min postthrombolysis infarct-related artery patency. As noted above, the predictive discrimination of this model is not especially strong (c index = 0.64). The false positive rate and false negative rate to be expected with this model are 40% and 33%, respectively. This suggests that this model should not be used as a clinicaltool to predict early successful reperfusion following thrombolytic therapy (28). Rather this model should be considered descriptive with regard to the main clinical predictors of early infarct-related artery patency and suggests problem areas that need to be addressed in designing future treatment strategies. It should be noted that the odds ratios and relative weights that each variable in the model contributes to the final prediction of infarct-related artery patency is dependent on the adequacy and completeness of the independent variables tested. A model with improved predictive discrimination with regard to early patency would probably need to include incompletely understood hematologic and metabolic variables related to patency but which were not available for incorporation into our model.
A second limitation is the subjective nature of quantifying the primary outcome variable of this study, i.e., patency. A change in classification of infarct-related artery status from TIMI 3 to TIMI 2 or vice versa could potentially alter the results of our multivariable model. However, repeating our analysis using an outcome variable of TIMI 2 + TIMI 3 flow failed to show marked differences from the results described above. In addition, there were no major differences between a proportional odds model and the logistic regression model presented above using a binary outcome. These results suggest that the degree of “misclassification” of either TIMI 2 or TIMI 3 flow did not contribute significantly to our findings.
The primary predictors of early infarct artery patency (TIMI 3 flow) are the use of an accelerated t-PA drug regimen, body weight, the particular infarct-related artery and patient smoking history. Our data suggests that nonsmoking patients with anterior infarctions and body weights greater than 85 kg are at a significant disadvantage with regard to achieving early successful reperfusion. Alternative strategies, including either augmented thrombolytic dose regimens or mechanical reperfusion methods, should be carefully considered in such patients.
☆ This study was supported by a combined grant from: Bayer (New York, New York), CIBA-Corning (Medfield, Massachusetts), Genentech (South San Francisco, California), ICI Pharmaceutical (Wilmington, Delaware) and Sanofi Pharmaceutical (Paris, France).
- Received March 7, 1997.
- Revision received April 21, 1998.
- Accepted May 11, 1998.
- American College of Cardiology
- for the GUSTO-I Investigators,
- Simes R.J,
- Topol E.J,
- Holmes D.R,
- et al.
- Chesebro J.H,
- Knatterud G,
- Roberts R,
- et al.
- Armitage P,
- Berry G
- Freund R.J,
- Littell R.C
- for the GUSTO-I Investigators,
- Lee K.L,
- Woodlief M.S,
- Topol E.J,
- et al.
- Miller M.E,
- Langefeld C.D
- ↵SAS Institute Inc. Logistic Regression Examples Using the SAS System. Version 6, 1st edition, Cary, NC: SAS Institute Inc., 1995.
- Neuhaus K.L,
- Feuerer W,
- Jeep-Tebbe S,
- et al.
- Anderson J.L,
- Sorenson S.G,
- Moreno F.L,
- et al.
- Pacouret G,
- Charbonnier B,
- Curien N.D,
- et al.
- Leizorovicz A,
- Boissel J.P,
- Robert F
- Tendera M.P,
- Campbell W.B,
- Tennant S.N,
- et al.
- Barbash G.I,
- Reiner J.S,
- White H.D,
- et al.
- for the APSAC Multicenter Investigators,
- Anderson J.L,
- Rothbard R.L,
- Hackworthy R.A,
- et al.
- for the RAPID II Investigators,
- Bode C,
- Smalling R.W,
- Berg G,
- et al.
- Lenderink T,
- Simoons M.L,
- Van ES G.A,
- et al.
- Reiner J.S,
- Lundergan C.F,
- Fung A,
- et al.
- Smalling R.W,
- Schumacher R,
- Morris D,
- et al.
- Gore J.M,
- Granger C.B,
- Simoons M.L,
- et al.
- LeGall J.R