TIMI Frame Count
A Quantitative Method of Assessing Coronary Artery Flow
Background Although the Thrombolysis in Myocardial Infarction (TIMI) flow grade is a valuable and widely used qualitative measure in angiographic trials, it is limited by its subjective and categorical nature.
Methods and Results In normal patients and patients with acute myocardial infarction (MI) (TIMI 4), the number of cineframes needed for dye to reach standardized distal landmarks was counted to objectively assess an index of coronary blood flow as a continuous variable. The TIMI frame-counting method was reproducible (mean absolute difference between two injections, 4.7±3.9 frames, n=85). In 78 consecutive normal arteries, the left anterior descending coronary artery (LAD) TIMI frame count (36.2±2.6 frames) was 1.7 times longer than the mean of the right coronary artery (20.4±3.0) and circumflex counts (22.2±4.1, P<.001 for either versus LAD). Therefore, the longer LAD frame counts were corrected by dividing by 1.7 to derive the corrected TIMI frame count (CTFC). The mean CTFC in culprit arteries 90 minutes after thrombolytic administration followed a continuous unimodal distribution (there were not subpopulations of slow and fast flow) with a mean value of 39.2±20.0 frames, which improved to 31.7±12.9 frames by 18 to 36 hours (P<.001). No correlation existed between improvements in CTFCs and changes in minimum lumen diameter (r=−.05, P=.59). The mean 90-minute CTFC among nonculprit arteries (25.5±9.8) was significantly higher (flow was slower) compared with arteries with normal flow in the absence of acute MI (21.0±3.1, P<.001) but improved to that of normal arteries by 1 day after thrombolysis (21.7±7.1, P=NS).
Conclusions The CTFC is a simple, reproducible, objective, and quantitative index of coronary flow that allows standardization of TIMI flow grades and facilitates comparisons of angiographic end points between trials. Disordered resistance vessel function may account in part for reductions in flow in the early hours after thrombolysis.
The TIMI flow-grading system classifies successful reperfusion after thrombolysis as either grade 2 (partial) or grade 3 (complete) flow.1 Although this widely used method of grading coronary flow has been a valuable tool for comparing the efficacy of thrombolytic agents and identifying patients at higher risk for adverse outcomes,2 3 4 5 6 7 the assessment of this angiographic end point is subjective. Variability in the visual assessment of this end point hampers comparisons of flow data between agents and between trials. Furthermore, in addition to its subjective nature, the conventional flow-grading system is categorical, and no continuous angiographic index of coronary flow currently exists. To standardize the assessment of coronary flow and develop a simple continuous index of coronary flow, the number of cineframes required for contrast material to first reach standard distal coronary landmarks (the TIMI frame count) was determined after thrombolysis in the TIMI 4 trial. The results of this study were published previously.8
The analysis was carried out on coronary angiograms from the previously reported randomized, double-blind TIMI 4 trial that compared the efficacy of three thrombolytic regimens8 : anistreplase (Eminase or APSAC), 30 U IV; rTPA (Activase or alteplase), 15 mg IV bolus followed by 0.75 mg/kg (up to 50 mg) IV over 30 minutes, then 0.5 mg/kg (up to 35 mg) IV over 60 minutes; or combination therapy, 20 U IV of APSAC plus a 15-mg bolus of rTPA followed by a 30-minute infusion of 0.75 mg/kg (up to 50 mg).
Angiography was performed immediately after arrival in the cardiac catheterization laboratory and at 60, 75, and 90 minutes after administration of thrombolytic therapy. Cardiac catheterization conditions at 90 minutes after therapy (angle and skew of the gantry, sequence of standard angiographic views, contrast agent, and cardiac catheterization suite used) were reproduced at 18 to 36 hours to minimize variability in imaging and hemodynamic parameters. Sublingual or IV nitroglycerin was administered at both time points and was repeated every 15 minutes to preserve a state of maximum vasodilation. All angiograms were filmed at 30 frames/s.
Quantitative Angiographic Analysis
The cinefilm reviewers were blinded to the treatment group assignment, the interpretation of the angiogram made by the clinical center, and the clinical outcome of the patient. Three consecutive frames from the same phase of the cardiac cycle (preferably end diastole) in the optimal single-plane projection that identified the stenosis in its greatest severity were chosen for quantitative angiographic analysis with a previously described and validated automated edge-detection algorithm.9
Qualitative Angiographic Analysis: TIMI Flow Grade Assessment
TIMI flow grade was assessed at the angiographic core laboratory as previously defined1 : Grade 0—No perfusion; no antegrade flow beyond the point of occlusion. Grade 1—Penetration without perfusion; contrast material passes beyond the area of obstruction but fails to opacify the entire coronary bed distal to the obstruction for the duration of the cineangiographic filming sequence. Grade 2—Partial perfusion; contrast material passes across the obstruction and opacifies the coronary artery distal to the obstruction. However, the rate of entry of contrast material into the vessel distal to the obstruction or its rate of clearance from the distal bed (or both) is perceptibly slower than its flow into or clearance from comparable areas not perfused by the previously occluded vessel (eg, opposite coronary artery or the coronary bed proximal to the obstruction). Grade 3—Complete perfusion; antegrade flow into the bed distal to the obstruction occurs as promptly as antegrade flow into the bed proximal to the obstruction, and clearance of contrast material from the involved bed is as rapid as clearance from an uninvolved bed in the same vessel or the opposite artery.
TIMI Frame Count
To objectively evaluate an index of coronary flow as a continuous quantitative variable, the number of cineframes required for contrast to first reach standardized distal coronary landmarks in the infarct-related artery (the TIMI frame count) was measured with a frame counter on the SONY SME 3500 cineviewer.10 The first frame used for TIMI frame counting is the first frame in which dye fully enters the artery. This occurs when three criteria are met: (1) A column of nearly full or fully concentrated dye must extend across the entire width of the origin of the artery; (2) Dye must touch both borders of the origin of the artery; and (3) There must be antegrade motion to the dye (Fig 1⇓). If the LAD is subselectively engaged and the LCx is the culprit vessel, the TIMI frame count begins when dye first touches both borders at the origin of the LCx. The same rule holds for subselective engagement of the circumflex artery.
The last frame is counted or included as one of the frames and is defined as the frame when dye first enters the distal landmark branch. Full opacification of the branch is not required. Often, the last frame is best determined by running the cinefilm past the initial opacification of the end-point branch and then moving frame by frame in reverse until the end-point branch disappears. Care must be taken to advance one frame forward once the dye disappears to identify the frame in which dye first appears.
The following distal landmark branches are used for analysis: the distal bifurcation of the LAD (ie, the “mustache,” “pitchfork,” or “whale’s tail”; Fig 2⇓); in the circumflex system, the distal bifurcation of the segment with the longest total distance that includes the culprit lesion (Fig 3⇓); and in the RCA, the first branch of the posterolateral artery (Fig 4⇓). Proper panning is essential for counting the number of cineframes required to first opacify the distal artery, particularly the LAD. The TIMI frame count of the LAD and circumflex arteries often is assessed best in either the right or left anterior oblique views with caudal angulation, and the RCA often is assessed best in the left anterior oblique projection with steep cranial angulation. The TIMI frame count in 77 nonculprit arteries in the setting of MI was also assessed in all films that could be evaluated from one center.
Data from 78 consecutive patients who presented to the West Roxbury Veterans Administration Hospital cardiac catheterization laboratory and who had not sustained a myocardial infarction, visually had normal flow in their arteries, and had technically adequate films for analysis were used to correct or normalize the frame count of the LAD for its increased length. The LAD frame count divided by the ratio of its frame count to that of the mean values for the RCA and the LCx (all in the noninfarct setting) was used to derive the corrected TIMI frame count (CTFC). Additionally, a previously described11 three-dimensional vector algebra computational model based on orthogonal views of the coronary arteries was used to determine the distance to the TIMI landmarks in the 37 normal-sized human hearts to confirm this ratio.
All analyses were performed with either the SAS statistical program12 or STATA.13 All continuous variable values were reported as the mean±SD. Analyses were performed with the χ2 test for categorical data and ANOVA or Student’s t test for continuous data. Nonparametric tests were used if the data were not normally distributed. Step-up multiple logistic regression and linear regression models were developed that required a value of P≤.10 for retention in the model. The κ statistic was used to calculate the magnitude and statistical significance of interobserver agreement between the interpretation of angiograms at participating clinical sites and the readings from the angiographic core laboratory.
CK values were sampled every 8 hours, integrated over the first 24 hours, and corrected by the upper limit of normal for the submitting center (multiples of the IU upper limit of normal at each center multiplied by hours). The rate of rise of CK was calculated by subtracting the first CK value collected from the peak CK value in the first 24 hours and dividing by the number of hours between these two CK values (IU/h).
Baseline Characteristics of Conventional TIMI Flow Grades
A total of 393 patients had lesions in the native coronary arteries that were analyzed in the present study (bypass graft lesions were excluded). There were no significant differences among the conventional TIMI flow grades in culprit arteries (assessed by the angiographic core laboratory) with respect to the baseline characteristics of sex, incidence of previous MI, and age. However, among patients with TIMI grade 2 flow, LAD lesions predominated (62.7% LAD versus 37.3% for other, P<.001) and conversely, among patients with TIMI grade 3 flow, RCA and circumflex lesions predominated (26.2% LAD versus 73.8% other, P<.001).
Interobserver Variability Between Clinical Center and Angiographic Core Laboratory Assessment of Conventional TIMI Flow Grades
The frequency of agreement between participating clinical centers and the angiographic core laboratory in the classification of TIMI flow grades was assessed with use of the κ statistic (range of values, −1 to +1). Values of κ>0.75 indicate excellent agreement beyond chance between two observers, whereas values <0.40 indicate poor agreement. Agreement was excellent in assessment of TIMI grade 0 or 1 flow, with an 89% rate of agreement between clinical centers and the angiographic core laboratory (Table 1⇓, κ=0.84±0.05). There was a moderate (79%) rate of agreement in assessment of TIMI grade 3 flow (κ=0.55±0.05). In contrast, for assessment of TIMI grade 2 flow, the rate of agreement was poor at 52% (κ=0.38±0.05). The overall value for assessment of all flow grades was κ=0.59±0.04, indicating moderate agreement.
Variability in the TIMI Frame-Counting Method
The reproducibility of the TIMI frame count was systematically studied in 85 consecutive pairs of injections of the infarct-related artery. The mean absolute value of the difference between two consecutive hand injections spaced apart by 1 to 2 minutes was 4.7±3.9 frames, with a range of 0 to 18 frames (coefficient of variation=9.0%). Reproducibility did not vary significantly by location of the infarct artery.
TIMI Frame Count in the Absence of Acute MI: The CTFC
In 78 consecutive patients presenting to the cardiac catheterization laboratory in the absence of acute MI, the TIMI frame count for the RCA (20.4±3.0 frames, n=38) did not differ significantly from that of the circumflex artery (22.2±4.1 frames, n=21), but the frame count of the LAD was significantly higher (36.2±2.6 frames, n=19, P<.001 for both comparisons) (Table 2⇓). This discrepancy in the distance to the distal arterial landmarks was corrected or adjusted for by dividing the TIMI frame count of the LAD by a factor of 1.7, the ratio of the unadjusted TIMI frame count for the LAD to that of the average of the RCA and circumflex, yielding a CTFC of 21.1±1.5 frames for the LAD. A previously described three-dimensional vector algebra computational model11 was used to determine that the approximate distance to the TIMI landmark in the average human LAD is 14.7 cm, the length of the RCA to the midpoint of the first third of the posterolateral branch (the usual location of the first small branch) is 9.8 cm, and the length of the LCx to the midpoint of the distal third of the third marginal branch is 9.3 cm, yielding comparable ratios of 1.5 (LAD/RCA) and 1.6 (LAD/LCx), for an average ratio of 1.55. Taken together, the mean CTFC for all 78 normal arteries was 21.0 frames with an SD of 3.1 frames, yielding 95% CIs for normal flow of ≥15 to ≤27 frames.
Relationship Between CTFC, Cardiac Catheterization, and Hemodynamic Parameters
Among patent arteries, there was no correlation between the number of injections before 90-minute angiography and the CTFC at 90 minutes (r=.055, P=.43). Furthermore, in a multiple regression model that controlled for infarct-artery location (a significant independent predictor of the CTFC, as discussed below), there remained no correlation. There was also no clear relationship between catheter size and CTFC (5F=32.2±10.6 frames, n=7; 6F=40.8±21.9 frames, n=106; 7F=39.5±19.5 frames, n=82; 8F=34.2±14.7 frames, n=23; four-way P=.40). In the subgroup of patients studied at one center, there was no correlation between the 90-minute CTFC and heart rate, systolic or diastolic blood pressure, right atrial pressure, difference between diastolic arterial blood pressure and right atrial pressure, pulmonary capillary wedge pressure, cardiac output, or cardiac index. When we corrected for infarct-artery location in a multivariable model, there was still no correlation with any of the above variables.
Clinical Center and Angiographic Core Laboratory Interpretation of Conventional TIMI Flow Grades Versus CTFC
When clinical centers interpreted TIMI flow grade, there was nearly total overlap in the distribution of the CTFC for TIMI grade 2 versus TIMI grade 3 flow (Fig 5A⇓). For the angiographic core laboratory, overlap was also present but appeared to be reduced compared with the clinical centers (Fig 5B⇓). This is demonstrated by the fact that among culprit lesions classified as having TIMI grade 2 flow, the CTFC was <30 (flow was normal or only modestly delayed) in 22 (28.6%) of the 77 culprit arteries assessed by the clinical center, which was significantly higher than the 4 (5.5%) of 73 culprit arteries assessed by the core laboratory (P=.003). Similarly, among culprit arteries classified as having TIMI grade 3 flow, the CTFC was >60 (ie, flow was actually slow) in 6 (4.4%) of the 134 culprit arteries assessed by the clinical center compared with none (0%) of the 148 culprits assessed by the core laboratory (P=.03).
CTFC in Nonculprit Arteries
In patients with acute MI, the mean CTFC in nonculprit arteries at 90 minutes after thrombolysis was 25.5±9.8 frames, which was significantly higher (reflecting slower flow) than that of normal arteries in the absence of acute MI (21.0±3.1 frames, P<.001) (Table 2⇑). By 18 to 36 hours, however, the CTFC of nonculprit arteries had improved significantly (−4.4±6.1 frames for pairs of films, P=.007) to nearly that of normal vessels in the absence of acute MI (21.7±7.1 versus 21.0±3.1 frames, P=NS).
CTFC in Culprit Arteries
The CTFC for culprit arteries at 90 minutes after thrombolysis was unimodally distributed with a single peak. Two distinct subpopulations of “slow” (TIMI grade 2 flow) and “fast” flow (TIMI grade 3 flow) did not exist (Fig 6⇓). The distribution was not normal, and therefore nonparametric tests were used in the statistical analyses of these data. The cumulative distribution function of the CTFC is shown in Fig 7⇓, which displays the proportion of the population on the y axis with a CTFC less than the value shown on the x axis. As can be seen from this plot, 32.6% of patients achieved a CTFC ≤27 frames, a value that can be taken as representative of the upper boundary of the CTFC for normal flow, as it was the upper bound of the 95% CI for the CTFC observed in the absence of acute MI (21.0+1.96×SD of 3.1). The mean 90-minute CTFC (read by the angiographic core laboratory) was 30.2±9.3 frames for TIMI grade 3 flow and 57.5±23.2 frames for TIMI grade 2 flow (P<.001).
The mean CTFC in all patent culprit arteries was 39.2±20.0 frames at 90 minutes, which improved significantly (−6.5±17.9 frames for all paired values, P<.001) to 31.7±12.9 frames by 18 to 36 hours (Table 2⇑). The magnitude of improvement in the CTFC over the course of the first day did not differ significantly between nonculprit (−4.4±6.1 frames) and culprit arteries (−6.5±17.9 frames, P=NS). However, whereas the CTFC had returned to normal in nonculprit arteries by 18 to 36 hours (21.7±7.1 versus 21.0±3.1 frames, P=NS), it remained persistently slower in culprit arteries (31.7±12.9 versus 21.0±3.1 frames, P<.001).
Relationship Between CTFC and Change in Lumen Diameter of Culprit Arteries Over Time
In the 33 patients in whom it could be assessed at all four time points (60, 75, and 90 minutes and 18 to 36 hours after thrombolysis), the CTFC decreased steadily over time, reflecting more rapid and improved dye transit (P=.005 for the linear trend over time) (Fig 8⇓). The CTFC improved by 18.5% between 60 and 90 minutes after thrombolysis (P=.008) and by an additional 14.8% by 18 to 36 hours (P=.001).
The minimum diameter of culprit arteries increased from 0.86±0.40 mm at 90 minutes to 1.09±0.49 mm at 18 to 36 hours (n=246, P<.001), and percent diameter stenosis improved from 73.8±9.9% at 90 minutes to 66.7±13.2% at 18 to 36 hours (n=246, P<.001). The fact that the state of vasomotor tone was reproduced at the time of initial and follow-up studies is confirmed by the observation that the 90-minute diameter of the normal reference segments (3.37±0.92 mm, n=246) was identical to that at the time of repeat angiography at 18 to 36 hours (3.37±0.95 mm, n=246, P=.97). There was no correlation between the improvement in CTFC between 90 minutes and 18 to 36 hours and the change in either the minimum lumen diameter (mechanical interventions excluded) (r=−.045, P=.59), the percent diameter stenosis (r=.097, P=.23), or the normal reference segment diameter (r=.10, P=.22).
Relationship Between CTFC and Infarct-Artery Location
At 90 minutes after thrombolysis, the mean CTFC for LAD culprit lesions (43.8±22.6 frames) was significantly higher (ie, slower flow) than for the RCA (37.2±19.3 frames, P=.029) and the circumflex artery (33.7±9.0 frames, P=.034) (Table 2⇑). Similarly, in nonculprit arteries, flow in the LAD was slower at 90 minutes after thrombolysis than in the other two locations combined (30.6±11.5 versus 23.1±7.9 frames, P=.001). In contrast, by 18 to 36 hours, the mean CTFC was similar among the three culprit artery locations (31.7±14.5 for LAD, 32.5±9.6 for LCx, and 31.6±12.8 for RCA, P=NS). There was a trend for the CTFC to improve more in LAD culprit arteries than in the two other locations combined (P=.098).
There were no differences in systolic blood pressure, time to treatment, reference segment diameter, minimum diameter, or percent stenosis for the three culprit arteries. Patients with an LAD culprit had larger integrated CK leaks (213.8±175.2 U) compared with RCA culprit arteries (131.6±83.8 U, two-way P<.001), and these CK leaks tended to be larger than those involving the LCx (173.2±106.4 U, P=.164).
Correlation of CTFC With Angiographic, Ventriculographic, and Enzymatic Outcomes
The 90-minute CTFC was only weakly correlated with 90-minute minimal lumen diameter (r=−.16, P=.016), percent diameter stenosis (r=.19, P=.006), and rate of rise of CK over the first 24 hours (r=.18, P=.009) and was inversely, weakly correlated with the predischarge radionuclide ventriculogram left ventricular ejection fraction (r=−.21, P=.002). There was a trend toward a correlation of the CTFC with the integrated CK over the first 24 hours (r=.13, P=.07). There was no correlation between the CTFC and the normal reference artery segment diameter or the elapsed time to treatment. There was no significant difference in the CTFC for patients with versus those without collaterals to the infarct-related artery.
Potential Pitfalls in the Conventional TIMI Flow-Grade System
The conventional TIMI flow-grade system1 has been widely used to compare thrombolytic agent efficacy and to stratify patients according to risk, but it has limitations. Whereas previous studies2 3 4 5 6 7 associated TIMI grade 2 flow after thrombolysis with suboptimal clinical outcomes, the present study demonstrates that there is a poor rate of agreement between an angiographic core laboratory and clinical centers in the assessment of TIMI grade 2 flow, which may limit the broad clinical applicability of this measure. Although the rate of agreement in assessment of the CTFC between the angiographic core laboratory and clinical centers remains to be determined, the variability at the core laboratory between two consecutive injections appears low (<5 frames, coefficient of variation=9.0%). Further evidence of the reproducibility and robust nature of this end point is the observation that the CTFC in one data set of normal arteries in the noninfarct setting (21.0±3.1) was nearly identical to the CTFC of a second independent data set of nonculprit arteries (21.7±7.1) 1 day after thrombolysis (Table 2⇑). Despite all the potential factors that introduce variability into the CTFC (variability in the distance to the landmark between patients, variability in rate of injection, intraobserver variability in counting frames, variability in catheter size, variability in the degree of engagement, variability in heart rate, variability in the phase of the cardiac cycle injected, etc) the SD of the CTFC in 78 normal arteries remained very low, at only 3.1 frames. Flow also has been estimated angiographically in the past by use of the mean contrast-appearance time, which involved calculating the time between peaks in contrast intensity at two points separated by known distances along the length of the artery. Unfortunately, this radiographic technique was limited by technical factors that confounded the videodensitometric analysis of angiograms, such as adhesion of the dye to the wall of the artery, scatter, veiling glare, variability in mask-mode subtraction, and motion artifact from patient movement.14 15
Even if excellent concordance could be demonstrated within or between experienced angiographic core laboratories in the assessment of TIMI grade flow, several major limitations remain when this categorical method is used to assess angiographic success after thrombolysis. First, the present study shows that coronary flow after thrombolysis is unimodally distributed (Fig 6⇑). Distinct subpopulations of patients with either slow (TIMI 2) or fast (TIMI 3) flow do not exist, and a categorical classification of coronary flow is, at best, arbitrary. Second, even if TIMI flow grading is proved to be reproducible, it may be inaccurate or may misclassify flow as a result of the use of a flawed gold standard to gauge “normal” flow. The original definition of TIMI grade 3 flow requires that flow into or clearance of contrast material from the involved bed be “as rapid as clearance from an uninvolved bed in the same vessel or the opposite artery.”1 However, the present study demonstrates that the velocity of contrast in an uninvolved bed may not be normal and is actually 21% slower (ie, the frame count is higher, 25.5 versus 21.0 frames, P<.001). This observation of reduced basal flow in nonculprit vessels at 90 minutes after thrombolysis extends the findings of Uren et al,16 who showed that at 1 week after MI, the vasodilatory response of nonculprit arteries remains reduced, which has been attributed to an abnormality in resistance vessel function. The fact that by 1 day after thrombolysis, these uninvolved arteries do eventually achieve flow virtually identical to that of arteries in the absence of acute MI (21.7 versus 21.0 frames) indicates that the delayed flow at 90 minutes was not an artifact of selecting a subpopulation of vessels with slower flow for analysis. In contrast to nonculprit vessels, flow in culprit vessels remained persistently reduced by 32% at 1 day after thrombolysis, which approximates the 26% flow reduction documented at 1 week after MI by positron emission tomography scanning.16 This persistent reduction in flow (which, paradoxically, is frequently classified as normal TIMI grade 3 flow) may be due either to the residual stenosis or to reduced oxygen consumption secondary to diminished contractility in the infarct zone. Changes in epicardial vasomotor tone do not appear to be responsible for these changes in frame counts over time, as the diameters of the normal reference segments were identical at 90 minutes and 18 to 36 hours (3.37 mm for both, P=.97).
An additional problem with the conventional TIMI flow-grading system is that uninvolved arteries that are used as the gold standard to classify the TIMI flow grade in the culprit artery may not all be slowed to the same degree, depending on their location. Flow in nonculprit LAD arteries was disproportionately slowed by 36% compared with that in uninvolved circumflex arteries, which confounds the classification of conventional TIMI flow grades. Although the present study shows that the CTFC in LAD culprits is, on the whole, higher (reflecting slower flow) than that in other locations at 90 minutes after successful thrombolysis, the CTFC for TIMI grade 3 flow was actually 32% lower for LAD culprits compared with the circumflex artery (25.7 versus 34.0 frames). This paradox is explained by the fact that TIMI grade 3 flow in culprit LAD arteries was gauged against faster flow in nonculprit circumflex arteries (22.5 frames), and consequently, few LAD culprits (26.2%) achieved a velocity rapid enough to be classified as achieving TIMI grade 3 flow. In contrast, flow in circumflex culprits was graded against the 36% slower flow in nonculprit LADs (30.6 frames), and consequently, the vast majority of circumflex arteries (92%) were classified as achieving TIMI grade 3 flow. Thus, the conventional notion that flow in uninvolved arteries is normal may be erroneous and may lead to the misclassification of TIMI flow grades. In the RCA, no adjacent comparative normal artery is even present. The problem of visual flow-grade assessment is further compounded by the fact that international cinefilms are filmed at a wide variety of speeds (12.5, 15, 25, 30, 50, or 60 frames/s). Another limitation of the conventional TIMI flow-grading system is that observers who grade flow might be biased by their concurrent assessment of the ejection fraction of the patient (which has been linked to clinical outcomes), and an objective frame-counting method might reduce this potential for bias.
The CTFC could facilitate the standardization of TIMI flow grades and flow assessment. However, devising a valid definition of normal flow is complicated. The mean CTFC for culprit arteries with normal TIMI grade 3 flow (30.2 frames) was significantly higher than for nonculprit arteries (25.5 frames, P<.001) and was higher yet (43.8%) than for normal arteries in the absence of acute MI (21.0 frames, P<.001). Thus, TIMI grade 3 flow in culprit arteries is not truly representative of normal flow. In our view, a valid threshold for defining normal flow would be to use the upper bound of the 95% CI for arteries that have normal flow in the absence of acute MI (27 frames with predominantly 6F and 7F catheters). By this definition, only one third (32.6%) of patients with a patent artery in the TIMI 4 trial actually achieved flow within the range observed in normal arteries (Fig 7⇑).
The use of either an objective classification system or the CTFC itself should facilitate the comparison of angiographic end points between trials. In addition, such an objective definition could improve on the large intraobserver variability experienced by submitting clinical centers, and such a comparison is currently under way. Use of the CTFC might permit meaningful analysis of the vast amount of angiographic data that are gathered by clinical centers and currently are not analyzed because a single core laboratory could not process tens of thousands of films from large clinical trials.1 2 This method is simple and should have broad applicability because it can be measured by any investigator with the frame counter that is present on most cineviewers. We have found that variability can be encountered in selecting the first frame for analysis, and care should be taken to ensure that all three criteria are fulfilled for the selection of the first frame as set forth in “Methods.”
The use of a continuous variable such as the CTFC for comparison of angiographic outcomes might be superior to the use of a categorical variable such as TIMI flow grade in terms of statistical power and sensitivity. As newer thrombolytic agents reportedly achieve a higher incidence of TIMI grade 3 flow, a categorical scale may fail to distinguish their efficacies, because there is a range of dye velocities that constitute TIMI grade 3 flow. Even if two agents result in the same proportion of TIMI grade 3 flow, there may be a difference in dye velocity between the two agents when analyzed as a continuous variable with the CTFC. For instance, two future thrombolytic agents may both achieve TIMI grade 3 flow in 75% of patients, but one agent may achieve a mean CTFC of 30 frames and the other a mean CTFC of 40 frames.
Differences in Flow Among the Three Coronary Arteries
The unadjusted TIMI frame count for LAD culprits was corrected to account for its longer length by dividing by 1.7, the ratio of the LAD TIMI frame count to the mean of normal LCx and RCAs in the absence of acute MI. This ratio is consistent with the mean ratio of 1.55 predicted by use of three-dimensional vector algebra devised by Dodge et al11 to calculate the distance to the TIMI landmarks in the normal-sized human heart. In addition to controlling for differences in artery length and minimizing the effect of discrepancies in the proportion of LAD culprit arteries between agents and between trials, this correction improves the power of this end point by reducing the SD of the TIMI frame count among patients with culprit arteries in the LAD versus other locations.
TIMI grade 3 flow more frequently was achieved in patients with an infarct-related artery other than the LAD in the present trial (74% versus 26%, P<.001), a finding that was also observed in the TEAM-2 study (61% versus 37%, P=.06).6 Further objective documentation of this observation is the fact that the CTFC (ie, the frame count already corrected for the difference in artery length) was higher (reflecting slower flow) in LAD culprit arteries compared with the other two arteries. Furthermore, flow in nonculprit LADs was also delayed compared with the other two arteries. This apparent overrepresentation of LAD culprit arteries with TIMI grade 2 flow and the preponderance of RCA culprit arteries with TIMI grade 3 flow has major implications when the clinical outcomes of the various flow grades are compared. In several recent analyses of angiographic data, it was demonstrated that patients with TIMI grade 2 flow had poorer outcomes than patients with TIMI grade 3 flow. However, data pertaining to the location of the infarct artery were neither presented nor corrected for in these analyses of outcomes.2 17 18 Thus, it would be appropriate that any analysis comparing the clinical, enzymatic, ventriculographic, or electrocardiographic outcomes of the various TIMI flow grades should correct for the large potential imbalances in infarct-artery location.
Despite differences in the CTFC in the three coronary arteries, there were no differences in their lumen dimensions, and there was no correlation between the CTFC and hemodynamic variables. This discrepancy in flow was transient and resolved by 18 to 36 hours. A question that therefore arises is whether slowed flow causes larger infarcts in the LAD or, conversely, if larger infarctions in the LAD cause slower flow. Seiler et al19 showed that the length of an artery is proportional to the myocardial mass subtended by the artery, and Kloner et al20 21 showed in turn that the myocardial mass injured is proportional to the magnitude of no reflow. Thus, the higher CTFC of LADs and the preponderance of LAD culprits among patients with TIMI grade 2 flow may be related to more extensive necrosis (LAD CK leaks were larger) as a result of the large myocardial mass subtended by this 1.7-times-longer coronary artery.
Although both the epicardial lesion and the microvasculature are determinants of flow after thrombolysis, the role of the microvasculature is supported by the following observations: (1) flow is slowed in uninvolved nonculprit arteries at 90 minutes after thrombolysis but returns to normal by 1 day after thrombolysis; (2) in culprit arteries, there is no relationship between the improvement in the CTFC and the change in minimal lumen diameter over the first day after thrombolysis (with no documented change in vasomotor tone); and (3) the improvement in flow over the course of the first day after thrombolysis does not differ significantly between nonculprit and culprit arteries. These findings implicating the microvasculature are consistent with the myocardial contrast echocardiography studies of Ito et al,22 in which no relationship was observed between epicardial stenosis severity and the incidence of myocardial no reflow after successful thrombolysis. That group also demonstrated that intracoronary injections of verapamil in the setting of acute MI may improve microvascular perfusion and left ventricular function.23 We again emphasize that in the current study, TIMI grade 3 flow (30.2 frames) was slower than flow in the absence of acute MI (21.0 frames), and the relative contribution of microvascular and epicardial resistance as well as reduced oxygen consumption to this delay in TIMI grade 3 flow remains to be determined.
Further studies are required to prospectively validate this new angiographic end point. The relationship of the CTFC to major clinical end points, such as mortality, remains to be determined. The correlation between angiographic core laboratories in the assessment of this end point will also need to be determined. Circumflex and saphenous vein graft lesions are underrepresented in thrombolytic trials, and additional data are required to characterize and fully evaluate the CTFC of these arteries. Currently, we use the same distal native landmarks to characterize flow in bypass grafts. Culprit lesions in the distal posterolateral branch or the distal posterior descending artery may occasionally lie distal to the TIMI landmark. Infrequently, there may only be a short distance of delayed flow beyond a distal lesion located before the TIMI landmark. Further prospective studies are needed to confirm that flow in LAD culprit arteries is slower than in other locations as well as to confirm the observation that flow in nonculprit arteries improves slightly over time.
Usually, the rate at which images are captured to cinefilm is the same rate at which images are captured into the video format in the cardiac catheterization laboratory. This rate is generally 30 frames/s (SONY, Phillips, and General Electric Inc). The number of frames can be counted on line in the cardiac catheterization laboratory or on a VCR by simply advancing the video one frame at a time. The frame acquisition and display rate should be confirmed with the manufacturer of the laboratory equipment. Alternatively, for videotapes, a studio production recorder can be used to record the length of time in fractions of a second to the landmark. It should be confirmed that 30 single frames elapse per second. The time (in seconds) can then be multiplied by 30 frames/s to convert to frames. In some centers, cinefilming may instead be performed at 12.5, 15, 25, 50, or 60 frames/s. Therefore, a correction must be made for this difference by multiplying the TIMI frame count observed in these films by a factor of 30 divided by the actual number of frames filmed per second. The rate of frame acquisition must be communicated to the angiographic core laboratory.
The CTFC will vary depending on the length of the artery. Fortunately, this variability appears to be fairly low, as the SD of the CTFC among normal arteries was only 3.1 frames. Furthermore, if comparisons of the CTFC are made over time, then length is controlled for within the same artery. It must be emphasized that larger numbers of patients need to be studied to fully ascertain the influence of catheter size on the CTFC, the effects of which may have been overshadowed by a variety of confounding variables in acute MI. Predominantly 6F and 7F catheters were used in the present study, and more comparative data in the absence of acute MI are needed for the use of 8F and 9F catheters in restenosis trials, for instance. The mean catheter size or the distribution of catheter sizes should be confirmed to be the same among treatment groups and should be reported so that comparisons across trials can be evaluated for their applicability. Although the rate of entry of dye into the coronary tree may be affected by the force of the injection, the rate of egress or washout of dye may be more independent of the rate of injection and warrants further investigation. The view that is optimal for frame counting may not be the view that is optimal for quantitative angiography.
The TIMI frame count could not be measured in 9% of consecutive acute MI patients, and a technique of more rapid panning may be required for interventional trials in which the flow is more brisk and the magnification is often higher in the 5-in mode. If panning is insufficient, we advocate that an injection be obtained of the culprit vessel on 9-in mode so that the frame count can be ascertained in the vast majority of patients.
In contrast to the conventional TIMI flow-grade system, the CTFC is quantitative rather than qualitative, is a continuous rather than a categorical variable, and is objective, reproducible, and sensitive to flow changes. This simple index of coronary flow allows calibration or standardization of flow grading and should facilitate comparisons of flow data between angiographic trials. Flow in nonculprit arteries routinely used to grade TIMI flow cannot be presumed to be normal, as demonstrated by the 21% reduction in flow in these arteries at 90 minutes after thrombolysis. Despite high rates of TIMI grade 3 flow reported in the literature, only a third of patients with an open artery actually achieve flow that is truly within the normal range (CTFC ≤27). Outcome analyses of TIMI flow grades should correct for imbalances in lesion location. The current study supports the idea that the disordered microvasculature tone plays a role in flow delays immediately after thrombolysis.
Selected Abbreviations and Acronyms
|APSAC||=||anisoylated plasminogen streptokinase activator complex|
|CTFC||=||corrected TIMI frame count|
|LAD||=||left anterior descending artery|
|LCx||=||left circumflex artery|
|RCA||=||right coronary artery|
|rTPA||=||recombinant tissue plasminogen activator|
|TIMI||=||Thrombolysis in Myocardial Infarction trial|
This study was supported by a grant from Smith Kline Beecham, Philadelphia, Pa. Additional support was provided by Genentech, Inc, South San Francisco, Calif.
Reprint requests to C. Michael Gibson, MS, MD, Director, Interventional Cardiology, West Roxbury Veteran’s Administration Hospital, 1400 VFW Parkway, West Roxbury, MA 02132.
- Received September 11, 1995.
- Revision received November 13, 1995.
- Accepted November 19, 1995.
- Copyright © 1996 by American Heart Association
Vogt A, von Essen R, Tebbe U, Feuerer W, Appel KF, Neuhaus KL. Impact of early perfusion status of the infarct-related artery on short-term mortality after thrombolysis for acute myocardial infarction: retrospective analysis of four German multicenter studies. J Am Coll Cardiol. 1993;21:1391-1395.
Karagounis L, Sorensen SG, Menlove RL, Moreno F, Anderson JL, for the TEAM-2 Investigators. Does thrombolysis in myocardial infarction (TIMI) perfusion grade 2 represent a mostly patent artery or a mostly occluded artery? Enzymatic and electrocardiographic evidence from the TEAM-2 study. J Am Coll Cardiol. 1992;19:1-10.
Anderson JL, Karagounis LA, Becker LC, Sorensen SG, Menlove RL, for the TEAM-3 Investigators. TIMI perfusion grade 3 but not grade 2 results in improved outcome after thrombolysis for myocardial infarction: ventriculographic, enzymatic, and electrocardiographic evidence from the TEAM-3 study. Circulation. 1993;87:1829-1839.
Cannon CP, McCabe CH, Diver DJ, Herson S, Greene RM, Shah PK, Sequeira RF, Leya F, Kirshenbaum JM, Magorien RD, Palmeri S, Davis V, Gibson CM, Poole WK, Braunwald E, and the TIMI 4 Investigators. Comparison of front-loaded recombinant tissue-type plasminogen activator, anistreplase and combination thrombolytic therapy for acute myocardial infarction: results of the Thrombolysis in Myocardial Infarction (TIMI) 4 trial. J Am Coll Cardiol. 1994;24:1602-1610.
Gibson CM, Cannon CP, Piana RN, Maher KA, Davis SF, Breall JA, Davis V, Diver DJ, for the TIMI 4 Study Group. Relationship of coronary flow to myocardial infarction size: two simple new methods to sub-classify TIMI flow grades. Circulation. 1992;86(suppl I):I-453. Abstract.
Dodge JT, Brown BG, Bolson EL, Dodge HT. Intrathoracic spatial location of specified coronary segments on the normal human heart: applications in quantitative arteriography, assessment of regional risk and contraction, and anatomic display. Circulation. 1988;78:1167-1180.
SAS/STAT Users Guide: Statistics (Version 6). 4th ed. Cary, NC: SAS Institutes Inc; 1990:851-889, 891-949, 1071-1092, 1195-1200, 1633-1636.
STATA Reference Manual. 4th ed. Santa Monica, Calif: Computing Resource Center; 1989.
Hodgson JM, LeGrand V, Bates ER, Mancini GBJ, Aueron FM, O’Neill WW, Simon SB, Beauman GJ, LeFree MT, Vogel RA. Validation in dogs of a rapid digital angiographic technique to measure relative coronary blood flow during routine cardiac catheterization. Am J Cardiol. 1985;55:188-193.
Ordoubadi FF, Huehns TY, Mohammad AA, Beatt KJ. TIMI grade 2 flow is not equivalent to TIMI 3: implications for the use of thrombolytic therapy—a meta-analysis of the trials. J Am Coll Cardiol. 1995;(suppl):401A. Abstract.
Kloner RA, Ganote CE, Jennings RB. The no-reflow phenomenon after temporary coronary occlusion in the dog. J Clin Invest. 1974;54:1496-1508.
Kloner RA, Alker KJ. The effect of streptokinase on intramyocardial hemorrhage, infarct size, and the no-reflow phenomenon during coronary reperfusion. Circulation. 1984;70:513-521.
Ito H, Tomooka T, Sakai N, Yu H, Higashino Y, Fujii K, Masuyama T, Kitabatake A, Minamino T. Lack of myocardial perfusion immediately after successful thrombolysis: a predictor of poor recovery of left ventricular function in anterior myocardial infarction. Circulation. 1992;85:1699-1705.
Taniyama Y, Ito H, Iwakura K, Takiuchi S, Maruyama A, Fujii K. Effects of intracoronary injection of verapamil and nicorandil on microcirculation and left ventricular function in patients with reperfused myocardial infarction. Circulation. 1995;92(suppl I):I-77. Abstract.