We conducted a retrospective observational study of a historic cohort of patients referred to a cardiac catheterization laboratory for coronary angiography for suspected coronary disease between January 1, 2008 and May 31, 2012. The patients were evaluated for revascularization by measuring the FFR across an intermediate coronary lesion (40%-70% diameter stenosis by visual estimation) using a pressure wire. We excluded patients with lesions > 20% in a segment distal or proximal to the target lesion. The FFR was not measured in vessels that perfused akinetic or previously infarcted territory. In patients with acute coronary syndrome, the FFR was only measured in nonculprit vessels.

All procedures were performed according to the usual protocol of the center conducting the study. After the decision was made to measure the FFR, the diagnostic catheter used for angiography was replaced by a 6-Fr guide catheter. This catheter was used to repeat the projections providing the best visualization of the lesions, with greater visual stenosis and without overlapping branches or loss of length because of curvature. All patients received 100 IU/kg intravenous sodium heparin before the procedure if not previously administered. Functional evaluation was performed with a 0.014-inch intracoronary pressure wire (Pressure Wire, Certus or Airis, Radi Medical Systems; Uppsala, Sweden, or PrimeWire Prestige Pressure Guide Wire, Volcano Corp.; San Diego, California, United States). The pressure wire was calibrated externally and then advanced to the distal end of the guide catheter while equalizing the pressures according to the system used to measure the FFR. After administration of 200 μg to 300 μg intracoronary nitroglycerin, the guide was advanced until the sensor was at least 20 mm distal to the lesion. We followed the standard procedure used in our hospital to obtain the FFR by administering 300 μg to 1200 μg intracoronary adenosine, while taking particular care to avoid wedging the catheter in the coronary ostium after bolus injection of the drug. The beat-to-beat ratio of the mean aortic pressure at the end of the guide catheter and the pressure distal to the lesion, obtained via the pressure wire under maximum hyperemia, were used to measure the FFR. We measured the FFR at least 3 times and used the lowest measurement. We successively administered 300 μg, 600 μg and 1200 μg intracoronary adenosine whenever the previous dose failed to produce a period of asystole ≥ 6 s.

The decision to revascularize was left to the operator's discretion based on the data obtained in the angiographic and functional study.

In our hospital, the routine method to obtain the FFR includes obtaining at least 1 projection that provides the best visualization of the lesion using the guide catheter, after the administration of intracoronary nitroglycerin. The diagnostic angiographic sequences of each procedure were separated from those obtained during the intervention (when applicable). The observers were only provided with diagnostic images and were blinded to the result of the FFR study when performing the digital quantification of lesion stenosis. Data were collected on the following variables: severe calcification (multiple opacification visible in more than 1 projection covering the entire vessel lumen at the site of the lesion); bifurcation (presence of a > 15-mm side branch originating at the site of the lesion); angulation > 45° (target lesion in a segment with angulation > 45°); ostial location (lesion at the origin of the vessel in the aorta); perfused myocardial territory (Duke jeopardy score18,19); and location of the lesion in the stent.

Digital quantification was performed using the QAngio XA version 7.1.43.0 postprocessing software package (Medis Medical Imaging Systems; Leiden, The Netherlands).

The lesions were analyzed by 2 experienced interventional cardiologists (more than 1000 coronary interventions using a pressure wire/measuring the FRR). One of them analyzed the lesions twice (OBS1A and OBS1B, with a 4-month interval between assessments), and the other analyzed them once (OBS2). They predicted whether the result of the pressure wire study was positive (FFR ≤ 0.80). The 2 observers were blinded to the previous predictions, the assessments made by the other, and the FFR results.

Based on their previous experience and published scientific evidence, the 2 observers took into account not only the degree of stenosis when making their predictions but also other parameters that have been associated with the FFR.16 In cases of disagreement, the mode of the 3 predictions was used. Overestimation was defined by a predicted FFR ≤ 0.80 vs a measured FFR > 0.80. Underestimation was defined by a predicted FFR > 0.80 vs a measured FFR ≤ 0.8. Disagreement was defined by the underestimation or overestimation of the FFR. We analyzed the factors associated with disagreement, overestimation, or underestimation.

All quantitative variables are expressed as mean±standard deviation and were compared using the Student t test. Qualitative variables are expressed as absolute values (percentages) and were compared using the chi-squared test. Logistic regression models were used to analyze the angiographic and clinical variables associated with an observed FFR ≤ 0.80, an estimated FFR ≤ 0.80, and disagreement (overestimation and underestimation) between the predicted and observed functional significance of the lesions. A P-value of <.5 was used as a cutoff for statistical significance. The Hosmer-Lemeshow test and the area under the ROC (receiver operating characteristic) curve were used to assess the model's goodness-of-fit and predictive power, respectively. All analyses were conducted using the SPSS 20.0 statistical package for Windows (SPSS Inc.; Chicago, Illinois, United States). Although this study compared 2 diagnostic tests, rather than assessing the reliability of 1 diagnostic test, the recommendations of the STARD initiative in data presentation were followed as closely as possible.20

The results of this study are consistent with those obtained in previous studies regarding low intraobserver and interobserver agreement in the visual assessment of the severity of intermediate coronary lesions. Matching values were higher than those of other studies based solely on the degree of stenosis.6,8 Despite being members of the same intervention team and using similar criteria for the estimations, κ<0.6 was obtained, which indicates the low reliability of the visual assessment of lesions.

Invasive measurement of FFR addresses characteristics of the coronary lesion apart from the degree of stenosis alone. Frequent use of the FFR in the cardiac catheterization laboratory may have an “instructive” effect that would lead the interventionist to modify the assessment of the significance of the lesion by taking into account aspects such as the lesion length, disease of the distal bed, or the size of the myocardium distal to the lesion, in addition to its degree of stenosis. In a recent study, Park et al.9 used the degree of angiographic stenosis (52% diameter stenosis) as the only classification parameter. They found an agreement of 66% with a sensitivity of 66%, a specificity of 67%, and positive and negative predictive values of 48% and 81%, respectively9. We used multiple clinical and angiographic parameters to visually estimate the functional significance of the lesion and observed a slightly higher agreement (70%), lower sensitivity, and greater specificity.

The percentage of lesions with an FFR > 0.80 considered significant by visual estimation was significantly lower than that observed by Park et al. (16% vs 57%). In contrast, a much higher percentage of lesions with an FFR ≤ 0.80 were visually classified as functionally nonsignificant (63% vs 16%). The frequent use of the FFR in intermediate lesions probably leads to the application of a “functional approach” to their assessment, without invasive measurement, which would lead to a more conservative assessment when considering a lesion functionally significant.

Together with the inability of angiography to predict the functional significance of coronary lesions, our study shows that certain angiographic characteristics of the lesion may be associated with underestimating or overestimating the significance of the stenosis. If these angiographic characteristics are present in an angiographically intermediate stenosis, the probability of error when decisions are made on whether to revascularize without a functional study can be even greater than in the general population.

In our study, moderate in-stent restenosis (40%-70% stenosis) was the factor most strongly associated with overestimating the significance of the lesion. The presence of bare-metal stents makes it difficult to determine the edges of the vessel lumen, which can lead to underestimation of its diameter and overestimation of the degree of restenosis. Although previous studies have demonstrated the limitations of angiography in assessing in-stent restenosis,21,22 recent clinical trials have continued to use the degree of angiographic stenosis to determine restenosis or to decide whether to revascularize the treated lesion.23–25 The use of FFR for the functional assessment of the lesion during follow-up in these studies could have led to fewer restenosis events being assigned to patients undergoing percutaneous coronary intervention.

The location of the lesion in an artery other than the anterior descending artery was one of the variables associated with overestimating the functional significance of the lesion. In the multivariable analysis, an adjustment was made for the myocardial territory perfused by the artery with the lesion. This adjustment may have been insufficient, such that the observed overestimation could have been due to a smaller myocardial area in lesions in locations other than the anterior descending artery with a greater FFR value for the same diameter stenosis value. This possibility is consistent with the finding of an association between a larger myocardial territory distal to the lesion and underestimating the functional significance of the lesion. Although the researchers were aware of the effect of perfused myocardial territory on the FFR value for a given lesion, the angiographic characteristics of the lesion continued to take precedence over the other characteristics when estimates were made on its functional significance. The location of the lesion in a bifurcation was also associated with the functional overestimation of the stenosis. Several studies have demonstrated the difficulty of assessing these lesions angiographically and have shown that less functional significance is found when the FFR is used than might be expected by the degree of stenosis observed both before and after treatment of the lesion.26,27

Male sex and severe calcification were associated with underestimation of the functional significance of lesions. As the researchers were blinded to the sex of the patients when assessing the lesions, they could not have taken this variable into account in their estimates. Our findings are in line with those of several studies that have found an association between female sex and a greater FFR value for the same degree of stenosis in men and women.28–30

Calcification hampers the visual and automated angiographic assessment of stenosis. In this study, in contrast to in-stent restenosis, calcification was associated with underestimating the significance of the lesion. Increased radiodensity due to increased vessel thickness can contribute to overestimating lumen diameter, which would lead to systematic errors in visual assessment. Moreover, coronary calcification can be a marker of more extensive and diffuse disease and could lead to diseased coronary segments being used as a reference, and thus lead to underestimation of the degree of stenosis or lesion length.

This study was conducted at a single hospital. Although the researchers who visually assessed the lesions were blind to the results of the FFR study and to each other's estimates, local factors may have influenced estimation of the functional significance of the lesions. The generalizability of our findings is supported by the low interobserver agreement found in this study, which is similar to that found other published studies, and agreement with the literature regarding the variables associated with the predictions. The use of the mode in the 3 estimates gave more weight to the observations of OBS1, who presented slightly worse agreement.