- Email: [email protected]
Reproducibility of Coronary Optical Coherence Tomography for Lumen and Length Measurements in Humans (The CLI-VAR [Centro per la Lotta contro l'Infarto-VARiability] Study)
Reproducibility of Coronary Optical Coherence Tomography for Lumen and Length Measurements in Humans (The CLI-VAR [Centro per la Lotta contro l’Infarto-VARiability] Study) Silvio Fedele, MDa, Giuseppe Biondi-Zoccai, MDa, Piotr Kwiatkowski, MDa, Luca Di Vito, MDa, Michele Occhipinti, MDa, Alberto Cremonesi, MDb, Mario Albertucci, MDa, Laura Materia, PharmDa, Giulia Paoletti, RTa, and Francesco Prati, MDa,c,* Frequency-domain optical coherence tomography (FD-OCT) is becoming a useful diagnostic tool for coronary imaging for quantitative coronary analysis. Second-generation FD-OCT produces detailed coronary lumen images. However, the reproducibility of coronary measurements using FD-OCT in humans has not been thoroughly explored. Our goal was to determine the intraobserver, interobserver, and interpullback reproducibility of the in vivo FD-OCT measurements of the lumen area and/or lesion length. Twenty-five patients undergoing coronary angioplasty were included. In all subjects, FD-OCT pullbacks (20 mm/s) were acquired twice from the same coronary segment different from the target lesion, at an interval of 5 minutes, with no other intervention. A total of 9,396 cross-sectional lumen area frames and the relative coronary lesion length of each pullback were analyzed off-line with dedicated software by 2 independent expert readers (A and B). We compared the lumen area and length measurements as follows: pullback 1, read by reader A twice at an interval of 7 days (intraobserver analysis); pullback 1, independently read by readers A and B (interobserver comparison); and pullback 1 versus pullback 2, read by reader A (interpullback comparison). The per-segment and per-frame analyses showed very high and significant correlation coefficients for the interobserver, intraobserver, and interpullback comparisons for the lumen area and lesion length (R >0.95 and p <0.001 in all cases). Accordingly, the Bland-Altman estimates of bias showed nonsignificant differences in the interobserver, intraobserver, and interpullback comparisons at all levels, with average biases never >0.150 mm2 for the lumen area or 0.200 mm for the lesion length. In conclusion, coronary imaging using FD-OCT showed excellent reproducibility, with low intraobserver, interobserver, and interpullback variability for both lumen area and lesion length measurements in humans. Thus, FD-OCT can be proposed for precise analysis in the catheterization laboratory to guide decision making and in clinical trials focusing on imaging end points. © 2012 Elsevier Inc. All rights reserved. (Am J Cardiol 2012;110:1106 –1112) Frequency-domain optical coherence tomography (FDOCT) is an ultrahigh-resolution intracoronary imaging technology able to provide detailed images of the coronary lumen.1,2 Second-generation FD-OCT introduced significant improvements, simplifying data acquisition by increasing the speed of pullbacks,3 and significantly reducing patient discomfort. Thus, FD-OCT is rapidly being adopted by catheterization laboratories worldwide for clinical use and research purposes.4 – 6 This technique has great potential to become an instrumental tool to address lumen and plaque changes in response to specific pharmacologic or interventional treatment. Although intravascular ultrasonography has been frequently used for plaque progression/regression studies, FD-OCT, given its superior resolution, could become complementary or an alternative to intravascular ula Centro per la Lotta contro l’Infarto Foundation, Rome, Italy; Sansavini Foundation, Cotignola, Italy; and cDivision of Cardiology, San Giovanni Hospital, Rome, Italy. Manuscript received March 4, 2012; revised manuscript received and accepted May 31, 2012. *Corresponding author: Tel: (⫹39) 0677208915; fax: (⫹39) 0645443331. E-mail address: [email protected] (F. Prati).
b
0002-9149/12/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjcard.2012.05.047
trasonography in such research settings. However, limited data are available on the reproducibility of such data between separate pullbacks and between or within operators.7–12 In particular, the variability between and within operators and/or pullbacks has not been formally tested. We hypothesized that FD-OCT will allow reproducible measurements of the coronary lumen area and lesion length. Thus, we conducted a prospective study to systematically determine the reproducibility of FD-OCT measurements for the coronary lumen area and lesion length in humans. Methods The present study was a prospective single-center study approved by the local review board. Twenty-five nonconsecutive patients scheduled to undergo angioplasty from April 2009 to June 2011 were selected for the study. Of the 25 patients, 6 had stable angina, 8 had non–ST-segment elevation coronary syndrome, and 11 had evidence of myocardial ischemia. The exclusion criteria were recent (⬍72 hours) acute myocardial infarction, hemodynamic instability, renal insufficiency (serum creatinine ⬎2.0 mg/dl), alwww.ajconline.org
Coronary Artery Disease/CLI-VAR Study
1107
Figure 1. Lumen area measurements at OCT by different readers, at different times, or during different pullbacks.
Figure 2. Segment length measurements at OCT during different pullbacks.
lergy to the contrast media, left main disease, ostial coronary artery lesion, bypass lesion, chronic total occlusion, and a tortuous vessel. Each of the selected patients underwent coronary angiography followed by the acquisition of 2 pullback images using FD-OCT with a minimum length of 18 mm (pullback 1, and, 5 minutes later, pullback 2). The optical coherence tomographic images were obtained in a coronary artery that was not found to be significantly diseased at angiography (diameter stenosis ⬍50% on visual estimation). All patients provided written informed consent.
FD-OCT is a high-resolution intracoronary imaging technique with a maximum spatial resolution of 10 to 15 m. The current generation C7 system (LightLab Imaging, Westford, Massachusetts) was used in the present study. The imaging catheters for FD-OCT were delivered over a 0.014-in. guidewire through ⱖ6F guiding catheters. For an effective clearing of blood from the imaging field, angiographic contrast medium was injected through the guiding catheter, as previously described,13 with image acquisition performed according to established methods.5,14
1108
The American Journal of Cardiology (www.ajconline.org)
In all patients, we performed a standard femoral catheterization approach. Unfractionated heparin was administered intravenously to maintain an activated clotting time ⬎300 seconds. After coronary angiography, we positioned a 6F guiding catheter into the coronary ostium and advanced a standard 0.014-in. guidewire into the coronary artery. The imaging catheter for FD-OCT was advanced into the coronary artery to the segment of interest after administration of 200 g intracoronary nitroglycerine. The blood was cleared by injection of iso-osmolar contrast (iodixanol 320, Visipaque, GE Healthcare, Dublin, Ireland) at 37°C with an injection pump (Acist, Bracco, Milan, Italy) through the guiding catheter. Two continuous motorized pullbacks (pullbacks 1 and 2) using FD-OCT were started as soon as the artery was cleared of blood. The images from each pullback were acquired at a speed of 20 mm/s from the same coronary segment until reaching the guiding catheter. The pullbacks were taken at an interval of 5 minutes without additional intervention between them. The coronary images acquired were saved on the OFDI system console and exported to a digital system for off-line analysis at an independent imaging core laboratory (Rome Heart Research, Rome, Italy). The analysis of the coronary images from FD-OCT was performed by 2 analysts (reader A and B) experienced in FD-OCT independently using a dedicated contour detection program (LightLab Imaging). First, we identified the coronary segment of interest in each pullback and verified the correspondence of the 2 pullbacks through side-by-side comparisons and using as reference landmarks the crosssectional views of the side branches. After identifying the coronary segment in each pullback, we calibrated the measurements using the 6F (2-mm) catheter. In each crosssectional coronary image, we calculated the lumen area using automatic contour tracing of the intima and the same detection program, with very occasional manual correction (Figure 1). The length of the coronary segment was calculated from the number of cross-sectional frames taken of that segment (Figure 2). To evaluate interobserver reproducibility, 2 independent readers (A and B) independently analyzed pullback 1. To determine intraobserver reproducibility, 1 reader (reader A) analyzed pullback 1 twice, with the second reading 1 week after the first. Finally, to assess interpullback reproducibility, 1 reader (reader A) analyzed pullbacks 1 and 2 during same work session. The analyses were conducted at the segmental (per-segment) level (thus including all frames obtained from the study subject) and frame (per-frame) level. Correlation and bivariate regression analysis were performed to determine the association between different readings using both Pearson’s correlation and Spearman’s rho tests. In addition, Bland-Altman methods were used to obtain the bias estimates (reported as the mean and 95% bootstrapped confidence interval stemming from 1,000 samples each) and accompanying plots. Assuming an 8- ⫾ 4-mm2 lumen area and 20- ⫾ 7-mm lesion length, 25 patients were deemed sufficient to achieve adequately small standard errors of the mean (0.8 mm2 and 1.3 mm, respectively) and suitably precise 95% confidence intervals (6.4 to 9.6 mm2 and 17.2 to 22.8 mm, respectively). Statistical significance was set at the 2-tailed 0.05 level, and p values unadjusted for multiplicity are
Table 1 Reproducibility analysis Variable
Bland-Altman Bias*
Lumen area (mm2) Per-segment analysis Interobserver 0.001 (⫺0.012, 0.009) Intraobserver 0.003 (⫺0.002, 0.009) Interpullback 0.150 (⫺0.371, 0.086) Per-frame analysis Interobserver 0.001 (⫺0.001, 0.002) Intraobserver 0.002 (0.001, 0.003) Interpullback ⫺0.091 (⫺0.139, ⫺0.040) Length (mm) Per-segment analysis Interpullback ⫺0.200 (⫺3.00, 2.00)
Regression Analysis† (R Value, p Value)
1.0, ⬍0.001 1.0, ⬍0.001 0.982, ⬍0.001 1.0, ⬍0.001 1.0, ⬍0.001 0.959, ⬍0.001 0.990, ⬍0.001
* Reported as mean (95% bootstrapped confidence interval stemming from 1,000 samples). † Similar results for statistical significance and magnitude were obtained with Spearman’s correlation test.
reported throughout. Analyses were performed with SPSS, version 19 (IBM, Armonk, New York). Results We were able to acquire 2 separate coronary pullbacks using FD-OCT from each of the 25 patients in the study. The image quality of the pullbacks was excellent, with artifacts in ⬍1% of the evaluated images. The manual correction of the automatic lumen area contour was performed in 25% of measurements from ⱖ1 reader. No imaging-related complications, such as coronary dissection, perforation, spasm, embolization, arrhythmia, or clinical adverse events, occurred. A total of 9,396 frames were included in the analysis, from which a longitudinal image of each patient coronary target lesion was reconstructed. The reproducibility of the coronary measurement analysis using FD-OCT in the present study was very high in all comparisons (Table 1). Per-segment and per-frame analysis of the lumen areas showed very high and significant correlation coefficients, using both parametric and nonparametric tests, for the interobserver, intraobserver and interpullback comparisons (R ⱖ0.95 and p ⬍0.001 for all cases; Figures 3 and 4). Accordingly, the Bland-Altman estimates of bias showed very small differences in the interobserver, intraobserver, and interpullback comparisons for the lumen area at all levels, with average biases never ⬎0.091 mm2 for the lumen area or 0.200 mm for the lesion length. As expected, the interobserver and intraobserver analyses for the lumen area showed even greater precision and correlation than the interpullback estimates. Specifically, the R values were invariably 1.0 for interobserver and intraobserver analyses compared to R values of 0.959 to 0.982 for the interpullback analyses. Accordingly, absolute BlandAltman estimates of bias ranged from 0.001 to 0.003 for the interobserver and intraobserver analyses compared to estimates of bias of 0.091 to 0.150 for the interpullback analyses. Finally, the interobserver and intraobserver analyses for lesion length were perfectly concordant and precise (R ⫽ 1.0, p ⬍0.001; Figure 5).
Coronary Artery Disease/CLI-VAR Study
1109
Figure 3. Reproducibility of per-segment measurements of lumen area at OCT (mm2). (A) Scatterplot and (B) Bland-Altman plot for interobserver reproducibility. (C) Scatterplot and (D) Bland-Altman plot for intraobserver reproducibility. (E) Scatterplot and (F) Bland-Altman plot for interpullback reproducibility.
Discussion The present study, which has originally and comprehensively reported on the variability of measurements using FD-OCT between different readers, different points of analysis, and different pullbacks, found the following. First, FD-OCT has very high precision and reproducibility for measuring the lumen area and lesion length. Second, the biases of FD-OCT were never ⬎0.15 mm2 for the lumen area or 0.2 mm for the lesion length. Finally, the differences between pullbacks were typically larger than the differences between readers or different points of analysis, because gating approaches might still require additional improvements. Nonetheless, the degree of reproducibility achieved suggests that FD-OCT can be used in current and future
research endeavors requiring comparisons of separately acquired pullbacks of the same coronary segment. Previous studies have shown a high reproducibility for measurements using FD-OCT regarding the coronary evaluation of stents, plaque, and intraluminal thrombosis in vitro and in vivo.7–12 Specifically, Tanimoto et al7 studied interobserver variability in OCT measurements in vitro and in vitro, showing adequate precision for the lumen area, stent area, and neointimal area. Notably, their results for the lumen area showed a lower precision than our results (0.11 vs 0.045 mm2, respectively), most likely because of the adoption of different imaging and analyses systems. The Thoraxcenter group has also provided important insights in the usefulness of fully automatic analysis systems,8 showing
1110
The American Journal of Cardiology (www.ajconline.org)
Figure 4. Reproducibility of per-frame measurements of lumen area at OCT (mm2). (A) Scatterplot and (B) Bland-Altman plot for interobserver reproducibility. (C) Scatterplot and (D) Bland-Altman plot for intraobserver reproducibility. (E) Scatterplot and (F) Bland-Altman plot for interpullback reproducibility.
that automated analyses are possible and can provide more precise estimates than nonautomated approaches. Comparisons between different software programs and pullbacks were reported for the lumen area and stent area by Okamura et al,11 showing a Bland-Altman bias of 0.06 mm2 for the mean lumen area. The interobserver and intraobserver estimates of FD-OCT precision were provided by Gonzalo et al10 in the context of drug-eluting stents implanted in human coronary vessels, with results similar to those from our study at the proximal and distal reference levels. Sihan et al12 have demonstrated that gating using automatic methods is superior to nongating approaches for area and volume measurements using FD-OCT. Another interesting study has been reported by Terashima et al.9 In that bench research project, FD-OCT was shown to provide precise stent
measurements compared with manufacturer nominal strut thickness data. Finally, insightful findings have been reported on the comparison between FD-OCT and intravascular ultrasonography in patients receiving bioabsorbable coronary stents, demonstrating that even in this setting, FD-OCT is significantly more precise and accurate than intravascular ultrasonography.15 Our work builds on these previous studies by systematically determining the reproducibility of coronary measurements using FD-OCT for the lumen area and atherosclerotic lesion length detected through comparisons at the intraobserver, interobserver and interpullback levels. The excellent reproducibility of coronary measurements using FD-OCT resulted from the high-resolution method and the optimization of data acquisition, allowing very high-quality images.
Coronary Artery Disease/CLI-VAR Study
1111
Figure 5. Reproducibility of per-segment measurements of segment length at optical coherence tomography (mm). (A) Scatterplot and (B) Bland-Altman plot for interpullback reproducibility.
Notably, the lumen area measurements were performed with a semiquantitative analysis and the lesion length was evaluated by calculating the number of cross-sections between the lesion edges and, thus, relying more directly on data acquisition and pullback. Furthermore, the assessment of the reproducibility of the extension of atherosclerotic coronary disease and the effects it has on coronary lumen reduction observed in our study tested the validity of coronary imaging using FD-OCT for the fundamental parameters useful to the interventional cardiologist in the catheterization laboratory. The findings from our reproducibility study, given the established safety of FD-OCT,16 contribute to the validation of FD-OCT for the quantitative assessment of coronary lesions. Thus, it has a possible role in the decision-making strategies during coronary interventional procedures (i.e., guiding us on when and how to intervene with a coronary lesion). It can also be used to appreciate the possible effects of new therapeutic treatments of coronary artery disease in a serial study design of the progression and regression of atherosclerotic plaques. In addition, the satisfactory reproducibility data, stemming from frames acquired during both systole and diastole, suggest that no electrocardiographic gating is most likely required with such precise and accurate technology. The present study had several limitations, including the relatively small sample size, single-center setting, and application of FD-OCT on nonsignificantly stenosed coronary segments. Moreover, we limited our appraisal of the variability of the measurements using FD-OCT to the lumen area and lesion length, because determining the variability of other measurements (e.g., stent apposition, dissection extent, or thrombus burden) was beyond the scope of our study.17,18 Additional studies are required to reliably establish the reproducibility of FD-OCT in these other settings. Finally, although the reproducibility of intravascular ultrasonography for area and length measurements is similar to that of FD-OCT,19 the superior spatial resolution of the latter clearly favors it whenever accurate intracoronary imaging is required, unless the vessel wall also needs to be explored thoroughly. 1. Prati F, Cera M, Ramazzotti V, Imola F, Giudice R, Giudice M, Propris SD, Albertucci M. From bench to bedside: a novel technique of acquiring OCT images. Circ J 2008;72:839 – 843.
2. Prati F, Jenkins MW, Di Giorgio A, Rollins AM. Intracoronary optical coherence tomography, basic theory and image acquisition techniques. Int J Cardiovasc Imaging 2011;27:251–258. 3. Prati F, Ramazzotti V, Fernandez B, Albertucci M. The non-occlusive modality of optical coherence tomography image acquisition: a new concept for wide clinical application. G Ital Cardiol 2009;10:644 – 649. 4. Prati F, Stazi F, Dutary J, La Manna A, Di Giorgio A, Pawlosky T, Gonzalo N, Di Salvo ME, Imola F, Tamburino C, Albertucci M, Alfonso F. Detection of very early stent healing after primary angioplasty: an optical coherence tomographic observational study of chromium cobaltum and first-generation drug-eluting stents: the DETECTIVE study. Heart 2011;97:1841–1846. 5. Prati F, Regar E, Mintz GS, Arbustini E, Di Mario C, Jang IK, Akasaka T, Costa M, Guagliumi G, Grube E, Ozaki Y, Pinto F, Serruys PW; Expert’s OCT Review Document. Expert review document on methodology, terminology, and clinical applications of optical coherence tomography: physical principles, methodology of image acquisition, and clinical application for assessment of coronary arteries and atherosclerosis. Eur Heart J 2010;31:401– 415. 6. Imola F, Mallus MT, Ramazzotti V, Manzoli A, Pappalardo A, Di Giorgio A, Albertucci M, Prati F. Safety and feasibility of frequency domain optical coherence tomography to guide decision making in percutaneous coronary intervention. EuroIntervention 2010;6:575– 581. 7. Tanimoto S, Rodriguez-Granillo G, Barlis P, de Winter S, Bruining N, Hamers R, Knappen M, Verheye S, Serruys PW, Regar E. A novel approach for quantitative analysis of intracoronary optical coherence tomography: high inter-observer agreement with computer-assisted contour detection. Catheter Cardiovasc Interv 2008;72:228 –235. 8. Sihan K, Botha C, Post F, de Winter S, Gonzalo N, Regar E, Serruys PJ, Hamers R, Bruining N. Fully automatic three-dimensional quantitative analysis of intracoronary optical coherence tomography: method and validation. Catheter Cardiovasc Interv 2009;74:1058 –1065. 9. Terashima M, Rathore S, Suzuki Y, Nakayama Y, Kaneda H, Nasu K, Habara M, Katoh O, Suzuki T. Accuracy and reproducibility of stentstrut thickness determined by optical coherence tomography. J Invasive Cardiol 2009;21:602– 605. 10. Gonzalo N, Garcia-Garcia HM, Serruys PW, Commissaris KH, Bezerra H, Gobbens P, Costa M, Regar E. Reproducibility of quantitative optical coherence tomography for stent analysis. EuroIntervention 2009;5:224 –232. 11. Okamura T, Gonzalo N, Gutiérrez-Chico JL, Serruys PW, Bruining N, de Winter S, Dijkstra J, Commossaris KH, van Geuns RJ, van Soest G, Ligthart J, Regar E. Reproducibility of coronary Fourier domain optical coherence tomography: quantitative analysis of in vivo stented coronary arteries using three different software packages. EuroIntervention 2010;6:371–379. 12. Sihan K, Botha C, Post F, de Winter S, Gonzalo N, Regar E, Serruys PW, Hamers R, Bruining N. Retrospective image-based gating of intracoronary optical coherence tomography: implications for quantitative analysis. EuroIntervention 2011;6:1098 –1103. 13. Prati F, Cera M, Ramazzotti V, Imola F, Giudice R, Albertucci M. Safety and feasibility of a new non-occlusive technique for facilitated
1112
The American Journal of Cardiology (www.ajconline.org)
intracoronary optical coherence tomography (OCT) acquisition in various clinical and anatomical scenarios. EuroIntervention 2007;3: 365–370. 14. Bezerra HG, Costa MA, Guagliumi G, Rollins AM, Simon DI. Intracoronary optical coherence tomography: a comprehensive review clinical and research applications. JACC Cardiovasc Interv 2009;2:1035–1046. 15. Gómez-Lara J, Brugaletta S, Diletti R, Gogas BD, Farooq V, Onuma Y, Gobbens P, Van Es GA, García-García HM, Serruys PW. Agreement and reproducibility of gray-scale intravascular ultrasound and optical coherence tomography for the analysis of the bioresorbable vascular scaffold. Catheter Cardiovasc Interv 2012;79:890 –902. 16. Barlis P, Gonzalo N, Di Mario C, Prati F, Buellesfeld L, Rieber J, Dalby MC, Ferrante G, Cera M, Grube E, Serruys PW, Regar E. A multicentre evaluation of the safety of intracoronary optical coherence tomography. EuroIntervention 2009;5:90 –95.
17. Prati F, Zimarino M, Stabile E, Pizzicannella G, Fouad T, Rabozzi R, Filippini A, Pizzicannella J, Cera M, De Caterina R. Does optical coherence tomography identify arterial healing after stenting? An in vivo comparison with histology, in a rabbit carotid model. Heart 2008;94:217–221. 18. Capodanno D, Prati F, Pawlowsky T, Cera M, La Manna A, Albertucci M, Tamburino C. Comparison of optical coherence tomography and intravascular ultrasound for the assessment of in-stent tissue coverage after stent implantation. EuroIntervention 2009;5:538 –543. 19. Huisman J, Hartmann M, Mintz GS, van Houwelingen GK, Stoel MG, de Man FH, Louwerenburg HW, von Birgelen C. Impact of analyzing fewer image frames per segment during offline volumetric radiofrequency-based intravascular ultrasound measurements of target lesions prior to percutaneous coronary interventions. Int J Cardiovasc Imaging 2012;28:479 – 489.
b
0002-9149/12/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjcard.2012.05.047
trasonography in such research settings. However, limited data are available on the reproducibility of such data between separate pullbacks and between or within operators.7–12 In particular, the variability between and within operators and/or pullbacks has not been formally tested. We hypothesized that FD-OCT will allow reproducible measurements of the coronary lumen area and lesion length. Thus, we conducted a prospective study to systematically determine the reproducibility of FD-OCT measurements for the coronary lumen area and lesion length in humans. Methods The present study was a prospective single-center study approved by the local review board. Twenty-five nonconsecutive patients scheduled to undergo angioplasty from April 2009 to June 2011 were selected for the study. Of the 25 patients, 6 had stable angina, 8 had non–ST-segment elevation coronary syndrome, and 11 had evidence of myocardial ischemia. The exclusion criteria were recent (⬍72 hours) acute myocardial infarction, hemodynamic instability, renal insufficiency (serum creatinine ⬎2.0 mg/dl), alwww.ajconline.org
Coronary Artery Disease/CLI-VAR Study
1107
Figure 1. Lumen area measurements at OCT by different readers, at different times, or during different pullbacks.
Figure 2. Segment length measurements at OCT during different pullbacks.
lergy to the contrast media, left main disease, ostial coronary artery lesion, bypass lesion, chronic total occlusion, and a tortuous vessel. Each of the selected patients underwent coronary angiography followed by the acquisition of 2 pullback images using FD-OCT with a minimum length of 18 mm (pullback 1, and, 5 minutes later, pullback 2). The optical coherence tomographic images were obtained in a coronary artery that was not found to be significantly diseased at angiography (diameter stenosis ⬍50% on visual estimation). All patients provided written informed consent.
FD-OCT is a high-resolution intracoronary imaging technique with a maximum spatial resolution of 10 to 15 m. The current generation C7 system (LightLab Imaging, Westford, Massachusetts) was used in the present study. The imaging catheters for FD-OCT were delivered over a 0.014-in. guidewire through ⱖ6F guiding catheters. For an effective clearing of blood from the imaging field, angiographic contrast medium was injected through the guiding catheter, as previously described,13 with image acquisition performed according to established methods.5,14
1108
The American Journal of Cardiology (www.ajconline.org)
In all patients, we performed a standard femoral catheterization approach. Unfractionated heparin was administered intravenously to maintain an activated clotting time ⬎300 seconds. After coronary angiography, we positioned a 6F guiding catheter into the coronary ostium and advanced a standard 0.014-in. guidewire into the coronary artery. The imaging catheter for FD-OCT was advanced into the coronary artery to the segment of interest after administration of 200 g intracoronary nitroglycerine. The blood was cleared by injection of iso-osmolar contrast (iodixanol 320, Visipaque, GE Healthcare, Dublin, Ireland) at 37°C with an injection pump (Acist, Bracco, Milan, Italy) through the guiding catheter. Two continuous motorized pullbacks (pullbacks 1 and 2) using FD-OCT were started as soon as the artery was cleared of blood. The images from each pullback were acquired at a speed of 20 mm/s from the same coronary segment until reaching the guiding catheter. The pullbacks were taken at an interval of 5 minutes without additional intervention between them. The coronary images acquired were saved on the OFDI system console and exported to a digital system for off-line analysis at an independent imaging core laboratory (Rome Heart Research, Rome, Italy). The analysis of the coronary images from FD-OCT was performed by 2 analysts (reader A and B) experienced in FD-OCT independently using a dedicated contour detection program (LightLab Imaging). First, we identified the coronary segment of interest in each pullback and verified the correspondence of the 2 pullbacks through side-by-side comparisons and using as reference landmarks the crosssectional views of the side branches. After identifying the coronary segment in each pullback, we calibrated the measurements using the 6F (2-mm) catheter. In each crosssectional coronary image, we calculated the lumen area using automatic contour tracing of the intima and the same detection program, with very occasional manual correction (Figure 1). The length of the coronary segment was calculated from the number of cross-sectional frames taken of that segment (Figure 2). To evaluate interobserver reproducibility, 2 independent readers (A and B) independently analyzed pullback 1. To determine intraobserver reproducibility, 1 reader (reader A) analyzed pullback 1 twice, with the second reading 1 week after the first. Finally, to assess interpullback reproducibility, 1 reader (reader A) analyzed pullbacks 1 and 2 during same work session. The analyses were conducted at the segmental (per-segment) level (thus including all frames obtained from the study subject) and frame (per-frame) level. Correlation and bivariate regression analysis were performed to determine the association between different readings using both Pearson’s correlation and Spearman’s rho tests. In addition, Bland-Altman methods were used to obtain the bias estimates (reported as the mean and 95% bootstrapped confidence interval stemming from 1,000 samples each) and accompanying plots. Assuming an 8- ⫾ 4-mm2 lumen area and 20- ⫾ 7-mm lesion length, 25 patients were deemed sufficient to achieve adequately small standard errors of the mean (0.8 mm2 and 1.3 mm, respectively) and suitably precise 95% confidence intervals (6.4 to 9.6 mm2 and 17.2 to 22.8 mm, respectively). Statistical significance was set at the 2-tailed 0.05 level, and p values unadjusted for multiplicity are
Table 1 Reproducibility analysis Variable
Bland-Altman Bias*
Lumen area (mm2) Per-segment analysis Interobserver 0.001 (⫺0.012, 0.009) Intraobserver 0.003 (⫺0.002, 0.009) Interpullback 0.150 (⫺0.371, 0.086) Per-frame analysis Interobserver 0.001 (⫺0.001, 0.002) Intraobserver 0.002 (0.001, 0.003) Interpullback ⫺0.091 (⫺0.139, ⫺0.040) Length (mm) Per-segment analysis Interpullback ⫺0.200 (⫺3.00, 2.00)
Regression Analysis† (R Value, p Value)
1.0, ⬍0.001 1.0, ⬍0.001 0.982, ⬍0.001 1.0, ⬍0.001 1.0, ⬍0.001 0.959, ⬍0.001 0.990, ⬍0.001
* Reported as mean (95% bootstrapped confidence interval stemming from 1,000 samples). † Similar results for statistical significance and magnitude were obtained with Spearman’s correlation test.
reported throughout. Analyses were performed with SPSS, version 19 (IBM, Armonk, New York). Results We were able to acquire 2 separate coronary pullbacks using FD-OCT from each of the 25 patients in the study. The image quality of the pullbacks was excellent, with artifacts in ⬍1% of the evaluated images. The manual correction of the automatic lumen area contour was performed in 25% of measurements from ⱖ1 reader. No imaging-related complications, such as coronary dissection, perforation, spasm, embolization, arrhythmia, or clinical adverse events, occurred. A total of 9,396 frames were included in the analysis, from which a longitudinal image of each patient coronary target lesion was reconstructed. The reproducibility of the coronary measurement analysis using FD-OCT in the present study was very high in all comparisons (Table 1). Per-segment and per-frame analysis of the lumen areas showed very high and significant correlation coefficients, using both parametric and nonparametric tests, for the interobserver, intraobserver and interpullback comparisons (R ⱖ0.95 and p ⬍0.001 for all cases; Figures 3 and 4). Accordingly, the Bland-Altman estimates of bias showed very small differences in the interobserver, intraobserver, and interpullback comparisons for the lumen area at all levels, with average biases never ⬎0.091 mm2 for the lumen area or 0.200 mm for the lesion length. As expected, the interobserver and intraobserver analyses for the lumen area showed even greater precision and correlation than the interpullback estimates. Specifically, the R values were invariably 1.0 for interobserver and intraobserver analyses compared to R values of 0.959 to 0.982 for the interpullback analyses. Accordingly, absolute BlandAltman estimates of bias ranged from 0.001 to 0.003 for the interobserver and intraobserver analyses compared to estimates of bias of 0.091 to 0.150 for the interpullback analyses. Finally, the interobserver and intraobserver analyses for lesion length were perfectly concordant and precise (R ⫽ 1.0, p ⬍0.001; Figure 5).
Coronary Artery Disease/CLI-VAR Study
1109
Figure 3. Reproducibility of per-segment measurements of lumen area at OCT (mm2). (A) Scatterplot and (B) Bland-Altman plot for interobserver reproducibility. (C) Scatterplot and (D) Bland-Altman plot for intraobserver reproducibility. (E) Scatterplot and (F) Bland-Altman plot for interpullback reproducibility.
Discussion The present study, which has originally and comprehensively reported on the variability of measurements using FD-OCT between different readers, different points of analysis, and different pullbacks, found the following. First, FD-OCT has very high precision and reproducibility for measuring the lumen area and lesion length. Second, the biases of FD-OCT were never ⬎0.15 mm2 for the lumen area or 0.2 mm for the lesion length. Finally, the differences between pullbacks were typically larger than the differences between readers or different points of analysis, because gating approaches might still require additional improvements. Nonetheless, the degree of reproducibility achieved suggests that FD-OCT can be used in current and future
research endeavors requiring comparisons of separately acquired pullbacks of the same coronary segment. Previous studies have shown a high reproducibility for measurements using FD-OCT regarding the coronary evaluation of stents, plaque, and intraluminal thrombosis in vitro and in vivo.7–12 Specifically, Tanimoto et al7 studied interobserver variability in OCT measurements in vitro and in vitro, showing adequate precision for the lumen area, stent area, and neointimal area. Notably, their results for the lumen area showed a lower precision than our results (0.11 vs 0.045 mm2, respectively), most likely because of the adoption of different imaging and analyses systems. The Thoraxcenter group has also provided important insights in the usefulness of fully automatic analysis systems,8 showing
1110
The American Journal of Cardiology (www.ajconline.org)
Figure 4. Reproducibility of per-frame measurements of lumen area at OCT (mm2). (A) Scatterplot and (B) Bland-Altman plot for interobserver reproducibility. (C) Scatterplot and (D) Bland-Altman plot for intraobserver reproducibility. (E) Scatterplot and (F) Bland-Altman plot for interpullback reproducibility.
that automated analyses are possible and can provide more precise estimates than nonautomated approaches. Comparisons between different software programs and pullbacks were reported for the lumen area and stent area by Okamura et al,11 showing a Bland-Altman bias of 0.06 mm2 for the mean lumen area. The interobserver and intraobserver estimates of FD-OCT precision were provided by Gonzalo et al10 in the context of drug-eluting stents implanted in human coronary vessels, with results similar to those from our study at the proximal and distal reference levels. Sihan et al12 have demonstrated that gating using automatic methods is superior to nongating approaches for area and volume measurements using FD-OCT. Another interesting study has been reported by Terashima et al.9 In that bench research project, FD-OCT was shown to provide precise stent
measurements compared with manufacturer nominal strut thickness data. Finally, insightful findings have been reported on the comparison between FD-OCT and intravascular ultrasonography in patients receiving bioabsorbable coronary stents, demonstrating that even in this setting, FD-OCT is significantly more precise and accurate than intravascular ultrasonography.15 Our work builds on these previous studies by systematically determining the reproducibility of coronary measurements using FD-OCT for the lumen area and atherosclerotic lesion length detected through comparisons at the intraobserver, interobserver and interpullback levels. The excellent reproducibility of coronary measurements using FD-OCT resulted from the high-resolution method and the optimization of data acquisition, allowing very high-quality images.
Coronary Artery Disease/CLI-VAR Study
1111
Figure 5. Reproducibility of per-segment measurements of segment length at optical coherence tomography (mm). (A) Scatterplot and (B) Bland-Altman plot for interpullback reproducibility.
Notably, the lumen area measurements were performed with a semiquantitative analysis and the lesion length was evaluated by calculating the number of cross-sections between the lesion edges and, thus, relying more directly on data acquisition and pullback. Furthermore, the assessment of the reproducibility of the extension of atherosclerotic coronary disease and the effects it has on coronary lumen reduction observed in our study tested the validity of coronary imaging using FD-OCT for the fundamental parameters useful to the interventional cardiologist in the catheterization laboratory. The findings from our reproducibility study, given the established safety of FD-OCT,16 contribute to the validation of FD-OCT for the quantitative assessment of coronary lesions. Thus, it has a possible role in the decision-making strategies during coronary interventional procedures (i.e., guiding us on when and how to intervene with a coronary lesion). It can also be used to appreciate the possible effects of new therapeutic treatments of coronary artery disease in a serial study design of the progression and regression of atherosclerotic plaques. In addition, the satisfactory reproducibility data, stemming from frames acquired during both systole and diastole, suggest that no electrocardiographic gating is most likely required with such precise and accurate technology. The present study had several limitations, including the relatively small sample size, single-center setting, and application of FD-OCT on nonsignificantly stenosed coronary segments. Moreover, we limited our appraisal of the variability of the measurements using FD-OCT to the lumen area and lesion length, because determining the variability of other measurements (e.g., stent apposition, dissection extent, or thrombus burden) was beyond the scope of our study.17,18 Additional studies are required to reliably establish the reproducibility of FD-OCT in these other settings. Finally, although the reproducibility of intravascular ultrasonography for area and length measurements is similar to that of FD-OCT,19 the superior spatial resolution of the latter clearly favors it whenever accurate intracoronary imaging is required, unless the vessel wall also needs to be explored thoroughly. 1. Prati F, Cera M, Ramazzotti V, Imola F, Giudice R, Giudice M, Propris SD, Albertucci M. From bench to bedside: a novel technique of acquiring OCT images. Circ J 2008;72:839 – 843.
2. Prati F, Jenkins MW, Di Giorgio A, Rollins AM. Intracoronary optical coherence tomography, basic theory and image acquisition techniques. Int J Cardiovasc Imaging 2011;27:251–258. 3. Prati F, Ramazzotti V, Fernandez B, Albertucci M. The non-occlusive modality of optical coherence tomography image acquisition: a new concept for wide clinical application. G Ital Cardiol 2009;10:644 – 649. 4. Prati F, Stazi F, Dutary J, La Manna A, Di Giorgio A, Pawlosky T, Gonzalo N, Di Salvo ME, Imola F, Tamburino C, Albertucci M, Alfonso F. Detection of very early stent healing after primary angioplasty: an optical coherence tomographic observational study of chromium cobaltum and first-generation drug-eluting stents: the DETECTIVE study. Heart 2011;97:1841–1846. 5. Prati F, Regar E, Mintz GS, Arbustini E, Di Mario C, Jang IK, Akasaka T, Costa M, Guagliumi G, Grube E, Ozaki Y, Pinto F, Serruys PW; Expert’s OCT Review Document. Expert review document on methodology, terminology, and clinical applications of optical coherence tomography: physical principles, methodology of image acquisition, and clinical application for assessment of coronary arteries and atherosclerosis. Eur Heart J 2010;31:401– 415. 6. Imola F, Mallus MT, Ramazzotti V, Manzoli A, Pappalardo A, Di Giorgio A, Albertucci M, Prati F. Safety and feasibility of frequency domain optical coherence tomography to guide decision making in percutaneous coronary intervention. EuroIntervention 2010;6:575– 581. 7. Tanimoto S, Rodriguez-Granillo G, Barlis P, de Winter S, Bruining N, Hamers R, Knappen M, Verheye S, Serruys PW, Regar E. A novel approach for quantitative analysis of intracoronary optical coherence tomography: high inter-observer agreement with computer-assisted contour detection. Catheter Cardiovasc Interv 2008;72:228 –235. 8. Sihan K, Botha C, Post F, de Winter S, Gonzalo N, Regar E, Serruys PJ, Hamers R, Bruining N. Fully automatic three-dimensional quantitative analysis of intracoronary optical coherence tomography: method and validation. Catheter Cardiovasc Interv 2009;74:1058 –1065. 9. Terashima M, Rathore S, Suzuki Y, Nakayama Y, Kaneda H, Nasu K, Habara M, Katoh O, Suzuki T. Accuracy and reproducibility of stentstrut thickness determined by optical coherence tomography. J Invasive Cardiol 2009;21:602– 605. 10. Gonzalo N, Garcia-Garcia HM, Serruys PW, Commissaris KH, Bezerra H, Gobbens P, Costa M, Regar E. Reproducibility of quantitative optical coherence tomography for stent analysis. EuroIntervention 2009;5:224 –232. 11. Okamura T, Gonzalo N, Gutiérrez-Chico JL, Serruys PW, Bruining N, de Winter S, Dijkstra J, Commossaris KH, van Geuns RJ, van Soest G, Ligthart J, Regar E. Reproducibility of coronary Fourier domain optical coherence tomography: quantitative analysis of in vivo stented coronary arteries using three different software packages. EuroIntervention 2010;6:371–379. 12. Sihan K, Botha C, Post F, de Winter S, Gonzalo N, Regar E, Serruys PW, Hamers R, Bruining N. Retrospective image-based gating of intracoronary optical coherence tomography: implications for quantitative analysis. EuroIntervention 2011;6:1098 –1103. 13. Prati F, Cera M, Ramazzotti V, Imola F, Giudice R, Albertucci M. Safety and feasibility of a new non-occlusive technique for facilitated
1112
The American Journal of Cardiology (www.ajconline.org)
intracoronary optical coherence tomography (OCT) acquisition in various clinical and anatomical scenarios. EuroIntervention 2007;3: 365–370. 14. Bezerra HG, Costa MA, Guagliumi G, Rollins AM, Simon DI. Intracoronary optical coherence tomography: a comprehensive review clinical and research applications. JACC Cardiovasc Interv 2009;2:1035–1046. 15. Gómez-Lara J, Brugaletta S, Diletti R, Gogas BD, Farooq V, Onuma Y, Gobbens P, Van Es GA, García-García HM, Serruys PW. Agreement and reproducibility of gray-scale intravascular ultrasound and optical coherence tomography for the analysis of the bioresorbable vascular scaffold. Catheter Cardiovasc Interv 2012;79:890 –902. 16. Barlis P, Gonzalo N, Di Mario C, Prati F, Buellesfeld L, Rieber J, Dalby MC, Ferrante G, Cera M, Grube E, Serruys PW, Regar E. A multicentre evaluation of the safety of intracoronary optical coherence tomography. EuroIntervention 2009;5:90 –95.
17. Prati F, Zimarino M, Stabile E, Pizzicannella G, Fouad T, Rabozzi R, Filippini A, Pizzicannella J, Cera M, De Caterina R. Does optical coherence tomography identify arterial healing after stenting? An in vivo comparison with histology, in a rabbit carotid model. Heart 2008;94:217–221. 18. Capodanno D, Prati F, Pawlowsky T, Cera M, La Manna A, Albertucci M, Tamburino C. Comparison of optical coherence tomography and intravascular ultrasound for the assessment of in-stent tissue coverage after stent implantation. EuroIntervention 2009;5:538 –543. 19. Huisman J, Hartmann M, Mintz GS, van Houwelingen GK, Stoel MG, de Man FH, Louwerenburg HW, von Birgelen C. Impact of analyzing fewer image frames per segment during offline volumetric radiofrequency-based intravascular ultrasound measurements of target lesions prior to percutaneous coronary interventions. Int J Cardiovasc Imaging 2012;28:479 – 489.