Author + information
- Giuseppe Ferrante, MD, PhD⁎ (, )
- Konstantinos Dimopoulos, MD, MSc, PhD,
- Rocco Antonio Montone, MD and
- Peter Barlis, MBBS, PhD
- ↵⁎Institute of Cardiology, Catholic University of the Sacred Heart, Largo Agostino Gemelli, 8, 00168 Rome, Italy
We read with great interest the article by Murata et al. (1), who compared the accuracy of optical coherence tomography (OCT) to histological analysis for the quantitative assessment of neointimal response after drug-eluting stent implantation in normal porcine coronary arteries. The authors should be commended for this study as the addressed issue is of clinical relevance, in view of the rapid widespread use of OCT in catheterization laboratories and the application of OCT end points in randomized clinical trials of intervention. Therefore, it is crucial to address the quantitative performance of OCT.
The authors conclude that OCT shows high correlation with histology for neointimal area, neointimal thickness, and luminal area measurement, but not for stent area assessment, based on squared correlation coefficient R2 between values obtained by the 2 techniques of approximately 0.8 and 0.3 for high and poor correlation, respectively.
However, the validity of the results reported by Murata and colleagues must be interpreted with caution due to the type of statistical analysis used. When comparing 2 different techniques, the Pearson correlation coefficient (R), used in this study, is not appropriate and may be misleading (2). Indeed, a high correlation coefficient suggests a strong relation between 2 variables, but not necessarily a good agreement between 2 methods. On a scatterplot of values measured by 1 technique against another, agreement is present if points lie along the line of equality, whereas correlation is good when points lie along any straight line. Bland and Altman plots with estimation of limits of agreement and repeatability coefficients is the recommended approach when comparing 2 techniques or for the assessment of intraobserver and interobserver variability (2–4), and has been previously used in OCT studies of strut apposition and neointimal coverage (5). Therefore, further statistical verification of the agreement between OCT and histology in the measurement of neointimal response to stent implantation may be useful to confirm the results of the present, elegant study. Finally, the use of Student t test for the comparison of neointimal thickness may have resulted in falsely low (significant) p values, as t test assumes independence of observations, whereas struts belonging to a single stent are not independent. Multilevel analysis accounting for clustering of struts in stents, lesions, and/or subjects is an appropriate, albeit computationally demanding, approach.
- American College of Cardiology Foundation
- Murata A.,
- Wallace-Bradley D.,
- Tellez A.,
- et al.
- Bland J.M.,
- Altman D.G.
- Barlis P.,
- Dimopoulos K.,
- Tanigawa J.,
- et al.