Send letter to journal:
Re: Multilevel analysis for OCT imaging

Dear Editor,

Chen et al., present the results of a non-randomised prospective study, in which individual struts in 28 stents implanted in 26 lesions in 24 patients were assessed by OCT. The authors report strut apposition and coverage as binary variables and pool strut-related observations to produce percentages of incompletely apposed/uncovered struts for stents types. It is evident that individual observations related strut coverage or strut-to-intima distance violate the fundamental assumption of independence of observations required for the application of classical statistical methods. Struts within lesions, in fact, share common characteristics, making them more similar to each other than to struts from different lesions. Ignoring this multilevel structure results in spuriously low estimated standard errors and p-values.

Similarly, when assessing neontimal thickness (NIT), Chen et al. measure maximal and minimal NIT for each strut and produce summary statistics at stent level by compiling these values into means. Subsequently, they recompile the means to produce “means of means” and compare these between stent groups using standard statistical tools. The use of summary statistics ignores the variation at strut level and the precision of the cluster estimate. Moreover, this can only be used for data with a 2-level structure and, thus, not in cases such as the present study in which more than one lesion belong to the same patient (3-level structure: struts clustered within lesions clustered within patients).

Optical coherence tomography (OCT) has revolutionised intracoronary imaging, providing a unique insight into all aspects of coronary stenting. Realisation of the multilevel structure of OCT data on coronary stenting is essential to the design, reporting and critical appraisal of all stent -related OCT research. To ignore this, could have disastrous effects on the validity of any statistical analysis and, eventually, on the credibility of the technique itself.

REFERENCES