Development of a Clinically Relevant Endoscopic Grading System for Chronic Rhinosinusitis Using Canonical Correlation Analysis

Published In

International Forum of Allergy & Rhinology

Document Type


Publication Date



Background Diagnostic nasal endoscopy is a routine measure of sinonasal inflammation in patients with chronic rhinosinusitis (CRS). Although multiple staging systems have been proposed and evaluated, evidence of association between concurrent symptoms and endoscopic findings remains discordant. The goal of this study is to identify the relevant endoscopic attributes associated with symptom burden, and to systematically derive a weighted endoscopic scale that optimizes prediction of concurrent symptoms. Methods Reported baseline symptom (22-item Sino-Nasal Outcome Test [SNOT-22]) and endoscopic evaluation scores (Lund-Kennedy [LK]) were obtained from patients with CRS enrolled in a prospective cohort study. Canonical correlation analysis of the SNOT-22 subdomains and LK variables was completed. Results A total of 629 patients were included in analysis including 343 with prior endoscopic sinus surgery. Significant canonical correlations outperformed aggregate correlations in explaining variance of the data (33% vs 3%, respectively). The first canonical correlation was dominated by the rhinologic symptom domain and the endoscopic polyp score (r = 0.54; p < 0.05) whereas additional significant canonical correlation was found between the extra-rhinologic symptom subdomain and the edema score in patients without prior ESS (r = 0.21; p < 0.05), and discharge in patients with prior ESS (r = 0.22; p < 0.05). All other domains and endoscopic variables did not significantly contribute to the canonical correlation. Conclusion Although aggregate symptoms and endoscopic scores demonstrate minimal correlation, a weighted combination of symptom domains and endoscopic attributes greatly improves this correlation. A simple approximation of the weights of each of the endoscopic variables of polyps, edema, discharge, scarring, and crusting, is an approximate ratio of 4:2:1:0:0, respectively.



Persistent Identifier