Commentary:H. pylorieradication therapy after subtotal gastrectomy for gastric cancer - can the qualitative be quantified? Authors' reply

Abstract

We appreciate Dr Genta for his valuable comments1 and agree with his perspective on our analysis.2 The updated Sydney system recommends that glandular atrophy and intestinal metaplasia be graded as absent, mild, moderate or marked.3 Such grading in this updated system is not a continuous scale, but an ordinal scale. Thus, the data evaluated in the updated Sydney system might not be suitable for comparing means and variances (standard deviations or standard errors of the mean) after conversion into numerical values. Several studies, however, have used means and variances to present changes both before and after Helicobacter pylori eradication.4-7 In our study, we similarly analysed our data using numerical analysis. However, to overcome the limitations of the numerical values, we incorporated Figure 3 into our paper, which provides a graphical representation of the distribution of the changes in gastritis severity both before and after H. pylori eradication.2 We also evaluated our data using a mixed-effects ordinal logistic regression model, which was suggested by Ohkusa et al.5 The results showed significant decreases in both glandular atrophy and intestinal metaplasia scores of the lesser curvature of the corpus in H. pylori-negative patients compared with those in H. pylori-positive patients, after considering H. pylori status over time (data not shown). Updated Sydney system grading was founded on a relatively objective visual analogue scale, using a graphic model and reducing textual explanation. However, as Dr Genta suggests, grading of such observations is still subject to significant inter- and intra-observer variabilities, particularly for atrophy.8, 9 Despite these limitations, this system is currently the mostly widely accepted and used grading system for gastritis. One specialised pathologist's examination could handle inter-observer variations, but is still subject to intra-observer variabilities. We believe that this dilemma might be resolved in part by a double-blind, randomised trial design. The authors’ declarations of personal and financial interests are unchanged from those in the original article.2