Dear Editor,I read with great interest the valuable article titled “Clinical and CT patterns to predict EGFR mutation in patients with non-small cell lung cancer: A systematic literature review and meta-analysis” published in the February 2022 issue of the European journal of radiology Open [1]. Authors have performed a valuable systematic review and meta-analysis to determine if the presence of specific clinical and computed tomography (CT) patterns are associated with epidermal growth factor receptor (EGFR) mutation in patients with non-small cell lung cancer. They have finally concluded that ground-glass opacities (GGO), air bronchogram, vascular convergence, pleural retraction, spiculated margins, early disease stage, female gender, and non-smoking status are significant risk factors for EGFR mutation. I would like to congratulate the authors for conducting this valuable meta-analysis; however, I wanted to mention an important issue concerning strategies for identifying and addressing heterogeneity in this study.The authors have calculated I2 values of 80.3% for GGO, 44.7% for air bronchogram, 0% for vascular convergence, 0% for pleural retraction, 51.2% for spiculation, 0% for cavitation, 55.2% for early disease stage, 64.6% for non-smoker status, and 67.7% for female gender. In the methods section, the authors have stated that, “The statistical heterogeneity of the studies included was explored for each clinical and CT pattern using Cochran’s Q test and I2 test. If the Cochran’s Q test was < 0.05, we considered that the meta-analysis presented a high degree of heterogeneity. If the I2 test value was < 50%, we used a fixed-effect model; nevertheless, we used a random effect model if the I2 test value was > 50%”. Accordingly, the authors have used random effects model for the majority of the evaluated variables, including GGO, air bronchogram, spiculation, early disease stage, non-smoker status, and female gender.According to the Cochrane handbook for systematic reviews of interventions, using a random effect model is indeed an accepted strategy for addressing heterogeneity [2]. However, the preferred option is to explore and determine the cause of observed heterogeneity [2]. The calculated I2 values demonstrate substantial heterogeneity in the evaluated variables particularly for GGO (80.3%), spiculation (51.2%), early disease stage (55.2%), non-smoker status (64.6%), and female gender (67.7%). Performing meta-analysis seems to be inappropriate for these variables (if the cause of observed heterogeneity cannot be explored and determined, a mere systematic literature review would be enough).To briefly provide examples, Han et al. [3] have provided evidence that tumors displaying GGOs on CT correlated positively with the lepidic-predominant subtype on histologic examination. In addition, GGOs are significantly more highly related to the exon 21 mutation than the exon 19 mutation (also demonstrated by Lee et al. [4]). These evidences might explain the observed considerable heterogeneity in the literature.Nevertheless, I again congratulate the authors for performing such a practical and comprehensive systematic review and meta-analysis. I believe this study is of high importance to practicing radiologists and oncologists.
Funding
This manuscript did not receive any grant from funding agencies in the public, commercial, or not-for-profit sectors.
Ethics
This is a letter to the editor and was drafted in accordance with ethical considerations.
Conflict of interest
The author declare that there was no conflict of interest.