Abstract
Background
Gastroesophageal adenocarcinomas expressing claudin 18.2 (CLDN18) are candidates for targeted therapies such as zolbetuximab. Immunohistochemical staining for CLDN18 above a certain threshold is used to establish treatment eligibility. The purpose of this study was to evaluate the interobserver variability among pathologists in scoring CLDN18 immunostaining.Design
Six gastrointestinal pathologists evaluated CLDN18 staining on digitally scanned slides from a nonconsecutive cohort of 47 cases and a consecutive cohort of 48 cases of gastroesophageal adenocarcinoma. Each rater estimated the percentage of tumor cells showing 0, 1+, 2+, or 3+ membranous staining intensity (producing an overall H-score) and also noted whether they would score the carcinoma as positive according to the cutoff for zolbetuximab eligibility (>75% of tumor cells with 2+ or 3+ staining). Factors potentially influencing scoring such as heterogeneity of staining and technical or sample factors were also noted. Interobserver agreement was assessed using kappa statistics (categorical scores) and intraclass correlation coefficients (ICCs) (continuous measures).Results
Interobserver agreement for overall H-scores was good, with an ICC of 0.85 and 0.91 in the two cohorts. There was substantial interobserver agreement (Fleiss kappa=0.63 and 0.72 for the two cohorts) regarding eligibility for zolbetuximab therapy, For the nonconsecutive cohort, there was complete concordance among all raters in 32 cases in the nonconsecutive cohort (68%) (8 cases eligible, 24 cases ineligible). For estimated proportions of tumor cells showing 0, 1+, 2+, and 3+ CLDN18 staining, the ICCs were 0.84, 0.29, 0.41, and 0.70, respectively. Cases with heterogeneous staining also showed greater variability (kappa 0.49 vs. 0.71 for zolbetuximab eligibility).Conclusion
The study demonstrated substantial interobserver agreement among gastrointestinal pathologists in evaluating CLDN18 staining for the purposes of zolbetuximab eligibility. Assessment of weaker staining intensities was a source of greater variability. These findings highlight the need for standardized scoring of CLDN18 staining for the purposes of treatment planning, ideally with a consensus approach in difficult cases.