Validation of central serous chorioretinopathy multimodal imaging-based classification system.

PURPOSE

Validation of a recently described central serous chorioretinopathy (CSCR) classification system and assessment of levels of agreement among 10 retina physicians.

METHODS

This was a cross-sectional (inter-reader agreement) study. Ten retina physicians (assigned a role of masked grader) were provided with a comprehensive dataset of 61 eyes of 34 patients of presumed CSCR. Relevant clinical details and multimodal imaging (fundus autofluorescence, fluorescein and indocyanine green angiography, optical coherence tomography) of both involved and fellow eye were electronically shared. Later, only the fellow eye images were resent to understand the influence of affected eye on the grading of the fellow eye. Multiple inter-grader agreement using Fleiss Kappa was performed to determine the level of agreement among the 10 graders. p value of ≤ 0.05 was considered statistically significant.

RESULTS

Sixty-one eyes of 34 patients were evaluated. There was moderate agreement for major criteria with Fleiss Kappa value of 0.50 (p < 0.0001) with a single outlier observer. After excluding that observer, the Fleiss Kappa value increased to 0.57 (p < 0.0001) with statistically significant p values among all categories, i.e., simple CSC ([Formula: see text] = 0.575), complex CSC ([Formula: see text] = 0.621), and no CSC ([Formula: see text] = 0.452). Overall, moderate to substantial agreement was noted among the subtypes (primary, recurrent, and resolved). The influence of the affected eye on fellow eye grading was studied. The global Fleiss Kappa coefficient ([Formula: see text] = 0.642, p < 0.0001) showed substantial agreement when observers were aware of the affected eye grading. However, without prior available information on the affected eye, the inter-grader agreement was significantly lower (global [Formula: see text] = 0.255, p < 0.0001).

CONCLUSION

A fair-moderate inter-grader agreement among the masked graders suggests a need for further refinement of this novel classification system. Disease grading should include both eyes as lack of information on affected eye has a bearing on fellow eye grading and inter-grader agreement as shown by a significant difference in global [Formula: see text] values.