DESIGN: Prospective cross-sectional study.
METHODS: Consecutive CSC patients were recruited from retina clinic. The reference standard for CNV was determined by interpretation of multimodal imaging with OCTA, structural OCT line scan, fluorescein angiography (FA), indocyanine green angiography (ICGA), ultra-widefield fundus photography and fundus autofluorescence (FAF). Two independent masked graders examined OCTA without FA and ICGA to diagnose CNV. Univariate and multivariate analyses were performed to evaluate factors associated with CNV.
RESULTS: CNV was detected in 69 eyes in 64 out of 277 CSC patients according to reference standard. The two masked graders who examined OCTA had sensitivity of 81.2% (95% Confidence Interval [CI], 71.9%-90.4%) and 78.3% (95% CI, 68.5%-88.0%), specificity of 97.3% (95% CI, 95.9%-98.8%) and 96.2% (95% CI, 94.5%-98.0%), positive predictive values of 82.4% (95% CI, 73.3%-91.4%) and 76.1% (95% CI, 66.1%-86.0%), and negative predictive values of 97.1% (95% CI, 95.6%-98.7%) and 96.7% (95% CI, 95.0%-98.3%). Their mean area under the receiver operating characteristic curve (AUC) was 0.88 with good agreement (Kappa coefficient 0.80 [95% CI, 0.72-0.89]). Flat irregular pigment epithelial detachment on structural OCT, neovascular network on OCTA and ill-defined late leakage on FA significantly correlated with CNV in CSC from multiple regression (P < 0.001, P < 0.001 and P = 0.005, respectively).
CONCLUSIONS: There is discordance between OCTA and multimodal imaging in diagnosing CNV in CSC. This study demonstrated the caveats in OCTA interpretation, such as small extrafoveal lesions and retinal pigment epithelial alterations. Comprehensive interpretation of OCTA with dye angiography and structural OCT is recommended.