Imaging-based detection of anterior chamber inflammation: a comparative diagnostic accuracy study

Purpose:
We investigated the impact of operator parameters on the diagnostic performance of anterior-segment optical coherence tomography (AS-OCT) in anterior uveitis.

Design:
Prospective comparative diagnostic analysis.

Methods:
Setting: Single site. Study population: Children younger than 18 years with anterior uveitis, recruited consecutively. Observation procedures: Index testing: Optovue RTVue80 AS-OCT using “low-volume” (LV, horizontal and vertical cross-sections) and “high-volume” (HV, 68 horizontal cross-sections) protocols. Reference testing: slitlamp examination with anterior chamber inflammation graded using standardization of uveitis nomenclature (SUN). Main outcome measure: Index test performance metrics (sensitivity, specificity, and likelihood ratios), utility for “ruling-in” and “ruling-out” disease (positive/negative predictive values, PPV/NPV), receiver operating characteristic curves to explore the impact of different imaging-derived metrics, multivariable multilevel regression analyses to quantify correlation of index to reference testing, and repeatability indices across protocols.

Results:
A total of 40 children (77 eyes: 51 eyes at SUN grade 0, 10 at SUN 0.5+, 8 at SUN 1+, and 8 SUN ≥2+ or higher) were included. There was high repeatability across protocols (0.98, P < .001, 95% CI: 0.75-1.0). OCT resulted in strong predictive values for “ruling-out” (LV-scan NPV 82.9%, 95% CI: 71.5%-90.4%; HV-scan NPV 100%, 95% CI: 3%-100%) but a less predictive value for “ruling-in” SUN ≥0.5+ (LV-scan PPV 52.8%, 95% CI: 41.5%-63.7%; HV-scan PPV 34.2%, 95% CI: 33.3%-35.1%). Detection of more than 1 cell within a cross-sectional scan was strongly suggestive of clinical activity, with an area under the curve of 0.76 (95% CI: 0.62-0.89) for SUN ≥0.5+ and 0.85 (95% CI: 0.73-0.98) for the detection of SUN ≥1+. Cell count correlated with SUN grades at higher levels of inflammation (SUN ≥2+ both protocols, SUN ≥1+ HV-scans). There was an independent positive association between age and AS-OCT cell (adjusted correlation coefficient 0.2 cells for each additional year of age).

Conclusions:
Operator-dependent factors impact the diagnostic and quantification performance of AS-OCT for anterior chamber inflammation. However, the strong, “dose-respondent” correlation of LV protocols with SUN grading promises clinical utility without the storage and analysis burden of HV approaches. Further work will involve exploration of the need for age-specific image metric interpretation.