Abstract
Purpose: :
To compare the abilities of the race-specific normative Heidelberg Retina Tomography (HRT)-3 databases (Caucasian, African-American, and Indian) to discriminate between normal and glaucomatous eyes in a Singapore population.
Methods: :
A prospective cross-sectional study was conducted in 56 eyes of 56 healthy volunteers and 27 eyes of 27 glaucoma patients. All participants underwent standardized interviews, ophthalmic examination, perimetry and HRT optic disc imaging. Area under the curve receiver operating characteristics (AROC), sensitivity and specificity were used to evaluate the diagnostic performances of 3 normative databases, for both Moorfields Regression Analysis (MRA) and Glaucoma Probability Score (GPS).
Results: :
Normal subjects were significantly younger than glaucoma patients (p<0.001). There were no significant differences in gender or ethnicity.There was no significant difference (p = 0.573) in MRA results among the 3 databases. There was a significantly higher percentage (64.6%) of patients with GPS results "within normal limits" when using the African-American database, as compared to 39.0% seen in both Caucasian and Indian databases (p = 0.006). Separate analysis of only Chinese subjects revealed similar findings for both MRA and GPS.For MRA, the highest sensitivity was obtained with either the Caucasian or African- American databases (70.4%, p <0.001), while the highest specificity was seen using the Indian database (94.6%, p < 0.001). The AROC for discrimination between glaucomatous and normal eyes by MRA Caucasian database, MRA African-American database, and MRA Indian database were 0.789 (95% CI, 0.691-0.887), 0.807 (0.712-0.903) and 0.732 (0.632-0.833) respectively. The MRA results are independent of optic disc area.For GPS, the highest sensitivity was obtained using either the Caucasian or Indian databases (69.2%, p<0.001). The highest specificity was seen with the African-American database (98.2%, p<0.001). The AROC for GPS Caucasian database, GPS African-American database, and GPS Indian database were 0.774 (95% CI, 0.674-0.875), 0.788 (0.692-0.884) and 0.774 (0.674-0.875) respectively.
Conclusions: :
The African-American database is preferable over the Caucasian and Indian databases in the Asian population. Future large-scale population studies are needed to address the lack of an Asian population-specific database.
Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • optic disc