Assessment of a race-specific normative HRT-III database to differentiate glaucomatous from normal eyes

Abstract

Purpose: To determine if a new, normative, race-specific database enhances the ability of confocal scanning laser ophthalmoscopy to differentiate normal from glaucomatous eyes.

Methods: One eye of eligible normal and glaucoma patients was enrolled. All subjects underwent a complete ophthalmologic examination, standard achromatic perimetry (SITA-SAP, 24-2), and confocal scanning laser ophthalmoscopy [Heidelberg retinal tomograph (HRT-II)] within 1 month of enrollment. Racial groups were defined by self-report. Glaucoma was defined by the existence of reproducible SAP loss (pattern standard deviation <5% and/or Glaucoma Hemifield Test outside normal limits) on 2 consecutive fields. Normal subjects had 2 normal visual fields (pattern standard deviation >5% and Glaucoma Hemifield Test within 97% normal limits) and a normal clinical examination. HRT-II examinations were exported to the HRT-III software, which includes a large race-specific normative database consisting of 733 white and 215 black eyes. Moorfields regression analysis (MRA) for the most abnormal optic disc sector was compared between the HRT-II (MRA2) and the HRT-III software before (MRA3-B) and after (MRA3-A) adjustment for race. Sectors outside the 99.9% confidence interval limits ("outside normal limits") were determined to be abnormal.

Results: We enrolled 124 black (52 glaucoma, 72 normal) and 96 white (32 glaucoma, 64 normal) subjects. Mean age was 51+/-13 years and 50+/-16 years for blacks and whites, respectively (P = 0.45). Visual field mean deviation was -7.3+/-6.7 db for glaucomatous eyes and -0.4+/-1.1 db for normal eyes (P < 0.001). Sensitivity and specificity for the HRT-II was 71.9% and 95.3%, respectively, for white subjects and 50.0% and 98.6%, respectively, for black subjects. Using the expanded HRT-III database, analysis yielded a sensitivity of 81.3% and specificity of 93.8% for whites and a sensitivity of 71.2% and specificity of 86.1% for blacks. After an adjustment for black ethnicity was made in the HRT-III program, the sensitivity and specificity for blacks was 65.4% and 90.3%, respectively.

Conclusions: A new, larger, race-specific HRT-III database increases sensitivity while maintaining specificity for whites and increases sensitivity but decreases specificity for blacks. New software and databases based on race require careful scrutiny before use in clinical practice.