optic nerve head was assessed within the image evaluation process to detect macropapillas. The criteria deÞned by Jonas [12] were used for diagnosis of glaucomatous optic nerve atrophy, which was based on the following signs: (1) rim inferior or rim superior was smaller than temporal rim,

(2) peripapillary hemorrhages, (3) localized nerve Þber layer defects, and (4) diffuse nerve Þber loss with glaucomatous cupping.

No visual Þeld test was performed. For the diagnosis of glaucomatous optic nerve atrophy, only the morphological pathology of the optic nerve head was examined. All images were stored electronically.

7.2.3Statistics

variables. The signiÞcance level for statistical tests was set to 0.05. The exact logistic regression models were computed with the statistical software package SAS 8.2, and the graphics were created with the freely available statistical programming environment R [13].

The correlation coefÞcient of reliability was determined by Cronbach alpha at the 95% conÞdence interval. The corresponding F-statistics were calculated as multivariate analysis from the Hotelling T2 test using SPSS software.

7.3Results

7.3.1Reliability of Image Evaluation

For statistical analysis, we compared several logistic regression models with different predictor

Two hundred and twelve monoscopic fundus images were randomly selected from the database mentioned above. Two ophthalmologists

assessed the images independently from each other two times according to the possible diagnoses: no glaucoma (grading 0Ð2), suspicious for glaucoma (grading 3Ð7), or deÞnite glaucoma (grading 8Ð10). The grading was based on subjective assessment and experience. If the difference between Þrst and second assessment differed less than two ranks from each other, the diagnosis was considered as conÞrmed. Alternatively, the diagnosis was classiÞed as not conÞrmed.

The intraobserver reliability was 0.884 and 0.840, respectively, for the two observers. The interobserver reliability of the Þrst cycle of evaluation was 0.740. Corresponding T2 for the intraobserver reliability was not signiÞcantly different (0.550, p = 0.459, and 0.033, p = 0.856, respectively), indicating that the mean values of the two cycles did not vary for each observer. The interobserver reliability, however, was (7.647, p = 0.006) with df1 = 1 and df2 = 211 each. This means that mean values of two cycles varied signiÞcantly between observers. Evaluation methods of images during all four assessment cycles did not increase the inner consistency in any noteworthy way (Cronbach alpha 0.870).

7.3.2Prevalence of Glaucomatous Optic Nerve Atrophy

We found 35 cases (0.36%) of glaucomatous optic nerve atrophy in the study population. Three cases have been previously diagnosed as glaucomatous. The proportion of glaucomatous optic nerve atrophy was smaller among women (9 cases: 0.24%) than among men (26 cases: 0.45%). The inßuence of age on the prevalence of glaucomatous optic nerve atrophy is illustrated in Table 7.1.

We found 2 cases (0.07%) of glaucomatous optic nerve atrophy in the age group 45Ð49 years, 12 cases (0.40%) in the group 50Ð54 years, 10 cases (0.45%) in the group 55Ð59 years, and 11 cases (0.82%) in the group 60Ð64 years (see Fig. 7.2).

To establish the age dependency of glaucomatous optic nerve atrophy, we Þtted a logistic regression model with age as the predictor variable and presence of glaucomatous optic nerve atrophy as the dependent variable. Because of the small case number, we computed the exact logistic regression [14] instead of the approximation, resulting in a highly signiÞcant model (p < 0.0001 for the estimated parameter). We examined dependency of glaucomatous optic nerve atrophy on gender by Þtting an exact logistic regression model using age group and gender as predictor variables. The resulting parameter estimates for gender and for the interaction of gender with age did not prove to be signiÞcantly different from zero (p = 0.14 and p = 0.20, respectively).

Two thousand two hundred and Þfty participants reported previous diagnosis of arterial hypertension. Among those were 13 participants with glaucomatous optic nerve atrophy. Four hundred and eight persons could not tell if they had arterial hypertension. Two hundred and ninety-four participants, including 20 glaucomatous optic nerve atrophy patients, reported previously raised IOP (148 women, 146 men). Two hundred and Þfty-four participants could not tell if their IOP was raised or not. Because these data were not measured but only reported and therefore open to doubt, we did not evaluate them statistically. One thousand four hundred and four persons reported to have a family history of glaucoma (see Table 7.3). Eight of these persons had glaucomatous optic nerve atrophy.

Fig. 7.2 Prevalence of glaucomatous optic nerve atrophy (own data).

The age groups <40, 40Ð44, and >64 are included for comparison with other studies

(see Table 7.2)

Table 7.2 Prevalence of open-angle glaucoma from the meta-analysis by the Eye Diseases Prevalence Research Group [16]

Table 7.3 Prevalence of glaucomatous optic nerve atrophy in subjects with and without family history of glaucoma (glaucoma in grandparents, parents, siblings, or children)

0.020

The numbers are generated from pooled data from the Baltimore Eye Survey, the Blue Mountains Eye Study, the Beaver Dam Study, the Rotterdam Study, and the Melbourne VIP, and the prevalence of glaucomatous optic nerve atrophy found in our study