Interobserver variability if evaluating cross-sectioanl images

Interobserver variability if evaluating cross-sectioanl images

Study 1

An online survey, including questions relating to qualification, practice environment, and diagnostic methods was completed by 1256 optometrists. Based on their responses, 208 (17%) were selected to undertake an online disc assessment exercise. Optometrists evaluated the same disc images previously assessed by European ophthalmologists as part of the European Optic Disc Assessment Trial (EODAT); the task was to state if the disc appeared healthy or glaucomatous. There were 110 stereoscopic disc images, of which 40 were healthy, 48 glaucomatous, and six ocular hypertensive, with 16 duplicates images. Sensitivity, specifisity and overall accuracy were calculated and compared between optometrist groups and with the EODAT ophthalmologists using permutation analysis.

Median sensitivity was 0.92 (95% CI: 0.70, 1.00) and median specificity was 0.74 (95% CI: 0.62, 0.88). Median overall accuracy was 80% (95% CI: 67%, 88%). Agreement between optometrists was moderate (Fleiss' κ: 0.57). Optometrists with higher qualifications did not have overall higher sensitivity than those without (p = 0.23), but had higher specificity (p = 0.001) and higher overall accuracy (p < 0.001). Optometrists displayed higher sensitivity but lower specificity than the EODAT ophthalmologists «Hadwin SE, Redmond T, Garway-Heath DF ym. Assessme...»1.

Study 2

The purpose of the study wasdetermine the diagnostic accuracy of judging optic disc photographs for glaucoma by ophthalmologists. A total of 243 of 875 (27%) invited ophthalmologists in 11 European countries classified 40 healthy eyes and 48 glaucomatous eyes with varying severity of the disease on stereoscopic slides. Duplicate slides were provided for determining intraobserver agreement «Reus NJ, Lemij HG, Garway-Heath DF ym. Clinical as...»2.

The intraobserver agreement (kappa) varied between -0.13 and 1.0 and was on average good (0.7).The overall diagnostic accuracy of ophthalmologists was 81% (standard deviation [SD], 6.8; range, 61%-94%).

Study 3

Two hundred seven subjects (109 glaucoma and 98 normal subjects) were evaluated to study the agreement of optic disc measurements obtained with the Cirrus high-density optical coherence tomography (HD-OCT) and the Heidelberg retina tomograph (HRT) and compare the intervisit, test-retest variability between the instruments.

One eye from each individual was selected randomly for optic disc imaging by the Cirrus HD-OCT and the HRT. Areas of the optic disc and the cup, cup volume, vertical cup-to-disc ratio and cup-to-disc area ratio were compared between the instruments. The OCT measurements were corrected for ocular magnification using the Littman's formula. The measurement agreement was evaluated with the Bland-Altman plots. The intervisit test-retest variability was examined in 17 randomly selected glaucoma patients who underwent optic disc imaging weekly for 8 consecutive weeks. The intraclass correlation coefficients (ICC) and the reproducibility coefficients of the optic disc parameters were computed.

The OCT measured smaller optic disc and rim areas and greater cup volume, vertical cup-to-disc ratio and cup-to-disc area ratio than the HRT did (all with P<0.001). There were proportional biases in the Bland-Altman plots between OCT and HRT optic disc measurements except for rim area and cup-to-disc area ratio. The 95% limits of agreement of rim area ranged between -0.28 and 0.88 mm(2) before, and between -0.22 and 0.92 mm(2) after correction for ocular magnification. Optic disc assessment by spectral-domain OCT and confocal scanning laser ophthalmoscopy demonstrates poor agreement «Yang B, Ye C, Yu M ym. Optic disc imaging with spe...»3.