The use of visual field indices in detecting changes in the visual field in glaucoma

We present results from 64 glaucoma patients and glaucoma suspects followed up for a median period of 7.4 yr who had a median of seven examinations using Program 31 on the Octopus perimeter. The patients also had manual visual field recorded on either the Tubinger or Goldman perimeter during the same period. By examining all manual fields over the follow-up, we classified 37 patients as deteriorating and 27 as nondeteriorating by using predetermined field criteria which we believed to be clinically significant. In a masked fashion, the indices mean defect (MD) and corrected loss variance (CLV), in addition to the three cluster analysis indices SIZ, CLUS, and PCLUS were computed for each patient and regressed on time. When a significant positive index/time slope (P < 0.05) was defined as indication of deterioration, all indices had remarkably poor sensitivities because their slopes did not reach statistical significance in the great majority of patients. When, regardless of statistical significance, positive slopes were defined as indication of deterioration and negative slopes as nondeterioration, the most sensitive index, PCLUS, still had a sensitivity of less than 65%. The indices were better in detecting the presence or absence of visual field deterioration in fields that were initially normal than in those that were initially abnormal. Since the testing modalities of manual and automated perimetry are different, our study was not designed to compare the sensitivity of one technique over the other. Our study does, however, demonstrate that the indices used currently may not be clinically reliable in the assessment of changes in the visual field.