Abstract
Purpose: :
Visual field asymmetry analysis has long been known to enhance diagnostic sensitivity and specificity for glaucoma diagnosis (Sponsel 1986). A new Pearson Product Coefficient asymmetry algorithm has been incorporated into the software of forthcoming HRT III software, prompted by observed weaknesses of the prevailing Gold–standard unilateral nerve fiber layer analysis method among some non–European populations. This study was necessary, since the desired positive and negative predictive values in a screening setting differ from those desired among clinical patients. Moreover, modern screenings refer with a visual field loss bias, while intra–clinical patient referrals tend to be derived from abnormal disc appearance or ocular hypertension.
Methods: :
200 days of screening for glaucoma were completed over a 2–year interval at multiple sites in Bexar County, Texas. Initial screening was performed by trained technicians with history, BP, acuity, and FDT C20–2. All screenees with ≥2FDT misses in either eye underwent HVF SITA 30–2, HRT scanning laser disc/nerve fiber layer analysis, and tonometry.
Results: :
4550 individuals were screened. Among those who screened positively for glaucoma or other eye disorders, bilateral Humphrey Visual Fields (HVF 24–2) were completed and classified in masked fashion (see Lindhorst, et al, 2006). Paired fields were categorized into five groups; I: both eyes normal, II: both eyes symmetrically abnormal, and III–V: asymmetric HVFs differing by 1, 2, or 3 HPA category levels, respectively. Among positive screenees, 97 provided bilateral HRT scans with intrinsic coefficient of variability of <50 microns in both eyes. Among these, HVF yielded 14, 22, 41, 13, and 7 individuals in categories I–V respectively. A sensitivity of 41%, specificity of 93%, positive predicative value (PPV) of 97% and negative predicative value (NPV) of 21% were obtained when the clinically–appropriate 25% asymmetry cutoff was used. Applying a more conservative HRT III asymmetry cutoff of 30% (to limit false–positive outcomes in a mass screening setting), sensitivity decreased to 34%, and NPV to 20%, while specificity and PPV increased to 100%.
Conclusions: :
For population screening, low specificity is paramount. With the new HRT III asymmetry algorithm, non–subjective screening on the basis of disc and peripapillary asymmetry may provide diagnostic cutoff criteria commensurate with superthreshold automated perimetry.
Keywords: imaging/image analysis: clinical • visual fields