N J Ellish1, E J Higginbotham. 1. Department of Ophthalmology, University of Maryland, Baltimore 21201, USA.
Abstract
OBJECTIVE: To examine the effects of the gold standard on sensitivity, specificity, and area under the ROC curve when determining the validity of a screening test. METHODS: Fifty-eight people were recruited from a Veterans' Administration eye clinic. Screening included testing with the 26-point oculokinetic perimeter (OKP). We used different gold standards to categorize patients as positive or negative for glaucoma. The glaucoma hemifield test (GHT) and corrected pattern standard deviation (CPSD) from the Humphrey visual field analyzer (HVF) were used as objective gold standards. Ophthalmologist review of the HVF, using two different methods for classifying visual field defects, was also used as a gold standard. RESULTS: Using the review of the HVF that included mild, moderate, or severe defects, the area under the ROC curve was not significantly different from 50%. However, for all the other gold standards, the area under the ROC curve was significantly different from 50%. CONCLUSIONS: In determining the ability of the OKP to differentiate between glaucoma and normal subjects, the choice of a gold standard led to different conclusions about its usefulness as a screening test for glaucoma. Furthermore, sensitivity and specificity varied widely depending on which gold standard was used, and what cutoff point was used for the OKP. Although clinician review of the HVF may be used when specific criteria are given to define visual field defects, using objective measures from the HVF is easier and will allow for better comparisons between studies.
OBJECTIVE: To examine the effects of the gold standard on sensitivity, specificity, and area under the ROC curve when determining the validity of a screening test. METHODS: Fifty-eight people were recruited from a Veterans' Administration eye clinic. Screening included testing with the 26-point oculokinetic perimeter (OKP). We used different gold standards to categorize patients as positive or negative for glaucoma. The glaucoma hemifield test (GHT) and corrected pattern standard deviation (CPSD) from the Humphrey visual field analyzer (HVF) were used as objective gold standards. Ophthalmologist review of the HVF, using two different methods for classifying visual field defects, was also used as a gold standard. RESULTS: Using the review of the HVF that included mild, moderate, or severe defects, the area under the ROC curve was not significantly different from 50%. However, for all the other gold standards, the area under the ROC curve was significantly different from 50%. CONCLUSIONS: In determining the ability of the OKP to differentiate between glaucoma and normal subjects, the choice of a gold standard led to different conclusions about its usefulness as a screening test for glaucoma. Furthermore, sensitivity and specificity varied widely depending on which gold standard was used, and what cutoff point was used for the OKP. Although clinician review of the HVF may be used when specific criteria are given to define visual field defects, using objective measures from the HVF is easier and will allow for better comparisons between studies.