Using panel tests in screening for congenital colour vision defects

The sensitivity and specificity of the Farnsworth Panel D-15, Lanthony’s desaturated D-15 and Farnsworth-Munsell (FM) 100-hue panel tests were compared in up to 201 congenital colour defective (37 ± 15 years) and 230 colour normal (36 ± 14 years) males. Results were analysed using the colour difference vector technique of Vingrys and King-Smith (Vingrys & King-Smith 1988). Vector analysis yields three indices which can be used to quantify a cap arrangement: the C-index (severity), the S-Index (polarity) and the angle (axis). ROC analysis was performed on the C-index in order to identify the optimum fail criterion for each test. Our data indicate that the efficiencies of the panel tests (D-15 = 1.63, desat D-15 = 1.67 and FM 100-hue = 1.72) are much poorer than the Ishihara plate test efficiency (1.98) for screening defects of colour vision, indicating that panels need to be supplemented with plate tests in screening protocols. Bayesian probabilities indicate that panel tests are successful in detecting colour defective vision (Pr {defective given fail at criterion} ≥ 0.99) when a C-index of 2.0 is used on the D-15 panel, 3.0 on the desaturated D-15 panel and 2.2 on the FM 100-hue test; at these criteria all panels have an equal probability of predicting normal vision (0.99). We conclude that none of the panels is adequate for screening congenital colour vision defects but that the desaturated D-15 panel test is the best clinical test since it gives the highest efficiency for the shortest test tim