The diagnostic accuracy of selected neurological tests.

The diagnostic value and reliability of selected neurological clinical tests was studied in control subjects with normal neuroimaging (n=42), and subjects with a focal brain lesion (n=38). The items were studied by two examiners blinded to group membership and using standardized protocols, and subsequently by a neurologist who was not blinded to diagnosis. The positive likelihood ratios ranged from 1.06 (pronator drift) to 22.11 (single leg stance with eyes open, while the negative likelihood ratios ranged from 0.47 (tandem gait) to 0.97 (pupil symmetry). Three items (single leg stance - eyes closed - firm surface; single leg stance - eyes open - foam surface; and tandem gait) successfully distinguished between the two groups (odds ratio p<0.05). The inter-rater reliability was generally poor, with only tandem gait showing excellent agreement (kappa [K]=0.92). Tandem gait was the only item to show noteworthy agreement (K=0.93) between the examiners and the neurologist. The tests varied considerably in their ability to detect radiologically demonstrated structural brain lesions, and several items were poorly reproducible, questioning their value as part of a routine neurological examination.