ACCURACY, REPRODUCIBILITY, AND VARIABILITY OF HAND-HELD DYNAMOMETRY IN MOTOR-NEURON DISEASE

A spring-loaded device that ''breaks'' at preset forces was used to as sess readings obtained by hand-held dynamometry by three raters with v arying experience in the method. Overall accuracy (3%), but not reprod ucibility or variability, was improved by greater experience. Readings obtained jointly by three raters had 53% greater variability than tho se obtained by a single rater. Nine muscle groups in 19 patients with motor neuron disease were assessed at 10 sessions (three replications per session) over six days by the experienced rater. Muscle force was expressed relative to that of 22 matched normal controls. The reproduc ibility was good with a mean % difference of 13.2 and repeatability co efficient of 2.17 kg-force for readings six days apart; the overall co rrelation coefficient was 0.98. The mean coefficient of variation (CV) of 10 readings was 9.9%. The poorer reproducibility and greater varia bility seen in clinically weaker muscles may account for differences i n patients with bulbar palsy and classical amyotrophic lateral scleros is; the degree of spasticity had no effect. The rater was estimated to contribute 37% of the total variability when testing patients. The us e of a composite score by combining normalised dynamometry readings of eight limb muscles improved mean % difference to 6.7 and mean CV to 5 .8%. The reproducibility and variability of hand-held dynamometry read ings obtained by a single rater compare well with those of fixed devic es. Readings from-single raters, irrespective of experience, have simi lar reproducibility and variability. If, however, multiple raters are used in longitudinal assessments of individual patients, as occurs in clinical trials, the variability of their combined readings should be estimated when calculating the sample size required.