The reliability of isokinetic dynamometers is extremely high. The studies which have examined the accuracy of peak torque, work and power have shown correlation coefficients between 0.93 and 0.99 (Magnusson et al. 1990, Montgomery et al. 1989 Bemben et al.1989 are just a few of the independent articles available on this topic). The reliability of testing various joints is discussed below:-
Shoulder
Greenfield et al (1990) and Perrin (1986) have demonstrated that internal and external rotation of the shoulder show greater reliability than flexion and extension. The reliability of shoulder extension and flexion testing is best when performed in a neutral position with reliability decreasing with shoulder abduction (Chan and Maffoulli, 1996). Trunk stabilization is important for reliable measures but nay have to be compromised for subjects safety (Frisiello et al, 1994).
Knee
Very high reliability as established by Kannus (1994). However, eccentric testing shows greater variability than concentric testing (Kramer, 1990).
Ankle
Correlation coefficients for concentric actions vary from 0.67-0.94 with little data available for eccentric actions (Cawthorn et al, 1991, Wennerberg, 1991, Karnofel et al, 1989).
Wrist
Lower reliability probably due to small ROM and torque production (thus variations in small torque values are obviously magnified particularly if the dynamometer does not register decimal places) (Chan and Maffulli, 1996).
Hip
Low reliability probably due to large ROM (Chan and Maffulli, 1996) However, Suomi et al (1993) did find correlation coefficients between 0.98 and 0.99 for hip abduction testing.
Elbow
Low reliability 0.72-0.83 (Griffin, 1987).
In short Chan and Maffuli (1996) have concluded that more advanced isokinetic machines do not produce higher reliability. Variables with greater numerical values e.g. peak torque of the knee show higher reliability than ones with lower values e.g. peak torque of the elbow and concentric results show greater reliability than eccentric ones.