In 1985 Mayhew and Rothstein published work on measurement of muscle performance with instruments. In their book emphasis was placed upon the issues of reproducibility and validity.
Most of the problem have now been solved, however, a few have not been circumvented. Advances in engineering and computer science have made systems more versatile and much more accurate (whilst at the same time holding back some systems). Eccentric activity is still a point of contention but the inclusion of gravity correction strength figures and ratios are no longer misused. Data smoothing options have ruined the 'pure' world of isokinetics (but have made it easier for day to day use). Results other than just pure strength have gone to further improve the situation but with so many results available it becomes harder to know which to look at!
Without reproducibility the issues of validity can be ignored because if data is not reproducible it can not be viewed as valid. There are 6 potential sources for error.
Machine linked inconsistencies (calibration)
Subject variations (motivation etc.)
Testing procedure errors (poor/inconsistent stabilization)
Protocol variations (different rest periods etc.)
Inter-tester reliability (difference between examiners)
Data processing factors (smoothing etc.)
Machine Linked Inconsistencies
There are 3 main problems with isokinetic test results for different models or systems.
Differences between individual machines of the same model type (software especially)
Differences between tests on the same machine during normal operation
Different models or systems
Most of us rely on the company that supplied our machine to calibrate it at necessary intervals. Normally calibration includes the force (moment) exerted on the dynamometer arm and recorded by the sensors as well as the angular position and velocity of the arm. To these, calibration of the damp setting should be added for absolute certainty.
Calibration of the amplifiers and analogue to digital converters should be performed. I hasten to add, these are rarely necessary. A detailed description of measuring the performance of an isokinetic dynamometer is given by Farrell and Richards (1986) and Bemben et al. (1988).
Unfortunately it is not always apparent how often the modules should be calibrated and manufacturers are not keen to divulge the differing extents to which the various sensors change over time. This issue has been examined by Olds (1981) and Sinacore et al. (1983). At this time they recommended calibration should be carried out on the moment measuring model every testing day and the velocity of the arm, every test velocity. The extent to which modern systems require calibration is difficult to tell, however, if you intend to do research on your machine you should learn how to calibrate and get the appropriate tools. The level of accuracy before compromise of interpretation is a difficult question. Some authors claim 10%, some 3%.
Stability (the machines own consistency from measurement to measurement) is dependent on the length of the period between measurements as the machine is likely to experience drift (deviations away from its own norms). Assessing stability is difficult and requires an intricate knowledge of the calibration process. Timm et al (1992) studied this and found that they were unable to demonstrate high enough consistency when calibrating the machine to actually assess stability.
This describes the ability of a tester to test a subject reliably. This has been examined at by Molczyk et al. (1991) who looked at testing using the same tester and subject on the same type of machine but at different sites. They found comparable strength and endurance values could be obtained. However, Byl et al. (1991) tried a similar study but with more testing speeds and found that the coefficient of variation between the machines would suggest that testing on the machine used originally would be more appropriate. This does appear to be specific to different tests as Byl and Sadowski (1993) have found that there are better intra-class correlation coefficients for the trunk muscle performance tests (flexion, extension, rotation and lifting).
This describes the difference between different types of machines. For more information see:
Francis and Hobler (1987) Cybex versus Lido
Reitz et al. (1988) Cybex versus Kin-Com
Gross (1991) Cybex versus Biodex
Timm (1990) Cybex versus Merac
The rule of thumb is findings obtained on one system cannot be compared to another.
These factors are extraordinarily difficult to assess quantitatively and are hard to assess and open to interpretation qualitatively. There are 3 factors to consider when performing repeated or multiple measures: