In this review the advantages and disadvantages of different variants of compression testing of trabecular bone are discussed. Factors affecting the precision and the accuracy of mechanical properties of trabecular bone derived from such tests are analysed. Below are listed some of the important conclusions which can be drawn. Conclusions based on the author's previous studies (I-IX) are shown in italic.

1) Trabecular bone is a viscoelastic solid.

2) Stiffness, strength, ultimate strain, and failure energy are derived from a standard compression test to failure. Viscoelastic properties such as energy dissipation and the relative energy loss (loss tangent) can be obtained from non-destructive cyclic tests.

3) A non-destructive test conducted between a lower load level (zero strain) and an upper strain limit of about 0.8% specimen strain has been developed. The reproducibility of such a test technique has been assessed at different conditions. The reproducibility was best after a number of conditioning cycles in order to achieve a viscoelastic steady state. Orthotropic properties can be determined by non-destructive testing in different directions of cubic specimens. The reproducibility of such testing has been established.

4) The stiffness derived from non-destructive tests will be lower than that obtained from a destructive test because of the non-linearity of the load-deformation curve, but the stiffnesses will be strongly correlated.

5) Stiffnesses derived from destructive and non-destructive tests have an elastic and a viscoelastic contribution. Since the viscoelastic contribution is time dependent, the results will be dependent on strain rate and loading frequency in cyclic tests.

6) Standard testing of small trabecular bone specimens is associated with systematic errors. The most significant of these errors are believed to be related to trabecular disintegrity at the surface of the specimen and to friction at the specimen-platen interface. Structural disintegrity causes an axial strain inhomogeneity resulting in a overestimation of axial strain and a corresponding underestimation of specimen stiffness. Friction at the interface causes an uneven stress and strain distribution in the layer nearest to the test platen resulting in a overestimation of stiffness. The net result of these systematic errors is a 20-40 per cent underestimation of stiffness.

7) The specimen geometry has a highly significant influence on mechanical properties such as stiffness, ultimate strain and energy absorption. A cube with a side length of 6.5 mm and a cylindrical specimen with a length of 6.5 mm and a diameter of 7.5 mm are suggested as standard geometries providing comparable results.