Ultrasonic assessment is a new approach to assess both quality and den
sity. Two ultrasonic parameters are measured on the os calcis: the att
enuation or broadband ultrasound attenuation (BUA) and the velocity or
speed of sound (SOS). The interunit variations in vitro and in vivo o
f an ultrasound instrument, the Lunar Achilles(R) system, used in a Fr
ench multicenter study named EPIDOS, were calculated and the stability
of these intruments over a 12-month period was evaluated. A third par
ameter called ''stiffness index,'' calculated from the SOS and BUA, wa
s also used in this study. The average CV in vitro for the BUA and SOS
was 0.92% and 0.12%, respectively, and the average CV in vivo for the
BUA, the SOS, and the stiffness index was 1.83%, 0.23%, and 1.9%, res
pectively. The interunit (or inter-machines) variations were calculate
d by a one-way analysis of variance. We detected small but significant
measurement differences among centers on a phantom for both SOS (maxi
mum significant difference 0.4%) and stiffness (maximum significant di
fference 3.5%) but not for BUA. Similar differences were found in vivo
. The precision over 12 months of the interunit variations in vitro wa
s evaluated by measuring a single phantom traveling from one center to
another several times. The range of the CV for the BUA (1.54- 0.51%),
for the SOS (0.25-0.14%), and for the stiffness index (2.26-1.10%) ar
e explained in part by technical failures. The variation among the fiv
e Achilles was estimated by the combined CV which was 1.42% for the BU
A, 0.32% for the SOS, and 2.33% for the stiffness index. In conclusion
, our findings indicate that equipment from one manufacturer appears t
o be consistent between machines for the BUA, but not completely for t
he SOS. The results for this stiffness index are necessarily influence
d by both SOS and BUA. The shortterm and long-term interunit precision
is good, both in vitro and in vivo. Such results provide increased co
nfidence in multicenter clinical trials where ultrasonic data are pool
ed.