DECLINING BONE LOSS RATE VARIABILITY WITH INCREASING FOLLOW-UP TIME

Bone loss rates are believed to vary substantially among post-menopausal women. The belief, however, might be largely based upon comparisons of loss rates measured with considerable error. This issue arises because the precision of the bone mass instrumentation (1-2% errors in measuring bone mass) is similar in magnitude to the loss rate (typically 1-2% of bone mass per year). The 'true' variability in loss rates, however, can be estimated by adjusting for instrument errors. An equation is presented which estimates the true variability in loss rates from a study sample. The adjustment was examined using a cohort of post-menopausal, Japanese-American women living in Hawaii. The study examined the calcaneus, distal, and proximal radius sites. The results suggest that measurement errors did not markedly inflate the bone loss variability. Even after adjustment for measurement errors considerable variability in loss rates remained. The variability was examined for various follow-up durations. Both the observed and adjusted loss rates had smaller standard deviations over long intervals than over short intervals. This result suggests that the more extreme rates of change in bone mass over a year or two were often not sustained. The implications of this finding for sample size calculations in longitudinal studies are examined. The results also indicate that the length of follow-up and instrument precision should be taken into consideration when comparing the proportion of fast losers between populations.