Background: The assessment of BMD change is currently based on the “least significant change” (LSC) rule, but this rule has many methodological deficiencies, and often generates wrong decisions. Here, I propose a new approach for assessing BMD change tailored to each individual.
Methods: For each individual, I estimated the 'true' average and standard deviation of BMD change, using the individual's measured change, the mean change in the population, and the intrasubject and intersubject variance. This approach implicitly removes the impact of regression-toward-the-mean effect. The probability of a real change in BMD is then estimated for each individual.
The longitudinal data from the Study of Osteoporotic Fracture (SOF) were used to test the utility of the new approach. Femoral neck BMD (FNBMD) was measured at baseline and subsequent visits using a densitometer (Hologic QDR1000 or QDR2000). From a published analysis, the measurement error of FNBMD on QDR1000 was 0.022 g/cm2, and the LSC threshold was 6.36%.
Results: The analysis was based on 5797 women aged 65 yr and older at baseline. The average (SD) baseline FNBMD was 0.65 (0.108) g/cm2. During an average duration of 3.5 years (range: 1.8 to 5.1) of follow-up, the annual rate of FNBMD change was -1.01% (SD 2.2%), but the adjusted average was 0.93%. Using the LSC rule, only 68 women (1.2%) were classified as 'decrease', and the rest was either 'unchanged' (98.3%) or 'increase' (0.5%). The new approach detected significant BMD change in 91% (n=5282) of women; of which, 42% (n=2447) showed a loss of at least 1% per year.
Conclusion: The approach introduced here allows the assessment of bone loss to be individualised rather than the one-size-fits-all LSC approach, and helps clinicians better identify people at high-risk of osteoporosis and fracture.