Differences of vertebral area in serial bone density measurements: a common source of potential error in...

Vertebral projected areas in serial BMD scans should not differ significantly to avoid measurement error due to apparent change in projected bone size. This criterion is rarely achieved in clinical practice. We analyzed 103 consecutive pairs of DXA reports to determine the frequency and magnitude of serial differences in vertebral area. Scans were performed at qualified community radiology sites and included if free from any technical errors, artifacts or rotation. We calculated the proportion of paired scans having at least 2 vertebrae differing in area by<2%,<3%,<4% or<5%. Using these differing sets of validity criteria, vertebrae not meeting the areal standard were removed form the analysis and the overall change in BMD recalculated. The new, recalculated BMD was compared to the original report to determine the frequency and magnitude by which the re-analysis would change the final report. Of the paired scans, 5%, 16%, 27% and 35% had all 4 vertebrae differ in area by less than 2%, 3%, 4% and 5% respectively. When only two vertebrae were required to meet acceptability criteria, 51%, 73%, 85% and 89% of scans met the 2%, 3%, 4% and 5% difference criteria. 11% of scans were non-comparable by even the least stringent criteria of two vertebra differing by<5% between scans. Re-analysis of BMD change in each group differed from the reported change by 0.012-0.015 g/cm(2). However, this amount was sufficient to change a clinical report from "significant change" to "non-significant change" in 26%, 27%, 21%, and 20% of scans in each of the four validity groups using a least significant change of 0.025 g/cm(2). Between 11%-17% of scans differed in the recalculated BMD change by an amount greater than the least significant change of 0.025 g/cm(2). Fewer serial BMD results were classified as non-acceptable when using the broader validity criteria of<5% area difference, but when corrected for areal differences, a similar and large proportion of scans would have a major change in the clinical interpretation of BMD change. These results do not change the interpretation of population BMD change in randomized trials but highlight the need for more caution in data analysis of serial densitometry results when used to make individual patient management decisions.