This study aimed to A) establish a protocol for measuring periprosthetic bone mineral density (BMD) of the proximal humerus following implant arthroplasty, and B) compare the differences in the periprosthetic BMD values 4–9 years after surface replacement and stemmed arthroplasties of the proximal humerus. The study design was of retrospective independent samples cohorts, of patients who had received a proximal humeral arthroplasty between January 1992 and December 1996 in a tertiary referral shoulder unit of a UK hospital. The exclusion criteria were A) patient unavailable for study, B) patient refusal, C) inadequate information of dependent and independent variables, or D) obvious measurement errors.

All available patients underwent DEXA scanning of the proximal humerus using a Lunar DPXL scanner fitted with Orthopaedic Hip software version 1.3. Replicable patient positioning with a special jig was used. On a predetermined format of Gruen equivalent zones and sub-zones, BMD values in gm/cm² were plotted. Cortical thickness on plain radiographs was recorded for each zone. Femoral neck DEXA scan was performed to obtain a proportionate value of BMD of the proximal humerus, in order to eliminate the effect of confounding variables. Confounding variables accounted for were age, gender, height, weight, activity level, indication for surgery, duration of implantation, dominance, type of arthroplasty (hemi or total) and use of cemented or uncemented stemmed implants. Statistical analysis was performed using Microsoft excel as well as SPSS software.

Initially, 58 shoulders in 52 patients were recruited. 6 patients declined to participate for ill health, 6 had moved out of the area and 8 did not attend or reply. Of the remaining 31 shoulders in 25 patients, 2 patients had an obviously erroneous DEXA reading. 29 shoulders in 23 patients were finally analysed, which included 10 male and 19 female shoulders. Average age was 67.5 years and average time since surgery was 6 years 2 months. Indications for surgery were RA in 14, OA in 8 and other reasons in 7. Of 29 shoulders, 20 received a total replacement, 9 a hemiarthroplasty.

The inevitable obliquity of the image caused some difficulties in maintaining accuracy and difficulties were observed due to limitation of the software to measure thin cortices and to distinguish between the cement and bone. There was no significant correlation between gross cortical thickness and BMD values, and the average periprosthetic BMD was 0.54 g/cm². Surface replacement implants were associated with relative preservation of proximal medial cortex and higher BMD values in this region. BMD values were consistently higher at the level of stem tip for the stemmed implants. No such phenomenon was observed for the surface replacement prostheses. Hemiarthroplasty was associated with relatively higher BMD values in the proximal medial cortical region than total arthroplasty.