Six RCTs (n=264) were included. The sample size ranged from 13 to 96 patients.
Inter-rater agreement was good for both study selection (kappa 0.85) and study quality (kappa 0.88).
In terms of study quality, 2 studies were blinded. There was complete follow-up in 4 studies.
Three studies assessed BMD at 3 months after surgery and all showed a decrease in both treatment groups.
There was significantly less periprosthetic bone loss at 3 months with bisphosphonates compared with the control; the WMD (n=152) was 3.3% (95% CI: 1.9, 4.7, P<0.01). The difference remained statistically significant at 6 and 12 months; the WMDs were 4.5% (5 studies, n=248; 95% CI: 1.6, 7.4, P<0.001) and 4.2% (4 studies, n=197; 95% CI: 1.5, 6.9, P=0.02), respectively.
Neither the type of bisphosphonate nor the quality score significantly altered the results.
The difference between bisphosphonate and control was significantly greater at 12 months (P<0.001) for cemented hip arthroplasties (WMD 7.5%, 95% CI: 4.3, 10.7) compared with uncemented hip arthroplasties (WMD 2.1%, 95% CI: 0.61, 3.6), but the results might have been confounded by the greater age of patients undergoing cemented arthoplasties (68 years for cemented versus 55 years for uncemented).
The difference between bisphosphonate and control was significantly greater at 6 months (P<0.001) for total knee arthroplasties (WMD 14.0%, 95% CI: 6.3, 21.7) compared with total hip arthroplasties (WMD 2.5%, 95% CI: 0.96, 4.1), but the difference was not statistically significant at one year.
The funnel plots showed no evidence of publication bias at 3, 6 or 12 months.