Fifteen trials (2,634 participants, range 30 to 643) were included in the review.
There were statistically significant reductions in vertebral fractures in patients who received bisphosphonate therapy compared to placebo-treated patients (RR 0.80, 95% CI 0.69 to 0.94; Ι²=0%; seven trials; NNT=166.6). For the analysis by drug type, only zoledronic acid (three studies) was shown to be associated with significant reductions in fractures (RR 0.77, 95% CI 0.65 to 0.91; Ι²=21%; NNT=14.9).
Reductions in risk of osteoporosis were observed for patients who received bisphosphonate therapy compared to placebo (RR 0.39, 95% CI 0.28 to 0.55; Ι²=0%; four trials; NNT=2.82 patients). There was a significant reduction in risk of osteoporosis from zoledronic acid (RR 0.36, 95% CI 0.16 to 0.83; one trial; NNT=2.68 patients), alendronate (RR 0.44, 95% CI 0.29 to 0.67; Ι²=0%; two trials; NNT=3.06) and clodronate (RR 0.31, 95% CI 0.15 to 0.64; one trial; NNT 2.49).
Bisphosphonate-treated patients had increases in bone mineral density observed in the lumbar spine (5.18±3.38), total hip (2.56±0.89), trochanter (3.70±0.82) and femoral neck (2.35±1.16). Patients who received placebo had observed decreases in bone mineral density at all classes of evaluation.
Treatment with clodronate was significantly associated with higher risks of serious adverse effects (grade 3 to 4) compared to placebo (RR 2.72, 95% CI 1.71 to 4.31; one trial) and alendronate was associated with reduced risk of grade 3 to 4 adverse events (RR 0.60, 95% CI 0.40 to 0.88; Ι²=0%; two trials).
There were no significant differences between bisphosphonate-therapy groups and placebo-treated groups for cardiovascular events, cancer risk and gastrointestinal events. Zoledronic therapy was associated with significantly increased risks of musculoskeletal pain (RR 1.35, 95% CI 1.01 to 1.80; four trials), anaemia (RR 1.38, 95% CI 1.01 to 1.90; three trials), fatigue (RR 1.19, 95% CI 1.01 to 1.41; four trials) and fever (RR 1.67, 95% CI 1.15 to 2.43; three trials).
Visual appraisals of funnel plots and the results of Begg and Egger tests showed no evidence of publication bias.