This search was conducted according to the updated preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) guidelines (Appendix S1).[11] First, a comprehensive search for systematic reviews or meta-analyses comparing outcomes between bisphosphonate and placebo in patients with osteoporosis on Cochrane Library (Cochrane Database of Systematic Review) was performed on November 13, 2016. The methodological quality of the included systematic reviews was independently assessed by the first two authors using the Assessment of Multiple Systematic Reviews (AMSTAR) tool.[12] Thereafter, we identified relevant randomized controlled trials (RCT) through the examination of reference lists from selected meta-analyses.[13,14,15,16]

We performed an additional updated searching of primary studies on PubMed-Medline and Cochrane Library search in November 2016 using key terms (bisphosphonate OR osteoporosis OR fracture). An overview of the search strategy and general characteristics about the four included systematic reviews are summarized in Appendix S2. Inclusion criteria were as follows: (1) study was a RCT, (2) patients were above 50 years old, and (3) articles that provided the rate of osteoporotic fracture. The intervention of at least one study group included the following bisphosphonates: alendronate, etidronate, ibandronate, risedronate, clodronate minodronate, pamidronate, tiludronate or zoledronic acid. We restricted our review to English and Korean studies, owing to translation difficulties and lack of resources for review. Basic science articles, comments, and letters were all excluded. When a published, updated study involving the same cohort of patients was identified, only the latest update was included in the analysis.

The primary outcome of interest was the rate of osteoporotic fracture including all fractures. Osteoporotic fractures were categorized into vertebral fractures and non-vertebral fractures. Additionally, we performed a subgroup analysis according to the type of bisphosphonate.

For each selected study, the following data were extracted and entered in a spread sheet by the two reviewers: the family name of the first author, the year of publication, enrollment period, study design, number of patients, number of osteoporotic fracture, mean age at the time of intervention, and the duration of follow-up.

Two of the authors (BJH, YBH) independently evaluated the quality of all the studies, using the criteria described in the Cochrane Handbook for Systematic Reviews of Interventions. The criteria included the following 7 items: (1) random sequence generation (2) allocation concealment (3) blinding of participants and personnel (4) blinding of outcome data (5) incomplete outcome data addressed (6) selective reporting (7) other biases. We assessed publication bias with Begg's funnel plot and Egger's test.

For each study, we calculated the odds ratio (OR) with a 95% confidence interval (CI) by using crude 2×2 tables, whenever possible, from the comparative studies.[17] The Mantel-Haenszel method was used to calculate the OR due to zero values in any cell count in a table.[18] Heterogeneity between comparable studies was tested with the chi-square and Higgins I² test. P-value of less than 0.1 or an I² value higher than 50% meant significant heterogeneity and a random-effects model should be applied. There was significant heterogeneity between the included studies (P=0.001). Thus, we reported the data from a randomized-effect. All analyses were performed using STATA version 14.0 software (Stata Corporation, College Station, TX, USA).

This study was exempted from institutional review board review since it did not involve any human subjects.

A primary search of databases yielded 2,755 records and a total of 24 studies were included in the final systematic review. From the initial searches, studies selected for final inclusion or exclusion are displayed in the flowchart (Fig. 1). Twenty-four randomized placebo controlled trials investigating the effects of alendronate (6 studies), risedronate (5 studies), etidronate (4 studies), zoledronic acid (4 studies), clodronate (2 studies), ibandronate (1 study), minodronate (1 study), and pamidronate (1 study) were identified with a systematic literature search.[6,7,8,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39] The characteristics of studies included are summarized in Table 1. A final total of 39,197 patients were included in our meta-analysis: 21,335 patients in the bisphosphonate group and 17,862 patients in the placebo group.

The overall rate of osteoporotic fracture was 7.7% (3,036/39,197): 5.9% (1,268/21,355) in the bisphosphonate group and 9.9% (1,768/17,862) in the placebo group. The overall rate of vertebral fracture was 5.9% (446/7,585) in the bisphosphonate group and 10.3% (741/7,190) in the placebo group. The overall rate of non-vertebral fracture was 6.0% (822/13,750) in the bisphosphonate group and 9.6% (1,027/10,672) in the placebo group.

In the random effects model for all studies, the use of bisphosphonate was associated with a decreased risk of osteoporotic fracture (OR 0.62; 95% CI 0.54 to 0.71 P<0.001). In vertebral fracture, the use of bisphosphonate was associated with a decreased risk of osteoporotic fracture (OR 0.55; 95% CI 0.44 to 0.69 P<0.001) and non-vertebral fracture (OR 0.73; 95% CI 0.67 to 0.81 P<0.001) (Fig. 2). In the subgroup analysis, zoledronic acid showed the lowest OR with significance for overall osteoporotic fractures (Table 2).

In terms of the methodological quality, subjects were randomized by established allocation sequence, and investigators and research assistants were all blind to allocation. However, it is unclear whether the included trials met all the quality assessment criteria (Fig. 3). The Begg's funnel plot was symmetrical, and P-value for bias was 0.192 in the Egger's test for the included trials (Fig. 4).