Studies were identified through a literature search of major databases. For this purpose, an expert documentarian worked and verified them. The terms used to capture and the results are shown in Table I (available in the electronic version of this article). The following databases were screened: Medline (from 1996 to July 20, 2010), EMBASE (1991 to July 20, 2010) and Cochrane Central. We also searched the abstracts of the EULAR and ACR congresses of the years 2008, 2009 and 2010. There were no language restrictions. All references retrieved were handled in EndNote X3 (Thomson Reuters). Finally, a manual search was conducted by reviewing the references of studies included.

The studies retrieved by the above search strategy were included if they met the following inclusion criteria. Patients had to be ≥18 years with OP of non-malignant etiology and treatment for OP with one of the BP (alendronate, etidronate disodium, ibandronic acid, pamidronate, risedronate and zoledronic acid) or strontium ranelate. We included meta-analyzes, SRs and CT. The comparator group in the RCT could be placebo or active drug. It should analyze the presence of ONJ as a measure of outcome. We excluded animal and basic science studies.

Two reviewers independently examined the titles and abstracts of the articles retrieved with the selection criteria. This process was conducted in 20min sessions. The 2 reviewers extracted the data from studies included using ad hoc standardized forms. All items collected were double and independent. One reviewer entered data forms into a spreadsheets. If, in doing so, the reviewer noticed a discrepancy between the information and the other reviewer, then a consensus was reached by reviewing the original article or by asking a mentor. Items which did not meet all the inclusion criteria or had insufficient data were excluded. To classify quality, we used a modification of the Oxford Centre for Evidence-based Medicine Levels of Evidence11 in which the levels of evidence correspond to: (1a) SR of RCTs with homogeneity, i.e. including studies with comparable results and the same objectives; (1b) individual RCT (with narrow confidence intervals); (1c) proven through clinical practice rather than experimentation; (2a) SR of cohort studies with homogeneity and (2b) individual cohort randomized clinical trials of low quality (<80% follow up); (2c) research studies and ecological “Results”; (3a) case–control studies with homogeneity; (3b) individual studies of cases and controls; (4) case-cohort and low-quality case–control series; (5) expert opinion without explicit critical appraisal, or based on physiology, bench research or “first principles”. Evidence tables were constructed. The meta-analysis was planned just in case there was sufficient homogeneity between the studies included.

In connection with the SR sample, one12 identified, until January 2006, 368 cases of ONJ with 15 (4%) occurring in patients with OP treated with oral BP. Another,13 based on extensive case findings, case series and observational studies in OP patients treated exclusively with BP, from 2003 to September 2006, identified 26 cases of ONJ in 11 studies, reflecting a very low incidence of ONJ compared to 190 million estimated prescriptions of BP for OP. Patients were mostly women (87%), with a highly variable age (39–83), and usually had a history of dental surgery or trauma (80%).

In 2007 a report of the American Society for Bone and Mineral Research (ASBMR) was published on BP-associated ONJ in which there was a SR of ONJ in OP patients treated with BP orally, which contributed a total of 57 cases, most associated to intravenous BP (iv).10 The report suggested that the actual incidence may be higher than those published so far, due to confounding factors.

During the following years, there was a progressive increase in cases of ONJ in patients receiving BP for the treatment of OP. Until September 2009, another review9 found 2408 cases of ONJ associated with BP, of which 178 corresponded to patients with OP.

Based on the results of these SR,9,10,12–17 it must be considered that in less than 25%, ONJ occurs spontaneously, is most frequently localized in the mandible (65%–70%) and can simultaneously affect both maxillary bones (5%–10%). Factors associated with its occurrence were mainly dental procedures and to a lesser extent, defective dentures, trauma, bone exostosis, drug therapy (corticosteroids), and so on. For some studies there was no clear time dependence of exposure,13 but others showed a minimum of 2 years and a mean of 4.6 years of exposure to BP for the development of ONJ.10,17 About 80% of patients were over 60 years and were predominantly female.13 There have been no cases of ONJ associated with the use of strontium ranelate in OP.

Only some of the latest clinical trials referred to jaw osteonecrosis as a possible adverse event of OP treatment with BP. In general, clinical trials measure efficacy and safety of at least 1 year in duration. Due to frequency and volume of patients we included those performed with zoledronate. Treatment for 5 years with 5mg iv annual zoledronate vs placebo in 2127 patients over 50 who had had a hip fracture in the last 3 months did not show any cases of ONJ.21 Neither reported any cases of ONJ, although active surveillance was performed in 66 patients with thalassemia and OP treated for 1 year with 2 different regimens of iv zoledronate (4mg/6 months or every 3 months) vs placebo19 or in 545 patients with glucocorticoid-treated OP for one year with a dose of 5mg zoledronate vs iv. 5mg oral risedronate.23 Finally, neither showed any cases in a RCT conducted in 7765 women between 65 and 89 years of age with postmenopausal OP, which studied the efficacy and safety of annual iv zoledronate 5mg vs placebo.20 The occurrence of ONJ was not a primary endpoint of the study, but with the ongoing study it was decided to assess maxillofacial adverse events in an objective and independent manner to identify a possible association. After reviewing the database for adverse events, 101 patients in the zoledronic group and 127 patients in the placebo group who had maxillofacial potential adverse events that met the criteria for an investigation were identified and subsequently attributed to ONJ. The independent adjudication committee consisted of 5 experienced specialists and decided that 2 patients (one from each treatment group) had an event that met the definition criteria of ONJ. In both cases there was a delay in healing after surgery.26

ONJ was also a primary endpoint in the trials with ibandronate. In 2 of them there was no communication of any cases of ONJ. One compared 2 intermittent iv regimens with daily oral ibandronate for 2 years in 1395 women between 55 and 80 years with posmenopausal22 OP and the other studied the efficacy and safety of ibandronate (150mg PO month) vs placebo in healthy young men with OP.24

The FLEX18 study compared the effect after treatment and was discontinued in 1099 women who had been treated with alendronate in the FIT study for 5 years. Although there was a significant number of women treated with alendronate for many years, there were no cases of ONJ. Among women who discontinued therapy after 5 years and followed until 10 years after treatment stop, there were no differences in the occurrence of ONJ.

In most cases of ONJ reported in BP treated OP, patients had undergone dental procedures of different kinds. An open clinical trial compared the outcome of 220 dental implants placed in 50 women with posmenopausal OP.25 Half were treated with weekly BP (alendronate 70mg or risedronate 35mg) and half received no treatment. The main endpoints were loss of implants and the occurrence of ONJ. There were no cases of ONJ or differences in the loss of implants.

Finally, there also were no cases of ONJ associated with BP in the meta-analyses of controlled clinical trials lasting at least one year for postmenopausal OP treatment with and risedronate and alendronate.17,18