This meta-analysis of observational studies evaluates the effectiveness of ibandronate in preventing osteoporotic nonvertebral fracture. There are reasons why this study stands out from a previous systematic review evaluating the antifracture benefit of ibandronate [20]. First, this meta-analysis is supported by an updated literature search which identified a new study [29]. Updating systematic reviews is a recommended procedure in clinical investigation, as newly identified studies can change the conclusions of previous publications [30]. Disregarding new evidence can threaten the validity of systematic reviews and misleading healthcare professionals, patients, and regulatory authorities at the time of decision-making [31, 32]. Second, the previous systematic review performed a qualitative evaluation of ibandronate antifracture effect. This meta-analysis provides a weighted average of the results of all individual studies and conducts a sensitivity analysis to assess the robustness of the results, helping readers understanding the conclusions. Statistical synthesis of the evidence adds value to this research topic since it provides objective risk estimates rather than qualitative descriptions [33].

RCTs provide the most robust evidence regarding the efficacy and short-term safety of pharmacological interventions. Among their advantages, the strict inclusion/exclusion criteria, random allocation, continuous monitoring of patients, and precise definition of endpoints reduces the risk of bias and confounding [34]. However, real-world studies are essential to provide evidence of treatment effectiveness, by including patient populations that may be representative of clinical practice, larger sample sizes, and extended follow-up times when compared to RCTs. In this context, evidence of efficacy and safety of bisphosphonates from RCTs may not predict their actual effectiveness in clinical practice because of clinical, demographic, and suboptimal persistence/adherence differences between the populations. Therefore, there is a rational to conduct a meta-analysis of observational studies evaluating the effectiveness of ibandronate in preventing osteoporotic nonvertebral fractures. Moreover, meta-analyses of nonexperimental studies are frequently designed to explore eventual sources of heterogeneity among studies rather than to find evidence of causative associations [35].

These results suggest that ibandronate reduces the risk of nonvertebral fractures in general. Regarding this outcome, all the included studies evaluated ibandronate in reducing nonvertebral fractures as a once-monthly 150 mg oral regimen. However, the analysis of the results should take into consideration the different comparators used in the studies. The sensitivity analysis demonstrated that once-monthly 150 mg oral ibandronate reduced the risk of nonvertebral fractures compared with once-monthly 150 mg oral risedronate [15]. No differences were found when once-monthly 150 mg oral ibandronate was compared to weekly risedronate or weekly alendronate [28]. The comparison between ibandronate and “no treatment” returned a reduced risk, but without statistical significance [17, 27].

Although the results did not demonstrate an overall risk reduction of hip fracture, oral ibandronate seems to have a comparable risk of hip fracture to other oral bisphosphonates. No risk differences were found between once-monthly 150 mg oral ibandronate versus the other oral bisphosphonates (either once-monthly 150 mg oral risedronate, or weekly risedronate or weekly alendronate), nor versus “no treatment” [15, 27, 28]. However, the results suggest that IV zoledronate (once yearly) is more effective than IV ibandronate (once quarterly) to prevent hip fractures [16].

Both risedronate and alendronate are approved to prevent hip fractures in women with osteoporosis [36, 37]. Since the risk of hip fractures was comparable between once-monthly 150 mg oral ibandronate and the other oral bisphosphonates (once-monthly 150 mg oral risedronate, or weekly risedronate, and weekly alendronate), whether ibandronate could also be recommended to prevent such type of fracture may be matter of discussion. These results are in line with those from previous meta-analyses of RCTs, where no differences were found between these three bisphosphonates on the risk of hip fracture [38, 39]. Therefore, this meta-analysis of observational studies may contribute to clarifying the comparative effectiveness of ibandronate versus other bisphosphonates in the prevention of nonvertebral fractures in real-world clinical practice.

This study assessed hip fractures as the only outcome for site-specific fractures. Hip fractures are one of most severe type of osteoporotic fractures and are associated with higher mortality, reduced quality of life, and increased health resources consumption [40]. As hip fractures are an important endpoint considered in the design of RCTs of new antiosteoporosis drugs, studies assessing the effectiveness of the bisphosphonates in reducing this type of fractures in real-world clinical practice are valued by decision-makers [32, 41, 42]. Furthermore, only the study of O’Kelly et al. reported results for other type of fracture (wrist/forearm fracture) [29]. Given this, it was not possible to conduct a meta-analysis including only one study.

A strength of this meta-analysis is including only studies reporting results on definite nonvertebral fracture outcomes, and not studies reporting results for surrogate-type endpoints (e.g., changes in BMD). Although the incidence of osteoporotic fractures is usually associated with changes in BMD, this is still a predictive risk factor that do not replace the measurement of definite fracture outcomes [34, 35]. The marketing authorisations of bisphosphonates to prevent fractures in osteoporosis were granted based on statistically significant risk reductions of definite fracture outcomes (rather than solely on changes in BMD) [36]. Therefore, the summary of product characteristics of ibandronate highlights that the efficacy of the drug is yet to be proved in the prevention of nonvertebral fractures.

Some limitations must be considered. First, only 6 studies verified inclusion criteria. There are several observational studies evaluating the effectiveness of bisphosphonates published in the scientific literature, but without disaggregating the results at drug level [37]. A thorough literature search has been conducted but only a small of group studies evaluating the risk of nonvertebral fractures associated with ibandronate were found. Second, the studies used different control groups. Not only different bisphosphonates were used as active comparators but also three studies comparing ibandronate with “no treatment” adopted different approaches when defining the control groups. In the study of Siris et al. patients with a medication possession ratio (MPR) of less than 50% were considered the referent ‘‘untreated’’ population [17]. The studies of O’Kelly et al. and Abelson and et al. evaluated the risk of nonvertebral fractures by comparing the incidence of those fractures during an initial 3-month period of therapy with the incidence of fractures observed in the subsequent first year of therapy [27, 29]. Authors argued that the baseline fracture incidence during the initial 3 months of therapy may accurately reflect the underlying risk of the cohort, as bisphosphonates may take up to 3 months to reach maximum effectiveness [38]. The different design approaches between the studies using “no treatment” as a control group may be the reason why they did not reach concordant nonvertebral fracture risk estimates, resulting in a non-statistically significant risk reduction in the meta-analysis. Third, the results were only stratified according to different comparators, with all studies having the same design and risk of bias assessment. The influence of additional risk factors in the results needs to be accounted for when analysing these findings. However, most studies did not detail the results according to known risk factors for nonvertebral fractures, like age, prior use of bisphosphonates, and previous fractures. Fourth, there is no currently satisfactory methodology to perform meta-analysis including a small number of studies, particularly because between-studies heterogeneity cannot be reliably estimated. The initial analysis heterogeneity was most likely underestimated by the frequentist method. Therefore, additional analyses were conducted to better account the uncertainty. The Knapp–Hartung method (combined with the Paule-Mandel estimator) is recommended by Cochrane Collaboration for this specific scenario and the Bayesian random-effects meta-analysis is an alternative approach when having a low number of studies [25]. Fifth, the covariates used for the adjustment and approaches to control confounders varied significantly between the studies. Therefore, studies were assessed as having low methodological quality. Sixth, there are additional differences between the studies regarding their methods and demographics. Three studies included patients aged ≥ 45 years old, while three considered patients aged ≥ 65 years old. Additionally, sample sizes, median time of follow-up, and proportion of patients with previous fracture varied greatly between studies. While the heterogeneity found in these meta-analyses was not excessive (maximum I2 = 27%), the possibility of studies’ demographic and methodological discrepancies have influence in the risk estimates should not be ruled out. Seventh, we only searched Pubmed and Embase. However, these are the two most comprehensive biomedical literature databases and comprised more than 35 and 44 million citations, respectively [49, 50]. Eight, no protocol of this study was previously published, which could increase transparency and reduce potential for bias. Nonetheless, guidance to conduct and report this meta-analysis was followed.