Introduction

Fragility fractures, resulting from a low-impact event such as a fall from a standing height, typically affect the elderly and individuals with poor bone quality [1]. The rising aging population in high-income countries is responsible for increasing rates of fragility fractures and their clinical and functional consequences [1, 2]. Indeed, older adults with an osteoporotic fracture are at higher risk of an imminent fracture, which is a subsequent event within 1–2 years after an index fracture [3, 4]. Therefore, secondary prevention should be adopted after an initial fracture to reduce the further risk of an imminent fracture [3]. In this regard, physicians should evaluate the fracture risk using an assessment tool [5] and are strongly encouraged to treat patients immediately after a sentinel event [6]. Thus, for patients with a high fracture risk, pharmacological agents ought to be promptly recommended as they could improve bone mineral density and reduce the incidence of subsequent fractures [4, 7].

Specifically, vertebral fragility fractures (VFFs), which are among the most common fragility fractures, (i) are the primary risk factor for the occurrence of further fractures affecting either vertebra or other sites [8,9,10,11], (ii) are associated with an increased risk of morbidity and mortality, and (iii) represent a significant economic burden on healthcare systems [12,13,14,15,16]. Although at least one in five persons aged > 50 years has ≥ 1 vertebral fracture [9, 16], detection of VFFs may be uneasy due to ambiguous terminology and the lack of diagnostic standards [2, 11, 14, 15, 17, 18]. Only a third of VFFs come to medical attention [16, 19], which might lead to inadequate patient care [18].

The current meta-analysis systematically reviewed randomized clinical trials (RCTs) investigating the efficacy of drugs for secondary prevention of refracture among patients who experienced vertebral fractures. We summarized data of patients who were randomly assigned to the control arm (i.e., who did not receive drugs for bone fragility care except calcium and vitamin D supplements), which are known to be ineffective for preventing fractures in community-dwelling adults [20]. Thus, we considered patients who received placebo as a proxy of patients with unrecognized fragility fractures. The aim of this systematic review and meta-analysis was to measure the implications of vertebral fractures on the risk of a new fracture in patients with unidentified frailty.

Methods

Search strategy and selection criteria

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist [21] for conducting and reporting this study. We did a systematic search of the Embase, PubMed (Medline), and Cochrane Library databases to cover primary studies, as well as systematic reviews published up to February 2022. A hand-checking search on clinicaltrial.gov was performed to detect additional eligible studies. The search strategy included keywords and/or corresponding MeSH terms related to “vertebral fragility fracture” and “subsequent fragility fractures”. Further details and search terms are listed in Supplemental Material.

Studies were eligible if they (i) were RTCs, (ii) reported data on refracture following a radiographically detected index fracture (or morphometric fracture) among (iii) patients with a VFF who, being randomly assigned to the comparator arm, did not have drug treatment for bone fragility. Vertebral and non-vertebral refracture occurring at the time point following the index fracture were considered primary and secondary outcomes, respectively. Studies were excluded if they (i) were not published in the English language, (ii) did not report original findings (i.e., letters and case reports), (iii) did not involve patients with at least one VFF at baseline, or (iv) did not evaluate the refracture risk. When data were published more than once, the most recent and complete paper was selected. Besides, if multiple articles were published on the same trial, all articles reporting different follow-up periods or different refracture sites were included.

Two independent authors (GP and AB) screened titles and abstracts according to the search strategy and then assessed the full text of all potentially relevant studies. Discrepancies between readers were resolved by conference. From each included RCT, the following information was extracted: (i) first author, year, and country of publication; (ii) type and characteristics of the target population; (iii) type of refracture; (iv) follow-up period.

Study quality

The quality of each RCT was evaluated using the Cochrane risk of bias (RoB) tool for RCTs [22]. The following domains of the Cochrane RoB tool were appraised: selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective reporting), and other bias (such as funding bias). Each domain was classified as “high” or “low” RoB. The latter was considered “unclear” if the publication did not provide sufficient information. The overall quality of each included study was judged as high, medium, or low if no high (and fewer than three unclear), at most one high (or more than three unclear), or more than one high RoB was found, respectively.

Statistical analysis

Only patients belonging to the comparator arm for whom a radiograph exam was performed to clinically recognize the index fracture (i.e., fractured patients at risk of developing further fractures) were considered in the current meta-analysis.

The refracture outcome was measured through both rate and risk. The refracture rate was calculated as the number of patients assigned to the comparator arm who experienced a subsequent bone fracture over the person-years (PYs) from them accumulated. The refracture rate was expressed as per 100 PYs of follow-up and was presented with 95% confidence intervals (CIs). Unless directly reported in the original publication, PYs were derived as years accumulated during follow-up by patients at risk of developing the outcome (by right censuring observations at outcome occurrence or lost to follow-up when feasible). The refracture risk was calculated as the number of patients assigned to the comparator arm who experienced the outcome within two, three, and four years over the number of patients randomized to the comparator arm and expressed as a percentage with corresponding 95% CIs.

Estimates were summarized if at least three studies reported the estimate of interest. In the case of < 3 studies per category, data were aggregated into larger classes.

Subgroup analyses were planned for (i) the baseline number of VFFs and (ii) specific sites of no vertebral fracture during follow-up.

Heterogeneity between study-specific estimates was tested using Chi-square statistics [23] and measured with the I2 index (heterogeneity measure across studies) [24]. Studies were combined to obtain a summary estimate using the DerSimonian random-effects model [25]. Potential publication bias was visually and statistically identified through funnel plots and Egger’s test [26]. Furthermore, influence analysis was performed to assess the impact of a single study on the overall pooled estimates by omitting one study at a time.

All tests were considered statistically significant for p-values < 0.05. The analyses and the correspondent graphical visualization of forest plots were performed using RevMan V.5.4 (Nordic Cochrane Center, Copenhagen, Denmark) and R Statistical Software (v4.1.2; R Core Team 2021).

Results

As shown in Fig. 1, a total of 1184 papers were initially extracted. Overall, after exclusion through title and abstract screening and further inclusion through papers referenced by systematic reviews [17, 27,28,29,30,31] and hand search on the topic [32], a total of 40 RCTs were included [32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72]. Summary characteristics of the 40 RCTs included in our meta-analysis are given in Table 1. Twenty-five and fifteen trials were classified into the categories of high or medium quality, respectively. Almost all RCTs had at least one unclear risk of bias, primarily regarding selection bias (random sequence generation or allocation concealment tools) or other bias. Moreover, 14 RCTs [34, 38, 40, 42, 44, 45, 48, 49, 55, 59,60,61, 63, 68] had a high risk of bias, mainly due to attrition (incomplete outcome data) (Supplemental Material, Figure S1–S2). Overall, 9,891 patients with at least one baseline VFF who did not receive drug treatment for correcting bone fragility except for supplements with calcium and/or vitamin D were considered to be summarized in our meta-analysis. These patients accumulated 22,990 PYs for the risk of refracture. Then, the majority of the included studies were focused on post-menopausal women.

Fig. 1
figure 1

Flow chart

Full size image
Table 1 Characteristics of included randomized controlled trials
Full size table

Figure 2 shows the forest plots of vertebral and non-vertebral refracture rates. Vertebral fracture rates ranged from 3 to 73 refractures per 100 PYs, being an overall estimate of 12 (95% CI 9–16) refractures per 100 PYs and having very high between-studies heterogeneity (I2 ≥ 90%) (Fig. 2, upper box). During the follow-up period, an increase in the rate of vertebral fractures was observed, specifically 4 (2–7) per 100 PY for those with 1 fracture at baseline and 13 (6–29) per 100 PY for those with more than 2 fractures at baseline (Supplementary Material, Figure S3). nVFFs rates ranged from 2 to 19 refractures per 100 PYs, being the overall estimate of 6 (95% CI: 5–8) refractures per 100 PYs and having high between-studies heterogeneity (I2 ≥ 75%) (Fig. 2, bottom box).

Fig. 2
figure 2

The incidence rate of new vertebral and non-VFFs, expressed as number of patients every 100 person-years

Full size image

The summary fracture rate of (i) upper limbs was 1.3 (95% CI 1.0–1.7), 2.5 (0.7–9.4), and 0.8 (0.3–2.0) per 100 PYs for wrist, arm or forearm, and humerus, respectively (Supplemental Material, Figure S4); (ii) lower limbs was 0.9 (0.7–1.2), 0.4 (0.2–0.6), 0.4 (0.2–0.6), and 0.8 (0.4–1.7) per 100 PYs for hip, ankle, pelvis, foot or metatarsal, respectively (Supplemental Material, Figure S5); (iii) torso was 1.4 (0.5–4.1) (Supplemental Material, Figure S6); and (iv) other fractures was 2.6 (1.6–4.3) (Supplemental Material, Figure S7). The between-studies heterogeneity was reduced for upper and lower limb fractures (I2 < 75%).

Publication bias was detected for vertebral fractures (p-value = 0.0124, Supplemental Material, Figures S8–S9), although there was no evidence of the influence of any individual study (Supplemental Material, Figure S10) for both vertebral and non-vertebral refractures.

Figures 3 and 4 depict the forest plots of vertebral and non-vertebral refracture risks, respectively. Within two, three, and four years from the index fracture, 16.6% (13.1–20.8%), 25.7% (18.8–34.1%), and 35.1% (24.4–47.7%) of patients, respectively, experienced at least a vertebral refracture, while the non-vertebral fracture risk was 8.0% (4.6–13.4%) within 2 years and 17.4% (14.1–21.4%) over 2 years.

Fig. 3
figure 3

Risk of new VFFs

Full size image
Fig. 4
figure 4

Risk of new nVFFs

Full size image

Discussion

In this meta-analysis of 40 original RCTs including almost 10,000 untreated patients affected by vertebral fracture at baseline, we found that, on average, a new vertebral and non-vertebral fracture occurred every year in 12 and 6 patients, respectively, per 100 patients who had previously experienced a vertebral fracture. Our meta-analysis found that 16.6% and 35.1% of patients experienced at least a vertebral refracture within two and four years from the index fracture, respectively, while non-vertebral fracture risk was 8.0% within 2 years and 17.4% over 2 years.

Our results are confirmed by the findings in the literature. Particularly, the UK clinical guideline for the prevention and treatment of osteoporosis revealed a doubled fracture risk related to the prior fracture, particularly for > 1 vertebral fracture [73]. Then, a summary of the literature reported a strong association, approximately 4 times greater, between prior and subsequent vertebral fractures than those without prior fractures, particularly within the next 2 years after the initial fracture [74]. The risk of further vertebral fractures appeared to increase with the number of prior vertebral fractures [74].

Early identification of VFFs might present a real opportunity to reduce the risk of a subsequent fracture [75]. However, there is considerable evidence that vertebral fractures might not be properly considered by clinicians and are under-reported by radiologists [14, 76], who might not use specific terminology and not alert the referring healthcare professionals. Thus, standardized radiographic acquisition and unambiguous radiological interpretation could contribute to reducing the further risk of VFFs [76].

Because the patients included in our meta-analysis received a drug therapy that should be considered ineffective for the treatment of fragility, all these findings strongly suggest that recognizing fragility as the cause or concomitant cause of the vertebral fracture should be considered a priority for the secondary prevention of fracture.

Strengths and limitations

The findings of this study should be interpreted considering its limitations. First, the analysis was not patient-centered but instead used summary data; therefore, an accurate assessment might be lacking due to the nature of meta-analysis. Second, this systematic review selected RCTs that included patients who might have different characteristics compared to the general population. Third, our results were affected by high heterogeneity. Particularly, there are certain concerns as to whether findings from selected studies could be combined into one conclusion since primary findings were obtained from studies including heterogeneous populations, definitions of vertebral fractures and adopting different study designs. Fourth, among trials in which PYs were not reported, we assumed that censorships occurred in a mean of half of the entire follow-up period and could estimate below or above the incidence rate. However, this assumption can reasonably be considered valid in the case of large data and/or time intervals of limited amplitude. At last, an unclear risk of bias was found in nearly all the included studies, primarily regarding attrition bias, and a high risk of bias was detected in 12 RCTs, mainly due to attrition bias.

Despite these limitations, this study had certain strengths. The exhaustive search strategy provided an overview of RCTs on the subsequent fractures among untreated patients with prior VFFs. In addition, the internal validity of the selected studies was assessed using the RoB tool for RCTs.

Conclusion

The average annual rate and short-long term risk of refracture among untreated patients with VFFs were estimated from this meta-analysis. Based upon the currently available evidence, further fractures are commonly observed in the following two years after the initial VFF. Early and accurate detection of VFFs should be conducted to reduce the risk of future fragility fractures and properly establish secondary prevention.