Introduction

A previous fracture is well-documented as a strong risk factor for the occurrence of future fracture [1,2,3,4]. The risk for subsequent fracture is at its greatest immediately following a fracture [3, 5, 6]; incidence of fracture after a prior fragility fracture has been reported as 7.6% in the first year and 11.6% in the first 2 years post-fracture [7]. Indeed, it has been reported that approximately half of recurrent fractures occur within the first 2 years after a fragility fracture [3, 7]. To reduce the risk of further fractures, it is therefore important to initiate treatment promptly after fracture [8].

However, despite being at imminent fracture risk, most patients are not prescribed treatments for osteoporosis in the year following a fracture [9, 10]. The reluctance to prescribe anti-osteoporotic medications immediately following a fracture may be due to concerns that they could interfere with fracture healing [11,12,13]. These concerns derive in part from the theory that suppression of bone remodeling, the mechanism of action for antiresorptive treatments, could affect remodeling and repair at the fracture site [14, 15]. However, the current clinical evidence indicates that there is no apparent detrimental effect of antiresorptive agents on fracture healing [14,15,16,17]. Similarly, prior studies of anabolic agents have shown no adverse effect on fracture healing [8].

Romosozumab is a humanized monoclonal antibody that binds to and inhibits sclerostin, thereby increasing Wnt signaling and exerting a dual effect on bone, increasing bone formation while decreasing bone resorption [18,19,20,21]. In the Fracture Study in Postmenopausal Women with Osteoporosis (FRAME) and Active-Controlled Fracture Study in Postmenopausal Women with Osteoporosis at High Risk (ARCH) trials, 12 monthly doses of 210 mg subcutaneous romosozumab treatment led to a reduction in the risk of vertebral and clinical (a composite of nonvertebral and symptomatic vertebral) fractures, as compared with either placebo (FRAME) or alendronate (ARCH) [21, 22]. Furthermore, in both trials, romosozumab treatment resulted in substantially greater gains in bone mineral density (BMD) at the total hip, femoral neck, and lumbar spine, compared with comparator treatments.

In this post hoc analysis of the FRAME and ARCH phase 3 trials, we evaluate the efficacy and safety of romosozumab administration in two subgroups of patients who, during the first 12 months of either trial, sustained a nonvertebral or a new or worsening vertebral fracture (including symptomatic and radiographically detected vertebral fractures). We evaluate safety during the first year and through the maximum available follow-up antiresorptive period (up to 36 months).

Methods

Study design

The post hoc analyses reported here are based on data from the FRAME (NCT01575834) and ARCH (NCT01631214) phase 3, international, randomized, double-blinded, parallel‑group trials, as reported previously [21, 22]. Informed consent was obtained from all participants in the FRAME and ARCH trials. Briefly, in both trials, patients were randomized 1:1 to receive either 210 mg monthly romosozumab subcutaneously or comparator (FRAME, placebo; ARCH, 70 mg weekly oral alendronate) for 12 months (Supplementary Fig. 1). In ARCH, patients also received matched placebo for either romosozumab or alendronate. At Month 12 in FRAME, all patients entered an open-label period and received 60 mg denosumab subcutaneously every 6 months for a year, followed by a 12-month extension study in which patients continued to receive denosumab every 6 months; therefore, end of study represents a total treatment period of 36 months. After 12 months in ARCH, all patients received open-label weekly oral alendronate (70 mg) until end of study (median total treatment period 33 months). On entering the open-label periods, blinding to the initial treatment was maintained in both studies.

Both trials enrolled postmenopausal women aged 55–90 years. Patients in FRAME had a T‑score of − 2.5 to − 3.5 at the total hip or femoral neck. Patients in ARCH had at least one of the following: (i) a T‑score ≤ –2.5 at the total hip or femoral neck and either ≥ 1 moderate or severe vertebral fracture, or ≥ 2 mild vertebral fractures; (ii) a T‑score ≤ –2.0 at the total hip or femoral neck and either ≥ 2 moderate or severe vertebral fractures, or a fracture of the proximal femur sustained 3–24 months prior to randomization. Exclusion criteria for FRAME included a history of hip fracture and any severe or more than two moderate vertebral fractures. In ARCH, patients were excluded for an inability to take alendronate oral tablets or contraindications to alendronate. Full exclusion criteria were described previously [21, 22].

Study population

In these analyses, we evaluated the efficacy and safety of romosozumab, administered in the immediate post‑fracture period, in two subgroups of women who experienced a fracture during the first 12 months of the FRAME or ARCH trials: those who sustained an on-study nonvertebral fracture, and those who sustained an on-study new or worsening vertebral fracture. Vertebral fractures were identified via lateral radiographs of the spine, performed at Month 12, using the Genant grading scale (grades range from 0 to 3, with higher grades indicating greater severity) at the central imaging vendor (BioClinica [now Clario]). Adjudicators were blinded to treatment group. New or worsening vertebral fractures were defined as an increase of ≥ 1 grade in previously normal, or previously fractured vertebrae, respectively.

Study outcomes

In the aforementioned subgroups of patients who sustained an on-study fracture by Month 12, we report the incidence and location of fracture and the percentage change in BMD from baseline (i.e., entry to the FRAME or ARCH trial) at Months 12 and 24 at the lumbar spine, total hip, and femoral neck in patients in who received at least one dose of study treatment. In the nonvertebral fracture subgroup, we additionally report the number and timing of romosozumab or romosozumab-matched subcutaneous placebo doses prior to and after the first occurrence of nonvertebral fracture.

The incidence (%) of treatment-emergent adverse events (TEAEs) in both subgroups is reported for Months 0–12 and Months 0–end of study; data through end of study are cumulative and include TEAEs that occurred during the double-blind and open-label periods. For patients in the nonvertebral fracture subgroup, TEAEs that occurred before the first nonvertebral fracture are excluded; in the vertebral fracture subgroup, all TEAEs that occurred by Month 12 are reported.

Overall TEAEs, serious TEAEs and TEAEs of interest are reported. Serious TEAEs were defined as TEAEs which met one or more of the following criteria: fatal, life threatening, requiring in-patient hospitalization or prolongation of existing hospitalization, resulting in persistent or significant disability or incapacity, or any other medically important serious event. TEAEs of interest, related to the fracture, included nonunion, malunion, delayed healing, chronic pain, and osteomyelitis. Adverse events are reported by preferred term, coded according to the Medical Dictionary for Regulatory Activities (MedDRA) version 19.1.

Study procedures and analysis

Full study procedures for the FRAME and ARCH phase 3 trials have been previously reported [21, 22]. Adverse events were reported by individual trial sites. Measurements of BMD at the lumbar spine and proximal femur were performed by dual‑energy X-ray absorptiometry (DXA) at baseline and every 12 months thereafter using Lunar or Hologic bone densitometers. DXA scans were processed and analyzed blinded to treatment at the central imaging vendor.

The analysis undertaken in this post hoc study was descriptive in nature and hypothesis testing was not undertaken. Descriptive statistics on continuous measurements comprise means, medians, ranges, and standard deviations (SD), while categorical data are summarized using counts and percentages. Data are standardly reported as observed data; missing new or worsening vertebral fracture status was imputed by carrying forward the last non-missing post-baseline value prior to the missing value.

Results

Patient baseline characteristics

In both FRAME and ARCH, the baseline characteristics of patients within each treatment group who experienced on‑study fracture within 12 months were well balanced (Table 1) and were largely similar to their respective full treatment populations [21, 22]. However, at baseline, patients in both FRAME and ARCH who subsequently experienced vertebral fracture by Month 12, had numerically lower mean (SD) T-score at the lumbar spine (FRAME: romosozumab: − 3.16 [0.76], placebo: − 3.25 [1.00]; ARCH: romosozumab: − 3.40 [1.21], alendronate: − 3.21 [1.25]) than their respective full treatment populations (FRAME: romosozumab: − 2.72 [1.04], placebo: − 2.71 [1.04]; ARCH: romosozumab: − 2.94 [1.25], alendronate: − 2.99 [1.24]) [21, 22]. Consistent with the inclusion and exclusion criteria for each trial, patients in ARCH had higher rates and grades of prevalent vertebral fracture and a higher baseline nonvertebral fracture rate than patients in FRAME. In addition, patients in ARCH were on average older and had generally lower T‑scores and higher Fracture Risk Assessment Tool (FRAX) scores.

Table 1 Baseline demographics and characteristics of patients with on-study fracture
Full size table

Incidence of first on-study fracture in FRAME and ARCH

Nonvertebral fracture subgroup

The incidence of nonvertebral fracture by Month 12 was numerically lower in romosozumab‑treated patients than those treated with comparator (Table 1). In FRAME, 56/3589 (1.6%) romosozumab-treated patients and 76/3591 (2.1%) patients treated with placebo experienced an on-study nonvertebral fracture by Month 12. In ARCH, 70/2046 (3.4%) romosozumab-treated patients and 95/2047 (4.6%) patients treated with alendronate experienced an on-study nonvertebral fracture by Month 12. The anatomic location of the first on-study nonvertebral fracture by Month 12 in FRAME and ARCH is shown in Table 2. In both trials, the radius was the most common nonvertebral fracture location.

Table 2 Incidence and locations of on-study fractures
Full size table

In patients who experienced nonvertebral fracture by Month 12, patients initially treated with romosozumab, then anti-resorptive therapy, had numerically lower rates of another nonvertebral fracture by end of study (up to 36 months) compared with patients initially treated with comparator (FRAME, 2/56 [3.6%] with romosozumab first then denosumab versus 7/76 [9.2%] with placebo first then denosumab; ARCH, 7/70 [10.0%] with romosozumab first then alendronate versus 12/95 [12.6%] with alendronate only; Table 2). Similarly, in patients with nonvertebral fracture by Month 12, rates of new or worsening vertebral fracture by end of study were numerically lower in those who initially received romosozumab, followed by antiresorptive therapy, compared with comparator (FRAME, 1/56 [1.8%] with romosozumab first then denosumab versus 3/76 [3.9%] with placebo first then denosumab; ARCH, 4/70 [5.7%] with romosozumab first then alendronate versus 8/95 [8.4%] with alendronate only; Table 2).

Overall, on-study nonvertebral fractures occurred with no specific pattern relative to the timing of the last administration of romosozumab prior to fracture. In patients who experienced on‑study nonvertebral fracture by Month 12, the timing of the last dose of romosozumab or romosozumab-matched placebo prior to the incidence of fracture was similar between treatment groups in both FRAME and ARCH (Supplementary Table 1). The mean and median number of days in study prior to the fracture occurrence was also similar between romosozumab and comparator groups. In both trials, a subsequent dose of romosozumab was received approximately 3 weeks following the occurrence of nonvertebral fracture (mean [SD] number of days in FRAME, 21.0 [26.6]; ARCH, 22.5 [34.4]). Further data regarding the timing of study treatment dosing before and after the first on‑study nonvertebral fracture are shown in Supplementary Table 1.

Vertebral fracture subgroup

Incidence of new or worsening vertebral fracture by Month 12 was numerically lower in patients treated with romosozumab than comparator-treated patients (Table 1). In FRAME, 17/3589 (0.5%) romosozumab-treated patients and 59/3591 (1.6%) patients treated with placebo experienced an on-study vertebral fracture by Month 12 (diagnosed by routine X-ray at Month 12 visit). In ARCH, 67/2046 (3.3%) romosozumab-treated patients and 101/2047 (4.9%) patients treated with alendronate experienced an on-study vertebral fracture by Month 12.

The total number of radiographic nonclinical vertebral fractures was numerically higher in both FRAME (romosozumab, n = 16; placebo, n = 49) and ARCH (romosozumab, n = 81; alendronate, n = 104) than the total number of clinical vertebral fractures (FRAME: romosozumab: n = 3; placebo: n = 21; ARCH: romosozumab: n = 12; alendronate: n = 20). The location of vertebral fractures by Month 12 in FRAME and ARCH is reported in Table 2; thoracic fractures, whether clinical or radiographic, were slightly more common than lumbar fractures.

In patients who experienced a vertebral fracture by month 12, the percentage who subsequently experienced another vertebral fracture by end of study was numerically lower in those who initially received romosozumab, then anti-resorptive therapy, versus comparator (FRAME: 0/17 [0.0%] with romosozumab first then denosumab, 7/59 [11.9%] with placebo first then denosumab; ARCH: 6/67 [9.0%] with romosozumab first then alendronate, 15/100 [15.0%] with alendronate only; Table 2). Similarly, the percentage who subsequently experienced a nonvertebral fracture by end of study was numerically lower in women who initially received romosozumab, followed by antiresorptive therapy, versus comparator (FRAME: 0/17 [0.0%] with romosozumab first then denosumab, 5/59 [8.5%] with placebo first then denosumab; ARCH: 5/67 [7.5%] with romosozumab first then alendronate, 16/100 [16.0%] with alendronate only; Table 2).

Safety outcomes in patients with an on-study fracture

In both the nonvertebral and vertebral fracture subgroups of FRAME and ARCH, incidence of TEAEs, serious TEAEs, and the proportion of patients who required surgery for their fracture were generally similar between patients initially treated with romosozumab versus those initially treated with comparator (Supplementary Table 2 and Supplementary Table 3).

There were no fracture-related complications (i.e., nonunion, malunion, delayed healing, chronic pain, osteomyelitis) reported in patients initially treated with romosozumab in either fracture subgroups of FRAME and ARCH. Among patients with nonvertebral fracture by Month 12, one fracture-related complication was reported in the group of patients who initially received placebo (FRAME; chronic osteomyelitis of the right fibula, occurring by Month 12) and alendronate (ARCH; osteomyelitis of the left leg, occurring by Month 12). In the nonvertebral fracture subgroups, the proportion of patients who required surgery for their fracture was similar between romosozumab and comparator groups in both trials. Safety outcomes are expanded upon within Supplementary Material 1, Supplementary Table 2, and Supplementary Table 3.

BMD change from baseline in patients with an on-study fracture

Greater gains in BMD were observed with romosozumab treatment versus placebo (FRAME) or alendronate (ARCH) in both the nonvertebral and vertebral fracture subgroups (Supplementary Fig. 2 and Supplementary Fig. 3, respectively). However, irrespective of the treatment initially received, percentage change in BMD at the total hip and femoral neck was slightly lower in both the nonvertebral and vertebral fracture subgroups than in the full trial populations, which included those who did not experience on‑study fractures.

Discussion

In these post hoc analyses of the FRAME and ARCH phase 3 trials, we evaluated the efficacy and safety of romosozumab in patients who experienced a nonvertebral, or a new or worsening vertebral fracture. In romosozumab-treated patients, incidence of first on-study fracture was numerically lower in the first year compared with placebo- or alendronate‑treated patients. Importantly, incidence of recurrent nonvertebral and vertebral fractures after the first fracture was also numerically lower in women who initially received romosozumab versus comparator in both FRAME and ARCH. Overall, adverse events were balanced in patients with on-study fracture in both romosozumab and comparator groups; continuing romosozumab treatment in the acute post-fracture period was not associated with fracture‑related complications or increased rates of post-fracture surgery. These numerical findings provide additional evidence underscoring the efficacy and safety of romosozumab administration in the acute post-fracture period.

Individuals who experience a fracture are at increased risk of future fractures [1,2,3,4,5]. One retrospective cohort study of postmenopausal women who had sustained a fracture reported the cumulative incidence of subsequent fracture within 1-year post fracture to be 9.8% for an initial fracture at any site and 7.2% for an initial fracture of the radius (being the most common fracture site in the present study) or ulna [23]. In older women, it has been reported that imminent fracture risk is greatest in those who initially experienced clinical vertebral fractures, with ~ 14% of these patients experiencing a subsequent fracture within 1 year; incidence of subsequent fracture was also high in those who sustained pelvic, clavicular, and humeral fractures [23]. In the current study, incidence of subsequent fracture, whether nonvertebral or vertebral, following either a nonvertebral or vertebral fracture occurring by Month 12, was numerically lower in those initially treated with romosozumab than those initially treated with comparator. Importantly, during Months 12–36, patients in both arms of FRAME or ARCH received the same open-label treatment (FRAME, denosumab; ARCH, alendronate) [21, 22]. Thus, the differences in incidence of recurrent fracture can be attributed to the difference in response during the 12 months of romosozumab treatment. These data provide support for the efficacy of romosozumab in patients who are at imminent fracture risk and suggest the importance of continuing treatment with romosozumab following a fracture.

Previous studies have provided no safety concerns for rapid initiation of an osteoporosis therapy after a fracture [24,25,26,27]. In one study of patients with tibial diaphyseal fractures, post‑operative administration of romosozumab on Day 1 and Weeks 2, 6, and 12 following the fracture was not associated with delayed healing when compared with placebo [28]; similar results have also been observed in romosozumab‑treated patients with intertrochanteric or femoral neck hip fractures [29]. The efficacy and safety of administration of other osteoporosis medications following a fracture have also been studied: in the FREEDOM trial, administration of denosumab within 6 weeks following a fracture was not associated with delayed healing or nonunion of the fracture site [30]. Similarly, a recent meta-analysis of the effectiveness of parathyroid hormone analogues, such as teriparatide, found no detrimental effect on fracture healing [31]. Zoledronic acid was also studied in a population of acute hip fracture patients, among whom no detrimental impact of zoledronic acid treatment was found on fracture healing [32]. Collectively, these data suggest that there is no need to delay the administration of osteoporosis treatment in patients who have recently experienced fracture.

In both fracture subgroups, greater gains in BMD were observed in romosozumab‑treated patients than placebo- (FRAME) or alendronate-treated (ARCH) patients. However, these gains were slightly lower at the total hip and femoral neck in patients with on‑study fracture versus the full trial populations, which included those who did not experience on-study fractures. This could be related to reduced mobility or activity post‑fracture or other inflammatory processes in the post-fracture period [33]. Indeed, acute bone loss after fracture has been described in multiple prior studies [33,34,35].

Limitations of the analyses undertaken in this study include the fact that the data are from clinical trials with defined patient inclusion criteria which may not fully reflect the characteristics of a real-world population. In addition, the FRAME and ARCH trials were not designed to evaluate the effects of romosozumab on fracture healing, and thus, the follow-up assessments employed may not have been sufficiently comprehensive to detect subtle differences between the romosozumab and comparator groups. Furthermore, fracture numbers were limited, and the sample size might not have been sufficient to observe an impact on fracture repair or complications. The relatively low rate of fracture events in the FRAME and ARCH trials precluded the ability to meaningfully conduct formal statistical testing; thus, results should be interpreted in the context of this limitation. Strengths of the study include the double-blind, parallel‑group study design and the robustness of assessments in a clinical trial setting.

Conclusions

In conclusion, 1 year of romosozumab treatment before antiresorptive therapy not only reduced the incidence of fracture in the first 12 months, but also reduced the incidence of recurrent nonvertebral and vertebral fractures for up to 3 years. This suggests that initial romosozumab treatment was effective at reducing imminent fracture risk versus comparator, and this benefit appeared to be sustained after patients switched from romosozumab to either denosumab or alendronate. Continuing romosozumab treatment in the acute post‑fracture period in patients who sustained a nonvertebral or vertebral fracture in the FRAME and ARCH phase 3 trials did not delay fracture healing or contribute to other skeletal adverse events. In addition, initial romosozumab treatment increased BMD at both hip and spine regions in patients with on-study fractures. These results, together with previously reported data which did not demonstrate adverse effects of romosozumab in the setting of tibial or proximal femoral fractures [28, 29], should help to give clinicians confidence that they can initiate romosozumab treatment promptly post-fracture and continue medication in patients who experience fracture while already receiving treatment.