Effects of Teriparatide Compared with Risedronate on the Risk of Fractures in Subgroups of Postmenopausal Women with Severe Osteoporosis: The VERO Trial

The 2‐year, randomized, double‐blind, active‐controlled fracture endpoint VERO study included postmenopausal women with established osteoporosis, who had at least 2 moderate or 1 severe baseline vertebral fractures (VFx), and bone mineral density (BMD) T‐score ≤–1.5. Patients were treated with either s.c. daily teriparatide 20 μg or oral weekly risedronate 35 mg. As previously reported, the risk of new VFx and clinical fractures (a composite of clinical VFx and nonvertebral fragility fractures [NVFFx]) was statistically significantly reduced with teriparatide compared with risedronate. Here we present the prospectively planned subgroup analyses of fracture data across subgroups, which were predefined by the following baseline characteristics: age, number and severity of prevalent VFx, prevalent nonvertebral fractures (NVFx), glucocorticoid use, prior osteoporosis drugs, recent bisphosphonate use, clinical VFx in the year before study entry, and baseline BMD. Heterogeneity of the treatment effect on the primary endpoint (new VFx), and the four key secondary endpoints (including clinical fractures and NVFFx) were investigated by logistic and Cox proportional hazards regression models. A total of 1360 women were randomized and treated (680 per group). Mean age was 72.1 years, mean (SD) number of prevalent VFx was 2.7 (2.1), 55.4% had a BMD T‐score <–2.5, 36.5% had a recent clinical VFx, 28.3% had a prior major NVFx, 43.2% were osteoporosis drug‐naïve, 39.3% were recent bisphosphonate users, and 9.3% were taking glucocorticoids at a prednisone‐equivalent dose of >5 mg/d. For most fracture endpoints, the risk reduction of teriparatide versus risedronate did not significantly differ in any of the subgroups analyzed (treatment‐by‐subgroup interaction p > 0.1), with most subgroups mirroring results from the total study population. In conclusion, in postmenopausal women with severe osteoporosis, the antifracture efficacy of teriparatide compared with risedronate was consistent in a wide range of patient settings, including treatment‐naïve and previously treated patients. © 2018 The Authors. Journal of Bone and Mineral Research Published by Wiley Periodicals Inc.


Introduction
T he choice of the drug treatment for secondary prevention of fragility fractures in patients with osteoporosis is a controversial topic. Most osteoporosis guidelines consider oral bisphosphonates as a first-line therapy, whereas other antiresorptives such as zoledronate or denosumab or the boneforming agent teriparatide are recommended as second-line therapies. We have recently reported the "VERtebral fracture treatment comparisons in Osteoporotic women" (VERO) trial, (1) the first active-controlled clinical trial adequately designed to show superiority in the incidence of new vertebral fractures (VFx) as the primary endpoint, comparing teriparatide with risedronate, an oral bisphosphonate that has shown significant antifracture efficacy results in postmenopausal women with low bone mass and prevalent VFx. (2)(3)(4)(5) In the VERO trial, 24-month treatment with teriparatide significantly reduced the incidence of VFx and clinical fractures (a composite of clinical VFx and nonvertebral fragility fractures [NVFFx]) compared with risedronate by 56% and 52%, respectively, in postmenopausal women with established osteoporosis. (1) Several major risk factors for osteoporotic fractures are recognized, including age, bone mineral density (BMD), glucocorticoid use, or a prior fracture. (6)(7)(8) These factors may influence the treatment effect of the anti-osteoporosis drugs. Using the data from the VERO trial, we examined whether the treatment effect on the incidence of new fractures observed in the entire population as reported by Kendler and colleagues (1) was heterogeneous across predefined subgroups by clinically relevant baseline characteristics. Similar subgroup analyses have been reported previously for alendronate, (9) risedronate, (4) zoledronic acid, (10) denosumab, (11) and abaloparatide (12) in the population of patients enrolled in the phase 3 trials but using placebo as the comparator, instead of an active comparator as utilized in the present study.

Study design
In this double-blind, multinational, multicenter trial, postmenopausal women with at least 2 moderate or 1 severe VFx, according to the classification of Genant and colleagues, (13) and a BMD T-score of -1.5 were enrolled. After their eligibility was confirmed, 1360 women were randomized to receive either 20 mg of s.c. teriparatide once daily plus oral weekly placebo (680 treated patients) or 35 mg of oral risedronate once weekly plus daily injections of placebo (680 treated patients) for up to 24 months. The primary study outcome was the incidence of new radiographic VFx. A new VFx was defined as a vertebral body height loss of at least 20% (and 4 mm) of a vertebra that was unfractured at baseline, based on a 6-point placement of the vertebral bodies from T 4 to L 4 , and confirmed by an increase by one or more severity grades according to the semiquantitative grading (SQ) scale by Genant and colleagues. (13) Key secondary outcomes included the incidence of pooled new and worsened VFx, clinical fractures (a composite of clinical VFx and NVFFx), NVFFx (excluding pathologic fractures and fractures of the skull, face, fingers, metacarpals, and toes), and a subgroup of major NVFFx (hip, radius, humerus, rib, pelvis, tibia, and femur) per the European Medicines Agency definition. (14) Worsening of a baseline preexisting spine fracture was diagnosed if the decrease in vertebral height was at least one severity grade in the SQ assessment . A clinical VFx was defined as an  episode associated with signs and symptoms highly  suggestive of a VFx, such as severe back pain of acute  onset; pain with little or no exertion; pain localized to  specific vertebra and associated with limited back mobility;  pain relieved by bed rest; worsened when upright, coughing, sneezing; limited back flexion; or paravertebral muscle tenderness secondary to spasms, confirmed with the detection of a new or worsened radiographic VFx by the centralized X-ray imaging readers. Analyses of nonvertebral and pooled clinical fractures were based on all randomized patients who received at least 1 dose of investigational product (full analysis set) and analyses of VFx on all patients with a baseline and at least 1 post-baseline spinal radiograph evaluable to assess the VFx status after 24 months (modified full analysis set). Further study design, entry criteria, and methodology details are described by Kendler and colleagues (1) in their article summarizing the trial's main efficacy and safety results.

Subgroup analyses
Subgroup analyses were prospectively planned and performed primarily to examine if the treatment effect differs across the different predefined subgroups and, therefore, consistent with the results observed in the entire population. A second aim was to explore the treatment effect within specific subgroups that were considered to be clinically relevant and that have not previously been reported in clinical trials. Subgroup analyses were exploratory and were done for the primary and the key secondary fracture endpoints as defined in the statistical analysis plan of the study that was approved before study unblinding. (1) The following subgroups were prespecified and analyzed: 1) by number of prevalent fragility VFx; 2) by severity of prevalent VFx using the classification by Genant and colleagues; (13) 3) by prevalent major NVFx, where major NVFx included fractures of the hip, radius, humerus, ribs, pelvis, tibia, and femur (excluding pathologic fractures); 4) by use of significant glucocorticoid treatment, defined as >5 mg/d of prednisone-equivalent dose at the baseline visit or during postbaseline visits, regardless of treatment duration; 5) by prior use of osteoporosis drugs using 3 different mutually exclusive categories: prior bisphosphonates users, prior users of osteoporosis medication other than a bisphosphonate, and osteoporosis treatment-na€ ıve patients. The category of prior bisphosphonate users included those patients who had received !3 months of oral (risedronate, alendronate, ibandronate, etidronate, neridronate), i.m. (clodronate) or i.v. (ibandronate, pamidronate, clodronate) bisphosphonates, or !1 dose of i.v. zoledronic acid. The category of nonbisphosphonate users included those patients who had received !3 months of any nonbisphosphonate antiresorptive (excepting denosumab) or !1 dose of s.c. denosumab. Osteoporosis treatment-na€ ıve patients included patients who had not received any osteoporosis medication, or <3 months of the antiresorptive drugs listed above. Calcium and vitamin D supplements were not considered osteoporosis medications; 6) by the lowest baseline T-score (<-2.5 or !-2.5 SDs) measured at the spine and the proximal femur; 7) by age categorized by tertiles; 8) by recent bisphosphonate use, where recent use was defined as a total of 6 or more months of treatment with any oral bisphosphonate-either intermittently or continuously-within 3 years before screening, or i.v. zoledronic acid at any dose within 2 years of screening, or i.v. ibandronate or pamidronate at any dose within 12 months before screening; and 9) by antecedent of recent clinical VFx within 12 months before screening. Because 97.3% of the patients in this study were white, analyses by the predefined ethnicity subgroups of whites versus non-whites were not considered feasible.

Statistical analyses
Fracture endpoints in subgroups were analyzed following the same prespecified statistical methods applied in the main analysis of the VERO trial. (1) Cochran-Mantel-Haenszel tests, adjusted for the two stratification factors used at randomization (ie, the antecedent of recent clinical VFx and recent bisphosphonate use) were employed to estimate the overall treatment effect between teriparatide and risedronate for the primary and secondary analyses of the morphometric (radiographic) VFx endpoints. For the subgroup analysis of the morphometric (radiographic) VFx endpoints, a logistic regression analysis with the adjusting factors treatment, subgroup, treatment-by-subgroup interaction, and the two stratification factors was fitted. These analyses were implemented in the modified full analysis set that included all randomized patients who received !1 dose of either investigational product and who had !1 post-baseline spine X-ray available to assess their fracture status during 24 months. Note: The full analysis set (FAS) included all randomized patients who received !1 dose of investigational product, and the modified full analysis set included all FAS patients who had !1 post-baseline spine X-ray available to assess their fracture status during 24 months. a Except for 2 patients, all patients in the FAS had at least 1 vertebral fracture (according to the central assessment) at baseline as per entry criteria. For these 2 patients, central fracture adjudication was not available, which was considered a major protocol violation. b Hip, radius, humerus, ribs, pelvis, tibia, and femur (excluding pathologic fractures). c Prednisone-equivalent doses of >5 mg/d at the baseline or any post-baseline visit. d Bisphosphonate: !3 months of bisphosphonates (ie, risedronate, alendronate, ibandronate, etidronate, neridronate, clodronate, or pamidronate) or !1 dose of i.v. zoledronic acid; nonbisphosphonate: !3 months of any nonbisphosphonate antiresorptive (except denosumab) or !1 dose of denosumab; treatment-na€ ıve: no osteoporosis medication or who received <3 months of the antiresorptive drugs listed above. Calcium and vitamin D supplements were not considered osteoporosis medications.
Stratified log-rank tests, adjusted for the two stratification factors, were employed for the analyses of the clinical fracture endpoints (NVFx and clinical VFx). The overall hazard ratio to estimate the treatment effect of teriparatide versus risedronate was obtained as part of the stratified log-rank test calculations. The corresponding subgroup analyses of the clinical fracture endpoints were performed using a Cox proportional hazards regression model with the same factors as specified above for the logistic regression model. These analyses were conducted in the full analysis set that included all randomized patients who received !1 dose of either investigational product.
The relative risk ratios (for VFx) or the hazard ratios (for clinical fractures) with the associated 95% CIs and p values for teriparatide versus risedronate at each level of the subgroup factor and overall were presented in forest plots, together with the incidence rates per treatment group within subgroup and the treatment-by-subgroup interaction p value. A treatment-bysubgroup interaction p value <0.1 was considered as evidence for the corresponding subgroup acting as an effect modifier, ie, the treatment effect of teriparatide versus risedronate would differ depending on the category of the subgroup.

Baseline characteristics
A total of 1360 patients were randomized and treated, 680 in each treatment group; 74.2% of patients overall completed the trial. Baseline demographics and clinical characteristics were similar between groups and have been described previously in Fig. 1. Relative risk ratio for new vertebral fractures by prespecified subgroups (modified full analysis set). BP ¼ bisphosphonates; N ¼ total number of patients; n ¼ number of patients in the specified category; SQ ¼ semiquantitative grading; VFx ¼ vertebral fracture. Ã Major nonvertebral fractures: hip, radius, humerus, ribs, pelvis, tibia, and femur (excluding pathologic fractures). ÃÃ Nonbisphosphonates include strontium ranelate, denosumab, calcitonin, fluoride, and vitamin D active metabolites. detail by Kendler and colleagues. (1) The numbers of patients per subgroup were balanced between treatment groups in the two analyzed data sets ( Table 1). The overall mean age was 72.1 years and 97.3% of patients were white. The mean (SD) number of prevalent VFx was 2.7 (2.1), with 1218 patients (89.6%) having !1 severe (SQ3) VFx at baseline. In the full analysis set, 496 patients (36.5%) had a clinical VFx within the 12 months before entering the study, 385 (28.3%) had a prior major NVFx, and 753 (55.4%) had a lowest BMD T-score of <-2.5 (Table 1). Furthermore, 588 patients (43.2%) were na€ ıve to osteoporosis medication, 534 (39.3%) had been recently treated with bisphosphonates (as per protocol definition), and 127 patients (9.3%) were taking glucocorticoids at a prednisone-equivalent dose of >5 mg/d.

Subgroup analysis results for new vertebral fractures
The forest plot in Fig. 1 illustrates the relative risk ratios between teriparatide and risedronate for the incidence of new VFx, the primary study endpoint, at 24 months within each subgroup as well as overall. The treatment effect found in the entire study population, with an incident rate of new VFx of 5.4% in the teriparatide group compared with 12.0% in the risedronate group (risk ratio 0.44; 95% confidence interval [CI] 0.29-0.68; p ¼ 0.000094 [ Fig. 1; see Kendler and colleagues (1) for details]), was homogeneous across all subgroups, ie, the treatment-bysubgroup interactions were not statistically significant (p ! 0.1) for any of the subgroups. Although none of the subgroups analyzed acted as an effect modifier, the significant treatment effect in favor of teriparatide found in the entire population was numerically greater in the subgroups of patients with 2 prevalent VFx (risk ratio 0.28; 95% CI 0.09-0.81), patients with a prior major NVFx (risk ratio 0.27; 95% CI 0.13-0.58), the oldest patient group (aged !76.8 years) (risk ratio 0.33; 95% CI 0.15-0.73), and patients with a recent clinical VFx within the 12 months before entering the study (risk ratio 0.35; 95% CI 0.20-0.62) (Fig. 1). Of note, the recent use of bisphosphonates did not influence the overall efficacy results with an almost identical, statistically significant, relative fracture risk reduction in patients with (risk ratio 0.46; 95% CI 0.24-0.88) or without (risk ratio 0.42; 95% CI 0.24-0.74) recent bisphosphonate use (treatment-by-subgroup p ¼ 0.85) (Fig. 1). The relative VFx risk reduction was also very similar and statistically significant in the patients with <-2.5 or !-2.5 of the lowest BMD T-score value (risk ratio 0.43 and 0.47, respectively, treatment-by-subgroup p ¼ 0.86) (Fig. 1).

Subgroup analysis results for pooled new and worsened vertebral fractures
As illustrated in the forest plot for the incidence of pooled new and worsened VFx (Fig. 2), the treatment effect found overall (risk ratio 0.46; 95% CI 0.30-0.68; p ¼ 0.000075 [see Kendler and colleagues (1) for details]) was homogeneous across subgroups, with all but two treatment-by-subgroup interactions not being statistically significant (p ! 0.1). Statistically significant treatment-by-subgroup interactions were observed for the subgroups by prior major NVFx and by recent clinical fragility VFx (treatment-by-subgroup p ¼ 0.08 each). The risk ratio within the group of patients with a prior major NVFx was 0.26 (95% CI 0.13-0.56), whereas in the group of patients without prior major NVFx, it was 0.57 (95% CI 0.35-0.92), suggesting a statistically larger reduction in new and worsened VFx incidence for teriparatide in patients with a prior NVFx. Similarly, within patients with a recent clinical fragility VFx, the risk ratio was 0.32 (95% CI 0.18-0.57), whereas in patients without a recent fragility VFx, it was 0.67 (95% CI 0.37-1.21).

Subgroup analysis results for pooled clinical fractures (clinical vertebral and nonvertebral fragility fractures)
The overall treatment effect for the occurrence of new clinical fractures was statistically in favor of teriparatide with a cumulative incidence of 4.8% in the teriparatide group compared with 9.8% in the risedronate group, corresponding to a hazard ratio between teriparatide and risedronate of 0.48 (95% CI 0.32-0.74; p ¼ 0.000869 [ Fig. 3; see Kendler and colleagues (1) for details]). The treatment effect in favor of teriparatide was homogeneous across subgroups. The treatment-by-subgroup interaction was not statistically significant (p ! 0.1) for any of the subgroups. Treatment effects within subgroups showed similar patterns as described above for the primary and first key secondary endpoints, with numerically greater treatment effects in favor of teriparatide observed for patients with 2 (hazard ratio 0.32; 95% CI 0.12-0.88) or more than 3 prevalent VFx (hazard ratio: 0.33; 95% CI 0.14, 0.77), and patients with recent clinical VFx (hazard ratio 0.38; 95% CI 0.20-0.75) (Fig. 3). The reduction in the risk of clinical fractures was statistically significant and very similar to the overall population in those patients who had received bisphosphonates as prior osteoporosis drugs (hazard ratio 0.48; 95% CI 0.26-0.88) or had a BMD T-score <-2.5 (hazard ratio 0.52; 95% CI 0.29-0.93). Fig. 4 shows the Kaplan-Maier curves of the cumulative incidence of clinical fractures observed within the groups of osteoporosis treatment-na€ ıve patients and prior bisphosphonate users.

Subgroup analysis results for nonvertebral fragility fractures
As illustrated in Fig. 5 for the VERO study population, the overall estimate of the hazard ratio for the incidence of NVFFx at 24 months between teriparatide and risedronate was 0.66 (95% CI 0.39-1.10; p ¼ 0.0990 [ Fig. 5; see Kendler and colleagues (1) for details]). The treatment-by-subgroup interactions were not statistically significant (p ! 0.1) for any of the subgroups analyzed. In all subgroups, the treatment difference was not statistically significant. The point estimates of the hazard ratios of all the analyzed subgroups were very similar to the overall estimate of the hazard ratio and numerically in favor of teriparatide, with the exception of the patients with 1 prevalent VFx (hazard ratio 1.06; 95% CI 0.49-2.29) (Fig. 5). We did not observe any increase in the risk of NVFFx fractures in patients with antecedent of prior bisphosphonate use,

Journal of Bone and Mineral Research
including those with recent bisphosphonate treatment. On the contrary, the statistically nonsignificant risk reduction in the prior bisphosphonate-treated patients was numerically similar to that in the bisphosphonate-na€ ıve group (Figs. 5 and 6A).

Subgroup analysis results for major nonvertebral fragility fractures
Similar to the results for all NVFFx, albeit numerically in favor of teriparatide, no statistically significant between-treatment difference for the incidence of major NVFFx was found (hazard ratio 0.58; 95% CI 0.32-1.05; p ¼ 0.0624 [ Fig. 7; see Kendler and colleagues (1) for details]). This nonsignificant treatment effect was homogeneous across subgroups. All treatment-by-subgroup interactions were not statistically significant (p ! 0.1).
Among prior bisphosphonate-treated patients, the statistically nonsignificant risk reduction was numerically similar to that in the bisphosphonate-na€ ıve group (Figs. 6B and 7).

Discussion
Our findings show that the overall treatment effects of teriparatide on the reduction of VFx and clinical fractures that were observed in the entire population of the VERO clinical trial were also consistently observed across all prespecified subgroups analyzed. The treatment-by-subgroup interaction terms were mostly not statistically significant (p ! 0.1), indicating lack of evidence in the data to conclude that the treatment effect was heterogeneous across the different categories of the subgroups. These results support and complement the subgroup analyses by baseline BMD, age, and number of prevalent VFx performed in the patient cohort included in a pivotal phase 3 placebocontrolled clinical trial, (15) where no interactions were observed.
Subgroup analyses of large clinical trials are useful because the beneficial treatment effects of osteoporosis drugs may vary across different categories of relevant risk factors, as has been found before in clinical trials with hip fracture as the study endpoint in patients treated with risedronate (5) or strontium ranelate. (16) For instance, there is a lack of previous data from randomized clinical trials on fracture outcomes in patients receiving teriparatide who were pretreated with bisphosphonates. Earlier studies analyzing the effects of switching from long-term treatment with a bisphosphonate to teriparatide on BMD and biochemical markers of bone turnover have suggested a "blunted" or "delayed" effect of subsequent teriparatide, at least during the early treatment period. (17)(18)(19)(20) Most recently, the STRUCTURE study randomized patients previously on long-term alendronate to either teriparatide or romosozumab, a monoclonal antibody against sclerostin. At 12 months, both total hip BMD and estimated hip strength significantly increased with romosozumab but declined with teriparatide, although no differences in fracture outcomes, analyzed as adverse events, between the two treatment groups were detected. (21) Our subgroup data show similar antifracture efficacy with teriparatide in bisphosphonate pretreated and treatment-na€ ıve patients, even in those patients with recent bisphosphonate treatment. This suggests that the antifracture efficacy may differ from that which the bone strength surrogate endpoints indicate, especially with agents differing in mechanism of action, and that the observed early decrease in BMD in cortical-rich bones, such as the forearm and the hip in patients switching to teriparatide from antiresorptives, does not seem to be associated with an increased risk of any type of fracture, including NVFx. These findings are of clinical importance because they are obtained from an active-controlled trial with risedronate, a bisphosphonate that has shown significant reduction in NVFx and hip fractures in two clinical trials that have analyzed patients with established osteoporosis at high risk of fracture, similar to the patients enrolled in the VERO study. (2,5) As long as a direct comparison of fracture risk is not available between romosozumab and teriparatide after bisphosphonate use, it remains speculative whether the differences in changes in hip BMD and calculated bone strength between these two drugs as reported in the STRUCTURE study (21) reflect differences in fracture risk during the early treatment after switching from bisphosphonates. On the other hand, the recently reported phase 3 trial of abaloparatide, a parathyroid hormone-related protein analog, in postmenopausal women with osteoporosis excluded patients on previous recent antiresorptive therapy; (22) thus, no conclusions on the fracture efficacy of abaloparatide in the clinical setting of prior bisphosphonate use can be drawn from this study.
Of particular importance is the residual burden of fracture risk in patients on risedronate in the VERO study. Thus, 12% of the patients treated with risedronate had a new VFx during the 24 months of follow-up and 9.8% a new clinical fracture. We have demonstrated that teriparatide can effectively reduce this residual fracture burden in these severely osteoporotic patients, with an absolute risk reduction in new VFx of 6.6% after 24 months of treatment and a consistent effect over the wide range of subgroups analyzed. These results confirm previous findings from two double-dummy, active-controlled trials that included the comparison of the fracture incidence between groups as secondary or exploratory endpoints. Thus, Hadji and colleagues showed a significant 5.0% absolute risk reduction in new VFx with teriparatide compared with risedronate in a randomized, double-blind, 18-month duration clinical trial in postmenopausal women with back pain and spine fractures. (23) In the study reported by Saag and colleagues, patients with glucocorticoid-induced osteoporosis treated with teriparatide had an absolute risk reduction of 5.5% in VFx compared with alendronate after the same treatment duration. (24) The VERO study also provides new information about antifracture efficacy in patients with near-term fracture risk (or imminent fracture risk) who are at the highest risk of subsequent fractures (ie, the patients with a prior clinical VFx in the year before entering the study). According to our results, the nearterm 2-year new VFx, new and worsened VFx, and clinical fracture risks in these patients were still high in the risedronatetreated group, but the fracture risks were significantly reduced with teriparatide by 65%, 68%, and 62% for new VFx, new and worsened VFx, and clinical fractures, respectively (Figs. 1-3). The reduction of NVFx remained not statistically significant in this subgroup (hazard ratio 0.60; 95% CI 0.28-1.27).
Also of special clinical and pharmacoeconomic interest is the finding that teriparatide was superior to risedronate in the reduction of clinical fractures, regardless of history of a prevalent major NVFx, with a relative risk reduction of 57% after 24 months of treatment in patients with a prior NVFx. This risk reduction in the VERO trial is similar to the 55% reduction in clinical fractures observed in the 3-year Fracture Intervention Trial (FIT) I trial, where alendronate was compared with placebo in a study cohort with similar clinical characteristics, including the presence of baseline VFx. (25) In the preplanned subgroup analyses of the phase 3 denosumab and abaloparatide trials, no Hazard ratio for major nonvertebral fragility fractures by prespecified subgroups (full analysis set). BP ¼ bisphosphonates; N ¼ total number of patients; n ¼ number of patients in the specified category; SQ ¼ semiquantitative grading; VFx ¼ vertebral fractures. Percentages indicate the cumulative incidence of fractures based on Kaplan-Meier methodology. Ã Major nonvertebral fractures: hip, radius, humerus, ribs, pelvis, tibia, and femur (excluding pathologic fractures). ÃÃ Nonbisphosphonates include strontium ranelate, denosumab, calcitonin, fluoride, and vitamin D active metabolites. significant antifracture effects of those two drugs on NVFx in patients with a history of this type of fracture were found. (11,12) The analysis by baseline NVFx was not considered in the zoledronic acid subgroup analysis. (10) Clinical fractures were analyzed as secondary fracture endpoints in the zoledronic acid and abaloparatide phase 3 trials, (22,26) but they were not reported in the subgroup analyses. (10,12) Our presented analysis has several limitations common to other subgroup study reports, including the limited power to detect interactions. The VERO study was primarily designed to detect a treatment effect difference between teriparatide and risedronate for the primary endpoint of new VFx, and therefore the detection of treatment-by-subgroup interactions would have required a significantly larger sample size. As a consequence, the possibility of a true differential treatment effect for some of the subgroups cannot be fully ruled out because of the inherent lack of power of the study to detect such effect in statistical tests. Additionally, some subgroups included very few patients, such as the glucocorticoid users (less than 10% of the entire population) or the prior nonbisphosphonate users (less than 5% of the entire population), or had very few fracture events. As a result, the CIs for the treatment effect estimated in these subgroups are considerably wider than the corresponding CI estimated using the overall population. Finally, the statistical significance was considered without adjustment for multiple comparisons. As another limitation, surrogate markers of bone quality such as BMD and biochemical markers of bone turnover were not measured as study endpoints.
The strengths of the study include the enrollment of patients with a high fracture risk, normally considered for osteoanabolic therapy, and numerous clinically relevant prespecified subgroups-including prior osteoporosis therapy, a recent clinical VFx, and number and severity of VFx at baseline-that have not previously been analyzed in subgroup analyses and that increase the clinical relevance to many patients that may be candidates for bone anabolic therapy. This is the first randomized clinical trial that assesses fractures as efficacy endpoints in a substantial subgroup of patients pretreated with bisphosphonates. We also included the analysis of relevant clinical outcomes, such as clinical fractures, that have not been reported before in subgroup analyses.
In conclusion, in postmenopausal women with severe osteoporosis, the antifracture efficacy of teriparatide compared with risedronate was consistent within the various predefined subgroups. There was no evidence of heterogeneity in treatment effect across the different categories of the subgroups in reducing the risk of new VFx or clinical fractures. Teriparatide antifracture efficacy is similarly superior to risedronate in both osteoporosis treatment-na€ ıve and prior bisphosphonate-treated patients. Data for the first time indicate the additional fracture benefit of using an anabolic compared with an antiresorptive across multiple patient clinical scenarios. This information should aid in positioning teriparatide bone anabolic therapy for reduction of residual fracture burden in patients at high fracture risk who are either treatment-na€ ıve or on bisphosphonate treatment.