Response to “Interpreting bone turnover responses and treatment persistence with once-weekly teriparatide” Authors: Jiemei Gu, Qi Zhao, Xin Zhang, Zhenlin Zhang Published: 2026 Dear Editor and Authors of the Letter, We thank Dr. Yanjun Li and colleagues for their interest in and comments on our recently published Phase III study (Gu et al., J Orthop Transl 2026). The letter raises four issues concerning the interpretation of bone turnover marker (BTM) data, the distinction between adherence and persistence, statistical methodology, and the characterization of fracture benefit. We address each in turn below, supplementing with additional background not fully elaborated in the original manuscript. Regarding bone turnover markers: monitoring and interpretation Dr. Li and colleagues note that the Week 12 P1NP in our study increased only 0.44% from baseline, substantially lower than the approximately 20% increase (at Week 4) reported in the Phase I/II study, and that the absence of Week 4 and Week 8 data cannot account for this discrepancy. The following background is relevant. SAL056 is a biosimilar of Teribone® (once-weekly teriparatide acetate, 56.5 μg; Asahi Kasei Pharma, marketed in Japan), sharing an identical amino acid sequence, formulation, strength, and route of administration with the reference product. SAL056 completed its first-in-human Phase I/II study in 2020 (study code G56W1A201, registration number CTR20181346). This was a multicenter, randomized, double-blind, active-controlled combined Phase I/II trial (N = 151) with head-to-head comparison against Teribone®. Both groups received weekly subcutaneous injections for 24 weeks, followed by an 8-week safety/PD follow-up period. BTMs (P1NP, s-CTX, BALP) were assessed as secondary endpoints at Weeks 4, 12, 24, 28, and 32. The results demonstrated highly concordant BTM time–change profiles between SAL056 56.5 μg and Teribone® 56.5 μg: P1NP peaked at Week 4 in both groups and gradually declined thereafter from Week 12 through Week 32. The LSMean (95% CI) percentage changes from baseline for SAL056 at Weeks 4, 12, 24, 28, and 32 were 19.54% (14.82%, 24.26%), 8.45% (2.81%, 14.09%), 2.79% (−5.03%, 10.62%), −6.34% (−14.44%, 1.77%), and −9.44% (−16.51%, −2.37%), respectively. We note that the statement in the original manuscript that P1NP “peaked at Week 12″ was incorrect; according to the complete Phase I/II time-course data, the P1NP peak actually occurred at Week 4. These head-to-head data establish that SAL056 produces a bone turnover pharmacodynamic response consistent with that of the reference Teribone®. P1NP peaked at Week 4 and then declined rapidly—by Week 12, it had already entered the descending phase. The P1NP data observed in the Phase III study are genuine objective results; the failure to capture the Week 4 P1NP increase is attributable not to any difference in drug effect but to the sampling design. As a confirmatory trial, the Phase III study collected BTM samples at baseline and Weeks 12, 24, and 48 to assess long-term trends, thereby missing the early peak at Week 4. The P1NP value recorded at Week 12, already near baseline, is not inconsistent with the complete Phase I/II trajectory of “an increase at Week 4 followed by gradual decline"—we were simply observing different segments of the same curve. The dense sampling of the Phase I/II studyneatly complemented the observational gap in the Phase III study during this period. The pharmacodynamic profile of once-weekly teriparatide differs fundamentally from that of the daily formulation. As shown by the P1NP results from the TOWER study [1] (Fig. 1), once-weekly teriparatide produces a transient early rise in bone formation markers (peaking at approximately Week 4), followed by a return to below baseline levels by around Week 12. This pattern is consistent with the P1NP trajectories observed across both the Phase I/II and Phase III studies: SAL056 and Teribone® both peaked at Week 4 and declined synchronously thereafter, reaching near-baseline levels by Week 12.Fig. 1Percentage change from baseline in P1NP in the tower study. Percentage change from baseline in P1NP in the tower study. In the phase III study of SAL056, the changes from baseline in P1NP at all time points hovered around the baseline [LSM (SE): Week 12 = 0.44% (2.075); Week 24 = −4.39% (2.071); Week 48 = −5.34% (2.456)], whereas the alendronate group showed sharp declines at all time points [LSM (SE): Week 12 = −54.09% (2.024); Week 24 = −60.99% (1.883); Week 48 = −60.19% (2.135)]. The between-group differences were statistically significant at all time points (P < 0.001), confirming fundamentally different pharmacodynamic effects on bone turnover. Alendronate reduces bone formation markers indirectly by inhibiting bone resorption, while SAL056, as a bone anabolic agent, maintained P1NP near baseline. Although the absence of a placebo arm precludes using absolute values to judge whether “maintaining baseline” represents an improvement over the natural course, the persistent and highly significant between-group difference (P < 0.001) demonstrates that SAL056 and alendronate exert fundamentally different effects on bone turnover. This directional divergence itself confirms that SAL056 possesses bone-formation-modulating activity distinct from that of antiresorptive agents. The letter states that “the sparse sampling and observed P1NP trajectory do not themselves demonstrate persistent stimulation of bone formation." The complete data from the TOWER trial [1] offer an additional perspective (Fig. 2). In TOWER, after 72 weeks of once-weekly teriparatide (56.5 μg) treatment, serum P1NP peaked at Week 4 and then declined continuously. Yet in the same trial, lumbar spine BMD increased progressively from baseline through Week 72. These data demonstrate that, under once-weekly dosing, a declining P1NP can coexist with a sustained increase in BMD. P1NP, as a marker of type I collagen synthesis, reflects the early phase of bone formation; its decline after an initial transient rise does not signify cessation of bone formation stimulation, as the subsequent mineralization process continues to drive BMD accretion. Therefore, using the P1NP trajectory at a single time point to dismiss persistent bone formation stimulation is unwarranted. The sustained and statistically significant increase in BMD during the treatment period constitutes stronger evidence for ongoing bone formation stimulation.Fig. 2Changes in BMD and P1NP over time in the TOWER study. Changes in BMD and P1NP over time in the TOWER study. Regarding the difference in Week 12 P1NP change between the Phase I/II and Phase III studies: An approximately 8% difference in Week 12 P1NP change was observed (Phase I/II: +8.45%; Phase III: +0.44%). The letter recommends disclosing Phase I/II time-course data, absolute values and variability, prior treatment history, assay methods, and the interval between blood draw and injection to assess the source of this difference. We address each item below.1)Assay method: The BTM assay conditions were identical across both studies. CTX and P1NP were both measured using Roche Diagnostics electrochemiluminescence assay kits on the Roche Cobas 8000 e602 automated clinical chemistry analyzer. Calibrators and quality control materials were all sourced from Roche Diagnostics.2)Blood sampling procedure: Both studies required blood collection prior to dosing at each visit, to preclude interference from post-dose PTH fluctuations on BTM levels.3)Prior anti-osteoporosis treatment history: see Table 1.Table 1Prior anti-osteoporosis treatment history.Table 1PhaseTreatment GroupProportion of Prior Anti-Osteoporosis TreatmentI/IISAL056 vs Teribone11.4% vs 10.0%IIISAL056 vs Alendronate16.1% vs 14.0% 1)Assay method: The BTM assay conditions were identical across both studies. CTX and P1NP were both measured using Roche Diagnostics electrochemiluminescence assay kits on the Roche Cobas 8000 e602 automated clinical chemistry analyzer. Calibrators and quality control materials were all sourced from Roche Diagnostics. Assay method: The BTM assay conditions were identical across both studies. CTX and P1NP were both measured using Roche Diagnostics electrochemiluminescence assay kits on the Roche Cobas 8000 e602 automated clinical chemistry analyzer. Calibrators and quality control materials were all sourced from Roche Diagnostics. 2)Blood sampling procedure: Both studies required blood collection prior to dosing at each visit, to preclude interference from post-dose PTH fluctuations on BTM levels. Blood sampling procedure: Both studies required blood collection prior to dosing at each visit, to preclude interference from post-dose PTH fluctuations on BTM levels. 3)Prior anti-osteoporosis treatment history: see Table 1.Table 1Prior anti-osteoporosis treatment history.Table 1PhaseTreatment GroupProportion of Prior Anti-Osteoporosis TreatmentI/IISAL056 vs Teribone11.4% vs 10.0%IIISAL056 vs Alendronate16.1% vs 14.0% Prior anti-osteoporosis treatment history: see Table 1.Table 1Prior anti-osteoporosis treatment history.Table 1PhaseTreatment GroupProportion of Prior Anti-Osteoporosis TreatmentI/IISAL056 vs Teribone11.4% vs 10.0%IIISAL056 vs Alendronate16.1% vs 14.0% Table 1Prior anti-osteoporosis treatment history.Table 1PhaseTreatment GroupProportion of Prior Anti-Osteoporosis TreatmentI/IISAL056 vs Teribone11.4% vs 10.0%IIISAL056 vs Alendronate16.1% vs 14.0% Prior anti-osteoporosis treatment history. In both studies, the proportion of prior treatment history was not significantly different between the study drug and control groups (Phase I/II: 1.43% difference; Phase III: 2.1% difference), indicating well-balanced randomization. Comparing across studies, the slightly higher proportion of prior treatment in the Phase III study likely reflects the natural increase in osteoporosis diagnosis and treatment rates between 2018 and 2021—as awareness of osteoporosis prevention grew and drug accessibility improved in China, more patients had received prior intervention before enrollment. Given that prior treatment history was balanced between groups within each study, it cannot account for the approximately 8% difference in P1NP change between the Phase I/II and Phase III studies. In summary, assay method, sampling procedure, and prior treatment history cannot explain the difference between the two studies. We attribute the difference to the following factors.1)Difference in study design 1)Difference in study design Difference in study design The two studies differ substantially in sample size. As an exploratory study, the Phase I/II enrolled only 70 subjects in the SAL056 56.5 μg group and 40 in the Teribone group; with such small samples, point estimates have limited precision, and one or a few extreme values can substantially influence the LS mean. The Phase III study, as a confirmatory trial, expanded the teriparatide group to 243 subjects, markedly improving estimation precision. Under large-sample conditions, the point estimate more closely approximates the true effect; therefore, the Week 12 P1NP change observed in the Phase III (+0.44%) more reliably reflects the true pharmacodynamic profile of the drug in the target population than that of the Phase I/II (+8.45%).2)Impact of sampling variability 2)Impact of sampling variability Impact of sampling variability Although both studies applied identical inclusion/exclusion criteria, the number of study centers differed fundamentally. The Phase I/II was conducted at only 12 centers, with subjects concentrated geographically; the Phase III expanded to 37 centers across broader geographic regions, resulting in a more diverse subject pool and significantly improved sample representativeness. While increased population heterogeneity may introduce greater inter-individual variability, it also enhances the external validity of the Phase III estimates, making them more reflective of the real-world target population. Regression of the point estimate toward the true mean with larger sample size is a normal statistical phenomenon. In summary, the difference in Week 12 P1NP change between the two studies reflects the gap between the sampling fluctuation inherent to a small exploratory study and the precise estimation afforded by a large confirmatory study, rather than a difference in the drug's intrinsic effect. Regarding adherence and persistence The letter draws a clear distinction between “medication adherence” (dose execution rate) and “treatment persistence” (whether the full treatment course was completed), and we fully accept this distinction. However, the choice of comparator is equally important. SAL056 was not developed to compete with oral alendronate on adherence—comparing subcutaneous injection with oral administration is inherently unfair. The problem it was designed to address is the poor persistence associated with daily teriparatide formulations due to the burden of daily injections. Indeed, inadequate treatment persistence with daily teriparatide is well documented in the literature. A 2020 systematic review in Osteoporosis International (29 teriparatide studies) reported 1-year persistence rates ranging from 10% to 87% (median 55%) and 2-year persistence rates of 10%–69% (median 29.5%) [2]. A Japanese real-world study showed a 365-day persistence rate of approximately 61.0% for the daily formulation [3]. These data consistently indicate that the daily injection regimen is a major factor undermining long-term patient commitment. Re-examining the treatment completion data in this light: despite the extreme constraint of COVID-19 lockdowns—which posed additional barriers to injectable therapies—the teriparatide group still completed 48 weeks of treatment at a rate of 72.0% (175/243), with a mean exposure of 38.9 weeks. Previously reported 1-year completion/persistence rates for daily teriparatide typically range from 50% to 60% [2,3]. Considering the pandemic's adverse impact, the 72% completion rate actually suggests that the weekly formulation may have improved long-term persistence by reducing injection frequency. The alendronate group served as an active comparator for the BMD primary endpoint, not as a benchmark for treatment adherence. Oral bisphosphonates and subcutaneous injections differ fundamentally in route of administration—under the close monitoring of a clinical trial, oral medications naturally achieve higher execution rates. Comparing the adherence of an injectable drug directly with that of an oral drug does not address the clinical question SAL056 was developed to answer. What SAL056 addresses is whether reducing the injection frequency of teriparatide from daily to weekly can yield persistence benefits within the same injectable route. Based on the above, we further clarify the specific reasons for the lower completion rate in the teriparatide group. The lower completion rate (175/243, 72.0% vs. 232/250, 92.8% for alendronate) and shorter mean treatment exposure (38.9 weeks vs. 45.8 weeks) were primarily attributable to COVID-19-related disruptions rather than drug intolerance or lack of efficacy: Subcutaneous teriparatide injections require either study center visits or third-party home injection services; during the study period (November 2021 to September 2023), both were severely affected by local lockdown policies. Alendronate, taken orally, was far less affected by accessibility constraints than subcutaneous injections. Among the 68 discontinuations in the teriparatide group (243 randomized − 175 completed), the leading cause was patient withdrawal (n = 50, 73.5% of all discontinuations), primarily due to COVID-19-related dosing accessibility difficulties and loss to follow-up. Withdrawals due to treatment-emergent adverse events (TEAEs) numbered 12 (5.0% of the teriparatide group), all judged drug-related. In the alendronate group, 18 discontinuations comprised 12 patient withdrawals and 3 TEAE-related withdrawals (1.2%, all drug-unrelated). Thus, drug-related withdrawals accounted for only 12/68 (17.6%) of all teriparatide-group discontinuations; the remainder were driven by external accessibility barriers. We agree that direct evidence regarding patient preferences, injection burden, quality of life, persistence, and healthcare resource utilization is currently lacking; these outcomes are better suited for assessment in post-marketing real-world studies. Regarding statistical analysis and applicability of ICH E9(R1) The study protocol was finalized on November 2, 2021. According to the announcement by the National Medical Products Administration (NMPA) on January 21, 2021, ICH E9(R1) applies to clinical trials initiated after January 21, 2022. Therefore, the study did not formally pre-specify estimands according to the ICH E9(R1) framework at the time of protocol development. To address the clinical interpretation questions raised, we have applied the estimand framework to retrospectively re-frame the clinical questions the study was designed to answer. Specifically: in postmenopausal women with osteoporosis at high fracture risk, the treatment effect of SAL056 combined with calcium + vitamin D versus alendronate combined with calcium + vitamin D on the percentage change from baseline in lumbar spine (L1–L4) BMD at Week 48 was estimated. This effect is summarized as the covariate-adjusted between-group mean difference with its 95% confidence interval. For subjects who permanently discontinued treatment or withdrew prematurely, the primary analysis employed ANCOVA with last observation carried forward (LOCF), consistent with a hypothetical strategy—estimating the between-group difference in Week 48 BMD under the assumption that the intercurrent event did not occur. For subjects who initiated alternative anti-osteoporosis therapy, the study continued to collect their Week 48 BMD data, consistent with a treatment-policy strategy: regardless of whether this intercurrent event occurred, the subsequently obtained Week 48 BMD values were included in the analysis. In this study, no subjects in either group initiated alternative anti-osteoporosis therapy; therefore, the handling of this intercurrent event had no impact on the actual analysis results. We further provide the Week 48 BMD data availability (Table 2). Of 242 randomized subjects in the SAL056 group, 183 (75.6%) had actual Week 48 BMD measurements; 59 permanently discontinued or withdrew prematurely. Of 248 subjects in the alendronate group, 229 (92.3%) had actual Week 48 BMD measurements; 19 permanently discontinued or withdrew.Table 2Availability of Week 48 BMD data.Table 2ItemSAL056AlendronateRandomized242248Actual BMD measurement at Week 48183(75.6%)229(92.3%)Permanent discontinuation/withdrawal59(24.4%)19(7.7%)Initiation of alternative therapy00Note: All data in this table are summarized based on the Full Analysis Set (FAS)."Randomized” refers to the number of randomized subjects included in the FAS, and the percentage is calculated using the number of FAS subjects in the corresponding treatment group as the denominator. Having an actual BMD measurement at Week 48 refers to the number of FAS subjects who obtained an actual BMD measurement value within the Week 48 visit window, reflecting the availability of BMD data at Week 48, and is not equivalent to the number of subjects who completed the study. For the number of subjects who completed the study and subject disposition, please refer to the subject flow chart. Table 2Availability of Week 48 BMD data.Table 2ItemSAL056AlendronateRandomized242248Actual BMD measurement at Week 48183(75.6%)229(92.3%)Permanent discontinuation/withdrawal59(24.4%)19(7.7%)Initiation of alternative therapy00Note: All data in this table are summarized based on the Full Analysis Set (FAS)."Randomized” refers to the number of randomized subjects included in the FAS, and the percentage is calculated using the number of FAS subjects in the corresponding treatment group as the denominator. Having an actual BMD measurement at Week 48 refers to the number of FAS subjects who obtained an actual BMD measurement value within the Week 48 visit window, reflecting the availability of BMD data at Week 48, and is not equivalent to the number of subjects who completed the study. For the number of subjects who completed the study and subject disposition, please refer to the subject flow chart. Availability of Week 48 BMD data. Note: All data in this table are summarized based on the Full Analysis Set (FAS)."Randomized” refers to the number of randomized subjects included in the FAS, and the percentage is calculated using the number of FAS subjects in the corresponding treatment group as the denominator. Having an actual BMD measurement at Week 48 refers to the number of FAS subjects who obtained an actual BMD measurement value within the Week 48 visit window, reflecting the availability of BMD data at Week 48, and is not equivalent to the number of subjects who completed the study. For the number of subjects who completed the study and subject disposition, please refer to the subject flow chart. The primary analysis using ANCOVA with LOCF yielded an adjusted between-group difference of 0.80% (95% CI: 0.10% to 1.50%, P = 0.025) favoring SAL056 over alendronate. We agree that LOCF imputes missing Week 48 BMD with the last available measurement, implicitly assuming no further BMD change after dropout; this method alone cannot serve as a direct estimate of the “treatment effect regardless of discontinuation." To evaluate the robustness of the primary conclusion under different missing-data assumptions and model specifications, we report sensitivity analyses (Table 3). The unimputed ANCOVA, multiple imputation-based ANCOVA, and two MMRM analyses primarily assessed consistency under the missing-at-random (MAR) assumption and varying model specifications; reference-based multiple imputation (J2R) served as a conservative sensitivity analysis under the missing-not-at-random (MNAR) scenario. In J2R, subjects with missing data in the SAL056 group were assumed to follow the outcome distribution of the alendronate control group; thus, J2R tests the robustness of the study conclusion under an unfavorable missing-data assumption rather than serving as a direct estimator for the primary estimand.Table 3Primary and sensitivity analyses for the percentage change from baseline in lumbar spine (L1–L4) BMD at Week 48.Table 3AnalysisInter-group difference between SAL056 and alendronate groups(95% CI)P-valuePrimary Analysis: ANCOVA + LOCF0.80(0.10,1.50)0.025ANCOVA, without imputation1.72(0.99,2.46)<0.001Reference-based imputation/J2R1.24(0.473,2.004)0.002ANCOVA + MI1.35(0.63,2.06)<0.001MMRM, center fixed effects1.49(0.78,2.21)<0.001MMRM, center random effects1.44(0.73,2.16)<0.001 Table 3Primary and sensitivity analyses for the percentage change from baseline in lumbar spine (L1–L4) BMD at Week 48.Table 3AnalysisInter-group difference between SAL056 and alendronate groups(95% CI)P-valuePrimary Analysis: ANCOVA + LOCF0.80(0.10,1.50)0.025ANCOVA, without imputation1.72(0.99,2.46)<0.001Reference-based imputation/J2R1.24(0.473,2.004)0.002ANCOVA + MI1.35(0.63,2.06)<0.001MMRM, center fixed effects1.49(0.78,2.21)<0.001MMRM, center random effects1.44(0.73,2.16)<0.001 Primary and sensitivity analyses for the percentage change from baseline in lumbar spine (L1–L4) BMD at Week 48. All analyses consistently showed greater lumbar spine BMD improvement in the SAL056 group: unimputed ANCOVA difference of 1.72% (95% CI: 0.99% to 2.46%, P < 0.001); J2R of 1.24% (95% CI: 0.473% to 2.004%, P = 0.002); multiple imputation ANCOVA of 1.35% (95% CI: 0.63% to 2.06%, P < 0.001); and MMRM analyses with center as fixed and random effects yielding 1.49% (95% CI: 0.78% to 2.21%, P < 0.001) and 1.44% (95% CI: 0.73% to 2.16%, P < 0.001), respectively. Therefore, although the effect size varies with missing-data assumptions and model specifications, all pre-specified sensitivity analyses are directionally consistent, and even under the conservative J2R scenario, SAL056 demonstrated a statistically significant BMD benefit, supporting the robustness of the primary conclusion against unfavorable missing-data assumptions. Regarding the characterization of fracture risk reduction The letter notes that fracture events in our study did not reach statistical significance and argues that concluding “reduced fracture risk” is unwarranted. We understand this rigorous stance but wish to explain the scientific basis for the fracture-benefit characterization in our manuscript across four dimensions. Regulatory positioning of fracture endpoints in osteoporosis drug trials An important piece of regulatory context is needed regarding why fracture was not selected as the primary endpoint. In January 2022, the Center for Drug Evaluation (CDE) of the NMPA issued the Technical Guideline for Clinical Trial Design of Teriparatide Injection Biosimilars. This guideline provides specific technical standards for the clinical development of teriparatide biosimilars, recommending the percentage change from baseline in lumbar spine BMD at the end of treatment as the primary efficacy endpoint, with a recommended study duration of at least 12 months (48 weeks), using a randomized, double-blind, parallel-controlled equivalence design. The SAL056 Phase III protocol was designed in November 2021—just before the guideline's release—but its key design elements (48-week treatment, lumbar spine BMD as primary endpoint, randomized parallel-controlled design) are fully consistent with the guideline's core recommendations. By using BMD rather than fracture incidence as the primary endpoint, the study followed the regulatory technical pathway for teriparatide biosimilar development. Indeed, BMD as a surrogate endpoint for osteoporosis drug efficacy has extensive precedent in biosimilar development. Currently marketed or Phase III-stage denosumab biosimilars have similarly used the percentage change in lumbar spine BMD at 12 months as the primary efficacy endpoint. BMD, as an objective quantitative measure of the net balance between bone formation and resorption, is widely accepted by regulatory authorities and the pharmaceutical industry. Scientific basis for BMD as a surrogate endpoint for fracture Regarding whether BMD improvement can be reasonably extrapolated to fracture risk reduction, the following evidence-based background is warranted. A well-validated, strong graded association exists between BMD and fracture risk. The FNIH-ASBMR-SABRE project [4] pooled individual patient data from 46,666 placebo-group participants across 25 randomized controlled trials (total follow-up 250,662 person-years), confirming that for every 1 SD decrease in total hip BMD, vertebral fracture risk increased (RR = 1.55/SD, i.e., approximately 55% increase) and hip fracture risk increased (HR = 2.27/SD, i.e., approximately 127% increase); for every 1 SD decrease in lumbar spine BMD, vertebral fracture risk increased (OR = 1.56/SD, i.e., approximately 56% increase). More importantly, a dose–response relationship exists between treatment-related BMD changes and fracture risk reduction. The 2024 analysis from the FNIH-ASBMR-SABRE project (Vilaca et al. [5], 22 RCTs, 122,235 participants) showed that changes in total hip BMD at 12 months were significantly associated with fracture risk reduction: r2 = 0.59 for vertebral fractures (95% CI: 0.19–0.75, p = 0.0003), r2 = 0.46 for all clinical fractures (95% CI: 0.11–0.65, p = 0.0007), and r2 = 0.27 for hip fractures (95% CI: 0.00–0.54, p = 0.03). The association strengthened with longer BMD measurement intervals (at 24 months: vertebral fracture r2 = 0.73, all clinical fractures r2 = 0.71), but even at 12 months, the r2 of 0.59 for vertebral fractures indicates that BMD changes are clinically meaningful predictors of fracture risk within the 48-week observation period of our study. Furthermore, the study defined the surrogate endpoint threshold effect (STE) [5]—the minimum BMD increase required to predict a significant fracture risk reduction: at 12 months, the STE was 0.84% for vertebral fractures and 1.76% for hip fractures. In our study, the SAL056 group's mean percentage changes from baseline in lumbar spine and total hip BMD were 4.66% and 2.74%, with LSM values of 5.01% and 3.27%, respectively—both exceeding the STE thresholds by either calculation method, further supporting the expectation of fracture benefit. Notably, these associations hold across anti-osteoporosis drugs with different mechanisms of action (including bisphosphonates, denosumab, PTH analogues, romosozumab, and SERMs), indicating that BMD as a surrogate endpoint for fracture has cross-class generalizability. Therefore, the statistically significant BMD improvement observed with SAL056 in the Phase III study itself constitutes compelling scientific evidence for fracture risk reduction. SAL056 as a Teribone biosimilar: whether Teribone's fracture efficacy data can be referenced Referencing Teribone's fracture data: basis for extrapolation + These data can be referenced, and the justification is well established. SAL056's clinical similarity to Teribone has been established through direct comparative studies; referencing the reference product's existing clinical evidence on this basis is a well-accepted, mature practice in the biosimilar field.(1)Clinical similarity between SAL056 and Teribone established through direct comparative studies (1)Clinical similarity between SAL056 and Teribone established through direct comparative studies Clinical similarity between SAL056 and Teribone established through direct comparative studies The clinical similarity between SAL056 and Teribone was verified through direct comparative studies: a Phase I bioequivalence study confirmed pharmacokinetic equivalence between the two drugs; and a Phase I/II head-to-head study (G56W1A201) directly compared SAL056 and Teribone under the same dosing regimen, patient population, and assay conditions, showing consistent BTM (P1NP, s-CTX) time–change profiles (see Part I of this response). This is the most direct evidence supporting SAL056's reference to Teribone clinical data—in a sensitive clinical comparison model, no difference in pharmacodynamic response was detected between the two drugs.(2)Referencing reference product clinical evidence is an internationally accepted scientific principle (2)Referencing reference product clinical evidence is an internationally accepted scientific principle Referencing reference product clinical evidence is an internationally accepted scientific principle Once a biosimilar demonstrates clinical similarity to its reference product through a sensitive clinical comparison model, referencing the reference product's existing clinical evidence—including efficacy endpoints not independently verified in the biosimilar's own confirmatory trial—is a widely accepted practice among international regulatory authorities. The EMA states that extrapolation “is a well-established scientific principle” whereby “efficacy and safety data can be extrapolated to other approved indications of the reference medicine” after clinical similarity has been demonstrated in one indication, a principle supported by “over 10 years of safe and effective use” [6]. China's CDE, in its 2021 Technical Guideline for Similarity Evaluation and Indication Extrapolation of Biosimilars, similarly stipulates that when a candidate drug demonstrates overall similarity to the reference product and direct comparative trials confirm clinical similarity in at least one indication, scientific justification based on relevant data and information can support the candidate drug's use in other approved indications of the reference product [7]. SAL056 satisfies the above extrapolation conditions: SAL056 and Teribone are both PTH 1-34 acting on the same PTH1 receptor, representing a single-receptor-mediated hormonal protein drug with a well-defined mechanism of action; the Phase I bioequivalence study demonstrated bioequivalence between the two drugs; and the Phase III study confirmed that SAL056's safety profile is consistent with the known characteristics of Teribone, with no new safety signals identified.(3)Evidence referencing under the same indication and dosing regimen (3)Evidence referencing under the same indication and dosing regimen Evidence referencing under the same indication and dosing regimen The context in which SAL056 references Teribone's fracture data is more direct than typical cross-indication extrapolation: both drugs target the same indication (osteoporosis), use the same dosing regimen (56.5 μg weekly subcutaneous injection), and act on the same target (PTH1 receptor). The fracture benefit demonstrated in the TOWER trial comes from the same molecule, same dose, and same dosing frequency as SAL056—referencing its fracture data is not a cross-class or cross-indication inference but a reasonable evidence reference built on proven clinical similarity. Teribone has been shown to reduce the risk of new vertebral fractures in the TOWER trial (3.1% vs. 14.5%, P < 0.01) [1]. SAL056 has been proven clinically similar to Teribone through direct comparative studies; on this basis, referencing Teribone's fracture efficacy data is scientifically well-founded. Our study's own fracture data (1.7% vs. 4.0%, P > 0.05), limited by low event numbers and short follow-up, are insufficient to independently prove fracture risk reduction. However, as a biosimilar with established clinical similarity, referencing the reference product's established fracture efficacy evidence is scientifically reasonable. Clarification of the original manuscript wording We have reviewed the original manuscript wording and confirm that the manuscript never stated that “SAL056 reduced the fracture rate." The manuscript's characterization of fractures was that “clinical fracture rates were numerically lower in the teriparatide group," explicitly noting that the difference did not reach statistical significance and acknowledging limitations including low event numbers and short follow-up. The context in which fracture benefit was mentioned was in discussing the overall clinical value of SAL056 as a Teribone biosimilar, where Teribone's known fracture benefit was introduced as background information—not a claim that our study independently demonstrated fracture risk reduction. In summary, the fracture-benefit characterization can be understood in layers: our study's own fracture data (1.7% vs. 4.0%, P > 0.05), limited by low event numbers and short follow-up, are insufficient to independently prove fracture risk reduction. However, for anti-osteoporosis drug development, using BMD as a surrogate endpoint for fracture incidence in clinical trials has both scientific and regulatory support. Furthermore, as a Teribone biosimilar, SAL056—on the basis of proven bioequivalence and clinical similarity—can reasonably reference Teribone's established fracture benefit evidence from the TOWER trial. More importantly, the original text does not draw a conclusion that SAL056 can reduce fracture risk, even if a reasonable inference can be made based on the complete evidence chain of “bioequivalence + clinical similarity + known efficacy of the reference drug". We once again thank the editorial team and the authors of the letter for their attention to and constructive feedback on our study. Section II. Statement of competing interests Qi Zhao and Xin Zhang are employees of Shenzhen Salubris Pharmaceuticals Co., Ltd., the sponsor of the original SAL056 phase III study discussed in this response. Jiemei Gu and Zhenlin Zhang declare no competing interests, and specifically they have no employment relationship with Shenzhen Salubris Pharmaceuticals Co., Ltd. Apart from the employment relationship of Qi Zhao and Xin Zhang, no author holds any other financial or personal relationship with people or organizations that could inappropriately influence this work, including consultancy, stock ownership, honoraria, paid expert testimony, patent applications/registrations, grants or other funding relevant to the preparation of this response letter. Author contributions Jiemei Gu: writing – review and editing, final approval of the version to be submitted. Qi Zhao: conceptualization, data interpretation, writing – original draft, final approval of the version to be submitted. Xin Zhang: data interpretation, writing – original draft, final approval of the version to be submitted. Zhenlin Zhang: conceptualization, supervision, writing – review and editing, corresponding author, final approval of the version to be submitted. All authors have read and agreed to the submitted version. Declaration of generative AI in scientific writing During the preparation of this manuscript, the authors used AI-assisted tools solely for language purposes, namely to calibrate the English translation and to polish the wording of text drafted by the authors. All scientific content — including the data analyses, interpretation, and argumentation — was conceived and written by the authors themselves. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of this article. Funding The original phase III study discussed in this response was sponsored by 10.13039/501100004791Shenzhen Salubris Pharmaceuticals Co., Ltd. No specific funding was received for the preparation of this response letter. Declaration of competing interests Jiemei Gu and Zhenlin Zhang declare no competing interests. Qi Zhao and Xin Zhang are employees of Shenzhen Salubris Pharmaceuticals Co., Ltd.