Prescribing Trends of Osteoporosis Medications, Adherence and Persistence with Denosumab versus Alendronate: A Real-World Retrospective Study in China Authors: Jun Chen, Yuliang Huang, Puqing Liu, Lei Fan Published: 2026 ABSTRACT AbstractIntroductionOsteoporosis, a common metabolic bone disease with high fracture risk and heavy global burden, requires long-term medication adherence and persistence for fracture prevention, yet both remain suboptimal. Prescribing trends, adherence, and persistence data for denosumab and alendronate remain limited, especially in Chinese populations.MethodsThis retrospective cohort study used hospital data. Patients aged ≥50 who newly initiated denosumab or alendronate in 2023–2025 were screened; 2025 initiators were excluded for insufficient 1-year follow-up. New users had no prior prescription of either drug since 2009. The proportion of days covered (PDC) was assessed, with good adherence defined as ≥80%. Persistence was time from initiation to discontinuation, with >60-day gap after coverage end defining discontinuation for both drugs. Propensity score matching (PSM) balanced baseline characteristics. Multivariable logistic regression and Kaplan-Meier survival analysis compared adherence and persistence.ResultsA shift in prescribing from oral to injectable formulations, with alendronate decreasing by 44% and denosumab increasing by 94% in 2023–2025. Denosumab demonstrated greater persistence and adherence than alendronate. The 1-year persistence rate was 46.28% (95% CI: 41.88%–50.80%) for denosumab vs 2.89% (95% CI: 1.60%–4.79%) for alendronate, with a 90.80% lower risk [HR: 0.092 (95% CI: 0.077–0.110), p < 0.001]. PDC was 0.530 (IQR: 0.493–0.975), adherence rate 46.90% (227/484) for denosumab vs 0.077 (IQR: 0.077–0.153), 2.48% (12/484) for alendronate; adherence was 41.29-fold higher with denosumab (95% CI: 22.41–76.07, p < 0.001). Multivariate analysis showed hypertension and fracture history with better adherence and persistence; inpatient status was associated with poorer persistence and adherence; and female sex with poorer persistence (all p < 0.05).ConclusionsA shift from oral to injectable formulations was observed. Denosumab was associated with higher persistence and adherence than alendronate. Hypertension and fracture history were associated with better persistence and adherence, whereas inpatient status was associated with poorer persistence and adherence. Female sex was associated with poorer persistence without affecting adherence.Keywords: denosumab, alendronate, adherence, persistence Introduction Osteoporosis is a common chronic bone disease worldwide, with a prevalence of 19.7%, significantly increasing the risk of fragility fractures.1 The one-year mortality rate after hip fracture is as high as 20%–30%, and approximately 60% of patients fail to regain independent living ability.2,3 With population aging, the global incidence of hip fracture is expected to double by 2050 compared with 2018, with Asia contributing half of this increase,4,5 and the associated medical economic burden continues to intensify,6,7 with US direct medical costs exceeding 25 billion and European costs surpassing €56 billion annually.8,9 Long-term standardized treatment is key to reducing fracture risk, but its clinical benefit highly depends on patient adherence and persistence.10,11 Evidence shows that good adherence can significantly reduce fracture risk, while poor adherence leads to deterioration of bone health and waste of medical resources.11,12 However, patient adherence is generally poor in real-world settings; a large cohort study in Korea showed that only 14.8% of new drug users maintained a medication possession ratio (MPR) ≥ 80% over two years,13 and similar patterns of early discontinuation or interruption have been observed in studies across Europe, North America, and the Asia-Pacific region. In terms of drug selection, although oral alendronate is recommended as first-line therapy, its strict dosing requirements (fasting, full glass of water, upright posture for 30 minutes) and gastrointestinal intolerance often affect adherence; denosumab, as a long-acting agent administered via subcutaneous injection every six months, offers convenient dosing but requires regular follow-up.14,15 However, direct comparison of these two dosing modalities in terms of adherence and persistence presents inherent challenges: adherence assessment for oral medications is heavily influenced by patients’ daily medication habits, whereas adherence to injectable agents is closely tied to healthcare accessibility and follow-up systems.16,17 Moreover, the two drugs differ in their applicable populations; for instance, denosumab can be used in patients with renal impairment, whereas bisphosphonates are contraindicated or require caution. Additionally, treatment selection itself may be confounded by multiple factors, including baseline bone mineral density (BMD), fracture severity, reimbursement policies, and drug costs, rendering direct comparisons in real-world settings more complex.18,19 In recent years, global prescribing trends have shifted from oral bisphosphonates to long-acting injectable agents; however, direct comparative evidence on adherence and persistence between these two drugs remains scarce, particularly lacking real-world data on Chinese populations and male patients. Therefore, this study aims to address these evidence gaps. Against the background of insufficient literature comparing weekly oral vs 6-monthly injectable dosing regimens and the lack of data on Chinese populations and male patients, this study intends to systematically analyze evolving trends in anti-osteoporosis drug prescriptions using real-world data, compare differences in adherence and persistence between denosumab and alendronate, and explore key influencing factors. This can provide evidence on the comparative effectiveness of different dosing regimens in long-term real-world applications, especially by extending the evidence to the Chinese population, thereby offering practical evidence for optimizing individualized treatment decisions and long-term management strategies. Materials and Methods Study Design and Data Source This investigation constitutes a single-center retrospective cohort study. Data were sourced from the hospital information system (HIS) of a Grade 3 Class A tertiary care hospital in Zhejiang Province, China. The study period encompassed January 1, 2023, to December 31, 2025. Study Population In this study, the anti-osteoporosis medications included were restricted to oral formulations (alendronate sodium tablets: 70 mg/tablet) and injectable formulations (denosumab injection: 60 mg/vial; zoledronic acid injection: 100 mL: 5 mg/vial; salmon calcitonin injection: 50 IU/vial) available in our hospital’s drug formulary. Prescription Trend Analysis Population (Not Limited to New Users) Encompassed all anti-osteoporosis prescriptions successfully issued at our hospital between 2023 and 2025. The primary statistical measure used to elucidate prescription trends was the annual total number of uses for each medication, with one use defined as one tablet (oral) or one vial (injection). Adherence and Persistence Analysis Population (New Users Only) ① Sampling procedure and new users definition: Patients who initiated alendronate sodium tablets or denosumab injections at our hospital between 2023 and 2025 were identified from the HIS. The date of the initial medication administration was designated as the index date. To verify new-user status, a retrospective review of all prescriptions from 2009 through the day preceding the index date was conducted. Patients with a history of using any anti-osteoporosis medications during this period were excluded. The final sample size was determined by the availability of eligible patients in the database. ② Follow-up duration definition: The follow-up duration commenced on the index date and continued until the date of treatment interruption (no new prescription was obtained for more than 60 days after the end of the drug coverage period of the last recorded prescription), or treatment discontinuation (364 days after the index date), with a maximum follow-up duration of 365 days (sufficient for assessment of early persistence, a key predictor of long-term outcomes). To ensure a complete 365-day follow-up duration, only patients with index dates between 2023 and 2024 were included in the study. ③ Inclusion criteria: Patients who received alendronate sodium tablets or denosumab injection between 2023 and 2025; new users who used only one of the two drugs throughout the entire follow-up duration; aged ≥ 50 years; with a follow-up duration of at least 365 days (index date no later than January 1, 2025); no history of malignant breast tumor. ④ Exclusion criteria: Aged < 50 years; index date after January 1, 2025; previously used anti-osteoporosis medications before the index date; breast cancer patients; used both alendronate and denosumab either at the same time or one after the other during the study period. Exposure Grouping Participants were categorized into the alendronate group or the denosumab group based on the medication received on the index date. Calculation of Persistence and Adherence Persistence Patients were classified as not experiencing treatment interruption during the 365-day follow-up duration. The criteria for defining an interruption were as follows: alendronate group: no new prescription was obtained for more than 60 days after the end of the drug coverage period (7 days multiplied by the number of tablets) of the last recorded prescription; denosumab group: no new injection was obtained for more than 60 days after the end of the drug coverage period (180 days) of the last recorded injection. Following any interruption, subsequent medication was excluded from the persistence analysis. The persistence rate was defined as the proportion of patients who did not experience treatment interruption during the follow-up duration, while persistence duration was defined as the number of days from medication initiation to the first interruption. The date of first interruption was defined as the day after the grace period [alendronate group: 60 days + drug coverage period (7 days multiplied by the number of tablets); denosumab group: 60 days + drug coverage period (180 days)] expired. Adherence The evaluation used the PDC metric. Calculated as total days with initial medication coverage during the follow-up duration, divided by 365. A PDC of 80% or greater was classified as adherent. For covered days, alendronate was calculated by multiplying the per-prescription quantity by 7. Denosumab was calculated as 180 days per injection. Covariates The following variables were extracted from the medical records at the time of the index-date visit: gender, age, type of visit (outpatient or inpatient), hypertension, diabetes, gastrointestinal disease, and history of fractures. Confounders such as BMD, Body mass index (BMI), renal function, reimbursement status, patient and physician treatment preferences, smoking status, alcohol consumption, physical activity level, and socioeconomic status were not included because of excessive missingness or unavailability in the database. Statistical Analysis Missing Data, Duplicate Records, and Loss to Follow-Up ① Missing data: All included covariates (gender, age, visit type, hypertension, diabetes, gastrointestinal disease, and fracture history) were completely ascertained without missing data. For outpatients, comorbidities were extracted from diagnosis records; for inpatients, comorbidities were extracted from discharge summaries. Documented diagnoses were coded as “yes”, and undocumented conditions were coded as “no”. BMI was originally considered as a covariate but was excluded due to a high proportion of missing values in routine clinical data. Therefore, no missing data handling was required for the final set of covariates. ② Duplicate records: Not encountered in this standardized electronic medical record system. ③ Loss to follow-up: In this retrospective database study, patients without subsequent prescriptions within the allowable gap period (eg, due to hospital transfer, emigration, death, or treatment discontinuation) were all classified as having treatment interruption. PSM A logistic regression model was constructed with treatment group as the dependent variable and gender, age, visit type, hypertension, diabetes, gastrointestinal disease, and fracture history as covariates to calculate propensity scores. A 1:1 nearest-neighbor matching without replacement was performed with a caliper of 0.1 times the standard deviation of the logit of propensity scores. The pre-matched cohort included 551 patients in the alendronate group and 966 patients in the denosumab group; after matching, 484 pairs were successfully matched. Analyses were conducted using Zstats version 1.0 (www.zstats.net) and R version 4.5.3 (R Core Team, 2026). Covariate balance post-matching was assessed by standardized mean difference (SMD), with SMD < 0.1 indicating adequate balance. All subsequent analyses were conducted in the propensity score-matched cohort. Descriptive Analysis Continuous variables were expressed as mean ± standard deviation (SD) or median and interquartile range based on normality (Shapiro–Wilk test), and compared using a t-test or Mann–Whitney U-test. Categorical variables were compared using the chi-square test. Persistence Analysis The Kaplan-Meier method was used to estimate median treatment duration and 1-year persistence rate, with between-group comparison by log-rank test. Multivariable Cox regression identified independent predictors of discontinuation, yielding hazard ratios (HRs) with 95% confidence intervals (CIs). The proportional hazards assumption of the Cox model was assessed based on Schoenfeld residuals, and no significant violation was observed (all P > 0.05). Adherence Analysis PDC was analyzed as a continuous variable, with between-group comparisons by independent-samples t-test or Mann–Whitney U-test as appropriate. Adherence was defined as PDC ≥ 80%, and between-group rates were compared by chi-square test. Multivariable logistic regression provided adjusted odds ratios (ORs) with 95% CIs. Sensitivity Analysis ① The PDC threshold was raised from ≥ 80% to ≥ 90%, and adherence outcomes were reassessed. ② The maximum permissible interval was shortened from 240 days to 210 days. The one-year persistence rate and HR for treatment interruption were recalculated. The allowable interval for alendronate remained unchanged. Statistical Software and Significance Level All statistical analyses, except PSM, were performed using JASP Version 0.97 (JASP Team, 2026). All tests were two-sided, with P <0.05 considered statistically significant. Results Trend Analysis Analysis of statistics on oral and injectable dosage forms used in our hospital from 2023 to 2025 reveals a consistent decline in the use of oral preparations, while the use of injectable agents has been variable. Specifically, consumption of injectable agents decreased slightly in 2024, then rebounded to the highest level recorded over the three years in 2025 (see Figure 1). Oral and injectable medication trends, 2023–2025. One use was defined as one tablet (oral) or one vial (injection). The sole oral preparation available in our hospital is alendronate, which demonstrated a cumulative reduction of approximately 44% over the three years. The injectable agents include zoledronate, denosumab, and salmon calcitonin. Zoledronate showed a steady decline, with a cumulative reduction of approximately 60% over the three years. Denosumab exhibited a consistent upward trend with a cumulative increase of approximately 94% during the same period. Salmon calcitonin demonstrated an initial decline followed by an increase, with limited fluctuations and an overall stable trend (see Table 1). Table 1Annual Consumption of Three Injectable Agents, 2023–2025YEARZoledronate (unit)Denosumab (unit)Salmon calcitonin (unit)202318198117,3432024142147016,550202572190518,723 Annual Consumption of Three Injectable Agents, 2023–2025 Screening of Patients and Baseline Characteristics Screening of Patients A cohort of 3225 patients prescribed either alendronate (n = 1059) or denosumab (n = 2166) between 2023 and 2025 was initially screened. Following the application of exclusion criteria, a total of 1517 patients were ultimately included in the study (alendronate group: 551; denosumab group: 966). The screening process is detailed in Figure 2. Flowchart for patient screening. Baseline Characteristics Before and After PSM To account for baseline differences between the two groups, PSM was utilized. In this analysis, the treatment group (denosumab vs alendronate) was the dependent variable, and age, gender, visit type, hypertension, diabetes, gastrointestinal disease, and history of fractures were included as independent variables in a logistic regression model to estimate propensity scores. A 1:1 nearest neighbor matching approach was implemented, with a caliper width defined as 0.1 times the standard deviation of the propensity score. Before matching, the study included 1517 patients, with 551 allocated to the alendronate group and 966 to the denosumab group. Notable differences were identified between the two groups concerning age, gender, type of visit, hypertension, diabetes, and history of fractures (P < 0.05). Specifically, patients in the denosumab group were older and exhibited higher proportions of females, hospitalizations, hypertension, diabetes prevalence, and history of fractures. No significant differences were observed between the groups regarding gastrointestinal disease (P = 0.093). Following the matching process, 484 pairs of patients were successfully matched, resulting in a total of 968 patients. This matching process achieved a good balance between the groups across all covariates, as indicated by SMD < 0.1, suggesting that the matching effectively mitigated intergroup differences. Further details are provided in Table 2. Table 2Baseline Characteristics Comparison Between Groups Before and After MatchingVariableBefore PSMAfter PSMTotal(n=1517)Alendronate(n=551)Denosumab(n=966)PSMDTotal(n=968)Alendronate(n=484)Denosumab(n=484)SMDAge(years),mean±SD70.87±10.1768.88±10.2872.01±9.93<0.0010.31067.87±9.5167.84±10.1267.90±8.870.006Gender,n(%)<0.001Male219(14.44)131(23.77)88(9.11)−0.510147(15.19)68(14.05)79(16.32)0.061Female1298(85.56)420(76.23)878(90.89)0.510821(84.81)416(85.95)405(83.68)−0.061Type of Visit,n(%)<0.001Outpatient827(54.52)414(75.14)413(42.75)−0.655683(70.56)347(71.69)336(69.42)−0.049Inpatient690(45.48)137(24.86)553(57.25)0.655285(29.44)137(28.31)148(30.58)0.049Hypertension,n(%)<0.001Yes346(22.81)89(16.15)257(26.60)0.237152(15.70)80(16.53)72(14.88)−0.046No1171(77.19)462(83.85)709(73.40)−0.237816(84.30)404(83.47)412(85.12)0.046Diabetes,n(%)0.033Yes136(8.97)38(6.90)98(10.14)0.10857(5.89)33(6.82)24(4.96)−0.086No1381(91.03)513(93.10)868(89.86)−0.108911(94.11)451(93.18)460(95.04)0.086Gastrointestinal,n(%)0.093Yes45(2.97)11(2.00)34(3.52)0.08328(2.89)11(2.27)17(3.51)0.067No1472(97.03)540(98.00)932(96.48)−0.083940(97.11)473(97.73)467(96.49)−0.067History of previous fracture,n(%)<0.001Yes732(48.25)231(41.92)501(51.86)0.199379(39.15)201(41.53)178(36.78)−0.099No785(51.75)320(58.08)465(48.14)−0.199589(60.85)283(58.47)306(63.22)0.099 Baseline Characteristics Comparison Between Groups Before and After Matching Persistence Analysis The Kaplan-Meier survival analysis revealed that treatment persistence was significantly higher in the denosumab cohort compared to the alendronate cohort (log-rank χ2 = 940.9, P < 0.001). In the alendronate group, the median treatment duration was 88 days (95% CI: 88–88), with a 1-year persistence rate of 2.89% (95% CI: 1.60%–4.79%). Conversely, in the denosumab group, the median treatment duration was 240 days (95% CI: 240–240), with a 1-year persistence rate of 46.28% (95% CI: 41.88%–50.80%). To assess independent factors influencing treatment discontinuation, a multivariable Cox proportional hazards regression model was employed, including group assignment, gender, age, hypertension, diabetes, visit type, history of fractures, and gastrointestinal disorders, with the alendronate group as the reference category. The findings indicated that the risk of treatment discontinuation in the denosumab group was approximately 90.80% lower compared to the alendronate group [HR: 0.092 (95% CI: 0.077–0.110), p < 0.001]. The specific impacts of each covariate are detailed in Table 3, and the survival curves and risk table are presented in Figure 3. Table 3The Specific Effects of Each Covariate in the Persistence AnalysisVariablesComparison (Reference)HR (95%CI)P valueGroupDenosumab vs Alendronate0.092(0.077–0.110)<0.001GenderFemale vs Male1.245(1.003–1.546)0.047AgePer one-year increase1.007(0.999–1.015)0.093Type of visitInpatient vs Outpatient1.480(1.212–1.809)<0.001HypertensionYes vs No0.726(0.579–0.911)0.006DiabetesYes vs No1.177(0.867–1.598)0.296History of fracturesYes vs No0.850(0.726–0.995)0.043Gastrointestinal diseaseYes vs No0.882(0.559–1.393)0.591 The Specific Effects of Each Covariate in the Persistence Analysis Kaplan-Meier survival curves for treatment persistence in the two groups. Kaplan-Meier survival curves comparing treatment persistence between the alendronate and the denosumab groups. The green solid line represents the alendronate group (n=484), and the black solid line represents the denosumab group (n=484). The log-rank test yielded a p-value < 0.001. The table below the graph shows the number at risk at each time point. Adherence Analysis Comparison of PDC The PDC data deviated from a normal distribution, as indicated by the Shapiro–Wilk test (p < 0.001). Consequently, the Mann–Whitney U-test was employed to assess the differences between the two groups. The analysis revealed that the median PDC, along with its interquartile range (IQR), was 0.530 (IQR: 0.493–0.975) in the denosumab group, which was significantly higher than the median PDC observed in the alendronate group [0.077 (IQR: 0.077–0.153)]. This difference was statistically significant (U = 6609, p < 0.001). Comparison of Adherence Rate (PDC ≥ 80%) Adherence was defined as PDC ≥ 80%. In the alendronate cohort, the adherence rate was 2.48% (12/484), whereas in the denosumab cohort, it was 46.90% (227/484). The chi-square test showed a statistically significant difference in adherence rates between the two groups (χ2 = 256.8, df = 1, p < 0.001). Multivariable Logistic Regression Analysis To adjust for potential confounders, group, age, gender, visit type, hypertension, diabetes, gastrointestinal disease, and history of fracture were simultaneously included in a binary logistic regression model. The results showed that the likelihood of adherence in the denosumab group was 41.29 times that in the alendronate group [OR: 41.29 (95% CI: 22.41–76.07), p < 0.001]. The effects of other covariates are detailed in Table 4. Table 4The Specific Effects of Each Covariate in the Adherence AnalysisVariablesComparison (Reference)OR (95%CI)P valueGroupDenosumab vs Alendronate41.29(22.410–76.070)<0.001GenderFemale vs Male1.159(0.638–2.106)0.628AgePer one-year increase1.013(0.990–1.037)0.272Type of visitInpatient vs Outpatient0.306(0.179–0.523)<0.001HypertensionYes vs No2.070(1.013–4.231)0.046DiabetesYes vs No0.796(0.284–2.235)0.666History of fracturesYes vs No1.583(1.058–2.370)0.026Gastrointestinal diseaseYes vs No0.531(0.188–1.500)0.232 The Specific Effects of Each Covariate in the Adherence Analysis Between-Group Comparison A comparison of persistence and adherence between the alendronate and denosumab groups is shown in Figure 4. As illustrated, the denosumab group showed significantly better persistence and adherence than the alendronate group. Specifically, persistence was 2.89% (95% CI: 1.60%–4.79%) in the alendronate group vs 46.28% (95% CI: 41.88%–50.80%) in the denosumab group [χ2=245.7, p<0.001, OR: 28.92 (95% CI: 16.52–50.55)]; adherence was2.48% (12/484) in the alendronate group vs 46.90% (227/484) in the denosumab group [χ2=256.8, p<0.001, OR: 34.74 (95% CI: 19.07–63.31)]. Comparison of medication persistence and adherence between alendronate and denosumab. Persistence: the proportion of patients without treatment interruption during the 365-day observation period (interruption: >60-day gap after drug coverage). Adherence: the proportion of patients with PDC ≥80%. 484 in both groups. Sensitivity Analysis Changing the Adherence Threshold To evaluate the robustness of the PDC threshold for defining adherence, the adherence criterion was raised from PDC ≥ 80% to PDC ≥ 90%, followed by re-evaluation of the two groups. The findings indicated that the adherence rate in the denosumab group was 43.60% (211/484), significantly higher than that in the alendronate group, which was 1.86% (9/484). This difference was statistically significant (χ2 = 240.0, p < 0.001). After further multivariate logistic regression (adjusting for age, gender, visit type, hypertension, diabetes, gastrointestinal disease, and history of fractures), the denosumab group was 48.29 times more likely to achieve high adherence (PDC ≥ 90%) than the alendronate group [OR: 48.29 (95% CI: 24.10–96.70), p <0.001], remaining significantly higher than the alendronate group. Compared with the primary analysis (PDC ≥ 80%, adjusted OR = 41.29), neither the direction of effect nor the significance level changed, indicating that the study conclusions do not depend on the specific choice of adherence threshold and that the results are robust. Changing the Permissible Gap Period for Denosumab To test the robustness of the “allowable gap” parameter in the persistence definition for denosumab injection, we shortened the allowable gap from 240 days (180 days + 60-day grace period) in the primary analysis to 210 days (180 days + 30-day grace period), redefined treatment discontinuation events, and repeated Kaplan-Meier survival analysis and Cox proportional hazards regression. The results showed that the median persistence duration in the alendronate group was 88 days (95% CI: 88–88), with a 1-year persistence rate of 2.89% (95% CI: 1.60%–4.79%); the median persistence duration in the denosumab group was 210 days (95% CI: 210–210), with a 1-year persistence rate of 41.94% (95% CI: 37.54%–46.34%). The log-rank test showed a statistically significant difference between the two survival curves (χ2 = 878.8, p < 0.001). Compared with the 1-year persistence rate of 46.28% (95% CI: 41.88%–50.80%) in the denosumab group in the primary analysis, the 210-day persistence rate decreased slightly but remained significantly higher than that in the alendronate group. The multivariate Cox proportional hazards regression model, adjusted for age, gender, hypertension, diabetes, visit type, history of fractures, and gastrointestinal disease, showed that the hazard ratio for treatment discontinuation in the denosumab group was 0.116 (95% CI: 0.098–0.137, p < 0.001), which was higher than that in the primary analysis [HR: 0.092 (95% CI: 0.077–0.110)], but the direction of effect and statistical significance remained unchanged, and the 95% CI was still well below 1. The effects of hypertension [HR: 0.721 (95% CI: 0.575–0.903), p = 0.004] and inpatients [HR: 1.436 (95% CI: 1.178–1.750), p < 0.001] were consistent with the primary analysis. Age reached statistical significance in the sensitivity analysis (HR = 1.008, p = 0.044), but the effect was minimal. None of the remaining covariates showed independent effects. In conclusion, the treatment persistence advantage of denosumab does not depend on the specific definition of the allowable gap, and the primary analysis conclusions are robust. Discussion This study is grounded in real-world clinical practice within our hospital. It was found that the prescription of oral medications has declined year by year, while injectable medications have shown an overall upward trend, and the growth magnitude of denosumab has exceeded the decline in alendronate prescriptions. This trend aligns with the findings reported by Jung Yoon Park et al20 who noted that “bisphosphonates were the dominant treatment, but their market share decreased by 11% over five years, while denosumab achieved a growth rate of up to 957.6% and has surpassed bisphosphonates”. In recent years, the phenomenon of gradual shift from oral to injectable medications for osteoporosis treatment has been mentioned in numerous studies, which may be related to the characteristics of the drugs themselves. For instance, alendronate has stringent administration requirements, and its rigorous administration protocols are associated with poor compliance, characterized by high rates of missed doses and treatment discontinuation. In contrast, denosumab requires only a single subcutaneous injection every six months, which is associated with lower dosing frequency. This streamlined approach enables patients to complete six-month medication coverage in a single visit, while avoiding direct irritation of the esophageal and gastric mucosa caused by oral bisphosphonates. Compared with alendronate, denosumab offers superior advantages in terms of adherence, convenience, and safety for patients with hepatic or renal impairment. Another finding is that denosumab showed markedly superior persistence and adherence compared to alendronate. Following PSM and multivariable adjustment, the denosumab group exhibited significantly longer median treatment duration [240 days (95% CI: 240–240) vs 88 days (95% CI: 88–88)] and higher 1-year persistence rate [46.28% (95% CI: 41.88%–50.80%) vs 2.89% (95% CI: 1.60%–4.79%)] compared with the alendronate group, corresponding to an approximate 90.80% lower risk of treatment discontinuation [HR: 0.092 (95% CI: 0.077–0.110), p < 0.001]; the median PDC [0.530 (IQR: 0.493–0.975) vs 0.077(IQR: 0.077–0.153)] and the adherence rate [46.90% (227/484) vs 2.48% (12/484)] were also notably higher, with adherence being 41.29 times higher in the denosumab group than in the alendronate group [OR: 41.29 (95% CI: 22.41–76.07), p < 0.001]. This result remained robust in sensitivity analyses. This perspective is corroborated by a multicenter prospective cohort study conducted in the Asia-Pacific region and a retrospective study in Germany.21,22 Although some regional variations exist, the findings consistently indicate that denosumab has superior persistence and adherence compared with alendronate. It should be noted that these adherence and persistence rates may be underestimated, as out-of-hospital medication purchases were not captured. Patients who obtained medications from other institutions would be recorded as non-adherent or non-persistent in our system, despite potential ongoing treatment. The extent of this misclassification could not be determined in this study, and warrants further investigation in future research with multi-center data linkage or comprehensive prescription monitoring systems. Multivariable analysis showed a bifurcated association of gender with treatment outcomes: female patients exhibited a significantly elevated risk of treatment discontinuation compared to male patients [HR: 1.245 (95% CI: 1.003–1.546), P = 0.047], yet their adherence was not inferior [OR: 1.159 (95% CI: 0.638–2.106), P = 0.628]. The mechanisms underlying this divergence remain unclear. Nevertheless, existing literature has yielded inconsistent conclusions regarding the relationship between gender and adherence or persistence. Orlando et al23 reported higher 1-year persistence in males (66.4% vs 44.7%). Conversely, Ahn et al13 reported significantly better female outcomes for both adherence [OR: 1.75 (95% CI: 1.70–1.80)] and persistence [OR: 1.81 (95% CI: 1.76–1.86)], and Nakatoh et al24 found higher adherence (64.7% vs 57.7%) and persistence (63.1% vs 55.3%) in females; all P < 0.001. These contradictory findings suggest that the association between gender and adherence or persistence may be moderated by multiple factors, including country, healthcare system, medication type, and patient characteristics, necessitating more comprehensive and long-term research. Therefore, clinicians and pharmacists might consider individualized follow-up approaches for all patients, with attention to treatment continuity; whether sex-specific interventions reduce discontinuation rates requires prospective evaluation. Multivariable analysis further identified hospitalization status, hypertension, and history of fractures as factors independently associated with persistence and adherence. Hospitalization status was associated with a heightened risk of treatment discontinuation [HR: 1.480 (95% CI: 1.212–1.809), P<0.001] and reduced adherence [OR: 0.306 (95% CI: 0.179–0.523), P<0.001], which may reflect the relatively more complex conditions, comorbidities, and concomitant medications in hospitalized patients, as well as unmeasured factors related to post-discharge care transitions. This finding is consistent with the study of Tanja Gram Petersen et al25 which identified “immobilization, comorbidities, and co-medication as the main determinants of treatment discontinuation”. In contrast, patients with hypertension [HR: 0.726 (95% CI: 0.579–0.911), P = 0.006; OR: 2.070 (95% CI: 1.013–4.231), P = 0.046] and fracture history [HR: 0.850 (95% CI: 0.726–0.995), P = 0.043; OR: 1.583 (95% CI: 1.058–2.370), P = 0.026] exhibited greater persistence and adherence. The effect of hypertension may be related to the development of habitual medication-taking behaviors among patients requiring long-term management for chronic conditions, with existing studies supporting this association. For instance, studies by Chung-Yuan Hsu et al26 in Taiwan and Ahn SH et al13 in South Korea both identified hypertension as a significant factor associated with increased adherence and persistence. Patients with prior fractures also showed higher persistence and adherence. However, existing literature remains inconsistent: Hadji et al27 found no statistically significant association between history of fractures and persistence in a real-world study (p > 0.05), and the Asia-Pacific study21 similarly observed no significant association. Some studies have even identified fracture history as an independent risk factor for low adherence. These contradictory findings suggest that the association between fracture history and persistence/adherence may be confounded by multiple factors, including country, healthcare system, and patient education. Patients with prior fractures exhibited higher persistence and adherence in this study, though whether this reflects differential post-fracture care, disease severity, physical limitations, or other unmeasured factors remains unclear. This inconsistency underscores the need for further validation of fracture history’s association with medication behavior. Multivariate analysis in this study revealed that age, diabetes, and gastrointestinal diseases were not independently associated with adherence and persistence, suggesting that these factors may not be primary predictors of adherence and persistence in osteoporosis. However, some literature presents differing views: regarding gastrointestinal diseases, Modi et al28 reported that patients with gastrointestinal symptoms had significantly lower adherence; regarding age, the study by Ahn SH et al13 suggested that advanced age is an important risk factor for reduced adherence; regarding diabetes, studies investigating it as an independent predictor remain scarce, with most literature integrating diabetes into broader analyses of comorbidities or concomitant medications, lacking targeted evidence. In summary, although this study did not identify independent associations for these three factors, some studies suggest that gastrointestinal symptoms and advanced age may be associated with reduced adherence, while the role of diabetes awaits further investigation. In addition to the factors examined above, several unmeasured variables may also have confounded the observed associations. For example, socioeconomic status (including patient income and education level) is associated with both medication access and health-seeking behavior. BMD test results, BMI, insurance coverage, and renal function may influence treatment decisions and outcomes. Disease severity and fracture risk assessment may also affect prescribing patterns and patient adherence. The absence of these variables precludes full control of confounding, and the observed associations should be interpreted with caution. This study systematically integrated three-year trends in anti-osteoporotic drug utilization with adherence and persistence analyses at a tertiary hospital, and explored the differential associations of visit type, history of fractures, comorbidities, age, and gender on medication behavior outcomes. The innovations of this study are as follows: First, unlike traditional research that focused solely on a single dimension, this study integrated changes in medication volume with medication behavior analyses, which may provide a more comprehensive reflection of clinical medication practices. Second, it systematically analyzed the associations of multiple factors, including visit type, disease status, and demographic characteristics, on adherence and persistence, which may inform future intervention research. Several limitations should be acknowledged. (1) Single-center data necessitates caution when extrapolating conclusions to other regions or healthcare facilities of varying types and levels; moreover, out-of-hospital medication purchase records were unavailable, potentially resulting in underestimated adherence and persistence rates. (2) The assessment of adherence and persistence was based on dispensing records and cannot confirm whether patients actually ingested their medications. (3) Due to constraints in data availability, variables such as patient income, BMD, BMI, and other relevant factors were not incorporated, which may impact the comprehensiveness of the study findings. (4) Direct validation of clinical outcomes is absent. This study did not verify whether improvements in these metrics translated into actual reductions in fracture incidence or hospitalization rates; therefore, the actual benefits to patient prognosis cannot be directly assessed. Conclusion We observed an increasing proportion of injectable relative to oral formulations in prescriptions at our single center; correspondingly, denosumab was associated with higher recorded medication persistence [46.28% (95% CI: 41.88%–50.80%) vs 2.89% (95% CI: 1.60%–4.79%)] and adherence [46.90% (227/484) vs 2.48% (12/484)] compared to alendronate in this cohort. These differences corresponded to an approximate 90.80% lower risk of treatment discontinuation [HR: 0.092 (95% CI: 0.077–0.110), p < 0.001] and 41.29 times higher likelihood of good adherence [OR: 41.29 (95% CI: 22.41–76.07), p < 0.001]. Between-group comparisons showed statistically significant differences in both persistence [χ2=245.7, p<0.001, OR: 28.92 (95% CI: 16.52–50.55)] and adherence [χ2=256.8, p<0.001, OR: 34.74 (95% CI: 19.07–63.31)]. For persistence, female, hypertension, hospitalization, and history of fractures showed significant associations, with female and hospitalization associated with a higher risk of treatment discontinuation, while hypertension and history of fractures were associated with a lower risk of treatment discontinuation. For adherence, hospitalization, hypertension, and history of fractures showed significant associations; hypertension and history of fractures were associated with better adherence, and hospitalization with poorer adherence. The associations between patient characteristics and medication persistence and adherence observed in this cohort may inform the development of individualized treatment strategies; however, given the single-center retrospective design, the generalizability of these findings is limited, and their clinical utility awaits further validation through prospective multi-center studies. Acknowledgments The authors thank the financial support of Zhejiang Provincial Administration of Traditional Chinese Medicine, “Zhejiang Provincial Traditional Chinese Medicine Science and Technology” program (Grant Number 2024ZL073). Funding Statement This study was supported by the Zhejiang Provincial Traditional Chinese Medicine Science and Technology Program (2024ZL073). Abbreviations PDC, proportion of days covered; PSM, propensity score matching; IQR, interquartile range; MPR, medication possession ratio; BMD, bone mineral density; HIS, hospital information system; BMI, body mass index; SMD, standardized mean difference; SD, standard deviation; HR, hazard ratio; CIs, confidence intervals; OR, odds ratio. Data Sharing Statement The dataset analyzed during the current study is available from the corresponding author upon reasonable request. Ethics Approval and Consent to Participate This study was conducted in accordance with the Declaration of Helsinki, and all methods complied with relevant guidelines and regulations. The study protocol was approved by the Ethics Review Committee of The Second Affiliated Hospital of Zhejiang Chinese Medical University (Approval No. 2026-096-01). Given the retrospective nature of the study and the use of de-identified data, the requirement for informed consent was waived by the Ethics Review Committee of The Second Affiliated Hospital of Zhejiang Chinese Medical University. Clinical trial number: not applicable. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors declare they have no conflicts of interest in this work. References Associated Data Data Availability Statement The dataset analyzed during the current study is available from the corresponding author upon reasonable request.