This meta-analysis was conducted in full compliance with the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (13). Our study was registered in INPLASY platform (number: INPLASY202530115). As all data were derived from previously published studies, ethical approval and review were not required for this research.

A comprehensive literature search was performed using PubMed, Embase, Cochrane Library, China National Knowledge Infrastructure (CNKI), and Wanfang databases through March 2025. The search strategy incorporated both subject headings and free-text terms. Keywords included combinations of “Chinese herbal medicine,” “Herbal therapy,” “Traditional Chinese medicine,” “Osteoporosis,” and “Fracture” (see Supplementary File 1). Randomized controlled trials (RCTs) evaluating the effects of CHM on osteoporosis or related fractures were systematically identified. No restrictions were placed on the language of publication. Despite searching international databases, all RCTs evaluating CHM for osteoporosis were conducted in China, likely reflecting the regional focus of such research. To ensure thoroughness, reference lists of included articles were also manually reviewed to capture additional relevant studies not found during the initial database search.

A double-blind, independent screening process was employed. Two researchers independently reviewed study titles, abstracts, and full texts in a stepwise manner. Any discrepancies between reviewers were resolved by a third researcher, who served as an arbitrator to reach a consensus.

Inclusion criteria were established according to the PICOS framework:

Two authors independently extracted data from each study. Extracted variables included: the first author’s surname, year of publication, study location, sample size, proportion of male participants, mean patient age, clinical condition, intervention protocol, control treatment, reported outcomes, and follow-up duration. Following data extraction, both authors independently evaluated the methodological quality of each study using the Cochrane Risk of Bias Tool. Assessment domains included: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, completeness of outcome data, selective reporting, and other potential sources of bias (15). Any inconsistencies in data extraction or quality ratings were resolved through discussion with a third reviewer, who made final determinations based on the original study texts.

Therapeutic outcomes of CHM were treated as continuous variables, with effect sizes expressed as weighted mean differences (WMDs) and corresponding 95% confidence intervals (CIs). Meta-analyses were conducted using a random-effects model to account for heterogeneity across studies (16, 17). Heterogeneity was assessed using the I² statistic and Cochran’s Q test, with significant heterogeneity defined as I² > 50% or a Q-test p-value < 0.10 (18, 19). Robustness of the results was evaluated via leave-one-out sensitivity analysis, in which each study was sequentially excluded to assess its impact on the overall effect size (20). Meta-regression were performed to identify potential source of heterogeneity on the basis of publication year, proportion of male participants, average patient age, clinical disease status, follow-up, and baseline BMD. Then subgroup analyses were performed and the differences between subgroups were tested using interaction t-tests, assuming normal distribution of data (21). Potential publication bias was assessed visually using funnel plot asymmetry and statistically using Egger’s linear regression test and Begg’s rank correlation test (22, 23). All analyses were conducted using STATA version 12.0 (StataCorp LLC, College Station, TX, USA), following Cochrane Collaboration guidelines for conducting and reporting meta-analyses. Two-tailed p-values were reported, with statistical significance set at α = 0.05.

A total of 2,240 relevant studies were initially identified through a comprehensive database search. After removing 868 duplicates using reference management software, 1,372 records remained for preliminary screening. Following title and abstract screening, 1,309 studies that did not meet the predefined eligibility criteria were excluded. The remaining 63 articles were subjected to full-text evaluation.

Through a double-blind, independent full-text review by two researchers, 18 studies were confirmed to meet the PICOS inclusion criteria and were included in the quantitative synthesis (24–41). To enhance the comprehensiveness of the literature coverage, reference lists of the included studies were manually searched. Two additional potentially eligible studies were identified. However, after duplicate verification using the CrossCheck database, both were confirmed to have already been included in the initial retrieval and were excluded due to redundancy. The final analysis included 18 studies (Figure 1).

Table 1 summarizes the basic characteristics of the included RCTs and the demographic details of the study populations. Across the 18 studies, a total of 1,816 patients diagnosed with osteoporosis were included. The median sample size of single-center studies was 98 participants (range: 62–140). Gender distribution was as follows: one study (5.6%) included only male patients, five studies (27.8%) enrolled only postmenopausal women, and 12 studies (66.6%) included mixed-gender cohorts. Regarding follow-up duration, six studies (33.3%) conducted short-term follow-up of 3 months, 11 studies (61.1%) had a 6-month follow-up, and one study (5.6%) reported outcomes at multiple time points (12 and 24 months).

Methodological quality, assessed using the Cochrane Risk of Bias Tool, indicated that the overall quality of the included studies was moderate (Table 2). The primary limitations involved allocation concealment and implementation of blinding procedures.

Pooled analysis of all included trials showed that CHM was associated with increased BMD at the lumbar spine compared to conventional therapy (WMD: 0.09; 95% CI: 0.04 to 0.13; p < 0.001; Figure 2). However, there was significant heterogeneity among studies (I² = 97.0%; p < 0.001). Sensitivity analysis indicated the overall result remained stable when each study was sequentially removed (Supplementary File 2). Meta-regression found publication year, proportion of male participants, average patient age, follow-up, and baseline BMD were not significant factors contributing to the association between CHM and BMD at the lumbar spine, while clinical disease status that contributed to the association between CHM and BMD at the lumbar spine (p = 0.021) (Supplementary File 3). Subgroup analyses revealed that CHM significantly improved lumbar spine BMD in most subgroups. However, no statistically significant differences were observed when the proportion of male participants was ≥ 50%, the mean patient age was < 65 years, or the patients had secondary osteoporosis (Table 3). No significant publication bias was detected for lumbar spine BMD (Egger’s test p = 0.457; Begg’s test p = 0.256; Supplemental File 4).

Eight, four, and four trials, respectively, reported on CHM’s effects on BMD at the femoral neck, greater trochanter of the femur, and Ward’s triangle area (Figure 3). Pooled results indicated CHM was associated with higher BMD at the femoral neck (WMD: 0.09; 95% CI: 0.02 to 0.17; p = 0.015) and Ward’s triangle area (WMD: 0.08; 95% CI: 0.01 to 0.15; p = 0.025). However, CHM was not associated with changes in BMD at the greater trochanter of the femur (WMD: 0.01; 95% CI: −0.03 to 0.05; p = 0.698). Significant heterogeneity was noted for BMD at the femoral neck (I² = 99.1%; p < 0.001), greater trochanter (I² = 68.0%; p = 0.025), and Ward’s triangle area (I² = 93.9%; p < 0.001). Sensitivity analyses suggested variable stability of results, likely due to the small number of included trials and wide confidence intervals (Supplementary File 2). Meta-regression analyses found publication year, proportion of male participants, average patient age, clinical disease status, follow-up, and baseline BMD were not significant factors contributing to the association between CHM and BMD at the femoral neck, greater trochanter of the femur, and Ward’s triangle area (Supplementary File 3). Subgroup analyses found that CHM significantly increased femoral neck BMD in studies published in 2010 or later, with <50% male participants, mean age ≥ 65 years, patients with primary osteoporosis, and across all follow-up durations. CHM did not affect BMD at the greater trochanter in any subgroup. For Ward’s triangle area, significant improvements were seen in studies published before 2010, with <50% male participants, mean age ≥ 65 years, patients with primary osteoporosis, and follow-up durations of 6 or 12 months (Table 3). No significant publication bias was observed for BMD at the femoral neck (Egger’s test p = 0.658; Begg’s test p = 0.711), greater trochanter (Egger’s p = 0.787; Begg’s p = 0.734), or Ward’s triangle area (Egger’s p = 0.965; Begg’s p = 1.000; Supplemental File 4).

Nine studies reported the effect of CHM on alkaline phosphatase. The pooled result showed no significant association between CHM and alkaline phosphatase levels (WMD: 0.98; 95% CI: −6.88 to 8.83; p = 0.808; Figure 4). Notably, there was significant heterogeneity across studies (I² = 94.2%; p < 0.001). Sensitivity analysis confirmed the overall conclusion was stable, with no individual study altering the pooled effect (Supplementary File 2). Meta-regression analyses found average patient age (p = 0.029) and disease status (p = 0.018) contributed to the association between CHM and alkaline phosphatase levels (Supplementary File 3). Subgroup analysis indicated CHM was associated with increased alkaline phosphatase levels in patients with secondary osteoporosis (Table 3). No significant publication bias was detected (Egger’s test p = 0.933; Begg’s test p = 0.917; Supplementary File 4).

Eight and six studies reported the effects of CHM on serum calcium and serum phosphorus, respectively (Figure 5). CHM had no significant effect on serum calcium (WMD: 0.08; 95% CI: −0.09 to 0.25; p = 0.372) or serum phosphorus (WMD: -0.05; 95% CI: −0.22 to 0.12; p = 0.574). Substantial heterogeneity was observed for both serum calcium (I² = 98.8%; p < 0.001) and phosphorus (I² = 97.2%; p < 0.001). Sensitivity analysis indicated stable pooled conclusions for serum calcium, while the effect on serum phosphorus was more variable (Supplementary File 2). Meta-regression analyses found publication year, proportion of male participants, average patient age, clinical disease status, follow-up, and baseline BMD were not significant factors contributing to the association of CHM with serum calcium and serum phosphorus (Supplementary File 3). Subgroup analyses showed CHM was not associated with serum calcium or phosphorus levels in any subgroup (Table 3). No significant publication bias was found for either outcome (serum calcium: Egger’s p = 0.832, Begg’s p = 0.536; serum phosphorus: Egger’s p = 0.694, Begg’s p = 0.707; Supplementary File 4).