We systematically searched four electronic databases—MEDLINE (PubMed), Embase, Web of Science, and China National Knowledge Infrastructure (CNKI)—from inception through 5 August 2024 for published RCTs evaluating the effects of probiotic supplementation (versus control or placebo) on bone mineral density (BMD) and bone turnover markers (BTMs) in postmenopausal women. We used the following search terms: ‘probiotics’, ‘probiotic*’, ‘lactobacillus’, ‘bifidobacterium’, ‘enterococcus’, ‘bone’, ‘bone mineral density’, ‘bone loss’, ‘bone turnover’, ‘osteoporosis’, ‘osteopenia’, ‘osteoporo*’, ‘osteopeni*’, ‘postmenopausal’, ‘post menopause’, and ‘postmenopause’ (see Supplementary Table 1 for details on the search strategy). Reference lists of original trials were manually examined to obtain additional relevant data. The language was restricted to English and Chinese.

The details regarding the PICOTS criteria are provided in Supplementary Table 2 .

Inclusion criteria are as follows: 1) RCTs focusing on postmenopausal women; 2) the use of probiotic (multiple-strain or single-strain) supplementation as interventions and use of placebo (or control) as a comparison and consideration of the change of BMD and/or BTMs as outcomes, or trials with multiple interventions (e.g., coadministered probiotics and vitamin D or calcium) were eligible if the study groups differed only by the use of probiotics; 3) trials utilized dual-energy X-ray absorptiometry (DXA) for the measurement of BMD in the lumbar spine and hip at baseline and trial’s end. Concurrently, BTMs were identified via blood analysis at the same time points; 4) probiotic supplementation for at least 3 months and 5) original articles are written in English or Chinese. In addition, when results from a study population were reported more than once, the results with the longest follow-up time were utilized.

The following types of studies were excluded: 1) cross-sectional, cohort or case-control studies, reviews or meta-analyses, case reports, and animal or cell experiments; 2) articles only reporting protocols, editorials, comments, letters, conferences or abstracts of meeting presentations and 3) absence of expected data for meta-analysis.

Two reviewers (FW and PJL) extracted independently the following information from each trial: the first author, year of publication, country, main participants’ characteristics (sample size, age, and body mass index), type of probiotics, intervention duration, other treatments, adherence with intervention, adverse effects, main outcomes including BMD in the lumbar spine and hip, collagen type 1 cross-l inked C-telopeptide (CTX), receptor activator of nuclear factor-κ B ligand (RANKL), osteocalcin (OC), osteoprotegerin (OPG), N-terminal propeptide of type 1 procollagen (P1NP), and bone-specific alkaline phosphatase (BALP). Additionally, the descriptions of the evidence of gut coloization for included studies after probiotic supplementation were presented in Supplementary Table 3 . In the current study, BMD in the lumbar spine and hip were defined as the primary outcomes, while BTMs were utilized as secondary outcomes.

Methodological quality was independently assessed by two investigators (W.W. and F.W.) by using the Cochrane Collaboration tool (33). All disagreements were resolved through consultation with a third investigator (P.J.L.). Bias in studies was appraised as low, high, or unclear, based on an evaluation of sequence generation, allocation concealment, participant and staff blinding, outcome assessor blinding, handling of incomplete data, selective outcome reporting, and other potential biases ( Supplementary Table 4 , Supplementary Figures 1A, B ).

Statistical analyses for this review were performed using STATA 14.0, and Review Manager 5.3 software. The impact of probiotic supplementation on bone status and BTMs was evaluated by examining the mean relative change from baseline to the conclusion of the intervention, alongside its standard deviation (SD). Direct usage of means and SDs of changes from baseline was prioritized; where not available, data were transformed using established methods (34–37). In cases where trials had multiple intervention arms of the same nature, they were combined into a single arm as per previous methods (36).

The pooled effects of the studies were expressed as standardized mean differences (SMD) with 95% confidence intervals (CIs). Heterogeneity among the studies was assessed using Cochrane’s Q test and quantified with the I² statistic, with thresholds for low, moderate, and high heterogeneity set at <25%, 25%-50%, and >50%, respectively (38). A random-effects model was employed for calculating pooled effect measures in the presence of any heterogeneity (I² > 0%). Sensitivity analyses were conducted to test the robustness of the results, involving the sequential omission of individual studies and further removal of studies with a high risk of bias to observe changes in heterogeneity. For the primary outcomes of lumbar spine and hip BMD, subgroup analyses were conducted to explore potential interactions based on the types of probiotic supplements, dosage of probiotic supplementation, intervention duration, geographical region, and participants’ baseline BMD indicators, including the presence or absence of osteoporosis (defined by a T-score of ≤ -2.5). Additionally, meta-regression analysis was utilized to determine if heterogeneity could be attributed to specific baseline characteristics such as age and BMI. Publication bias was assessed in meta-analyses with at least 10 studies using funnel plots and the Egger test (39). The trim-and-fill method was applied to identify and adjust for potential publication bias in the effect estimates.

Our initial search strategies across four databases yielded a total of 346 papers. After the removal of 112 duplicate records and screening the titles or abstracts of the remaining 234 records, we excluded 199 records that were obviously not relevant. The full text of the 35 eligible reports was read, which helped us to identify one additional article. Finally, we identified twelve RCTs that involved 1183 postmenopausal women (635 in the intervention group and 548 in the control group) as eligible for meta-analyses (20–31). A detailed overview of the selection process is provided in the PRISMA flow diagram in Figure 1 .

The general characteristics of the included trials published between 2017 and 2024 were presented in Table 1 . These studies were parallel-design, single-center trials conducted in the China (25, 29–31), Denmark (20), Iran (21), Japan (22), Poland (26), Sweden (23, 24, 27), and Thailand (28). Of these studies, four trials (22, 24, 26, 27) employed a single-strain probiotic as their intervention approach, in contrast to the rest that used multiple-strain probiotic formulations (20, 21, 23, 25, 28–31). The duration (3~24 months) of intervention varied across the studies, with three studies implementing interventions for a period of three months (25, 26, 28), five studies with interventions spanning six months (21, 22, 29–31), and four studies conducting interventions for a timeframe exceeding one year (20, 23, 24, 27). Additionally, eleven trials reported the results of lumbar spine BMD (20–27, 29–31), and ten studies described the results of hip BMD (20–27, 29, 30). According to the dosage of probiotics, seven studies were defined as the high-dose group (≥ 1 x 10^9 CFU/d) (20–27) and three as the low-dose group (< 1 x10^8 CFU/d) (29–31).

For the primary outcomes included in the meta-analysis, there was obvious heterogeneity (I² > 50%). Subsequently, subgroup, sensitivity analyses and meta-regressions were performed to try to identify potential sources of heterogeneity.

The sensitivity analysis for lumbar spine BMD showed that the I² statistic did not fall below 50% after excluding any single trial from the eleven, and the hip BMD analysis also resulted in a similar finding. In addition, the I² statistic for both lumbar spine and hip BMD continued to exceed 50% even after the removal of trials that were reported in the Chinese language. Despite this, the meta-analysis results for both lumbar spine and hip BMD were robust.

Subgroup analyses revealed that probiotic supplementation, regardless of whether it involves single- or multiple-strain formulations, resulted in significant increases in both lumbar spine and hip BMD in trials that specifically included postmenopausal women with osteopenia. This enhancement in BMD was notably more pronounced in comparison to trials involving women who have osteoporosis (characterized by a T-score of more than or equal to -2.5) ( Tables 2 , 3 ). Furthermore, our results indicated that high-dose supplementation of probiotics could improve both lumbar spine and hip BMD more effectively than low-dose supplementation ( Tables 2 , 3 ). In addition, our findings showed that extended probiotic supplementation (at least one year) indicated greater benefits for enhancing lumbar spine BMD (SMD=1.03, 95%CI [0.06, 2.00], P=0.037; I² = 96.4%) compared to supplementation regimens that last for six months or less ( Table 2 ). Meta-regressions did not reveal any significant correlation between the effects of probiotics on BMD and the age and BMI of postmenopausal women ( Supplementary Figures 2A, B ).

Potential publication bias was detected using funnel plots and the Egger test. The funnel plots of lumbar spine BMD ( Supplementary Figure 4A ) and hip BMD ( Supplementary Figure 4B ) were displayed in the Supplementary Materials . Funnel plots were not created for the other markers as there were fewer than ten included trials in their meta-analyses (39). The Egger test results revealed publication bias for the effects of probiotic supplementation on hip BMD (t=2.42, P=0.042; Supplementary Figure 5B ), while the results regarding lumbar spine BMD did not show publication bias (t=1.52, p=0.162; Supplementary Figure 5A ). We used the trim-and-fill method to detect and adjust for publication bias regarding the results of hip BMD, but the updated overall effect estimate did not show significant changes (z=2.945, P=0.003; Supplementary Figure 5C ). Consequently, more studies with large number are further needed.

This meta-analysis of twelve RCTs involving 1183 postmenopausal women sheds new light on the influence of probiotic supplementation on bone health in this demographic. The analysis revealed that supplementation with probiotics, especially at a dose of at least 1 x 10^9 CFU per day, positively impacted bone health, with stronger evidence in women with osteopenia than osteoporosis, as indicated by improvements in lumbar spine and hip BMD. Additionally, probiotic supplementation correlated with reduced levels of CTX and BALP, pointing to a potential anti-osteoporotic effect of probiotics. Despite significant heterogeneity across the included studies, the findings are supported by objective primary and secondary outcome measures and a robust random-effects analysis model. The credibility of the results is further supported by the stability of the meta-analysis outcomes following sensitivity analysis.

Postmenopausal women are disproportionately affected by osteoporosis due to estrogen deficiency (40, 41). The current understanding of osteoporosis is not yet sufficient to develop pharmaceuticals capable of completely preventing or stopping the disease’s progression (42). However, research has shown a link between the gut microbiota and bone mass reduction, as well as osteoporosis prevalence (13–16, 43, 44). The gut microbiota is believed to influence bone metabolism by affecting the balance between osteoclast and osteoblast activity, thereby impacting the host’s metabolism and immune system (9, 10, 45) (The diagram illustrating the potential impact of gut microbiota regulation on promoting bone health is presented in Figure 4 ). This has led to the exploration of modulating the intestinal microbiome, such as through probiotic supplementation, as a treatment for osteoporosis or osteopenia in postmenopausal women (20–31). However, the effects of probiotics on the bone health in postmenopausal women are still inconsistent.

To our knowledge, only one previous meta-analysis has specifically examined the effects of probiotic supplementation on bone health in postmenopausal women (19). That study found positive effects on BMD in the spine but no observed benefits in the hip region. Another meta-analysis (46) evaluated the impact of probiotics on bone health among postmenopausal women and other individuals suffering from senile or diabetic osteoporosis, and it reported similar findings regarding BMD in the lumbar spine and hip among postmenopausal women as Yu et al.’s study (19). The discrepancy in the effects of probiotics on BMD in different areas among postmenopausal women may be due to the small sample size of studies included in these two meta-analyses (19, 46). Our current meta-analysis reaffirms the previous findings on lumbar spine BMD and further expands on these results by demonstrating positive effects of probiotics on hip BMD as well.

Additionally, subgroup analyses were conducted and revealed that the improvement in BMD was more pronounced in postmenopausal women with mild bone loss (osteopenia) (20–24, 26–31), when treated with probiotics, as opposed to those with osteoporosis (25, 29, 30). This may be due to the limited inclusion of studies in the osteoporosis group. Therefore, further research is needed on probiotic intervention for postmenopausal osteoporosis. When the included trials were divided into subgroups according to intervention duration (≤6 months or > 12 months), we found extended durations of probiotic intervention (> 12 months) (20, 23, 24, 27), in contrast to brief periods of treatment, demonstrated superior enhancement in lumbar spine BMD (P=0.037), suggesting that longer-term probiotic supplementation may yield more benefits in bone health. Additionally, the improvement in hip BMD did not reach statistical significance (SMD= 1.10, 95%CI [-0.02, 2.22], P=0.054). We also conducted subgroup analysis based on study region (Europe and Asia), and found that probiotics appear to be more effective in enhancing lumbar spine BMD in postmenopausal women from Europe compared to those from Asian backgrounds (21, 22, 25, 28–31). This observation could be attributed to potential differences in gut microbiome composition, genetic factors, lifestyle, or dietary habits between these populations, although further research is needed to clarify these regional disparities in response to probiotic treatment. Of note, our subgroup analysis also suggested that there was no significant difference in the impact on BMD between single- and multiple-strain probiotics.

CTX and P1NP are both widely recognized bone turnover markers (BTMs) in clinical use, with CTX indicating bone resorption and P1NP indicating bone formation (47). Our study’s findings that probiotic supplementation can lead to a reduction in CTX levels are in line with two previous meta-analyses (19, 46), neither of which reported the effects of probiotics on P1NP due to the limited data. By synthesizing our findings on both P1NP and CTX, it is clear that the mechanism through which probiotic supplementation may help in preventing bone loss is likely due to its effect on inhibiting bone resorption by suppressing osteoclast activity (22). BALP, a marker traditionally associated with osteoblast proliferation, is recognized as a bone formation indicator (45). However, there is a growing body of evidence suggesting that BALP should be reclassified as a marker of bone turnover rather than solely bone formation (48, 49). In our study, it was found that levels of BALP could be decreased by supplementation with probiotics. This is consistent with the findings of other studies, which have reported a decrease in BALP following the consumption of symbiotic products (50) or specific probiotic strains such as Lactobacillus reuteri (51). These collective findings suggest that the classification of BALP as a bone formation marker may need to be reconsidered in light of its association with the broader process of bone turnover. However, the meta-analysis by Yu et al. showed no significant changes in BALP (19). This is most likely due to the limited number of studies included.

In addition our study also examined other bone turnover markers, including OC, OPG, and RANKL. Our findings are in line with previous meta-analyses (19, 46), which did not find any significant differences in the levels of these markers between postmenopausal women who received probiotics and those in the control group. However, it is important to highlight a preclinical study that demonstrated the supplementation of heat-killed Lacticaseibacillus paracasei GMNL-653 could lead to a specific reduction in the mRNA level of RANKL in ovariectomized mice; Whole-genome sequencing and comparative genomics analysis indicated that genes associated with the transport and metabolism of carbohydrates, as well as the biogenesis of cell walls, membranes, and envelopes, might play a role in the anti-osteoporotic effects of GMNL-653 (52). This observation prompts the hypothesis that the impact of different probiotic strains on bone turnover markers might vary. To substantiate these preliminary findings, further research is essential to explore the potential variability in effects among different strains of probiotics on BTMs.

Our study has several limitations. Firstly, there is high heterogeneity between the included studies, although random-effects model was used to calculate the results, complemented by suitable subgroup analyses and meta-regressions. Secondly, we had to calculate SMD rather than the weighted mean difference due to the inconsistent units describing BMD change among the included studies. Third, in several trials (20–22, 25, 27, 28, 30, 31), co-interventions were used as treatment methods, rather than probiotics alone. This might lead to potential bias; however, after we conducted sensitivity analysis, the robustness of the meta-analysis findings was maintained. Lastly, based on the current trials included, it was not possible for us to discern which probiotic strains specifically improve BMD or BTMs.

On the other hand, this study also has several strengths. Firstly, most of the included trials were of high quality. Secondly, we performed adequate subgroup analyses; sensitivity analysis and meta-regressions were also conducted to minimize heterogeneity between the included studies. Thirdly, compared with previous meta-analyses and reviews, this meta-analysis includes a larger number of trials and a wider population.