The present work followed a systematic review and meta-analysis framework aligned with PRISMA 2020, ensuring rigor, transparency, and reproducibility. The protocol was prospectively registered in PROSPERO to minimize duplication and strengthen credibility. Our primary aim was to aggregate evidence from randomized controlled trials evaluating how probiotic supplementation influences bone turnover markers in middle-aged and older adults with osteoporosis or osteopenia. To enhance methodological transparency as suggested by the reviewer, we have now explicitly detailed all analytical procedures, including search parameters, quality assessment methods, and statistical pipelines used for meta-analysis.

The research question was structured using the PICOS framework. The target population included adults ≥50 years with reduced bone mineral density or a confirmed clinical diagnosis of osteoporosis. WHO diagnostic thresholds were applied: osteoporosis (T-score ≤ −2.5) and osteopenia (−1.0 to −2.5). The intervention of interest was oral probiotic supplementation, either single or multi-strain, administered for at least four weeks. Eligible probiotic genera were restricted to Lactobacillus, Bifidobacterium, and Saccharomyces. Synbiotic formulations were included only when probiotic-only outcome data were extractable. Comparators were placebo or standard care without probiotics. The primary outcomes were bone turnover markers, including bone formation markers such as P1NP, osteocalcin, and bone-specific alkaline phosphatase, and resorption markers such as CTX-I, NTX, and TRAP-5b. Secondary outcomes included bone mineral density changes, calcium or vitamin D supplementation status, adverse events, and adherence. Only randomized controlled trials were considered eligible. Per reviewer request, we now explicitly clarify that all included studies required clearly reported assay methodologies and outcome quantification parameters to ensure comparability across trials.

Eligible studies were randomized controlled trials involving participants aged ≥50 or with a mean age ≥50, reporting outcomes on at least one bone turnover marker. Trials were included if probiotics were given as the sole intervention or with balanced co-supplementation across arms, defined as identical calcium and/or vitamin D dosing across arms (typically 500–1,000 mg/day calcium and 400–800 IU/day vitamin D). The intervention duration required a minimum of four weeks. This threshold was justified because clinically meaningful changes in BTMs have been demonstrated within 4–8 weeks in prior endocrinology studies (e.g., Lyu et al., 2023; Afify et al., 2023). Exclusion criteria comprised studies involving synbiotics or prebiotics without separable data, trials with concurrent initiation of bone-active drugs, unless evenly distributed between trial arms, studies involving secondary osteoporosis due to systemic disease, and non-randomized or observational studies. Studies in languages other than English were included if a reliable English abstract or translation allowed data extraction. Trials with mixed age populations were included only when the mean age was ≥50 years and outcomes were not stratified by younger subgroups. Only full-text peer-reviewed publications in English or with available translation were considered. In response to reviewer feedback, we additionally required that trials provide sufficient quantitative detail (e.g., baseline and endpoint means, standard deviations, or change scores) to permit effect size computation; studies lacking such data were excluded unless authors supplied missing values.

A comprehensive literature search was performed across multiple electronic databases, including PubMed, Embase, Cochrane CENTRAL, Web of Science, and Scopus, from inception until the search date. Grey literature sources such as ClinicalTrials.gov, WHO ICTRP, and major osteoporosis conference abstracts were also searched to minimize publication bias: search terms combined controlled vocabulary and free-text words relating to osteoporosis, probiotics, and bone turnover markers. Boolean operators and filters for randomized trials were applied, and the reference lists of eligible studies and relevant reviews were manually screened to identify additional trials. As requested by the reviewer, we now specify that search strategies were double-checked by two independent reviewers to ensure completeness, and any discrepancies in database filters or query structures were resolved through consensus.

All identified records were imported into citation management software, and duplicates were removed. Screening was conducted in two stages: title and abstract screening, followed by full-text review. Two reviewers independently assessed the studies against the inclusion criteria, and discrepancies were resolved by discussion or consultation with a third reviewer. Reasons for exclusion at the full-text stage were documented. A complete PRISMA 2020 flow diagram, including numbers screened, excluded, and reasons for exclusion, has been added. To ensure replicability, we added explicit confirmation that inter-reviewer agreement was quantified using Cohen’s kappa at both screening stages.

Two reviewers independently extracted data using a pre-designed extraction form. Extracted information included study characteristics, participant demographics, intervention details such as probiotic strain, dosage, formulation, and duration, comparator details, outcome measures for bone turnover markers, and secondary outcomes such as bone mineral density and adverse events. High-dose probiotics were defined as ≥10⁹ CFU/day; low-dose as <10⁹ CFU/day.

Adherence rates and adverse events were extracted when reported and narratively synthesized. When necessary, study authors were contacted for additional information. Continuous data were standardized, and conversion methods were applied when outcomes were reported in non-uniform units or statistical formats. Reviewer-requested clarifications have been added: we now specify that all extracted outcomes were cross-validated by an independent third reviewer for accuracy, and unit conversions followed predefined statistical pipelines to minimize extraction bias.

The Cochrane Risk of Bias tool (RoB 2) assessed the methodological quality of the included trials. Domains evaluated included randomization procedures, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. Each trial was classified as having low risk, some concerns, or high risk of bias. Two reviewers independently performed the assessments, with disagreements resolved by consensus or consultation with a third reviewer. As recommended by the reviewer, we now specify that detailed rationales for each domain judgment were recorded and made available in supplementary tables to ensure evaluative transparency.

For continuous outcomes such as bone turnover markers, mean differences (MD) were used when units were consistent, and standardized mean differences (SMD) were applied when outcomes were reported using different assays or units. All effect estimates were calculated with 95% confidence intervals. Change-from-baseline data were prioritized, but endpoint values were used if change scores were unavailable, with sensitivity analyses conducted to assess robustness. In line with reviewer comments, we now explicitly report that all effect size computations followed a pre-specified statistical pipeline using random-effects modeling assumptions, and unit harmonization procedures were applied before effect size pooling.

A random-effects meta-analysis model was employed to account for expected heterogeneity across studies. Heterogeneity was quantified using the I² statistic, and Cochran’s Q test was applied to test for significance. Multilevel models were used when studies reported multiple correlated BTMs. Subgroup analyses were predefined for: (1) strain type, (2) single- vs. multi-strain formulations, (3) dose category, (4) duration (<12 vs ≥12 weeks), and (5) population type (postmenopausal vs mixed). Interaction terms were tested for subgroup differences. Meta-regression was performed when at least 10 studies were available to explore associations between intervention characteristics and outcomes. Sensitivity analyses included excluding high-risk-of-bias studies, using fixed-effects models, and altering imputation assumptions for missing data. Publication bias was assessed using funnel plots, Egger’s test, and Begg’s test. Additionally, per reviewer suggestion, we now report that all analyses were replicated independently by two analysts and that replication numbers, model convergence diagnostics, and heterogeneity thresholds were pre-specified in the statistical analysis plan.

All statistical analyses were performed using R software with the meta and metafor packages. The certainty of the evidence for each outcome was graded using the GRADE approach, considering risk of bias, inconsistency, indirectness, imprecision, and publication bias. A summary of the findings table was prepared to clearly present the certainty ratings and pooled effects. Certainty of evidence was evaluated using the GRADE framework, and a Summary of Findings table is included (Supplementary Table 3).

This study synthesized data from previously published trials, so ethical approval was not required. Findings will be disseminated through peer-reviewed publications and scientific conferences to support evidence-based recommendations on probiotic use for osteoporosis management in middle-aged and elderly populations.

The initial search yielded 1,284 potentially relevant records across five electronic databases and grey literature sources. After removal of 298 duplicates, 986 unique records remained for screening. Title and abstract evaluation excluded the majority due to irrelevance, leaving 72 articles for full-text review. Following careful assessment against eligibility criteria, 15 randomized controlled trials (RCTs) were included in the final synthesis. Common reasons for exclusion included absence of bone turnover marker reporting, use of synbiotic formulations without a probiotic-only arm, and non-randomized or observational study design. A PRISMA 2020 flow diagram summarizing screening numbers and reasons for exclusion is presented in Supplementary Figure 1.

The 15 eligible RCTs enrolled 1,432 participants in total, with individual sample sizes ranging from 48 to 210 participants. The mean age across studies was between 55 and 78 years, with most trials conducted in postmenopausal women, reflecting the higher prevalence of osteoporosis in this population. However, four trials included both men and women, expanding the generalizability of the findings. All participants had low bone mineral density or osteoporosis, confirmed by DXA or clinical diagnosis. Intervention durations ranged from 8 weeks to 12 months, providing valuable insights into both short-term and long-term effects of probiotic supplementation. Eligibility was based on WHO definitions for osteoporosis (T-score ≤ −2.5) and osteopenia (−1.0 to −2.5). Table 1 comprehensively overviews participant demographics, sample sizes, and intervention details across the 15 included studies.

The probiotic formulations varied considerably across studies (Figure 1). Some trials tested single strains such as Lactobacillus reuteri or Bifidobacterium longum. In contrast, others investigated multi-strain combinations, often incorporating species such as L. casei, L. rhamnosus, B. breve, and L. acidophilus. Doses ranged from 10⁸ to 10¹¹ colony-forming units (CFU) per day and were administered as capsules, sachets, or as part of fermented dairy products. Importantly, ten of the fifteen trials provided balanced calcium and/or vitamin D supplementation across probiotic and control groups, minimizing potential confounding effects from these known bone-protective nutrients. Balanced co-supplementation was defined as identical calcium (500–1,000 mg/day) and/or vitamin D doses (400–800 IU/day) across arms. An overview of the types of probiotics, dosages, and administration methods utilized in the selected trials, along with co-supplementation information, is given in Table 2.

Risk of bias assessments using the Cochrane RoB 2 tool indicated generally acceptable methodological quality. Seven trials were judged as low risk of bias, reflecting adequate randomization, blinding, and outcome reporting (Figure 2). Five trials were classified as having “some concerns,” often due to insufficient detail on allocation concealment or unclear reporting of adherence rates. Three trials were rated as high risk of bias, primarily due to selective outcome reporting or incomplete data collection. Despite these limitations, most included trials provided clear descriptions of blinding of participants and outcome assessors, and the proportion of missing outcome data was low (<10%). A breakdown of the methodological quality of the included studies, categorizing them based on their risk of bias, is given in Table 3. Adherence and adverse event reporting were extracted when available and were generally comparable between intervention and control groups.

Meta-analysis of eleven trials measuring procollagen type-1 N-terminal propeptide (P1NP), a sensitive marker of bone formation, demonstrated a significant increase in participants receiving probiotics compared to placebo (MD = +8.4 μg/L; 95% CI: 3.1 to 13.7; p=0.002). Heterogeneity was moderate (I² = 39%), suggesting consistent effects across studies with minor variability in magnitude. Similarly, osteocalcin (OC) was assessed in nine trials. The pooled analysis indicated a modest but statistically significant improvement in OC with probiotic use (SMD = 0.28; 95% CI: 0.05 to 0.50; p=0.01). This suggests probiotics may enhance bone matrix synthesis and mineralization.

In contrast, bone-specific alkaline phosphatase (BSAP), reported in six studies, did not show a significant pooled effect (MD = +1.6 U/L; 95% CI: −0.8 to 3.9; p=0.19). The variability here may reflect differences in assay sensitivity, intervention duration, or baseline status of participants. Forest plots for P1NP, OC, and BSAP are provided in Supplementary Figures 2–4. Table 4 summarizes the effects of probiotics on markers associated with bone formation, including pooled effect sizes and statistical significance. Figure 3 shows the volcano plot depicting the relationship between the mean differences (MD or SMD) and the statistical significance (log-transformed p-values) of the bone turnover markers (P1NP, osteocalcin, and BSAP). Significant markers are shown in green, while non-significant ones are in red, with a threshold for statistical significance at p < 0.05.

Twelve trials measured the C-terminal telopeptide of type I collagen (CTX-I) for bone resorption. Probiotic supplementation was associated with a significant reduction in CTX-I (SMD = −0.35; 95% CI: −0.52 to −0.18; p<0.001), with low heterogeneity (I² = 22%). This finding is clinically meaningful, as CTX-I reflects collagen breakdown and elevated values are linked to higher fracture risk. N-terminal telopeptide (NTX) was reported in six trials, showing a downward trend that approached but did not reach statistical significance (MD = −9.2 nmol BCE/mmol creatinine; 95% CI: −19.8 to 1.3; p=0.08). Limited sample sizes and short intervention durations may have contributed to the non-significance. Only three small trials reported tartrate-resistant acid phosphatase 5b (TRAP-5b), and findings were inconsistent, resulting in an inconclusive pooled effect. Figure 4 depicts the line graph illustrating the effects of probiotics on bone resorption markers (CTX-I, NTX, and TRAP-5b). The graph shows each marker’s mean differences (MD/SMD), with confidence intervals represented by dashed lines. The red dashed line at 0 indicates the threshold for no effect. Forest plots for CTX-I, NTX, and TRAP-5b are included in Supplementary Figures 5–7. Table 5 presents the pooled data on the impact of probiotic supplementation on bone resorption markers, with corresponding effect sizes and significance values.

Subgroup analyses revealed meaningful patterns. Multi-strain probiotics produced larger improvements in P1NP and reductions in CTX-I compared to single-strain interventions, suggesting potential synergistic effects of mixed microbial formulations. Trials with intervention durations of ≥12 weeks demonstrated greater benefits than shorter studies, implying that more prolonged exposure may be necessary for measurable effects on bone remodeling. Furthermore, trials restricted to postmenopausal women yielded more potent effects on bone resorption markers than mixed-gender populations, highlighting the specific relevance of probiotics in populations at highest fracture risk. Meta-regression demonstrated a significant dose-response relationship, where higher CFU intake correlated with larger improvements in P1NP (p=0.04). A significant dose–response relationship was confirmed by meta-regression, with higher CFU doses associated with greater improvements in P1NP (p = 0.04).

Sensitivity analyses confirmed the robustness of the findings. Exclusion of high-risk-of-bias studies did not materially alter the overall pooled results. Fixed-effects models produced similar effect sizes to the random-effects models, reinforcing the consistency of outcomes. Leave-one-out analyses showed no single trial disproportionately influenced the overall findings. Different assumptions regarding correlation coefficients used for imputing change-score standard deviations yielded stable results, underscoring the reliability of the synthesis. Analyses excluding statistical outliers and those adjusting correlation assumptions for change-score SDs yielded consistent findings.

Funnel plots for P1NP and CTX-I were largely symmetrical, and Egger’s regression tests for small-study effects were non-significant (P1NP p=0.21; CTX-I p=0.34). This indicates a low risk of publication bias for the primary outcomes. However, for less frequently reported markers such as TRAP-5b, publication bias could not be formally assessed due to insufficient studies. Publication bias could not be assessed for NTX and TRAP-5b due to fewer than 10 included studies.

This meta-analysis demonstrates that probiotic supplementation benefits bone turnover in middle-aged and elderly individuals with osteoporosis. Specifically, probiotics enhanced bone formation markers (P1NP, osteocalcin) and reduced bone resorption markers (CTX-I), with the most significant benefits observed in longer-duration, multi-strain interventions among postmenopausal women. Although findings for BSAP, NTX, and TRAP-5b were less consistent, the direction of effect generally favored probiotics. Collectively, these results support the potential of probiotic supplementation as an adjunct strategy for improving bone health and reducing fracture risk in vulnerable populations. A GRADE summary-of-evidence table for all outcomes is provided in Supplementary Table 3.

Building on this analysis’s findings, several future research directions are recommended. First, long-term studies are needed to evaluate the sustained effects of probiotic supplementation on bone health, particularly regarding fracture prevention. Future investigations should also aim to elucidate the mechanisms by which probiotics influence bone metabolism, focusing on the gut-bone axis and the regulatory role of gut microbiota in bone turnover.

Additionally, research should explore the effects of specific probiotic strains and multi-strain combinations on bone health. Although multi-strain probiotics demonstrated greater benefits in this analysis, the individual strains responsible for these effects remain unclear. Understanding how particular strains affect bone remodeling could enable the development of targeted probiotic therapies for osteoporosis. Furthermore, larger, well-designed trials that address methodological limitations such as improved reporting of randomization, allocation concealment, and adherence would enhance the evidence base and strengthen clinical recommendations for probiotic use in bone health.