Effectiveness and safety of probiotics in treating knee osteoarthritis: an updated systematic review and meta-analysis of randomized controlled trials
Authors: Keye Chen, Yan Sun, Longkang Cui, Yuchen Zhu, Bingbing Zhang, Lianguo Wu
Published: 2026

ABSTRACT
Knee osteoarthritis (KOA) is a prevalent degenerative joint disease, and gut microbiota dysbiosis in its pathogenesis is a research hotspot. This study performed a traditional meta-analysis to systematically evaluate probiotics’ efficacy (for KOA symptoms, strain-specific superiority) and safety, using RCT data from Chinese and English databases. Stata 17.0 and RevMan 5.4 analyzed outcomes including patient-reported pain (Visual Analog Scale (VAS), WOMAC pain), joint stiffness, function limitation, serum High-sensitivity C-reactive protein (hs-CRP), and safety. Probiotics showed no significant adverse event differences vs. placebo, with no serious safety issues. Results suggest probiotics have potential for KOA, especially Saccharomyces boulardii and LatiLactobacillus sakei LB-P12. However, small sample sizes and limited studies weaken evidence; further research is needed.

Introduction

KOA is one of the most frequent degenerative joint diseases among middle-aged or older people which leads to disabilities by cartilage degeneration. Synovitis, subchondral bone changes, and signs including pain, stiffness and impaired mobility. Existing therapies are unsatisfactory because prolonged use of Non-steroidal anti-inflammatory drugs (NSAIDs) causes damage to many organ systems, glucosamine is of uncertain effectiveness, there exist underdevelopment of Disease-modifying antirheumatic drugs (DMARDs), late stage joint replacement comes with complications (1, 25).

The “gut-joint axis” theory suggests that the imbalance of gut microbiota causes a series of systemic inflammatory responses through the “microbiota-gut-immune” axis—pathogens secrete Lipopolysaccharide (LPS), which passes through the mucosa of the gastrointestinal tract, activates Toll-like receptor 4 (TLR4) signaling, and promotes the release of proinflammatory mediators [Interleukin-6 (IL-6), Tumor necrosis factor-alpha (TNF-α)] in order to accelerate joint inflammation and cartilage degradation (2, 3). KOA patients have gut dysbiosis (higher Bacteroidetes/Firmicutes, lower Lactobacillus) and this is related to disease severity (4), suggesting intervention opportunity.

Theoretically, probiotics treat KOA by colonizing the gut and inhibiting pathogens, strengthen the mucosal barrier, regulate immunity, and reduce inflammation. Basic research demonstrates that Lactobacillus and Bifidobacterium (24) decrease intestinal endotoxin, down-regulate systemic inflammation, and downregulate synovial MMP-13; other strains, such as LatiLactobacillus sakei LB-P12, also control Short-chain fatty acid (SCFA) production and Nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) through the gut–joint axis (5). Probiotic is safe in use for long-term period on GI disease (6).

Nevertheless, there are still controversies or inconsistencies about the conclusion for KOA with regard to probiotics interventional treatment among existing RCTs and meta-analysis. Particularly, recent meta-analysis draw different conclusion from the same or similar data set. For instance, although Tian et al. (7) find taking pro-biotic could effective to raise WOMAC score and VAS score, Moyseos et al. (8) and Zeng et al. (9) concluded that there was not enough evidence at present to recommend the use of probiotics because of a high risk of bias and methodological limitations. Furthermore, the previous meta-analysis included few trials (only one or two to five), were based on one strain, lacked sufficient heterogeneity tests; thus, it is difficult to draw a conclusion about whether probiotics can be used clinically in patients with KOA.

This research will comprehensively analyze and quantitatively integrate the effectiveness and security of probiotics for KOA based on the conventional meta-analysis method, providing a good reference to use in the clinic or further study.

Materials and methods

Study design

This narrative and meta-analysis followed the PRISMA statement, using published data without original clinical data collection or participant consent, so no ethics approval was required.

Inclusion and exclusion criteria

The inclusion criteria including: KOA patients (ACR/EULAR criteria, Kellgren-Lawrence ≥ 1, no age/gender/ethnicity limits) receiving single/multi-strain probiotics; control group without probiotics; studies including primary (VAS, WOMAC pain/function) and secondary [serum/synovial Interleukin-1 beta (IL-1β), TNF-α, C-reactive protein (CRP)] outcomes; English/Chinese RCTs, including formally published journal articles and publicly accessible thesis papers approved by academic institutions. The exclusion criteria showed as follows: Non-RCTs, unpublished gray literature (e.g., unfiled internal reports, thesis papers not approved by academic institutions), animal/in vitro studies, and intervention groups without probiotics.

Literature search strategy

The literature search strategy was following the PICO principle. P (patients): knee osteoarthritis, knee OA, gonarthrosis; I (Intervention): probiotics, lactobacillus, bifidobacterium; C (Comparator): placebo, usual treatment; O (Outcomes): VAS, WOMAC pain/function, serum/synovial interleukin-1 beta (IL-1β), TNF-α, C-reactive protein (CRP).

The following databases were searched: English (PubMed, Cochrane Library, Web of Science); Chinese (CNKI, Wanfang Data). Period: January 1, 2015–December 30, 2025. The PubMed queries were as follows: ((((knee osteoarthritis) OR (knee OA)) OR (gonarthrosis)) AND (((probiotics) OR (lactobacillus)) OR (bifidobacterium))) AND (((((((VAS) OR (WOMAC function)) OR (WOMAC pain)) OR (serum interleukin-1 beta)) OR (IL-1β)) OR (TNF-α)) OR (C-reactive protein (CRP))).

Supplementary: Included study references and recent systematic reviews/meta-analyses were manually reviewed to identify missed RCTs.

Supplementary explanation: Publicly accessible thesis papers were included in the search process. These papers have been reviewed by the academic committee of the institution, with data collection and analysis processes complying with academic standards, and providing complete research design and original data, which is of great value for supplementing the evidence of this study.

Literature screening and data extraction

Two investigators independently screened titles/abstracts, then full texts, resolving conflicts via discussion or third researcher input. A pre-designed form extracted study characteristics (authors, year, design), participant demographics (sample size), intervention details (strain, dosage, duration), and outcomes (inflammatory markers, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), VAS).

Quality assessment of included studies

Most included studies had low bias risk in key domains (random sequence generation, allocation concealment, blinding, selective reporting) per detailed methodological descriptions [e.g., (10, 11)]. A small subset [e.g., (12, 13)] had unclear risk for attrition (incomplete missing data reporting) and other bias (limited confounding variable details). No high bias risk studies were found (Figure 1).

Panel (a): Risk of bias summary; Panel (b): Risk of bias graph.

Statistical analysis

RevMan 5.4 and Stata 17.0 were used to analyze data. Heterogeneity was tested via Q test and I2: fixed-effects was applied if I2 ≤ 50% and p ≥ 0.10, otherwise, the random-effects. For continuous outcomes, weighted mean difference (WMD) was used for outcomes measured by the same tools or standardized mean difference (SMD) when outcomes were measured by different tools. For dichotomous outcomes (adverse events), the Relative risk (RR) was used, all with 95%CI. Sensitivity analysis (sequential study exclusion) and publication bias were assessed by Stata. For core outcomes (VAS pain score, CRP level), publication bias was assessed via Egger’s test and small-study effect regression plots. Funnel plots and trim-and-fill method were assessed when enough studies were included. p < 0.05 indicated significant bias.

Results

Literature search and screening results

Records identified through the search processes (n = 180) and other sources (n = 20: manual = 9, reference tracing = 11). After removal of duplicates (n = 40), 160 studies were included in the screening process. After 125 irrelevant studies excluded from study, 35 studies were left for full-text assessment. Of theses, 27 were excluded because of inappropriate study design, incomplete data, or because they were meta-analyses excluded from data synthesis. Finally, eight studies met the criteria (Figure 2).

Literature search and screening results. Initial search and supplementary search yielded 180 + 20 = 200 records. After excluding 40 duplicate records, 160 records were screened; 125 records were excluded due to irrelevant topics/designs, unpublished gray literature, etc., leaving 35 for full-text evaluation; 27 were excluded due to inappropriate study design, incomplete data, or being meta-analyses excluded from data synthesis, and finally 8 studies were included.

Basic characteristics of included studies

Eight RCTs (10–17) whose basic demographic data were similar and be included in the analysis of this review. The intervention group used one kind of strain (Lactobacillus casei Shirota), multi-strains (Vivomix 8), or a combination of probiotics and glucosamine/physical therapy. The control using placebo with adjunct intervention. Duration of the intervention was 3–6 months with the main results including hs-CRP, WOMAC, VAS (Table 1).

Table 1Basic characteristics of included studies.Author (year)Sample size (N)InterventionPeriodOutcomesStudyControlStudyControlLei et al. (12)215218 Lactobacillus casei Shirota Placebo6 monthsHigh-sensitivity C-reactive protein (hsCRP); WOMAC; Visual Analog Scale (VAS)Lyu et al. (13)3730TCI633Placebo12 weekshs-CRP; WOMACWang (14)3728Bifidobacterium lactis + calcium+ chondroitinPlacebo + calcium+ chondroitin4 monthsWOMAC; Inflammatory factors; Bone densityHan (15)3234Bifidobacterium + glucosaminePlacebo + glucosamine3 monthsWOMAC; VAS; Inflammatory factors; Body mass index (BMI)Dolatkhah et al. (16)3231S. boulardii + physical therapyPlacebo + physical therapy12 weekshs-CRP; WOMAC; VAS; Inflammatory factorsShine et al. (11)4547 LatiLactobacillus sakei LB-P12 Placebo (carboxymethyl cellulose)12 weeksWOMAC; VAS; Inflammatory factors; Joint space widthKarim et al. (17)b6558Vivomix (8strains)aPlacebo (fruit powder)12 weeksWOMAC; VAS; Inflammatory factors; Bone densityKarim et al. (10)4649Vivomix (8strains)aPlacebo16 weeksVAS; Physical function; Inflammatory factorsaVivomix (8 strains): Streptococcus thermophilus DSM 24731, Bifidobacterium longum DSM 24736, B. breve DSM 24732/24737, Lactobacillus spp. DSM 24735/24730/24733, L. delbrueckii subsp. bulgaricus DSM 24734.bThe study labeled as “Rizwan Qaisar 2024” in the forest plots corresponds to the same clinical trial as Karim et al. (17), with identical patient cohorts and outcome data. The discrepancy in first authorship is due to different author orderings in the publication and dataset presentation.

Basic characteristics of included studies.

Vivomix (8 strains): Streptococcus thermophilus DSM 24731, Bifidobacterium longum DSM 24736, B. breve DSM 24732/24737, Lactobacillus spp. DSM 24735/24730/24733, L. delbrueckii subsp. bulgaricus DSM 24734.

The study labeled as “Rizwan Qaisar 2024” in the forest plots corresponds to the same clinical trial as Karim et al. (17), with identical patient cohorts and outcome data. The discrepancy in first authorship is due to different author orderings in the publication and dataset presentation.

Meta-analysis, publication bias and small-study effect analysis

Pain score improvement

VAS pain

Five studies showed that probiotics significantly reduced VAS scores vs. controls (MD = −0.11, 95%CI: −0.15 to −0.08, Z = 6.24, p < 0.00001; Figure 3).

VAS pain scores.

Publication bias and small-study effect (VAS pain)

Egger’s test for VAS pain (Table 2) suggested no obvious publication bias (p = 0.054 > 0.05). Small-study effect regression analysis (Figure 4) indicated no obvious small-study bias (p = 0.828 > 0.05). Combined with the results of sensitivity analysis, the robustness of the results of this outcome indicator was initially supported, but it still needed to be interpreted with caution.

Table 2Results of Egger’s regression test for publication bias.MetricCoefficientStandard errort priceP price95% confidence intervalSlope0.14780.04813.070.054−0.0053 ~ 0.3008Bias−2.33471.1536−2.020.136−6.0058 ~ 1.3365Test of H0: no small-study effects, p = 0.136.

Results of Egger’s regression test for publication bias.

Test of H0: no small-study effects, p = 0.136.

Small-study effect regression analysis of VAS pain scores. Each scatter point represents one RCT, and the red line is the regression fit line. The regression analysis showed p = 0.828, indicating no significant small study effect.

Supplementary explanation: Due to the number of RCTs included for core outcome indicators (VAS pain score, CRP level) being only 5 and 4 studies respectively, and the applicability of funnel plots and the trim-and-fill method is limited, so these methods were not used for publication bias assessment. The results of the publication bias analysis in this study are limited by the sample size and have certain limitations; future studies need to include more high-quality RCTs for further verification.

WOMAC pain

Two studies (268 participants) showed probiotics lowered WOMAC pain scores (MD = −2.48, 95%CI: −4.55 to −0.40, Z = 2.34, p = 0.02) with high heterogeneity (I2 = 91%, p = 0.001; random-effects model; Figure 5).

WOMAC pain scores.

Joint function score improvement

WOMAC total

Two studies (268 participants) showed slight but significant WOMAC total score reduction with probiotics (random-effects model; Figure 6).

WOMAC total scores.

WOMAC stiffness

Two studies (268 participants) showed probiotics reduced stiffness scores (MD = −0.26, 95%CI: −0.35 to −0.17, Z = 5.38, p < 0.00001; Figure 7).

WOMAC stiffness score.

WOMAC physical function

Two studies (268 participants) showed no significant improvement (MD = −10.16, 95%CI: −21.54 to 1.22, Z = 1.75, p = 0.08) with high heterogeneity (I2 = 96%, p < 0.00001; random-effects model; Figure 8).

WOMAC physical function score.

Inflammatory cytokine changes

Four studies (350 patients) showed probiotics reduced serum CRP (MD = −1.08, 95%CI: −1.73 to −0.42, Z = 3.23, p = 0.001) with high heterogeneity (I2 = 94%, p < 0.00001; random-effects model; Figure 9).

Changes in inflammatory cytokine levels.

Publication bias and small-study effect (CRP)

Egger’s test for CRP (Table 3) showing no obvious publication bias (p = 0.680 > 0.05). Small-study effect regression analysis (Figure 10) indicating no small-study bias (p = 0.680 > 0.05). The robustness of the results needs to be further confirmed with more research data.

Table 3Results of Egger’s regression test for publication bias.MetricCoefficientStandard errort priceP price95% confidence intervalSlope−0.65030.5665−1.150.37−3.0878 ~ 1.7872Bias−2.08144.3528−0.480.68−20.8104 ~ 16.6475Test of H₀: no small-study effects, p = 0.680.

Results of Egger’s regression test for publication bias.

Test of H₀: no small-study effects, p = 0.680.

Small-study effect regression of CRP. Each scatter point represents one RCT, and the red line is the regression fitted line. The regression analysis showed p = 0.680, indicating no significant small—study effect.

Adverse event analysis

Eight studies (1,129 patients) showed no significant difference in adverse event incidence between probiotics and placebo (RR = 0.71, 95%CI: 0.23–2.18, Z = 0.60, p = 0.55). The result was shown in Figure 11.

The forest plot of adverse event analysis.

Strain-specific efficacy

Subgroup analysis of 8 RCTs (strains: Lactobacillus brevis LB-P12, Streptococcus thermophilus TCI633, Vivomix, Bifidobacterium infantis BX-01, Lactobacillus casei Shirota, Saccharomyces boulardii, Bifidobacterium lactis Probio-M8) showed that.

Saccharomyces boulardii had the strongest analgesic effect (WOMAC pain MD = -3.57, 95%CI: −4.66 to −2.49) (16). Bifidobacterium infantis BX-01 (VAS MD = −2.11, hs-CRP MD = −1.89) and LatiLactobacillus sakei LB-P12 (WOMAC total MD = −1.52, VAS MD = −1.48, IL-1β MD = −1.32) excelled in symptom relief + inflammation regulation (11). Moderate heterogeneity (e.g., WOMAC/OKS I2 = 93.7%) was linked to strain, dosage, and duration (Figure 12).

Comparative efficacy among different strains.

Sensitivity analysis and heterogeneity of WOMAC outcomes

Leave-one-out sensitivity was shown in Figure 13 (N = 2). Result showed a very high sensitivity after removing Lei et al. (12) causing effect toward null, and after removing Shine et al. (11) the effect changed toward the opposite. Such dramatic change suggested that there was large amount of clinical heterogeneity instead of chance noise. In order to figure out what caused such large heterogeneity, we compared the main features between those two critical papers (Table 4).

Leave-one-out analysis (A: WOMAC Pain Index; B: WOMAC Physical Function; C: WOMAC Total Score).

Table 4Sources of clinical and methodological heterogeneity.DomainIndicatorLei et al. (12)Shine et al. (11)Impact on heterogeneityBaseline severityWOMAC scores (total, pain, function)HighLowMarked baseline disparity across all dimensions induced a systemic ceiling effect, uniformly limiting absolute improvement in the Shine cohortDisease stageKL gradeGrade 2–3Grade 1–2Earlier structural stage correlates with attenuated therapeutic responsiveness in both pain and function domainsInflammatory mechanismPrimary biomarkerhs-CRP ↓ (systemic)IL-1β ↓ (local)Divergent mechanistic pathways (systemic vs. local) create non-comparable biological endpointsIntervention durationTimeframe24 weeks12 weeksShorter duration likely precludes cumulative functional benefits, contributing to variance in long-term outcomes

Sources of clinical and methodological heterogeneity.

For the main features compared in Table 4, we found the following four reasons causing heterogeneity: (1) Baseline severity: more severe state of illness existing in study of Lei et al. causing “ceiling effect” in Shine et al.; (2) Stage of disease: different fibrosis (K-L grade 2 to 3) or different levels of inflammation; (3) Type of marker: inflammatory factors were hsCRP or cytokine (IL-1b); (4) Treatment duration: 24 or 12 weeks. All these factors may account for a part of the high between-study variability.

Subgroup and meta-regression analysis

There were only two studies included in the comparison of WOMAC scores (pain, function, total) with considerable heterogeneities, and the subgroup analyses were unable to perform. For CRP, we did the meta-regression analysis, and the result showed that the mean ages were the source of the heterogeneity (Z = 1.99, p = 0.046).

In subgroup analysis of CRP, the treatment effect was different to some extent regarding the initial metabolism burden (Figure 14). In high burden group, there were more obvious reductions of CRP (MD = −1.76; 95% CI: −2.77a to −0.75d; I2 = 86%) while in the normal-load subgroup, we found a moderate homogeneous effect (MD = −0.53; 95% CI: −0.70 to −0.37; I2 = 0%). The heterogeneity between subgroups (p = 0.02) suggested pre-existing inflammatory status accounting for most variability.

Sources of heterogeneity in CRP outcomes: Subgroup analysis by baseline metabolic load.

To understand the heterogeneity (I2 = 86%) within the high load group, we found three sources of evidence (Table 5), (1) Taxonomy: different probiotics in study group (bacteria L. casei vs. fungi S. boulardii); (2) Time course: different therapy time (24 weeks vs. 12 weeks) and (3) Initial BMI: different adipokines-induced inflammatory responses between the restricted-weight group vs. non-restricted weight group. This could explain some of the inconsistency.

Table 5Clinical heterogeneity analysis in subgroup with high metabolic/inflammatory load.ParameterLei et al. (12)Dolatkhah et al. (16)Proposed mechanism for CRP heterogeneityProbiotic strain Lactobacillus casei Shirota Saccharomyces boulardii Distinct immunomodulatory pathways between bacterial and fungal kingdomsIntervention duration24 weeks (6 months)12 weeks (3 months)Longer duration may enhance cumulative anti-inflammatory effectsBaseline BMI statusNot restrictedOverweight/obeseAdipokine-mediated chronic inflammation may drive divergent CRP responsesCRP, C-reactive protein.

Clinical heterogeneity analysis in subgroup with high metabolic/inflammatory load.

CRP, C-reactive protein.

Discussion

Probiotics decreased VAS/WOMAC pain/stiffness score and modulated inflammation markers (CRP, IL-6, TNF-α) similar to the pain/function results (3, 18). This indicates that probiotics block the systemic inflammation in order to reduce the joint inflammation and cartilage damage, to achieve “anti-inflammation→analgesia→function improvement.”

Mechanistically, this is in line with a “gut-joint axis,” which describes how probiotics may provide systemic immunomodulation and metabolic reprogramming. The mechanisms seem to involve at least two main pathways: First, immunomodulation by T-cell polarization: some strains such as Lactobacillus casei Shirota induce the differentiation of regulatory T cells (Tregs), which contributes to the establishment of an antiinflammatory environment (6). S. boulardii improve intestinal permeability in order to avoid translocation of inflammatory pathogens triggering macrophages activation (19). Secondly, metabolic signal through microbiota derived metabolite: Interestingly, LatiLactobacillus sakei LB-P12 regulates short chain fatty acids (SCFAs) production and SCFAs down regulate NF-κB pathway by blocking histone deacetylases (HDACs), thus inhibiting catabolic enzyme (e.g., MMP-13), proinflammatory cytokine (e.g., IL-1ß, IL-6) production in the joint thus slowing down the rate of cartilage degradation.

Probiotics did not significantly impact WOMAC physical function (MD = −10.16, p = 0.08) likely due to the fact that functional recovery may require structural recovery (e.g., cartilage) and muscle strength improvement—probiotics affect inflammation/symptoms—not structure; short duration of interventions (8–24 weeks)—also might contribute indicating a requirement of longer follow-ups.

Therapeutic potential and underlying mechanisms of probiotics in KOA

Vasomotor symptoms were also decreased by probiotics as was the inflammation reflected in CRP, IL-6, and TNF-α; these findings paralleled those seen for pain/function (3, 18). Thus, it appears that the action of probiotics consists on inhibiting systemic inflammation aiming at reducing joint inflammation and cartilage damage, achieve “anti-inflammatory → analgesia → function improvement (26, 27).” The effects of probiotics were not statistically significant in WOMAC physical function (MD = −10.16, p = 0.08) possibly because there was high clinical heterogeneity. Different baseline severity and different structural stage among trials would dilute the pooled estimate. Probiotics mainly targeting inflammation instead of structural repair mayand the duration of interventions (8–24 weeks), were low, detectable functional improvements were limited.

Clinically, simultaneous amelioration on both pain and stiffness with reduction of CRP indicated that the application of probiotics may be able to modulate multiple aspects simultaneously in KOA. Although the overall magnitude of effects was relatively small, the larger effect of CRP lowering seen with the “higher metabolic / inflammatory burden” group (MD = −1.76), indicates that there may be differential responses based on particular clinical phenotype, and also preliminary safety data (RR = 0.71) suggest that the administration of probiotics is well tolerated as add on treatment. These results show a possible use for their application to reduce general inflammatory response, the ultimate clinical benefit needs to be validated by high quality trials.

Clinical significance and practical implications

The reported decreases of WOMAC total score (MD = −13.16; 95% CI: −26.29 to −0.04), and stiffness (MD = −1.13; 95% CI: −1.43 to −0.83) seem plausible given the Minimal clinically important difference (MCID) range (−7.0 to 14.6 points on total score and −0.5 to −1.6 point on stiffness) (20). This implies that the probiotic treatment might provide an effective degree of symptom alleviation, which could be experienced by KOA patients. Although this improvement of physical function (MD = −10.16) was not statistically significant (p > 0.05), it is still inside the range of MCID, which may indicate a potential clinical tendency. In addition, despite the relatively small numeric change of VAS pain (MD = −0.11 cm), the large reduction in hsCRP (SMD = −1.14) indicates that the mechanism by which probiotics work could be via systemic anti-inflammation and not directly as a strong analgesic agent. Therefore, as a clinical-relevant adjuvant treatment, it is possible that probiotics could be valuable in the multi-dimensional and long-term management of KOA.

Strain-specificity

Subgroup analysis identified differential strain effects: Lactobacillus casei Shirota (VAS MD = −2.65, WOMAC total MD = −2.70, low heterogeneity) was optimal (9, 12); Saccharomyces boulardii focused on analgesia (16).

Bifidobacterium infantis BX-01 and LatiLactobacillus sakei LB-P12 combined symptom relief and anti-inflammation (11). These differences may relate to metabolites, gut colonization, and immunomodulation (e.g., Lactobacillus casei Shirota regulates Treg cells; Saccharomyces boulardii protects intestinal mucosa) (6, 19).

Sensitivity analysis and reliability

Sensitivity analysis (leave-one-out) shows that VAS and WOMAC estimations are stable without any single study having a strong impact on the results; heterogeneity mainly comes from different intervention protocols (strain, dose, and duration); trials with consistent strain-specific protocols [e.g., (9)] exhibit lower variability. Furthermore, the favorable safety profile (AE incidence: probiotics 0.71% vs. placebo 1.07%, p = 0.55) suggests good tolerance for probiotic usage. Although promising regarding the alleviation of symptoms more research is necessary to obtain consistency and reproducibility of large scale RCTs.

Strengths, limitations, and future directions

Strengths

Compared with other recent reviews (7–9), this review may give broader perspective because it includes more studies in term of nRCTs (n = 8), which can increase power to detect KOA specific responses. One of our strengths will be an effort to bring together the areas of pooled efficacy with those of clinical use in a finer grain, strain-specific manner (e.g., S. boulardii vs. LB-P12). In addition, through defining “high metabolic/inflammatory load” group as a high-response group (CRP SMD = −1.76), we provided some initial information about individualized clinical decisions which are not widely discussed in previous meta-analyses (Table 6).

Table 6Comparison of methodological characteristics and key findings between the present study and previous meta-analyses.Methodological dimensionZeng et al. (9)Moyseos et al. (8)Tian et al. (7)Present studyScope and focusUmbrella (8 arthritis types)OA/RA specificKOA specificKOA specificNo. of KOA RCTs135 8 Strain-specific analysisNone (single strain)Limited (LcS focus)None (pooled effect) Yes (SB vs. LB-P12) Heterogeneity managementSuperficialIdentified biasSensitivity only In-depth (subgroups) Key conclusionInsufficient evidenceLcS effectiveOverall efficacy Strain-specific + mechanism SB, Saccharomyces boulardii; LB-P12, LatiLactobacillus sakei LB-P12. Bold text indicates the unique strengths and novel contributions of the current study relative to prior research.

Comparison of methodological characteristics and key findings between the present study and previous meta-analyses.

SB, Saccharomyces boulardii; LB-P12, LatiLactobacillus sakei LB-P12. Bold text indicates the unique strengths and novel contributions of the current study relative to prior research.

Limitations

Limited sample size of some outcome (e.g., BMI/BMD), few study about specific strain (e.g., Streptococcus thermophilus TCI633) (13), and no long-term KOA progressions. In addition, only eight RCTs were included (<10), therefore the publication bias was tested by Egger’s test and small-study effect regression plot, with trim-and-fill analysis (≥10 studies) unavailable. Another limitation, the evaluation of publication bias in this article also exists some shortcomings. For core outcomes indicator, there are less than 10 RCTs included, leading to poor statistical power of methods like Egger’s test that are unable to fully exclude possible influence from publication bias, and one has to be careful in interpreting the corresponding result. Therefore, it is important to note that any beneficial effects seen in symptoms cannot be taken as a true demonstration of efficacy.

Future research should

Conduct large-sample, long-term RCTs on priority strains (Lactobacillus casei Shirota, Bifidobacterium infantis BX-01) to assess long-term joint protection (11, 12, 28).

Use gut microbiome sequencing/metabolomics to clarify strain-specific molecular mechanisms (21, 22).

Explore personalized interventions (age, KL grade, baseline microbiota) and build clinical decision models for strain selection (9, 23).

Conclusion

Probiotic relieves the pain of KOA (VAS, WOMAC pain), improves joint function, reduces serum inflammation, and has a safety similar to placebo, with Saccharomyces boulardii (16) and LatiLactobacillus sakei LB-P12 (11) showing promising results. However, smaller study size and fewer trials dilute strength; further they are clinically relevant only to a lesser extent due to smaller magnitudes of effects. Thus, there is a need for large-sample, high quality RCT’s that can help confirm both efficacy and strain-specificity.