Efficacy of Probiotics in Rheumatoid Arthritis and Spondyloarthritis: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background: We aimed to provide a systematic review and meta-analysis of randomized controlled trials assessing the effect of probiotics supplementation on symptoms and disease activity in patients with chronic inflammatory rheumatic diseases (rheumatoid arthritis (RA), spondylarthritis (SpA), or psoriatic arthritis). Methods: A systematic literature review and meta-analysis from RA and SpA randomized controlled trials were conducted searching for articles in MEDLINE/PubMed and abstracts from recent international rheumatology meetings. The control group was a placebo or another dietary intervention. The risk of bias of the selected studies was evaluated using the Cochrane Collaboration tool and the Jadad scale. Results: The initial search yielded 173 articles. Of these, 13 studies were included in the qualitative synthesis, 8 concerning a total of 344 RA patients and 2 concerning a total of 197 SpA patients. Three meta-analyses were also analyzed. Probiotic strains and quantities used were different among trials (5 studies using Lactobacillus sp., 1 trial Bacillus coagulans and the others a mix of different probiotic strains). Time to assess response ranged from 8 weeks to one year. Two studies associated probiotic supplementation with a dietary intervention. Meta-analysis showed a statistically significant decrease of C-reactive protein (CRP) concentration (mean difference (MD)) −3.04 (95% CI −4.47, −1.62) mg/L, p < 0.001; I2 = 20%, n patients = 209) with probiotics in RA. However, after excluding high-risk-of-bias trials of meta-analysis, there was no difference between probiotics and placebo on DAS28 (standard MD −0.54; 95% CI −1.94 to 0.85, p = 0.45, I2 93%, n patients = 143). The two studies on SpA patients showed no efficacy of probiotics. Conclusions: Probiotic supplementation might decrease RA activity with a moderate decrease effect on CRP, but lack of evidence and studies’ heterogeneity do not allow us to propose them to patients with inflammatory arthritis to control their disease. Further RCTs are required in the future to determinate the efficacy of probiotics and the optimal administration design.


Introduction
The World Health Organization in 2002 defined probiotics as "living organisms in food and dietary supplements that upon ingestion can improve the health of the host beyond their inherent basic nutritional content" [1]. Naturally presents in fruits, raw vegetables, dairy products (in particular fermented ones), they are an integral part of the gut microbiota as a component of commensal flora. The main probiotic microorganisms used in human nutrition are lactic acid bacteria such as Lactobacillus and Bifidobacterium [2].
Compared with healthy people, patients with chronic inflammatory diseases, in particular rheumatoid arthritis (RA) and spondyloarthritis (SpA), have an altered gut microbiota called dysbiosis with an increased permeability allowing luminal antigens or bacteria to interact with the host immune system more readily [3,4]. This gut inflammation correlates with systemic inflammation and could be a trigger in developing some autoimmune diseases and participating in their severity [5][6][7][8].
Evidence from mice and human studies revealed that probiotics modulate locally and systemically the immune system, leading to a reduction in mucosal inflammation and pro-inflammatory cytokines [9]. They also alleviate joint inflammation in mice [10]. They can access the intestinal mucosal immune system, persist for a certain amount of time, and initiate a specific immune response. The interaction between probiotic strains and enterocytes is important for the controlled production of cytokines and chemokines secreted by epithelial cells. Indeed, it has been shown that some probiotic organisms can modulate the in vitro expression of pro-and anti-inflammatory molecules in a strain-dependent manner. Indeed, treatment with some Lactobacillus strains reduced gut permeability and decreased arthritic severity [11,12]. The impact of probiotics has already been well studied in atopic diseases and Crohn's disease, for which they have not shown any real benefit.
Today, the growing interest of patients in the use of complementary therapies is justified by the existence of numerous side effects of usual disease-modifying anti-rheumatic drugs (DMARDS) and symptomatic treatments such as non-steroidal anti-inflammatory drugs (NSAIDs) and glucocorticoids (GCs). Probiotics could represent an alternative and complementary therapy to the standard drugs we already use to control rheumatic activity. However, the effect is not well documented in patients suffering from RA, SpA or psoriatic arthritis (PsA), especially recently, where studies are rare.
The available data from randomized controlled trials are highly heterogeneous in terms of the populations included, the characteristics of the rheumatism, and the results regarding activity scores and inflammatory markers. The three meta-analyses published attempted to provide conclusions on the efficacy of probiotics in patients with inflammatory rheumatism, with disease activity score in 28 joints (DAS28) and C-reactive protein (CRP) as the common primary endpoints. Two focused on RA, and the third included patients with SpA and juvenile idiopathic arthritis (JIA). The number and type of studies included as well as the population studied were therefore not homogeneous, and the results are discordant and difficult to interpret. The meta-analysis we performed is the first to pool only studies without high risk of bias and therefore to provide more conclusive results. It is also the first systematic review of the literature to examine the efficacy of probiotics in patients with SpA.
We aimed to provide a systematic literature review (SLR) of randomized controlled trials (RCT) and meta-analyses assessing the effect of probiotics on inflammatory rheumatic diseases' symptoms and activity. This SLR was used to inform the recommendations of the French Society of Rheumatology (SFR) on diet in inflammatory rheumatic diseases [13].

Materials and Methods
This SLR of RCT and meta-analyses was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [14].

Search Strategy
Relevant articles were found using the MEDLINE database via PUBMED from inception to June 2020. A bibliography of selected retrieved articles was assembled and reviewed for inclusion by 2 researchers and a manual search was carried out. Conference abstracts from selected rheumatology meetings (European League Against Rheumatism (EULAR) and the American College of Rheumatology (ACR)) and from nutrition meetings (International Congress of Nutrition, European Nutrition Conference, American Society of Nutrition) from 2017 to 2019 were searched and manually reviewed for inclusion too.
Screening title and abstract took place initially followed by full-text screening. Search results were retrieved and duplicates were removed using Covidence software, V.1. developed by Veritas Health Innovation Ltd. (Melbourne, Australia). Disagreement was resolved by discussion and consensus between reviewers and senior researchers.

Eligibility Criteria
We included any open-label or blinded randomized controlled studies concerning adult patients with diagnosis of SpA, (PsA) or RA for whom an oral supplementation of probiotics was administered and disease activity assessed. This probiotic supplementation could be mixed with another diet in the intervention group. Control could be a placebo or another dietary intervention. We also included previous meta-analysis in our systematic literature review, to compare with ours and to make a synthesis of the available data.
We excluded any uncontrolled study, case reports, case series, letters, literature reviews, editorial comments, theses, reviews, book chapters, news, or isolated abstracts. We also excluded papers if their data were insufficient to be extracted or if not written in English or French.
Outcome measures included RA clinical activity markers such as disease activity score of 28 joints (DAS-28), EULAR or ACR response, Health Assessment Questionnaire Disability Index (HAQ), number of tender or swollen joints (TJC and SJC), visual analog scale (VAS) for disease activity by the patient, VAS for pain and global health score (GH score). Concerning SpA and PsA, outcomes were Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Bath Ankylosing Spondylitis Functional Index (BASFI), VAS for disease activity, quality of life (ASQoL), SJC, TJC, global well-being and bowel symptoms. Laboratory markers were C-reactive protein (CRP) and/or erythrocyte sedimentation rate (ESR). Disagreement in the determination of the eligibility of each study was resolved by consensus.

Data Extraction
Data of interest were collected from the included articles using the standardized extraction form. The extracted variables were publication date, journal, disease characteristics (including activity score, treatments such as DMARDs and symptomatic medications (GCs and NSAIDs), duration of the disease and positivity of rheumatoid factor (RF) and/or anti-citrullinated protein antibodies positive (ACPA)), study design, inclusion and noninclusion criteria, probiotic strains, sample size, results in terms of disease activity, side effects and adherence.

Quality Assessment
Risk of bias assessments used the Cochrane Collaboration tool for assessing risk of bias [15] and the Jadad scale [16]. Records limited to abstracts were not assessed because of the lack of information about study design. Two reviewers evaluated the quality of the studies independently (P.S. and C.D.).

Statistical Analysis
For the meta-analyses, the outcomes were the variation of DAS28 and CRP between inclusion and evaluation endpoint comparing the two groups of treatment (with or without probiotics). A narrative synthesis has been carried out to describe data extracted from articles that could not be included in the meta-analyses.
A standardized mean difference was estimated for DAS28 calculations due to heterogeneity in DAS28 scores across the included studies: some studies calculated the DAS28-ESR, and some calculated the DAS28-CRP. A mean difference (MD) was used for CRP calculations.
This provided a common weighted MD (or standardized MD) estimate with a 95% CI, taking into account weighting of the different samples. MD and 95% CIs are expressed as forest plots. Statistical heterogeneity of the selected studies was tested using the Qtest (χ 2 ), applying a 0.05 statistical significance cutoff, and reported with the I 2 statistic in which high values of I 2 , ranging from 0% to 100%, represent strong heterogeneity. In case of a significant heterogeneity, a random effect model was applied to take into account heterogeneity. All computations were performed using the RevMan V.5.3 software package developed by Nordic Cochrane Centre (Review Manager (computer program), V.5.3. Copenhagen, Denmark: The Nordic Cochrane Centre, the Cochrane Collaboration, 2011). p values lower than 0.05 were considered significant. Publication bias was checked through Egger's test.

Study Selection
The literature review recognized a total of 172 records and one additional article was identified manually as shown in Figure 1. Of those, 137 were excluded after screening. Then, 23 studies were excluded after review because of wrong type of literature (n = 16), outcome (n = 1) or population (n = 5). One study was excluded because of the overlap with another study published by the same researchers at the same time [17]. Finally, 13 studies were included in the qualitative synthesis (including 10 RCTs and 3 meta-analyses) and 3 in our meta-analysis. No relevant unpublished studies were obtained from abstract meetings.  [18][19][20], 2 studies assessed a mix of two Lactobacillus strains [21,22], one study assessed Bacillus coagulans [23], the others assessed a mix of different probiotic types [24][25][26][27]. Time to assess response ranged from 8 weeks to 1 year. The comparator was placebo except in Vadell et al., which compares an anti-inflammatory diet rich in probiotics versus a typical Swedish diet, and in Nenonen et al., which compares an uncooked vegan diet rich in Lactobacilli versus a normal diet.  For RA, most studies assessed the efficacy on probiotic supplementation in reducing its activity in patients with active disease except one in which mean DAS-28 scores were below 2.6 [18]. Of note, three studies reported serum high-sensitivity CRP (hs-CRP) level change [18,24,25].

Study Characteristics and Search Results
The outcome measurement corresponds to the date at which the primary and secondary endpoints were assessed, which also corresponds for all trials to the duration of probiotic or placebo/control supplementation.
The main inclusion criteria was RA diagnosed according to the 1987 ACR/EULAR criteria [18,19,24,25], other inclusion criteria were a disease duration of more than 1 year [18,19,23], more than 6 months for Zamani's et al. and more than 2 years for Vadell et al. Treatments had to be at stable doses from 1 to 3 months prior to inclusion [18][19][20][21]. Two trials did not meet precisely the minimum activity score required [18,19] whereas all others referred to at least mild activity according to the DAS28.
Concerning SpA studies, patients were included only if they had a sacroiliitis (radiographic or magnetic resonance imaging (MRI)) in one of these studies [27]; the second one assessed patients with SpA without distinction on clinical phenotypes (i.e., axial, peripheral and/or enthesitic) [26].
Mean disease duration was between 4.7 and 19 years for RA and between 7.9 and 20.3 years for SpA. Little information was available concerning the proportion of patients taking targeted therapies; one study in RA specified the use of DMARD as a non-inclusion criteria [19] whereas another one in RA specified the use of NSAIDs as a non-inclusion criteria [18].

Risk of Bias within Studies
The risk of bias assessment ranged from low to high and is illustrated in Figure 2 and Table S1 (Supplementary Materials). All studies were double-blinded except 2 singleblinded studies whose probiotics were not the unique intervention, thus had a high risk of bias with Cochrane's tool and a Jadad score of 3 [20] and 2 [22]. Two other studies were considered to have a high risk of bias because of attrition bias [18] and inappropriate data reporting [23]. Most studies were monocentric. Two studies had a Jadad score of 5 [26,27].

Results in Rheumatoid Arthritis
Study outcomes are summarized in Table 3, sorted according to significant results in favor of probiotic supplementation or not. Results from former meta-analyses [28][29][30] are detailed in Table 4; data from our meta-analysis are shown in Figure 3.      • DAS28 Six studies provided data on DAS28 [18,[20][21][22]24,25]. A significant improvement of DAS28 was achieved by three trials comparing probiotic supplementation with placebo [18,24,25]. On other hand, three trials concluded with no significant results [20][21][22], in particular Vadell et al. comparing an anti-inflammatory diet (rich in fatty acid, fibers and probiotics) versus typical Swedish diet, and Nenonen et al. comparing an uncooked vegan diet (fermented wheat drink, wheat grass drink, dietary fiber) with normal diet. The fact that probiotics supplementation was mixed with another dietary intervention in these studies may have been a confounding factor and altered the quality of the results because the efficacy of probiotics is difficult to individualize. Moreover, the cross-over design is questionable, especially because the washout period is arbitrary and does not allow to clearly establish that the first diet taken had no influence on the microbiota composition during the second period of the study (Table 3).
Our meta-analysis included three studies after exclusion of high-risk-of bias studies [18,20,22], and revealed probiotic supplementation did not show a significant effect in reducing DAS28-CRP with a SMD (95% CI) of −0.  [17,21] (Table 4). In contrast, the meta-analysis from Mohammed et al. concluded there was no significant effect with SMD (95% CI) 0.023 (−0.584 to 0.631), p = 0.94, I 2 = 73%, n patients = 132, but included a high-risk-of-bias study [18]. Lactobacillus was the only strain used in all trials included, except Zamani RCTs, which pooled different probiotic strains.

Inflammatory markers
We found eight studies provided data on CRP and/or ESR and 3 of them concerned hs-CRP (Table 3) [18,24,25]. In our SLR, three studies provided significant reduction of hs-CRP whereas results of four studies indicated no significant beneficial effects of probiotic supplementation on CRP [19,[21][22][23]. No significant decrease concerning ESR was reported.
Our meta-analysis pooled five RCTs and revealed a significant effect in reducing CRP with a MD (95% CI) of −3.04 mg/L (−4.47 to −1.62 mg/L) (p < 0.001) by analyzing a total of 209 patients ( Figure 3B). The heterogeneity was moderate (p = 0.29; I 2 = 20%). Meta-analysis from Mohammed  . However, two trials with SpA and JIA patients were added and post-treatment values rather than variations were compared between intervention and comparator groups [26,31]. In the sub-population analysis on RA patients, no benefit of probiotics was found (Table 4).

Results in Spondyloarthritis
Studies' results are shown in Table S2 (Supplementary Materials). Two studies concluded no significant decrease in any disease activity markers. Intestinal chronic inflammatory disease was an exclusion criteria for Jenks et al. and about 75% of patients had been taking NSAIDs in both groups at baseline, which may have altered gut microbiota.

Tolerance Data
Five studies concerning RA reported the absence of any side effects related to probiotic supplementation [18,21,[23][24][25]. One RCT did not report any information on this subject [19]. Vadell et al. described a 30% proportion of patients during the intervention periods who experienced moderate digestive intolerance such as nausea, diarrhea, stomach pain, heartburn and gas but no dropouts were related and most of them were present only at the start of a diet period.
Adverse events from spondyloarthritis RCTs were well documented: Vadell et al. mentioned the occurrence of nausea, increased flatus, diarrhea, vomiting and abdominal pain in the same proportion between the two groups. Brophy et al. reported 5 people in the placebo group and 6 in the probiotic group with adverse events, mainly stomach cramps (3) and indigestion (1) in both groups.

Discussion
This work updated the SLR and meta-analysis on probiotics in chronic inflammatory rheumatism and found: (1) a non-significant decrease of DAS28-CRP, (2) a significant decrease of CRP levels with RA patients under probiotics, (3) the absence of efficacy in patients with SpA. This work was used to inform the French Society of Rheumatology guidelines on diet in chronic inflammatory rheumatisms.
Bifidobacterium and Lactobacillus are known to provide many beneficial effects, in mice and humans. A recent meta-analysis in 2020 reported significant effects in the reduction of blood pressure, lipid profile, BMI and serum glucose [32]. Other evidence has shown their power of reducing intestinal permeability and modulating immune function through direct interaction with the mucosal immune system [33,34]. Immunological studies revealed that probiotics have dose and duration-dependent immunomodulatory effects on B and T cell proliferation and affect pro-inflammatory and anti-inflammatory cytokine regulation [35]. Probiotics administration can restore the normal mucosal barrier function through keeping the balance between intestinal microflora and resistance against harmful bacterial colonization, adherence and translocation [36]. These effects are dependent on the species and strain of bacteria [37]. For instance, oral administration of Lactobacillus rhamnosus attenuates various types of experimental arthritis [38]. Increasing evidence suggests that gut dysbiosis in RA and SpA patients favors inflammation, participating in disease activity and severity [39,40]. However, the data on gut microbiota composition are controversial and the strains contained in probiotics are not specifically decreased.
Concerning RA, although a statistically significant decrease in CRP was identified, cautious interpretation is required before inferring clinical significance. First, the reduction in CRP with a MD (95% CI) of −3.04 mg/L (−4.47 to −1.62 mg/L) may not represent a clinically meaningful change. On the other hand, some studies reported normal baseline values of both CRP and ESR, but they did not provide the data and it could not be extracted a posteriori. That may partially explain the absence of significant results regarding DAS28, of which inflammatory markers are one of the components. This systematic review of the literature provided very different baseline characteristics and inclusion criteria for both RA and SpA patients; e.g., only women were included in Alipour et al. [18], the minimum duration of disease progression at inclusion was variable, it had to be either more than 6 months for Zamani's et al., more than 2 years for Vadell et al., one year for the others or unspecified. Despite this, the average disease duration was more than 5 years in all studies, provided longstanding RA and SpA. Therefore, as studies did not provide newly diagnosed RA and SpA, the results cannot be extrapolated to this population. Another limitation of this meta-analysis is the small sample size, which may affect the reliability and validity of the results. Besides, four studies were monocentric [18,20,22,23], once again limiting the generalization of the results. Those are important limitations to propose the routine use of probiotics in patients with RA.
Concerning DAS28, three studies showed a significant improvement comparing probiotic supplementation with placebo, but results concerning TJC and SJC are not clinically pertinent [18,24,25]. In fact, in Alipour et al., patients in both groups had no or few TJC or SJC. It is important to note that studies included in meta-analyses for evaluation of DAS28, TJC, SJC, CRP, ESR and HAQ are different, which affects the comparability of the results. Furthermore, the control group was different with two studies comparing probiotics with another dietary intervention, no significant effect on DAS-28, CRP, ESR, TJC and SJC was observed in these trials, suggesting a potential confusion bias. Finally, we may suggest that it is not surprising that meta-analysis of Rudbane et al. reports a significant improvement of DAS28 by the fact that the only two studies which reported significant findings were included.
Concerning meta-analyses, as a primary outcome, we analyzed the variation of the values before and after supplementation, as Mohammed  The strengths of our meta-analysis are that we compared variations in outcome measures (such as DAS28 and CRP) between the two groups and included only studies which were RCTs with low or moderate risk of bias and homogeneous disease populations and added new studies. We were interested in three types of rheumatism because they are the focus of most of the literature data currently available on probiotics and represent the largest proportion of patients with inflammatory rheumatic disease. Concerning DAS28 and CRP, Alipour  Another strength of our meta-analysis is its methodological quality. We used the "A MeaSurement Tool for Assessing systematic Reviews" (AMSTAR2), designed to carry out rapid and reproducible assessments of the quality of conduct of systematic reviews of RCTs. It represents one of the most widely used instruments to date. Analysis of the methodological quality according to this tool revealed that it was critically low for the three previous meta-analyses. Ours had the advantage of being of higher quality because we removed RCTs at high risk of bias and conducted an adequate investigation of publication bias.
The results of Egger's test revealed the absence of publication bias for DAS28 (p value = 0.96) and a publication bias for CRP (p value = 0.02). This bias disappeared when sensitivity analyses were performed by removing the studies of Pineda et al. (p value = 0.07) or Mandel et al. (p value = 0.08) (data not shown). Ours to date is the only meta-analysis clearly describing publication bias regarding probiotics supplementation in RA, from which it is possible to conclude the reliability of the results regarding DAS28. The search for publication bias was theoretically limited by the small number of RCTs available in the literature, but our search strategy was intended to be as complete as possible, including searching for conference abstract data. The number of RCTs included is consistent with previous meta-analyses or higher. This limits the risk of not having included relevant RCTs.
Regarding spondyloarthritis, only two RCTs assessed the efficacy of probiotics supplementation versus placebo on SpA activity. No significant decrease of activity score or well-being was found. It is important to note the disparity between the type of patients included, from peripheral phenotype to ankylosing spondylitis and sacroiliitis confirmed by MRI. It would have been interesting to evaluate the effect of probiotic supplementation in a more homogeneous group of patients because these phenotype differences may have affected the final results. More studies are needed to assess the efficacy of probiotics in this selected population.
Tolerance data are reassuring; however, adverse events were not primary endpoints, therefore this data may have suffered from a lack of collection and precision in its measurement. All studies reported an excellent observance except that of Brophy et al., which suffered from less than 70% of tablets being taken. Furthermore, it has been shown that drugs influence the intestinal microbiota composition and as such might impact response to probiotics. At baseline, in general, stable antirheumatic medication between one and three months before inclusion were one of the inclusion criteria in RA studies. Almost all patients with RA appeared to be treated with either DMARD or symptomatic treatments such as glucocorticoids or NSAIDs. Only two studies specify the proportion of patients not taking any medication, that of Mandel et al. where it was equal to 6% in the probiotic group (n = 1/15) and 7% in the control group (n = 1/14), and Pineda et al. with 4% in the probiotic group (n = 1/23) and 13.6% in the control group (n = 3/22). However, most of the studies did not mention whether any modification of medication occurred during follow-up, especially biological DMARD (bD-MARD), corticosteroids and NSAIDs use, which would be important confounding factors.
Disparity in publication years also influence the medications taken by patients, in particular bDMARD. Baseline rheumatism activities were also very different between studies, which influences microbiota composition and possibility of activity variations with intervention.
It is currently difficult to conclude whether probiotics are efficient or not because of a high heterogeneity in studies' design due to the use of different strains, quantities and duration of supplementation. In regard to the heterogeneity present between the RCTs, a large heterogeneity exists between the previous meta-analyses, which can explain the differences in their results concerning the DAS28 and the CRP. Indeed, the main limitation is the wide variety of probiotic strains, administration dose and duration among studies with insufficient power to perform subgroup analysis. No study argued for the dosage of probiotics based on a possible pathophysiological rationale or evidence-based medicine. In addition, strains were not adapted to the profile of the patient's initial microbiota; if they were targeted to a possibly deficient one it would have been interesting to see if the effect was significantly increased, which is currently analyzed with the new generation probiotics.
In addition to the limitations of the literature studied, our systematic review may have been impacted by the fact that the initial screening for inclusion and exclusion was undertaken by two reviewers, introducing the risk of human error.
In summary, according to the available evidence, a literature review provided data on probiotic supplementation in patients with established rheumatism and low markers of inflammation, impacting the generalizability of the results. There has been shown to be a significant decrease of CRP concerning RA patients, but clinical relevance is not demonstrated. Probiotic supplementation might decrease RA activity but lack of evidence and heterogeneity lead us to conclude they should not be proposed to patients with inflammatory arthritis to control their disease. However, the good tolerance and potential cardiovascular benefits should be noted and patients who are willing to take them should be informed but not discouraged. Further RCTs and meta-analysis are required in the future to determinate the efficacy of probiotics and in which administration schema.