Based on the PICOt (Population Intervention Comparator Outcome Time or Type of Study) framework, a search strategy was generated with MeSH terms and free terms, detailed in Appendix A. The MeSH terms included subheadings related to: drug therapy, drug effects, therapy, therapeutic use, and physiopathology.

One of the study investigators (NF) proposed a strategy, which was supervised by another investigator (LC). Based on this, a systematic literature search was conducted in the Medline (via PubMed), Embase, and Cochrane Library databases with no start date and up to February 18, 2024 (Appendix A). The only restriction applied was to limit the search to human studies. Articles written in a language other than English were included and translated.

The inclusion criteria were: studies that analysed the composition of HM in biological samples (1) from patients with SARDs, with the exception of spondyloarthritis because this information has already been extensively reviewed,11 (2) and compared it based on the IS treatment used or the response to it. (3) No limits were set regarding disease stage, geographic location, or publication date of the studies. The titles and abstracts of the identified works were reviewed. In the case of oral communications, an extension search was conducted after publication, retaining the most recent. The exclusion criteria were as follows: reviews, guidelines, or editorials; animal studies, in vitro or ex vivo; studies on SARDs in children and adolescents; and studies that only included patients with spondyloarthritis.

The results obtained were transferred to the Rayyan® software tool. After eliminating duplicate studies, two researchers (PS, NF) independently screened the titles and abstracts, eliminating those that did not meet the inclusion criteria and that presented any exclusion criteria. For studies that raised concerns, the full text was extracted to further assess their inclusion. Studies with disagreement regarding their selection were reviewed jointly with the assistance of a third researcher (LC).

Following this, the full texts of all selected studies were compiled and manually assessed by one of the researchers (NF) to confirm their final inclusion or exclusion, and any concerns were confirmed with the other researchers. The reasons for exclusion were noted at this stage.

Data extraction was performed by one reviewer (NF) using the ELICIT® tool, a research assistant powered by artificial intelligence.

The data collected in Excel® tables included 1) publication information: title; first author; year of publication; format; language; citation, and DOI (digital object identifier); 2) general characteristics of the study: objective and design; type of autoimmune disease studied; classification criteria and IS treatment administered to the patient; type of biological sample collected for the study; patient selection criteria; study duration; sample size; number of patients; mean age; gender, and demographic data; 3) study methods and results: sample analysis and study techniques and results; association of outcome measures with microbiota study variables (richness, structure, and composition) of a niche at different times and in different individuals in relation to IS treatment; association measures used and their corresponding p-value, together with other relevant findings from the studies documented in free text format. Microbiota-related prognostic factors were categorised into alpha-diversity; beta-diversity; abundance, and taxonomic composition parameters12 (Table 1).

Regarding the study characteristics and results for which information was not available, after checking Appendix A for additional data, they were left blank, except for the geographic location, which was assumed to be the first author's work centre.

Following this, a qualitative analysis and assessment of study heterogeneity was performed based on study design; methodology and analysis; disease and subgroups; measures and outcomes, and study quality.

The methodological quality of the included studies was assessed by one reviewer (NF), with any uncertainties resolved by another (LC) using the Newcastle-Ottawa Score (NOS). To maintain uniformity in the assessment of the studies' risk of bias, the NOS was applied to all studies, including one randomised controlled trial (RCT).

Due to the limited number of comparable studies by disease, methodological design, and outcome measures, publication bias was not assessed.

Of the 15 included studies, 13 collected faecal samples; 2 of these also included saliva and dental plaque samples. The remaining 2 studies used plasma and oral swab samples. The limited representation of samples from the oral cavity and other types of samples limits their comparative analysis. Consequently, subsequent analysis will focus on faecal samples from the 13 studies.

Most of these studies recruited patients with recent-onset and/or DMARD-naïve RA.17–21 Eight studies were observational, prospective (7) and retrospective (1), with one RCT providing more robust data. The remaining three studies were cross-sectional. All studies analysed the interaction of GM with cDMARDs, and specifically MTX in six of these studies.

Seven of the 13 studies directly exclusively analysed the interaction of IS therapy with GM. Two studies measured changes in GM based on changes in clinical activity in patients under treatment, allowing for inferences about the relationship between GM and IS therapy, and four studies employed direct and indirect outcome measures.

The most commonly used measure of association to measure alpha diversity was the Shannon index, followed by the Simpson index and the Chao index. Of the 9 studies with available alpha-diversity data, 89% presented statistically significant data, 5 directly and 3 indirectly.

Most studies use the Bray-Curtis Distance to measure beta diversity, with some studies using the UniFrac Distance and OTU (operational taxonomic unit). Of the 9 studies with available beta diversity data, 78% presented statistically significant results, 6 directly and one indirectly.

All studies analysed abundance changes using different measures, with the results being statistically significant in the 9 studies that defined it: 7 directly and one indirectly. Several studies agree on the existence of a baseline GM profile that could predict treatment response. In patients responding to treatment, Prevotella, Bacteriodetes, and Clostridiales species show a tendency to decrease in several studies. One study analysed the Firmicutes/Bacteriodetes ratio, and another compared two populations of RA patients based on the predominance of Prevotella or Bacteroides.

In the studies by Marazzato et al.17 and Úbeda et al.,18 patients who initiated treatment with MTX with or without CS reported a decrease in GM diversity and composition 3 mo after starting the IS treatment17 and small changes in GM diversity. They concluded that GM characteristics predict clinical response to MTX, including overgrowth of unclassified Coriobacteriaceae and related Coprococcus species.18

The results of Isaac et al.19 also reveal lower diversity, both alpha and beta, in patients who respond to MTX treatment.

In the 2024 study by Danckert et al.,20 stool samples were taken at three time points: baseline, 6 weeks after, and 12 weeks after the start of cDMARDs. Changes in the patient's microbiota were found, showing clinical improvement after the start of cDMARDs and a reduction in the relative abundance of microorganisms predominant in the microbiota of RA patients: Prevotella spp. at 6 weeks and Streptococcus spp. at 12 weeks. These findings suggest that treatment with DMARDs restores the immune system to a eubiotic state and, therefore, that the biodiversity characteristics of the immune system could be used to predict the likelihood of response to DMARDs.

Artacho et al.21 evaluated the clinical response to MTX at 16 weeks of treatment, demonstrating significant differences in the diversity and composition of the baseline immune system between responders and non-responders. Their findings suggest that the immune system may be involved in MTX metabolism, modulating its bioavailability and therapeutic efficacy. Based on these results, the authors proposed the development of a model based on the prediction of non-response to MTX based on GM diversity parameters.

The work published in 2023 by Koh et al.22 classified patients based on their disease activity into moderate and high according to the Disease Activity Score 28 (DAS28), and the treatment they received, cDMARDs or bDMARDs, with a DAS28 reassessment at 6 mo. Differences in GM diversity were studied based on the type of IS treatment (monotherapy or combination therapy). No differences in GM diversity were identified based on the IS administered, except for sulfasalazine (SSZ), which significantly reduces GM richness and uniformity. When evaluating the taxonomic composition based on the response to cDMARDs, responding patients showed an increase in Fusicatenibacter, Subdoligranulum, and uncultured Clostridia genus 014; and a decrease in Faecalitalea. The combination of Fusicatenibacter and Subdoligranulum distinguishes responders from non-responders to a second line of treatment.

The retrospective study by Gupta et al.23 analyses therapy with cDMARDs, bDMARDs, and SC, in which GM diversity is associated with disease activity, and microbial composition with cDMARD treatment.

The only registered RCT, by Mei et al.,14 compares the effects of treatment with MTX + TCM (Huayu-Qiangshen-Tongbi formula) versus MTX + leflunomide (LFM) in GM. Both treatments reduce various clinical indices of disease activity (number of tender joints, number of swollen joints, and morning stiffness) at 3 mo in the case of TCM and at 6 mo in the case of LFM. However, microbial diversity does not differ significantly over time, except for beta diversity, which increases in patients treated with LFM when compared at 1 mo and 6 mo after starting treatment. Regarding the relative abundance of species, there were no significant changes over time in patients treated with LFM, while there were in the TCM group. However, Prevotella spp., the predominant microorganism in RA patients, did not show significant differences between the two groups.

Zhang et al.13 compared the relative abundance of various niches of the oral microbiota and GM. Dysbiosis of both microbiotas was observed in RA patients when compared to healthy controls, but it partially recovered upon initiation of IS treatment. Furthermore, there were concordant changes associated with RA between the two microbiotas, suggesting an overlap in the abundance and function of species in different body locations. Treatment with TCM, specifically glycosides of the Tripterygium wilfordii component, reduced the number of MLGs (metagenomic linked groups) enriched in the gut, H. filiformis and Bacteroides spp., to a greater extent than treatment with MTX or MTX + TCM.

Regarding the studies conducted in patients with SLE, two analysed faecal samples and one analysed mouthwash. Two of the studies had a cross-sectional design, and the third was prospective. The GM studies determined alpha and beta diversity, as well as relative abundance, using the same measures for alpha diversity (Shannon index) and beta diversity (unweighted UniFrac distance and its respective PCoA expression measure).24,25 The results of Guo et al.24 showed greater alpha diversity in patients without corticosteroid treatment, while Li et al.25 found no relationship between IS treatment and microbial diversity, partly explained by the small number of patients recruited without IS treatment.

The work of Guo et al.24 highlights the variation in the Bacteroidetes/Firmicutes ratio depending on treatment with/without CS, with the Firmicutes population being restored when CS was administered. These results suggest that the Firmicutes/Bacteroidetes ratio could be used as a diagnostic and treatment response biomarker. In the taxonomic composition analysis by Gou et al., some differentially abundant taxa were recognised in each of the groups. In SLE patients treated with CS, there was an increase in Bifidobacterium and Streptococcus, and a decrease in Enterobacteriales; and in the SLE group not treated with CS, there was an increase in Bacteroidetes and Bilophila.

In the study by De Araújo et al.,16 mouthwash samples were processed using plate culture, which may have led to an underestimation of microbiota diversity and abundance, as well as limiting comparisons with other studies. Their results did not reveal significant differences based on IS treatment.

The cross-sectional observational study by Tan et al.26 in patients with SSc did not generate significant results, but a tendency toward decreased alpha diversity was inferred in patients receiving IS treatment, moving away from the profile of untreated patients and closer to that of healthy controls.

Wang et al.27 recruited patients with pSS and collected stool, saliva, and vaginal swab samples. An increase in microbiota richness was recorded in all three niches at 3 and 6 mo after starting treatment, but only at the intestinal level was this increase sustained at 6 and 12 mo. The overall microbiota composition at all three sites showed significant changes since the start of treatment, which remained unchanged in the GM and oral areas, but was still different from that of healthy controls. Certain phyla showed significant changes in relative abundance; Fusobacteria stands out, initially increasing and then decreasing.