The protocol was registered in the international prospective registry of systematic reviews (PROSPERO). Registration number: CRD 42019137746.

Narrative literature reviews, SRs of non-RCTs, and meta-analyses without SR were excluded.

A literature search was carried out in the following databases: MEDLINE (via PubMed), Embase (via Elsevier), Cochrane Central, Virtual Health Library, and Dialnet. A search was also carried out in the gray literature database Open Gray and in Google Scholar. We searched references of studies read in full and contacted experts and study authors to identify additional SRs.

[Osteoarthritis (MeSH Terms)] OR [Osteoarthritides (Text Word)] OR [Osteoarthrosis (Text Word)] OR [Osteoarthritis (Text Word)] OR; [Degenerative Arthritides; (Text Word)] OR [Osteoarthroses (Text Word)] OR; [Degenerative Arthritis; (Text Word)] OR [Arthroses (Text Word)] OR [Arthrosis (Text Word)] OR; [Osteoarthrosis deformans (Text Word)] AND [OZONE (Text Word)] OR; ozone injections; OR; ozone injection; OR; ozone therapy; OR [OZONE (MeSH Terms)].

The search strategies for the other bases are found in (Supplementary Table S1). To identify any non-published or updated review, we contacted the authors of recent SRs and one expert in OT in Brazil.

The relevant SRs were selected in two stages. In the first one, two review authors (VL/CAFA) independently screened the titles and abstracts identified during the search phase to identify the eligibility criteria. In the second stage, the same authors independently read the selected studies in full. A consensus meeting resolved any discrepancies. If the two authors did not reach a consensus, a third review author would act as the arbiter (NCPR).

A data extraction form was previously prepared. After identifying the eligible SRs, the same two reviewers (VL/CAFA) independently extracted the following data:

General characteristics: language, country of origin, year of publication, funding, and setting.

The same two authors independently evaluated the methodological quality of the included SRs using the instrument AMSTAR2 (Shea et al., 2017). The overall rating is based on the weakness of the following critical domains: Item 2) protocol registered before the commencement of the review; Item 4) adequacy of the literature search; Item 7) justification for excluding individual studies; Item 9) risk of bias from individual studies being included in the review; Item 11) appropriateness of meta-analytical methods; Item 13) consideration of the risk of bias when interpreting the results of the review; and Item 15) assessment of publication bias.

The overall confidence in the results of the UR was rated as follows:

High—0 or one non-critical weakness (NCW)—SR provides a precise and clear description of the results of the studies that deal with the question of interest.

Moderate—more than one NCW—the investigation has more than one weakness but no critical flaws. It may provide a precise and clear description of the results of the studies included in the review. Many NCW may reduce the confidence in SR. It may be suitable to move the classification down from moderate to low confidence.

Low—one critical flaw with or without NCW—SR has a critical flaw and may not provide a precise and clear description of the results of the studies included in the review.

Critically low—more than one critical flaw with or without NCW—the investigation has more than one critical flaw and should not be relied on to give a precise and clear description of the available studies (Shea et al., 2017).

There was a high degree of clinical heterogeneity (e.g., differences in the methods of application of OT related to concentration, volume of ozone, comparators, and the time of follow-up), which made quantitative synthesis impossible. We decided to report the data in a narrative synthesis, according to the following items: 1) the overall confidence in the results of SRs based on AMSTAR2; 2) the characteristics of included studies; 3) the many different primary studies included in the selected SRs; 4) the efficacy of OT; and 5) the main conclusions of the diverse SRs.

An overlap in reviews results from the use of multiple identical primary studies in similar reviews and indicates the degree to which reviews address the same literature of primary research. To measure the level of publication overlap, the corrected covered area (CCA) has been introduced as a quantitative metric. In this methodological approach, researchers initiate the process by constructing a citation matrix that organizes primary publications in rows, and the different systematic reviews were included in the umbrella review in columns. Subsequently, determining the frequency with which a particular study is cited across systematic reviews becomes a straightforward calculation (Pieper et al., 2014).

The citation matrix was constructed in order to calculate the “corrected covered area” as follows: CA = N rc , CCA = N − r / rc − r , CA = covered area ;  CCA = Corrected covered area . N = number of included publications, including double counting. This is the sum of ticked boxes in the citation matrix.

r = number of rows (number of index publications).

c = number of columns (number of reviews).

The degree of overlap in studies was rated according to Pieper et al. (2014): CCA = 0–5: slight; 6–10: moderate; 11–15: high; and >15: very high.

In the eight SRs evaluated in this umbrella review, the quality assessment of included studies was analyzed by the authors of each SR. RCTs with a lower risk of bias could result in higher confidence in SR’s conclusions.

The literature search identified 253 records. After removing 115 duplicates, titles and abstracts of 138 records were independently screened and 127 were excluded. Eleven studies were read in full, and three were excluded (Anzolin and Bertol, 2018; Arias-Vázquez et al., 2019a; Noori-Zadeh et al., 2019) (Table 1). The two reviewers reached the consensus of all studies. One contact with two different authors was necessary (Li et al., 2018; Arias-Vázquez et al., 2019a). Eight studies were included in this research. Figure 1 shows the process according to the PRISMA flow diagram.

All SRs were published between 2015 and 2020. They analyzed a total of 15 RCTs that enrolled 3,685 patients. VAS and WOMAC scales were the preferred tools for the assessment of pain and/or function in all studies. The majority of SRs utilized the Kellgren–Lawrence grading system (grades 0–IV) for OA in their assessment (Raeissadat et al., 2018b; Costa et al., 2018; Li et al., 2018; Arias-Vázquez et al., 2019a; Oliviero et al., 2019; Sconza et al., 2020). As per this grading system, a significant portion of the studies focused on patients in the early or mid-stages of knee OA. The most frequent comparator was hyaluronic acid. The follow-up time for the outcome pain ranged from less than 1 month to 12 months. Most SRs included RCTs that applied the intervention to patients with mild to moderate stages of KOA and did not mention the settings of the studies. In relation to the strategy for data synthesis, more than half of the reviews performed quantitative synthesis. Table 1 presents the characteristics of included and excluded SRs. The results of quantitative synthesis and the safety of interventions are described in Supplementary Table S2.

No SR was excluded based on the methodological quality criteria. According to AMSTAR2, the overall confidence in the results of all eight studies was critically low. Only two SR established the review methods prior to conducting the review and registering the studies. Regarding the literature search, no SR performed all the steps to the comprehensive search. Furthermore, no SR provided a list of excluded studies. The critical appraisal results for each of the eight systematic reviews are summarized in Table 2.

Six countries accounted for all the 15 RCTs: India, China, Iran, Italy, Turkey, and Brazil (Table 3). The citation matrix presents the level of overlapping in the reviews. Considering that N = 47 (number of included publications), r = 15 (number of index RCTs), and c = 8 (number of included SRs), the corrected covered area was 30.05, demonstrating a very high degree of overlap in this UR. Table 3 presents the citation matrix for SRs.

Most SRs assessed the risk of bias of RCTs using the Cochrane risk-of-bias (RoB) tool for RCTs (Higgins et al., 2011). Only one SR did not present the assessment of the quality of the selected RCTs (Oliviero et al., 2019). The quality assessments are presented in Supplementary Table S3.

Considering all SRs included in this UR, we selected the more and the less rigorous evaluation of each study classified by RoB. At best, three RCTs showed a low risk of bias (Lopes de Jesus et al., 2017; Raeissadat et al., 2018a; Babaei-Ghazani et al., 2018). All the others presented a high risk of bias.

The frequency of interventions most commonly reported in RCTs was weekly injections, with three to eight sessions in total. The duration of treatments varied from 1 to 12 weeks. The volume of IA ozone and the concentration per dose varied from 5 to 20 mL and 15 to 40 mcg/mL, respectively. Periarticular injections were administered with 5–10 mcg/mL per anatomical point around the joint (Table 2).

Three SRs reported statistically significant efficacy in groups treated with ozone, when compared to placebo for the outcome “pain reduction” (Raeissadat et al., 2018b; Costa et al., 2018; Arias-Vázquez et al., 2019a), but the intervention was not superior to other treatments in three reviews (Zhu et al., 2015; Li et al., 2018; Javadi Hedayatabad et al., 2020). Regarding function, only one investigation reported the efficacy of OT when compared to placebo (Arias-Vázquez et al., 2019a). One SR recommended OT as an efficient non-surgical treatment (Raeissadat et al., 2018b). All but one study examined safety, and OT was considered a safe procedure with a low incidence of mild adverse events reported. Table 4 presents the main conclusions and observations about the included SRs.

Umbrella reviews are employed when there are many SRs on the same subject, aiming to summarize their results and support clinical and health planning decisions (Pollock et al., 2023). This UR synthesized the available evidence on the efficacy of OT in relieving KOA pain and function improvement. It was based on a thorough literature search and assessment of study quality. All eight SRs presented critically low methodological quality according to AMSTAR2 (Shea et al., 2017). As most SRs included in this UR did not present a protocol prior to conducting the review, we highlight the need for systematic reviews’ authors to pay more attention to this process.

None of the SRs met all the requirements of a comprehensive bibliographic search. Considering that OT is practiced in more than 40 countries worldwide, if the researcher does not search studies in different languages and in the gray literature, the SR may omit important research in that area (Quintero and Schwartz, 2012).

No SR presented a list of potentially relevant RCTs with justification for the exclusion of each one. Reasons to exclude a study are related to inappropriate population, intervention, or control. Sufficient knowledge on the study characteristics can help readers decide whether the study should be selected as unjustified exclusion may bias the review findings (Shea et al., 2017).

Different problems were identified in SRs that performed quantitative synthesis, including the absence of analysis of the investigation of risk of bias, (Raeissadat et al., 2018b; Li et al., 2018; Oliviero et al., 2019; Javadi Hedayatabad et al., 2020), heterogeneity in RCTs, (Oliviero et al., 2019), and the use of the same substance in both arms of the RCTs (Arias-Vázquez et al., 2019a). In addition, the level of overlapping in studies across SRs was very high, meaning that many RCTs appeared several times, leading to unintentionally stronger weighting in any meta-analyses (Pieper et al., 2014).

Based on the RoB assessment, the majority of RCTs examined in the SRs utilized in this UR displayed a substantial risk of bias. Among those that did meet the predefined quality criteria, their findings can serve as a foundation for informing and shaping the design of upcoming studies (Lopes de Jesus et al., 2017; Raeissadat et al., 2018a; Babaei-Ghazani et al., 2018).

Despite the previously mentioned aspects that limit the quality of SRs, no review was excluded based on the criteria of methodological quality. It is important not to interpret low-quality evidence as the evidence of no effect. Low-quality evidence means unclear evidence, and findings should stimulate more research (Jamtvedt et al., 2008). Most weaknesses resulting from flaws in the methodology of SR could be avoided if the authors used a tool like AMSTAR or AMSTAR2 before starting the investigation.

Despite SRs highlighting methodological limitations in RCTs, the majority conclude that OT enhances short-term (3–6 months) pain management in mild-to-moderate KOA (Raeissadat et al., 2018b; Costa et al., 2018; Li et al., 2018; Arias-Vázquez et al., 2019a; Oliviero et al., 2019). In these instances, OT appears superior to the placebo but not surpassing the intra-articular injections of hyaluronic acid (Raeissadat et al., 2018b; Li et al., 2018; Javadi Hedayatabad et al., 2020). Results concerning the improvement of physical function are less consistent. Ozone therapy reduces the release of proteolytic enzymes and pro-inflammatory cytokines, leading to a decrease in intra-articular edema. Additionally, it stimulates the synthesis of chondrocytes and fibroblasts (Raeissadat et al., 2018a), demonstrating biological plausibility for modifying the disease course. OT has proven to be a safe procedure (Arias-Vázquez et al., 2019a), with no serious adverse events reported in any of the SRs in this study. A recent evidence map on OT has similarly revealed no serious adverse effects (Serra et al., 2023). In summary, these findings suggest that intra-articular OT could be considered a safe and effective short- to mid-term treatment for KOA patients. However, further high-quality studies are needed to enhance the scientific understanding of this promising conservative intervention.

This study has some positive aspects. To the best of our knowledge, it is the first attempt to review the existing evidence of ozone therapy in KOA using a standardized methodology for an overview. The protocol was pre-specified, and the variables of interest were clearly defined. Methodological quality was evaluated using AMSTAR2 as a guiding framework, which is known for its reliability, construct validity, and feasibility. Furthermore, the assessment of the amount of overlapping in SRs contributed to the improvement of the methodological quality of this UR.

This study has a number of limitations. Our focus was restricted to systematic reviews, and we did not reference primary research articles. Consequently, our synthesis was constrained to the authors’ interpretation of the primary data encompassed in the review. Like any umbrella review, definitive conclusions regarding the sources of heterogeneity and other potential biases cannot be firmly established. Lastly, while the majority of the included reviews were recently published, it is crucial to acknowledge that reviews may become outdated swiftly due to the rapid emergence of new research. Therefore, insights should be interpreted in the context of this dynamic research landscape.