This study presents the main report of methodological and reporting surveillance on SRMAs of physical activity intervention/exposure assessed by the SEES Initiative in the year of 2019. Across the seven domains operationalized to aggregate the items/recommended practices by the PRISMA Statement, ROBIS, and AMSTAR 2 tools, some of the most adopted practices were in the completeness domain: identification as a systematic review, meta-analysis, or both in the study title (97.1%) and more 2 items with 95.1%; outcome domain: description of the main outcome of interest in the abstract (95.1%); and methodological rigor domain: description of RoB assessment (95.1%). In opposite, the existence of a publicly available methodological protocol (4.9%) (transparency domain), discussion of the results in light of RoB in individual studies (32.0%) (critical appraisal domain), and statement on data sharing (35.9%) (transparency domain) were among the practices least adhered to, therefore indicating possible emphases that can be addressed through educational resources and editorial policies.
Practices most and least present in assessed articles
In the transparency domain, modest existence of reviews registered in public databases and only a minimal fraction (5/103) presenting a detailed methodological protocol in addition to the registration may reduce the potential for methodological reproducibility [
30]. Although most SRMAs retrieve disseminated and aggregated data, data sharing statements were still less than modest. There are similar results indicating low adherence to practices of registration [
1,
24] or data sharing [
31] in other disciplines, denoting that transparent practices need more implementation in varied research fields.
We observed that most studies adhered to items under the completeness, participants, and interventions/exposures domains. Taken together, these three domains included 14 items, which allow stakeholders to understand the scope of the study, the research question, its applicability, and, finally, the replication of the study if desired [
21]. However, our results indicate that the information on the eligibility criteria for the selection of individual studies was not adequately reported in a relevant number of SRMAs, either in abstracts (25.2%) or in methods (31.1%). Furthermore, 22.4% of SRMAs did not provide a clear description of participants or main conditions under study in the abstract section. We underscore the importance of such information since it is key for stakeholders to decide whether to access (possibly by payment) the article.
In domains of methodological rigor, outcome reporting, and critical appraisal, we combined items (practices) that could more directly affect the interpretation of evidence generated by SRMAs. Surprisingly, several reviews did lack duplicate processes in data extraction and risk of bias assessments. Although similar results were observed in meta-research studies in health sciences [
1,
24,
32], it is widely recommended that these procedures be performed by two independent researchers or primarily conducted by one researcher and fully verified by another one [
22,
29]. Additionally, we found that nearly two-thirds of SRMAs (60.3%) considered only the English language to study eligibility. Such restriction is commonly applied in literature searches [
33], but can introduce a systematic error, a language bias, possibly modifying meta-analytic estimates [
33,
34]. In the outcome domain, items such as individual results for each analyzed study and meta-analytic summary estimates still need substantial reporting improvement to facilitate the understanding and replicability of meta-analyses. These items were considered as properly reported when all recommended elements were addressed; therefore, our assessment may have been very strict. However, to note, previous studies have also reported similar findings in the biomedical sciences [
31,
35].
In the critical appraisal domain, more than 80% of our sample presented the risk of bias assessment; however, only a third (33/103) discussed their review findings in the context of potential biases identified through the assessment. Part of our results are in line with data by Buttner et al. [
36], who focused only on the risk of bias assessment, and also described a high percentage of these assessments in systematic reviews and SRMAs of exercise interventions/exposures published in a single journal. Moreover, studies in biomedical literature have also reported a reduced use of this type of assessment in the discussion of SRMAs [
9,
32], therefore, in agreement with our findings. In contrast, Buttner et al. [
36] found more favorable results, indicating that risk of bias assessments were incorporated in the interpretation/discussion sections of 86% of their sample. Although both studies, Buttner et al. [
36] and ours, have distinct samples and criteria to assess this practice, we reason that, especially when the risk of bias is high in individual studies, such results should be thoroughly addressed in
Discussion sections. Finally, we assessed whether SRMAs reported methodological changes during the course of the review, which were not planned as described in the registry or methodological protocol. These changes in design, conduct, or analysis can introduce bias in the review findings and therefore should be reported to the readers’ knowledge [
22,
29]. Unfortunately, our results show that 17.5% (18/103) of SRMAs made unplanned changes and did not describe them in the final publication. Additionally, in 41.7% (43/103) of SRMAs, it was not possible to carry out this verification due to the lack of a public registration record, which again reinforces the need for transparent practices at the pre-study stage.
The results observed show the importance of this type of initiative, with a characteristic of “post-publication criticism,” for the realization of a call for attention and action for the different stakeholders. It is noteworthy that the SEES Initiative could have a significant impact if authors and editors had actively engaged in the cordial dialogue about the results of their manuscript assessments, proposed by this initiative in the emails sent to these key actors. However, the reluctance to respond to criticism does not seem to be unique to the field of exercise science [
37].
Prospective registration and open access publication
The results of our exploratory analysis indicate that prospective registration is associated with better adherence to some relevant items/recommended practices in the domains of transparency, completeness, methodological rigor, and critical appraisal. To our knowledge, there are no studies on the influence of the prospective registration on the methodological and reporting standards of SRMAs in exercise sciences, but a recent study investigated this topic in systematic reviews of healthcare interventions [
38]. In a similar way to our study, they evaluated a sample (
n = 150) of systematic reviews published in 1 year (i.e., 2015) through PRISMA and AMSTAR tool. However, unlike us, they used total scores for these analyses. As main results, they observed that the total scores on PRISMA and AMSTAR were higher for those systematic reviews registered a priori, being significant only for the total scores on AMSTAR. Therefore, they concluded that prospective registration may at least indirectly improve the overall methodological quality of systematic reviews. Regarding open access, Pastorino et al. [
39], evaluating only 47 systematic reviews with and without meta-analysis in the field of oncology (open access,
n = 15 and no open access,
n = 32), published in 2013, observed that the overall methodological and reporting standards were comparable between studies with and without open access. Similar to our study, they found a significant difference in only one item (description of the methods used to assess the risk of bias in individual studies). However, one must be aware of the large amount of statistical comparisons, which increases the probability that a difference is significant by chance. In addition, these stratified analyses were not planned. Therefore, studies specifically designed to investigate these issues (registration and open access) must be conducted to confirm these findings.
Strengths and limitations
Several strategies exist to improve the methodological and reporting quality of SRMAs in the biomedical literature [
18,
20,
24,
32,
40‐
42]. Our study highlights that solely relying on the availability of these documents/tools could be insufficient for adoption. In exercise sciences, there is a need for most journals to provide endorsement of the guidelines, which may trigger quality improvements [
37,
38]. The use of guidelines in the peer review process has also been proposed [
24,
32]. We point out that the main strength of this study is the “translational approach,” since we used established guidelines in a lively project, giving feedback to editors and authors (despite the low number of replies). In methodological grounds, we did not homogenize our assessments through a total score, allowing readers a more detailed view of adherence to items and domains. Indeed, generating such scores is not recommended, since it assumes that all items have the same weight [
22]. Finally, we included some ROBIS items in our assessment form, therefore addressing more directly the methodological quality assessment of the SRMAs.
Some limitations should be considered. First, our sample of SRMAs represents only the year 2019 and only a few pre-selected scientific journals. Therefore, the results observed in our assessments are not promptly generalizable to other SRMA publications in exercise sciences. However, it seems unlikely that the quality of reporting of previous SRMAs is superior to the current ones, given that important documents providing guidance and recommendations for planning, conducting, and reporting SRMAs were published a few years ago (i.e., PRISMA in 2009, AMSTAR 2 in 2017, ROBIS in 2016); our results are in line with several studies of this nature in other health disciplines. In addition, the selected journals had features such as connection to a professional or scientific societies, considerable public reach, and impact factor. Since these features increase visibility, we believe the quality of studies is a common priority among the selected journals. Finally, unlike other studies of this kind, we proposed to assess SRMAs prospectively and systematically (i.e., monthly basis), as well as to provide direct feedback to authors and journal editors, encouraging them to question the assessments when they do not agree with these. Thus, focusing on some journals became essential to increase our chances of being able to contemplate this proposal. Second, a possible limitation is the fact that our assessment form has not been formally validated. Although the form was elaborated on the basis of a reporting guideline for systematic reviews (i.e., PRISMA) and well-planned methodological quality assessment tools (i.e., AMSTAR and ROBIS), testing the properties of our form would be important.