Thank you for the opportunity to review your systematic review investigating the effectiveness of perturbation-based training (PBT) on balance and balance confidence in patients with stroke. While you have concluded that ‘PBT was effective in improving the balance on patients with stroke’, I think that this conclusion is overstating the results and the GRADE of the evidence needs to be considered when making your conclusions on the effect of PBT. At best, the evidence for this finding is low quality which suggests that PBT MAY be effective in improving balance in people with stroke; further evidence is required to confirm the effectiveness of PBT. The conclusion does acknowledge the need for further high-quality studies, but does not clearly indicate that the effect is uncertain due to the quality of evidence. You report a range of quality of evidence (from very low to moderate) of the studies included in the review which has the potential to mislead the reader as to what the actual quality of evidence is for each finding. You need to state the quality of evidence for each particular finding, rather than the range for studies included in the review. Currently, GRADE is not provided for the two main meta-analyses (i.e., those reported in the abstract). This is my major concern with the review. However, I also have several other concerns that I believe need to be addressed.

Page 6 – Search strategy: Did you use subject headings in your search strategy? Particularly for databases such as MEDLINE?
Also, did you pilot the search strategy prior to conducting the review (i.e., preliminary searches)? Or consult a research librarian in the development of the search strategy?
I think that the search strategy could have been more comprehensive and this is reflected by the fact that almost 50% of the articles (n=3 out of 7) included in your review were identified by ‘other methods’. This is a possible limitation of your review.

Page 8 – Evaluation of methodological quality: Were methodological ratings (using the PEDro scale) determined independently by the two reviewers? This should be clearly stated, as per the PRISMA reporting guidelines.

Page 9 – Quality of evidence: The specific GRADE criteria that were applied should be explained here in detail. There are no uniform rules for determining what the threshold is for ‘marking down’ the quality of a body of evidence. For example, what was the threshold 95%CI for marking down quality of evidence for imprecision?
Note that criteria should be provided for all five dimensions (i.e., study limitations, imprecision, inconsistency, indirectness, publication bias).

Page 10 – Study characteristics: You have reported the number of studies that reported on falls, yet this was not outlined as an outcome of interest in the methods. Ideally, only data on the outcomes of interest should be reported on in the results. It would be reasonable to include falls as an outcome, but you need to be clear whether the decision to include falls as an outcome was part of your protocol or post-hoc.
Similarly, reporting on other measures performed that are not intended outcome measures for your review is not required (e.g. quality of life, ADLs, patient satisfaction, physical function). It ultimately distracts the reader from the purpose of your review.

Page 12 – Effect of PTB on balance: What is the GRADE for the ‘overall effect’ of PTB on balance meta-analysis. This is the main outcome of your review, along with the ‘overall effect’ of PTB on balance confidence meta-analysis. However, you do not report the GRADE for these meata-analyses, only the sub-group analyses (i.e., PTB only and PTB + EX). You need to report the GRADE for all meta-analyses, not just the sub-group analyses. Especially considering you are reporting these main analyses (and not the sub-group analyses) in the abstract. These should be reported in Table 4 alongside with the sub-group GRADE.

Page 13 - Effect of PBT on balance: The sub-group differences test adds little information to the interpretation of your results. The number of studies is too low in each sub-group, as is the number of participants, to justify inferential statistics. I suggest removing this test and presenting only SMD/MD and 95% CI for each sub-group; this is sufficient information for the reader to determine whether there are any true differences between sub-groups without the unnecessary presentation of a p-value.

Page 14 – Effect of PBT on balance confidence: my previous two comments apply to these results as well. In both sections, too much focus is given to the sub-group analyses, particularly when your main conclusions are drawn from the main analyses. There seems to be unjustified prioritisation given to the sub-group analyses.
Please not that I am not critical of the sub-group analyses – they are much appreciated and add to the review. However, the main analyses should be clearly reported (i.e., the focus of the results) before moving onto the sub-group analyses. This includes reporting the GRADE/quality of evidence, which should be applied to all analyses.

Page 13/14 – You have reported the stage of stroke as ‘sub-acute’ and ‘chronic’, however, on page 9 you have not provided a definition for each ‘stage of stroke’. Please provide a clear definition for what was classified as ‘sub-acute’ and what was classified as ‘chronic’ stroke.

Page 14 – Discussion: As stated above, given the low quality of evidence, I believe that you are overstating the effect of PBT on balance in people with stroke. At best, it MAY improve balance and further high-quality studies are required to prove this result. Please also state the quality of evidence for each finding rather than a range of quality of evidence – for example, ‘We found low quality evidence that PBT may improve balance in people with stroke’. The purpose of GRADE is to indicate the strength of evidence for each finding of a systematic review (i.e., the quality of a body of evidence that has been synthesised to produce a single result).
Interpretation of results in the discussion and conclusion needs to reflect the quality of evidence supporting each result.

Discussion: Throughout the discussion there is no mention of the clinical significance of the results. While you have provided an interpretation of the SMD in the methods, there is no reference to the size of effect anywhere in the discussion. Similarly, is the point estimate of the MD for balance confidence clinically significant? Does the size of the point estimate justify further research? Could it be clinically meaningful, despite not reaching statistical significance? Interpretation of clinical meaningfulness of the results is lacking.

Page 15 – Discussion: It is not clear how your results shed any light on the discussion points raised in the ABC scale paragraph. Was the MD of your meta-analysis (i.e., MD 1.96 [95% CI -4.25 to 8.16]; P =0.54) clinically significant? Would this change in confidence be significant enough to overcome the psychological barriers that limit everyday activities and increase risk of falls?
Maybe balance retraining alone is not enough to improve confidence? Maybe psychological interventions (e.g., motivational interviewing or cognitive behavioural therapy) would be more effective? Or maybe they could be used in conjunction with balance retraining?
A more nuanced discussion is warranted, with consideration of whether the effect size is clinically significant.

Page 17 – Discussion: Please consider discussion of the limitations of the design of your review. Particularly the search strategy and inclusion of outcomes such as falls that were not prospectively considered as outcomes of interest. Also, it should be made clear that there was a deviation in protocol, with the ‘additional outcome’ changed from functional mobility to balance confidence.

Thank you for addressing my comments. Please find some minor suggestions to improve the quality of your reporting:

Abstract: Page 2, line 55: '...which suggests further high-quality trials', I suggest amending to 'which suggests further high-quality trials are required.'

Conclusion: Page 18 line 37: I suggest making the same amendment to the conclusion in the main text.

Discussion: Page 18, line 3: You state 'Moreover, presence of
inconsistency in the terminology regarding PBT in literature causing difficulties in detecting all eligible studies by simple searching. 'I am not sure what you mean by 'simple searching'. Are you referring to searching online databases? Please clarify.