Pain Trajectories in Knee Osteoarthritis—A Systematic Review and Best Evidence Synthesis on Pain Predictors

Round 1

Reviewer 1 Report

Review of

Pain trajectories in knee osteoarthritis – A systematic review and best evidence synthesis on pain predictors

JCM-868717

Manuscript summary

This study looked at the pain progression in patients with knee osteoarthritis by reviewing articles which provided information on this topic. The authors examined 454 articles and resulted in only 7 meeting their criteria for inclusion. These represented 7155 patients who reported through their last follow-up. Six of the studies had follow-ups of at least 5 years. For mid-term OA patients, 84% showed a steady pain trajectory, with 7% reduction and 9% increase in pain. Low education, comorbidities, and depression are patient-related predictors of worsening knee OA pain.

Overall comments

The manuscript is well written and does a good job of examining the literature on this topic. As the authors state, there are not a lot of articles that look at this issue and there are some limitations in how these studies were done (limited assessments, not diverse populations). Consequently, the results of the analysis are minimal in that most patients did not see a change in pain, although this in itself is interesting and supports the “wait and see” approach, as the authors point out. The predictors with moderate levels of evidence and the conflicting findings also suggest that more studies need to be done in this area.

Specific comments

Line 156                59,5 % should have a decimal point

Author Response

Reviewer 1

Manuscript summary

This study looked at the pain progression in patients with knee osteoarthritis by reviewing articles which provided information on this topic. The authors examined 454 articles and resulted in only 7 meeting their criteria for inclusion. These represented 7155 patients who reported through their last follow-up. Six of the studies had follow-ups of at least 5 years. For mid-term OA patients, 84% showed a steady pain trajectory, with 7% reduction and 9% increase in pain. Low education, comorbidities, and depression are patient-related predictors of worsening knee OA pain.

Overall comments

The manuscript is well written and does a good job of examining the literature on this topic. As the authors state, there are not a lot of articles that look at this issue and there are some limitations in how these studies were done (limited assessments, not diverse populations). Consequently, the results of the analysis are minimal in that most patients did not see a change in pain, although this in itself is interesting and supports the “wait and see” approach, as the authors point out. The predictors with moderate levels of evidence and the conflicting findings also suggest that more studies need to be done in this area.

• We are happy to see the reviewer understood the purpose of our study, which is to investigate the role of pain trajectories in knee OA patients, in order to stimulate research in this interesting field, being the pain the main symptom of knee OA and main target of treatment for this condition. Moreover, we further implemented our manuscript to encourage its publication.

 Specific comments

Line 156                59,5 % should have a decimal point

• Done (line 168).

Reviewer 2 Report

Author Response

Reviewer 2

Thank you for submitting this well-designed article. As a systematic review, it shows the current status regarding pain trajectories in knee osteoarthritis.

The route to the seven articles used is not entirely clear. It is particularly interesting why 249 articles were sorted out on the way. Perhaps you give the Reader different categories, why the articles could not be used.

Otherwise the paper is well structured, and the limitations are clearly shown.

• Dear reviewer, thank you for your positive revision. We agree with the considerations. Thus, we have implemented the section on study selection (lines 120-130).

Reviewer 3 Report

Summary of Manuscript:

            The authors performed a literature review of pain trajectory in individuals who have knee OA yet have not undergone surgical intervention. They found that most patients pain levels were stable over time, with few patients who had increasing or decreasing pain levels over time. Further, they found only two significant contributors to pain evolution which were low education levels and presence of comorbidities.

Comments:

Abstract

1. Lines 19-20: Why were KL grades 0 and 4 not included? These are your endpoint conditions and it seems like they would greatly impact how we understand pain evolution. For example, is pain evolution an exponential growth curve after KL 3? And are there other factors that are connected to this growth curve? These are things you need to evaluate to get the entire picture of pain evolution in knee OA

Introduction

1. Lines 35-36: You are correct in saying that pain evolution is often misunderstood or poorly understood at best. However, this is only one symptom. It might be a major driver for a significant portion of folks, but the development of pain in knee OA is highly multifactorial. For example, how a patient moves can greatly impact both the function of the knee and the pain a patient experiences. Yet, your paper has failed to even mention these as possible confounding factors. We also know that treating pain in knee OA has no impact on the actual state of the disease, simply what is tolerable within the current disease state. Accordingly, I remain unconvinced that this ability to categorize patients based on purely pain evolution has much clinical value.
2. Line 43: Given pain is such a multifactorial experience, are you really convinced that you can distill different OA categories by solely pain? This seems unlikely and clinically of low value.
3. Line 44: Yes, TKA is an invasive surgery, however it is also extremely effective. TKA and many other arthroplasty procedures are some of the highest value (as measured by QALY) interventions available in all of medicine. Given this efficacy, the work herein almost seems more like an academic exercise than something we can use clinically.
4. Is pain the best metric to utilize? It is so multi factorial yet you are treating it monolithically. Can you incorporate other metrics into your clinical decision-making tool? For example, other PROMs, kinematic info, etc.? I am unconvinced that pain can be the end all be all your paper appears to purport it to be without at least giving credence to other important variables.

Methods

1. Line 67: Over what time period were these studies published? It’s unclear what the total population of patients might have been during this time period (i.e. what is the denominator of your fraction) versus the numerator (i.e. the patients used in this present manuscript). This will really dictate the validity of the sample you’ve extracted. In other words, is the sample size of patients large enough to really represent reality?
2. Line 99-106: This method seems a bit weak given the variety of statistical methods at your disposal. You could use multiple regressions, or change score analyses, or any number of evaluations. This choice appears very subjective with an already subjective metric as your metric of choice.
3. Have you thought of looking at a patient’s change score for this? Presumably you have the data from these studies at multiple time points. This would allow you to see if the change in pain is more important than the overall pain level. This may be important because we know individual pain tolerance is variable as well. So, someone who has a very high pain tolerance (typically older women in our clinic), may start at higher pain, but not be impacted by this at all. Yet if they change to a markedly higher pain level they may begin to complain about that symptom.

Results

1. Line 115: This seems at first glance like a really small number. Can you expand this to incorporate other articles that include pain and still get representative information?
2. Line 125: 7747 is a nice number, but we have no idea how many patients could have been seen in that same time period for knee OA. This number could in fact be tiny. For example, each year in the US we perform somewhere near 700,000 TKA. If your study uses articles that transverse a decade we have to assume then that at least 7,000,000 people have undergone TKA and had knee OA which could have been included in this study. This would mean your sample size is just 0.1% of the possible population. Is this representative? Can you support this statistically? If using this small of a sample size, what is your risk of T1 and T2 error? These are things that really need evaluated before this is publishable.
3. Line 140: It seems like if pain is constant this might be a good thing. However, it also seems to me like this also makes using it as a clinical differentiator difficult. You also have failed to account for the idea that even if someone’s pain remains unchanged, if it remains unchanged above a specific threshold level that might be too much for certain people. How does this impact your results?
4. Line 169: These two results are interesting but need further defining. What does low education level really mean? Less than high school? No college degree? No education? And what comorbidities were present. Nearly every patient that presents in our clinic with knee OA has at least one comorbidity. Some of them are life threatening, others much more benign. It seems like the type and severity of comorbidity will be important to each patient’s pain evolution.

Discussion

1. Line 194: If a patient has a constant pain level over time, this would indicate they don’t need invasive treatments because what we are doing in terms of treatment is working. Yet, many individuals with static pain levels over a long period of time tire of that pain level and desire greater intervention. How does this information inform us in this situation?
2. Line 197: This statement is a significant reach to me. We have no idea what the OA phenotype really is, only what the pain level is doing. This disease is so multifactorial that it seems like distilling it down to pain and pain alone is unwise and clinically impossible.

Author Response

Reviewer 3

Summary of Manuscript:

            The authors performed a literature review of pain trajectory in individuals who have knee OA yet have not undergone surgical intervention. They found that most patients pain levels were stable over time, with few patients who had increasing or decreasing pain levels over time. Further, they found only two significant contributors to pain evolution which were low education levels and presence of comorbidities.

• Dear reviewer, thank you for your positive revision. We modified our manuscript following your suggestions and advices. Please see the following point by point answers to your comments and the related changes.

Comments:

 Abstract

Lines 19-20: Why were KL grades 0 and 4 not included? These are your endpoint conditions and it seems like they would greatly impact how we understand pain evolution. For example, is pain evolution an exponential growth curve after KL 3? And are there other factors that are connected to this growth curve? These are things you need to evaluate to get the entire picture of pain evolution in knee OA

• We agree with the reviewer, this is a thoughtful comment, it would be certainly extremely interesting to know something about all KL grades. However, this manuscript is a systematic review of the literature about pain trajectories in OA patients, and the studies retrieved by our search only included these kind of patients with Kellgren-Lawrence grade between 1 and 3, excluding patients with grade 0 and 4. Thus, it is not possible to overcome this limitation, as it is not a search limitation but actually represents a limitation of the available literature, but we added this aspect in the discussion as requested (lines 292-295).

Introduction

Lines 35-36: You are correct in saying that pain evolution is often misunderstood or poorly understood at best. However, this is only one symptom. It might be a major driver for a significant portion of folks, but the development of pain in knee OA is highly multifactorial. For example, how a patient moves can greatly impact both the function of the knee and the pain a patient experience. Yet, your paper has failed to even mention these as possible confounding factors. We also know that treating pain in knee OA has no impact on the actual state of the disease, simply what is tolerable within the current disease state. Accordingly, I remain unconvinced that this ability to categorize patients based on purely pain evolution has much clinical value.

• The considerations of the reviewer are certainly correct and reflect the complexity of the field: even though pain is the main symptom of knee OA and it is the main target of treatment for this condition, it is only one of the features characterizing OA and its development is highly multifactorial, with different possible confounding factors. We added this aspect in the introduction as requested (lines 33-37). On the other hand, the current study is a review specifically focusing on pain trajectories in OA, and the included studies allow us to draw some conclusions about some of such confounding factors (low education, comorbidities, and depression, which are described in the results section (lines 180-192). We tried to better take into account these aspects in the limitation section of the study (lines 281-287). Nevertheless, we believe that a better knowledge of pain characteristics and in particular of pain trajectories represents potentially an outstanding tool not only in the clinical practice, allowing physicians to better manage patients, but also in the design of future trials about OA treatments, from injective treatments to arthroplasty. These studies will have to include the activity level as well, as we now implemented in the discussion according to the reviewer indication.

Line 43: Given pain is such a multifactorial experience, are you really convinced that you can distill different OA categories by solely pain? This seems unlikely and clinically of low value.

• Reviewer criticism is legitimate: even though pain is the main symptom of knee OA and it is the main target of treatment for this condition, it is only one of the symptoms characterizing OA. On the other hand, the current study is a review specifically focusing on pain trajectories in OA, and the included studies were able to distill different OA categories by pain trajectories with a statistical methodology. The authors of the included studies, together with the authors of the current manuscript, believe that a better knowledge of pain characteristics and in particular of pain trajectories could represent an important tool both for the clinical practice, allowing physicians to better manage patients, and in the design of future trials about OA treatments, from injective treatments to arthroplasty. We added these aspects in the introduction and among study limitations as requested (lines 33-37 and 281-287).

Line 44: Yes, TKA is an invasive surgery, however it is also extremely effective. TKA and many other arthroplasty procedures are some of the highest value (as measured by QALY) interventions available in all of medicine. Given this efficacy, the work herein almost seems more like an academic exercise than something we can use clinically.

• We agree with the reviewer, TKA is without doubt an effective and outstanding procedure. Nevertheless, current guidelines reserve its indication after the failure of conservative treatments, and it is undeniable that it is not free from complications, especially in young patients. In fact, a recent database study on 54,000 knee arthroplasties showed that joint replacement in patients younger than 55 years increased the lifetime risk for revision up to 35%, bringing into question the increasing use of knee replacement in younger patients [Bayliss LE, et al. Lancet. 2017]. For this reason, the understanding of different OA categories based on pain evolution represents something more than an academic exercise, since a tailored indication for conservative treatment or TKA may be also determined from the knowledge of the disease progression pattern, and in this sense pain also represents one of the main factors leading to the TKA indication. Still, we agree that pain is only one aspect of a complex multifactorial disease like knee OA, and we acknowledged these aspects in the introduction and among study limitations as requested (lines 33-37 and 281-287).

Is pain the best metric to utilize? It is so multi factorial yet you are treating it monolithically. Can you incorporate other metrics into your clinical decision-making tool? For example, other PROMs, kinematic info, etc.? I am unconvinced that pain can be the end all be all your paper appears to purport it to be without at least giving credence to other important variables.

• Thank you for your comment, we agree that pain is only one aspect of a complex multifactorial disease like knee OA. On the other hand, the current study is a review specifically focusing on pain trajectories in OA, and it can not claim to be able to propose a clinical decision-making tool. Certainly, the best clinical decision-making tool should take into consideration several other metrics. Our aim was to shed some light about the main symptom and the main target of treatment for knee OA, and a deeper knowledge of pain and especially pain trajectories may contribute in the future to better categorize different OA profiles. Still, we modified the introduction and add these aspects among study limitations as requested (lines 33-37 and 281-287).

Methods

Line 67: Over what time period were these studies published? It’s unclear what the total population of patients might have been during this time period (i.e. what is the denominator of your fraction) versus the numerator (i.e. the patients used in this present manuscript). This will really dictate the validity of the sample you’ve extracted. In other words, is the sample size of patients large enough to really represent reality?

• The included studies were published from 2014 to 2018 (we searched the literature with no time limitation), as reported now in the results section (line 129) and in Table 2 (line 145). This study is a systematic review specifically focusing on pain trajectories in OA, thus the sample size evaluation is not applicable. Anyway, the included studies analyzed patients’ registries reporting data for over 7000 patients (a pretty large number considering other types of literature analysis in this field) and, despite the limitations of this tool, they represent real-life data that may be considered reliable in describing a specific population or pathology.

Line 99-106: This method seems a bit weak given the variety of statistical methods at your disposal. You could use multiple regressions, or change score analyses, or any number of evaluations. This choice appears very subjective with an already subjective metric as your metric of choice.

• We agree that the methods used to rate the level of evidence of the results may appear somehow subjective. This is probably due to the fact that most of the methods used to rate the level of evidence are qualitative and not quantitative. However, to increase the validity of our evaluation, two different authors rated the risk of bias of the included studies and the level of evidence of the results, following the Cochrane guidelines. Unfortunately, the methods suggested (multiple regression, change score analysis), despite useful and powerful, are used to detect correlations and the evolution of symptoms overtime, respectively, and cannot be used to evaluate the level of evidence.

Have you thought of looking at a patient’s change score for this? Presumably you have the data from these studies at multiple time points. This would allow you to see if the change in pain is more important than the overall pain level. This may be important because we know individual pain tolerance is variable as well. So, someone who has a very high pain tolerance (typically older women in our clinic), may start at higher pain, but not be impacted by this at all. Yet if they change to a markedly higher pain level they may begin to complain about that symptom.

• Looking at patient’s change score to understand how single patients’ characteristics may have impacted pain evolution would be undoubtedly interesting. However, when performing a systematic review, it is not always possible to retrieve single patient data and all the studies analyzed, each one including a large number of patients, provided only aggregated data. Being our study a review, we based our results about the predictors of pain evolution on the data reported in the single studies analyzed. The influence of pain tolerance on pain trajectories was unfortunately not analyzed in the included studies but, interestingly, patients’ age seems to be not correlated to the evolution of pain over time while the evidence about patients’ sex is conflicting. Following your suggestion, we now discussed the possible importance of pain tolerance on pain trajectories in knee OA (line 252-255) and we added the fact that we were not able to evaluate this point in the limitation section (line 281-287).

Results

Line 115: This seems at first glance like a really small number. Can you expand this to incorporate other articles that include pain and still get representative information?

• We agree, this is a relatively small number, and we believe it represent the main limitation of our review, which is anyway the first review dealing with this topic. Other studies on pain trajectories are needed to better delineate a comprehensive picture about knee OA. We added all this aspect in the discussion as requested (lines 281-287).

Line 125: 7747 is a nice number, but we have no idea how many patients could have been seen in that same time period for knee OA. This number could in fact be tiny. For example, each year in the US we perform somewhere near 700,000 TKA. If your study uses articles that transverse a decade we have to assume then that at least 7,000,000 people have undergone TKA and had knee OA which could have been included in this study. This would mean your sample size is just 0.1% of the possible population. Is this representative? Can you support this statistically? If using this small of a sample size, what is your risk of T1 and T2 error? These are things that really need evaluated before this is publishable.

• We agree with the reviewer, a higher number of OA subjects to be included in databases is desirable. We added this aspect in the limitation section as requested (lines 281-287). Nevertheless, the current number is still higher than other reviews published in the field, the current study is a review of studies specifically focusing on pain trajectories in OA, and the numerosity of our review is superior of every already published study. Moreover, this systematic review allowed to obtain preliminary information on a still early topic like pain trajectories, which could be helpful for future studies.

Line 140: It seems like if pain is constant this might be a good thing. However, it also seems to me like this also makes using it as a clinical differentiator difficult. You also have failed to account for the idea that even if someone’s pain remains unchanged, if it remains unchanged above a specific threshold level that might be too much for certain people. How does this impact your results?

• Very interesting consideration: a steady pain above a specific threshold level might not be tolerable for certain people. Unfortunately, the current review on pain trajectories cannot make hypothesis about specific thresholds, which would be extremely different among patients and among cultures. Nevertheless, it should also be considered that patients with intolerable pain are likely to undergo some kind of treatment, and the included studies did not considered patients undergoing treatments. So in this way, even though extremely important from a clinical point of view, where a steady pain may not represent a good thing, this does not impact our results. We better specified these aspects in discussion as requested (lines 295-298).

Line 169: These two results are interesting but need further defining. What does low education level really mean? Less than high school? No college degree? No education? And what comorbidities were present. Nearly every patient that presents in our clinic with knee OA has at least one comorbidity. Some of them are life threatening, others much more benign. It seems like the type and severity of comorbidity will be important to each patient’s pain evolution.

• Interesting consideration: patient-related factors identified as predictors of a severe or progressing pain trajectory with a strong level of evidence were low education level (college degree versus no college degree) and presence of comorbidities. Only one study analyzed the type of comorbidity influencing pain trajectories, while the others only evaluated the presence of comorbidity. We added now this aspect in the results section (lines 180-191).

Discussion

Line 194: If a patient has a constant pain level over time, this would indicate they don’t need invasive treatments because what we are doing in terms of treatment is working. Yet, many individuals with static pain levels over a long period of time tire of that pain level and desire greater intervention. How does this information inform us in this situation?

• We agree with the reviewer, a patient with a constant pain level over time could tire and desire greater intervention. However, it is also a useful information that pain does not increase in many patients, at least at mid-term. Nevertheless, it should also be considered that patients with intolerable pain are likely to undergo some kind of treatments, and the included studies did not considered patients undergoing treatments. Thus, even though extremely important from a clinical point of view, where a steady pain may not represent a good thing, this does not impact our results. We better specified these aspects in discussion as requested (lines 295-298).

Line 197: This statement is a significant reach to me. We have no idea what the OA phenotype really is, only what the pain level is doing. This disease is so multifactorial that it seems like distilling it down to pain and pain alone is unwise and clinically impossible.

• We agree with the reviewer, even though pain is the main symptom of knee OA and it is the main target of treatment for this condition, it is only one of the symptoms characterizing OA. On the other hand, the current study is a review specifically focusing on pain trajectories in OA, and the included studies were able to distill different OA categories by pain trajectories with a statistical methodology. The authors of the included studies, together with the authors of the current manuscript, believe that a better knowledge of pain characteristics and in particular of pain trajectories could contribute in the future to better categorize different OA profiles, representing an important tool both for the clinical practice, allowing physicians to better manage patients, and in the design of future trials about OA treatments. We modified the introduction and ed these aspects among study limitations as requested (lines 33-37 and 281-287).

Reviewer 4 Report

This interestin study concluded that low education, comorbidities, and depression are patients-related predictors of severe/worsening knee OA pain. Conversely, age, alcohol, smoking, pain coping strategies, and medications were unrelated to pain evolution. Conflicting/no evidence was found for alljoint-related factors, such as baseline radiographic severity. As a systematic review, this study needs to adress the following major issues:

1) Statistical analyses for heterogeneity are necessary to determine risk ratio estimates (I² coefficient). These analyses need to be adressed in both methods and results section.

2) Foresplot figures are necessary. Also, these analyses need to be adressed in both methods and results section.

3) The experimental section needs to be expanded and subdivided into sub-section including all subsection required by the PRISMA criteria. Please, be accurate with all sub-sections and requirements of this checklist. Please, include the checklist of PRISMA as a supplemental file for review purposes

Author Response

Reviewer 4

This interesting study concluded that low education, comorbidities, and depression are patients-related predictors of severe/worsening knee OA pain. Conversely, age, alcohol, smoking, pain coping strategies, and medications were unrelated to pain evolution. Conflicting/no evidence was found for all joint-related factors, such as baseline radiographic severity. As a systematic review, this study needs to address the following major issues:

1) Statistical analyses for heterogeneity are necessary to determine risk ratio estimates (I2 coefficient). These analyses need to be addressed in both methods and results section.

• Despite being originally planned, a meta-analysis of prevalence (with forest plots and I2 coefficients) for the different pain trajectories was not possible due to the paucity of the data retrieved from the systematic search of the literature. In fact, some of the trajectories are represented in only one or two of the included courts, thus limiting the statistical strength of a meta-analysis with such data. Moreover, the statistical heterogeneity of the analyzed data further limited the possibility to perform a statistically strong meta-analysis. Thus, we preferred to perform only a systematic review and express data as simple proportions with range in order to avoid presenting data as more powerful than they really are (less experienced readers perceive meta-analyses always as supported by a high level of evidence). We now addressed this point in the limitation section to further underline the need of more high-level studies on this interesting topic (lines 291-292).

2) Foresplot figures are necessary. Also, these analyses need to be addressed in both methods and results section.

• Since the present study is not a meta-analysis (for the reasons explained at point 1), we cannot present forest plots.

 3) The experimental section needs to be expanded and subdivided into sub-section including all subsection required by the PRISMA criteria. Please, be accurate with all sub-sections and requirements of this checklist. Please, include the checklist of PRISMA as a supplemental file for review purposes.

• We agree with the reviewer, therefore we expanded and subdivided the experimental section followed the PRISMA criteria as requested (line 69). Moreover, we added the requested supplementary file with the PRISMA checklist.

Round 2

Reviewer 4 Report

Despite authors affirm that "Despite being originally planned, a meta-analysis of prevalence (with forest plots and I2 coefficients) for the different pain trajectories was not possible due to the paucity of the data retrieved from the systematic search of the literature. In fact, some of the trajectories are represented in only one or two of the included courts, thus limiting the statistical strength of a meta-analysis with such data. Moreover, the statistical heterogeneity of the analyzed data further limited the possibility to perform a statistically strong meta-analysis. Thus, we preferred to perform only a systematic review and express data as simple proportions with range in order to avoid presenting data as more powerful than they really are (less experienced readers perceive meta-analyses always as supported by a high level of evidence). We now addressed this point in the limitation section to further underline the need of more high-level studies on this interesting topic (lines 291-292)"; this reviewer believes that this data should be presented as a meta-analysis including in the limitation section the problems for interpretation. Thus, please add statistical analyses for heterogeneity to determine risk ratio estimates (I2 coefficient). These analyses need to be addressed in both methods and results section. Thanks.

Again, please add foresplots in addition to the limitations of this meta-analysis.

Finally, despite authors affirm that "We agree with the reviewer, therefore we expanded and subdivided the experimental section followed the PRISMA criteria as requested", authors have not subdivided the methods section into all subsecction recommended by the PRISMA criteria. Please, add these subsections. Thanks.

Author Response

Reviewer 4

Despite authors affirm that "Despite being originally planned, a meta-analysis of prevalence (with forest plots and I2 coefficients) for the different pain trajectories was not possible due to the paucity of the data retrieved from the systematic search of the literature. In fact, some of the trajectories are represented in only one or two of the included courts, thus limiting the statistical strength of a meta-analysis with such data. Moreover, the statistical heterogeneity of the analyzed data further limited the possibility to perform a statistically strong meta-analysis. Thus, we preferred to perform only a systematic review and express data as simple proportions with range in order to avoid presenting data as more powerful than they really are (less experienced readers perceive meta-analyses always as supported by a high level of evidence). We now addressed this point in the limitation section to further underline the need of more high-level studies on this interesting topic (lines 291-292)"; this reviewer believes that this data should be presented as a meta-analysis including in the limitation section the problems for interpretation. Thus, please add statistical analyses for heterogeneity to determine risk ratio estimates (I2 coefficient). These analyses need to be addressed in both methods and results section. Thanks.

• In order to avoid any doubt that we are not willing to perform a meta-analysis, we did, and we confirmed the problems we anticipated. Anyway, according to the reviewer’s request we added the methods and the results of the meta-analysis of prevalence performed (lines 121-130 and lines 203-207). As explained in the method section, the limited number of included studies hindered the possibility to solve the issue of heterogeneity. Thus, since a meta-analysis with these data may lead the reader to a biased and misleading conclusions, we still prefer to synthesize the results as proportions with range and I² coefficient, while we report the results of this weak meta-analysis in the supplementary files for completeness (but not as primary results as it would be misleading for many readers), to comply with the reviewer request.

Again, please add foresplots in addition to the limitations of this meta-analysis.

• Forest plots have been added in the supplementary files as requested.

Finally, despite authors affirm that "We agree with the reviewer, therefore we expanded and subdivided the experimental section followed the PRISMA criteria as requested", authors have not subdivided the methods section into all subsecction recommended by the PRISMA criteria. Please, add these subsections. Thanks.

• Although adding all the subchapters makes the manuscript heavy and is not in line with other manuscripts of this journal (we have been editors of this journal ourselves…), we added all the subsections requested by the reviewer.