Dextrose Prolotherapy for Symptomatic Grade IV Knee Osteoarthritis: A Pilot Study of Early and Longer-Term Analgesia and Pain-Specific Cytokine Concentrations

Round 1

Reviewer 1 Report (Previous Reviewer 2)

The current study has a relevant topic and is within the aim and scope of the journal. It is of interest to the readers of this journal. The manuscript is clearly written in professional, unambiguous language throughout. The author studied the effect of dextrose (D-glucose) injection for therapeutic purposes (dextrose prolotherapy; 18 DPT) in painful knee osteoarthritis (KOA) and whether the injection would favorably affect synovial fluid neurocytokine concentrations. The author has addressed my comments.

The correlation between painful knee osteoarthritis and concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ has been discussed in the manuscript.

A brief introduction of the way how the synovial fluid neurocytokines discussed in this manuscript affect painful knee osteoarthritis has been included in the article. The roles of neurocytokines (SP, CGRP, and NPY) has been discussed as well.

WOMAC0 in table 1 has been corrected to WOMAC.

Author Response

Comment 1: The current study has a relevant topic and is within the aim and scope of the journal. It is of interest to the readers of this journal. The manuscript is clearly written in professional, unambiguous language throughout. The author studied the effect of dextrose (D-glucose) injection for therapeutic purposes (dextrose prolotherapy; 18 DPT) in painful knee osteoarthritis (KOA) and whether the injection would favorably affect synovial fluid neurocytokine concentrations. The author has addressed my comments.

Reply 1: Thank you 

Comment 2: The correlation between painful knee osteoarthritis and concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ has been discussed in the manuscript.

Reply 2: Thank you.

Comment 3: brief introduction of the way how the synovial fluid neurocytokines discussed in this manuscript affect painful knee osteoarthritis has been included in the article. The roles of neurocytokines (SP, CGRP, and NPY) has been discussed as well.

Response 3: Thank you

WOMAC0 in table 1 has been corrected to WOMAC.

Response 4: Thank you. 

I very much appreciate the very helpful changes to the manuscript that were initiated by reviewer 1, and have been completed to his or her satisfaction 

Kind regards, 

K. Dean Reeves, M.D. , corresponding author

Reviewer 2 Report (New Reviewer)

Dears,

I invite you to read the attached file of reviews. The article is very interesting and deserving of attention, but it requires heavy revision work.

Review clinpract-2016583

In the proposed work, the authors hypothesize that dextrose (D-glucose) injection may cause in painful knee osteoarthritis. The results of statistical analysis show that the proposed treatment has benefits for patients. The statistical analysis, however, needs to be heavily revised, as some of the analyses conducted are inconsistent with the nature of the data. I found the introduction, materials and methods section straightforward and easy to understand.

Main comments:

1.      Since the variables being studied (NRS and WOMAC) are scales rather than non-numerical variables, it is impossible to compute their means, standard deviations, or ensuing t-tests. They can be evaluated in terms of the median because they are ordinal variables. Check out the references below for suggestions on the types of analyses that can be performed [1]. All tests, therefore, must be retested.

2.      Figure 2 is incorrect from a theoretical point of view since it cannot average a scale. Instead, I would propose to report boxplots at various time instants. Also, add in the title what kind of pain scale is depicted.

3.      Reading from the text, it is unclear what analysis you conducted. Initially, report in Table 2 the results of a t-test, then write in the footnote that you have "adjusted" the observations by setting the NRS scale equal to 0, which is the variables investigated. How would the groups be adjusted? Also, on page 5 and lines 172-173, you refer to a "General Linear Model." Where are the results of the model presented? This part should be better explained in the text and consistent with the presentation of the result. Finally, you cannot use a "simple" generalized linear model if you want to use a modelling approach. For example, using a mixed-effects model, you should consider the structure of repeated measurements.

4.      In a very small sample, a huge number of tests were run! After performing so many tests, it becomes necessary to "correct" the nominal alpha level. Therefore, I suggest using a correction like Bonferroni's for very large test batteries.

Minor comments:

1.      On table 1, I would remove the absolute number of women and report it in the text. If you want to report the presence of women with test, do so in terms of proportions (even if the two groups have the same numerosity).

2.      To make the text more readable, I would summarize the names of all subsections of the results section.

3.      Page 7, line 220: you cannot report the percentage improvement of a scale variable. Consider reporting the improvement in terms of median change.

4.      Page 7, line 237: a parenthesis is omitted.

5.      Page 8, line 249: you reported a p-value without a point (p = 033).

6.      In Table 4, the third row of each variable reported is not the mean but the difference between the means. Correct the name.

7.      I recommend dividing Table 2 into two tables, one for NRS results and one for WOMAC.

References

[1] Scaturro, D.; Vitagliani, F.; Terrana, P.; Cuntrera, D.; Falco, V.; Tomasello, S.; Letizia Mauro, G. Intra-Articular Hybrid Hyaluronic Acid Injection Treatment in Overweight Patients with Knee Osteoarthritis: A Single-Center, Open-Label, Prospective Study. Appl. Sci. 2021, 11, 8711. https://doi.org/10.3390/app11188711

Author Response

Comment 1: Since the variables being studied (NRS and WOMAC) are scales rather than non-numerical variables, it is impossible to compute their means, standard deviations, or ensuing t-tests. They can be evaluated in terms of the median because they are ordinal variables. Check out the references below for suggestions on the types of analyses that can be performed [1]. All tests, therefore, must be retested.

Response 1:

Although treating a 0-10 NRS score as a Likert scale is a point of debate among statisticians, it is perfectly reasonable to do so.  Table 2 has been modified to list medians and interquartile ranges as suggested.

However, the 0-100 WOMAC composite score has a comparatively large response range.  For that reason, it is appropriate to treat WOMAC data as interval data, and to use that data for calculation of means and standard deviation, unless the data is non-parametric.  Therefore, we consider that the approach we took to analysis and presentation of WOMAC data to be the most common approach, and consistent with standard practice.   

Comment 2:  Figure 2 is incorrect from a theoretical point of view since it cannot average a scale. Instead, I would propose to report boxplots at various time instants. Also, add in the title what kind of pain scale is depicted.

Response 2:

Thank you.  For figure 2, given the small size of the study, with only 10 in each group, we thought it optimally clear for readers to present a simple bar graph depicting the medians at various time instants.  We did place in the title that the NRS pain scale is depicted.

Comment 3.      Reading from the text, it is unclear what analysis you conducted. Initially, report in Table 2 the results of a t-test, then write in the footnote that you have "adjusted" the observations by setting the NRS scale equal to 0, which is the variables investigated. How would the groups be adjusted? Also, on page 5 and lines 172-173, you refer to a "General Linear Model." Where are the results of the model presented? This part should be better explained in the text and consistent with the presentation of the result. Finally, you cannot use a "simple" generalized linear model if you want to use a modelling approach. For example, using a mixed-effects model, you should consider the structure of repeated measurements.

Response 3:

We modified our analysis paragraph to accommodate changes merited by the different analysis of NRS data, and to maximize clarity.  We modeled the outcome at each time point separately to identify the change, and present the results better to readers. Therefore, we didn’t use a mixed-effect model.  

Here is how the analysis section of method reads

“Previous data on neurocytokine changes after DPT were not available. Sample size was determined by convenience and resource limitations, and was limited to 10 in each group in this pilot data collection. Data were analyzed using PASW Statistics 18, Release 18.0.0, IBM and SAS version 9.4. Analysis was performed by intention-to-treat. A between-group analysis for differences in baseline characteristics was performed using t-tests for normal data, Mann-Whitney tests for Likert scale data, and Pearson chi-square test for categorical variables. Baseline characteristics that met or approached a significant difference between-group were included as covariates in the follow up analyses. Mann-Whitney tests were performed for NRS raw score to test the difference from baseline to different time points. For the NRS raw score, differences between baseline and each follow-up time point (20 minutes, one week, 3, 6 and 9 months) were calculated and treated as outcome variables in Proportional Odds models to estimate the significance of between-group differences. Differences between baseline and each follow-up time point (3 and 6 months) for WOMAC were calculated and treated as outcome variables in General Linear model analyses. Change scores in Cytokine data were not normally distributed. P-values for changes in cytokine concentrations were calculated by Mann-Whitney testing for between-group comparisons from 0 to 1 week, and by Wilcoxon matched-pair signed-ranks test for within-group analysis from time 0 to 3 months.”

Comment 4.      In a very small sample, a huge number of tests were run! After performing so many tests, it becomes necessary to "correct" the nominal alpha level. Therefore, I suggest using a correction like Bonferroni's for very large test batteries

.

Response 4: Thank you for this suggestion. We analyzed eight cytokines. Therefore, we added the following statement onto the end of the analysis paragraph, and included this is the footnotes for tables 3 and 4.

  “A Bonferroni corrected alpha of .006 (.05/8) was utilized for determination of significance for within and between-group analysis of change scores for the eight cytokines analyzed.”

Comment 5:  On table 1, I would remove the absolute number of women and report it in the text. If you want to report the presence of women with test, do so in terms of proportions (even if the two groups have the same numerosity).

Response 5:  The proportion of women was removed from table 1 and placed in the first paragraph of results as follows “The gender distribution was equally balanced in each group (5/10 females in Dextrose 1st and 6/10 females in the Aspiration 1st group.”

Comment 6: To make the text more readable, I would summarize the names of all subsections of the results section.

Response 6:  Excellent suggestion. The names of subsections were substantially simplified.

Comment 7: Page 7, line 220: you cannot report the percentage improvement of a scale variable. Consider reporting the improvement in terms of median change.

Response 7:  We chose to remove percentage improvements from the tables and discussion to simplify and characterize the presentation of results. Thank you for your suggestion which led to this useful simplification.

Comment 8:  Page 7, line 237: a parenthesis is omitted.

Response 8:  Parentheses were placed about second p values within that sentence

Comment 9: Page 8, line 249: you reported a p-value without a point (p = 033).

Response 9: That line was deleted, as the change in that cytokine was no longer significant after Bonferroni correction.

Comment 10: In Table 4, the third row of each variable reported is not the mean but the difference between the means. Correct the name.

Response 10: Thank you so much for catching this.  The same change needed to be made in Table 3.  Means were changed to “mean difference” and a footnote was placed under the table explaining the abbreviation”

Comment 11: I recommend dividing Table 2 into two tables, one for NRS results and one for WOMAC.

Response 11:  The WOMAC results have been placed into a new table, table 5.

Round 2

Reviewer 2 Report (New Reviewer)

Dear authors. Thank you for considering my comments. 

I only have one note to make: the scale of the WOMAC scale, as the name suggests, is ordinal. Therefore, it must be treated as such. That it is treated as continuous in the literature does not justify an incorrect treatment of it. The results and substance will not change, but what makes the difference is the authority and elegance of the analysis. 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Very interesting and well-written pilot study. However, I have 2 issues.

1.  I suggest that an introductory paragraph on KOA followed by dextrose prolotherapy should be added to the introduction section.

2. Why was cytokine analysis not done at 6 and 9 months?

Author Response

Reviewer 1

Comment 1.  I suggest that an introductory paragraph on KOA followed by dextrose prolotherapy should be added to the introduction section.

Response 1:  Thank you. We need to set the stage more clearly for why this study was considered necessary. The following paragraph on KOA and dextrose prolotherapy was added to the introduction section, requiring a few supportive citations.

“Knee osteoarthritis (KOA) affects around 30% of adults by age 65 [1], and is responsible for substantial chronic pain and disability worldwide [2]. Identifying conservative and safe care that complements current strategies is a top priority in research [3]. Dextrose prolotherapy (DPT), the injection of hypertonic dextrose for therapeutic purposes,  has outperformed both anesthetic [4], steroid [5], saline injections [6, 7], and exercise controls [8, 9] in multiple randomized trials for improvement in both function and pain. Why dextrose injection would reduce pain in KOA is unknown. 

Comment 2. Why was cytokine analysis not done at 6 and 9 months?

Response 2: For this largely-self-funded study we did not have funds for additional ELISA kits for analysis at 6 and 9 months. This needs to be clear and is a limitation that should be mentioned.   We modified our primary limitation statement as follows (new area underlined)

The primary limitations of our study were its small size, a limited array of cytokines, absence of ELISA measurements at 6 and 12 months, and non-blinding of the injector and assessor. Expense constraints limited the frequency of ELISA measurement to three points in time, including baseline. Given the extreme inflation in Argentina, associated with a concern of the general population about losing jobs due to taking unnecessary time off work, we were concerned about loss of aspirate data at 6 and 12 months. Therefore, we elected to aspirate for ELISA analysis at baseline, one week, and three months. Strengths of our study include a randomized design with a masked control for the initial phase, a long-term serial evaluation of clinical outcomes, and the assessment of neurocytokine concentration as an outcome for the first time.

Reviewer 2 Report

The current study has a relevant topic and is within the aim and scope of the journal. It is of interest to the readers of this journal. The manuscript is clearly written in professional, unambiguous language throughout. In this study, 20 participants with grade IV symptomatic KOA received synovial fluid aspiration followed by dextrose or simulated dextrose injection. At day 0, day 1, and 3-month time points, the concentration of neurocytokines (SP, CGRP, and NPY) was analyzed. This study has a masked control at the first phase. It provided a long-term serial evaluation of clinical outcomes and the assessment of neurocytokine concentration. I have minor comments below.

Please discuss the correlation between painful knee osteoarthritis and concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ.

What is the reason why the concentration of neurocytokines was stopped at three months instead of continuing at six-month time point?

What is the correlation between NRS and the concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ?

What is the correlation between WOMAC and the concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ?

What is the meaning of WOMAC0 in table 1? WOMAC is used throughout the manuscript.

Author Response

Reviewer 2

Comment 3: Please discuss the correlation between painful knee osteoarthritis and concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ.

Response 3:  Here is a table (please see the attachment) that includes results from Pearson bivariate two-tailed  correlation between pain with KOA (NRS from 0-20 min) and cytokine changes in dextrose and control participants (Comment 3),  Pearson correlations between NRS and cytokine changes in all participants at 3 months (Comment 5),  and  Pearson correlations between WOMAC and cytokine changes at 3 months (Comment 6).

This table indicates only one positive correlation, which did not occur in an area where there was a significant difference between groups. This one positive correlation means little in isolation without a corroborating/corresponding between-group difference by independent sample T testing for the same variables. Therefore, these findings appear to be non-contributory, and potentially confusing. For that reason, we would prefer to not add another table to the document. The following paragraph could be added by request, but our first inclination is to not include it for sake of clarity.

Pearson bivariate two-tailed correlation revealed no significant correlation between either pain (NRS) or WOMAC changes from baseline to short-term or medium-term follow-up and cytokine levels changes over the same time periods except between NPY and NRS from baseline to week 1 for the control group.  This lone significant correlation was not accompanied by a corroborative finding of a significant between-group difference by an independent samples T-test.

Comment 4: What is the reason why the concentration of neurocytokines was stopped at three months instead of continuing at six-month time point?

Response 4:  See response to comment 2.

Comment 5: What is the correlation between NRS and the concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ?

Response 5:  See response to comment 3

Comment 6: What is the correlation between WOMAC and the concentration changes in neurocytokines (SP, CGRP, and NPY) and selected non-neurocytokines (MMP-3, TIMP-1, IL-6, IGF, and TGFβ?

 Response 6: See response to comment 3

Comment 7: What is the meaning of WOMAC0 in table 1? WOMAC is used throughout the manuscript.

Response 7: We altered the title of table 1 to “Baseline Demographic Comparison Between Groups  and altered the “Measures” column to  WOMAC from WOMAC0

Author Response File: Author Response.docx