Targeted Microbial Shifts and Metabolite Profiles Were Associated with Clinical Response to an Anti-Inflammatory Diet in Osteoarthritis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Manuscript s well written but there are aspects that should be improved:

- interpretation of causality in this article is too strong: statements connected with the diet caused microbial changes should be softened to "associated with"

- 21 participants completed the study what may be underpowered- this is one of limitations and places work as the pilot

- figures sometimes emphasize uncorrected p-values, which may be misleading

- please correct - please justify 66.2% adherence described as "high" because usually >80% is considered high

- please correct grammar because sentences are too long. Breaking them would improve readability

- WOMAC total score baseline is eported as "452.61 ± 13.97" - is it a typo or unconverted raw value

- unify terminology "non-responders" and "no responders"

- BMI given as "31.57 (± 10.03)" — SD of 10 seems large

 

Comments on the Quality of English Language

Manuscript s well written but there are aspects that should be improved:

- interpretation of causality in this article is too strong: statements connected with the diet caused microbial changes should be softened to "associated with"

- 21 participants completed the study what may be underpowered- this is one of limitations and places work as the pilot

- figures sometimes emphasize uncorrected p-values, which may be misleading

- please correct - please justify 66.2% adherence described as "high" because usually >80% is considered high

- please correct grammar because sentences are too long. Breaking them would improve readability

- WOMAC total score baseline is eported as "452.61 ± 13.97" - is it a typo or unconverted raw value

- unify terminology "non-responders" and "no responders"

- BMI given as "31.57 (± 10.03)" — SD of 10 seems large

 

Author Response

Reviewer 1 - Comment 1

Interpretation of causality in this article is too strong: statements connected with the diet causing microbial changes should be softened to “associated with.”

Response:
We thank the reviewer for this valuable observation. In response, we have revised the manuscript to ensure that statements do not imply direct causality. Specifically:

• In the Abstract, we changed “demonstrates that a short-term anti-inflammatory dietary intervention can lead to meaningful” to “was associated with meaningful,” and “Lachnospiraceae family emerged as key taxa linked to pain reduction and anti-inflammatory metabolite production” to “associated with pain reduction and anti-inflammatory metabolite production.”
• In the Introduction, we changed “The study found that while the diet led to a reduction in the pro-inflammatory cytokine” to “was associated with a reduction in the pro-inflammatory cytokine.”
• In the Limitations section, we changed “where sham interventions have produced significant symptomatic relief” to “where sham interventions were associated with significant symptomatic relief.”

These edits ensure the language reflects associations rather than implying causation, in line with the observational nature of this pilot study.

Reviewer 1 - Comment 2

21 participants completed the study which may be underpowered – this is one of the limitations and places the work as a pilot.

Response:
We agree with the reviewer. This point is already acknowledged in the first paragraph of the Limitations section, where we state that this was a pilot study with a small sample size (n = 21), which limits generalizability and increases the risk of type II error.

Reviewer 1 - Comment 3

Figures sometimes emphasize uncorrected p-values, which may be misleading.

Response:
We thank the reviewer for this comment. To address this, we have revised the figure legends for Figures 3, 5, 6, and 7 to clarify the interpretation of p-values. In the updated legends, results that remain significant after false discovery rate (FDR) correction (q<0.05) are explicitly labeled FDR and visually emphasized. Significant results without multiple testing correction (p<0.05) are labeled as NonFDR, described as exploratory, and not presented as primary findings. Non-significant results are labeled NS (p>0.05). This standardized labeling and legend clarification ensure that corrected results are prioritized and reduce the potential for misinterpretation of uncorrected p-values.

Reviewer 1 - Comment 4

Please correct – please justify 66.2% adherence described as “high” because usually >80% is considered high.

Response:

We thank the reviewer for this observation. We agree with this point — the description of 66.2% adherence as “high” in the abstract was an oversight. This has been corrected to “good adherence” in the revised version.

Reviewer 1 - Comment 5:

Please correct grammar because sentences are too long. Breaking them would improve readability.

Response: We thank the reviewer for this helpful suggestion. We have revised the manuscript extensively to improve clarity, conciseness, and readability. Specifically, we shortened overly long sentences, reduced redundancy in statistical reporting, and ensured consistent formatting of p-values, test names, and feature nomenclature. In the Results section, we restructured the gut microbiome, metabolome, and plasma metabolome subsections to clearly separate responder (R) and non-responder (NR) findings, reorganized the “delta” analysis for better methodological flow, and clarified descriptions in the DIABLO network analysis. For the salivary microbiome and metabolome, we grouped results by comparison type (post-intervention, baseline, endpoint, and trajectories) and moved supplementary figure references to the start of each subsection for easier navigation. These changes improve the manuscript’s readability while preserving the accuracy and integrity of the results.

Examples of revisions:

Section	Before	After
Gut microbiome (Figure 5A)	“In the gut microbiome (Figure 5A), R and NR did not exhibit a significantly different alpha diversity (alpha diversity: p=0.779, Wilcoxon Rank-Sum; beta diversity: p=0.358, PERMANOVA), but an enrichment of Clostridium and a depletion of Anaerococcus, Cloacibacterium (Weeksellaceae), Anaerostipes, Limivivens, and Parabacteroides was observed in R (Wilcoxon Signed-Rank p<0.05).”	“In the gut microbiome (Figure 5A), R and NR did not differ in alpha diversity (p=0.779, Wilcoxon Rank-Sum) or beta diversity (p=0.358, PERMANOVA). However, R showed an enrichment of Clostridium and a depletion of Anaerococcus, Cloacibacterium (Weeksellaceae), Anaerostipes, Limivivens, and Parabacteroides (p<0.05, Wilcoxon Signed-Rank).”
Delta analysis	“For each dataset, we first calculated the difference (delta) between pre- and post-intervention relative abundance values separately within R samples only and NR samples only. We then compared these differences between groups (testing the differences in distributions of deltas), identifying features where the degree of change was significantly different in R versus NR. This comparison revealed notable distinctions. In the gut microbiome (Figure 7A), Anaerostipes, Limivivens, Anaerococcus, and Cloacibacterium (Weeksellaceae) exhibited a significantly more positive change in R compared to NR (p<0.05), while Lachnospiraceae COE1 showed a more pronounced negative shift in R than NR.”	Paragraph 1: “For each dataset, we calculated the change (delta) between pre- and post-intervention values separately for R and NR.” Paragraph 2: “We then compared these deltas between groups, identifying features where changes differed significantly between R and NR. In the gut microbiome (Figure 7A), Anaerostipes, Limivivens, Anaerococcus, and Cloacibacterium (Weeksellaceae) increased more in R (p<0.05), whereas Lachnospiraceae COE1 decreased more in R than in NR.”
DIABLO analysis	Long paragraph describing method, baseline results, post-intervention results, and taxa-metabolite associations together.	Three shorter paragraphs: one on the method and visualization approach, one summarizing baseline results, and one summarizing post-intervention changes.
Salivary microbiome	“In the salivary microbiome (Supplementary Figure 5A), we did not observe significant changes in diversity metrics post-intervention (alpha diversity: p=0.266, Wilcoxon Signed-Rank; beta diversity: p=0.889, PERMANOVA), but did observe a significant depletion of Veillonella and Nanosyncoccus after dietary intervention. In the salivary metabolome (Supplementary Figure 5B), there were no significant changes in alpha or beta diversity (alpha diversity: p=1, Wilcoxon Signed-Rank; beta diversity: p=0.798, PERMANOVA), with a depletion in Nicotinuric acid and Theobromine (Wilcoxon Signed-Rank p<0.05)...”	Results separated into four short paragraphs, each covering one comparison type (post-intervention, baseline, endpoint, trajectories), with statistical results clearly separated from interpretation.

 

Reviewer 1 - Comment 6

WOMAC total score baseline is reported as "452.61 ± 13.97" - is it a typo or unconverted raw value

 

Response:

We thank the reviewer for noticing this. This was a typographical error in the original version. The correct baseline WOMAC total score is 52.61 ± 13.97, and this has been updated in the revised manuscript.

 

 

Reviewer 1 - Comment 7

Unify terminology “non-responders” and “no responders.”

Response:
We thank the reviewer for noting this inconsistency. We have revised the manuscript to use the term “non-responders” uniformly throughout the text, figures, and tables.

 

Reviewer 1 - Comment 8

BMI given as "31.57 (± 10.03)" — SD of 10 seems large.

Response:
We appreciate the reviewer’s observation. The relatively large standard deviation in BMI reflects the heterogeneity of our study population, which included participants with a wide range of body weights. For example, some participants had BMI values in the normal range (~21), while others had severe obesity (BMI up to ~59). This broad distribution is consistent with the inclusion of patients across the full BMI spectrum and naturally results in a higher variability measure (SD). We have clarified this in the manuscript by noting the wide range of BMI values.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. The study used an open label single arm design and could not rule out the effects of placebo, regression to mean, or natural course. Pain score (VAS/WOMAC) is a highly subjective indicator that is susceptible to expected bias interference.
2. The chord diagram in Fig 8 does not label metabolite names (only HMDB IDs), resulting in poor readability.
3. The methodology section should be segmented according to different content to enhance the readability of the article
4. Changes in the Lachnospiraceae family were observed to be associated with pain improvement (Fig 8D-G), but confounding factors such as BMI and comorbidities were not controlled for.
5. Short term effects of 4-week intervention not discussed: changes in microbiota/metabolites may not have reached steady state, and long-term effects are unknown.
6. Saliva omics results are negative, but the reasons have not been thoroughly analyzed (such as whether the sampling reflects systemic inflammation?).

Author Response

Reviewer 2 – Comment 1:

The study used an open-label single-arm design and could not rule out the effects of placebo, regression to the mean, or natural course. Pain score (VAS/WOMAC) is a highly subjective indicator that is susceptible to expectation bias interference.

Response:
We thank the reviewer for this important observation. We agree that the lack of a placebo or control group limits our ability to differentiate the true effects of the intervention from expectation-related responses, regression to the mean, or the natural course of symptoms. This limitation has now been explicitly addressed in the revised manuscript. In particular, the Limitations section has been updated to highlight the potential influence of placebo effects and expectation bias when interpreting subjective pain outcomes such as VAS or WOMAC scores. Yet, we would like to note that the WOMAC pain subscale is the standard and widely accepted primary outcome measure in osteoarthritis clinical trials, providing a validated framework for assessing pain despite its inherent subjectivity. These additions emphasize the exploratory nature of our findings and the need for future placebo-controlled designs.

Updated Limitations section (revised paragraph): “Third, the absence of a placebo or control group precludes fully separating the effects of the intervention from placebo responses, regression to the mean, or the natural course of symptoms. Placebo effects are well documented in pain research and can arise from participants’ expectations, increased clinical attention, or perceived novelty of the intervention, leading to symptom improvement independent of the intervention’s biological mechanism. Regression to the mean is also a concern in studies enrolling participants with high baseline pain, as extreme values tend to move closer to the population average upon repeated measurement, regardless of intervention This is particularly relevant for subjective pain measures such as WOMAC pain score, which is susceptible to expectation bias. While we attempted to minimize bias through clear inclusion and exclusion criteria and adjusted for demographic and clinical covariates (e.g., sex, BMI, diabetes) in generalized linear models, the lack of blinding and a control arm remains an important consideration when interpreting the results. Larger, placebo-controlled trials with longer follow-up are warranted to confirm and expand upon these findings.”

Reviewer 2 – Comment 2:

The chord diagram in Fig. 8 does not label metabolite names (only HMDB IDs), resulting in poor readability.

Response:
We appreciate the reviewer’s feedback regarding figure readability. In the revised version, we have replaced the HMDB IDs with the corresponding metabolite names and increased the font size of the labels to improve clarity. These changes enhance the interpretability of Fig. 8 and facilitate a more intuitive understanding of the metabolite associations

 

 

Reviewer 2 – Comment 3:

The methodology section should be segmented according to different content to enhance the readability of the article.

Response:
We thank the reviewer for this suggestion. In the revised manuscript, we have reformatted the Materials and Methods section to improve clarity and navigation. Specifically, we have organized the section using numbered subheadings that clearly separate each methodological component, as follows:

• 1. Study Participants and Study Design
• 2. Primary and Secondary Outcomes
• 3. Metabolome Sample Preparation for LC-MS/MS Analyses
• 4. Microbiome Data Acquisition
• 5. Microbiome Data Processing
• 6. Statistical Analyses

This restructuring, together with paragraph breaks, enhances the visual organization of the methods and allows readers to quickly locate specific details without navigating a single large block of text.

Reviewer 2 – Comment 4:

Changes in the Lachnospiraceae family were observed to be associated with pain improvement (Fig. 8D–G), but confounding factors such as BMI and comorbidities were not controlled for.

Response:

We sincerely thank the reviewer for this insightful comment. As described in lines 409–421 of the revised manuscript and in Supplementary Table 11, we performed generalized linear models (GLMs) including BMI and comorbidities as covariates to address potential confounding effects. Notably, the observed associations remained consistent after this adjustment.

Reviewer 2 – Comment 5:

Short term effects of 4-week intervention not discussed: changes in microbiota/metabolites may not have reached steady state, and long-term effects are unknown.

Response:
We thank the reviewer for this valuable observation. We have now expanded the Limitations section to explicitly acknowledge that the short duration of the intervention may not allow microbiota and metabolite profiles to reach a new steady state, and that the long-term effects remain unknown. The revised text reads as follows (highlighted in yellow in the manuscript):

“Second, the intervention period was relatively short (4 weeks), which may not capture long-term effects on pain, metabolic outcomes, or microbiome profiles. Changes in microbiota and metabolite composition observed within this timeframe may not have reached a new steady state, and it is unknown whether these effects would persist, diminish, or evolve with prolonged dietary intervention.”

This addition clarifies the temporal limitations of our findings and the need for future studies to evaluate their long-term impact.

Reviewer 2 – Comment 6:

Saliva omics results are negative, but the reasons have not been thoroughly analyzed (such as whether the sampling reflects systemic inflammation?).

Response:
We thank the reviewer for this insightful comment. To address this point, we have expanded the Discussion to acknowledge possible reasons for the negative saliva-omics findings. Specifically, we have added the following paragraph:

“Saliva-omics analyses in our cohort yielded null findings. Whole saliva may be a relatively insensitive matrix for detecting systemic or joint inflammation, particularly in cohorts with low periodontal disease burden and low inflammatory activity. Prior studies, especially in RA, have reported stronger and more consistent oral microbiome alterations when sampling subgingival plaque rather than saliva, with taxa such as Porphyromonas gingivalis, Prevotella, and Veillonella linked to systemic inflammatory markers, autoantibody production, and even bacterial product translocation to the joint space [61]. In OA, reported associations with the oral microbiome are more heterogeneous and generally weaker, though potential pathogenic taxa and functional pathways (e.g., lipopolysaccharide biosynthesis) have been identified in saliva [62]. The differences across oral niches, combined with the high short-term variability of salivary communities and the limitations of single time-point sampling, likely reduced our ability to detect consistent associations in this study.”

This addition incorporates evidence from prior literature on both RA and OA and provides a clearer rationale for why saliva may not have reflected systemic inflammation in our study.

Author Response File: Author Response.pdf