Review Reports - Hyperuricemia as a Biomarker for Circadian Syndrome: A Cross-Sectional Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this study, the authors investigate the relationship between hyperuricemia (HUA) and circadian syndrome (CircS). Data were drawn from the National Health and Nutrition Examination Survey (NHANES) between 2005 and 2020, involving 27,410 adults. Their findings showed that HUA is the strongest independent predictor of CircS, alongside other risk factors such as chronic kidney disease, female sex, and smoking. The study concludes that HUA is associated with an increased risk of CircS and suggests that it could serve as an early biomarker for risk detection. Overall, this study is interesting, and the results generally align with existing knowledge and reinforce findings from prior studies using NHANES data. I only have a few suggestions for improvements that came to my mind.

Comment #1: I am missing the sense of urgency about the current research in the introduction. What are the specific gaps and limitations in the literature to further assess the relationship between HUA and CircS? Why is there a need for reanalyzing the NHANES dataset or doing a new analysis here? Please present scientific hypotheses and how the analysis was performed to verify these hypotheses. Also, highlight the key findings of this study at the end of the introduction. Furthermore, I think that the study presents the findings in a way that could be seen as overstating the causation or disease progression, particularly given that NHANES is a cross-sectional survey; the data establishes associations, not causation. Please revise related phrases, e.g., Line 62 and Lines 267-268.

Comment #2: In lines 80-84, the authors note that of 45,980 adults (>18 y), 28,577 were excluded for missing uric acid and 18,570 for missing CircS status. In my opinion, this is a massive exclusion rate. Please provide a brief supplementary table or paragraph comparing the baseline characteristics of the included versus excluded adults to prove that this missingness is missing at random and does not introduce selection bias. Moreover, at line 83, the final retained sample size is n = 27,410, which matches Table 1 (line 178). However, the title of Table 2 (Line 211) states that the multivariable logistic regression is based on n = 13,782. This means almost 50% of the study population was dropped from the main regression model. Please also clarify whether the appropriate NHANES complex-survey weights were applied to the multivariable logistic regression model to account for the complex sampling design.

Comment #3: For Figure 1, the legend contains very limited information. I recommend expanding some information, sample size, error, abbreviations, etc., when applicable. Also, refer to the figure in the main text.

Comment #4: The discussion section elaborates on the repetition of results. Consider refining or merging the text there and throughout the manuscript to avoid repeating the same information. At lines 279-280, the mean continuous uric acid levels for participants with and without CircS were 348.1 vs. 306.3 μmol/L, respectively. However, this data is strangely absent from the baseline characteristics in the results. Also, the dose-response analysis shows that HUA strongly predicts metabolic components but weakly predicts purely circadian components, such as short sleep and depression (Lines 243-246). Does this imply HUA is a biomarker of circadian syndrome, or does it merely remain a strong biomarker for traditional metabolic syndrome?

Author Response

Comment #1:

I am missing the sense of urgency about the current research in the introduction. What are the specific gaps and limitations in the literature to further assess the relationship between HUA and CircS? Why is there a need for reanalyzing the NHANES dataset or doing a new analysis here? Please present scientific hypotheses and how the analysis was performed to verify these hypotheses.

Also, highlight the key findings of this study at the end of the introduction.

Furthermore, I think that the study presents the findings in a way that could be seen as overstating the causation or disease progression, particularly given that NHANES is a cross-sectional survey; the data establishes associations, not causation. Please revise related phrases, e.g., Line 62 and Lines 267-268.

Response 1:

We thank the reviewer for highlighting the importance of clarifying the novelty and scientific significance of our study. We have revised the Introduction to emphasize the urgency of investigating the relationship between hyperuricemia (HUA) and Circadian Syndrome (CircS), as well as to identify existing gaps in the literature. While a recent NHANES-based study published in Scientific Reports examined the association between CircS and serum uric acid, our study extends this work by incorporating non-alcoholic fatty liver disease (NAFLD), a key hepatic manifestation of metabolic dysfunction. The inclusion of NAFLD enhances the clinical and biological relevance of the CircS framework, as it represents a critical link between uric acid metabolism, insulin resistance, chronic inflammation, and circadian disruption. Unlike previous analyses, our study evaluates HUA as a potential biomarker of an expanded circadian, metabolic phenotype that includes hepatic involvement.

Unlike prior NHANES analyses that operationalized CircS using a 7-component definition, our study extends the phenotype by incorporating NAFLD, a clinically important hepatic manifestation of metabolic dysfunction that is biologically linked to both hyperuricemia and circadian disruption. Because NAFLD is a major organ-level manifestation of metabolic dysfunction, excluding it may underestimate the metabolic burden captured by CircS.
At the end of intro In this nationally representative cross-sectional analysis of NHANES data, hyperuricemia emerged as a strong and independent predictor of Circadian Syndrome. Elevated serum uric acid levels were significantly associated with increased odds of CircS, supporting its potential utility as an accessible biomarker for early risk stratification.

Comment #2:

In lines 80-84, the authors note that of 45,980 adults (>18 y), 28,577 were excluded for missing uric acid and 18,570 for missing CircS status. In my opinion, this is a massive exclusion rate. Please provide a brief supplementary table or paragraph comparing the baseline characteristics of the included versus excluded adults to prove that this missingness is missing at random and does not introduce selection bias.

Moreover, at line 83, the final retained sample size is n = 27,410, which matches Table 1 (line 178). However, the title of Table 2 (Line 211) states that the multivariable logistic regression is based on n = 13,782. This means almost 50% of the study population was dropped from the main regression model. Please also clarify whether the appropriate NHANES complex-survey weights were applied to the multivariable logistic regression model to account for the complex sampling design.

Response 2:

Thank you for this important comment. We agree that the exclusion of participants due to missing data could introduce selection bias. To address this concern, we conducted a comparison of baseline characteristics between included and excluded participants, which has now been added as Supplementary Table 1.

We found that excluded individuals were older, more likely to be female, and differed in several socioeconomic and clinical characteristics, including a higher prevalence of hyperuricemia and chronic kidney disease. These findings suggest that missingness may not be completely at random.

We have now acknowledged this as a limitation in the revised manuscript. Importantly, the consistency of our results across multiple sensitivity analyses supports the robustness of the observed associations despite these differences.

The discrepancy in sample size arises because the multivariable logistic regression analysis included only participants with complete data for all variables incorporated into the model. Specifically, individuals with missing information on exposure, outcomes, or covariates were excluded through a complete-case analysis. As a result, the final analytic sample for the regression was reduced to n = 13,782 from the initially retained n = 27,410. This approach ensured the accuracy and validity of the model estimates by avoiding bias introduced by incomplete responses.

Furthermore, the multivariable logistic regression analysis appropriately applied the NHANES complex survey weights, along with stratification and clustering variables, to account for the multistage probability sampling design. This ensures that the findings are nationally representative and statistically robust.

This clarification has been added to the revised manuscript to enhance transparency and methodological rigor.

Comment #3:

For Figure 1, the legend contains very limited information. I recommend expanding some information, sample size, error, abbreviations, etc., when applicable. Also, refer to the figure in the main text.

Response #3:

Edits have been made to the figure legend, and the Figure was addressed in the manuscript line 130.

Comment #4:

The discussion section elaborates on the repetition of results. Consider refining or merging the text there and throughout the manuscript to avoid repeating the same information.
At lines 279-280, the mean continuous uric acid levels for participants with and without CircS were 348.1 vs. 306.3 μmol/L, respectively. However, this data is strangely absent from the baseline characteristics in the results.
Also, the dose-response analysis shows that HUA strongly predicts metabolic components but weakly predicts purely circadian components, such as short sleep and depression (Lines 243-246). Does this imply HUA is a biomarker of circadian syndrome, or does it merely remain a strong biomarker for traditional metabolic syndrome
Most importantly, study critically lacks objective circadian markers: claims of a specific hyperuricemia-circadian disruption link are weakened without actigraphy, melatonin, or light exposure data. Observed associations may reflect general metabolic dysfunction rather than circadian-specific effects. Unmeasured confounders like chronotype, shift work, or genetic factors (e.g., circadian clock and melatonin receptors gene variants) could distort results.

The Healthy Eating Index overlooks circadian-relevant factors like meal timing and chrono nutrition. Its null association with circadian syndrome likely stems from measurement limitations, not absence of effect (e.g., as shown in several publications by Panda’s and Garaulet’s groups).

Statistical and modelling concerns: Supplemental Figure 2 reveals a non-linear age-circadian syndrome relationship (U-shaped), which may confound or bias the age-adjusted odds ratios provided in Supplemental Figure 3. Consider spline models or stratified analyses to address this.

Response 4:

Thank you for this valuable suggestion. We have carefully revised the Discussion section to eliminate redundancy and avoid repetition of statistical findings already presented in the Results. Specifically, detailed numerical comparisons—such as mean uric acid levels—have been removed or summarized, and repetitive descriptions of sensitivity analyses have been condensed. The Discussion now emphasizes interpretation, comparison with existing literature, and clinical implications. These revisions have improved the clarity, conciseness, and overall readability of the manuscript.

This discrepancy was due to a typographical error. The omission of the mean continuous uric acid levels for participants with and without CircS from the baseline characteristics table has now been corrected.

The dose-response analysis indeed demonstrated that hyperuricemia (HUA) strongly predicts metabolic components while showing weaker associations with purely circadian components, such as short sleep duration and depression (Lines 243–246). However, this does not diminish its relevance as a biomarker of circadian syndrome (CircS).

CircS is a multidimensional construct that integrates both metabolic and circadian-related components. In this study, our focus was on CircS as a composite condition rather than on its individual components in isolation. The stronger association between HUA and metabolic disturbances reflects the well-established pathophysiological links between uric acid, insulin resistance, oxidative stress, and systemic inflammation. Since metabolic dysfunction constitutes a central pillar of CircS, these findings support the role of HUA as an indicator of the syndrome as a whole.

The comparatively weaker associations with purely circadian components, such as short sleep and depression, suggest that HUA may not directly reflect circadian disruptions but rather their downstream metabolic consequences. Therefore, HUA should be interpreted as a biomarker of CircS in its entirety, primarily driven by its metabolic dimension, rather than as a marker of isolated circadian disturbances or solely traditional metabolic syndrome.

We agree that the absence of objective circadian markers, such as actigraphy, melatonin levels, or light exposure data, limits our ability to directly assess circadian rhythm disruption. We have now clarified in the revised manuscript that the observed associations may partly reflect broader metabolic dysfunction rather than circadian-specific effects.

We also acknowledge that unmeasured confounders, including chronotype, shift work, and genetic variation in circadian-related pathways, could influence the observed relationships. These factors have now been explicitly discussed in the limitations section.

Importantly, our study was conducted within a predictive modelling framework, which prioritizes predictive performance rather than causal inference. While confounding is less critical in this context for prediction purposes, it remains relevant when interpreting potential biological mechanisms. We have clarified this distinction in the revised discussion and have moderated our language accordingly to avoid overstating circadian-specific interpretations.

Regarding the Healthy Eating Index (HEI-2015), we agree that it does not capture circadian-relevant dietary behaviours such as meal timing or chrono nutrition. We have revised the discussion to emphasize that the lack of association likely reflects measurement limitations rather than absence of a true effect and have clarified this point in the manuscript.

We agree that the relationship between age and Circadian Syndrome appears non-linear, as illustrated in Supplementary Figure 2.

We would like to clarify that age was modelled using restricted cubic splines in the primary multivariable analysis to appropriately account for this non-linearity, as reflected in the main regression model and supplementary analyses. The use of a linear age term in Supplementary Figure 3 (forest plot) was done solely for visual simplicity and ease of interpretation as already stated in lines (362-363). However, we are willing to revise or remove the simplified visualization if deemed necessary.

Reviewer 2 Report

Comments and Suggestions for Authors

This is a well-designed, methodologically rigorous cross-sectional epidemiological study that advances understanding of the association between hyperuricemia and circadian-metabolic syndrome.

However, there a re several major concerns that should be properly addressed:

Most importantly, study critically lacks objective circadian markers: claims of a specific hyperuricemia-circadian disruption link are weakened without actigraphy, melatonin, or light exposure data. Observed associations may reflect general metabolic dysfunction rather than circadian-specific effects. Unmeasured confounders like chronotype, shift work, or genetic factors (e.g., circadian clock and melatonin receptors gene variants) could distort results.
The Healthy Eating Index overlooks circadian-relevant factors like meal timing and chrononutrition. Its null association with circadian syndrome likely stems from measurement limitations, not absence of effect (e.g., as shown in several publications by Panda’s and Garaulet’s groups).
Statistical and modeling concerns: Supplemental Figure 2 reveals a non-linear age-circadian syndrome relationship (U-shaped), which may confound or bias the age-adjusted odds ratios provided in Supplemental Figure 3. Consider spline models or stratified analyses to address this.
The weak physical activity-circadian syndrome link contradicts robust evidence of beneficial effects of activity and excericse (including meta-analyses). This likely arises from low sensitivity of the questionnaire; objective actigraphy would be necceassary.
Specify the reference group for race/ethnicity and explain why all other groups show lower odds (e.g., potential confounding by unmeasured cultural/dietary factors?).
Self-reported data on sleep duration, alcohol intake, and diet quality are susceptible to recall and social desirability biases. The PHQ-9, while validated, is a screening tool rather than a diagnostic instrument, risking misclassification of depression.
Excluding a substantial proportion of participants due to incomplete data may introduce bias if such exclusions are non-random (e.g., related to socioeconomic, health, or even directly chonotype-related factors).

Given these limitations, I would suggest toning down circadian claims (e.g., from "associated with circadian disruption" to "associated with circadian-metabolic syndrome indicators"), adding robustness checks, and prioritizing objective circadian phenotyping in extensions. Causal inference and mechanistic insights are also inherently limited. Authors are invited to explicitly discuss the issues above and consider prospective designs in future work.

Author Response

Comment #1:

The weak physical activity-circadian syndrome link contradicts robust evidence of beneficial effects of activity and exercise (including meta-analyses). This likely arises from low sensitivity of the questionnaire; objective actigraphy would be necessary.

Response 1:

We agree with the reviewer's assessment. Our results showed a statistically significant association between physical activity and circadian syndrome. We acknowledge that the effect size appears modest to the evidence found in clinical trials and meta-analysis. We believe that this discrepancy likely stemmed from the use of self-reported questioners in NHANES which are prone to recall bias and often fail to capture light-intensity activity. As suggested, objective actigraphy would provide a more robust measurement of the relationship between physical activity and circadian health. We will add this as a specific limitation in the discussion

Comment #2:

Specify the reference group for race/ethnicity and explain why all other groups show lower odds (e.g., potential confounding by unmeasured cultural/dietary factors?).

Response #2:

Thank you for this insightful comment. In the multivariable logistic regression analysis, Hispanic participants were used as the reference group for race/ethnicity, as indicated in Table 2. Compared with this group, Black and Other/Multiracial participants demonstrated significantly lower odds of Circadian Syndrome (CircS), while no statistically significant difference was observed among White participants.

These findings should be interpreted with caution, as they may reflect residual confounding from unmeasured or imperfectly measured variables. Differences in cultural practices, dietary habits, sleep behaviors, healthcare access, and other social determinants of health across racial and ethnic groups may influence both serum uric acid levels and the components of CircS. Although our analysis adjusted for key demographic, socioeconomic, lifestyle, and clinical factors, including education, poverty income ratio (PIR), diet quality, smoking status, alcohol consumption, physical activity, and chronic kidney disease, NHANES data may not fully capture all relevant environmental and behavioral determinants.

Accordingly, the observed associations likely represent a complex interplay of biological, social, and environmental influences rather than causal relationships. To clarify this point, we have revised the manuscript (Lines 100–103 and 233–238) to explicitly state the reference group and discuss the potential for residual confounding.

Comment #3:

Self-reported data on sleep duration, alcohol intake, and diet quality are susceptible to recall and social desirability biases. The PHQ-9, while validated, is a screening tool rather than a diagnostic instrument, risking misclassification of depression.

Response 3:

We appreciate the reviewer’s insightful comment regarding the limitations of the self-reported data and PHQ-9 nature. We agree that self-reported data is susceptible to recall and social desirability biases which may lead to an underestimation or overestimation of their association with circadian syndrome. In addition, we agree that PHQ-9 is used as a screening tool rather than to make clinical diagnosis, which introduces a risk of misclassification of depression.

In response to this, we have expanded our limitations section to explicitly discuss these factors. We have also emphasized that while those tools are standard in large scale epidemiological surveys like NHANES, they require cautious interpretation of their effect sizes.

Comment #4:

Excluding a substantial proportion of participants due to incomplete data may introduce bias if such exclusions are non-random (e.g., related to socioeconomic, health, or even directly chronotype-related factors).

Response 4:

Thank you for this comment. We agree that exclusion of participants due to incomplete data may introduce bias if the missingness is non-random. To address this, we conducted an additional analysis comparing baseline characteristics between included and excluded participants, which has now been presented in Supplementary Table 1.

This comparison showed that excluded individuals differed in several demographic and clinical characteristics, suggesting that missingness may not be completely at random. We have therefore retained and expanded this statement in the manuscript and explicitly acknowledged the potential for selection bias in the limitations section.

Comment #5:

Given these limitations, I would suggest toning down circadian claims (e.g., from "associated with circadian disruption" to "associated with circadian-metabolic syndrome indicators"), adding robustness checks, and prioritizing objective circadian phenotyping in extensions. Causal inference and mechanistic insights are also limited. Authors are invited to explicitly discuss the issues above and consider prospective designs in future work.

Response #6:

Thank you for this insightful and constructive suggestion. We fully agree with the reviewer and have revised the manuscript accordingly to ensure a more cautious and precise interpretation of our findings.

First, we have moderated circadian-related claims throughout the manuscript, replacing stronger statements implying direct circadian disruption with more accurate terminology such as “circadian–metabolic syndrome indicators.” These revisions have been incorporated into the main text to reflect the proxy nature of the circadian components used in the study (Lines 207–208).

Second, we have expanded the Limitations section to explicitly acknowledge the absence of objective circadian biomarkers, such as melatonin secretion profiles, actigraphy-based sleep assessments, cortisol rhythms, and light exposure data. We also clarified that the cross-sectional design precludes causal inference and limits mechanistic interpretation. Additionally, we addressed the potential influence of unmeasured confounders, including chronotype, shift work, and genetic variation in circadian-related pathways (Lines 312–324).

Third, we emphasized the robustness of our findings by highlighting the multiple validation approaches conducted, including dose–response analyses, sensitivity analyses using alternative hyperuricemia definitions, and validation using a 10% random subsample. These additions strengthen the reliability and internal validity of the study (Lines 326–342).

Finally, we revised the Strengths and Conclusions sections to recommend future prospective and longitudinal studies incorporating objective circadian phenotyping to elucidate causal pathways and underlying biological mechanisms (Lines 426–427).

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I do not have any further comments. In my opinion, all my concerns have been addressed.

Author Response

Dear Editor,

Thank you very much for your kind email and for your time and consideration throughout the review process.

We sincerely appreciate the valuable feedback from you and the peer reviewers, as it helped us improve our manuscript significantly. Thank you for the opportunity to revise our work and for your thoughtful handling of the submission.

Kind regards,

Reviewer 2 Report

Comments and Suggestions for Authors

The authors provided adequate responses to main concerns. I have no further comments at this time.

Author Response

Thank you very much for your kind email and for your time and consideration throughout the review process.