Review Reports - Macronutrient-Based Dietary Patterns and the Risk of Muscle Mass Decline in Middle-Aged and Older Adults: A Prospective Community-Based Cohort Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript examined the association between macronutrient-based dietary patterns and the decline in skeletal muscle mass in a prospective cohort of Korean adults aged 40–69 years. The author reported that HCHO and HP dietary patterns were associated with a higher risk of muscle mass decline in middle-aged and older adults, although average longitudinal skeletal muscle mass trajectories did not differ significantly across dietary patterns. Based on the KoGES prospective cohort, a long follow-up period, repeated dietary assessments, and repeated measurements of muscle mass, the study provides valuable observational evidence on the association between macronutrient intake proportions and muscle mass decline. However, the main concern is also clear. Please find my specific comments below:

The author reported that HP dietary patterns were associated with a higher risk of muscle mass decline in middle-aged and older adults, which is contradictory to the fact that high protein intake increases muscle mass and that dietary guidelines recommend higher protein intake for the elderly. Although discussion and the cited literature has demonstrated that other protein-related factors affect skeletal muscle mass, none of these studies have investigated the independent effects of the proportional intake of dietary protein on skeletal muscle. The discussion is still inefficient.

Dietary quality represents a major confounding factor; adjustment for this variable in the covariate model is therefore recommended.

The HF/HP group had a very small sample size, and the conclusions regarding HP are insufficiently stable.

Title: The phrase “a national prospective study” may be overstated,since the first sentence in Methods describes the study as a KoGES community-based cohort.

Introduction: Paragraph 3, please further clarify the gap linking macronutrient proportion classifications, repeated dietary exposure assessments, and the risk of muscle mass decline.

Dietary assessments: Citation [16,17] should be checked. The FFQ validation citations in the Methods section appear to be inconsistent with the reference list, as the current references 16 and 17 are articles on dietary patterns/protein intake rather than validation studies of the KoGES FFQ.

Dietary assessments: The mutually exclusive classification rules for HCHO, HF, HP, and ND are insufficiently described.For example, if a participant had carbohydrate intake >65% and protein intake >20%, would this participant be classified as HCHO or HP? If a participant had fat intake >30% and protein intake >20%, would this participant be classified as HF or HP? If there was overlap among HCHO, HF, HP, and ND, what was the priority rule? The authors should explain how they were handled. If no such cases occurred, this should also be explicitly stated.

Incidence of muscle mass decline: The outcome was defined as any decline below baseline.The event rate of 73.3% suggests that this outcome may capture normal fluctuation, measurement error, or age-related decline, rather than clinically meaningful muscle loss. You should clearly emphasize that this outcome does not represent sarcopenia. To strengthen the clinical interpretability of the outcome, you may add sensitivity analyses using alternative definitions, such as a ≥3% or ≥5% decline, or a decline exceeding the threshold of measurement error.

Statistical analyses:Medians and interquartile ranges are reported, but t-tests were used. The statistical methods are inconsistent with the descriptive statistics.

Statistical analyses:The authors mentioned “Four separate sensitivity analyses were performed to confirm the robustness of the findings”. What are these sensitivity analyses?

The authorsused Cox models in this study, please report the assessment of proportional hazards assumption.

Results 3.5: In the age-stratified analysis, the HP group aged ≥65 years had only 7/9 events, yet an aHR of 2.28 was still highlighted. This estimate is highly unstable.

The Figure 1 legend is suggestive to include the reason categories for the excluded participants (n = 514).

Figure 2 requires clarification. The authorsmentioned that dietary patterns were treated as time-varying exposures. Standard Kaplan–Meier curves and log-rank tests generally assume that group assignment is defined at baseline. If Figure 2A is based on baseline dietary patterns, this should be explicitly stated in the caption. If it is based on time-updated dietary patterns, the authors should explain the specific method used, as a standard KM curve does not directly accommodate changes in exposure group during follow-up.

Figure 3.The title of fig 3 is the same as fig 2. Please revise the title to accurately reflect the content of Figure 3.

Table 4. The sample sizes are very smallin some subgroup analyses, raising concerns about the statistical reliability of the estimates.

Supplementary table title format error

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents a study on macronutrient-based dietary patterns and the risk of muscle mass decline in middle-aged and older adults.

Overall, the study is of potential interest; however, several aspects require clarification and further development before the manuscript can be considered for publication.

My detailed comments are listed below.

Abstract:

The percentage of energy from macronutrients should be added (line 17).
What was the median time of follow-up? (line 22)
The HR (with 95% CI) should be added (lines 23-26).

Introduction:

What does it mean moderate dietary fat intake in the sentence in lines 60-62?

Materials and methods:

Please check the references for accuracy in the paragraph in lines 93-101.
In lines 103-105 it is written that carbohydrates were multiplied by 4 kcal. Do the authors mean digestible carbohydrates here? If yes, how dietary fiber was calculated (0 kcal or 2 kcal – it should be explained)? If not, if total carbohydrates were multiplied by 4 kcal the calculation is wrong and it must be corrected.
In point 2.4 (covariates) nothing is mentioned about daily energy intake.
In lines 134-137 it is written that “comorbidities… diagnosed between 2001 and 2002 were recorded as baseline conditions.” What about comorbidities diagnosed before the study?
It is written that t-tests were used for continuous variables. The normality of the distribution should be verified, and then the appropriate parametric or non-parametric tests should be used.

Results:

Table 1. Only the data for current smoking are shown. Since this variable comprises three categories, the percentages for all three groups should be reported. The same principle should be applied to other variables with more than two categories.
Table 1. Waist circumference (cm) should be shown separately for men and women.
Table 1. The data about muscle mass decline (%) is not clear. Is this the percentage of participants who experienced a decline in muscle mass, or does it represent the mean loss of muscle mass?
Point 3.3. It would be valuable and informative to report the protein intake per kg body weight for each of the analyzed dietary patterns.
What values are shown in Table 3 in the columns "baseline" and "follow-up 1-8"? Only numerical values are shown, but the corresponding units are missing.
Line 238. In my opinion Table S1 can be shown in the main manuscript. More important results from this table could be mentioned in the point 3.4.
Table 4. It is not clear what is reported in the fourth column. Do these values correspond to the number of events/ total number of participants?

References:

The DOI number should be added wherever possible.

The general comment:

The disproportion between the number of participants in each group with different dietary patterns is very big, so comparing the results between them may be questionable. Have the authors considered a different grouping?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

MATERIAL METOD:
The major methodological concern is the definition of the outcome as any decrease in BIA-derived muscle mass relative to baseline. Given the known biological and technical variability of bioelectrical impedance measurements, this definition may classify trivial measurement fluctuations as incident muscle loss and likely contributes to the exceptionally high event rate of 73.3%. The validity and clinical relevance of the primary outcome should therefore be reconsidered, and sensitivity analyses using clinically meaningful or measurement error-adjusted thresholds are necessary.

Lines 81–86: The dietary groups are extremely imbalanced, with only 48 participants in the HF group and 96 in the HP group compared with 8,055 in the HCHO group. This imbalance raises concerns regarding statistical power, unstable hazard estimates, sparse-data bias, and generalizability. The authors should report power or precision considerations and interpret the HF and HP findings with substantial caution.

Lines 129–130: Physical activity is an important confounder in the relationship between diet and muscle mass, yet its assessment is described only as “evaluated using MET.” The questionnaire, calculation method, units, validity, and whether physical activity was updated over time should be reported.

Lines 117–124 and 153–159: Defining muscle mass decline as any decrease from baseline is a major methodological concern. Small fluctuations in bioelectrical impedance-derived muscle mass may reflect hydration status, biological variability, device error, or regression to the mean rather than clinically meaningful muscle loss. The authors should justify this threshold and consider clinically relevant alternatives, such as a predefined percentage decline, measurement error-based threshold, or consensus-derived low muscle mass criteria.

Lines 112–118: Two different bioelectrical impedance devices were used across the KoGES follow-up period. The authors should clarify at which examinations each device was used, whether the devices were cross-calibrated, and whether device type was adjusted for in the analyses. Without evidence of comparability, apparent longitudinal changes may partly reflect measurement-system differences.
RESULTS:

The authors appear to have presented their findings well. Their use of visuals and the design of their tables are well done. Additionally, the findings are well reported. I did notice some shortcomings in certain presentations, which I have noted in the attached file. Please review the attached file carefully.

Discussion ve Conclusion

The Discussion is generally thoughtful and appropriately cautious; however, it does not sufficiently confront the central inconsistency between the Cox and mixed-effects results. Because the primary event was defined as any BIA-derived decline from baseline, the reported hazard associations may partly reflect measurement variability rather than clinically meaningful muscle loss. The HCHO and especially HP findings should therefore be framed as exploratory, and the absence of differences in longitudinal muscle-mass trajectories should receive greater emphasis in both the Discussion and Conclusion.

271–282 ve 310–320: The inconsistency among the main findings has not been sufficiently addressed

Lines 323–332: An event occurred in 73.3% of the participants. Such a high rate suggests that measurement variability may have been misinterpreted as an event rather than actual clinical muscle loss.

Lines 283–296: The authors partially acknowledge this, but the first sentence still emphasizes the metabolic effects of a high-carbohydrate diet. However, the group consists of older, lower-income individuals with higher rates of hypertension. There is a strong possibility of residual confounding.

Satırlar 297–309: There are only 96 individuals in the HP group. The adjusted HR is 1.20 and the 95% GA is 1.01–1.42. This is based on both a borderline and a small sample size. Furthermore, although the protein content is high, the protein source, quality, distribution, and energy balance are unknown. For this reason, a finding of high protein levels should be interpreted with much greater caution

Lines 271–309:The results are correlational, not causal

Comments for author File: Comments.pdf

Author Response

Answer)

We thank the reviewer for this important comment. We agree that the primary outcome, defined as any observed decline in muscle mass relative to baseline, may capture normal biological fluctuation and technical measurement variability in addition to clinically meaningful muscle loss, which likely contributes to the high event rate observed for this definition. We have clarified in the manuscript that this outcome does not represent consensus-defined sarcopenia.

To strengthen the clinical interpretability of our findings, we conducted three additional sensitivity analyses using more stringent outcome definitions:

A decline of ≥3% from baseline (5,754 events, 60.5% of participants)
A decline of ≥5% from baseline (4,634 events, 48.7% of participants)
A decline exceeding the threshold of measurement error, defined using the least significant change (LSC₉₅) calculated from the standard error of measurement (SEM) of the InBody 720 (reproducibility ICC = 0.98–0.99): 2.32 kg for men and 1.70 kg for women (4,803 events, 50.5% of participants). Because this threshold is defined in absolute kilograms, this analysis used raw muscle mass (kg) rather than height-adjusted muscle mass (kg/m²), and baseline muscle mass was likewise entered as a covariate in kilograms.

(Reference) McLester, C.N.; Nickerson, B.S.; Kliszczewicz, B.M.; McLester, J.R. Reliability and Agreement of Various InBody Body Composition Analyzers as Compared to Dual-Energy X-Ray Absorptiometry in Healthy Men and Women. J Clin Densitom 2020, 23, 443-450, https://doi.org/10.1016/j.jocd.2018.10.008

Lines 81–86: The dietary groups are extremely imbalanced, with only 48 participants in the HF group and 96 in the HP group compared with 8,055 in the HCHO group. This imbalance raises concerns regarding statistical power, unstable hazard estimates, sparse-data bias, and generalizability. The authors should report power or precision considerations and interpret the HF and HP findings with substantial caution.

Answer)

Although the number of participants classified into the HF and HP dietary pattern groups at baseline was relatively small, the primary analysis was performed using an extended Cox proportional hazards model with time-varying dietary patterns based on the counting-process formulation. Consequently, the analysis was conducted using 37,285 person-interval observations, including 168 and 355 person-interval observations in the HF and HP groups, respectively. Although the time-dependent Cox model incorporates accumulated person-time and events during follow-up, the imbalance in exposure categories remains. Therefore, we have added this limitation to the revised manuscript, and the findings for the HF and HP groups should be interpreted with appropriate caution.

Answer)

Thank you for this important comment. Physical activity was assessed at baseline through a face-to-face, interviewer-administered questionnaire used in the KoGES cohort. The questionnaire assessed time spent in occupational and leisure-time activities across five intensity levels. Following the original KoGES scoring algorithm, the assigned values were 0 MET for lying down, 1.5 METs for sedentary activities, 3 METs for light activities, 5 METs for moderate activities, and 7 METs for vigorous activities. Total physical activity was calculated by summing the products of the assigned MET value and the reported time spent at each activity level and was expressed as MET-minutes per week. This measure therefore represents total daily physical activity rather than leisure-time exercise alone. We have added the content with reference in the Methods section.

We used the baseline physical activity value as a fixed covariate and did not update it during follow-up, because the physical activity questionnaire was modified after the baseline examination, preventing the consistent calculation of the same MET measure across all follow-up waves. Although the questionnaire was administered by trained personnel according to the standardized KoGES protocol, we did not identify a separate criterion-validation study comparing this specific five-level total-physical-activity score with accelerometry or another objective measure.

Lines 117–124 and 153–159: Defining muscle mass decline as any decrease from baseline is a major methodological concern. Small fluctuations in bioelectrical impedance-derived muscle mass may reflect hydration status, biological variability, device error, or regression to the mean rather than clinically meaningful muscle loss. The authors should justify this threshold and consider clinically relevant alternatives, such as a predefined percentage decline, measurement error-based threshold, or consensus-derived low muscle mass criteria.

Answer)

We agree with the reviewer that small fluctuations in BIA-derived muscle mass may reflect hydration status, biological variability, device error, or regression to the mean rather than true muscle loss. As described in our response to the related comment above regarding the high event rate of the primary outcome, we have addressed this concern by conducting sensitivity analyses using three clinically relevant alternative thresholds—a ≥3% decline, a ≥5% decline, and a decline exceeding a measurement error-based threshold derived from the InBody 720's reproducibility (ICC = 0.98–0.99). Regarding consensus-derived low muscle mass criteria, we were unable to apply these directly because appendicular lean mass and muscle strength were not available in the KoGES dataset; this is noted as a limitation in the Discussion.

Answer)

We thank the Reviewer for this careful reading.

The original Methods stated that skeletal muscle mass was assessed using the "InBody 3.0 and 720". However, all body composition measurements analyzed in the KoGES were obtained with a single device, the InBody 3.0 (Biospace Co., Seoul, South Korea). No device transition occurred within the analytic window, and all within-person comparisons were therefore made on the same measurement system. We have revised the content in the Methods sections and added the related references.

RESULTS:

Answer)

We thank the reviewer for the positive assessment of our data presentation and for the detailed comments provided in the attached file. We have carefully reviewed each point noted in the attachment and revised the manuscript accordingly. We have attached the file with answers below.

Discussion ve Conclusion

Answer)

We thank the reviewer for this comment. We have revised the Discussion to state explicitly that the primary outcome's high event rate likely reflects, in part, measurement variability rather than clinically meaningful muscle loss, and that the HCHO and HP findings were attenuated and no longer significant under more stringent thresholds (≥3%, ≥5%, least significant change). We have stated these findings as exploratory rather than confirmatory.

271–282 ve 310–320: The inconsistency among the main findings has not been sufficiently addressed

Answer)

We thank the reviewer for this comment. As described in our response above, we have substantially revised the Discussion to confront this inconsistency directly, explicitly stating that the Cox and linear mixed-effects models assess different aspects of muscle mass change, that the HCHO and HP associations were attenuated under more stringent outcome thresholds, and that these findings should accordingly be interpreted as exploratory rather than confirmatory.

Answer)

We agree with the reviewer. We have acknowledged in the Discussion that the primary outcome's high event rate (73.3%) likely reflects, in part, biological fluctuation and measurement variability inherent in BIA-based body-composition assessment rather than clinically meaningful muscle loss. As noted above, we addressed this directly through sensitivity analyses using more stringent thresholds (≥3%, ≥5%, and a decline exceeding the least significant change), under which the HCHO and HP associations were attenuated and no longer statistically significant (Table S1).

Lines 283–296: The authors partially acknowledge this, but the first sentence still emphasizes the metabolic effects of a high-carbohydrate diet. However, the group consists of older, lower-income individuals with higher rates of hypertension. There is a strong possibility of residual confounding.

Answer)

We thank the reviewer for this comment. We agree that the sentence overemphasized the metabolic effects of high carbohydrate intake without adequately accounting for the sociodemographic and clinical profile of the HCHO group. We have revised this paragraph to lead with these characteristics—the HCHO group was older, more likely to have lower household income, and had a higher prevalence of hypertension—and have removed the unsupported statement regarding metabolic effects, instead explicitly noting that residual confounding by dietary quality, carbohydrate sources, insulin sensitivity, inflammation, or socioeconomic characteristics cannot be excluded

Lines 297–309: There are only 96 individuals in the HP group. The adjusted HR is 1.20 and the 95% GA is 1.01–1.42. This is based on both a borderline and a small sample size. Furthermore, although the protein content is high, the protein source, quality, distribution, and energy balance are unknown. For this reason, a finding of high protein levels should be interpreted with much greater caution

Answer)

We thank the reviewer for this comment. We agree that the HP finding (aHR 1.20; 95% CI, 1.01–1.42) is based on a small sample size (n = 96) and a confidence interval bordering the null, and that this finding should be interpreted with substantial caution. We have addressed this in the discussion. he HP association may reflect these unmeasured characteristics rather than an adverse effect of protein intake per se. We have also explicitly stated that, given the small size of the HP group, this finding requires confirmation in larger studies.

Lines 271–309: The results are correlational, not causal

Answer)

We agree with the reviewer that our findings are correlational rather than causal, given the observational design of the study. To avoid overstating causality, we have revised the Discussion to replace causal expressions (e.g., "effect," "determinant," "explain") with terms denoting association.

Author Response File: Author Response.pdf