Dietary plant-to-animal protein ratio and risk of cardiovascular disease in 3 prospective cohorts

Table of Contents

Overall Summary

Overview

This study found a correlation between higher plant-to-animal protein ratios and reduced CVD (HR 0.81) and CAD (HR 0.73) risk in three large US cohorts over 30 years. No association was found with stroke. Substituting red/processed meat with plant protein, especially nuts, showed the greatest benefit. The association was stronger with higher protein density. These findings should be interpreted cautiously due to the observational design and reliance on FFQs.

Key Points

P:A Ratio and CVD/CAD Risk (written-content)
Higher plant-to-animal protein ratios correlated with lower CVD and CAD risk, but not stroke. The highest ratio showed a 19% lower CVD risk and 27% lower CAD risk compared to the lowest.
Section: Results
Correlation vs. Causation (written-content)
The observational design limits causal conclusions. The association may be due to confounding factors (e.g., healthier lifestyles in those with higher P:A ratios).
Section: Discussion
Prospective Cohort Design (written-content)
The use of three large prospective cohorts with long follow-up minimizes recall bias and allows for examination of long-term dietary patterns.
Section: Methods
FFQ Limitations (written-content)
Reliance on FFQs for dietary assessment introduces potential measurement error and recall bias, affecting the accuracy of P:A ratio estimations.
Section: Methods
Substitution Analysis (written-content)
Substituting red and processed meat with nuts showed the most significant CVD/CAD benefit, offering practical dietary guidance.
Section: Results
Practical Dietary Advice (written-content)
Lack of specific dietary recommendations limits practical applicability. The discussion should provide clear guidance on achieving beneficial P:A ratios.
Section: Discussion
Dose-Response Visualization (graphical-figure)
Figure 2 clearly illustrates the non-linear dose-response relationship, showing diminishing returns at higher P:A ratios.
Section: Results
Interaction Visualization (graphical-figure)
Figure 3 lacks clear visualization of the interaction between P:A ratio and protein density, hindering interpretation of their joint effect.
Section: Results

Conclusion

This analysis of prospective cohort studies suggests a correlation between higher plant-to-animal protein ratios and reduced CVD/CAD risk, but not stroke. While the large sample size, long follow-up, and sensitivity analyses are strengths, the observational nature limits causal inference. Residual confounding and reliance on FFQs are key limitations. Substituting red/processed meat with plant protein, especially nuts, appears most beneficial. However, translating these findings into practical dietary advice requires caution. Further research, including randomized controlled trials, is needed to confirm causality and establish optimal P:A ratios for diverse populations. While the findings support existing recommendations to favor plant protein, individual dietary needs and preferences should be considered.

Section Analysis

Abstract

Key Aspects

Strengths

Suggestions for Improvement

Introduction

Key Aspects

Strengths

Suggestions for Improvement

Methods

Key Aspects

Strengths

Suggestions for Improvement

Results

Key Aspects

Strengths

Suggestions for Improvement

Non-Text Elements

FIGURE 1. Trends in the median plant-to-animal protein ratio in the 3...
Full Caption

FIGURE 1. Trends in the median plant-to-animal protein ratio in the 3 prospective cohort studies. Secular population trends of plant-to-animal protein ratio indicate changing from ~1:3 to ~1:2 across all cohorts over time. NHS, Nurses' Health Study; HPFS, Health Professionals Follow-up Study.

First Reference in Text
The median plant-to-animal protein ratio increased from ~0.36 (1:3) to ~0.50 (1:2) over follow-up (Figure 1).
Description
  • Content and Organization: The graph displays the change in the median plant-to-animal protein ratio over time in three prospective cohort studies: NHS, NHSII, and HPFS. The x-axis represents time (year), spanning from 1982 to 2017. The y-axis represents the plant-to-animal protein ratio, ranging from roughly 0.1 to 0.7. Each cohort is represented by a different colored line (NHS-blue, NHSII -green, HPFS- orange), showing the median ratio at different time points during the follow-up period.
  • Technical Terms: A prospective cohort study is a type of longitudinal research where a group of individuals (the cohort) is followed over time to observe the incidence of a particular outcome (like a disease) and its potential risk factors. The plant-to-animal protein ratio is calculated by dividing the amount of protein consumed from plant sources by the amount consumed from animal sources.
  • Ratios and Statistics: The median is the middle value in a set of numbers when arranged in ascending or descending order. It is used to represent the central tendency of the data. The ratio of ~1:3 signifies that for every one unit of plant protein, three units of animal protein were consumed, and ~1:2 signifies that for every one unit of plant protein, two units of animal protein were consumed. The ratios are calculated based on the energy contribution of these proteins to the diet.
Scientific Validity
  • Use of Statistics: The presentation of median values for the ratio is appropriate for summarizing the central tendency in each cohort. However, providing a measure of variability (e.g., interquartile range) would enhance the interpretation and understanding of the distribution of the ratio within each cohort.
  • Data Collection and Limitations: The secular trend observed in the figure is valuable for demonstrating the dietary shift toward increased plant protein consumption over time. However, the study should clarify how the protein intake data was collected (e.g., food frequency questionnaires) and address potential limitations related to dietary assessment, such as recall bias or measurement error. This is particularly important since dietary assessment is a key aspect of the study.
  • Specificity of Protein Sources: The rationale for focusing on the median plant-to-animal protein ratio is sound. However, exploring the distribution of protein sources within both plant and animal categories would provide additional insights into the types of proteins driving the observed trend. For example, are legumes and nuts replacing red and processed meat or are refined grains and starchy vegetables contributing more to the change?
Communication
  • Clarity and Visual Appeal: The figure effectively communicates the increasing trend of the plant-to-animal protein ratio over time. The use of different colors for different cohorts aids in visual differentiation. The axes are clearly labeled, and the caption provides context.
  • Context and Interpretation: While the figure clearly shows an increasing trend, it could be enhanced by adding a reference line indicating the recommended or optimal ratio, if one exists. This would provide additional context and strengthen the visual message.
TABLE 1 Baseline characteristics of participants according to select deciles of...
Full Caption

TABLE 1 Baseline characteristics of participants according to select deciles of the plant-to-animal protein ratio.

First Reference in Text
Table 1 shows the baseline characteristics by cohort according to
Description
  • Content and Organization: Table 1 presents the baseline characteristics of participants from three prospective cohort studies (NHS, NHSII, and HPFS) categorized by deciles of their plant-to-animal protein ratio. Deciles divide the participants into ten groups based on their protein ratio, from the lowest (Decile 1) to the highest (Decile 10). Each decile represents 10% of the participants. The characteristics listed include demographic information (age, race/ethnicity), lifestyle factors (physical activity, smoking, alcohol use), health conditions (diabetes, hypertension, hypercholesterolemia, family history of myocardial infarction), dietary habits (total energy intake, macronutrient intake, specific food consumption like dairy, red meat, nuts etc., and micronutrient intake), and use of medications (multivitamin and aspirin). The data is presented as mean (standard deviation) for continuous variables and as percentages for categorical variables. A plant-to-animal protein ratio is calculated by dividing the amount of protein a person gets from plants by the amount they get from animals.
  • Units and Abbreviations: MET-h/wk refers to Metabolic Equivalent Task-hours per week, a measure of physical activity intensity. AHEI stands for Alternate Healthy Eating Index, a scoring system used to assess diet quality. EPA and DHA are omega-3 fatty acids. The "servings/d" for food groups represents the average number of servings consumed per day.
Scientific Validity
  • Justification and Methodology: Presenting baseline characteristics stratified by the exposure variable (plant-to-animal protein ratio) is crucial for assessing potential confounding and identifying any systematic differences between groups. This information helps to support the validity of the subsequent analyses examining the relationship between protein ratio and cardiovascular disease outcomes. The use of deciles is a reasonable approach for categorizing the exposure, allowing for the examination of trends across the range of protein intake ratios. It provides more granular information than simply splitting into two or three groups.
  • Missing Data and Confounding: The authors should clarify how missing data were handled for the baseline characteristics. Were participants with missing values excluded from the analysis? If so, could this exclusion introduce bias? Imputation methods, if used, should be described in detail. It's also important to ensure the baseline characteristics are balanced across the deciles, after adjusting for potential confounders in the main analysis.
  • Relevance of Included Variables: The table presents numerous characteristics, some of which may not be directly relevant to the primary research question. The rationale for including each characteristic should be clearly stated, and the authors should consider prioritizing the most relevant variables in the table to enhance clarity and focus.
Communication
  • Overall Presentation and Clarity: The table is generally well-organized and presents a comprehensive set of baseline characteristics. The use of standard deviations (SD) provides a measure of variability for continuous variables. The inclusion of both percentages and mean values where appropriate aids in understanding the distribution of characteristics within each decile.
  • Missing Ratio Values: While the table provides deciles of the plant-to-animal protein ratio, it doesn't explicitly state the actual ratio values for each decile. Including these values would make the table more informative and easier to interpret. For instance, instead of just saying "Decile 1", it would be beneficial to have "Decile 1 (Ratio: 0.21)" or a similar format.
  • Visual Presentation and Structure: The table could benefit from some visual enhancements, such as using boldface for the column headers or adding separators between the different cohorts to improve readability. Additionally, considering the length of the table, splitting it into smaller, more focused tables might improve clarity.
TABLE 2 Pooled associations of deciles of the plant-to-animal protein ratio...
Full Caption

TABLE 2 Pooled associations of deciles of the plant-to-animal protein ratio with the cardiovascular disease outcomes.

First Reference in Text
Table 2 shows the results of the pooled findings for deciles of the plant-to-animal protein ratio for total CVD, CAD, and stroke; findings by each cohort are provided in Supplemental Tables 4 6.
Description
  • Content and Organization: Table 2 displays the association between deciles of plant-to-animal protein ratio and the risk of cardiovascular disease outcomes (total CVD, coronary artery disease (CAD), and stroke). The table presents hazard ratios (HRs) which indicate the relative risk of developing the outcome in each decile compared to a reference group (Decile 1 - lowest plant-to-animal protein ratio). A HR less than 1 indicates a lower risk, while a HR greater than 1 indicates a higher risk. The table includes results for three different models: age-adjusted (Model 1) and two multivariable-adjusted models (Models 2 and 3) that control for various confounding factors like demographic factors, lifestyle factors, and other health conditions. The table also includes the number of cases and person-years for each outcome. The Plant-to-animal protein ratio is calculated by dividing the energy from plant protein by energy from animal protein. Deciles split the sample population into ten equal groups, from lowest to highest, based on their plant-to-animal protein ratio.
  • Statistical Measures: A hazard ratio (HR) represents the relative risk of an event (in this case, a cardiovascular disease outcome) occurring in one group compared to another group over a specific period of time. A 95% confidence interval (CI) provides a range of values within which the true HR is likely to fall. The p-trend indicates the statistical significance of the trend in HRs across the deciles. Person-years is the sum of the time each participant contributed to the study before experiencing the outcome or dropping out.
Scientific Validity
  • Pooling of Data: Pooling the data from three cohorts increases the statistical power of the analysis and improves the generalizability of the findings. However, the authors should justify the decision to pool the data and discuss any potential heterogeneity between cohorts, such as differences in baseline characteristics or follow-up periods. This could be done by conducting and reporting the results of a heterogeneity test.
  • Model Specifications: The choice of multivariable adjustment models (Models 2 and 3) seems appropriate for controlling for potential confounders. However, the authors should clearly specify all the covariates included in each model, ideally within the table or its footnotes. This would improve transparency and reproducibility.
  • Decile Analysis Limitations: While examining deciles allows for exploring trends, it might not be the most informative approach for identifying a potential threshold effect or optimal range for the plant-to-animal protein ratio. The authors should consider alternative analyses, such as using splines or categorizing the ratio based on clinically or biologically relevant cutoffs.
  • Multiple Comparisons: The lack of adjustment for multiple comparisons could inflate the risk of type I error. While the authors acknowledge this limitation in the methods, they should consider applying a correction method (e.g., Bonferroni, Benjamini-Hochberg) to the p-values to address this issue.
Communication
  • Clarity of Results: The table effectively communicates the main findings of the pooled analysis. The use of hazard ratios (HRs) and 95% confidence intervals (CIs) allows for clear comparisons between deciles. The inclusion of a p-value for the trend further strengthens the presentation of the results.
  • Reference Group Definition: While the table presents HRs for each decile, the reference group (Decile 1) is somewhat unclear. Explicitly stating the actual plant-to-animal protein ratio for Decile 1 (and ideally for all deciles) in the table would enhance clarity and interpretation.
  • Model Description: The table could be improved by including a brief explanation of the multivariable models used in the analysis (Models 1-3) within the table or caption. This would help readers understand the adjustments made and interpret the HRs more accurately.
  • Ratio Presentation: The presentation of the ratio in two formats (decimal and ratio) could be confusing. Maintaining consistency in presenting the ratio (e.g., only as a decimal or only as a ratio) would improve readability. In this case, it would be better to present it as a ratio to improve understanding.
FIGURE 2. Dose-response relationship of the plant-to-animal protein ratio with...
Full Caption

FIGURE 2. Dose-response relationship of the plant-to-animal protein ratio with risk of cardiovascular outcomes. (A) Total cardiovascular disease, (B) coronary artery disease, and (C) stroke. Analysis was conducted after combining all 3 cohorts. Multivariable model was adjusted for age, race, smoking, menopausal status and postmenopausal hormone use, oral contraceptive use, multivitamin use, regular aspirin use, physical activity, family history of myocardial infarction, family history of diabetes, marital status, BMI, alcohol intake, total energy intake, modified AHEI score, socioeconomic status, baseline hypercholesterolemia, hypertension, and diabetes. Dose-response relationships were determined using restricted cubic splines. Solid lines represent hazard ratios and dotted lines represent 95% confidence intervals.

First Reference in Text
Figure 2 shows the pooled dose-response relationships of the plant-to-animal protein ratio with risk of cardiovascular outcomes.
Description
  • Content and Organization: Figure 2 illustrates the dose-response relationship between the plant-to-animal protein ratio and the risk of three cardiovascular outcomes: total cardiovascular disease (CVD), coronary artery disease (CAD), and stroke. Each panel (A, B, and C) represents a separate outcome. The x-axis represents the plant-to-animal protein ratio, and the y-axis represents the hazard ratio (HR) for each outcome. The solid lines depict the fitted dose-response curves generated using restricted cubic splines, a statistical method used to model non-linear relationships. The dotted lines represent the 95% confidence intervals around the HR estimates. P-values are provided for curvature (non-linearity) and linearity of the relationships. A plant-to-animal protein ratio means the energy you get from plant protein divided by the energy you get from animal protein. For example, if you get 10% of your energy from plant protein and 20% of your energy from animal protein, the plant-to-animal protein ratio is 10/20 = 0.5. The hazard ratio shows how much more or less likely someone is to have the event compared to people in the reference group (lowest protein ratio).
  • Technical Terms: Restricted cubic splines are a flexible statistical method used to model non-linear relationships between variables. They involve dividing the range of the predictor variable (here, plant-to-animal protein ratio) into sections (knots) and fitting separate cubic polynomials within each section, ensuring smoothness at the knots. The hazard ratio (HR) is a measure of relative risk, indicating the likelihood of developing an event (like CVD, CAD, or stroke) in one group compared to another group (usually a reference group with the lowest exposure). A 95% confidence interval (CI) provides a range within which we are 95% confident that the true HR lies.
Scientific Validity
  • Spline Methodology: Using restricted cubic splines is a valid approach for exploring and visualizing non-linear dose-response relationships. The authors appropriately provide p-values for linearity and curvature to assess the nature of the relationships. However, the choice of knots for the splines should be justified. Did the authors use pre-specified knots based on prior knowledge, or were they determined data-dependently? Sensitivity analyses using different knot locations would strengthen the robustness of the findings.
  • Cohort Heterogeneity: Pooling data from three cohorts increases statistical power. However, potential heterogeneity among cohorts could affect the dose-response relationship. Did the authors assess for interactions between the protein ratio and cohort membership? Stratified analyses or inclusion of interaction terms in the model would address potential heterogeneity and provide more nuanced results. Additionally, more details about the multivariable adjustments would help.
  • Model Details and Validation: The caption mentions a multivariable model but lacks details on model selection, diagnostic checks, and assessment of model fit. Providing these details would enhance the transparency and rigor of the analysis. Specifically, what criteria were used for selecting the variables in the model? Were any interactions between covariates considered?
Communication
  • Overall clarity and presentation: The figure clearly presents the dose-response relationships for the three cardiovascular outcomes. The use of separate panels for each outcome facilitates comparison and interpretation. The x and y axes are labeled clearly, and the caption provides sufficient context. The visual presentation of the splines, confidence intervals, and p-values for curvature and linearity effectively communicates the key findings.
  • Highlighting Key Findings: While the figure effectively presents the trends, visually highlighting key points or areas of interest (e.g., points where the curves plateau or change direction) could enhance interpretation. Annotations or callouts on the graph could guide the reader's attention to these important aspects of the dose-response relationship.
  • Clarity of X-axis Units: The figure could benefit from a clearer explanation of the units used for the plant-to-animal protein ratio on the x-axis. While the caption mentions it is a ratio, providing example values or specifying if it's based on protein intake in grams or percentage of total calories would enhance understanding.
FIGURE 3. Joint associations of the plant-to-animal protein ratio and protein...
Full Caption

FIGURE 3. Joint associations of the plant-to-animal protein ratio and protein density with cardiovascular outcomes. Hazard ratios (HRs) were determined using Cox regression models. Analysis was conducted after combining all 3 cohorts. The reference group was the lower ratios and lower protein density group. Multivariable model was adjusted for the same covariates as model 3 in Table 2. No corrections for multiple tests were applied.

First Reference in Text
In the joint analysis by protein density (Figure 3), the inverse as-sociations with CVD (HR: 0.72; 95% CI: 0.64, 0.82) and CAD (HR: 0.64; 95% CI: 0.55, 0.75) were stronger when higher ratios (>0.50)
Description
  • Content and Organization: Figure 3 presents the joint association of plant-to-animal protein ratio and protein density with cardiovascular outcomes (total CVD, coronary artery disease [CAD], and stroke). It shows how the combined effect of these two factors influences the risk of each outcome. The figure is divided into three panels, one for each outcome. Within each panel, the x-axis represents combinations of plant-to-animal protein ratio categories (lower, middle, higher) and protein density categories (lower, medium, higher). The y-axis represents the hazard ratio (HR), a measure of relative risk. Each bar represents a specific combination of protein ratio and density, with its height corresponding to the HR. Error bars indicate 95% confidence intervals (CIs). The reference group (lowest risk) is the combination of lower plant-to-animal protein ratio and lower protein density. A plant-to-animal protein ratio is the energy from plant protein divided by the energy from animal protein. Protein density refers to the percentage of total daily calories that come from protein.
  • Statistical Measures: A hazard ratio (HR) represents the relative risk of experiencing the outcome in a specific group compared to the reference group. An HR less than 1 indicates a reduced risk, an HR greater than 1 indicates an increased risk, and an HR of 1 indicates no difference in risk. The 95% confidence interval (CI) provides a range of values within which the true HR is likely to fall 95% of the time. If the CI doesn't cross 1, the result is considered statistically significant.
Scientific Validity
  • Interaction Assessment: Investigating the joint effects of protein ratio and density is crucial for understanding their combined influence on cardiovascular risk. The use of Cox regression models is appropriate for analyzing time-to-event data. However, the authors should clearly state the interaction term used in the Cox model and report its p-value. This is essential for determining whether the combined effect is statistically significant and differs from the individual effects of protein ratio and density.
  • Covariate Specification: The caption mentions adjusting for covariates from Model 3 in Table 2. However, it's important to explicitly list these covariates in the figure caption or a supplementary table. This enhances transparency and allows for a better understanding of potential confounding factors considered.
  • Multiple Comparisons: The figure caption notes that no corrections for multiple tests were applied. While this is acknowledged, the authors should explain their rationale for this decision. Given the multiple comparisons made in the figure, a correction method (e.g., Bonferroni, Benjamini-Hochberg) should be considered to minimize the risk of false positives. The rationale for not applying a correction should be explicitly stated and justified.
  • Ratio Categorization: The cutoff of >0.50 for defining "higher ratios" needs justification. Is this cutoff based on prior research, clinical guidelines, or a data-driven approach? Providing a clear rationale for this categorization would strengthen the analysis.
Communication
  • Clarity of Presentation: The figure is generally clear in presenting the joint associations of protein ratio and density with cardiovascular outcomes. The use of different colors and groupings helps differentiate the combined categories. Displaying HRs with confidence intervals provides a clear comparison across the groups.
  • Y-axis Label Specificity: The y-axis label "Hazard Ratio (HR)" could be improved by specifying the outcome to which the HR refers. Since there are three separate panels for different outcomes, adding the specific outcome (CVD, CAD, or Stroke) to each y-axis label would improve clarity.
  • Visualizing Interaction: While the figure shows the joint effects, it doesn't visually represent the interaction itself. Including a panel or separate figure demonstrating the interaction between protein ratio and density for each outcome would enhance understanding and highlight the interaction's significance.
  • Protein Density Values: Consider adding the exact protein density percentages for each category (low, medium, high) in the figure legend or labels. This would avoid reliance on the text and make the figure more self-explanatory.

Discussion

Key Aspects

Strengths

Suggestions for Improvement

Non-Text Elements

FIGURE 4. Subgroup analyses for cardiovascular outcomes for 1-SD increase in...
Full Caption

FIGURE 4. Subgroup analyses for cardiovascular outcomes for 1-SD increase in the plant-to-animal protein ratio. Hazard ratios (HRs) were determined using Cox regression models. Pooled multivariable-adjusted HRs (indicated by the blue dots) and 95% CIs (indicated by black horizontal lines) of all 3 cohorts are shown. The same covariates were adjusted as model 3 in Table 2, with the exception of not adjusting for the categorical covariate when it was used as a strata. No corrections for multiple tests were applied. AHEI, Alternative Healthy Eating Index; CAD, coronary artery disease; CI, confidence interval; CVD, car-diovascular disease.

First Reference in Text
The associations were consistent across most subgroups and risk of cardiovascular outcomes, however, the association between the plant-to-animal protein ratio and CVD risk was significantly stronger among participants with hypercholesterolemia (P interaction < 0.01) and ever smokers (P interaction < 0.01) (Figure 4).
Description
  • Content and Organization: Figure 4 displays the results of subgroup analyses examining the association between a 1-standard deviation (SD) increase in the plant-to-animal protein ratio and the risk of cardiovascular outcomes (CVD, CAD, and stroke). The figure uses a forest plot format, where each row represents a different subgroup. The blue dots represent the pooled, multivariable-adjusted hazard ratios (HRs) for each subgroup, and the black horizontal lines represent the corresponding 95% confidence intervals (CIs). The vertical line at HR = 1 represents the null hypothesis (no association). If a CI crosses this line, the result is not statistically significant. P-values for interaction are provided for each subgroup, indicating whether the effect of the plant-to-animal protein ratio differs significantly across the subgroup categories. Plant-to-animal protein ratio is the energy you get from plant protein divided by the energy you get from animal protein. An increase of 1-SD in the plant-to-animal protein ratio represents a change in the ratio equal to one standard deviation of that ratio in the study population. A standard deviation represents how spread out the ratio is across the study population.
  • Statistical Measures: A hazard ratio (HR) is a measure of relative risk, indicating the likelihood of developing the outcome in one group compared to another. A 95% confidence interval (CI) provides a range of values within which we are 95% confident that the true HR lies. A p-value for interaction tests whether the effect of the plant-to-animal protein ratio is different across subgroups. A p-value less than 0.05 usually suggests that the interaction is statistically significant.
Scientific Validity
  • Subgroup Selection Rationale: Conducting subgroup analyses is essential for assessing the consistency of the main findings across different population characteristics. However, the choice of subgroups should be based on a priori hypotheses and justified based on biological or clinical plausibility. The authors should provide a clear rationale for the selected subgroups and explain why these specific factors were considered potential effect modifiers. Are the subgroups selected based on some pre-defined hypotheses or post hoc analysis?
  • Handling of Subgroup Variables in the Model: The caption mentions using the same covariates as Model 3 in Table 2. However, it also states that the categorical covariate being investigated was not included in the model when used as a strata. This needs clarification. If the subgroup variable is part of the multivariable adjustment in the main analysis, it should be removed from the model when performing subgroup analysis for that specific variable to avoid over-adjustment. The authors should explicitly state how they handled this and ensure the analysis is performed correctly.
  • Heterogeneity Across Cohorts Within Subgroups: The figure presents pooled HRs, which assumes no substantial heterogeneity among cohorts. However, the authors should explicitly state whether they tested for heterogeneity of effects across cohorts within each subgroup. If significant heterogeneity exists, presenting pooled estimates may not be appropriate, and stratified results should be presented or a random effects model used.
  • Multiple Comparisons: The authors should address the issue of multiple comparisons. While they acknowledge that no corrections were applied, they need to discuss the implications of this decision and the potential for inflated type I error rates. Providing adjusted p-values, or at least discussing the potential impact of multiple comparisons on the interpretation of results, would strengthen the analysis.
Communication
  • Clear Visual Representation: The forest plot effectively visualizes the results of the subgroup analyses, making it easy to compare the effects of a 1-SD increase in plant-to-animal protein ratio across different subgroups. The consistent use of blue dots for HRs and black lines for CIs facilitates quick interpretation. The inclusion of p-values for interaction directly on the plot enhances understanding of the significance of subgroup differences.
  • Direction of Interaction: While the figure presents p-values for interaction, clarifying the direction of these interactions would be beneficial. For instance, indicating whether the association is stronger or weaker in specific subgroups (e.g., using annotations or different colors) would enhance interpretation.
  • Subgroup Ordering: Consider reordering the subgroups based on the magnitude or significance of the interaction effect. This would help highlight the most relevant findings and improve the visual flow of information.
  • Defining Median Cut-offs: The figure caption should clearly define what constitutes "below" and "above median" for continuous variables like physical activity, BMI, age, and AHEI. Providing these cut-off values directly in the caption would enhance interpretation.
FIGURE 5. Substitution analyses for 3% energy replacement of animal for plant...
Full Caption

FIGURE 5. Substitution analyses for 3% energy replacement of animal for plant protein and top animal protein sources for plant protein sources and the cardiovascular outcomes. Pooled multivariable-adjusted HRs (indicated by the colored dots) and 95% CIs (indicated by the horizontal lines) of all 3 cohorts are shown. HR (95% CI) for each cardiovascular outcome associated with replacing 3% energy from animal protein with plant protein and 1 daily serving of poultry, red meat, and dairy replaced with refined grains and potatoes, whole grains, nuts, and legumes are shown. The Cox proportional hazards models included all protein foods simultaneously and models were stratified by cohort, age in moths, and follow-up period and adjusted for race, smoking, menopausal status and postmenopausal hormone use, oral contraceptive use, multivitamin use, regular aspirin use, physical activity, family history of myocardial infarction, family history of diabetes, marital status, BMI, alcohol intake, total energy intake, modified AHEI score, socioeconomic status, and baseline hypercholesterolemia, hypertension and diabetes. For the energy substitution, the leave-one-out approach to examine the isocaloric substitution of animal for plant protein and included protein density, carbohydrate density, and fat density in the models instead of the modified AHEI. For each substitution of 1 food item for another, we exponentiated the difference between the β-coefficients of the 2 foods to estimate the HR, and we used the variances and covariance of the 2 food items to estimate the 95% CI. No corrections for multiple tests were applied. Red meat included unprocessed and processed sources. AHEI, Alternative Healthy Eating Index; CI, confidence interval.

First Reference in Text
Replacing 3% energy from animal protein with the same amount of energy from plant protein was associated with an 18% and 24% lower risk of CVD and CAD, respectively, with no association for stroke (Figure 5).
Description
  • Content and Organization: Figure 5 presents the results of substitution analyses, which estimate the change in cardiovascular disease risk associated with replacing specific animal protein sources with plant protein sources. It shows two types of substitutions: 1) Replacing 3% of total energy from animal protein with plant protein (isocaloric substitution); 2) Replacing one serving per day of a specific animal protein food (poultry, red meat, or dairy) with a serving of a plant protein food (refined grains and potatoes, whole grains, nuts, or legumes). The results are presented as hazard ratios (HRs) and their 95% confidence intervals (CIs) for each substitution and outcome (CVD, CAD, and stroke). A plant-to-animal protein ratio is the energy from plant protein divided by the energy from animal protein. An isocaloric substitution means that the same number of calories are maintained when the proteins are substituted.
  • Statistical Measures and Methods: A hazard ratio (HR) less than 1 indicates a reduced risk of the outcome when the substitution is made, while an HR greater than 1 indicates an increased risk. The 95% confidence interval (CI) provides a range of plausible values for the true HR. A leave-one-out analysis is a statistical method used in substitution analysis where one food item is removed, and another is added to the diet, while keeping total energy intake constant. β-coefficients (beta coefficients) are values from a statistical model (here, Cox proportional hazards model) representing the association between a predictor variable (food item) and the outcome. Exponentiating the difference between β-coefficients provides the HR for the substitution. Variances and covariances are measures of statistical variability and association, respectively, used to calculate the CI of the HR.
Scientific Validity
  • Isocaloric Substitution: Substitution analyses are valuable for assessing the potential impact of dietary changes on health outcomes. However, it's crucial that these analyses maintain isocaloric conditions to accurately estimate the effects of replacing one food with another. The authors should carefully consider the potential impact of residual energy and ensure that the substitutions are truly isocaloric, especially when using servings/d which don't perfectly account for energy differences between foods.
  • Justification for Substitution Amounts: The choice of 3% energy replacement and one serving/d substitutions should be justified. Are these values based on realistic dietary changes or existing dietary guidelines? Providing a rationale for these specific amounts would strengthen the validity and relevance of the findings.
  • Confounding Assessment: The caption mentions numerous covariates included in the Cox proportional hazards model. However, it's important to justify the inclusion of each covariate and explain how potential confounding was addressed. Were there any covariates strongly correlated with both protein intake and cardiovascular outcomes? Providing a clear description of the confounding assessment and control would strengthen the analysis.
  • Multiple Comparisons: The authors acknowledge that no corrections for multiple comparisons were applied. Given the numerous substitutions and outcomes examined, there is a high risk of type I error. The authors should consider adjusting the p-values for multiple comparisons or at least discuss the implications of not doing so.
  • Selection of Top Animal Protein Sources: The caption provides limited information about the "top animal protein sources." Specifying the criteria used for selecting these top sources (e.g., most commonly consumed, highest contribution to total animal protein intake) would enhance transparency.
Communication
  • Overall Clarity and Organization: The figure effectively communicates the results of the substitution analyses. The use of a forest plot format with colored dots for HRs and horizontal lines for CIs allows for clear comparison of the effects of different substitutions. The arrangement of substitutions by animal protein source (poultry, red meat, dairy) improves clarity and facilitates interpretation.
  • Explanation of Substitution Analysis: While the figure caption provides a detailed description of the statistical methods, it could benefit from a concise explanation of what a substitution analysis entails. A brief explanation in the caption or a separate note could clarify the purpose and interpretation of these analyses for a broader audience.
  • Visual Grouping of Results: Visually separating or grouping the results for the 3% energy replacement and the individual food substitutions would enhance clarity and make it easier to distinguish between these two types of analyses.
  • Reference Line for Null Hypothesis: Consider adding a reference line at HR=1 to visually emphasize the null hypothesis and facilitate the quick identification of statistically significant results.
↑ Back to Top