Now they have put in more usage limits at OpenAI it takes about 30% of my 5 hour chatGPT Astra max subscription to do one paper review. I hope people find this useful. I find the reviews quite good and they find things that maybe should have been found in peer review.
This is a large descriptive study with modest novelty. It finds that adults reporting several cardiovascular conditions eat whole fruit slightly less frequently, while the findings for fruit juice are mixed. It cannot establish whether either dietary pattern caused disease. Two reporting errors also warrant attention.
The paper is by Kesler and colleagues: Examining the Differences in Frequency of Whole Fruit or 100% Fruit Juice Intake among United States Adults With and Without Cardiovascular Disease Risk Factors , published in Current Developments in Nutrition in 2026. PIIS2475299126018822(1)
The study compares dietary frequency between people with and without reported cardiovascular diagnoses. It analyses 371,038 adults from the 2019 US Behavioral Risk Factor Surveillance System, a telephone survey.
Participants reported how often they consumed whole fruit and 100% fruit juice, and whether a health professional had ever diagnosed:
- High blood pressure.
- High cholesterol.
- Heart attack.
- Angina or coronary heart disease.
- Stroke.
The researchers used survey weights and accounted for the sampling design. Adjusted analyses included age, education, income and smoking. The methods also mention race and ethnicity, although the table footnotes omit them.
Crucially, the dietary measure was consumption occasions, not portion size, grams of fruit or millilitres of juice . Average reported frequency was 1.06 occasions per day for whole fruit and 0.30 for juice.
The adjusted results were as follows. Differences compare participants reporting each condition with those not reporting it; negative numbers mean less frequent consumption.
| Reported condition |
Whole-fruit difference, occasions/day |
Adjusted p-value |
Juice difference, occasions/day |
Adjusted p-value |
| High blood pressure |
-0.087 |
<0.0001 |
+0.002 |
1.000 |
| High cholesterol |
-0.081 |
<0.0001 |
-0.023 |
<0.0001 |
| Heart attack |
-0.062 |
<0.0001 |
+0.043 |
0.0187 |
| Angina or coronary heart disease |
-0.053 |
0.0034 |
+0.005 |
1.000 |
| Stroke |
-0.037 |
0.1343 |
+0.029 |
0.0700 |
The p-values incorporate the authors’ Bonferroni correction for five cardiovascular comparisons. Figures are rounded from the detailed results accompanying Tables 6 and 7.
Whole-fruit consumption remained significantly lower for four of the five conditions. For juice, only two associations remained significant: lower frequency with high cholesterol and higher frequency with a history of heart attack . The adjusted stroke associations were not statistically significant for either fruit or juice.
The novelty is primarily the scale and specific US population comparison. The paper supplies a large set of estimates describing fruit and juice consumption among people with different cardiovascular histories.
However, separating whole fruit from juice in cardiovascular research is already established. The earlier EPIC-NL study followed 34,560 participants for an average of 14.6 years and examined both exposures against subsequent cardiovascular events. It reported favourable associations for moderate juice consumption, although residual confounding remained possible. Cambridge Core
A subsequent EPIC-NL analysis also explicitly modelled substituting juice for whole fruit or sugar-sweetened beverages. Substitution of juice for fruit showed no statistically significant association with cardiovascular disease after adjustment. That does not establish equivalence, but demonstrates that more directly comparative analyses already existed. Cambridge Core
The authors emphasise their approach of treating cardiovascular status as the predictor and dietary frequency as the outcome. That changes the descriptive question being asked, but reversing the regression direction does not establish which came first . There is no new biological mechanism, intervention or causal method here.
The study has useful strengths. Its large sample provides precise estimates of average differences. The analysis incorporates survey weights, clustering and stratification, separates juice from whole fruit, and corrects for multiple comparisons. The authors also acknowledge many limitations and use publicly available data.
My main criticisms concern interpretation, measurement and reporting.
- **The timing of diet and disease is unresolved.**Current reported eating habits are compared with diagnoses that could have occurred years earlier. Participants might have changed their diet after a heart attack or cholesterol diagnosis. Disease, disability or changes in appetite could also affect food choices.Consequently, the findings are compatible with diet influencing disease, disease influencing diet, or other factors influencing both. They do not estimate the effect of fruit consumption on future cardiovascular risk.The discussion suggests that a health scare might explain some findings. That is plausible, but the disappearance of an association after adjustment does not demonstrate a post-diagnosis change in behaviour.
- **Frequency is a weak measure of dietary exposure.**Someone drinking a small glass of juice once daily and someone drinking a large glass once daily receive the same frequency score.Without quantities, the study cannot establish a dose-response relationship, distinguish moderate from high-volume juice intake, or quantify sugar, energy, fibre or polyphenol exposure.It also cannot determine whether juice replaced whole fruit, another beverage or nothing. Participants could consume both fruit and juice, so these are not mutually exclusive dietary groups.
- **The absolute differences are small, despite very low p-values.**The statistically significant adjusted whole-fruit differences correspond to approximately 0.37 to 0.61 fewer consumption occasions per week .For juice, the heart-attack association corresponds to approximately 0.30 additional occasions per week , and the cholesterol association to 0.16 fewer occasions per week .These differences may have descriptive population value, but their clinical significance is unknown. A very large sample can make small differences statistically convincing without showing that they meaningfully affect health. The study provides no estimate of cardiovascular risk reduction or increase attributable to these differences.
- **Adjustment leaves substantial scope for confounding.**Sex is absent from the reported adjustment set. Physical activity, BMI, broader dietary patterns and sugar-sweetened beverage consumption were also not controlled; the authors acknowledge these latter omissions.Whole-fruit consumption could therefore be acting partly as a marker of other behaviours or circumstances. The appropriate adjustment strategy would depend on a clearly specified causal question, which this descriptive analysis does not resolve.Reporting is also inconsistent about whether race and ethnicity were included. A complete model specification and analysis code would improve reproducibility.
- **Self-reported diagnoses and selection limit interpretation.**Reporting that a clinician has diagnosed high cholesterol is different from having a measured current LDL cholesterol concentration. It depends on testing, access to care, recall and treatment history. Some people in the comparison group may have undiagnosed conditions.The study also concerns surviving, noninstitutionalised survey respondents. It cannot capture people who died from cardiovascular disease, and missing dietary responses led to substantial exclusions. Survey weighting helps address known sampling differences but cannot guarantee elimination of selection bias.
- **The proposed mechanisms were not tested.**The discussion places considerable emphasis on fibre, polyphenols and differences in their bioavailability between fruit and juice.These are reasonable hypotheses, but the study measured none of those exposures directly. It did not measure relevant metabolites, microbiome changes or physiological responses. It therefore cannot identify which constituent, if any, explains the associations.
There are also specific reporting problems that should be corrected:
| Location |
Problem |
| Abstract |
States that lower whole-fruit frequency remained significant for all five conditions after adjustment. For stroke, Table 7 reports p = 0.1343, which is not significant. |
| Discussion, page 7 |
One sentence says both high cholesterol and heart attack were associated with greater juice frequency. The cholesterol association was actually with lower frequency. |
| Titles of Tables 6 and 7 |
Describe the “impact” of intake on cardiovascular risk factors, although the stated models predict intake frequency from cardiovascular status. This misstates the model direction and implies more causality than the design supports. |
These discrepancies do not by themselves invalidate the numerical analyses, but they can materially mislead readers relying on the abstract or discussion.
The paper discloses that the first author was employed by VDF FutureCeuticals and received tuition reimbursement from the company. USDA funding is also reported. This is relevant context for assessing interpretation, although it is not evidence that the results are biased.
A stronger follow-up would measure food and juice quantities repeatedly before and after diagnosis, validate cardiovascular outcomes, and explicitly examine dietary substitutions. Testing the proposed mechanisms would additionally require measured nutrient exposures and biological responses.
