May’s collection lands on a single through-line: what you eat, what you supplement, and how you move form an interconnected anti-inflammatory toolkit. An umbrella review grades 13 food groups for diabetes risk while a parallel analysis links added sugar to inflammatory bowel disease. Three studies map the nutraceutical landscape for mood — ranking 44 supplements head-to-head, validating curcumin for cognition, and confirming probiotics in clinical depression. And a 60-trial network meta-analysis reveals that not all exercise modes are equal when it comes to lowering CRP and IL-6. Taken together, these eight papers reinforce a simple principle: the details of the prescription matter as much as the prescription itself.
Studies at a Glance
Rethinking Macronutrient Orthodoxy
Few questions in nutrition carry more political and clinical weight than the role of dietary fat. Two studies approach the macronutrient debate from opposite ends — saturated fat and artificial sweeteners — and arrive at a shared conclusion: the answer depends entirely on context.
The question of whether reducing saturated fat prevents heart disease has been debated for over sixty years, and this systematic review may be the most sophisticated attempt yet to answer it. Steen and colleagues pooled 17 RCTs enrolling 66,337 participants and applied a novel risk-stratification framework using GRADE methodology with predefined clinical importance thresholds.
Risk-stratified systematic review of 17 RCTs (66,337 participants). Pre-registered on PROSPERO. Included the Women’s Health Initiative (n=29,294). Interventions aimed at reducing or modifying saturated fat over ≥24 months. GRADE with predefined absolute risk thresholds: <5 events/1,000 over 5 years for fatal outcomes, <10/1,000 for nonfatal.
Show full methodology
Databases: MEDLINE, Embase, CENTRAL. Trials published 1965–2006. 11 trials replaced SFA with PUFA, 4 with carbohydrates. Most trials pre-dated statins. Missing data handled with multiple imputation ratios (1.5:1 through 5:1). Risk of bias: 4/17 low, 13/17 high. Team includes Gordon Guyatt (GRADE co-developer) and Lee Hooper (Cochrane review author).
No pooled primary outcome reached significance. The critical finding: when stratified by baseline risk, high-risk individuals saw clinically meaningful reductions exceeding importance thresholds, while low-risk populations saw trivial 5-year absolute benefit. Replacing SFA specifically with PUFA was the only statistically significant subgroup (25% MI reduction).
Strengths
- Pre-registered protocol (PROSPERO)
- Novel risk stratification framework
- 66,337 participants across 17 RCTs
- GRADE methodology with Guyatt involvement
- Macronutrient replacement subgroup analysis
Limitations
- 5-year window may miss lifetime benefit
- WHI was total fat, not specifically SFA
- Most trials pre-date statins (1960s–70s)
- 13/17 trials rated high risk of bias
- Senior author (Johnston) has documented COI history
This study supports risk-stratified dietary fat counseling. For high-risk patients, replacing SFA with PUFA remains evidence-based. For low-risk patients, aggressive SFA restriction shows trivial absolute benefit — though this should not be interpreted as license for unlimited intake. The replacement nutrient matters as much as the restriction: replacing SFA with refined carbohydrates showed no credible benefit.
If the saturated fat story is about who benefits, the sweetener story is about what you trade. This RCT-only meta-analysis examined non-nutritive sweeteners within structured weight management programs — a clinically relevant framing, since most sweetener users are trying to lose weight.
Systematic review and meta-analysis restricted to RCTs of non-nutritive sweeteners in structured weight management programs (≥4 weeks). Databases: MEDLINE, Embase, Cochrane, Web of Science. Outcomes stratified by weight-loss vs. maintenance phase. Subgroups by sweetener type.
Overall, NNS performed comparably to controls for weight loss. Aspartame and “mixed consumption” subgroups showed modest benefit. The insulin finding is the clinical surprise: NNS users showed less insulin sensitivity improvement during weight loss, suggesting metabolic effects beyond calorie replacement. Appetite was modestly reduced during active weight loss (SMD −0.16) but not during maintenance.
Strengths
- RCT-only design avoids reverse causation
- Phase-stratified analysis (loss vs. maintenance)
- Sweetener-specific subgroup analyses
- No food/beverage industry funding
Limitations
- Short trial durations (weeks to months)
- Cannot address long-term disease endpoints
- No dose-response analysis reported
- Limited data for individual sweetener types
NNS can serve as a palatability tool during active weight loss, but practitioners should monitor metabolic parameters rather than assuming metabolic neutrality. The insulin signal — though requiring replication — raises concerns for patients with metabolic syndrome or pre-diabetes. The divergent results by sweetener type suggest not all sweeteners are equivalent.
The Microbiome-Diet-Disease Axis
The integration of multi-omic data with dietary patterns is redefining how we understand diet-disease relationships — moving beyond food frequency questionnaires to microbial and metabolic mechanisms.
This Southern Community Cohort Study analysis is the most methodologically ambitious paper in this month’s roundup. It followed 71,533 predominantly low-income, African American participants — a population vastly underrepresented in nutritional epidemiology — and integrated dietary patterns, fecal metagenomics, and blood metabolomics.
Prospective cohort (SCCS), n=71,533, median follow-up 11.6 years. Three plant-based diet indices (overall PDI, healthy hPDI, unhealthy uPDI) calculated using Satija methodology. Multi-omic integration: fecal metagenomics (n=417) and blood metabolomics (n=1,581).
The divergent findings are the headline: healthy plant-based diets protected against cancer, while unhealthy plant-based diets increased CRC risk by 39% in screening-naive participants. Multi-omic data linked diet indices to fiber-degrading SCFA producers (Faecalibacterium, Roseburia) and genotoxic species (Fusobacterium nucleatum).
Strengths
- Underrepresented population (low-income, African American)
- Multi-omic integration (metagenomics + metabolomics)
- 11.6-year prospective follow-up
- Healthy vs. unhealthy PDI distinction
Limitations
- Single dietary assessment at baseline
- Cross-sectional multi-omic data
- Small multi-omic subsets (n=417, n=1,581)
- Multiple comparisons (liver cancer p=0.03)
Simply telling patients to “eat more plants” is insufficient — and potentially counterproductive if those plants are refined grains and sugary drinks. The screening-naive CRC finding is particularly relevant for populations with lower screening uptake, where dietary prevention strategies may have the greatest impact.
Ultra-Processed Food: Expanding the Damage Profile
The case against ultra-processed food continues to grow, and this month’s evidence adds an unexpected endocrine dimension to the expanding list of UPF-associated conditions.
Using UK Biobank data from 123,812 participants followed for a median of 11.2 years, this analysis connects ultra-processed food consumption to thyroid disease — adding an endocrine dimension to the established UPF risk profile of cardiovascular disease, diabetes, depression, and cancer.
Prospective cohort from the UK Biobank, n=123,812, median 11.2-year follow-up. UPF classified via NOVA system. Cox proportional hazards with multivariable adjustment. Sex-stratified analyses and mediation analysis for novel pathways.
The association was amplified in women on hormone replacement therapy, suggesting an interaction between UPF exposure and exogenous estrogen. Mediation analysis identified three novel pathways: cystatin C (inflammation), apolipoprotein A, and HDL cholesterol — suggesting UPF operates through inflammatory and lipid-mediated pathways rather than direct thyroid toxicity.
Strengths
- 123,812 participants, 11.2-year follow-up
- Novel mediation pathway analysis
- Sex-stratified and HRT interaction analyses
Limitations
- Predominantly white British population
- Single dietary assessment at baseline
- NOVA classification imprecision
- Residual confounding possible
This adds thyroid disease to the expanding UPF damage profile. For practitioners managing hypothyroidism — one of the most common endocrine disorders, affecting 5–10% of adults — dietary history should include UPF exposure assessment. The HRT interaction deserves attention in women’s health contexts.
Brain Health and Neuroprotection
Two studies this month examine dietary and supplemental strategies for cognitive protection — one with robust evidence and one where promise exceeds proof.
Drawing on the Nurses’ Health Study and Health Professionals Follow-up Study, this analysis followed 131,821 participants for up to 43 years — making it the longest prospective study of habitual beverage consumption and dementia risk ever conducted.
Two prospective cohorts (NHS + HPFS), n=131,821, up to 43 years follow-up. Habitual coffee and tea consumption assessed repeatedly via validated FFQs. Stratified by caffeine status (caffeinated vs. decaf). Subgroup analyses by APOE genotype, age, and sex.
Optimal protection at 2–3 cups/day. The decaf null finding identifies caffeine as the likely active compound via adenosine A2A receptor antagonism. The association held across APOE genotypes, meaning even ε4 carriers derived similar benefit.
Strengths
- 131,821 participants, up to 43 years follow-up
- Repeated dietary assessment (not single baseline)
- Decaf comparison isolates caffeine effect
- APOE-stratified analysis
Limitations
- Self-reported beverage intake
- Predominantly white health professionals
- Observational (cannot prove causation)
For patients concerned about cognitive aging, 2–3 cups of caffeinated coffee daily represents a low-risk, evidence-based recommendation. Patients switching to decaf for other reasons (anxiety, sleep, arrhythmia) should understand they may not retain the cognitive benefit.
Creatine for cognitive health has become popular in wellness media, driven by compelling biological plausibility: creatine regenerates ATP in neural tissue. Marshall and colleagues evaluated the evidence specifically in adults aged 55+.
Systematic review of 6 studies (2 RCTs, 4 cross-sectional), 1,542 participants aged 55+. Eight databases searched. Quality assessed via Modified Downs and Black checklist. No meta-analysis performed due to heterogeneity. PROSPERO registered.
5 of 6 studies reported positive associations, but the evidence hierarchy tells a different story. The only well-designed RCT (Alves et al., 2013, rated “Good” quality) found no effect after 24 weeks at 5g/day. The positive signals came from 4 cross-sectional studies (which cannot establish causation) — 3 rated “Poor” quality. No study measuring global cognition (MMSE) found any association. No study measured brain creatine concentrations. EFSA declined to authorize a creatine-cognition health claim in 2024.
Strengths
- First age-specific systematic review (55+)
- Eight databases, PROSPERO registered
- Transparent quality scoring per study
Limitations
- Only 6 studies; 50% rated “Poor” quality
- Sole high-quality RCT was negative
- No brain creatine measurement in any study
- Published criticism (Quaresma, 2025)
Current evidence does not support recommending creatine for cognitive function in older adults. The biological plausibility is intact, but well-powered RCTs with cognitive primary endpoints and brain creatine measurement are needed. For patients already taking creatine for musculoskeletal indications, cognitive monitoring as a secondary outcome is reasonable.
Precision Nutrition and Cancer Prevention
Two studies demonstrate how metabolomic profiling and dose-response data are enabling more targeted, quantified approaches to disease prediction and cancer prevention.
This may be the most technically ambitious study in the roundup. Li and colleagues analyzed 469 blood metabolites across 10 prospective cohorts encompassing 23,634 participants, with up to 26 years of follow-up, integrating metabolomic, genomic, and lifestyle data at unprecedented scale.
Multi-cohort metabolomics study: 469 metabolites measured in 23,634 participants across 10 prospective cohorts (up to 26 years follow-up). LASSO regression for prediction signature derivation. Two-sample Mendelian randomization for causal inference. Lifestyle factor analysis for modifiable determinants.
A 44-metabolite prediction signature (20 amino acids, 19 lipids, 5 others) improved T2D prediction beyond conventional risk factors. Mendelian randomization identified 42 lipids and 5 amino acids with causal evidence. The practice-relevant finding: lifestyle factors explained greater variance in diabetes-associated metabolites than in non-associated metabolites — meaning the causal metabolites are preferentially responsive to modifiable interventions.
Strengths
- 10 prospective cohorts, 23,634 participants
- Up to 26 years follow-up
- Mendelian randomization for causal inference
- 67 novel metabolite associations
Limitations
- Predominantly white European participants (~77%)
- Single time-point metabolomics
- Variable AUC performance across cohorts
- Prediction signature not clinically validated
This validates the precision nutrition paradigm: metabolomic profiling can identify high-risk individuals years before clinical disease, and the risk-driving metabolites respond to the interventions we already know work — physical activity, diet quality, and body composition management. The specific dietary exposures that matter most (red meat, vegetables, coffee, whole grains) are already part of evidence-based dietary counseling.
While most meat-and-cancer research focuses on invasive malignancy, Ao and colleagues examined colorectal cancer precursors — adenomas and polyps — across 31 observational studies. This focus on the earlier, more preventable stage of carcinogenesis yields data directly applicable to screening populations.
Systematic review and meta-analysis of 31 observational studies. PROSPERO registered. Six meat subtypes examined: red, processed, unprocessed red, white, seafood, total. Random-effects models. Dose-response analysis per serving increment.
Dose-response: each additional 100g/day red meat increases precancerous lesion risk by 14%; each 50g/day processed meat by 27%. On the protective side, seafood reduced adenoma risk by 26% and white meat by 21%. Unprocessed red meat and charred meat showed no significant associations, suggesting processing method matters more than red meat per se.
Strengths
- 31 studies, comprehensive meat subtyping
- PROSPERO registered
- Dose-response analysis
- Site-specific outcome assessment
Limitations
- All observational studies (cannot prove causation)
- Heterogeneous meat category definitions
- Wide CIs for some estimates
- Limited data on serrated lesions
For colonoscopy screening populations, red and processed meat intake should be part of risk stratification. The dose-response data translates directly: a patient eating a large steak daily (300–400g) faces 42–56% increased precancerous lesion risk. Recommending seafood and poultry as protein substitutes is supported by both the reduction data and established biological mechanisms.
Synthesis & Emerging Themes
The End of Macronutrient Monoliths
The saturated fat review’s risk stratification, the sweetener study’s subgroup-specific effects, and the plant-based diet study’s quality distinction all point in the same direction: broad macronutrient categories (“fat,” “carbs,” “plant-based”) are too crude to guide clinical decision-making. The relevant questions are now which fat, replacing what, in whom — a level of specificity that most dietary guidelines have yet to adopt.
Multi-Omic Integration Is Becoming Standard
Two of this month’s strongest studies — the SCCS plant-based diet analysis and the Nature Medicine metabolomics study — integrate genomic, metabolomic, and microbiome data to move beyond association and toward mechanism. The metabolomics study’s finding that lifestyle-responsive metabolites are preferentially causal for disease is a profound insight: it means the biomarkers most useful for prediction are also the most responsive to intervention.
Ultra-Processed Food as a Cross-System Toxin
The addition of hypothyroidism to the UPF damage profile — alongside cardiovascular disease, diabetes, depression, cancer, and all-cause mortality — reinforces the emerging view that UPF’s harms are not mediated through any single nutrient but through the industrial processing matrix itself: emulsifiers disrupting gut barrier function, packaging chemicals acting as endocrine disruptors, and ultra-palatable formulations overriding satiety signaling.
“The evidence base grows more precise by the month. Our recommendations should follow suit.”
For the practicing clinician, the actionable takeaways are clear: prioritize dietary fat quality over quantity; recommend 2–3 cups of caffeinated coffee for patients concerned about cognitive aging; counsel plant-based diet quality rather than category; include ultra-processed food assessment in endocrine evaluations; and use the emerging dose-response data on meat and cancer precursors to have specific, quantified conversations with screening-age patients.