Food--vs--supplement-based-very-low-energy-diets-and-gut-micro_2025_Cell-Rep.pdf

Article
Food- vs. supplement-based very-low-energy diets
and gut microbiome composition in women with high
body mass index: A randomized controlled trial
Graphical abstract
Highlights
• 47 women with obesity randomized to food- or supplement-
based very-low-energy diets
• Three-week diets matched at ∼800–900 kcal/d, but whole vs.
industrial ingredients differ
• Food vs. supplement: greater increases in species-level
alpha diversity (Shannon)
• Food vs. supplement: greater richness, smaller beta diversity
shifts, and preserved taxa
Authors
Melissa M. Lane, Amelia J. McGuinness,
Mohammadreza Mohebbi, ...,
Sally Beattie, Wolfgang Marx,
Felice N. Jacka
Correspondence
m.lane@deakin.edu.au
In brief
Lane and McGuinness et al. show that in
47 women following 3 weeks of very-low-
energy diets, a food-based approach
leads to greater gains in gut microbial
diversity and better preservation of
beneficial taxa than a supplement-based
approach, highlighting the importance of
diet format in gut health during energy
restriction.
Lane et al., 2025, Cell Reports Medicine 6, 102417
October 21, 2025 © 2025 The Author(s). Published by Elsevier Inc.
https://doi.org/10.1016/j.xcrm.2025.102417
ll
Article
Food- vs. supplement-based very-low-energy diets 
and gut microbiome composition in women with high 
body mass index: A randomized controlled trial
Melissa M. Lane,1,12,14,* Amelia J. McGuinness,1,12 Mohammadreza Mohebbi,4 Mojtaba Lotfaliany,1 Amy Loughman,11
Martin O’Hely,2,10 Adrienne O’Neil,1 Jessica Batti,1 Mark Kotowicz,1,7,8,10 Michael Berk,1,3,10 Lucy Saunders,10
Richard Page,8,9 Sally Beattie,9 Wolfgang Marx,1,13 and Felice N. Jacka1,5,6,13
1Deakin University, the Institute for Mental and Physical Health and Clinical Translation (IMPACT), Food & Mood Centre, School of Medicine 
and Barwon Health, Geelong, VIC, Australia
2Murdoch Children’s Research Institute, Parkville, VIC, Australia
3Orygen, The National Centre of Excellence in Youth Mental Health, Centre for Youth Mental Health, Florey Institute for Neuroscience and 
Mental Health and the Department of Psychiatry, The University of Melbourne, Melbourne, VIC, Australia
4Deakin University, Faculty of Health, Biostatistics Unit, Geelong, VIC, Australia
5Centre for Adolescent Health, Murdoch Children’s Research Institute, Melbourne, VIC, Australia
6College of Public Health, Medical & Veterinary Sciences, James Cook University, Townsville, QLD, Australia
7Department of Medicine-Western Health, The University of Melbourne, St Albans, VIC, Australia
8Barwon Health, University Hospital, Geelong, VIC, Australia
9Barwon Centre of Orthopaedic Research and Education, Deakin University, Geelong, VIC, Australia
10Deakin University, the Institute for Mental and Physical Health and Clinical Translation (IMPACT), School of Medicine and Barwon Health, 
Geelong, VIC, Australia
11University of Melbourne, Melbourne School of Psychological Sciences, Melbourne, VIC, Australia
12These authors contributed equally 
13Senior author 
14Lead contact
*Correspondence: m.lane@deakin.edu.au
https://doi.org/10.1016/j.xcrm.2025.102417
SUMMARY
In a single-blind, two-arm, randomized controlled-feeding trial (May 2021–February 2022), 47 women (30–65 
years, BMI 30–45 kg/m2) are randomized to either a food-based or a supplement-based very-low-energy diet 
(VLED: 800–900 kcal/d) for 3 weeks. The food-based VLED comprises pre-packaged meals (∼93% whole- 
food ingredients), while the supplement-based VLED comprises shakes, soups, bars, and desserts (∼70% 
industrial ingredients). The primary outcome is species-level alpha diversity (Shannon index). Secondary out­
comes include species richness, beta diversity, taxonomic composition, functional potential, anthropomet­
rics, serum biomarkers, mental health, sleep, and gastrointestinal symptoms. Modified intention-to-treat 
(mITT) analyses (n = 45) assess diet group × time interactions as beta coefficients (β) with 95% confidence 
intervals (CIs). A between-group differential change is observed for the Shannon index, with a greater in­
crease in the food-based group (mITT β: 0.37, 95% CI: 0.15–0.60). The food-based group also shows greater 
species richness, smaller beta diversity shifts, and compositional changes preserving fiber-degrading, 
health-associated taxa.
INTRODUCTION
The gut microbiome is intricately connected to human health and 
disease, and understanding the influence of diet on its composi­
tion and function may inform gut-focused treatment strategies.1
The gut microbiome is shaped by both short- and long-term di­
etary exposures.2–5 Dietary interventions, including high-fiber 
and Mediterranean-style diets, have been shown to beneficially 
alter the gut microbiome.6,7 This includes increasing bacterial di­
versity and the abundances of bacterial species considered 
beneficial for health, enhancing carbohydrate breakdown by mi­
crobiome enzymes, and reducing inflammation.4 Conversely, 
more ‘‘Westernized’’ diets, characterized by higher intakes of 
sugar, fat, and protein and lower intakes of fiber, are linked to 
reduced gut microbiome diversity and functional capacity, 
higher body mass index (BMI), increased inflammatory markers, 
and elevated risk of diseases such as cancer.8,9
Although dietary interventions have been linked to alterations 
in gut microbiome composition and potential function,4,6,7,10
the effects of food processing on the gut microbiome have not 
been thoroughly examined. Heavily processed foods comprise 
largely of industrially derived constituents, including colorants, 
Cell Reports Medicine 6, 102417, October 21, 2025 © 2025 The Author(s). Published by Elsevier Inc. 
1 
This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
ll
OPEN ACCESS
emulsifiers, flavors, microparticles, stabilizers, and thickeners, 
with few whole foods.11 While human data are lacking, emerging 
evidence in mice suggests that these industrially derived constit­
uents might alter the gut microbiome and increase the risk for 
immune-mediated diseases.12 Additionally, the often lower nutri­
tional quality of extensively processed foods, especially in terms 
of fiber diversity and quantity, may negatively affect gut bacterial 
metabolism of short-chain fatty acids, which are crucial for main­
taining gut barrier integrity.13 Although diets high in extensively 
processed foods typically have poorer nutritional profiles 
compared to those consisting mainly of whole foods,14 highly 
processed items such as supplement-based shakes, bars, and 
soups used in very-low-energy diet (VLED) programs, are in­
tended to meet nutritional needs. These products aim for an 
adequate ratio of macronutrients (protein, fat, and carbohydrate) 
and essential vitamins and minerals, despite their extensive pro­
cessing. Targeted at individuals with a high BMI of 30 kg/m2 or 
more, VLEDs limit energy intake to approximately 800–900 kcal 
per day, and are shown to effectively reduce weight and improve 
markers of type 2 diabetes and cardiovascular disease.15,16
While these VLEDs demonstrate clear benefits in managing 
metabolic-related outcomes, the impact of highly processed 
supplement-based VLEDs on the gut microbiome is not well un­
derstood, especially compared to less processed, food-based 
VLEDs.
A systematic review examining the impact of supplement- 
based VLEDs on gut microbiome composition reported mixed 
findings for diversity and inconsistent shifts within the Firmicutes 
and Bacteroidetes phyla (now referred to as Bacillota and Bac­
teroidota, respectively), including changes in both potentially 
beneficial and opportunistically pathogenic taxa—likely reflect­
ing methodological limitations in the included studies.17 While 
metabolic parameters generally improved (e.g., reductions in 
anthropometric measures) or remained unchanged, most 
studies lacked randomization and used single-group designs 
without comparators, with further confounding introduced by 
concurrent lifestyle changes.17 Only two of seven trials used a 
parallel group design: one compared a VLED with surgical 
intervention and the other against two control groups based on 
BMI, where neither control group received any intervention.17
Additionally, controlled ad libitum feeding studies have shown 
that minimally processed, whole-food-based diets—whether 
differing in macronutrient composition18 or designed to be 
matched in nutrient profiles19 to heavily processed diets—can 
alter gut microbiome composition,18 modulate immune cell ac­
tivity,18 reduce energy intake (∼508 kcal/day),19 induce sponta­
neous weight loss (∼0.9 kg),19 and lower inflammatory markers 
(∼45%).19 Despite these insights, no experimental studies 
have directly compared how supplement- and food-based 
VLEDs, which are designed to be isocaloric and similar in their 
nutritional profiles but different in their level of food processing, 
affect the gut microbiome. Understanding these potential differ­
ential impacts is crucial, given the significant role of the gut mi­
crobiome in overall health,1 and will aid in evaluating the compre­
hensive risk-benefit profile of these VLED formats.
This study explored the effects of a food- versus supplement- 
based VLED program on gut microbiome composition in women 
with high BMI. Our primary outcome was species-level alpha di­
versity, assessed using the Shannon index. Secondary gut mi­
crobiome outcomes included species richness, beta diversity, 
taxonomic composition, and functional potential. Secondary 
clinical outcomes included anthropometric measures, serum 
biomarkers, mental health, sleep, and gastrointestinal symp­
toms. We hypothesized that a food-based VLED, primarily 
comprising whole-food ingredients, would increase gut micro­
biome diversity and the relative abundances of health-associ­
ated bacteria compared to a supplement-based VLED, predom­
inantly consisting of highly processed industrial ingredients. We 
also hypothesized that those randomized to the food-based 
VLED would experience more beneficial metabolic and mental 
health outcomes relative to those on the supplement- 
based VLED.
RESULTS
Recruitment and trial retention
We screened 102 participants for eligibility, of whom 40 were 
initially randomized (Figure 1). Due to participant withdrawal/ 
loss to follow-up (n = 4) and missing fecal samples (n = 4), we 
aimed to recruit an additional eight participants to reach a sam­
ple size of 40 with complete data. Overall, 47 participants were 
randomized, including 23 in the food-based group and 24 in 
the supplement-based group. Of these, 45 were included in 
mITT analyses (food-based: n = 23, supplement-based: n = 
22), and 39 in complete case analysis (food-based: n = 22, sup­
plement-based: n = 17) of the primary outcome.
Baseline characteristics
On average, participants in the food-based group were 
less commonly married or employed compared to the supple­
ment-based group (Table 1). More participants in the food- 
based group were taking medication, and they also had higher 
average BMI, body weight, physical activity levels, and 
waist and hip circumferences, compared to the supplement- 
based group.
Adherence and safety
Diet adherence
Of the complete cases (n = 39), on average, dietary intake was 
recorded on 20 out of the requested 21 days. The average daily 
energy intake of 825 kcal was within the 800–900 kcal VLED 
target range (Table S1). Both groups reported similar average 
daily intakes of carbohydrate, protein, fiber, sugar, and sodium. 
The food-based group reported higher average total energy 
intake (903 vs. 748 kcal/d) and total fat intake (40.1 vs. 22.5 
g/d, equivalent to ∼158 kcal) compared to the supplement- 
based group. The difference in average energy intake between 
the two groups closely matches the energy difference attribut­
able to the variation in fat intake.
Adverse events
Participants in the supplement-based group reported more 
adverse events compared to the food-based group (19 vs. 8, 
respectively) (Table S2). The most common adverse event 
reported for both groups was headaches (5 vs. 3 events, 
respectively). No serious adverse events were reported in either 
group.
2 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
(∼45%).19 Despite these insights, some experimental studies g/d, equivalent to ∼158 kcal) compared to the supplement-
Primary outcome
Species-level alpha diversity (Shannon index)
We observed a statistically significant between-group differen­
tial change over time in the Shannon index (mITT β: 0.37, 95% 
CI: 0.15–0.60) (Table S3), with the food-based group experi­
encing a greater increase (mean change: 0.27, 95% CI: 0.09– 
0.44) compared to the statistically non-significant decrease 
observed in the supplement-based group (mean change: 
−0.11, 95% CI: −0.27 to 0.05) (Figure 2A). Results were similar 
across complete case analyses (Table S3) and sensitivity ana­
lyses that adjusted for prognostic covariates and removed a 
sample with low read count (see quantification and statistical 
analysis section). Additional adjustment for age and postmeno­
pausal status gave similar results, while adjusting for alcohol 
intake increased the effect size (mITT β: 0.45, 95% CI: 0.18– 
0.72) (Table S10).
Secondary outcomes
Species-level alpha diversity (richness)
We observed a between-group differential change in species 
richness (mITT β: 27.9, 95% CI: 12.1–43.7), with the food-based 
Figure 1. CONSORT flow diagram 
Main analysis refers to the modified intention-to- 
treat (mITT) approach, which included all ran­
domized participants who provided baseline data, 
regardless of follow-up completion.
group experiencing a greater increase 
(mean change: 23.2, 95% CI: 12.7– 
33.7) compared to the statistically 
non-significant decrease exhibited in the 
supplement-based group (mean change: 
−4.59, 95% CI: −17.0 to 7.78) (Table S3; 
Figure 2B).
Beta diversity (Aitchison distance)
We observed a between-group differen­
tial change over time in beta diversity 
(complete case r2 = 0.051, p = 0.001) 
(Table S4), suggesting that 5.1% of the 
difference in the shift in beta diversity 
observed between groups could be ex­
plained by diet. Visual inspection of the 
data using PCA suggested a greater shift 
in beta diversity in the supplement- 
versus food-based group (Figure 2C). 
The variance in Aitchison distances ex­
plained by time point was 1.8% in the 
food-based group and 4.5% in the sup­
plement-based group (Table S4).
Taxonomic composition (species, 
genus, family, and phylum)
We observed between-group differential 
changes over time in the centered log-ra­
tio (CLR)-transformed relative abun­
dance of 72 bacterial species (p < 0.05); 
however, these did not survive adjust­
ment for multiple comparisons (Table S5). There were be­
tween-group differential changes in 56 genera (Table S6), eight 
families (Table S7), and one phylum (Table S8); of these, eight 
genera and one family survived adjustment for multiple compar­
isons (Table 2; Figures S1 and S2); however, no phyla met the 
significance threshold after correction. Results of complete 
case analyses were similar (Tables S5, S6, S7, and S8).
For brevity, and given the large number of taxa tested and sub­
stantial within-genus variability, we report detailed results for the 
three most statistically significant and well-characterized genera 
that remained significant after correction for multiple compari­
sons: Lachnospira (mITT β: 2.17, 95% CI: 0.99–3.34), Anaeros­
tipes (mITT β: −1.73, 95% CI: −2.67 to −0.80), and Ruminococ­
cus_A (mITT β: −1.05, 95% CI: −1.68 to −0.43) (Figures 2D–2F). 
For Lachnospira, the food-based group experienced a greater, 
but non-significant, increase (mean change: 0.55, 95% CI: 
−0.08 to 1.19) compared to a statistically significant decrease 
in the supplement-based group (mean change: −1.67, 95% CI: 
−2.68 to −0.66). For Anaerostipes, the food-based group expe­
rienced a non-significant decrease (mean change: −0.41, 95% 
CI: −1.01 to 0.20) compared to a statistically significant increase 
Cell Reports Medicine 6, 102417, October 21, 2025 
3 
Article
ll
OPEN ACCESS
in the supplement-based group (mean change: 1.34, 95% 
CI: 0.58–2.09). For Ruminococcus_A, the food-based group 
experienced a statistically significant decrease (mean change: 
−0.86, 95% CI: −1.35 to 0.37) compared to a non-significant 
increase in the supplement-based group (mean change: 
0.19, 95% CI: −0.23 to 0.62). Additionally, the family that 
remained statistically significant after correction for multiple 
comparisons was an uncharacterized family of the Christense­
nellales order, GCA_900066905 (mITT β: −1.19, 95% CI: −1.81 
to −0.57) (Table S8; Figure S2). The food-based group experi­
enced a statistically significant decrease in Christensenella­
les_GCA_900066905 (mean change: −0.33, 95% CI: −0.60 to 
−0.06) compared to a statistically significant increase in the sup­
plement-based group (mean change: 0.87, 95% CI: 0.29–1.44).
Functional potential (MetaCyc pathways)
No significant between-group differential changes in functional 
alpha diversity were observed (Table S3). We observed 
between-group differential changes in six MetaCyc groups 
(p < 0.05); however, none of these remained statistically signifi­
cant after adjustment for multiple comparisons (Table S9).
Secondary clinical outcomes
For secondary clinical outcomes, there was a between-group 
differential change over time (p < 0.05) in hip circumference 
(mITT β: 3.29 cm, 95% CI: 0.17–6.42), albumin concentration 
(mITT β: −1.60 g/L, 95% CI: −2.86 to −0.33), and constipation 
symptoms (mITT β: 19.5, 95% CI: 3.59–35.5) (Table S3); howev­
er, these did not remain statistically significant after correction 
for multiple comparisons. For hip circumference, the food-based 
group experienced a smaller, non-significant decrease (mean 
change: −1.18 cm, 95% CI: −3.40 to 1.03) compared to a larger, 
statistically significant decrease in the supplement-based group 
(mean change: −4.53 cm, 95% CI: −6.87 to −2.20). For albumin 
concentration, the food-based group experienced a smaller, sta­
tistically significant increase (mean change: 1.32 g/L, 95% CI: 
0.48–2.15) compared to a larger, statistically significant increase 
in the supplement-based group (mean change: 2.91 g/L, 95% CI: 
1.90–3.92). For constipation symptoms (where higher scores 
reflect more favorable ratings), the food-based group experi­
enced a greater, non-significant improvement (mean change: 
1.95, 95% CI: −6.20 to 10.1) compared to a statistically signifi­
cant worsening in the supplement-based group (mean change: 
−18.5, 95% CI: −33.1 to −3.98).
Post hoc exploratory analyses
Correlation analyses
Changes in species-level Shannon index showed a weak inverse 
association with changes in serum leptin (r = −0.37, p = 0.024), 
suggesting that greater increases in the Shannon index tended 
to be associated with either smaller increases or greater de­
creases in serum leptin concentrations. Weak positive associa­
tions were observed between changes in the Shannon index 
and changes in bloating/flatulence (r = 0.31, p = 0.059) and 
abdominal pain (r = 0.31, p = 0.061), suggesting that greater in­
creases in the Shannon index tended to be associated with 
greater increases (more favorable ratings) or smaller decreases 
in these gastrointestinal symptoms (Figure 3).
Species-level contributions to the Shannon index
Decomposition of the Shannon index into species-level 
contributions revealed modest differences between groups. 
The food-based group showed a slightly greater number of 
Table 1. Baseline characteristics of participants randomized to food- vs. supplement-based very-low-energy diets
Food-based VLED (n = 23)
Supplement-based VLED (n = 24)
Total (n = 47)
Sociodemographic factors
Age, years—M (SD)
46.5 (10.3)
48.3 (9.91)
47.4 (10.0)
Born in Australia, n (%)
22 (96%)
19 (79%)
41 (87%)
Married, n (%)
15 (65%)
19 (79%)
34 (72%)
Post-secondary school education, n (%)
21 (91%)
20 (83%)
41 (87%)
Employed, n (%)
18 (78%)
23 (96%)
41 (87%)
Household income above $74,999, n (%)a
13 (57%)
13 (54%)
26 (55%)
Health factors
Current smoker, n (%)
1 (4%)
2 (8%)
3 (6%)
Alcohol intake, grams/day—M (SD)b
4.8 (6.5)
3.0 (4.8)
3.9 (5.7)
Any medication use, n (%)
18 (78%)
13 (59%)
31 (69%)
Stool consistency—M (SD)
3.8 (1.1)
4.3 (1.2)
4.0 (1.2)
Body mass index—M (SD)
36.8 (3.6)
34.8 (3.2)
35.8 (3.5)
Weight, kilograms—M (SD)
101 (10.1)
94.1 (10.3)
97.3 (10.6)
Waist circumference, centimeters—M (SD)
107 (8.7)
103 (9.1)
105 (9.1)
Hip circumference, centimeters—M (SD)
124 (9.1)
119 (7.1)
122 (8.5)
Low physical activity, n (%)
13 (57%)
18 (75%)
31 (66%)
Peri/menopausal, n (%)c
7 (30%)
11 (46%)
18 (38%)
Note: stool consistency measured using the Bristol Stool Form Scale; low physical activity measured using the International Physical Activity Ques­
tionnaire. Abbreviations: M, Mean; n (%), number and proportion of participants; VLED, very low-energy diet; SD, standard deviation.
aData available for 22 in the food-based group and 23 in the supplement-based group (n = 45).
bData available for 18 in the food-based group and 19 in the supplement-based group (n = 37).
cData available for 19 in the food-based group and 22 in the supplement-based group (n = 41).
4 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
species with increased contributions to diversity (16 increased 
contribution vs. 14 decreased contribution) compared to the 
supplement-based group (13 increased contribution vs. 17 
decreased contribution) (Figure 4). Across both groups, 45 
distinct species were identified as major contributors. Of these, 
15 species were shared between groups, and each of their con­
tributions changed in the same direction in both groups. The re­
maining 30 species were group specific: 15 unique to the food- 
based group and 15 unique to the supplement-based group, 
with both increases and decreases in contribution observed.
To capture broader community shifts beyond these top contrib­
utors, we examined the net contribution of all other species (rep­
resented by the ‘‘Other_net’’ bar) in each group. The food-based 
group showed a large positive contribution from other species, 
suggesting a more diffuse restructuring of the community beyond 
just the most prominent taxa that contributed to the increase in 
Shannon index observed in this group. In contrast, in the supple­
ment-based group, this contribution was small and negative, 
showing that the top 30 species contributed the most to the 
modest reduction in the Shannon index observed in this group.
To further contextualize these shifts, we classified the top 30 
species in each group as either dominant (≥1% mean relative 
abundance) or subdominant (<1%) at baseline. In both groups, 
most species were subdominant at baseline (20 in the food- 
based group and 21 in the supplement-based group) rather 
than dominant (10 in the food-based group and 9 in the supple­
ment-based group), indicating that changes in the Shannon in­
dex were largely contributed to by less abundant members of 
the community in both groups.
Phylum-level taxonomic composition
Aggregating species-level relative abundances at the phylum 
level showed the gut microbiome was broadly dominated by 
bacteria belonging to Firmicutes (now Bacillota) and Bacteroi­
dota (previously Bacteroidetes) in both groups and at both 
time points (Figure 5). Overall, phylum-level profiles remained 
relatively stable across groups and over time, with no substantial 
compositional shifts observed at this taxonomic level.
DISCUSSION
We present evidence showing differential effects of food- versus 
supplement-based VLEDs on gut microbiome alpha diversity, 
beta diversity, and the relative abundances of eight genera and 
one family in women with high BMI. Consistent with our hypoth­
eses, and despite comparable weight loss and nutrient intakes 
(carbohydrate, protein, sugar, sodium, and fiber), the food- 
based group exhibited greater increases in the species-level 
Shannon index compared to the supplement-based group. The 
food-based group also showed more pronounced increases in 
species richness, smaller shifts in beta diversity, and, similar to 
Group
SB−VLED
FB−VLED
3.6
3.8
4.0
4.2
Baseline
Final
A
     Shannon index
140
150
160
170
180
190
Baseline
Final
B
         Richness
−15
−10
−5
0
5
10
15
−10
0
10
20
PC1: 9.15%
PC2: 6.24%
Legend
Baseline FB−VLED
Final FB−VLED
Baseline SB−VLED
Final SB−VLED
C
2
3
4
5
Baseline
Final
Lachnospira
D
4
5
Baseline
Final
Anaerostipes
E
2.0
2.5
3.0
3.5
4.0
Baseline
Final
Ruminococcus_A
F
Figure 
2. Between-group 
differential 
changes in gut microbiome alpha diversity, 
beta diversity, and genus-level composition 
during food- vs. supplement-based VLEDs 
over three weeks 
(A and B) Interaction plots for the estimated mar­
ginal means of (A) species-level Shannon index 
(primary outcome) and (B) species richness, in 
participants who received a food-based very-low- 
energy diet (FB-VLED; purple) versus supplement- 
based very-low-energy diet (SB-VLED; blue) from 
baseline to week 3 (final). Vertical bars represent 
95% confidence interval. 
(C) Principal-component analysis of Aitchison 
distances in those that received a food-based 
very-low-energy diet (FB-VLED; purple) versus 
supplement-based very-low-energy diet (SB- 
VLED; blue) at baseline and week 3 (final). The 
scatterplot shows the first two principal compo­
nents (PC1 and PC2), explaining 9.15% and 
6.24% of the variance, respectively. Each point 
represents one biological replicate (an individual 
sample). Ellipses around each group represent the 
95% confidence interval, illustrating the spread 
and central tendency of the samples within each 
group. The ellipses provide a visual indication of 
the multivariate normal distribution of the data 
points, helping to identify the degree of overlap or 
separation between different groups. 
(D–F) Interaction plots for the estimated marginal 
means of the three most statistically significant 
and well-characterized bacterial genera that re­
mained significantly different between groups 
over time after correction for multiple compari­
sons (q < 0.1): Anaerostipes, Lachnospira, and 
Ruminococcus_A. (A, B, D–F) Points represent estimated marginal means derived from all biological replicates (individual participant samples) within each group 
and time point, with vertical bars representing 95% confidence intervals. No technical replicates were performed.
Cell Reports Medicine 6, 102417, October 21, 2025 
5 
Article
ll
OPEN ACCESS
alpha diversity patterns, largely divergent directional changes 
across genus- and family-level taxa. Although we hypothesized 
more favorable changes in the food-based group, we observed 
no statistically significant differences between the two diet 
groups in terms of changes to other secondary outcomes, 
including species- and phylum-level composition, functional po­
tential, or clinical outcomes such as anthropometrics, serum bio­
markers, mental health, sleep, or gastrointestinal parameters. 
Hence, the observed differences in microbial diversity and com­
munity structure offer preliminary support for potential micro­
biome-mediated effects of VLED format, with further research 
needed to assess the persistence and clinical relevance of these 
changes over longer time frames.
Diversity differences and clinical implications
To our knowledge, no randomized trials have compared the ef­
fects of food- versus supplement-based VLEDs on the gut micro­
biome. While previous studies have examined the impact on the 
gut microbiome of supplement-based VLEDs alone,17 as well as 
supplement- and food-based weight loss interventions in general 
(not VLEDs),20,21 we directly investigated the role of diet format 
within a VLED context. We observed differential impacts of these 
two types of VLEDs on alpha diversity, with consistent increases 
in the food-based group across both the Shannon index and rich­
ness, while the supplement-based group showed non-significant 
decreases, reflecting divergent ecological responses. A recent 
meta-analysis reported a non-significant pooled increase in 
alpha diversity following food-based weight loss interventions,22
aligning in direction with the statistically significant increases in 
the Shannon index and richness observed in our food-based 
VLED group. In contrast, supplement-based diets have been 
associated with statistically significant increases in alpha diver­
sity in prior single-arm studies,22 which diverges from the 
modest, non-significant reductions observed in our supple­
ment-based VLED group, despite similar weight loss. Notably, 
the between-group difference in the Shannon index persisted 
across sensitivity analyses and was even more pronounced after 
adjustment for alcohol intake, suggesting the robustness of our 
findings and the potential influence of other non-study foods or 
drinks. As both groups in our study achieved comparable weight 
loss, the divergent alpha diversity trajectories are more likely 
attributable to diet format than to energy restriction alone. Our 
exploratory decomposition of the Shannon index provides 
some insight into potential underlying contributors to this diver­
gence. In both groups, most top-contributing species were sub­
dominant at baseline, consistent with previous findings that low- 
abundance taxa contribute substantially to diversity shifts.23
However, based on visual inspection, the distribution of these 
contributions appeared to differ between groups. In the supple­
ment-based group, changes to the Shannon index (i.e., alpha di­
versity) were largely accounted for by a small number of species, 
with limited influence from the broader microbial community. 
Conversely, the food-based group exhibited a broader pattern 
of change, suggesting that shifts extended beyond the top 30 
species and were more evenly distributed across the community. 
This may reflect greater evenness and/or the emergence of pre­
viously low-abundance species not captured among the top con­
tributors, in response to the food-based intervention. These find­
ings underscore the value of decomposing diversity metrics to 
disentangle community-level shifts from the contributions of 
Table 2. Modified intention-to-treat (n = 45) analyses of the within-group changes from baseline, and between-group differential 
changes over time, in bacterial genera and families after consuming a food-based versus supplement-based very-low-energy diet for 
3 weeks
Taxon
Level
Food-based VLED 
within-group change 
from baseline to 
week 3, (mean, 
95% CI)
Supplement-based 
VLED within-group 
change from baseline 
to week 3, (mean, 
95% CI)
Between-group differential 
change from baseline to 
week 3 (unadjusted β, 
95%CI)
p value
q value
GCA_900066905
genus
−0.30 (−0.54 to −0.07)
0.90 (0.32–1.49)
−1.21 (−1.82 to −0.60)
0.000
0.056
Lachnospira
genus
0.55 (−0.08 to 1.19)
−1.67 (−2.68 to −0.66)
2.17 (0.99–3.34)
0.001
0.056
Anaerostipes
genus
−0.41 (−1.01 to 0.20)
1.34 (0.58–2.09)
−1.73 (−2.67 to −0.79)
0.001
0.056
Lachnospiraceae_ 
MIC9331
genus
−0.08 (−0.31 to 0.16)
1.27 (0.50–2.04)
−1.35 (−2.13 to −0.57)
0.001
0.079
Ruminococcus_A
genus
−0.86 (−1.35 to −0.37)
0.19 (−0.23 to 0.62)
−1.05 (−1.68 to −0.43)
0.002
0.079
UBA11774
genus
1.58 (0.54–2.61)
−0.59 (−1.45 to 0.26)
2.12 (0.81–3.43)
0.002
0.079
CAG_103
genus
0.15 (−0.51 to 0.82)
−1.70 (−2.71 to −0.69)
1.91 (0.73–3.10)
0.002
0.079
Clostridium_A
genus
0.26 (−0.47 to 1.00)
2.06 (1.19–2.92)
−1.78 (−2.89 to −0.68)
0.002
0.079
Christensenellales_ 
GCA_900066905
family
−0.33 (−0.60 to −0.06)
0.87 (0.29–1.44)
−1.19 (−1.81 to −0.57)
0.000
0.028
The taxa presented in the table (n = 9) are those with evidence (q < 0.1) of between-group differential changes over time, based on the modified inten­
tion-to-treat analyses. 
Note: unadjusted model: ∼group*time point. 
To avoid ambiguity in taxa sharing the identifier GCA_900066905, we refer to the genus-level taxon as GCA_900066905 and the corresponding family- 
level taxon as Christensenellales_GCA_900066905. 
Adjusted q value: the p value adjusted for multiple comparisons using the Benjamini-Hochberg procedure. 
Abbreviations: β, beta-coefficient; CI, confidence interval.
6 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
specific taxa. They also suggest that the food-based VLED may 
have promoted broader microbial restructuring compared to 
the supplement-based VLED.
In our study, we also observed shifts in beta diversity in both 
groups, with a greater shift observed in the supplement-based 
group. While there is inconsistent evidence of the effects of sup­
plement-based VLEDs or food-based weight loss diets on gut 
microbiome beta diversity,17,20,21 the greater shift in the supple­
ment-based group reflects a more pronounced compositional 
response to this type of diet. In contrast, the smaller beta diver­
sity shift in the food-based group suggests internal restructuring 
toward more diverse and evenly distributed communities, 
without marked divergence between individuals. Although beta 
diversity does not directly measure microbial stability, the 
smaller shifts in the food-based group indicate a broader micro­
bial response and potentially less ecological disruption. 
Together, the alpha and beta diversity findings reflect funda­
mental differences in food processing between the two VLEDs, 
such as food matrix complexity, fiber diversity, or additive 
exposure, warranting further investigation. They also highlight 
the potential of food-based VLEDs to promote a more favorable 
microbial profile even under substantial energy restriction.
Taxonomic differences and clinical implications
During the 3-week study period, we observed between-group 
differential changes in eight genera and one family. For 12 of 
r = −0.37, p = 0.024
−100
−50
0
−1.0
−0.5
0.0
0.5
1.0
1.5
Change in Shannon Index
Change in Leptin Concentration
group
FB−VLED
SB−VLED
Leptin Concentration vs. Shannon
r = 0.31, p = 0.059
−50
0
50
−1.0
−0.5
0.0
0.5
1.0
1.5
Change in Shannon Index
Change in Bloating/Flatulence
group
FB−VLED
SB−VLED
Bloating/Flatulence vs. Shannon
r = 0.31, p = 0.061
−30
0
30
60
−1.0
−0.5
0.0
0.5
1.0
1.5
Change in Shannon Index
Change in Abdominal Pain
group
FB−VLED
SB−VLED
Abdominal Pain vs. Shannon
Figure 3. Associations between changes in 
species-level Shannon index and changes 
in clinical outcomes over the 3-week inter­
vention period 
Scatterplots show Spearman correlation co­
efficients (r) and corresponding p values between 
changes in gut microbial alpha diversity (Shannon 
index) and changes in leptin concentration, 
bloating/flatulence symptoms, and abdominal 
pain symptoms. Each point represents one bio­
logical replicate (the paired change in an individual 
participant). No technical replicates were per­
formed. Data were pooled across intervention 
groups (food-based very-low-energy diet [FB- 
VLED, purple]; supplement-based very-low-en­
ergy diet [SB-VLED, blue]).
these genera and one family, within- 
group estimates showed opposite pat­
terns of change between the food- and 
supplement-based groups, mirroring the 
divergence observed in alpha diversity. 
As noted in the results section, we 
focused on the three most significant 
and well-characterized genera (Lachno­
spira, Anaerostipes, and Ruminococ­
cus_A) to enhance interpretability and 
avoid overinterpreting changes in less 
well-defined taxa.
The between-group difference for 
Lachnospira appeared to be driven by 
the more substantial, statistically signifi­
cant decrease in the supplement-based group, versus the 
smaller, non-significant increase in the food-based group. Lach­
nospira is a pectinolytic genus that ferments complex plant- 
derived polysaccharides such as pectin and polygalacturonic 
acid24—dietary fibers abundant in fruits, vegetables, and le­
gumes but largely absent from highly processed foods. Its abun­
dance is directly associated with greater fruit and vegetable 
intake25–27 and is typically higher in healthier populations.28–30
The marked reduction in Lachnospira in the supplement-based 
group may reflect reduced availability of fermentable substrates 
in that diet, while the preservation in the food-based group may 
indicate protection of fiber-degrading niches through greater fi­
ber diversity and intact plant matrices. These shifts reinforce 
the possibility that certain fiber-degrading species were better 
maintained in the context of the food-based VLED, suggesting 
that lower levels of food processing may help preserve fiber-de­
grading taxa relevant to microbial resilience.
Anaerostipes is a butyrate-producing genus within the Lach­
nospiraceae family that ferments carbohydrates into short-chain 
fatty acids31 such as acetate, lactate, and butyrate32–34—metab­
olites that contribute to gut barrier integrity and have anti-inflam­
matory effects.35,36 In our study, the between-group difference 
for Anaerostipes appeared to be primarily influenced by the 
more pronounced, statistically significant increase in the supple­
ment-based group, compared to the smaller, non-significant 
decrease in the food-based group. The increase in Anaerostipes 
Cell Reports Medicine 6, 102417, October 21, 2025 
7 
Article
ll
OPEN ACCESS
within the supplement-based group may be explained by the in­
clusion of inulin—a fermentable prebiotic fiber known to selec­
tively promote this genus in feeding studies.37 While butyrate 
production is generally considered beneficial,35,36 this increase 
occurred alongside broader reductions in diversity in the 
supplement-based group, suggesting that targeted enrichment 
of specific genera may not compensate for wider ecological dis­
ruptions. The non-significant decrease in Anaerostipes in the 
food-based group may reflect lower levels of isolated prebiotic 
fibers, but this was offset by broader preservation of microbial 
diversity, highlighting the complexity of dietary influences on mi­
crobial ecology.
Ruminococcus_A is a taxonomically diverse genus known for 
fermenting complex carbohydrates into various fermentation 
products including acetate, formate, and succinate.38 Some spe­
cies within this genus also degrade host-derived mucins,39 poten­
tially influencing gut barrier function. In our study, the food-based 
group experienced a statistically significant decrease in Rumino­
coccus_A, while the supplement-based group showed a small, 
non-significant increase. Although the functional significance of 
this genus remains uncertain due to its heterogeneity, some Rumi­
nococcus taxa have been linked to inflammatory diseases,40
whereas others are considered fiber degraders.41,42 The reduction 
observed in the food-based group may therefore reflect suppres­
sion of potentially mucin-degrading or pro-inflammatory strains, 
although further strain-level resolution is needed. These findings 
add to the growing recognition that shifts in gut microbiome 
composition must be interpreted in the context of both functional 
potential and broader community dynamics.
While several species and predicted functional pathways 
showed between-group differences, none passed correction 
for multiple comparisons and are not discussed further due to 
high false-positive risk in small-sample, high-dimensional set­
tings. Similarly, no between-group differences were observed 
for MetaCyc alpha diversity, and no functional pathways passed 
multiple testing correction, which may reflect the well-recog­
nized functional redundancy of the gut microbiome, where taxo­
nomic shifts do not always lead to measurable differences in 
SB−VLED
FB−VLED
0.0
0.1
0.2
0.3
0.0
0.1
0.2
0.3
s__Agathobacter.rectale
s__Bifidobacterium.adolescentis
s__Bacteroides.eggerthii
s__Agathobacter.faecis
s__Blautia_A.massiliensis
s__Roseburia.intestinalis
s__Bifidobacterium.longum
s__Ruminococcus_E.bromii_B
s__Bacteroides_B.massiliensis
s__Gemmiger.sp003476825
s__Ruminococcus_A.sp003011855
s__Faecalibacterium.prausnitzii_K
s__Dorea.longicatena
s__Bacteroides.stercoris
s__Lachnospira.sp900316325
s__Alistipes.putredinis
s__CAG.313.sp003539625
s__Bacteroides_B.dorei
s__Bacteroides.thetaiotaomicron
s__Methanobrevibacter_A.smithii
s__Bacteroides.caccae
s__Parabacteroides.distasonis
s__Barnesiella.intestinihominis
s__Blautia_A.obeum
s__PeH17.sp000435055
s__Akkermansia.muciniphila
s__Akkermansia.muciniphila_B
s__Lachnospira.eligens_B
s__UBA11524.sp000437595
s__Bacteroides_B.vulgatus
Other_net
s__Agathobacter.faecis
s__Agathobacter.rectale
s__Ruminococcus_E.bromii_B
s__Ruminococcus_D.bicirculans
s__Roseburia.intestinalis
s__Gemmiger.sp003476825
s__Fusicatenibacter.saccharivorans
Other_net
s__Dorea.longicatena
s__Faecalibacterium.prausnitzii_D
s__Roseburia.inulinivorans
s__Bacteroides.xylanisolvens
s__Faecalibacterium.prausnitzii_K
s__Blautia_A.massiliensis
s__Bacteroides.fragilis
s__Bacteroides.eggerthii
s__Bacteroides_A.plebeius_A
s__Escherichia.coli
s__Bacteroides_B.dorei
s__Methanobrevibacter_A.smithii
s__UBA1417.sp003531055
s__CAG.74.MIC8062
s__Eisenbergiella.tayi
s__Parabacteroides.distasonis
s__Escherichia.flexneri
s__Ruthenibacterium.lactatiformans
s__Parabacteroides.goldsteinii
s__Bacteroides.uniformis
s__Bacteroides.thetaiotaomicron
s__Akkermansia.muciniphila
s__Anaerostipes.hadrus
Change in Shannon Contribution (Final − Baseline)
Taxon / Grouped Contribution
Group
FB−VLED
Other_net
SB−VLED
Top Taxa and 'Other Net' Contribution to Shannon Index Changes
Figure 4. Species-level decomposition of the Shannon index over the 3-week intervention period 
Bar plots show the 30 species with the largest absolute changes in contribution to Shannon diversity, alongside the grouped contribution of all other taxa 
(‘‘Other_net’’), separately for each group (supplement-based very-low-energy diet [SB-VLED, blue]; food-based very-low-energy diet [FB-VLED, purple]). 
Positive values indicate increased contribution to Shannon diversity at week 3 relative to baseline, and negative values indicate decreased contribution. Species 
are ordered by net change within each group. Bars represent group-level averages calculated across all biological replicates (individual participants). No 
technical replicates were performed.
8 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
predicted functional capacity.43 At a broader taxonomic level, 
post hoc analyses showed that phylum-level profiles remained 
relatively stable across both groups and time points, with no 
marked shifts in dominant phyla observed. These limitations 
highlight the need for larger, longer-term studies to validate 
and extend the current findings.
Clinical outcomes and implications
Preliminary analyses revealed between-group differences in hip 
circumference, constipation symptoms, and albumin concentra­
tion before correction for multiple comparisons. Specifically, the 
food-based group showed a smaller and non-significant reduc­
tion in hip circumference, a marginal and non-significant 
improvement in constipation symptoms, and a less pronounced 
increase in albumin concentration. In contrast, the supplement- 
based group exhibited statistically significant and more pro­
nounced changes in all three parameters, namely a greater 
reduction in hip circumference, a worsening in constipation 
symptoms, and a larger increase in albumin concentration. 
However, none of these differences remained significant after 
multiple comparisons correction, likely reflecting the nature of 
the interventions (both active rather than active versus placebo), 
the short study duration, small sample size, and low baseline 
symptom burden in this non-clinical sample. Although our trial 
was not powered for secondary clinical outcomes, we highlight 
these three outcomes with nominal statistical significance as po­
tential signals, explored below as hypothesis-generating leads. 
Notably, adverse events were twice as common in the supple­
ment-based group, suggesting broader health impacts that 
may not have been captured by our selected measures and war­
rant further investigation.
Hip circumference, which reflects both adiposity and lower- 
body muscle mass, is inversely associated with cardiometabolic 
risk and premature mortality according to observational 
studies.44–46 The relatively minor reduction in hip circumference 
in the food-based group may suggest better preservation of 
lower-body muscle mass and/or lower reduction in gluteal fat 
mass compared to the supplement-based group, which showed 
a larger reduction. While further body composition data are 
required, preservation of muscle mass aligns with prior evidence 
SB−VLED
Baseline
SB−VLED
Final
FB−VLED
Baseline
FB−VLED
Final
BBC2365
BBC2385
BBC2387
BBC2395
BBC2402
BBC2405
BBC2413
BBC2419
BBC2424
BBC2426
BBC2432
BBC2433
BBC2436
BBC2441
BBC2442
BBC2450
BBC2451
BBC2455
BBC2457
BBJ4933
BBJ4996
BBJ5029
BBC2372
BBC2390
BBC2394
BBC2399
BBC2404
BBC2406
BBC2409
BBC2410
BBC2411
BBC2414
BBC2417
BBC2420
BBC2421
BBC2422
BBC2435
BBC2448
BBC2458
BBJ4941
BBJ4942
BBJ5020
BBC2367
BBC2368
BBC2373
BBC2376
BBC2377
BBC2378
BBC2379
BBC2382
BBC2384
BBC2396
BBC2397
BBC2415
BBC2416
BBC2437
BBC2438
BBC2444
BBC2446
BBC2454
BBC2460
BBC2461
BBJ4904
BBJ4925
BBJ4927
BBC2381
BBC2386
BBC2389
BBC2400
BBC2401
BBC2407
BBC2412
BBC2423
BBC2425
BBC2427
BBC2434
BBC2439
BBC2447
BBC2452
BBC2453
BBC2456
BBJ4899
BBJ4916
BBJ4940
0.00
0.25
0.50
0.75
1.00
Sample
Relative Abundance
phylum
Actinobacteriota
Bacteroidota
Blastocystidae_phylum
Campylobacterota
Cyanobacteria
Desulfobacterota_A
Euryarchaeota
Firmicutes
Firmicutes_A
Firmicutes_B
Firmicutes_C
Proteobacteria
Synergistota
Verrucomicrobiota
Figure 5. Relative abundance of major gut microbial phyla, stratified by group (SB-VLED; FB-VLED; purple) and time point (baseline; week 3 
[final]) 
Each stacked bar represents one biological replicate (an individual participant sample) after compositional transformation. Colors denote phylum identity as 
indicated in the legend. Taxonomy follows Microba’s database (mixture of older and Genome Taxonomy Database [GTDB] nomenclature); Firmicutes_A–C and 
Desulfobacterota_A are GTDB-defined clades within their respective former phyla. No technical replicates were performed.
Cell Reports Medicine 6, 102417, October 21, 2025 
9 
Article
ll
OPEN ACCESS
linking whole-food-derived protein to muscle maintenance47 and 
highly processed food intake to muscle loss.48 Similarly, the 
worsening of constipation symptoms in the supplement-based 
group, despite similar reported fiber intake, may reflect a limited 
diversity of soluble fiber sources, consistent with previous re­
ports of higher constipation prevalence on supplement-based 
VLEDs49 and evidence that fiber type and diversity influence 
gut function.50,51 Since the supplement-based VLED included 
vitamin- and mineral-fortified products, the smaller increase in 
albumin concentrations, a marker of liver function and nutritional 
status,52 in the food-based group may reflect differences in 
micronutrient content and fortification.52,53 Higher serum albu­
min concentrations can also reflect dehydration,54 which may 
partially account for the greater increases observed in the sup­
plement-based group, given that dehydration is a commonly 
reported side effect of supplement-based VLEDs. However, al­
bumin changes over short periods are typically minimal in 
healthy populations, making the clinical relevance unclear.
To further explore potential links between gut microbiome 
changes and clinical outcomes, we conducted post hoc correla­
tion analyses between changes in the species-level Shannon 
index and changes in selected outcomes. Small-to-moderate in­
verse correlations with leptin and positive correlations with 
bloating/flatulence and abdominal pain were observed. The sta­
tistically significant inverse correlation of alpha diversity with lep­
tin is particularly noteworthy, given leptin’s links to metabolic 
flexibility and inflammation,55 though causal pathways remain 
speculative. While the other correlations did not reach statistical 
significance (p > 0.05), their consistency in direction and magni­
tude with observed microbial and preliminary clinical shifts sug­
gests potential relevance that warrants further investigation.
Future directions
Our study adds to the growing evidence suggesting that changes 
in gut microbiota composition may represent one pathway 
through which diets high in ultra-processed foods (UPFs) affect 
health. While our study was not specifically designed to investi­
gate the impact of UPFs, as defined by the widely used Nova 
food classification system,11 the supplement- and food-based 
VLEDs can broadly be characterized as high and low UPF, 
respectively. Existing epidemiological evidence links high-UPF 
diets to poorer health outcomes, including greater risks of cardio­
vascular disease, type 2 diabetes, common mental disorders, and 
mortality, as well as gut-related conditions such as inflammatory 
bowel diseases, functional gastrointestinal disorders, and intesti­
nal cancers.12,56 However, an oft-cited limitation of this literature is 
the lack of causal and mechanistic evidence explaining these 
associations.57 In 2023, the UK’s Scientific Advisory Committee 
on Nutrition highlighted ongoing uncertainty as to whether the 
adverse health effects attributed to UPFs are independent of their 
typically poor nutritional profile and noted the scarcity of evidence 
in specific population subgroups.57 Our study helps address this 
gap by focusing on individuals with high BMI using two nutrition­
ally adequate VLEDs that differed in processing-related charac­
teristics. The supplement-based VLED was comprised almost 
entirely of highly processed industrial ingredients, including 
macronutrient isolates (e.g., calcium caseinate, sodium caseinate, 
and medium chain triglycerides) and food additives (e.g., emulsi­
fiers and nonsugar sweeteners such as citric and fatty acid esters 
of glycerol [472c] and aspartame [951], respectively). Conversely, 
the food-based VLED consisted primarily of ingredients contain­
ing intact, minimally processed foods that retained their physical 
structure (e.g., chopped and cooked vegetables, legumes, and 
whole grains), with minimal inclusion of heavily processed indus­
trial ingredients and none in approximately half of the meals. 
These differences may partly explain the observed divergence in 
gut microbial composition between groups, despite both diets’ 
similar nutrient and low-energy profiles, highlighting the gut micro­
biome as a potential intermediary between dietary processing 
level and health outcomes.
Emerging mechanistic evidence from in vitro and animal 
studies, and a small number of human studies, suggests that 
the characteristic ingredients of UPFs, such as emulsifiers, 
nonsugar sweeteners, colors, and micro- and nano-particles, 
may reduce gut microbial diversity, increase intestinal perme­
ability, and promote intestinal inflammation.12,58,59 In contrast, 
intact food structures and matrices, along with polyphenols, 
phytonutrients, and diverse fibers found in whole foods, may 
positively modulate the gut microbiome.60,61 While much of this 
research has focused on individual components, UPFs typically 
comprise complex mixtures of food additives and other industrial 
ingredients that may interact in cumulative or synergistic ways, 
while containing little to no whole foods. For example, a recent 
in vitro study showed that food additive mixtures commonly 
consumed in UPFs can elicit combined effects not detected in sin­
gle-substance testing.62 These findings underscore the need to 
consider the broader food matrix and processing-related features 
of diets when evaluating their health effects. Although research on 
UPF-rich diets and the human gut microbiome is expanding, few 
intervention studies have tested their effects under controlled 
conditions. Our study contributes to this effort by demonstrating 
that two VLEDs, designed to be isoenergetic and nutritionally 
matched but with varying processing characteristics, were asso­
ciated with distinct gut microbial responses. Further research is 
required to isolate the specific components responsible for these 
shifts, elucidate their mechanisms of action, and inform the design 
of microbiome-supportive dietary interventions using VLEDs.
Conclusion
Our study provides evidence of the differential impacts of food- 
versus supplement-based VLEDs on the diversity of the gut mi­
crobiome of women with high BMI, even over a short 3-week 
intervention. Despite similar weight loss and intakes of carbohy­
drate, protein, sugar, sodium, and fiber, participants in the food- 
based group experienced significantly greater increases in spe­
cies-level alpha diversity (Shannon index and richness), smaller 
shifts in beta diversity, and compositional changes consistent 
with enhanced preservation of fiber-degrading and health-asso­
ciated taxa, compared to the supplement-based group. These 
findings suggest that food-based VLEDs, richer in whole-food 
components and lower in highly processed industrial ingredi­
ents, may more favorably modulate gut microbiome composition 
compared to supplement-based alternatives. Further investiga­
tion in adequately powered, longer-term trials is needed to 
confirm these findings and identify the specific dietary elements 
responsible for the observed effects.
10 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
Limitations of the study
Several limitations must be considered when interpreting our 
findings. While the single-blind nature of the trial likely had little 
impact on the objective gut microbiome measures, knowledge 
of intervention allocation may have influenced self-reported 
gastrointestinal and mental health symptoms. Our study was 
conducted in women with a BMI at or above 30 kg/m2 and 
over a relatively short time frame, reducing generalizability to 
other populations and typical VLED durations of 12 weeks or 
more. Broader generalizability will require studies with more 
diverse populations and extended follow-up, particularly to 
assess the durability of observed effects.
We randomized 47 participants instead of the planned 40 to 
address missing data, withdrawals, and loss to follow-up; how­
ever, mITT principles were maintained, and LMER models, 
which partially mitigate bias from missing data, showed consis­
tent results aligned with complete case analyses. The large 
number of outcomes assessed increases the potential for 
type I errors; however, primary and secondary outcomes 
were grouped into related categories, and adjustments for mul­
tiple comparisons were made for each analysis. There were 
baseline differences between the groups in country of birth, 
marital status, employment, medication use, stool consistency, 
BMI, physical activity, menopause status, and alcohol intake, 
likely due to the small sample size that was not large enough 
to balance all prognostic covariates. While sensitivity analyses 
adjusting for these prognostic covariates yielded similar results 
to the primary analysis, larger trials are required to confirm 
findings.
Dietary adherence was measured by self-reported intake us­
ing the Easy Diet Diary app. Although we employed strategies 
to maximize adherence, self-reported dietary assessment 
methods have limitations that may have influenced results.63
Future inpatient or metabolic ward settings may help to address 
these limitations. While the two VLED formats were intended to 
be isocaloric and nutritionally equivalent, the food-based group 
reported ∼150 kcal/d higher average energy intake and ∼18 
g/d higher fat intake compared to the supplement-based group. 
Notably, higher dietary fat intake is often associated with lower 
microbial diversity,64,65 yet our food-based group showed 
increased alpha diversity. Conversely, the supplement-based 
group had lower dietary fat intake yet showed non-significant re­
ductions in alpha diversity. This finding is somewhat consistent 
with a prior supplement-based VLED study that reported similar 
patterns despite lower fat and energy intake compared to a 
higher-energy, higher-fat conventional weight-loss diet phase.66
These findings suggest that energy and fat intake alone may not 
explain microbiome shifts. Lastly, other program differences 
included a recommended restriction on nonsugar sweeteners 
in the food-based group and the inclusion of the probiotic spe­
cies Lactobacillus plantarum in some of that group’s discre­
tionary snack items. Although present at less than 1% in those 
options and not detected in any participant fecal samples, sug­
gesting limited probiotic exposure, the possible implications of 
these dietary differences on our findings remain unclear.
In terms of clinical relevance, our study was not powered for 
clinical endpoints, and the primary outcomes were gut micro­
biome measures rather than clinical outcomes. Preliminary evi­
dence suggested nominal between-group differences in hip 
circumference, constipation symptoms, and albumin concentra­
tion; however, none remained statistically significant after 
correction for multiple comparisons, underscoring the limited 
clinical utility of these findings at this stage. The higher rate of 
adverse events in the supplement-based group likewise repre­
sents a hypothesis-generating signal only. More broadly, our trial 
reflects the current state of the field, in which promising micro­
biome science has yet to be translated into routine clinical 
care. As highlighted in a recent perspective (‘‘The microbiome 
for clinicians’’), multiple biological, methodological, and transla­
tional barriers have prevented the integration of microbiome 
endpoints into clinical medicine.67 Although we followed the 
standardized microbiome reporting initiative STORMS (Stren­
gthening the Organizing and Reporting of Microbiome Studies), 
alongside CONSORT (Consolidated Standards of Reporting Tri­
als)68,69 (see STAR Methods), to enhance the translatability and 
reproducibility of our work, larger, longer-term, and clinically 
focused trials will be required to determine whether the gut mi­
crobiome differences observed here translate into meaningful 
patient outcomes.
RESOURCE AVAILABILITY
Lead contact
Requests for further information and resources should be directed to and will 
be fulfilled by the lead contact, Dr. Melissa M. Lane (m.lane@deakin.edu.au).
Materials availability
This study did not generate new unique reagents.
Data and code availability
• Raw shotgun metagenomic sequencing data are stored on Deakin 
University’s Research Data Store, a secure institutional repository 
managed by Deakin University. These data cannot be deposited in an 
external public repository because participant consent for open sharing 
was not obtained at recruitment. Participant-level data can be re­
quested from the lead contact with a detailed research proposal outlin­
ing the intended use of the data. Data access will only be provided once 
an appropriate Human Research Ethics Committee (HREC) or equiva­
lent institutional approval is in place and will be managed in line with 
Barwon Health (primary sponsor) and Deakin University (secondary 
sponsor) governance requirements.
• This paper does not report original code.
• Any additional information required to reanalyze the data reported in this 
work is available from the lead contact upon request.
ACKNOWLEDGMENTS
We would like to thank Dr. Jerry Lai for their assistance with REDCap, Dr. 
Jessica Davis and Samantha Kernaghan for their assistance with data collec­
tion, and Chantelle Erwin for their assistance with data cleaning of dietary 
data. Be Fit Food provided financial support for the collection and analysis 
of fecal gut microbiome and biochemical samples. Investigational products 
used in the trial were also supplied in kind by Be Fit Food. The study was 
supported and sponsored by Barwon Health (primary sponsor) and Deakin 
University (secondary sponsor), with additional support from a Deakin Uni­
versity PhD stipend. Be Fit Food had no role in the study design, collection, 
analysis, or interpretation of data; the writing of the report; or the decision to 
submit the paper for publication. The academic sponsors (Barwon Health 
and Deakin University) had no role beyond their institutional and ethical over­
sight responsibilities.
Cell Reports Medicine 6, 102417, October 21, 2025 
11 
Article
ll
OPEN ACCESS
Notably, lower dietary fat intake is often associated with lower an appropriate Human Research Ethics Committee (HREC) or equiva-
AUTHOR CONTRIBUTIONS
Conceptualization, M.M.L., M.M., A.L., A.O’.N., W.M., and F.N.J. Data cura­
tion, M.M.L. and A.J.M. Formal analysis, M.M.L., A.J.M., M.M., M.L., A.L., 
and M.O’.H. Funding acquisition, F.N.J. and W.M. Investigation, A.J.M., 
J.B., and L.S. Methodology, M.M.L., M.M., A.L., A.O’.N., W.M., and F.N.J. 
Project administration, M.M.L. and A.J.M. Resources, R.P. and S.B. Supervi­
sion, M.M., A.L., A.O’.N., M.K., M.B., W.M., and F.N.J. Visualization, A.M., M. 
M., M.L., and M.O’.H. Writing—original draft, M.M.L. and A.J.M. Writing—re­
view & editing: all authors.
DECLARATION OF INTERESTS
M.M.L. is a committee member (2019–present) and former secretary (2022– 
2024) of the Melbourne Branch Committee of the Nutrition Society of Australia 
(both unpaid) and has received travel funding support from the International 
Society for Nutritional Psychiatry Research, the Nutrition Society of Australia, 
the Australasian Society of Lifestyle Medicine, and the Gut Brain Congress. A. 
J.M. is immediate past secretary for the International Society for Nutritional 
Psychiatry Research (unpaid) and is funded through the National Health and 
Medical Research Council (NHMRC)-supported CREDIT CRE, the Center for 
Research Excellence for the Development of Innovative Therapies. M.B. is 
supported by an NHMRC Senior Principal Research Fellowship (1156072). 
M.B. has received grant/research support from the NIH, Cooperative 
Research Center, Simons Autism Foundation, Cancer Council of Victoria, 
Stanley Medical Research Foundation, Medical Benefits Fund, National Health 
and Medical Research Council, Medical Research Futures Fund, Beyond Blue, 
Rotary Health, the a2 Milk Company, Meat and Livestock Board, Woolworths 
Limited, and Avant and the Harry Windsor Foundation and has been a speaker 
for Abbot, Astra Zeneca, Janssen and Janssen, and Lundbeck and Merck and 
served as a consultant to Allergan, Astra Zeneca, Bioadvantex, Bionomics, 
Collaborative Medicinal Development, Janssen and Janssen, Lundbeck and 
Merck, Pfizer, and Servier—all unrelated to this work. A.L. has received grant, 
research, or travel support from Deakin University, the University of Mel­
bourne, RMIT University, the National Health and Medical Research Council, 
Australian Academy of Science, the Jack Brockhoff Foundation, Epilepsy 
Foundation of Australia, and American Epilepsy Society and has received 
speakers’ honoraria from the European Space Agency and Swisse 
Australia—all unrelated to this work. A.L. is a named inventor on a patent 
relating to Prevotella. M.M. has received grant/research support from the 
NHMRC, Deakin University School of Medicine, Deakin Biostatistics Unit, 
Institute for Mental and Physical Health and Clinical Translation, Stroke Foun­
dation, and Medibank Health Research Fund. W.M. is currently funded by an 
NHMRC Investigator grant (#2008971). W.M. has previously received funding 
from the Cancer Council Queensland and university grants/fellowships from La 
Trobe University, Deakin University, the University of Queensland, and Bond 
University. W.M. has received funding and/or has attended events funded 
by Cobram Estate Pty. Ltd. and Bega Dairy and Drinks Pty Ltd. W.M. has 
received travel funding from the Nutrition Society of Australia. W.M. has 
received consultancy funding from Nutrition Research Australia and Parachu­
teBH. W.M. has received speakers’ honoraria from VitaFoods, the Cancer 
Council Queensland, and the Princess Alexandra Research Foundation. F.N. 
J. has received competitive grant/research support from the Brain and 
Behavior Research Institute, the NHMRC, Australian Rotary Health, the Gee­
long Medical Research Foundation, the Ian Potter Foundation, and the Univer­
sity of Melbourne; industry support for research from Meat and Livestock 
Australia, Woolworths Limited, the a2 Milk Company, and Be Fit Food; philan­
thropic support from the Fernwood Foundation, Wilson Foundation, the JTM 
Foundation, the Serp Hills Foundation, the Roberts Family Foundation, and 
the Waterloo Foundation; and travel support and speakers’ honoraria from Sa­
nofi-Synthe´ labo, Janssen Cilag, Servier, Pfizer, Network Nutrition, Angelini 
Farmace´ utica, Eli Lilly, and Metagenics. F.N.J. has written two books for com­
mercial publication. She is on the scientific advisory board of the Dauten Fam­
ily Center for Bipolar Treatment Innovation and Zoe Limited. She is currently 
supported by an NHMRC Investigator grant L1 (#1194982). The Food & 
Mood Center has received grant/research support from the a2 Milk Company, 
Be Fit Food, Meat and Livestock Australia, and Woolworths Limited and phil­
anthropic support from the Fernwood Foundation, Wilson Foundation, the 
JTM Foundation, the Serp Hills Foundation, the Roberts Family Foundation, 
and the Waterloo Foundation. A.O’N. receives funding support from the 
NHMRC (#2009295).
DECLARATION OF GENERATIVE AI AND AI-ASSISTED 
TECHNOLOGIES IN THE WRITING PROCESS
During the preparation of this work, the authors used ChatGPT to assist with 
refining text and statistical code. After using this tool, the authors reviewed 
and edited the content as needed and take full responsibility for the final con­
tent of the publication.
STAR★METHODS
Detailed methods are provided in the online version of this paper and include 
the following:
• KEY RESOURCES TABLE
• EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS
• METHOD DETAILS
○Recruitment and participants
○Randomization, allocation, and blinding
○Interventions
○Sample size
○Sample collection
○DNA extraction
○Library preparation
○Shotgun metagenomic sequencing
○Species profiles and functional potential
○Data preparation
○Alpha diversity, beta diversity, taxonomic composition, and func­
tional potential
○Secondary clinical outcomes
• QUANTIFICATION AND STATISTICAL ANALYSIS
○Post hoc exploratory analyses
• ADDITIONAL RESOURCES
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j. 
xcrm.2025.102417.
Received: November 27, 2024
Revised: May 28, 2025
Accepted: September 19, 2025
Published: October 10, 2025
REFERENCES
1. Lynch, S.V., and Pedersen, O. (2016). The human intestinal microbiome 
in health and disease. N. Engl. J. Med. 375, 2369–2379. https://doi.org/ 
10.1056/NEJMra1600266.
2. Rothschild, D., Weissbrod, O., Barkan, E., Kurilshikov, A., Korem, T., 
Zeevi, D., Costea, P.I., Godneva, A., Kalka, I.N., Bar, N., et al. (2018). 
Environment dominates over host genetics in shaping human gut micro­
biota. Nature 555, 210–215. https://doi.org/10.1038/nature25973.
3. Flint, H.J., Duncan, S.H., and Louis, P. (2017). The impact of nutrition on 
intestinal bacterial communities. Curr. Opin. Microbiol. 38, 59–65. 
https://doi.org/10.1016/j.mib.2017.04.005.
4. Wastyk, H.C., Fragiadakis, G.K., Perelman, D., Dahan, D., Merrill, B.D., 
Yu, F.B., Topf, M., Gonzalez, C.G., Van Treuren, W., Han, S., et al. 
(2021). Gut-microbiota-targeted diets modulate human immune status. 
Cell 184, 4137–4153.e14. https://doi.org/10.1016/j.cell.2021.06.019.
5. David, L.A., Maurice, C.F., Carmody, R.N., Gootenberg, D.B., Button, J. 
E., Wolfe, B.E., Ling, A.V., Devlin, A.S., Varma, Y., Fischbach, M.A., et al. 
12 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
(2014). Diet rapidly and reproducibly alters the human gut microbiome. 
Nature 505, 559–563. https://doi.org/10.1038/nature12820.
6. Fu, J., Xu, K., Ni, X., Li, X., Zhu, X., and Xu, W. (2022). Habitual dietary fi­
ber intake, fecal microbiota, and hemoglobin A1c level in Chinese pa­
tients with type 2 diabetes. Nutrients 14, 1003. https://doi.org/10.3390/ 
nu14051003.
7. Ghosh, T.S., Rampelli, S., Jeffery, I.B., Santoro, A., Neto, M., Capri, M., 
Giampieri, E., Jennings, A., Candela, M., Turroni, S., et al. (2020). Medi­
terranean diet intervention alters the gut microbiome in older people 
reducing frailty and improving health status: the NU-AGE 1-year dietary 
intervention across five European countries. Gut 69, 1218–1228. 
https://doi.org/10.1136/gutjnl-2019-319654.
8. Vangay, P., Johnson, A.J., Ward, T.L., Al-Ghalith, G.A., Shields-Cutler, R. 
R., Hillmann, B.M., Lucas, S.K., Beura, L.K., Thompson, E.A., Till, L.M., 
et al. (2018). US immigration westernizes the human gut microbiome. 
Cell 175, 962–972.e10. https://doi.org/10.1016/j.cell.2018.10.029.
9. O’Keefe, S.J., Li, J.V., Lahti, L., Ou, J., Carbonero, F., Mohammed, K., 
Posma, J.M., Kinross, J., Wahl, E., Ruder, E., et al. (2015). Fat, fibre 
and cancer risk in African Americans and rural Africans. Nat. Commun. 
6, 6342. https://doi.org/10.1038/ncomms7342.
10. Walker, A.W., Ince, J., Duncan, S.H., Webster, L.M., Holtrop, G., Ze, X., 
Brown, D., Stares, M.D., Scott, P., Bergerat, A., et al. (2011). Dominant 
and diet-responsive groups of bacteria within the human colonic micro­
biota. ISME J. 5, 220–230. https://doi.org/10.1038/ismej.2010.118.
11. Monteiro, C.A., Cannon, G., Levy, R.B., Moubarac, J.C., Louzada, M.L., 
Rauber, F., Khandpur, N., Cediel, G., Neri, D., Martinez-Steele, E., et al. 
(2019). Ultra-processed foods: what they are and how to identify them. Public 
Health Nutr. 22, 936–941. https://doi.org/10.1017/S1368980018003762.
12. Whelan, K., Bancil, A.S., Lindsay, J.O., and Chassaing, B. (2024). Ultra- 
processed foods and food additives in gut health and disease. Nat. 
Rev. Gastroenterol. Hepatol. 21, 406–427. https://doi.org/10.1038/ 
s41575-024-00893-5.
13. Maki, K.A., Sack, M.N., and Hall, K.D. (2024). Ultra-processed foods: 
increasing the risk of inflammation and immune dysregulation? 
Nat. Rev. Immunol. 24, 453–454. https://doi.org/10.1038/s41577-024- 
01049-x.
14. Martini, D., Godos, J., Bonaccio, M., Vitaglione, P., and Grosso, G. 
(2021). Ultra-processed foods and nutritional dietary profile: a meta-anal­
ysis of nationally representative samples. Nutrients 13, 3390. https://doi. 
org/10.3390/nu13103390.
15. Mustajoki, P., and Pekkarinen, T. (2001). Very low energy diets in the 
treatment of obesity. Obes. Rev. 2, 61–72. https://doi.org/10.1046/j. 
1467-789X.2001.00026.x.
16. Sellahewa, L., Khan, C., Lakkunarajah, S., and Idris, I. (2017). A system­
atic review of evidence on the use of very low calorie diets in people 
with diabetes. Curr. Diabetes Rev. 13, 35–46. https://doi.org/10.2174/ 
1573399812666151005123431.
17. Lane, M., Howland, G., West, M., Hockey, M., Marx, W., Loughman, A., 
O’Hely, M., Jacka, F., and Rocks, T. (2020). The effect of ultra-processed 
very low-energy diets on gut microbiota and metabolic outcomes in indi­
viduals with obesity: A systematic literature review. Obes. Res. Clin. 
Pract. 14, 197–204. https://doi.org/10.1016/j.orcp.2020.04.006.
18. Link, V.M., Subramanian, P., Cheung, F., Han, K.L., Stacy, A., Chi, L., 
Sellers, B.A., Koroleva, G., Courville, A.B., Mistry, S., et al. (2024). Differ­
ential peripheral immune signatures elicited by vegan versus ketogenic 
diets in humans. Nat. Med. 30, 560–572. https://doi.org/10.1038/ 
s41591-023-02761-2.
19. Hall, K.D., Ayuketah, A., Brychta, R., Cai, H., Cassimatis, T., Chen, K.Y., 
Chung, S.T., Costa, E., Courville, A., Darcey, V., et al. (2019). Ultra-pro­
cessed diets cause excess calorie intake and weight gain: an inpatient 
randomized controlled trial of ad libitum food intake. Cell Metab. 30, 
67–77.e3. https://doi.org/10.1016/j.cmet.2019.05.008.
20. Jian, C., Silvestre, M.P., Middleton, D., Korpela, K., Jalo, E., Broderick, 
D., de Vos, W.M., Fogelholm, M., Taylor, M.W., Raben, A., et al. (2022). 
Gut microbiota predicts body fat change following a low-energy diet: a 
PREVIEW intervention study. Genome Med. 14, 54. https://doi.org/10. 
1186/s13073-022-01053-7.
21. Bliesner, A., Eccles-Smith, J., Bates, C., Hayes, O., Ho, J.Y., Martins, C., 
Truby, H., and Nitert, M.D. (2022). Impact of food-based weight loss in­
terventions on gut microbiome in individuals with obesity: a systematic 
review. Nutrients 14, 1953. https://doi.org/10.3390/nu14091953.
22. Koutoukidis, D.A., Jebb, S.A., Zimmerman, M., Otunla, A., Henry, J.A., 
Ferrey, A., Schofield, E., Kinton, J., Aveyard, P., and Marchesi, J.R. 
(2022). The association of weight loss with changes in the gut microbiota 
diversity, composition, and intestinal permeability: a systematic review 
and meta-analysis. Gut Microbes 14, 2020068. https://doi.org/10.1080/ 
19490976.2021.2020068.
23. Shade, A., Jones, S.E., Caporaso, J.G., Handelsman, J., Knight, R., Fi­
erer, N., and Gilbert, J.A. (2014). Conditionally rare taxa disproportion­
ately contribute to temporal changes in microbial diversity. mBio 5, 
e01371. https://doi.org/10.1128/mBio.01371-14.
24. Cornick, N.A., and Stanton, T.B. (2015). Lachnospira. In Bergey’s Manual 
of Systematics of Archaea and Bacteria (Wiley), pp. 1–6. https://doi.org/ 
10.1002/9781118960608.gbm00647.
25. De Angelis, M., Ferrocino, I., Calabrese, F.M., De Filippis, F., Cavallo, N., 
Siragusa, S., Rampelli, S., Di Cagno, R., Rantsiou, K., Vannini, L., et al. 
(2020). Diet influences the functions of the human intestinal microbiome. 
Sci. Rep. 10, 4247. https://doi.org/10.1038/s41598-020-61192-y.
26. Manor, O., Dai, C.L., Kornilov, S.A., Smith, B., Price, N.D., Lovejoy, J.C., 
Gibbons, S.M., and Magis, A.T. (2020). Health and disease markers 
correlate with gut microbiome composition across thousands of people. 
Nat. Commun. 11, 5206. https://doi.org/10.1038/s41467-020-18871-1.
27. Barber, C., Mego, M., Sabater, C., Vallejo, F., Bendezu, R.A., Masihy, M., 
Guarner, F., Espı´n, J.C., Margolles, A., and Azpiroz, F. (2021). Differential 
effects of Western and Mediterranean-type diets on gut microbiota: a 
metagenomics and metabolomics approach. Nutrients 13, 2638. 
https://doi.org/10.3390/nu13082638.
28. Arrieta, M.C., Stiemsma, L.T., Dimitriu, P.A., Thorson, L., Russell, S., Yu­
rist-Doutsch, S., Kuzeljevic, B., Gold, M.J., Britton, H.M., Lefebvre, D.L., 
et al. (2015). Early infancy microbial and metabolic alterations affect risk 
of childhood asthma. Sci. Transl. Med. 7, 307ra152. https://doi.org/10. 
1126/scitranslmed.aab2271.
29. Lun, H., Yang, W., Zhao, S., Jiang, M., Xu, M., Liu, F., and Wang, Y. 
(2019). Altered gut microbiota and microbial biomarkers associated 
with chronic kidney disease. MicrobiologyOpen 8, e00678. https://doi. 
org/10.1002/mbo3.678.
30. Brandt, E., Koivisto, A., Pereira, P., Mustanoja, E., Auvinen, P., Saari, T., 
Lehtola, J.M., Hannonen, S., Rusanen, M., Leinonen, V., et al. (2024). Gut 
microbiome changes in patients with idiopathic normal pressure hydro­
cephalus. Alzheimer Dis. Assoc. Disord. 38, 133–139. https://doi.org/ 
10.1097/WAD.0000000000000613.
31. Editorial Board (2015). Anaerostipes. In Bergey’s Manual of Systematics 
of Archaea and Bacteria (Wiley), pp. 1–4. https://doi.org/10.1002/ 
9781118960608.gbm00638.
32. Zeevi, D., Korem, T., Godneva, A., Bar, N., Kurilshikov, A., Lotan-Pom­
pan, M., Weinberger, A., Fu, J., Wijmenga, C., Zhernakova, A., and Segal, 
E. (2019). Structural variation in the gut microbiome associates with host 
health. Nature 568, 43–48. https://doi.org/10.1038/s41586-019-1065-y.
33. Doumatey, A.P., Adeyemo, A., Zhou, J., Lei, L., Adebamowo, S.N., 
Adebamowo, C., and Rotimi, C.N. (2020). Gut microbiome profiles are 
associated with type 2 diabetes in urban Africans. Front. Cell. Infect. Mi­
crobiol. 10, 63. https://doi.org/10.3389/fcimb.2020.00063.
34. Singh, V., Lee, G., Son, H., Koh, H., Kim, E.S., Unno, T., and Shin, J.H. 
(2022). Butyrate producers, ‘‘the sentinel of gut’’: their intestinal 
significance with and beyond butyrate, and prospective use as microbial 
Cell Reports Medicine 6, 102417, October 21, 2025 
13 
Article
ll
OPEN ACCESS
therapeutics. Front. Microbiol. 13, 1103836. https://doi.org/10.3389/ 
fmicb.2022.1103836.
35. Abdugheni, R., Wang, W.-Z., Wang, Y.-J., Du, M.-X., Liu, F.-L., Zhou, N., 
Jiang, C.-Y., Wang, C.-Y., Wu, L., Ma, J., et al. (2022). Metabolite profiling 
of human-originated Lachnospiraceae at the strain level. iMeta 1, e58. 
https://doi.org/10.1002/imt2.58.
36. Jeong, H., Chang, D.-H., and Kim, B.-C. (2024). Agathobaculum. In Ber­
gey’s Manual of Systematics of Archaea and Bacteria (Wiley), pp. 1–10. 
https://doi.org/10.1002/9781118960608.gbm01657.
37. Vandeputte, D., Falony, G., Vieira-Silva, S., Wang, J., Sailer, M., Theis, S., 
Verbeke, K., and Raes, J. (2017). Prebiotic inulin-type fructans induce 
specific changes in the human gut microbiota. Gut 66, 1968–1974. 
https://doi.org/10.1136/gutjnl-2016-313271.
38. Ezaki, T. (2015). Ruminococcus. In Bergey’s Manual of Systematics of 
Archaea and Bacteria, M.E. Trujillo, S. Dedysh, P. DeVos, B. Hedlund, 
P. Ka¨ mpfer, F.A. Rainey, and W.B. Whitman, eds. (Wiley), pp. 1–5. 
https://doi.org/10.1002/9781118960608.gbm00678.
39. Schaus, S.R., Vasconcelos Pereira, G., Luis, A.S., Madlambayan, E., Ter­
rapon, N., Ostrowski, M.P., Jin, C., Henrissat, B., Hansson, G.C., and 
Martens, E.C. (2024). Ruminococcus torques is a keystone degrader of 
intestinal mucin glycoprotein, releasing oligosaccharides used by 
Bacteroides thetaiotaomicron. mBio 15, e0003924. https://doi.org/10. 
1128/mbio.00039-24.
40. Crost, E.H., Coletto, E., Bell, A., and Juge, N. (2023). Ruminococcus gna­
vus: friend or foe for human health. FEMS Microbiol. Rev. 47, fuad014. 
https://doi.org/10.1093/femsre/fuad014.
41. Chassard, C., Delmas, E., Robert, C., Lawson, P.A., and Bernalier-Dona­
dille, A. (2012). Ruminococcus champanellensis sp. nov., a cellulose-de­
grading bacterium from human gut microbiota. Int. J. Syst. Evol. Micro­
biol. 62, 138–143. https://doi.org/10.1099/ijs.0.027375-0.
42. Ze, X., Duncan, S.H., Louis, P., and Flint, H.J. (2012). Ruminococcus bro­
mii is a keystone species for the degradation of resistant starch in the hu­
man colon. ISME J. 6, 1535–1543. https://doi.org/10.1038/ismej.2012.4.
43. McGuinness, A.J., Stinson, L.F., Snelson, M., Loughman, A., Stringer, A., 
Hannan, A.J., Cowan, C.S.M., Jama, H.A., Caparros-Martin, J.A., West, 
M.L., et al. (2024). From hype to hope: considerations in conducting 
robust microbiome science. Brain Behav. Immun. 115, 120–130. 
https://doi.org/10.1016/j.bbi.2023.09.022.
44. Snijder, M.B., Zimmet, P.Z., Visser, M., Dekker, J.M., Seidell, J.C., and 
Shaw, J.E. (2004). Independent and opposite associations of waist and 
hip circumferences with diabetes, hypertension and dyslipidemia: 
the AusDiab Study. Int. J. Obes. Relat. Metab. Disord. 28, 402–409. 
https://doi.org/10.1038/sj.ijo.0802567.
45. Parker, E.D., Pereira, M.A., Stevens, J., and Folsom, A.R. (2009). Associ­
ation of hip circumference with incident diabetes and coronary heart dis­
ease: the Atherosclerosis Risk in Communities study. Am. J. Epidemiol. 
169, 837–847. https://doi.org/10.1093/aje/kwn395.
46. Cameron, A.J., Romaniuk, H., Orellana, L., Dallongeville, J., Dobson, A. 
J., Drygas, W., Ferrario, M., Ferrieres, J., Giampaoli, S., Gianfagna, F., 
et al. (2020). Combined influence of waist and hip circumference on 
risk of death in a large cohort of European and Australian adults. 
J. Am. Heart Assoc. 9, e015189. https://doi.org/10.1161/JAHA.119. 
015189.
47. Alexandrov, N.V., Eelderink, C., Singh-Povel, C.M., Navis, G.J., Bakker, 
S.J.L., and Corpeleijn, E. (2018). Dietary protein sources and muscle 
mass over the life course: the Lifelines Cohort Study. Nutrients 10, 
1471. https://doi.org/10.3390/nu10101471.
48. Kong, W., Xie, Y., Hu, J., Ding, W., and Cao, C. (2024). Higher ultra-pro­
cessed foods intake is associated with low muscle mass in young to 
middle-aged adults: a cross-sectional NHANES study. Front. Nutr. 11, 
1280665. https://doi.org/10.3389/fnut.2024.1280665.
49. Ard, J.D., Lewis, K.H., Rothberg, A., Auriemma, A., Coburn, S.L., Cohen, 
S.S., Loper, J., Matarese, L., Pories, W.J., and Periman, S. (2019). Effec­
tiveness of a total meal replacement program (OPTIFAST program) 
on weight loss: results from the OPTIWIN study. Obesity 27, 22–29. 
https://doi.org/10.1002/oby.22303.
50. Aslam, H., Lotfaliany, M., So, D., Berding, K., Berk, M., Rocks, T., Hock­
ey, M., Jacka, F.N., Marx, W., Cryan, J.F., and Staudacher, H.M. (2024). 
Fiber intake and fiber intervention in depression and anxiety: a system­
atic review and meta-analysis of observational studies and randomized 
controlled trials. Nutr. Rev. 82, 1678–1695. https://doi.org/10.1093/nu­
trit/nuad143.
51. Yang, J., Wang, H.P., Zhou, L., and Xu, C.F. (2012). Effect of dietary fiber 
on constipation: a meta-analysis. World J. Gastroenterol. 18, 7378–7383. 
https://doi.org/10.3748/wjg.v18.i48.7378.
52. Moman, R.N., Gnanenthiran, P., and Varacallo, M.A. (2022). Physiology, 
albumin. In StatPearls [Internet] (StatPearls Publishing). https://www. 
ncbi.nlm.nih.gov/books/NBK459198/.
53. Mah, J.Y., Choy, S.W., Roberts, M.A., Desai, A.M., Corken, M., Gwini, S. 
M., and McMahon, L.P. (2020). Oral protein-based supplements versus 
placebo or no treatment for people with chronic kidney disease requiring 
dialysis. Cochrane Database Syst. Rev. 5, CD012616. https://doi.org/10. 
1002/14651858.CD012616.pub2.
54. Levitt, D.G., and Levitt, M.D. (2016). Human serum albumin homeostasis: 
a new look at the roles of synthesis, catabolism, renal and gastrointes­
tinal excretion, and the clinical value of serum albumin measurements. 
Int. J. Gen. Med. 9, 229–255. https://doi.org/10.2147/IJGM.S102819.
55. La Cava, A., Alviggi, C., and Matarese, G. (2004). Unraveling the multiple 
roles of leptin in inflammation and autoimmunity. J. Mol. Med. 82, 4–11. 
https://doi.org/10.1007/s00109-003-0492-1.
56. Lane, M.M., Gamage, E., Du, S., Ashtree, D.N., McGuinness, A.J., Gauci, 
S., Baker, P., Lawrence, M., Rebholz, C.M., Srour, B., et al. (2024). Ultra- 
processed food exposure and adverse health outcomes: umbrella review 
of epidemiological meta-analyses. BMJ 384, e077310. https://doi.org/ 
10.1136/bmj-2023-077310.
57. Scientific Advisory Committee on Nutrition (SACN) (2023). SACN state­
ment on processed foods and health. Available at: https://www.gov.uk/ 
government/groups/scientific-advisory-committee-on-nutrition
58. Chassaing, B., Compher, C., Bonhomme, B., Liu, Q., Tian, Y., Walters, 
W., Nessel, L., Delaroque, C., Hao, F., Gershuni, V., et al. (2022). 
Randomized controlled-feeding study of dietary emulsifier carboxymeth­
ylcellulose reveals detrimental impacts on the gut microbiota and metab­
olome. Gastroenterology 162, 743–756. https://doi.org/10.1053/j.gastro. 
2021.11.006.
59. Suez, J., Cohen, Y., Valde´ s-Mas, R., Mor, U., Dori-Bachash, M., Federici, 
S., Zmora, N., Leshem, A., Heinemann, M., Linevsky, R., et al. (2022). 
Personalized microbiome-driven effects of non-nutritive sweeteners on 
human glucose tolerance. Cell 185, 3307–3328.e19. https://doi.org/10. 
1016/j.cell.2022.07.016.
60. Srour, B., Kordahi, M.C., Bonazzi, E., Deschasaux-Tanguy, M., Touvier, 
M., and Chassaing, B. (2022). Ultra-processed foods and human health: 
from epidemiological evidence to mechanistic insights. Lancet Gastro­
enterol. Hepatol. 7, 1128–1140. https://doi.org/10.1016/S2468-1253 
(22)00169-8.
61. Plamada, D., and Vodnar, D.C. (2021). Polyphenols–gut microbiota inter­
relationship: a transition to a new generation of prebiotics. Nutrients 14, 
137. https://doi.org/10.3390/nu14010137.
62. Recoules, C., Touvier, M., Pierre, F., and Audebert, M. (2025). Evaluation 
of the toxic effects of food additives, alone or in mixture, in four 
human cell models. Food Chem. Toxicol. 196, 115198. https://doi.org/ 
10.1016/j.fct.2024.115198.
63. Ravelli, M.N., and Schoeller, D.A. (2020). Traditional self-reported dietary 
instruments are prone to inaccuracies and new approaches are needed. 
Front. Nutr. 7, 90. https://doi.org/10.3389/fnut.2020.00090.
64. Wan, Y., Wang, F., Yuan, J., Li, J., Jiang, D., Zhang, J., Li, H., Wang, R., 
Tang, J., Huang, T., et al. (2019). Effects of dietary fat on gut microbiota 
14 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
and faecal metabolites, and their relationship with cardiometabolic risk 
factors: a 6-month randomised controlled-feeding trial. Gut 68, 1417– 
1429. https://doi.org/10.1136/gutjnl-2018-317609.
65. Schoeler, M., Ellero-Simatos, S., Birkner, T., Mayneris-Perxachs, J., Ols­
son, L., Brolin, H., Loeber, U., Kraft, J.D., Polizzi, A., Martı´-Navas, M., 
et al. (2023). The interplay between dietary fatty acids and gut microbiota 
influences host metabolism and hepatic steatosis. Nat. Commun. 14, 
5329. https://doi.org/10.1038/s41467-023-41074-3.
66. von Schwartzenberg, R.J., Bisanz, J.E., Lyalina, S., Spanogiannopoulos, 
P., Ang, Q.Y., Cai, J., Dickmann, S., Friedrich, M., Liu, S.Y., Collins, S.L., 
et al. (2021). Caloric restriction disrupts the microbiota and colonization 
resistance. Nature 595, 272–277. https://doi.org/10.1038/s41586-021- 
03663-4.
67. Porcari, S., Ng, S.C., Zitvogel, L., Sokol, H., Weersma, R.K., Elinav, E., 
Gasbarrini, A., Cammarota, G., Tilg, H., and Ianiro, G. (2025). The micro­
biome for clinicians. Cell 188, 2836–2844. https://doi.org/10.1016/j.cell. 
2025.04.016.
68. Mirzayi, C., Renson, A., Genomic Standards Consortium; Massive Anal­
ysis and Quality Control Society; Zohra, F., Elsafoury, S., Geistlinger, L., 
Kasselman, L.J., Eckenrode, K., van de Wijgert, J., et al. (2021). Report­
ing guidelines for human microbiome research: the STORMS checklist. 
Nat. Med. 27, 1885–1892. https://doi.org/10.1038/s41591-021-01552-x.
69. Schulz, K.F., Altman, D.G., and Moher, D.; CONSORT Group (2010). 
CONSORT 2010 statement: updated guidelines for reporting parallel 
group randomised trials. PLoS Med. 7, e1000251. https://doi.org/10. 
1371/journal.pmed.1000251.
70. Steinegger, M., and Soding, J. (2017). MMseqs2 enables sensitive pro­
tein sequence searching for the analysis of massice data sets. Nat. Bio­
technol. 35, 1026–1028. https://.doi.org/10.1038/nbt.3988.
71. Boyd J.A., Woodcroft B.J., Tyson G.W. (2019). Comparative genomics 
using EnrichM. https://github.com/geronimp/enrichM.
72. Harris, P.A., Taylor, R., Minor, B.L., Elliott, V., Fernandez, M., O’Neal, L., 
McLeod, L., Delacqua, G., Delacqua, F., Kirby, J., et al. (2019). The 
REDCap consortium: building an international community of software 
platform partners. J. Biomed. Inform. 95, 103208. https://doi.org/10. 
1016/j.jbi.2019.103208.
73. Harris, P.A., Taylor, R., Thielke, R., Payne, J., Gonzalez, N., and Conde, J. 
G. (2009). Research electronic data capture (REDCap)—a metadata- 
driven methodology and workflow process for providing translational 
research informatics support. J. Biomed. Inform. 42, 377–381. https:// 
doi.org/10.1016/j.jbi.2008.08.010.
74. Asher, R.C.Z., Burrows, T.L., and Collins, C.E. (2013). Very low-energy di­
ets for weight loss in adults: a review. Nutr. Diet. 70, 101–112. https://doi. 
org/10.1111/j.1747-0080.2012.01628.x.
75. Food Standards Australia New Zealand (2014). AUSNUT 2011–2013 
Food Composition Database. Available at: https://www.foodstandards. 
gov.au/science/monitoringnutrients/ausnut/pages/default.aspx
76. Food Standards Australia New Zealand (2015). Nutrient Tables for Use in 
Australia (NUTTAB 2010). Available at: http://www.foodstandards.gov. 
au/science/monitoringnutrients/nutrientables/nuttab/pages/default.aspx
77. Ambrosini, G.L., Hurworth, M., Giglia, R., Trapp, G., and Strauss, P. 
(2018). Feasibility of a commercial smartphone application for dietary 
assessment in epidemiological research and comparison with 24-h die­
tary recalls. Nutr. J. 17, 5. https://doi.org/10.1186/s12937-018-0315-4.
78. Wing, R.R., and Jeffery, R.W. (2001). Food provision as a strategy to pro­
mote weight loss. Obes. Res. 9, 271S–275S. https://doi.org/10.1038/ 
oby.2001.130.
79. Noakes, M., Foster, P.R., Keogh, J.B., and Clifton, P.M. (2004). Meal re­
placements are as effective as structured weight-loss diets for treating 
obesity in adults with features of metabolic syndrome. J. Nutr. 134, 
1894–1899. https://doi.org/10.1093/jn/134.8.1894.
80. Ruiz, A., Cerdo´ , T., Ja´ uregui, R., Pieper, D.H., Marcos, A., Clemente, A., 
Garcı´a, F., Margolles, A., Ferrer, M., Campoy, C., and Sua´ rez, A. (2017). 
One-year calorie restriction impacts gut microbial composition but not its 
metabolic performance in obese adolescents. Environ. Microbiol. 19, 
1536–1551. https://doi.org/10.1111/1462-2920.13713.
81. Kers, J.G., and Saccenti, E. (2021). The power of microbiome studies: 
some considerations on which alpha and beta metrics to use and how 
to report results. Front. Microbiol. 12, 796025. https://doi.org/10.3389/ 
fmicb.2021.796025.
82. Bindels, L.B., Watts, J.E.M., Theis, K.R., Carrion, V.J., Ossowicki, A., Sei­
fert, J., Oh, J., Shao, Y., Hilty, M., Kumar, P., et al. (2025). A blueprint for 
contemporary studies of microbiomes. Microbiome 13, 95. https://doi. 
org/10.1186/s40168-025-02091-0.
83. Schloss, P.D. (2024). Waste not, want not: revisiting the analysis that 
called into question the practice of rarefaction. mSphere 9, e00355-23. 
https://doi.org/10.1128/msphere.00355-23.
84. R Core Team (2017). R: A Language and Environment for Statistical 
Computing. Available at: https://www.R-project.org/
85. Bastiaanssen, T., Quinn, T., and Loughman, A. (2022). Treating bugs as 
features: a compositional guide to the statistical analysis of the micro­
biome–gut–brain axis. Preprint at arXiv. https://doi.org/10.48550/arXiv. 
2207.12475. 
86. Oren, A., and Garrity, G.M. (2021). Valid publication of the names of forty- 
two phyla of prokaryotes. Int. J. Syst. Evol. Microbiol. 71, 005056. https:// 
doi.org/10.1099/ijsem.0.005056.
87. Lovibond, P.F., and Lovibond, S.H. (1995). The structure of negative 
emotional states: comparison of the Depression Anxiety Stress Scales 
(DASS) with the Beck Depression and Anxiety Inventories. Behav. Res. 
Ther. 33, 335–343. https://doi.org/10.1016/0005-7967(94)00075-U.
88. Topp, C.W., Østergaard, S.D., Søndergaard, S., and Bech, P. (2015). The 
WHO-5 Well-Being Index: a systematic review of the literature. Psy­
chother. Psychosom. 84, 167–176. https://doi.org/10.1159/000376585.
89. Soldatos, C.R., Dikeos, D.G., and Paparrigopoulos, T.J. (2000). Athens 
Insomnia Scale: validation of an instrument based on ICD-10 criteria. 
J. Psychosom. Res. 48, 555–560. https://doi.org/10.1016/S0022-3999 
(00)00095-7.
90. Bengtsson, M., Persson, J., Sjo¨ lund, K., and Ohlsson, B. (2013). Further 
validation of the visual analogue scale for irritable bowel syndrome after 
use in clinical practice. Gastroenterol. Nurs. 36, 188–198. https://doi.org/ 
10.1097/SGA.0b013e3182945881.
91. Bengtsson, M., Ohlsson, B., and Ulander, K. (2007). Development and 
psychometric testing of the Visual Analogue Scale for Irritable Bowel 
Syndrome (VAS-IBS). BMC Gastroenterol. 7, 16. https://doi.org/10. 
1186/1471-230X-7-16.
92. Blake, M.R., Raker, J.M., and Whelan, K. (2016). Validity and reliability of 
the Bristol Stool Form Scale in healthy adults and patients with diarrhoea- 
predominant irritable bowel syndrome. Aliment. Pharmacol. Ther. 44, 
693–703. https://doi.org/10.1111/apt.13746.
93. IPAQ Research Committee (2002). Guidelines for data processing and 
analysis of the International Physical Activity Questionnaire (IPAQ)–Short 
Last 7 Days Self-Administered Format. Available at: http://www.ipaq. 
ki.se
94. RStudio Team (2015). RStudio: Integrated Development Environment for 
R [Internet]. Available at: http://www.rstudio.com/
95. Bates, D., Ma¨ chler, M., Bolker, B., and Walker, S. (2015). Fitting linear 
mixed-effects models using lme4. J. Stat. Softw. 67, 1–48. https://doi. 
org/10.18637/jss.v067.i01.
96. Greenland, S., Senn, S.J., Rothman, K.J., Carlin, J.B., Poole, C., 
Goodman, S.N., and Altman, D.G. (2016). Statistical tests, P values, 
confidence intervals, and power: a guide to misinterpretations. Eur. J. 
Epidemiol. 31, 337–350. https://doi.org/10.1007/s10654-016-0149-3.
97. Buuren, S.v., and Groothuis-Oudshoorn, K. (2011). mice: Multivariate 
imputation by chained equations in R. J. Stat. Softw. 45, 1–67. https:// 
doi.org/10.18637/jss.v045.i03.
Cell Reports Medicine 6, 102417, October 21, 2025 
15 
Article
ll
OPEN ACCESS
98. Lenth, R.V. (2024). emmeans: Estimated Marginal Means, aka Least- 
Squares Means. R package version 1.11.2-80003. Available at: https:// 
rvlenth.github.io/emmeans/.
99. Oksanen, J., Simpson, G.L., Blanchet, F.G., Kindt, R., Legendre, P., Min­
chin, P.R., O’Hara, R.B., Solymos, P., Stevens, M.H.H., Szo¨ cs, E., et al. 
(2024). vegan: Community Ecology Package. R package version 2.7-1. 
Available at: https://github.com/vegandevs/vegan.
100. Haynes, W. (2013). Benjamini–Hochberg method. In Encyclopedia of 
Systems Biology, W. Dubitzky, O. Wolkenhauer, K.-H. Cho, and H. Yo­
kota, eds. (Springer), p. 78. https://doi.org/10.1007/978-1-4419-9863- 
7_1215.
101. Stuart, E.A., Lee, B.K., and Leacy, F.P. (2013). Prognostic score-based 
balance measures can be a useful diagnostic for propensity score 
methods in comparative effectiveness research. J. Clin. Epidemiol. 66, 
S84–S90.e1. https://doi.org/10.1016/j.jclinepi.2013.01.013.
102. Akoglu, H. (2018). User’s guide to correlation coefficients. Turk. J. Emerg. 
Med. 18, 91–93. https://doi.org/10.1016/j.tjem.2018.08.001.
103. Preska Steinberg, A., Wang, Z.G., and Ismagilov, R.F. (2019). Food poly­
electrolytes compress the colonic mucus hydrogel by a Donnan mecha­
nism. Biomacromolecules 20, 2675–2683. https://doi.org/10.1021/acs. 
biomac.9b00442.
104. Han, G., and Vaishnava, S. (2023). Microbial underdogs: exploring the 
significance of low-abundance commensals in host–microbe interac­
tions. Exp. Mol. Med. 55, 2498–2507. https://doi.org/10.1038/s12276- 
023-01120-y.
105. McMurdie, P.J., and Holmes, S. (2013). phyloseq: an R package 
for reproducible interactive analysis and graphics of microbiome 
census data. PLoS One 8, e61217. https://doi.org/10.1371/journal. 
pone.0061217.
106. Lahti, L., and Shetty, S. (2012–2019). microbiome: R package. Available 
at: http://microbiome.github.io
16 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
STAR★METHODS
KEY RESOURCES TABLE
REAGENT or RESOURCE
SOURCE
IDENTIFIER
Biological samples
Human fecal and blood samples
This paper
N/A
Critical commercial assays
FLOQSwab-ADT (active drying tube 
with internal desiccant)
Copan Italia SPA
https://www.copangroup.com/
Alcohol wipes (70% isopropyl alcohol)
Briemarpak
https://www.briemarpak.com.au/
BD Vacutainer® tubes (yellow top)
Becton Dickinson
https://www.bd.com/en-anz/
Butterfly needles and barrels
Becton Dickinson
https://www.bd.com/en-anz/
BDTM Cytometric Bead Array assay 
platform
BD Biosciences
https://www.bdbiosciences.com/
Radioimmunoassay (Iodine 125) 
kit for leptin
Merck Millipore
https://www.sigmaaldrich.com/AU/en/
ADVIA® Chemistry kits
Siemens Healthineers
https://www.siemens-healthineers. 
com/en-au/
DNeasy 96 PowerSoil Pro QIAcube HT Kit
Qiagen
https://www.qiagen.com/au/
QIAcube HT DNA extraction system
Qiagen
https://www.qiagen.com/au/
PowerBead Pro Tubes
Qiagen
https://www.qiagen.com/au/
Quant-ITTM PicoGreenTM dsDNA Assay Kit
ThermoFisher Scientific
https://www.thermofisher.com/
Illumina DNA Prep (M) Tagmentation Kit
Illumina
https://www.illumina.com/
IDT for Illumina DNA/RNA UD Index 
Sets A-D
Illumina
https://www.idtdna.com/
QIAxcel DNA High Resolution Kit
Qiagen
https://www.qiagen.com/au/
NovaSeq 6000 Sequencing System
Illumina
https://www.illumina.com/
Software and algorithms
REDCap
REDCap
https://redcap.deakin.edu.au/
FoodWorks Professional
Xyris Software Pty Ltd
https://foodworks.online/editions/ 
foodworks-professional/
R
R Project for Statistical Computing
https://www.r-project.org/
RStudio
Posit
https://posit.co/
Microba Gene and Pathway Profiler 
(MGPP) v1.0
Microba Life Sciences
https://www.microba.com/
Microba Genome Database (MGDB) v1.0.3
Microba Life Sciences
https://www.microba.com/
MMSeqs2 Release 10-6d92c
Steinegger & So¨ ding, 201770
https://github.com/soedinglab/MMseqs2
enrichM
Woodcroft et al., 202071
https://github.com/geronimp/enrichM
UniProt ID mapping service
UniProt Consortium
https://www.uniprot.org/
Other
Clinical Trial Registration
Australian New Zealand Clinical 
Trials Registry
ACTRN12620000301965 at http://anzctr. 
org.au/
−80◦C secure freezer facility
Barwon Health, Geelong, Australia
N/A
Clinical pathology services (blood sample 
processing and biochemical assays)
Australian Clinical Labs, Clayton, 
Australia
https://www.clinicallabs.com.au/
Clinical pathology services (serum 
inflammatory marker assays)
SA Pathology, Adelaide Women’s & 
Children’s Hospital, Australia
https://www.sapathology.sa.gov.au/
Dietary Questionnaire for Epidemiological 
Studies v3.2 scoring service
Cancer Council Victoria, Nutritional 
Assessment Office
https://www.cancervic.org.au/research/ 
epidemiology/
Cell Reports Medicine 6, 102417, October 21, 2025 
e1 
Article
ll
OPEN ACCESS
EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS
The MicroFit Study was a three-week, investigator-initiated, single-blind, two-arm, parallel-group randomized controlled-feeding trial 
with computer-generated 1:1 randomization. Women aged 30–65 years with BMI 30–45 kg/m2 were recruited from a community- 
based sample in the south-west region of Victoria, Australia. Participants were screened against eligibility criteria online using 
REDCap,72,73 and research assistants confirmed eligibility and obtained informed consent via telephone. At the baseline 
assessment, research assistants collected anthropometric measurements to verify participants’ eligibility based on BMI. Eligible par­
ticipants were randomized to a meal replacement program comprising primarily food-based or supplement-based VLED options. A 
uniform calorie limit for the VLEDs was chosen to ensure consistency, simplify adherence and monitoring, and enable a more 
controlled comparison than varying caloric restriction based on individual energy needs. Participants attended in-person study visits 
at baseline and week three at Australian Clinical Labs (Geelong, Australia), where a phlebotomist collected serum samples after an 
overnight fast. Participants were provided kits to collect fecal samples at home at baseline and week three, which they sent via 
prepaid mail to Microba Pty Ltd (Brisbane, Australia) for analysis. Questionnaire data, including sociodemographic, dietary, and 
health-related factors, were self-reported at both study timepoints by participants at home using REDCap online. Adverse events 
were monitored from consent until study completion (week 3), recorded by the study team when reported by participants, and 
assessed for severity and relatedness to the intervention according to Barwon Health research monitoring guidelines. An a priori 
sample size calculation was conducted based on 40 participants (20 in each arm). However, by study completion, we randomized 
47 participants to account for missing baseline data and higher than anticipated dropout, and the power calculation was adjusted 
accordingly. A schematic overview of the study design, including participant characteristics, intervention arms, and assessment 
timepoints, is presented in Figure S3.
Participants were informed that the purpose of the study was to examine whether VLEDs influence the type and number of gut 
microbiota in the gastrointestinal tract, and to assess associations between gut microbiota and BMI, weight, waist circumference, 
blood markers, symptoms of depression, anxiety and stress, perceived wellbeing, sleep patterns, stool consistency, and gastroin­
testinal symptoms.
This trial was conducted following the principles of the Declaration of Helsinki, received ethical approval from the Barwon Health 
(19/112) and Deakin University (2018/211) Human Research Ethics Committees and was registered on the Australian New Zealand 
Clinical Trials Registry (ACTRN12620000301965). This manuscript is presented as per the Consolidated Standards of Reporting 
Trials (CONSORT) statement and checklist69 and gut microbiome data are reported as per the Strengthening the Organizing and Re­
porting of Microbiome Studies (STORMS) checklist.68 The completed checklists and additional details are provided in the 
supplemental information (see Tables S11 and S12). The trial and manuscript development did not involve patients or the public 
owing to the absence of funding to support consumer engagement for this research.
METHOD DETAILS
Recruitment and participants
Community-based recruitment was conducted from May 2021 to February 2022 using online platforms hosted by Deakin University 
and Barwon Health (e.g., Facebook, Instagram, Twitter, Barwon Health online newsletter), distributing flyers to local general 
practitioner offices, and through a variety of both paid and free online advertising services.
Inclusion criteria were: female sex, chosen to reduce interindividual heterogeneity; with a BMI between 30 and 45 kg/m2 to align 
with guidelines recommending VLEDs for individuals in this BMI range74; aged 30–65 years; able to commit to all study procedures, 
including attending in-person appointments and consuming only the investigational products and recommended extras for the study 
duration; able to understand study materials and directions presented in English; with access to the internet and a computer, smart­
phone, or tablet; and able to consent to not to enrolling in another clinical trial while taking part in the study.
Exclusion criteria were: currently consuming VLED products; having a diagnosed food allergy or food intolerance; receiving treat­
ment with medications related to obesity; confirmed/suspected/planned pregnancy, or lactating; diagnosed with or having 
commenced a new treatment for, anxiety and/or depression within one month before baseline; having gastrointestinal disease or 
history of major gastrointestinal surgery; having a pre-existing cardiometabolic conditions; having had a heart attack within the 
past six months; having a diagnosed eating disorder; having other major medical conditions likely to have systemic effects or deemed 
unfit for study participation by the research team (e.g., type 2 diabetes, prediabetic, insulin resistance); regularly (subjective, as 
defined in our study protocol and not quantified due to individual variations in interpretation) using opioid-based medications; 
regularly using recreational or illicit drugs; regularly using sodium-glucose co-transporter-2 inhibitors (i.e., gliflozins); having used an­
tibiotics, prebiotics, and/or probiotics in the month before baseline; and having been enrolled in another clinical trial within the past 
three months.
Randomization, allocation, and blinding
Eligible participants were randomly assigned in a 1:1 ratio to either the food-based or supplement-based VLED using a computer- 
generated randomization sequence with randomly ordered blocks of sizes 2 and 4. This sequence was created by a study statistician 
and input into REDCap online by an independent researcher to ensure allocation concealment from the study investigators. An 
e2 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
unblinded trial coordinator enrolled and informed participants of the VLED program to which they had been assigned. The outcome 
assessors (research assistants) and all other study investigators, including the analyzing statisticians, remained blinded to the group 
allocations until data analysis completion. As the study was single-blind, participants were aware of their group allocations and in­
structed not to discuss their diet with the outcome assessors to preserve blinding.
Interventions
The VLEDs were intended to be matched in overall energy (800–900 kcal per day), macronutrient profiles, sugar, sodium, and fiber. 
The supplement-based VLED comprised three daily total meal replacement options (16 items). Participants chose from a selection of 
powdered shakes in banana, caramel, chocolate, coffee, mocha, strawberry, and vanilla flavors; soups in chicken and vegetable fla­
vors; bars in almond butter, berry, cappuccino, chocolate, and cereal flavors; and, desserts in chocolate and lemon cre` me flavors. On 
average, approximately 70% of the composition of these options was made of extracted, refined, fractionated, modified, and/or iso­
lated proteins (e.g., calcium caseinate), carbohydrates (e.g., maltodextrin), fats (e.g., medium chain triglycerides), and fibers (e.g., 
fructo-oligosaccharide), as well as added vitamins (e.g., vitamin B1) and minerals (e.g., potassium citrate), and additives like emul­
sifiers (e.g., citric and fatty acid esters of glycerol [472c]), non-sugar sweeteners (e.g., aspartame [951]), flavors (e.g., unspecified 
‘‘flavor’’), colors (e.g., curcumin [100]), thickeners and stabilizers (e.g., vegetable gum [414]). The remaining 30% consisted primarily 
of whole powdered milk. The food-based VLED comprised three daily total meal replacement options (55 items) and a discretionary 
snack (11 items). Participants chose from a selection of pre-prepared meals, including various porridge, egg, and vegetable dishes 
for breakfast, with lunches and dinners composed of baked and stuffed items, casseroles, curries, meatballs, pasta, rice dishes, stir- 
fries, and soups. On average, approximately 93% of the composition of these options was made of vegetables (e.g., green cabbage), 
fruits (e.g., banana), whole grains (e.g., oats), beans (e.g., cannellini beans), legumes (e.g., chickpeas), lean meats (e.g., chicken), 
dairy cheeses (e.g., ricotta), nuts (e.g., almond meal), seeds (e.g., flaxseed), herbs (e.g., parsley), and spices (e.g., cinnamon). The 
remaining 7% of the composition consisted primarily of protein isolates (e.g., whey protein isolate), as well as additives such as emul­
sifiers (e.g., soy lecithin [322]), non-sugar sweeteners (e.g., stevia [960]), flavors (e.g., vanilla extract), thickeners and stabilizers (e.g., 
guar gum [412]), with approximately less than 1% of the composition of discretionary snacks including an added and isolated fiber 
(e.g., oligofructose) and the probiotic Lactobacillus plantarum (now Lactiplantibacillus plantarum). The full list of the options in each 
group and their ingredients are detailed in the supplemental information (see Table S13).
Participants in both groups selected their meal replacements online weekly, with all consumables and home delivery provided free 
of charge. They could choose any three meal replacements to consume at any time throughout the day. Participants were also 
permitted to include additional ‘recommended extras’ foods from a predetermined list (see Table S14). The supplement-based group 
was recommended to consume at least two cups of low starch vegetables. The food-based group was recommended to include one 
additional fruit or protein snack and three serves of side salads or vegetables. Non-sugar sweeteners, diet jelly desserts, and sugar- 
free lollies and gum were not recommended for daily use in the food-based group, with diet cordial and diet soft drinks recommended 
as occasional options. The supplement-based group were recommended to consume these items ad libitum, with no restrictions. 
Participants logged their daily food consumption using the Easy Diet Diary application (app) throughout the entire study period. These 
data were used to monitor adherence to the approximately 900 kcal/d target and to assess macronutrient intake, which was analyzed 
using Australian food composition databases via the FoodWorks Professional nutrient analysis software (Xyris Software Pty Ltd, 
Brisbane, Australia.75,76 Several strategies were employed to optimize reporting accuracy and minimize burden within the practical 
constraints of a free-living trial. Participants received standardized and group-specific dietary guidance, used the app daily 
(pre-populated with trial consumables and featuring ‘search as you type’ and barcode functionality, with demonstrated feasibility, 
acceptability, and relative validity77), and were supported by weekly staff contact. Additionally, the provision of food, particularly 
meal replacements, can enhance dietary adherence by increasing convenience and allowing for tighter control compared to 
participant-prepared meals.78,79
Sample size
An a priori sample size calculation was conducted, informed by a similar study available at the time of protocol development that used 
alpha diversity metrics, including the Shannon index.80 The Shannon index was selected as the primary outcome for this manuscript 
due to its inclusion in the sample size calculation, as well as its widespread use and sensitivity in detecting between-group 
differences in microbiome research.81 A sample size of 40 participants (20 in each arm) had 80% power with an alpha of 5% to 
detect large between-group difference effect sizes (d = 0.91) in the Shannon index. An a posteriori power calculation based on 
45 participants had 80% power to detect effect sizes of d = 0.85.
Sample collection
Fecal samples (∼15g) were collected by participants at baseline and week three using a Copan Italia SPA FLOQSwab in an active 
drying tube, including an internal desiccant to preserve samples at room temperature for up to four weeks. Participants were in­
structed to collect their fecal sample within 48 h after their in-person appointments. Fecal samples were sent directly to Microba 
via post at room temperature where they were stored at −80◦C until further processing. Fasted blood samples (40mL) were collected 
and stored at ACL as per standard procedures.
Cell Reports Medicine 6, 102417, October 21, 2025 
e3 
Article
ll
OPEN ACCESS
Fecal samples (∼15g) were collected by participants at baseline and week eight using a Copan Italia SPA FLOQSwab in an active
DNA extraction
Fecal samples were extracted using the DNeasy 96 PowerSoil Pro QIAcube HT Kit (Qiagen 47021) in a 2mL deep well plate format 
with a modified initial processing step on the QIAcube HT DNA extraction system (Qiagen 9001793). Mechanical lysis was performed 
with PowerBead Pro beads (Qiagen 19311). DNA was quantified using a high-sensitivity dsDNA fluorometric assay (QuantIT, 
ThermoFisher, Q33120), with samples needing to reach a minimum of 0.2 ng/μL for quality control.
Library preparation
Libraries were constructed using the Illumina DNA Prep (M) Tagmentation Kit (Illumina, 20018705) with IDT for Illumina DNA/RNA UD 
Index Sets A-D (Illumina 20027213-16), modified to accommodate processing in a 384-plate format. Individual libraries were pooled 
in equimolar amounts and assessed using a high-sensitivity dsDNA fluorometric assay (QuantIT, ThermoFisher, Q33120) and visu­
alized with capillary gel electrophoresis using the QIAxcel DNA High Resolution Kit (Qiagen, 929002).
Shotgun metagenomic sequencing
Samples were sequenced on the NovaSeq6000 (Illumina) using v1.5 300bp PE sequencing reagents to a target depth of 3Gb, with a 
minimum of 2Gb (approximately 7–16 million paired-end reads per sample). Sequence data were reviewed for yield and quality, with 
known control samples included in each run to monitor for background contamination. Paired-end DNA sequencing data were de­
multiplexed and adaptor trimmed using Illumina BaseSpace Bcl2fastq2 (v2.20) with one mismatch allowed in index sequences. 
Reads were quality trimmed and residual adaptors removed using Trimmomatic v0.39 with parameters: -phred33 LEADING:3 TRAIL­
ING:3 SLIDINGWINDOW:4:15 CROP:100000 HEADCROP:0 MINLEN:100. Human DNA was removed by aligning reads to the human 
genome reference assembly 38 (GRCh38.p12) using bwa-mem v0.7.17 with a minimum seed length of 31 (-k 31). Alignments were 
filtered using SAMtools v1.7, and reads mapping to the human genome with >95% identity over >90% of the read length were flagged 
as human DNA and removed.
Species profiles and functional potential
Species profiles were obtained using the Microba Community Profiler (MCP) v1.0 and the Microba Genome Database (MGDB) 
v1.0.3, with reads assigned to genomes within MGDB to estimate and report the relative cellular abundance of species clusters. 
Quantification of gene and pathway abundance was performed using the Microba Gene and Pathway Profiler (MGPP) v1.0 against 
the Microba Genes (MGENES) database v1.0.3. Open reading frames from genomes in MGDB were clustered against Uniref. 
90 (release 2019/04) using MMSeqs2 with 90% identity over 80% of read length. Gene clusters were annotated with Uniref. 90 iden­
tifiers and linked to Enzyme Commission and Transporter Classification Database annotations via the UniProt ID mapping service. 
Enzyme Commission annotations were then used to determine MetaCyc pathway encoding using enrichM, with pathways classified 
as encoded if completeness exceeded 80%. DNA sequencing read pairs aligning with gene sequences from any protein within an 
MGENES protein cluster were summed. The abundances of encoded pathways for species detected by MCP were calculated by 
averaging the read counts of all genes for each enzyme in the pathway.
Data preparation
Data normalization was performed by down-sampling to a standardized number of reads (7,000,000) before profiling within the MCP. 
One sample (food-based, week 3) had low read count (2,779,002) and was removed in sensitivity analyses. As per recommendations 
to account for variation in sequencing depth,82,83 alpha diversity was calculated using raw count data, with rarefaction applied to 
match the smallest total number of prokaryotic reads across all samples (4,139,908). All other secondary gut microbiome outcome 
analyses, including beta diversity, taxonomic composition, and functional potential, were conducted using non-rarefied data. CLR 
transformations were utilized before conducting beta-diversity and differential abundance statistical tests (species, genus, family, 
and phylum) given the compositional and non-normal nature of microbiome relative abundance data.
Alpha diversity, beta diversity, taxonomic composition, and functional potential
Alpha diversity, which summarizes community structure within a sample, was evaluated at the species-level using two metrics: the 
Shannon index (primary outcome) and richness (secondary outcome). The Shannon index provides a comprehensive measure of 
community structure by considering both the number of detected species (richness) and how evenly distributed the species are 
(evenness); communities with higher numbers of detected species and more even distributions of these species will result in a higher 
Shannon index. Richness, which quantifies only the number of different species present in each sample, was chosen as a secondary 
outcome measure of alpha diversity due to its more limited capacity to capture community structure. In addition, alpha diversity was 
assessed at the functional level using MetaCyc pathway annotations as a secondary outcome. Beta diversity, also a secondary 
outcome, which summarizes between-sample differences in community structure, was calculated using Aitchison distances, defined 
as the Euclidean distance between CLR-transformed samples, via the stats84 package. Principal Component Analysis (PCA) was 
used to reduce data dimensionality for visualization in two dimensions to identify patterns or clusters of samples within the dataset 
using the Tjazi85 package. Additionally, taxonomic composition was assessed by identifying and quantifying bacterial species, 
genera, families, and phyla using shotgun metagenomic sequencing. The analysis of functional potential, focused on MetaCyc 
pathways, was also conducted to assess the microbiome’s functional capabilities. Throughout, we report traditional phylum- and 
e4 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS
genus-level taxonomic names for clarity and continuity, while acknowledging recent taxonomic reclassifications.86 For certain taxa 
(e.g., GCA-900066905), the same identifier is used at both the family and genus levels due to incomplete taxonomic classification in 
reference databases.
Secondary clinical outcomes
Anthropometric measurements included height (stadiometer), weight (electric scales), and hip and waist circumferences (measuring 
tape). BMI was calculated as: BMI = (weight, kg)/(height, m).2 Serum inflammation markers (homocysteine, IL-β, IL-6, and TNF-α) 
were assayed using the BDtm Cytometric Bead Array platform (SA Pathology, Adelaide Women’s and Children’s Hospital). Serum 
leptin was measured using Merck Millipore radioimmunoassay kits (Royal Prince Alfred’s Central Sydney Pathology Services). Other 
serum biomarkers (glucose, insulin, liver function markers (ALT, GGT, ALP, AST, total bilirubin, albumin, protein, and globulin), and 
lipid markers (total cholesterol, HDL, LDL, non-HDL, LDL/HDL ratio, cholesterol/HDL ratio, triglycerides) were analyzed using 
Siemens’ ADVIA Chemistry kits (Australian Clinical Labs, Victoria). Self-reported measures included: mental health symptoms using 
the Depression Anxiety Stress Scale-21 (DASS-21),87 with higher scores indicating more severe symptoms; perceived well-being 
using the World Health Organization Wellbeing Scale (WHO-5),88 with higher scores indicating better well-being; sleep-related dif­
ficulties using the Athens Insomnia Scale (AIS),89 with higher scores indicating more severe issues; gastrointestinal symptoms using 
the Visual Analogue Scale for Irritable Bowel Syndrome (VAS-IBS),90,91 with higher scores indicating more favorable symptom ratings 
(‘‘very good’’); stool consistency using the Bristol Stool Form Scale (BSFS),92 a 7-point scale spanning from firmest (1) to softest (7) 
stool, with a mean score calculated for each participant across one week; and physical activity using the International Physical 
Activity Questionnaire-Short Form (IPAQ-SF),93 with categorical scores estimated (i.e., low, moderate, high).
QUANTIFICATION AND STATISTICAL ANALYSIS
All statistical analyses were conducted using R in the RStudio environment.94 Participant baseline characteristics were summarized 
with mean and standard deviation for continuous variables (or median and interquartile range where appropriate) and frequency and 
percentage for categorical variables. To assess both primary and secondary outcomes, we conducted our main analyses using a 
modified intention-to-treat (mITT) approach, which included all participants who provided baseline data, regardless of follow-up 
completion. We also conducted secondary complete case analyses, which included only participants who provided both baseline 
and week three data, to confirm the robustness of the results. We used linear mixed-effects regression (LMER) models to estimate 
between-group (food-based vs. supplement-based) differential changes (week three vs. baseline) in species-level alpha diversity, as 
measured by the Shannon index (primary outcome), and secondary outcomes using the lme495 package. The models included 
participant as a random effect, and group allocation, nominal time point, and the interaction between diet group and time point 
(i.e., diet group x time point) as fixed effects. The interaction estimated the between-group differential changes from baseline to 
week three using beta-coefficient point estimates (β) with 95% confidence intervals (95%CI) and p-values (two-tailed; p < 0.05 for 
significance).96 The supplement-based VLED group was set as the reference group. A positive β indicates a greater increase or 
smaller decrease in the food-based group compared to the supplement-based group. Conversely, a negative β reflects a greater 
increase or smaller decrease in the supplement-based group. When both groups change in the same direction (e.g., both increase 
or both decrease), the β reflects a difference in the magnitude of change. When groups change in opposite directions, the β reflects 
the divergence in trend, with the sign indicating which group changed more in the positive (or less in the negative) direction. Impor­
tantly, the β alone does not convey the absolute direction of change in either group; this must be determined by examining within- 
group estimates or visualized using change-from-baseline plots. Mean within-group changes (with 95%CIs) are provided to infer 
directionality, however should be interpreted with caution due to the limited sample size and low statistical power. Missing covariate 
and secondary outcome data were imputed using the mice97 package (five imputations), with predictive mean matching employed 
for continuous outcomes (alcohol intake) and proportional odds logistic regression employed for ordered factors (household income, 
menopause status). Interaction plots of the estimated marginal means were created using the emmeans98 package.
For beta diversity, we used complete cases to create individual CLR component-wise change scores (baseline minus week three) 
and then used permutational analysis of variance using adonis2 via the vegan99 package with 999 permutations to calculate the be­
tween-group differential change (i.e., beta diversity change ∼group). We report the R-squared (r2) statistic, providing a measure of 
the proportion of variance explained by the grouping factor (i.e., diet group) in the model.
We applied the Benjamini-Hochberg procedure100 to adjust for multiple comparisons. Outcomes were grouped into related 
categories (e.g., gastrointestinal outcomes) and multiple comparisons testing was conducted within each category (q < 0.1 for sig­
nificance given the small sample size). In post hoc sensitivity analyses for the primary outcome, we adjusted for prognostic covariates 
individually due to the limited sample size. These included country of birth, marital status, employment, medication use, stool 
consistency, BMI, physical activity, menopause status, and alcohol intake. Adjustments were based on baseline group imbalances, 
assessed using Standardized Mean Differences (>0.25, a threshold considered reasonable for acceptable standardized biases101). 
BMI was the sole anthropometric covariate considered due to its high correlation with weight and hip and waist circumferences. 
Additional models adjusted for age and postmenopausal status (SMDs <0.25), included to confirm the robustness of findings. 
One sample with a low read count was also removed.
Cell Reports Medicine 6, 102417, October 21, 2025 
e5 
Article
ll
OPEN ACCESS
Post hoc exploratory analyses
All post hoc analyses described below were exploratory and descriptive in nature. Consequently, no formal statistical inference was 
conducted, and no correction for multiple testing was applied. p-values are reported for transparency where relevant but should be 
interpreted with caution. We first examined post hoc Spearman’s correlations (r) between changes in our primary outcome (species- 
level Shannon index) and changes in secondary clinical outcomes over the three-week intervention period. Data were pooled across 
intervention groups. Only associations with r values ≥0.30 in magnitude were reported, consistent with thresholds indicating small- 
to-moderate or greater effect sizes.102
To explore the microbial contributors of within-sample alpha diversity change, we decomposed the Shannon diversity index into 
species-level contributions at both baseline and week three. Raw species-level metagenomic counts were transformed to relative 
abundances within each sample. For each taxon, its contribution to the Shannon index was calculated as −p⋅log(p), where p is 
the relative abundance. Mean contributions were then averaged within each group and timepoint, and the change in contribution 
(week three minus baseline) was computed for each taxon. To facilitate interpretation and reduce potential distortion from low-impact 
species, we identified the top 30 taxa with the largest absolute changes in Shannon contribution within each group. All remaining 
species were aggregated into a single ‘‘Other_net’’ category, capturing the summed net contribution of taxa outside the top 30, 
thus preserving the total change in diversity. Bar plots were generated to visualize the top 30 contributing species and the Other_net 
category within each intervention group. This descriptive visualization was used to qualitatively examine patterns in taxon-level con­
tributions to Shannon diversity change across groups.
To further provide some ecological context, the top 30 contributing species were also classified as dominant or subdominant 
based on their mean relative abundances at baseline across all samples in both arms: taxa with ≥1% mean relative abundance 
were considered dominant, while those <1% were classified as subdominant.103,104 This threshold was applied to distinguish be­
tween more prevalent community members and rarer taxa, helping to contextualize whether the top contributors to the change in 
Shannon index in each group were relatively abundant (i.e., dominant) or less abundant (i.e., subdominant) species at baseline.
Finally, to assess overall microbial community structure, taxonomic data were aggregated from species to phylum level using the 
tax_glom() function in the phyloseq105 package. To facilitate compositional comparisons across samples, counts were transformed 
to relative abundances by dividing each taxon’s count by the total count for that sample, using the microbiome106 package. The re­
sulting phylum-level relative abundances were visualized using stacked bar plots generated with ggplot2, stratified by intervention 
group (food-based vs. supplement-based) and time point (baseline vs. week three).
ADDITIONAL RESOURCES
Australian New Zealand Clinical Trials Registry Identifier: ACTRN12620000301965.
e6 
Cell Reports Medicine 6, 102417, October 21, 2025 
Article
ll
OPEN ACCESS