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