Fried-food-consumption-related-gut-microbiota-is-associate_2025_The-American.pdf

Original Research Article
Fried food consumption-related gut microbiota is associated with obesity, 
fat distribution, and cardiometabolic diseases: results from 2 large 
longitudinal cohorts with sibling comparison analyses
Yiting Duan 1,2, † , Yuwei Li 1,2, † , Chengquan Xu 1,2 , Wenjie Wang 1,2 , Xiaoyan Wang 1,2,
Weifang Zheng 3 , Julianna C Hsing 4 , Joyce Wu 5 , April Myers 5 , Ann W Hsing 4,5,6 , Wei He 1,2,7,* ,
Shankuan Zhu 1, 2, 7,**
1 Department of Nutrition and Food Hygiene, Children’s Hospital, Zhejiang University School of Medicine, National Clinical Research Center for 
Child Health, Hangzhou, Zhejiang, China; 2 Chronic Disease Research Institute, School of Public Health, School of Medicine, Zhejiang University, 
Hangzhou, Zhejiang, China; 3 Lanxi Hospital of Traditional Chinese Medicine, Lanxi, Zhejiang, China; 4 Department of Epidemiology and Population 
Health, Stanford School of Medicine, Palo Alto, CA, United States; 5 Department of Medicine, Stanford Prevention Research Center, Stanford School of 
Medicine, Palo Alto, CA, United States; 6 Stanford Cancer Institute, Stanford School of Medicine, Palo Alto, CA, United States; 7 Binjiang Institute of 
Zhejiang University, Hangzhou, Zhejiang, China
A B S T R A C T
Background: In prospective cohort studies, the relationship between fried food consumption, gut microbiota, obesity, and cardiometabolic diseases 
remains unknown.
Objectives: We aimed to explore associations of fried food consumption with gut microbiota and associations of fried food consumption-related 
microbiota with obesity and related disorders.
Methods: We analyzed 6637 individuals from the Wellness Living Laboratory China cohort (baseline 2016–2019) and 3466 from the Lanxi cohort 
(baseline 2017–2019), with follow-up until 24 June, 2024. Face-to-face interviews provided data on fried food consumption and other covariates. 
Analysis of 16S ribosomal ribonucleic acid data from fecal samples collected at baseline identified microbial genera. Body composition was evaluated 
using dual-energy x-ray absorptiometry. The microbiome multivariable associations with linear models helped identify genera associated with the 
frequency of fried food consumption in the cross-sectional analysis. Cox regression models examined the relationship of fried food consumption-related 
microbiota with cardiometabolic diseases during follow-up. Sibling comparison analyses were used to control for unmeasured familial confounders 
using the between-within model.
Results: Twenty-five microbial genera were significantly associated with fried food consumption frequency. Using these genera, we constructed a fried 
food consumption-related microbiota index. Meta-analysis of both cohorts found a positive relationship of this index with overall adiposity measures 
[body mass index (kg/m 2 )] [β coefficient: 0.26; 95% confidence interval (CI): 0.19, 0.32] and central fat distribution parameters [including android-
gynoid fat ratio (β: 1.48; 95% CI: 1.14, 1.82)]. Longitudinal analyses indicated that a higher fried food consumption-related microbiota index was 
linked to a higher risk of developing cardiometabolic diseases, with adjusted hazard ratios (95% CI) of 1.16 (1.07, 1.27) for diabetes and 1.16 (1.06, 
1.26) for major adverse cardiovascular events. Sibling comparison analyses yielded similar results.
Conclusions: Fried food consumption-related microbiome is associated with a higher risk of obesity, central fat distribution, and cardiometabolic 
diseases, emphasizing the importance of dietary choices in the management and prevention of chronic diseases.
Keywords: fried food consumption, gut microbiota, obesity, central fat distribution, cardiometabolic disease
Abbreviations: AOI, android-gynoid fat ratio; CI, confidence interval; FFQ, food frequency questionnaire; FMI, fried food consumption-related microbiota index; HbA1c, 
glycated hemoglobin; HR, hazard ratio; MACE, major adverse cardiovascular events; MASLD, metabolic dysfunction-associated steatotic liver disease; MetS, metabolic syndrome; 
OR, odds ratio; WELL-China, Wellness Living Laboratory China.
* Corresponding author.
** Corresponding author.
E-mail addresses: zjuhewei@zju.edu.cn (W. He), zsk@zju.edu.cn (S. Zhu).
† YD and YL are equal first authors.
journal homepage: https://ajcn.nutrition.org/
https://doi.org/10.1016/j.ajcnut.2025.06.025
Received 12 December 2024; Received in revised form 17 May 2025; Accepted 30 June 2025; Available online 2 July 2025
0002-9165/© 2025 The Authors. Published by Elsevier Inc. on behalf of American Society for Nutrition. This is an open access article under the CC BY license (http://creativeco 
mmons.org/licenses/by/4.0/).
The American Journal of Clinical Nutrition 122 (2025) 733–743
Results: Twenty-five microbial genera were dramatically associated with fried food consumption frequency. Using these genera, we constructed a fried
Introduction
Fried food, a ubiquitous component of many diets worldwide, has 
garnered considerable attention in the realm of public health due to its 
potential negative impact on human health [1]. It is reported that 
25–36% of adults in North America have food from fast food res-
taurants on a daily basis, usually fried food [1]. Multiple studies have 
provided evidence about the relationships between fried foods and risk 
of central fat accumulation, metabolic syndrome (MetS), as well as 
cardiometabolic diseases [2–4]. Understanding the complex interplay 
of fried food with health outcomes is crucial for developing effective 
strategies to reduce the diet-associated diseases burden and promote 
healthier dietary habits on a global scale.
Diet is a vital source of nutrients not only for humans but also for 
the gut microbiota, which consists of trillions of microbes located in 
the gastrointestinal tract [5]. The different components of our 
diet—fibers, proteins, fats, and micronutrients—shape the diversity, 
population, and activity of these microbial communities [6–8]. 
Research indicated that fried foods influence the presence of certain 
microbial genera [9–11]. For instance, a study with 117 overweight 
participants revealed that consuming fried meat reduced Flavoni-
fractor and Lachnospiraceae prevalence, whereas increasing that of 
Veillonella, Dorea, and Dialister [9]. However, population-based 
studies have provided inconsistent evidence about the relationship 
between fried foods and gut microbiota α-diversity [9,10,12], and little 
is known about the association of fried food consumption-related 
microbiome with obesity and cardiometabolic diseases.
To address these knowledge gaps, we utilized both cross-sectional 
and longitudinal methods to analyze data from 2 Chinese cohorts: the 
Wellness Living Laboratory China (WELL-China) cohort (Hangzhou 
City, Zhejiang Province) [13] and the Lanxi cohort (Lanxi City, 
Zhejiang Province) [14]. The primary aim of our study was to inves-
tigate 3 key questions: 1) Is there a relationship between fried food 
consumption and gut microbiota composition? 2) Is the 
fried-food-related microbiota associated with obesity and body fat 
distribution? 3) Is fried-food-related microbiota associated with the 
incidence rate of cardiometabolic diseases?
As a secondary aim, considering that gut microbiota, obesity, and 
cardiometabolic diseases often cluster within families [15], we 
examined whether genetic factors and shared early life environments 
might play a role in these associations. We performed sibling com-
parison analyses to take genetic and early life environment factors into 
consideration and obtain more reliable and robust results.
Methods
Data sources
This study used data from 2 prospective cohorts: the WELL-China 
cohort and the Lanxi cohort. Comprehensive details about the WELL-
China cohort [13] and the Lanxi cohort [14] have been published 
previously. In brief, the WELL-China cohort included 10,268 in-
dividuals aged 18–80 y recruited from 3 districts of Hangzhou, China. 
The WELL-China cohort recruited 3069 participants from Xihu dis-
trict from November 2016 to April 2017, 3055 participants from 
Shangcheng district from August 2017 to March 2018, and 4144 
participants from Gongshu district from September 2018 to May 2019. 
The Lanxi cohort, including Lanxi urban cohort and Lanxi rural 
cohort, included 4503 participants aged 18–80 y from urban (N = 
2698) and rural (N = 1805) areas of Lanxi, China. The recruitment 
periods for the Lanxi urban cohort and Lanxi rural cohort were from
June 2019 to August 2019 and from July 2017 to August 2017, 
respectively. The baseline investigation of the WELL-China cohort 
and Lanxi cohort both included a questionnaire survey, physical ex-
amination, dual-energy x-ray absorptiometry scan, and fecal and 
blood sample collection. The subsequent health conditions, including 
diabetes, cardiovascular diseases, cancer, and mortality, of all partic-
ipants in the 2 cohorts were monitored through the disease registry 
system.
The institutional review boards of Zhejiang University (number 
ZGL201507-3) and Stanford University (institutional review board-
35020) approved the protocol for the WELL-China cohort study. 
The Lanxi cohort study obtained approval from the ethics committee 
of the School of Public Health, Zhejiang University (number 
ZGL201905-1). All participants in both cohorts provided written 
informed consent.
Study population
The participant selection process for both cohorts is shown in 
Figure 1. After excluding individuals with abnormal energy intake 
(<500 kcal or >4000 kcal), inflammatory bowel diseases, baseline 
cancers, or missing information on gut microbiota, 6637 individuals 
from the WELL-China cohort were included for the final analyses. The 
selection process for the Lanxi cohort was similar to the above, with 
3466 participants included in the final analyses.
For sibling comparison analyses, we restricted to 431 participants 
from 175 families (Lanxi urban cohort: 151 siblings from 66 families; 
Lanxi rural cohort: 280 siblings from 109 families) where >2 siblings 
were involved in the Lanxi cohort. Sibling information was collected 
through a questionnaire survey.
Frequency of fried food consumption
In the WELL-China cohort, the information on habitual dietary 
consumption was evaluated through a validated food frequency 
questionnaire (FFQ) consisting of 26 items through face-to-face in-
terviews, as previously described [16,17]. The test-retest reliability of 
FFQ26 was assessed by comparing food intake measurements at 2 
time points separated by a 2-wk interval. The validity of FFQ26 was 
evaluated through comparative and correlational analyses of food 
intake, and energy and nutrient intake between FFQ26 and a 
comprehensive dietary frequency questionnaire (FFQ146), as well as 
between FFQ26 and a 3-d 24-h dietary record [16,17]. The question 
about the frequency of consumption of fried foods [fried breadstick 
(Youtiao), Chinese oil pancake, French fries, etc.] had 9 choices of 
responses, which ranged from never to daily. In the Lanxi urban 
cohort, habitual dietary consumption was evaluated through the China 
Health and Nutrition Survey (CHNS) 2002 FFQ, consisting of 46 
items, which was obtained from the China Health and Nutrition Sur-
vey [18]. This FFQ has been widely used in the literature [19,20]. 
Participants were requested to recall how many times they had fried 
wheat-based foods in the past year: _____ times/day; _____ time-
s/week; _____ times/month; _____ times/year. In final analyses, the 
frequency of consumption in the 2 cohorts was divided into 3 cate-
gories: <1 time/mo; 1~3 times/mo, and ≥4 times/mo, based on pre-
vious publications [21]. Dietary assessment and fecal and blood 
sample collection were all conducted at baseline.
Fried food consumption-related microbiota index
In baseline surveys in both cohorts, fecal samples were self-
collected by participants with guidance from cohort staff and placed 
in dry ice within 4 h and then transferred to long-term storage at
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
734
–80 ◦ C. Microbial DNA was extracted from each sample using the 
DNeasy PowerSoil kit (Qiagen) following the manufacturer’s in-
structions. DNA concentrations were determined using the NanoDrop 
1 spectrophotometer (NanoDrop Technologies) and Qubit 3.0 fluo-
rometer (Life Technologies). The V3-V4 hypervariable region of the 
16S rRNA gene was amplified from genomic DNA using primers 
341F (5'-CCTACGGGNGGCWGCAG-3’) and 805R (5'-GAC-
TACHVGGGTATCTAATCC-3’).
Amplicon sequencing was performed on the Illumina Nova-
Seq6000 (Illumina). Fastq files were demultiplexed, merge-paired, and 
quality filtered by 64-bit ultra-fast sequence analysis software (version 
7.0.1234) [22]. To obtain effective reads, marker gene Illumina 
sequence data, chimeric sequences, and low-quality regions of the 
sequences were detected and filtered. Filtered sequences were clus-
tered into amplicon sequence variants with 97% similarity. Amplicon 
sequence variants were annotated to the genus level using the SILVA 
ribosomal RNA database (version 138) [23].
After obtaining gut microbial genera for each participant, based on 
the 25 identified microbial genera, we created the fried food 
consumption-related microbiota index (FMI), which offers a 
comprehensive view of gut bacterial profiles linked to the frequency of 
fried food consumption (Supplemental Method 1).
Obesity data at baseline
Body weight, height, waist circumference, and hip circumference 
were assessed by trained personnel for both cohorts. BMI was
calculated as weight (kilogram)/height (meter) 2 . Waist-to-hip ratio was 
calculated as waist circumference (centimeters)/hip circumference 
(centimeters). Body fat mass and regional fat mass were evaluated via 
dual-energy x-ray absorptiometry scans (software version 11.40.004, 
GE Lunar Prodigy; GE Healthcare) [14]. For the android fat region, 
the upper boundary was a horizontal line drawn 20% of the way be-
tween the pelvis and the head, and the lower boundary was at the 
pelvis. For the gynoid fat region, the upper boundary begins slightly 
below the pelvic line, extending down 1.5 times the android region’s 
height [24]. Android fat mass percentage and gynoid fat mass per-
centage were calculated as android fat mass/total fat mass and gynoid 
fat mass/total fat mass, respectively. Additionally, the android-gynoid 
fat ratio (AOI) was calculated as android fat mass/gynoid fat mass.
Cardiometabolic conditions at baseline
Blood pressure, lipid profiles, fasting glucose, and glycated he-
moglobin (HbA1c) were examined at baseline. Glucose, total tri-
glycerides, HDL cholesterol, and total cholesterol in serum were 
measured on an automated analyzer (Roche cobas 8000 c701). LDL 
cholesterol and HbA1c were measured on an automated analyzer 
(Roche cobas 8000 c502). Hypertension was defined as having a 
systolic blood pressure of ≥140 mmHg, a diastolic blood pressure of
≥90 mmHg, or currently taking antihypertensive medications. Dys-
lipidemia was defined as having HDL cholesterol concentrations 
<1.0 mmol/L, LDL cholesterol concentrations ≥4.1 mmol/L, tri-
glyceride 
concentrations 
≥2.3 
mmol/L, 
total 
cholesterol
FIGURE 1. Flow diagram in the WELL-China cohort, the Lanxi cohort, and the Lanxi sibling subcohort. In the Lanxi urban cohort, after excluding 191 
participants with diet energy intake <500 kcal or >4000 kcal, 1811 participants were included in the final analyses, including 1069 participants with fried food 
consumption frequency <1 times/mo, 437 participants with fried food consumption frequency 1~3 times/mo, and 305 participants with fried food consumption 
frequency ≥4 times/mo. We did not collect the fried food consumption frequency in the Lanxi rural cohort. rRNA, ribosomal ribonucleic acid; WELL-China, 
wellness living laboratory China.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
735
TABLE 1
Baseline characteristics of the participants according to frequency of fried food consumption in the wellness living laboratory China cohort and the Lanxi urban 
cohort 1.
All participants
Fried food consumption frequency, times/month
<1
1~3
≥4
WELL-China cohort 
Participants, n (%)
6637 (100.0)
4171 (62.8)
1549 (23.3)
917 (13.8)
Age, years
56.0 (12.5)
58.5 (11.0)
52.3 (13.5)
50.7 (13.9)
Sex, n (% of males) 
2724 (41.0)
1721 (41.3)
593 (38.3)
410 (44.7)
Marital status, n (%)
Married
5930 (89.3)
3777 (90.6)
1365 (88.1)
788 (85.9)
Others
707 (10.7)
394 (9.4)
184 (11.9)
129 (14.1)
Education level, n (%)
≤6 y
1377 (20.7)
1013 (24.3)
233 (15.0)
131 (14.3)
6–12 y
3873 (58.4)
2505 (60.1)
851 (54.9)
517 (56.4)
≥ 12 y
1383 (20.8) 
651 (15.6)
463 (29.9)
269 (29.3)
Income level (Chinese yuan/year/person), n (%)
≤50000
4307 (64.9)
2884 (69.1)
881 (56.9)
542 (59.1)
50001–109999 
1921 (28.9)
1078 (25.8)
541 (34.9)
302 (32.9)
≥110000
404 (6.1) 
206 (4.9)
125 (8.1) 
73 (8.0)
Smoking status, n (%) 
Nonsmokers
4876 (73.5)
3067 (73.5)
1171 (75.6)
638 (69.6)
Former smokers
514 (7.7)
344 (8.2)
113 (7.3)
57 (6.2)
Current smokers
1243 (18.7)
758 (18.2)
263 (17.0)
222 (24.2)
Alcohol intake, n (%) 
Nondrinker
3630 (54.7)
2346 (56.2)
828 (53.5)
456 (49.7)
Occasional
1646 (24.8)
940 (22.5)
443 (28.6)
263 (28.7)
Regular
1361 (20.5)
885 (21.2)
278 (17.9)
198 (21.6)
Physical activity level, n (%) 2 
Low activity
1085 (16.3)
638 (15.3)
281 (18.1)
166 (18.1)
Moderate activity
3228 (48.6)
2044 (49.0)
753 (48.6)
431 (47.0)
High activity
2319 (34.9)
1487 (35.7)
512 (33.1)
320 (34.9)
Diet energy intakes, kcal 3 
1354.7 (531.1)
1325.7 (510.6)
1320.5 (500.4)
1544.4 (625.8)
Living area, n (%)
Xihu district
1176 (17.7)
726 (17.4)
278 (17.9)
172 (18.8)
Shangcheng district
2279 (34.3)
1382 (33.1)
569 (36.7)
328 (35.8)
Gongshu district
3182 (47.9)
2063 (49.5)
702 (45.3)
417 (45.5)
BMI, kg/m 2 
23.9 (3.2) 
23.7 (3.1) 
24.0 (3.4) 
24.3 (3.5)
Waist circumference, cm
83.9 (9.8)
83.6 (9.6)
83.9 (10.1) 
85.2 (10.4)
Waist-hip ratio, %
90.3 (7.6)
90.4 (7.4)
89.8 (7.5) 
90.9 (8.4)
Android fat percentage, %
10.9 (1.8)
11.0 (1.8)
10.7 (1.8)
11.0 (1.8)
Gynoid fat percentage, %
16.6 (3.0)
16.5 (3.0)
16.9 (3.2)
16.7 (3.1)
Android-gynoid fat ratio, %
69.2 (21.5)
69.9 (21.3)
67.2 (21.7)
69.1 (22.0)
16S rRNA Sequencing batch, n (%) 
First
779 (11.7)
503 (12.1)
178 (11.5)
98 (10.7)
Second
4648 (70.0)
2908 (69.7)
1081 (69.8)
659 (71.9)
Third
1148 (17.3)
721 (17.3)
277 (17.9)
150 (16.4)
Fourth
62 (0.9)
39 (0.9) 
13 (0.8) 
10 (1.1)
16S rRNA sequencing depth
150,590.7 (59,494.0)
150,586.7 (60,650.4)
151,080.7 (56,845.8)
149,780.7 (58,602.2)
Lanxi urban cohort
Participants, n (%)
1811 (100.0)
1069 (59.0)
437 (24.1)
305 (16.8)
Age, years
57.3 (11.5)
59.4 (10.7)
55.6 (12.0)
51.9 (11.8)
Sex, n (% of males)
651 (35.9)
358 (33.5)
157 (35.9)
136 (44.6)
Marital status, n (%)
Married
1646 (90.9)
973 (91.0)
396 (90.6)
277 (90.8)
Others
165 (9.1) 
96 (9.0) 
41 (9.4) 
28 (9.2)
Education level, n (%)
≤6 y
555 (30.6)
366 (34.2)
125 (28.6)
64 (21.0)
6–12 y
982 (54.2)
584 (54.6)
229 (52.4)
169 (55.4)
≥12 y
274 (15.1)
119 (11.1)
83 (19.0)
72 (23.6)
Income level (Chinese yuan/year/family), n (%)
≤50000
568 (31.4)
352 (32.9)
143 (32.7)
73 (23.9)
50001–109999
630 (34.8)
380 (35.5)
149 (34.1)
101 (33.1)
≥110000
613 (33.8)
337 (31.5)
145 (33.2)
131 (43.0)
Smoking status, n (%) 
Nonsmokers
1462 (80.7)
898 (84.0)
339 (77.6)
225 (73.8)
Former smokers
113 (6.2)
67 (6.3)
27 (6.2)
19 (6.2)
Current smokers 
236 (13.0)
104 (9.7)
71 (16.2)
61 (20.0)
Alcohol intake, n (%)
(continued on next page)
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
736
concentrations ≥6.2 mmol/L, or using lipid-lowering medications. 
MetS was defined according to revised criteria set by the Interna-
tional Diabetes Federation [25]. Specifically, MetS is diagnosed 
based on a waist circumference of ≥90 cm in males and ≥80 cm in 
females, along with the presence of any 2 of the following criteria: 
elevated triglycerides (>150 mg/dL or 1.7 mmol/L), decreased HDL 
cholesterol (<40 mg/dL or 1.03 mmol/L in males, <50 mg/dL or 
1.29 mmol/L in females), specific treatment for triglycerides or HDL 
cholesterol abnormality, elevated blood pressure (systolic ≥130 
mmHg or diastolic ≥85 mmHg), treatment of previously diagnosed 
hypertension, elevated fasting plasma glucose (≥100 mg/dL or 5.6 
mmol/L), and previously diagnosed type 2 diabetes. Diabetes was 
diagnosed based on fasting blood glucose concentrations ≥7.0 
mmol/L, HbA1c concentrations ≥6.5% (48 mmol/mol), or currently 
using diabetes medications. Metabolic dysfunction-associated stea-
totic liver disease (MASLD) was identified based on a consensus 
review of ultrasound images conducted by skilled operators. The 
detailed methodology of the ultrasound image consensus review has 
been previously published [26,27].
Cardiometabolic diseases during the follow-up
The deaths and hospitalizations of the participants were monitored 
through an established disease registration system, including the inpa-
tient system and the outpatient system. Participants were monitored from 
baseline (2016–2019) until the occurrence of cardiometabolic disease, 
mortality, or 24 June, 2024, whichever came first. The International 
Statistical Classification of Diseases and Related Health Problems, 10th 
Revision (ICD-10) diagnostic codes were used to define cardiometabolic 
diseases, such as incident diabetes (E11–E14) and major adverse car-
diovascular events (MACE), including death, myocardial infarction 
(I20–I25), heart failure (I50), or stroke (I60–I64, I69) [28].
Covariates
Data on age (years), sex (male or female), residential area (Xihu 
district, Shangcheng district, or Gongshu district), physical activity
levels (low, moderate, or high), smoking status (nonsmokers, 
former smokers, or current smokers), alcohol intake (nondrinker, 
occasional, or regular), antibiotic use (yes or no), annual income 
(≤50000 CNY, 50000–109999 CNY or ≥110000 CNY), marital 
status (married or others), and education level (≤6 y, 6–12 y, or
≥12 y) were gathered through standardized structured question-
naires [29,30]. Dietary energy consumption was evaluated via a 
validated FFQ consisting of 26 items in the WELL-China cohort 
[16] and a validated FFQ consisting of 46 items in the Lanxi urban 
cohort [18].
For categorical covariates, missing data were assigned to a separate 
“Missing” category, as detailed in Table 1 and Supplemental Table 1. 
For the continuous covariate 16S rRNA sequencing depth, 1 missing 
sample each in the WELL-China cohort and the Lanxi cohort was 
imputed using the mean value. Diet energy intake in the Lanxi rural 
cohort was imputed using the mean value from the Lanxi urban cohort 
due to the unavailability of dietary data.
Statistical analysis
Microbiome analyses
To explore the relationships of fried food consumption frequency 
with gut microbiota α-diversity, multivariable linear regression was 
employed. Permutational multivariate analysis of variance with 999 
permutations was used to investigate the relationship of fried food 
consumption frequency with gut microbiota β-diversity [31]. Micro-
biome multivariable associations with linear models [32] were utilized 
to identify potential gut microbial genera linked to fried food con-
sumption frequency. To mitigate false discoveries from multiple tests, 
the Benjamini-Hochberg method was employed, setting a Q value 
(false discovery rate-adjusted P value) below 0.25 as statistically 
significant. To validate the collective effects of genera (FMI) con-
structed by the above-identified gut microbial genera, the association 
between fried food consumption frequency and FMI was investigated 
through multivariable linear regression.
TABLE 1 (continued )
All participants
Fried food consumption frequency, times/month
<1
1~3
≥4
Nondrinker
1136 (62.7)
713 (66.7)
260 (59.5)
163 (53.4)
Occasional
417 (23.0)
212 (19.8)
120 (27.5)
85 (27.9)
Regular
258 (14.2)
144 (13.5)
57 (13.0)
57 (18.7)
Physical activity level, n (%) 2 
Low activity
459 (25.3)
268 (25.1)
113 (25.9)
78 (25.6)
Moderate activity
984 (54.3)
596 (55.8)
224 (51.3)
164 (53.8)
High activity
366 (20.2)
205 (19.2)
100 (22.9)
61 (20.0)
Diet energy intakes, kcal 3
2209.9 (605.6)
2,173.8 (612.8)
2201.4 (576.4)
2348.4 (603.4)
BMI, kg/m 2 
23.6 (3.2) 
23.3 (2.9) 
23.7 (3.5) 
24.4 (3.6)
Waist circumference, cm
83.9 (9.6)
83.3 (9.3)
84.2 (10.2) 
85.8 (9.9)
Waist-hip ratio, %
91.4 (7.7)
91.3 (7.9)
91.4 (7.6) 
92.0 (7.2)
Android fat percentage, %
10.9 (1.6)
10.8 (1.5)
10.9 (1.5)
11.0 (1.6)
Gynoid fat percentage, %
16.2 (2.9)
16.3 (2.8)
16.2 (2.9)
15.9 (2.9)
Android-gynoid fat ratio, %
70.3 (20.3)
69.5 (20.0)
70.4 (20.1)
72.6 (21.5)
16S rRNA Sequencing depth
198,688.9 (58,289.7)
197,969.9 (57,780.5)
198,366.4 (62,322.7)
201,668.7 (54,027.6)
Abbreviations: BMI, body mass index; rRNA, ribosomal ribonucleic acid; SD, standard deviation; WELL-China, wellness living laboratory China.
1 Data were presented as means (SDs) for continuous variables and numbers (percentage) for categorical variables. The missing data (normally ≤5) were not 
presented in the table.
2 Physical activity was classified into 3 groups according to the international physical activity questionnaire.
3 Diet energy intakes in the WELL-China cohort were measured via a 26-item validated food frequency questionnaire (FFQ). The 46-item validated FFQ was 
used in the Lanxi urban cohort.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
737
Cross-sectional analyses
Multivariable linear regression was utilized to explore the rela-
tionship of FMI with obesity and fat distribution. In addition, multi-
variable logistic regression was applied to explore the association of 
FMI with cardiometabolic conditions, including hypertension, MetS, 
dyslipidemia, diabetes, and MASLD.
Longitudinal analyses
Longitudinal analyses were performed using a Cox regression 
model to examine the association of FMI with the onset of car-
diometabolic diseases, such as diabetes and MACEs. The Schoenfeld 
residual test was used to test the proportional assumption of the Cox 
regression (in the WELL-China cohort, the P values corresponding to 
the diabetes model, MACE model, and diabetes-MACE combined 
model were 0.741, 0.236, and 0.533, respectively. In the Lanxi cohort, 
the P values for the 3 models were 0.035, 0.287, and 0.025, respec-
tively.). To investigate the role of adiposity in the relationship between 
FMI and cardiometabolic diseases, we additionally controlled for BMI 
and AOI on the basis of the fully adjusted model.
Meta-analyses
We conducted meta-analyses using the Mantel-Haenszel method 
to derive combined effect estimates from the WELL-China cohort 
and the Lanxi cohort. Cochran’s Q test was utilized to assess het-
erogeneity within the meta-analysis. A fixed-effects model was 
utilized.
Sibling comparison analyses
To address confounding factors common within families, sibling 
comparison analyses were conducted. The analyses of FMI with 
obesity and cardiometabolic conditions were repeated using the 
between-within model [33]. This was implemented through the lmer 
and glmer functions in the “lme4” package (version 1.1.27) in R. The 
between-within model divides the covariate effect into 2 independent 
components: the between-cluster effect and the within-cluster effect 
to eliminate the bias caused by the generalized linear mixed model 
[33,34].
We used R version 4.1.2 for statistical analysis, and a 2-sided P 
value <0.05 was considered statistically significant.
FIGURE 2. Fried food consumption and β-diversity in the participants of the WELL-China cohort and the Lanxi urban cohort. Dietary assessment and fecal 
samples collection were all conducted at baseline. Principal coordinates analysis(PCoA) was conducted on the WELL-China cohort (N = 6637) and the Lanxi 
urban cohort (N = 1811) using Bray-Curtis dissimilarity for (A 1 and B 1 ), and weighted Unifrac dissimilarity for (A 2 and B 2 ). A PERMANOVA test was 
employed, utilizing 999 permutations, to explore the relationship of the frequency of fried food consumption with β-diversity, whereas controlling for age, sex, 
residential area, annual income, marital status, education level, alcohol consumption, physical activity levels, smoking habits, diet energy intakes, antibiotics 
use (specific to the WELL-China cohort), sequencing batch, and sequencing depth. PERMANOVA, permutational multivariate analysis of variance; WELL-
China, wellness living laboratory China.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
738
Results
Population characteristics
Table 1 presents the baseline characteristics by frequency of fried 
food consumption in the WELL-China cohort and the Lanxi urban 
cohort. Supplemental Table 1 presents the baseline characteristics for 
the Lanxi rural cohort and the Lanxi sibling subcohort (both part of the 
Lanxi cohort).
Fried food consumption and gut microbiota
When comparing higher to lower frequency of fried food con-
sumption, a significant difference in β-diversity (gut microbial 
composition) was observed in both the WELL-China cohort (R 2 = 
0.00035, P < 0.001) and the Lanxi urban cohort (R 2 = 0.00127, P = 
0.002) (Figure 2A 1 and B 1 ). Analysis of β-diversity, considering 
phylogenetic relationships using weighted Unifrac distances, supported 
these findings (Figure 2A 2 and B 2 ). Twenty-five microbial genera were
FIGURE 3. Fried food consumption and identified microbial genera in the participants of the WELL-China cohort. Dietary assessment and fecal sample 
collection were all conducted at baseline. (A) Multivariate analysis by linear models were used to identify microbial genera associated with frequency of fried 
food consumption in the WELL-China cohort (N = 6637), adjusted for age, sex, residential area, annual income, marital status, education level, alcohol 
consumption, physical activity levels, smoking habits, diet energy intakes, antibiotics use, sequencing batch, and sequencing depth. The Q values (false 
discovery rate-adjusted P value) were calculated using the Benjamini-Hochberg method (Q value <0.25). (B) Multivariable linear regression models were 
utilized to explore the relationship of Fried food consumption-related microbiota index (FMI) (per SD unit increase) with frequency of fried food consumption 
in participants of the WELL-China cohort (left) and the Lanxi urban cohort (right), adjusted for age, sex, residential area, annual income, marital status, 
education level, alcohol consumption, physical activity levels, smoking habits, diet energy intakes, antibiotics use (specific to the WELL-China cohort), 
sequencing batch, and sequencing depth. In the WELL-China cohort (N = 6637), compared with participants whose consumption frequency of fried food was 
<1 time/mo, participants whose consumption frequency of fried food were 1~3 times/mo and ≥4 times/mo had higher FMI, respectively (1~3 times/mo: β = 
0.18; 95% CI: 0.12, 0.24; ≥4 times/mo: β = 0.21; 95% CI: 0.14, 0.29). In the Lanxi urban cohort (N = 1811), compared with the <1 time/mo group, β (95% 
CI) for the 1~3 times/mo group and the ≥4 times/mo group were 0.07 (–0.04, 0.18) and 0.15 (0.02, 0.29), respectively. CI, confidence interval; SD, standard 
deviation; WELL-China, wellness living laboratory China.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
739
When comparing lower to lower frequency of fried food con-
statistically significantly linked to the fried food consumption fre-
quency (false discovery rate <0.25) in the WELL-China cohort 
(Figure 3A, Supplemental Table 2). After adjusting for the frequency of 
fruit intake and vegetable intake, only 1 unannotated genus-level bac-
terium was removed, demonstrating the reliability of our results 
(Supplemental Table 3). However, associations of the 3 groups cate-
gorized by fried food consumption frequency with α-diversity (gut 
microbial richness) were not significant (Supplemental Figure 1).
FMI
The FMI, comprising 25 identified microbial genera, was associ-
ated with the fried food consumption frequency in the WELL-China 
cohort (Figure 3B). Compared with participants whose consumption 
frequency of fried food was <1 time/mo, participants whose con-
sumption frequency of fried food were 1~3 times/mo and ≥4 times/mo 
had higher FMI, respectively [1~3 times/mo: β: = 0.18; 95%
confidence interval (CI): 0.12, 0.24; ≥4 times/mo: β = 0.21; 95% CI: 
0.14, 0.29] (Figure 3B). This association was also observed in the 
Lanxi urban cohort, underscoring the robustness of the FMI 
(Figure 3B).
FMI and obesity
Table 2 presents the relationship of FMI with obesity measures. No 
significant heterogeneity was found between the cohorts. A meta-
analysis found a strong positive relationship of FMI with obesity 
indicators including BMI (β: 0.26; 95% CI: 0.19, 0.32), waist 
circumference (β: 0.79; 95% CI: 0.61, 0.97), waist-to-hip ratio 
(β: 0.43; 95% CI: 0.29, 0.56), android fat percentage (β: 0.09; 95% CI: 
0.06, 0.11), and AOI (β: 1.48; 95% CI: 1.14, 1.82), whereas a negative 
relationship with gynoid fat percentage (β: –0.23; 95% CI: –0.28, 
–0.17). Similarly, in sibling comparisons, significant associations were 
observed across these obesity measures.
TABLE 2 
Fried food consumption-related microbiota index and prevalent obesity in the participants of the wellness living laboratory China cohort, Lanxi cohort, and 
Lanxi sibling subcohort.
Prevalent
Total cohort analysis 1
Sibling comparison analysis 2
WELL-China cohort
Lanxi cohort
Meta-analyzed
Lanxi sibling subcohort
N
β (95% CI)
N
β (95% CI)
β (95% CI)
P heterogeneity
N
β (95% CI)
BMI (kg/m 2 )
6629
0.23 (0.15, 0.31)
3446
0.31 (0.20, 0.42)
0.26 (0.19, 0.32)
0.249
431
0.39 (0.12, 0.66)
Waist-circumference (cm)
6628
0.81 (0.59, 1.03)
3451
0.74 (0.43, 1.05)
0.79 (0.61, 0.97)
0.718
431
1.17 (0.37, 1.98)
Waist-hip ratio (%)
6627
0.47 (0.31, 0.64)
3451
0.34 (0.11, 0.57)
0.43 (0.29, 0.56)
0.368
431
0.89 (0.35, 1.43)
Android percentage (%)
6568
0.09 (0.06, 0.13)
3408
0.08 (0.03, 0.12)
0.09 (0.06, 0.11)
0.731
430
0.10 (− 0.02, 0.22)
Gynoid percentage (%)
6568
–0.22 (–0.28, –0.15)
3408
–0.24 (–0.33, –0.15)
–0.23 (–0.28, –0.17)
0.724
430
–0.39 (–0.61, –0.16)
Android-gynoid ratio (%)
6568
1.43 (1.01, 1.85)
3408
1.58 (0.98, 2.18)
1.48 (1.14, 1.82)
0.688
430
2.32 (0.69, 3.94)
Abbreviations: BMI, body mass index; CI, confidence interval; FMI, fried food consumption-related microbiota index; SD, standard deviation; WELL-China, 
wellness living laboratory China.
1 Multivariable linear regression models were utilized to explore the relationship of FMI (per SD unit increase) with obesity in the total cohort analysis, 
adjusted for age, sex, residential area, annual income, marital status, education level, alcohol consumption, physical activity levels, smoking habits, diet energy 
intakes, antibiotics use (specific to the WELL-China cohort), sequencing batch, and sequencing depth.
2 Between-within models were employed to assess the association between FMI (per SD unit increase) and obesity in a sibling comparison analysis, adjusted 
for age, sex, residential area, annual income, marital status, education level, alcohol consumption, physical activity levels, smoking habits, diet energy intakes, 
antibiotics use (specific to the WELL-China cohort), sequencing batch, and sequencing depth.
TABLE 3
Fried food consumption-related microbiota index and prevalent cardiometabolic conditions in the participants of the wellness living laboratory China cohort,
Lanxi cohort, and Lanxi sibling subcohort 1.
Prevalent
Total cohort analysis 2
Sibling comparison analysis 3
WELL-China cohort
Lanxi cohort
Meta-analyzed
Lanxi sibling subcohort
N
OR (95% CI)
N
OR (95% CI)
OR (95% CI)
P heterogeneity
N
OR (95% CI)
Hypertension
6605
1.11 (1.05, 1.18)
3461
1.07 (0.99, 1.15)
1.09 (1.05, 1.15)
0.449
429
1.37 (1.09, 1.71)
Dyslipidemia
6628
1.08 (1.02, 1.14)
3461
1.18 (1.10, 1.27)
1.12 (1.07, 1.17)
0.056
430
1.21 (0.97, 1.51)
Metabolic syndrome
6620
1.21 (1.14, 1.28)
3448
1.21 (1.12, 1.30)
1.21 (1.16, 1.27)
1.000
428
1.27 (1.01, 1.60)
Diabetes
6629
1.29 (1.19, 1.40)
3462
1.27 (1.13, 1.42)
1.28 (1.20, 1.37)
0.827
430
1.48 (1.03, 2.12)
MASLD 4
2547
1.16 (1.07, 1.26)
3466
1.25 (1.16, 1.34)
1.21 (1.15, 1.28)
0.113
431
1.18 (0.94, 1.47)
Abbreviations: CI, confidence interval; FMI, fried food consumption-related microbiota index; MASLD, metabolic dysfunction-associated steatotic liver
disease; OR, odds ratio; SD, standard deviation; WELL-China, wellness living laboratory China.
1 Prevalent cardiometabolic conditions were defined using baseline data.
2 Multivariate logistic regression models were employed to assess the relationships between FMI (per SD unit increase) and prevalent cardiometabolic 
conditions in the total cohort, adjusted for age, sex, residential area, annual income, marital status, education level, alcohol consumption, physical activity levels, 
smoking habits, diet energy intakes, antibiotics use (specific to the WELL-China cohort), sequencing batch, and sequencing depth. The estimates were subjected 
to meta-analysis using a fixed-effect model.
3 Between-within models were utilized to evaluate the relationship between FMI (per SD unit increase) and prevalent cardiometabolic conditions in sibling 
comparison analysis, adjusted for age, sex, residential area, annual income, marital status, education level, alcohol consumption, physical activity levels, 
smoking habits, diet energy intakes, antibiotics use (specific to the WELL-China cohort), sequencing batch, and sequencing depth.
4 The diagnosis of MASLD was conducted exclusively in the Gongshu district of the WELL-China cohort.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
740
FMI and prevalent cardiometabolic conditions
Table 3 presents the relationship between FMI and prevalent car-
diometabolic conditions. No significant heterogeneity was found be-
tween the 2 cohorts. A meta-analysis found a positive relationship of 
FMI with cardiometabolic conditions, with each SD unit increase in 
FMI associated with higher risk of hypertension [odds ratio (OR): 
1.09; 95% CI: 1.05, 1.15], dyslipidemia (OR: 1.12; 95% CI: 1.07, 
1.17), MetS (OR: 1.21; 95% CI: 1.16, 1.27), diabetes (OR: 1.28; 95% 
CI: 1.20, 1.37), and MASLD (OR: 1.21; 95% CI: 1.15, 1.28). 
Consistent results were found in sibling analyses for hypertension 
(OR: 1.37; 95% CI: 1.09, 1.71), MetS (OR: 1.27; 95% CI: 1.01, 1.60), 
and diabetes (OR: 1.48; 95% CI: 1.03, 2.12).
FMI and incident cardiometabolic diseases
Table 4 presents the relationship of FMI with incident car-
diometabolic diseases. No significant heterogeneity was found be-
tween the 2 cohorts. A meta-analysis found a positive relationship of 
FMI with incident cardiometabolic diseases, with each SD unit in-
crease in FMI associated with higher risk of incident diabetes [hazard 
ratio (HR): 1.16; 95% CI: 1.07, 1.27] and MACEs (HR: 1.16; 95% CI: 
1.06, 1.26). The association of FMI with cardiometabolic diseases 
diminished after adjusting for obesity (BMI) and central fat distribu-
tion (AOI), suggesting that part of the association between FMI and 
cardiometabolic diseases may be mediated through fat and fat distri-
bution (Supplemental Figures 2 and 3). Fried food consumption-
related microbiota was independently associated with incident 
cardiometabolic diseases after controlling for diet energy intake or 
dietary diversity score (Supplemental Tables 4 and 5).
Discussion
In this study involving 2 population-based cohorts in China, we 
observed a significant association between gut microbiota β-diversity 
and fried food consumption frequency. We detected no associations 
between fried food consumption frequency and gut microbiota 
α-diversity. Further analyses identified 25 genera related to the fried 
food consumption frequency in the WELL-China cohort. FMI, based 
on these 25 genera, was linked to the fried food consumption fre-
quency in the WELL-China cohort, and the association was validated 
in the Lanxi urban cohort. We found that FMI had a strong association
with obesity, central fat distribution, and various cardiometabolic 
conditions in both cross-sectional and longitudinal analyses. Results of 
sibling comparison analyses provided consistent evidence, confirming 
that gut microbiomes associated with fried food consumption are 
associated with obesity and cardiometabolic diseases, independent of 
genetic and early life environmental factors.
Several limitations should be considered. First, residual con-
founding still persists despite adjustment for multiple covariates. 
Second, we did not gather extensive details on the specific type of oil 
(vegetable or animal oil) used for frying, the method (deep or pan) 
used for frying, and other details about fried food consumption that 
could potentially affect results. Third, in our microbial analysis, the 
exclusive use of 16S rRNA sequencing rather than metagenomic 
sequencing may have limited functional insights into the microbiome. 
Fourth, in the prospective analysis, the statistical power was con-
strained by insufficient incident case numbers, precluding robust 
assessment of specific MACE endpoints (such as stroke, heart failure). 
Fifth, the observational design used to investigate the association be-
tween fried food consumption and gut microbiota precludes robust 
causal inference. Last, as both of our cohorts are Chinese, caution 
needs to be taken when generalizing our findings.
Three previous studies, limited by small sample sizes, have exam-
ined the link between fried foods and gut microbiota α-diversity, and 
have shown inconsistent results [9,10,12]. A randomized controlled trial 
in 117 adults revealed no changes in Chao1/Shannon diversity indices 
between high fried meat intake and control groups [9], whereas a 
cross-sectional analysis of 862 adults revealed an inverse link between 
fried food consumption and α-diversity indices [10]. The inconsistent 
findings could stem from short trial durations and/or limited sample 
sizes. Our observational study on gut microbiota in nearly 10,000 in-
dividuals has greater statistical power compared to previous studies, 
suggesting that our findings of no relationship of frequency of fried food 
consumption with the overall richness of gut microbiota are reliable. 
Our findings of significant variations in gut microbe β-diversity 
associated with fried food consumption are in line with prior studies 
[10,11], implying that fried food consumption is significantly related 
to the gut microbiota overall structure. We identified 25 genera asso-
ciated with fried food consumption, suggesting selective associations 
between fried food consumption and specific gut microbiota genera. 
Although the mechanisms by which fried foods alter gut microbiota 
remain unclear, frying-induced chemical reactions (e.g., polymer
TABLE 4
Fried food consumption-related microbiota index and incident cardiometabolic diseases in the participants of the wellness living laboratory China cohort and 
Lanxi cohort 1.
Incident
WELL-China cohort
Lanxi cohort
Meta-analyzed
Case/N
HR (95% CI) 2
Case/N
HR (95% CI) 2
HR (95% CI) 3
P heterogeneity
Diabetes 4
419/5751
1.14 (1.03, 1.26)
140/3039
1.23 (1.04, 1.46)
1.16 (1.07, 1.27)
0.450
Major adverse cardiovascular events
356/6637
1.14 (1.02, 1.27)
209/3412
1.19 (1.03, 1.36)
1.16 (1.06, 1.26)
0.635
Diabetes and major adverse cardiovascular events
620/5724
1.13 (1.04, 1.22)
285/3011
1.21 (1.07, 1.36)
1.15 (1.08, 1.23)
0.352
Abbreviations: CI, confidence interval; HR, hazard ratio; SD, standard deviation; WELL-China, wellness living laboratory China.
1 Cardiometabolic disease incidents were identified based on the disease register data from the inpatient system and the outpatient system.
2 Cox models were utilized to explore the relationship of per SD unit increase in FMI with risk of diabetes, major adverse cardiovascular events, and car-
diometabolic diseases (diabetes and major adverse cardiovascular events), adjusted for age, sex, residential area, annual income, marital status, education level, 
alcohol consumption, physical activity levels, smoking habits, diet energy intakes, antibiotics use (specific to the WELL-China cohort), sequencing batch, and 
sequencing depth. Patients were monitored from the beginning of the study until the occurrence of cardiometabolic diseases, death, or the end of the follow-up on 
24 June, 2024, whichever came first. The median follow-up time was 6 y in the WELL-China cohort and 5 y in the Lanxi cohort.
3 Estimates were meta-analyzed using a fixed-effect model.
4 Diabetes: In the longitudinal analysis of the onset of diabetes, we excluded participants with diabetes at baseline. The exclusion criteria were self-reported use 
of diabetes medication or a baseline fasting blood glucose concentration >7.0 mmol/L or a baseline diagnosis of diabetes in the disease registry system.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
741
CI: 1.20, 1.37), and MASLD (OR: 1.21; 95% CI: 1.15, 1.28). Second, we did gather extensive details on the specific type of oil
formation, trans-fatty acid) [35] may affect the health and abundance 
of some microbiota. A clinical trial has shown that the intake of fried 
meat not only affects the abundance of some bacteria but also causes 
significant changes in fecal metabolites [9]. In an association analysis 
of 19 dietary variables with gut microbiota β-diversity, fried products 
explained the second-highest proportion of variation in gut microbiota 
composition, surpassed only by raw fruit intake, suggesting its po-
tential microbial impacts [10].
Since the effect of individual microbial genera is limited, we 
constructed an index representing the fried food consumption-related 
gut microbiota, termed FMI, to quantify the cumulative effect of 
variation across multiple identified genera. We found that FMI was 
linked to overall obesity and central fat accumulation, with the latter 
showing a stronger association with cardiometabolic risk compared to 
overall obesity [36]. Fried food consumption-related microbiota were 
independently associated with central fat distribution after controlling 
for overall obesity (BMI) (Supplemental Table 6), suggesting fried 
food consumption-related microbiota may contribute to the accumu-
lation of abdominal fat, and, by extension, greater cardiometabolic 
risk, independent of BMI.
In our cross-sectional and longitudinal analyses, we found a posi-
tive association between FMI and cardiometabolic diseases, including 
MACEs and diabetes. Identifying new gut microbiota related to fried 
food consumption may help elucidate the mechanism determining the 
relationship of fried food consumption with cardiometabolic risk. 
Notably, the association between FMI and cardiometabolic diseases 
was reduced when controlled for BMI and AOI, as indicated in Sup-
plemental Figures 2 and 3. This suggests that overall obesity and 
central fat distribution may play a mediating role in the relationship of 
fried food consumption-associated gut microbiota with car-
diometabolic diseases.
The sibling comparison design reinforced our results. Recent 
studies have found extensive sharing of bacterial strains among in-
dividuals, with patterns of intrahousehold and intrapopulation trans-
mission [15]. Obesity and cardiometabolic diseases are affected by 
common environmental and genetic factors that are difficult to capture 
in cohort studies. The sibling comparison design is a powerful method 
for mitigating these confounding factors [37], thereby boosting the 
reliability of our results. Given the limited sample size in our study, 
investigating whether siblings discordant in fried food consumption 
exhibit differences in their gut microbiota composition also presents a 
valuable direction for future research.
In conclusion, our research identifies measurable associations be-
tween fried food consumption frequency and specific gut microbiota 
profiles that link to higher adiposity, central fat distribution, and a 
higher likelihood of cardiometabolic diseases. These observations call 
for further investigation into microbial mediators of metabolic out-
comes and emphasize the potential value of monitoring dietary pat-
terns in population health strategies.
Acknowledgments
We thank the Chronic Disease Research Institute of Zhejiang 
University for their contribution in both the Wellness Living Labo-
ratory China and Lanxi cohorts.
Author contributions
The authors’ responsibilities were as follows – YD, YL, WH, SZ: 
designed the study; YD, YL: performed the data analyses and prepared
the figures; YD, YL, WH, SZ: wrote the paper; CX, WW, XW, WZ, 
JCH, JW, AM, AWH: provided constructive analytical suggestions; 
WH, SZ: was primarily responsible for the final content; and all au-
thors: read and approved the final manuscript.
Conflict of interest
WH reports financial support was provided by the National Key 
R&D Program of China. WH reports financial support was provided 
by Pioneer and Leading Goose R&D Program of Zhejiang. SZ reports 
financial support was provided by the National Key R&D Program of 
China. SZ reports financial support was provided by grants from the 
Nutrilite Health Institute Wellness Fund. SZ reports financial support 
was provided by the Cyrus Tang Foundation. SZ reports financial 
support was provided by the China Medical Board (CMB) and the 
Hsun K. Chou Fund of Zhejiang University Education Foundation. All 
other authors report no conflicts of interest.
Funding
WH is supported by the National Key R&D Program of China 
(grant number 2022YFC2705303) and “Pioneer” and “Leading 
Goose” R&D Program of Zhejiang (grant number 2024C03180). SZ is 
supported by the National Key R&D Program of China (grant number 
2022YFC2705300) and grants from the Nutrilite Health Institute 
Wellness Fund, the Cyrus Tang Foundation, the China Medical Board 
(CMB), and the Hsun K Chou Fund of Zhejiang University Education 
Foundation.
Data availability
Data described in the manuscript, code book, and analytic code will 
be made available upon request pending application and approval by 
the corresponding authors.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi. 
org/10.1016/j.ajcnut.2025.06.025.
References
[1] Y. Sun, B. Liu, L.G. Snetselaar, J.G. Robinson, R.B. Wallace, L.L. Peterson, et 
al., Association of fried food consumption with all cause, cardiovascular, and 
cancer mortality: prospective cohort study, BMJ 364 (2019) k5420, https://doi. 
org/10.1136/bmj.k5420.
[2] P. Guallar-Castillon, F. Rodríguez-Artalejo, N.S. Fornes, J.R. Banegas, P.
A. Etxezarreta, E. Ardanaz, et al., Intake of fried foods is associated with 
obesity in the cohort of Spanish adults from the European Prospective 
Investigation into Cancer and Nutrition, Am. J. Clin. Nutr. 86 (1) (2007) 
198–205, https://doi.org/10.1093/ajcn/86.1.198.
[3] V.S. Beretta, W.R. Tebar, C.C. Scarabottolo, L.D. Delfino, B.T. Saraiva, A.
B. Santos, et al., Association between diet and adiposity in adults: influence of 
sedentary behavior patterns, Healthcare (Basel) 11 (8) (2023) 1157, https://doi. 
org/10.3390/healthcare11081157.
[4] P.L. Lutsey, L.M. Steffen, J. Stevens, Dietary intake and the development of 
the metabolic syndrome: the Atherosclerosis Risk in Communities study, 
Circulation 117 (6) (2008) 754–761, https://doi.org/10.1161/ 
CIRCULATIONAHA.107.716159.
[5] L. Fan, Y. Xia, Y. Wang, D. Han, Y. Liu, J. Li, et al., Gut microbiota bridges 
dietary nutrients and host immunity, Sci. China Life Sci. 66 (11) (2023) 
2466–2514, https://doi.org/10.1007/s11427-023-2346-1.
[6] A. Paoli, L. Mancin, A. Bianco, E. Thomas, J.F. Mota, F. Piccini, Ketogenic 
diet and microbiota: friends or enemies? Genes 10 (7) (2019) 534, https://doi. 
org/10.3390/genes10070534.
[7] R.D. Hills Jr., B.A. Pontefract, H.R. Mishcon, C.A. Black, S.C. Sutton, C. 
R. Theberge, Gut microbiome: profound implications for diet and disease, 
Nutrients 11 (7) (2019) 1613, https://doi.org/10.3390/nu11071613.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
742
[8] J. Zhu, J. Lyu, R. Zhao, G. Liu, S. Wang, Gut macrobiotic and its metabolic 
pathways modulate cardiovascular disease, Front Microbiol 14 (2023) 
1272479, https://doi.org/10.3389/fmicb.2023.1272479.
[9] J. Gao, X. Guo, W. Wei, R. Li, K. Hu, X. Liu, et al., The association of fried 
meat consumption with the gut microbiota and fecal metabolites and its impact 
on glucose homoeostasis, intestinal endotoxin levels, and systemic 
inflammation: A randomized controlled-feeding trial, Diabetes Care 44 (9) 
(2021) 1970–1979, https://doi.org/10.2337/dc21-0099.
[10] V. Partula, S. Mondot, M.J. Torres, E. Kesse-Guyot, M. Deschasaux,
K. Assmann, et al., Associations between usual diet and gut microbiota 
composition: results from the milieu Interieur cross-sectional study, Am. J 
Clin. Nutr. 109 (5) (2019) 1472–1483, https://doi.org/10.1093/ajcn/nqz029.
[11] Z. Zhou, Y. Wang, Y. Jiang, Y. Diao, P. Strappe, P. Prenzler, et al., Deep-fried 
oil consumption in rats impairs glycerolipid metabolism, gut histology and 
microbiota structure, Lipids Health Dis 15 (2016) 86, https://doi.org/10.1186/ 
s12944-016-0252-1.
[12] K.M. Kemp, C.A. Orihuela, C.D. Morrow, S.E. Judd, R.R. Evans, S. Mrug, 
Associations between dietary habits, socio-demographics and gut microbial 
composition in adolescents, Br. J Nutr 131 (5) (2024) 809–820, https://doi.org/ 
10.1017/S0007114523002271.
[13] Y. Min, X. Zhao, R.S. Stafford, X. Ma, S.H. Chen, D. Gan, et al., Cohort 
profile: WELL Living Laboratory in China (WELL-China), Int. J Epidemiol 
50 (5) (2021) 1432–1443, https://doi.org/10.1093/ije/dyaa283.
[14] C. Wei, S. Ye, Y. Ru, D. Gan, W. Zheng, C. Huang, et al., Cohort profile: the 
Lanxi Cohort study on obesity and obesity-related non-communicable diseases 
in China, BMJ Open 9 (5) (2019) e025257, https://doi.org/10.1136/bmjopen-
2018-025257.
[15] M. Valles-Colomer, A. Blanco-Míguez, P. Manghi, F. Asnicar, L. Dubois,
D. Golzato, et al., The person-to-person transmission landscape of the gut and 
oral microbiomes, Nature 614 (7946) (2023) 125–135, https://doi.org/10.1038/ 
s41586-022-05620-1.
[16] Y. Ru, N. Wang, Y. Min, X. Wang, V. McGurie, M. Duan, et al., 
Characterization of dietary patterns and assessment of their relationships with 
metabolomic profiles: A community-based study, Clin. Nutr. 40 (5) (2021) 
3531–3541, https://doi.org/10.1016/j.clnu.2020.12.006.
[17] J. Gao, Association of dietary patterns and physical activities with total body fat 
proportions and metabolic syndrome among middle-aged and elderly people: a 
cross-sectional study [dissertation], Fudan University, Shanghai, China, 2012.
[18] L.M. Li, K.Q. Rao, L.Z. Kong, C.H. Yao, H.D. Xiang, F.Y. Zhai, et al.,
A description on the Chinese national nutrition and health survey in 2002, 
Zhonghua Liu Xing Bing Xue Za Zhi 26 (7) (2005) 478–484.
[19] M. Lin, N. Shivappa, J.R. Hebert, H. Huang, L. Cai, J. Liang, et al., Dietary 
inflammatory index and cardiorenal function in women with diabetes and 
prediabetes, Nutr. Metab. Cardiovasc. Dis. 31 (8) (2021) 2319–2327, https:// 
doi.org/10.1016/j.numecd.2021.05.011.
[20] Y. Shi, J. Kan, W. Wang, Y. Cao, Y. Wu, X. Chen, et al., Nut consumption, 
gut microbiota, and body fat distribution: results of a large, community-based 
population study, Obesity (Silver Spring) 32 (9) (2024) 1778–1788, https://doi. 
org/10.1002/oby.24099.
[21] Y. Kang, J. Kim, Association between fried food consumption and 
hypertension in Korean adults, Br. J Nutr 115 (1) (2016) 87–94, https://doi.org/ 
10.1017/S000711451500402X.
[22] Y.X. Liu, L. Chen, T. Ma, X. Li, M. Zheng, X. Zhou, et al., EasyAmplicon: an 
easy-to-use, open-source, reproducible, and community-based pipeline for 
amplicon data analysis in microbiome research, Imeta 2 (1) (2023) e83, https:// 
doi.org/10.1002/imt2.83.
[23] P. Yilmaz, L.W. Parfrey, P. Yarza, J. Gerken, E. Pruesse, C. Quast, et al., The 
SILVA and “All-species Living Tree Project (LTP)” taxonomic frameworks, 
Nucleic Acids Res 42 (2014) D643–D648, https://doi.org/10.1093/nar/ 
gkt1209.
[24] J. Mai, Q. Wu, H. Wu, C. Zeng, Y. Li, J. Shang, et al., Assessment of whole-
body and regional body fat using abdominal quantitative computed 
tomography in Chinese women and men, Lipids Health Dis 23 (1) (2024) 47, 
https://doi.org/10.1186/s12944-024-02034-y.
[25] K.G. Alberti, P. Zimmet, J. Shaw, IDF Epidemiology Task Force Consensus 
Group, The metabolic syndrome—a new worldwide definition, Lancet 366 
(9491) (2005) 1059–1062, https://doi.org/10.1016/S0140-6736(05)67402-8.
[26] Y. Yang, J. Liu, C. Sun, Y. Shi, J.C. Hsing, A. Kamya, et al., Nonalcoholic 
fatty liver disease (NAFLD) detection and deep learning in a Chinese 
community-based population, Eur. Radiol. 33 (8) (2023) 5894–5906, https:// 
doi.org/10.1007/s00330-023-09515-1.
[27] J.C. Bae, No More NAFLD: The Term Is Now MASLD, Endocrinol. Metab. 
(Seoul). 39 (1) (2024) 92–94, https://doi.org/10.3803/EnM.2024.103.
[28] P.E. Sigvardsen, A. Fuchs, J.T. Kühl, S. Afzal, L. Køber, B.G. Nordestgaard, 
et al., Left ventricular trabeculation and major adverse cardiovascular events: 
the Copenhagen General Population Study, Eur. Heart J Cardiovasc. Imaging. 
22 (1) (2021) 67–74, https://doi.org/10.1093/ehjci/jeaa110.
[29] C. Wei, S. Ye, J.R. Sheng, X. Ma, Y. Ru, L. Zhang, et al., Associations of 
nicotine dependence and fat distribution in Chinese male adults: a cross-
sectional study in Lanxi, China, BMJ Open 9 (3) (2019) e022465, https://doi. 
org/10.1136/bmjopen-2018-022465.
[30] C.L. Craig, A.L. Marshall, M. Sj€ostr€om, A.E. Bauman, M.L. Booth, B.
E. Ainsworth, et al., International physical activity questionnaire: 12-country 
reliability and validity, Med. Sci. Sports Exerc. 35 (8) (2003) 1381–1395, 
https://doi.org/10.1249/01.MSS.0000078924.61453.FB.
[31] P. Dixon, VEGAN, a package of R functions for community ecology, J. Veg. 
Sci. 14 (6) (2003) 927–930, https://doi.org/10.1111/j.1654-1103.2003.tb02228.x.
[32] H. Mallick, A. Rahnavard, L.J. McIver, S. Ma, Y. Zhang, L.H. Nguyen, et al., 
Multivariable association discovery in population-scale meta-omics studies, 
PLOS Comput. Biol. 17 (11) (2021) e1009442, https://doi.org/10.1371/journal. 
pcbi.1009442.
[33] A. Sj€olander, T. Frisell, S. € 
Oberg, Sibling comparison studies, Annu. Rev. Stat.
Its. Appl. 9 (1) (2022) 71–94, https://doi.org/10.1146/annurev-statistics-
040120-024521.
[34] P.B. Axelsson, T.D. Clausen, A.H. Petersen, I. Hageman, A. Pinborg, L.
V. Kessing, et al., Investigating the effects of cesarean delivery and antibiotic 
use in early childhood on risk of later attention deficit hyperactivity disorder, J 
Child Psychol, Psychiatry 60 (2) (2019) 151–159, https://doi.org/10.1111/ 
jcpp.12961.
[35] Q. Qi, A.Y. Chu, J.H. Kang, J. Huang, L.M. Rose, M.K. Jensen, et al., Fried 
food consumption, genetic risk, and body mass index: gene-diet interaction 
analysis in three US cohort studies, Br. Med. J (Clin. Res. Ed.). 348 (2014) 
g1610, https://doi.org/10.1136/bmj.g1610.
[36] C. Tilves, T. Tanaka, M.K. Differding, A.P. Spira, C.W. Chia, L. Ferrucci, et 
al., The gut microbiome and regional fat distribution: findings from the 
Baltimore Longitudinal Study of Aging, Obesity (Silver Spring) 31 (5) (2023) 
1425–1435, https://doi.org/10.1002/oby.23717.
[37] W. He, P. Sparen, F. Fang, V. Sengpiel, B. Strander, K. Czene, Pregnancy 
outcomes in women with a prior cervical intraepithelial neoplasia Grade 3 
diagnosis: A nationwide population-based cohort study with sibling 
comparison design, Ann. Intern. Med. 175 (2) (2022) 210–218, https://doi.org/ 
10.7326/M21-2793.
Y. Duan et al.
The American Journal of Clinical Nutrition 122 (2025) 733–743
743