Received: 29 August 2021 | Revised: 11 November 2021 | Accepted: 22 November 2021

DOI: 10.1111/cas.15215

ORIGINAL ARTICLE

Low-­carbohydrate diet and risk of cancer incidence: The Japan

Public Health Center-­based prospective study

Honglin Cai1 | Tomotaka Sobue1 | Tetsuhisa Kitamura1 | Junko Ishihara2 |

Akiko Nanri3,4 | Tetsuya Mizoue4 | Motoki Iwasaki5 | Taiki Yamaji5 |
Manami Inoue5 | Shoichiro Tsugane5,6 | Norie Sawada5

1
Department of Environmental Medicine
and Population Sciences, Graduate School Abstract
of Medicine, Osaka University, Suita,
Epidemiological evidence on the effects of a long-­term low-­carbohydrate diet (LCD)
Japan
2
Department of Food and Life Science,
on cancer incidence remains sparse. We investigate the association between LCD and
Azabu University, Sagamihara, Japan the risk of overall and specific cancer site incidence in a Japanese population-­based
3
Department of Food and Health prospective cohort study among 90 171 participants aged 45-­74. Cox proportional
Sciences, International College of Arts and
Sciences, Fukuoka Women’s University, hazards models were used to estimate hazard ratios (HRs) and 95% confidence inter-
Fukuoka, Japan vals (CIs). During a median 17.0 y of follow-­up, we identified 15 203 cancer cases. A
4
Department of Epidemiology and
higher overall LCD score was associated with increased overall cancer risk (HR = 1.08
Prevention, Center for Clinical Sciences,
National Center for Global Health and [CI: 1.02-­1.14], P-­trend = .012), while it was associated with decreased gastric cancer
Medicine, Tokyo, Japan
5
(GC) risk (0.81 [0.71-­0.93], P-­trend = .006). A higher animal-­based LCD score was as-
Epidemiology and Prevention Group,
Center for Public Health Sciences, sociated with higher risk of overall cancer (1.08 [1.02-­1.14], P-­trend = .003), colorectal
National Cancer Center, Tokyo, Japan cancer (CRC) (1.11 [0.98-­1.25], P-­trend = .018), rectal cancer (RC) (1.24 [1.00-­1.54], P-­
6
National Institute of Health and
trend = .025), lung cancer (LC) (1.16 [1.00-­1.34], P-­trend = .042), and lower risk of GC
Nutrition, National Institutes of
Biomedical Innovation, Health and (0.90 [0.79-­1.01], P-­trend = .033). Furthermore, we found that plant-­based LCD score
Nutrition, Tokyo, Japan
was related to lower GC incidence (0.87 [0.77-­0.99], P-­trend = .031). Additionally,
Correspondence adjusted for plant fat intake amplified the adverse associations (overall cancer: 1.08
Tomotaka Sobue, Department of
[1.02-­1.14] vs. 1.11 [1.05-­1.18]; CRC: 1.08 [0.95-­1.22] vs. 1.13 [0.99-­1.30]; LC: 1.14
Environmental Medicine and Population
Sciences, Graduate School of Medicine, [0.98-­1.33] vs. 1.19 [1.01-­1.41]). We conclude that LCD enriching with animal products
Osaka University, 2-­2 Yamadaoka, Suita,
was associated with increased overall cancer, CRC, and LC incidence. These adverse
Osaka, 565-­0 871 Japan.
Email: tsobue@envi.med.osaka-u.ac.jp associations could be attenuated by plant fat consumption. LCD reduces the risk of
developing GC. Long-­term adherence to LCD without paying attention to the balance
Funding information
National Cancer Center for Research and between animal and plant food source consumption might cause adverse overall can-
Development Fund; Ministry of Health,
cer incidence consequences.
Labor and Welfare of Japan

Abbreviations: BMI, body mass index; CIs, confidence intervals; CRC, colorectal cancer; ER−, estrogen receptor negative; FFQ, food frequency questionnaire; GC, gastric cancer; H.
pylori, Helicobacter pylori; HCAs, heterocyclic amines; HPFS, Health Professionals Follow-­up Study; HRs, hazard ratios; IGF-­1, insulin-­like growth factor-­1; JPHC, Japan Public Health
Center-­based Prospective Study; LC, lung cancer; LCD, low-­c arbohydrate diet; LCHP, low carbohydrate and high protein; NHS, Nurses’ Health Study; NOCs, N-­nitroso compounds;
PAHs, polycyclic aromatic hydrocarbons; PHC, public health center; RC, rectal cancer.

This is an open access article under the terms of the Creat​ive Commo​ns Attri​butio​n-­NonCo​mmerc​ial-­NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-­commercial and no modifications or adaptations are made.
© 2021 The Authors. Cancer Science published by John Wiley & Sons Australia, Ltd on behalf of Japanese Cancer Association.

744 | 
wileyonlinelibrary.com/journal/cas Cancer Science. 2022;113:744–755.
CAI et al. | 745

KEYWORDS
Asian population, cancer incidence, low-­carbohydrate diet, prospective cohort study, specific
cancer site incidence

1 | I NTRO D U C TI O N We then excluded 1074 participants who did not respond to the
food intake questions; 2514 participants who reported or were di-
Although a balanced diet has been recommended for health through agnosed with cancer before the 5 y follow-­up questionnaire survey;
various studies,1 diet low in carbohydrates and high in protein is still and 4744 participants with energy intake at the upper or lower 2.5%.
a popular option for weight loss and weight control. Such a LCD Finally, 90 171 participants were included in the present study.
emphasizes the reduction of carbohydrate intake while encourag-
ing increased intake of high-­protein animal products that therefore
contain high amounts of fat. When the intake of one macronutrient 2.2 | Food frequency questionnaire
is high, the others will become low. Carbohydrates, protein, and fat
are the 3 main macronutrients. Their effect on health should be eval- The FFQ included 138 food items, and 9 beverage items, and was
uated as a whole rather than only focus on a single macronutrient. used to assess the average dietary food and beverage intake.
Therefore, a simple LCD summary score approach based on the per- Participants were asked about the frequency and portion size for
2
centage of energy from carbohydrate, protein, and fat were raised. each item consumed over the previous year.7 The daily food con-
As is well known, cancer is a disease that develops with years of sumption (g/d) was calculated by multiplying the consumption fre-
potentially dangerous exposure to factors, including dietary habits. quency by the typical portion size. Food and nutrient intake was
Several previous studies have investigated the association between estimated using the Standard Table of Food consumption in Japan
3-­5
a LCD and cancer morbidity or mortality. The NHS in the USA (7th revised and enlarged edition).8
suggested that LCD with high plant protein and fat was associated The validity of the FFQ was assessed using either 14-­d or 28-­d
with a decreased incidence of ER− breast cancer in postmenopausal dietary records. Spearman correlation coefficients between energy-­
women.3 Moreover, cohort studies in the USA demonstrated that adjusted intake for carbohydrate, fat, and protein derived from the
a higher overall LCD score and a higher animal-­based LCD score FFQ, and those derived from dietary records were 0.66-­0.69, 0.55-­
are related to higher cancer mortality.4 In contrast, the JPHC study 0.57, and 0.30-­0.31, respectively, in men and 0.45-­0.47, 0.39-­0.46,
showed no association between LCD and cancer mortality.5 To date, and 0.24-­0.33, respectively, in women.9 The reproducibility of es-
the long-­term safety of LCD remains controversial, and the evidence timations for intake of carbohydrate, fat, and protein between the
on how LCD affects cancer incidence remains sparse. 2 FFQs administered 1 y apart was 0.45-­0.55, 0.47-­0.57, and 0.47-­
Therefore, in this large Japanese population-­based cohort study, 0.57, respectively, in men, and 0.41-­0.50, 0.38-­0.52, and 0.32-­0.54,
we used the LCD score to evaluate the association between LCD and respectively, in women.10,11 Furthermore, we estimated protein and
the risk of overall and specific cancer site incidence. fat intakes from animal and plant sources separately. Animal food
included fish and shellfish, meat and processed meat, egg, milk and
dairy products, and butter, and plant food included foods other than
2 | M ATE R I A L S A N D M E TH O DS animal food. When we assessed the validity and reproducibility of
animal or plant protein and fat derived from FFQ, the Spearman
2.1 | Study population correlation coefficients between % energy of animal protein, ani-
mal fat, plant protein, and plant fat derived from the FFQ, and those
The JPHC study was initiated in 1990 for cohort I and in 1993 for derived from the dietary records were 0.21, 0.42, 0.59, and 0.39,
cohort II, at 11 PHC areas.6 In the baseline study, 140 420 partici- respectively, in men and 0.26, 0.42, 0.49, and 0.22, respectively, in
pants were informed of the objectives of the study, and the comple- women. The corresponding values between the 2 FFQs were 0.49,
tion of the survey questionnaire was regarded as providing consent 0.53, 0.60, and 0.64, respectively, in men and 0.48, 0.53, 0.58, and
to participate. A self-­administered questionnaire was administered 0.54, respectively, in women.12
at the baseline, 5-­y, and 10-­y follow-­ups. In this study, we took the
5-­y follow-­up survey as the starting point because it includes more
comprehensive information on food intake. 2.3 | Assessment of LCD score
Initially, the participants from the Tokyo area were not included
because information on cancer incidence was unavailable (n = 7097). The method used to assess LCD score has been described else-
After excluding ineligible participants (non-­Japanese nationality, late where.12 Briefly, according to the percentage of energy from car-
report of migration occurring before the start of the study, incorrect bohydrate, protein, or fat, participants were equally divided into
birth date, or lost to follow-­up), 130 777 participants remained. Of 11 categories. For carbohydrate, participants from the lowest
these, 98 503 participants returned the 5-­y questionnaire survey. to highest category scored 10-­0 points, while for protein and fat,
746 | CAI et al.

they scored 0-­10 points. The LCD score was calculated as the total (never, past, current with <20 cigarettes, 20-­4 0 cigarettes, ≥40 cig-
score of carbohydrate, protein, and fat, ranging from 0 to 30 points. arettes); alcohol consumption (none, occasional, regular of 1-­150,
A higher LCD score represented a lower carbohydrate intake with 150-­300, 300-­450, >450 g alcohol/wk); BMI (<23, 23-­25, 25-­27,
higher protein and fat intake. We then created separate scores for ≥27 kg/m2), history of diabetes mellitus (yes or no), total physical
animal protein, animal fat, plant protein, and plant fat. Similarly, the activity levels (Met-­h/d, quartiles), total energy intake (kcal/d, quin-
animal-­based LCD score was defined as the total score of carbohy- tiles), green tea consumption (never, <1 cup/d, 1 cup/d, 2-­3 cups/d,
drate, animal protein, and animal fat. The plant-­based LCD score was ≥4 cups/d), and coffee consumption (never, <1 cup/d, 1 cup/d,
the total score of carbohydrate, plant protein, and plant fat. 2-­3 cups/d, ≥4 cups/d). For breast cancer in women, Model 2 sim-
plified the categories for smoking status (never, past, current) and
alcohol consumption (none, occasional, regular of 1-­150, >150 g al-
2.4 | Follow-­up and case identification cohol/wk), and contained 2 other covariables: use of exogenous fe-
male hormones (yes or no) and menopausal status (premenopausal,
We followed the study participants from the date of the 5-­y follow- natural menopause, surgical menopause). Based on Model 2, Model
­up questionnaire survey until the date of moving out of the study 3 was further adjusted for sodium intake (quintiles) for GC. We
area, date of death, date of diagnosis with cancer, or the end of tested the interaction for each LCD score with sex before analyz-
follow-­up (December 31, 2012, for Osaka; December 31, 2013, for ing the association between LCD score and risk of overall cancer
Kochi and Nagasaki areas; December 31, 2015, for the other areas), and specific cancer site. To examine the effect of protein and fat
whichever occurred first. intakes on cancer risk, we further adjusted for animal protein, animal
The JPHC study incidence data were obtained from medical re- fat, plant protein, and plant fat (% energy, quintiles). The correla-
cords and cancer registries with permission from the respective local tion coefficients among these 4 macronutrients were tested before
governments of each study area. Death certificates were used as the adjustment. In sensitivity analyses, the above analyses were re-
supplementary sources. According to the Japan cancer statistics in peated after excluding cancer cases that were diagnosed in the first
2018, we selected the top 10 cancer sites (excluding malignant lym- 3 y. Additionally, 32 335 participants from cohort II provided blood
phoma) and 2 most common gender-­related cancer sites (prostate specimens at the date of baseline survey. Of them, 17 507 partici-
and breast) for specific cancer sites analyses. Cancer identification pants in our current study had undergone a H. pylori infection test
by site was assigned according to the International Classification and had atrophic gastritis status. We described the GC case distri-
of Diseases for Oncology, 3rd edition13 as follows: GC (C16), CRC bution for this subpopulation, and then conducted subgroup analy-
(C18-­C20), colon cancer (C18), RC (C19; C20), liver cancer (C22.0), ses for the relationship between LCD score and GC risk in H. pylori
pancreatic cancer (C25), LC (C34), esophageal cancer (C15), biliary antibody-­positive participants (N = 11 934) with further adjustment
tract cancer (C22.1; C23; C24), kidney cancer (C64), bladder cancer for H. pylori antibody concentration (tertiles) and atrophic gastritis
(C67), upper urinary tract cancer (C65; C66), prostate cancer (C61), status (none, moderate, and severe) based on Model 2.
and breast cancer (C50).

3 | R E S U LT S
2.5 | Statistical analysis
Of 90 171 participants, we ascertained 15 203 cancer cases during
Study participants were grouped into quintiles of overall LCD score, a median 17.0 y of follow-­up (1 418 371 person years). Participants
animal-­based LCD score, and plant-­based LCD score. Cox propor- in the highest quintile of any kind of LCD score tended to have a
tional hazards models were used to estimate HRs and 95% CIs to history of diabetes, higher total energy intake and consumed more
verify overall cancer and specific cancer site risk. The test for a linear coffee and green tea. Participants with higher overall LCD score or
trend was performed by entering the median value of each category animal-­based LCD score consumed more animal protein, animal fat,
into the model. All P-­values were two-­sided, and all statistical analy- and plant fat, but less plant protein. Participants with higher plant-­
ses were performed using SAS statistical software (version 9.4; SAS based LCD score had higher protein and fat consumption, but the
Institute Inc). We imputed missing data for covariates (BMI, smoking amounts and gradients were lower than those in the overall LCD
status, alcohol consumption, physical activity, coffee consumption, score and animal-­based LCD score (Table 1).
and green tea consumption, use of exogenous female hormones) Table 2 shows the association between LCD score and the risk
(women only), and menopausal status (women only) by including all of overall cancer and site-­specific cancer. Higher overall LCD score
covariates, follow-­up duration, and outcome in the model for multi- was associated with increased overall cancer risk (HR = 1.08 [CI:
ple imputations (SAS PROC MI).14 We performed 10 rounds of impu- 1.02-­1.14], P-­trend = .012), while it was associated with decreased
tation, then combined the estimates and P-­trend values according to GC risk (0.81 [0.71-­0.93], P-­trend = .006). A null association was
the Rubin rule (SAS PROC MIANALYZE).14,15 observed in other cancers. Furthermore, a higher animal-­based
We adjusted for age (continuous), sex, and area in Model 1. LCD score was associated with higher risk of overall cancer (1.08
Model 2 was further adjusted for the following: smoking status [1.02-­1.14], P-­trend = .003), CRC (1.11 [0.98-­1.25], P-­trend = .018),
TA B L E 1 Characteristics of participants according to quintiles of LCD score

Overall LCD scorea Animal-­based LCD scorea Plant-­based LCD scorea

CAI et al.

Q1 Q3 Q5 Q1 Q3 Q5 Q1 Q3 Q5

No. of subjects 17 410 17 685 17 495 19 030 16 663 19 125 18 531 16 304 18 000
Median score (range) 4 (2-­5) 15 (14-­16) 26 (24-­28) 3 (1-­5) 15 (14-­16) 26 (25-­28) 8 (6-­9) 15 (14-­16) 22 (21-­24)
Age (y), mean ± SD 58.1 ± 8.2 57.0 ± 7.8 57.6 ± 7.8 58.1 ± 8.1 57.1 ± 7.8 57.3 ± 7.8 57.9 ± 8.2 57.1 ± 7.8 57.6 ± 7.7
Sex (men, %) 44.7 50.4 46.3 46.0 49.5 46.8 50.2 48.9 49.7
BMI (kg/m2), 23.4 ± 3.1 23.5 ± 3.0 23.6 ± 3.0 23.4 ± 3.1 23.5 ± 3.0 23.6 ± 3.1 23.4 ± 3.1 23.4 ± 3 23.6 ± 3.0
mean ± SD
Current smoker (%) 23.3 24.9 21.0 22.9 24.8 22.7 27.3 24.0 21.5
Alcohol consumption 32.4 41.4 35.6 30.5 41.5 39.0 40.8 39.0 36.5
(≥1 time/wk, %)
Physical activity 32.4 ± 6.4 32.6 ± 6.3 32.2 ± 6.2 32.4 ± 6.3 32.5 ± 6.3 32.2 ± 6.2 32.2 ± 6.4 32.5 ± 6.3 32.4 ± 6.2
(MET-­h/d)
History of diabetes 3.6 4.9 5.9 3.9 5.1 5.5 4.1 4.6 6.4
(yes, %)
Coffee consumption 27.4 34.9 34.0 27.1 35.0 35.0 30.0 33.4 34.1
(≥1 cup/d, %)
Green tea consumption (≥1 cup/d, %)
Sencya 51.0 59.0 60.0 54.3 58.6 56.6 46.0 58.9 64.6
Bancya 43.2 42.4 44.5 43.0 43.1 43.9 42.8 43.0 44.4
Dietary intake
Total energy intake 1705.8 ± 548.5 1997.7 ± 568.1 2292.3 ± 681.8 1745.3 ± 565.3 1989.3 ± 573.4 2280.3 ± 675.9 1809.5 ± 587.6 2030.0 ± 607.3 2135.5 ± 644.2
(kcal/d)
Carbohydrate (% 66.0 ± 6.0 54.5 ± 5.9 44.1 ± 5.9 65.6 ± 5.7 54.3 ± 5.9 43.9 ± 6.2 59.8 ± 9.6 54.2 ± 9.0 50.2 ± 6.9
energy/d)
Protein (% energy/d)
Animal protein 4.6 ± 1.5 7.5 ± 1.5 11.3 ± 2.6 4.5 ± 1.4 7.5 ± 1.3 11.3 ± 2.4 6.6 ± 2.7 7.9 ± 2.8 8.2 ± 2.6
Plant protein 7.2 ± 1.1 6.7 ± 1.3 6.1 ± 1.7 7.7 ± 1.3 6.7 ± 1.2 5.6 ± 1.2 6.0 ± 1.0 6.6 ± 1.3 7.8 ± 1.5
Fat (% energy/d)
Animal fat 8.0 ± 3.0 13.6 ± 3.2 21.3 ± 5.7 7.6 ± 2.5 13.5 ± 2.5 22.1 ± 5.2 12.2 ± 5.6 14.6 ± 6.1 14.6 ± 5.2
Plant fat 8.9 ± 2.8 11.3 ± 3.2 13.0 ± 3.7 10.1 ± 3.6 11.3 ± 3.5 11.8 ± 3.3 7.3 ± 1.9 11.0 ± 1.8 15.3 ± 3.2
Red meat and 27.2 ± 19.5 46.9 ± 27.2 75.3 ± 47 25.3 ± 17.7 46.1 ± 25.1 79.2 ± 47.8 38.6 ± 28.3 52.5 ± 37.7 51.7 ± 35.9
processed meat
(g/d) b
Vegetables (g/d) b 192.7 ± 134.3 223.9 ± 134.7 230.6 ± 132.4 220.3 ± 159.7 221.8 ± 128.0 207.8 ± 116.0 145.7 ± 87.7 215.8 ± 110.9 296.7 ± 169.1
|
747
748 | CAI et al.

RC (1.24 [1.00-­1.54], P-­trend = .025), and LC (1.16 [1.00-­1.34], P-­

211.2 ± 140.1
trend = .042), and lower risk of GC (0.90 [0.79-­1.01], P-­trend = .033).

14.3 ± 4.2
Another, we found that plant-­based LCD score were related to lower
GC incidence (0.87 [0.77-­0.99], P-­trend = .031). No interactions for

Number of missing were 2234 for BMI, 5284 for smoking status, 2042 for alcohol consumption, 3016 for physical activity, 4802 for coffee consumption, 3568 for sencya and 3916 for bancya
73.7
3.1
Q5 LCD score with sex were observed (P-­interaction > .1). The results
215.2 ± 158.9 remained unchanged after further adjustment for sodium intake for
12.5 ± 12.3 GC or after excluding 1654 cancer cases diagnosed in the first 3 y in
sensitivity analyses.
Table 3 shows the association between the overall LCD score

72.0
Plant-­based LCD scorea

2.4
Q3

and the risk of overall cancer, GC, CRC, and LC after additional ad-

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by square of height in meters); LCD, low-­c arbohydrate diet; MET, metabolic equivalents.
justment for animal protein, animal fat, plant protein, or plant fat
based on Model 2. Pearson correlation coefficients among these
216.4 ± 211.1

4 macronutrients were 0.75 for animal protein and animal fat, and
9.7 ± 3.1

<0.5 for any other 2. Adjustment for animal protein intake atten-
71.6
2.5
Q1

uated the adverse associations (overall cancer: 1.08 [1.02-­1.14],

P-­trend = .012 vs. 1.03 [0.95-­1.13], P-­trend = .604; CRC: 1.08 [0.95-­
1.22], P-­trend = .176 vs. 1.02 [0.83-­1.25], P-­trend = .798; LC: 1.14
167.0 ± 119.2
12.8 ± 14.9

[0.98-­1.33], P-­trend = .170 vs. 1.00 [0.78-­1.29], P-­trend = .850). In

contrast, adjusted for plant fat intake amplified the adverse associa-
70.4
2.6
Q5

tions (overall cancer: 1.08 [1.02-­1.14], P-­trend = .012 vs. 1.11 [1.05-­
1.18], P-­trend = .001; CRC: 1.08 [0.95-­1.22], P-­trend = .176 vs. 1.13
[0.99-­1.30], P-­trend = .040; LC: 1.14 [0.98-­1.33], P-­trend = .170 vs.
214.4 ± 154.5

1.19 [1.01-­1.41], P-­trend = .055).

12.1 ± 3.7

Among the subgroup, among the 17 507 participants who had

consumption, 2752 women were missing for use of exogenous female hormones, and 2963 for menopausal status.
Animal-­based LCD scorea

72.1

H. pylori infection test and atrophic gastritis status, approximately

2.7
Q3

two-­thirds (n = 11 934) were H. pylori antibody-­positive. In total, 391

cases (92.2%) were H. pylori antibody-­positive among the 424 cases
262.1 ± 224.4

of GC. Although the number of cases was limited, a substantially de-

11.8 ± 4.3

creased GC risk was linked to a higher overall LCD score and animal-­
73.3

based LCD score (0.67 [0.47-­0.95], P-­trend = .029; 0.68 [0.49-­0.96],

2.4
Q1

P-­trend = .021, respectively) in the H. pylori antibody-­positive sub-

population (Table S1).
176.1 ± 120.7
13.4 ± 10.6

4 | DISCUSSION
71.7
2.6
Q5

In this population-­based cohort study, the overall LCD score was

216.0 ± 153.5

associated with increased overall cancer risk and reduced GC risk.

12.3 ± 3.7

When considering the LCD score based on animal or plant sources

Adjusted for total energy intake using residual method.

of protein and fat, we found the animal-­based LCD score was cor-
71.8
2.5
Q3

related with increased overall cancer risk, marginally significant in-

Overall LCD scorea

crease in CRC, RC, and LC risk, and a marginally significant decrease

256.9 ± 231.0

in GC risk. Furthermore, a higher plant-­based LCD score was associ-

11.0 ± 4.0

ated with a decreased incidence of GC.

To the best of our knowledge, only a few studies have investi-
72.4
2.4
Q1

gated the association between LCD and cancer incidence. Our study
TA B L E 1 (Continued)

is the first prospective study to evaluate the association between

hormones (yes,
Use of exogenous

Postmenopausal

LCD and subsequent cancer incidence in Asia. To date, there have

b
Sodium (g/d)
Fruits (g/d) b

(yes, %)c

been only 3 prospective studies that have assessed the association

female

between LCD and cancer incidence.3,16,17 The Nurse Health Study

Women only.
%)c

observed that a diet moderate in carbohydrate and high in plant pro-

tein and fat was related to a decreased ER− breast cancer incidence.3
However, the other 2 studies from Sweden only considered LCHP
b
a

c
TA B L E 2 Hazard ratio (95% confident interval) of incidence of overall cancer and site-­specific cancer according to quintiles of LCD score

Overall LCD score Animal-­based LCD score Plant-­based LCD score

CAI et al.

Cancer type Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda

No of subjects 17 410 17 685 17 495 19 030 16 663 19 125 18 531 16 304 18 000
Median score 4 (2-­5) 15 (14-­16) 26 (24-­28) 3 (1-­5) 15 (14-­16) 26 (25-­28) 8 (6-­9) 15 (14-­16) 22 (21-­24)
(range)
Person years 271 389 277 554 276 294 276 205 261 155.02 301 490.86 281 982 258 195.43 285 758.38
Overall cancer, 2891 2997 3051 3195 2826 3322 3177 2766 3100
cases
Model 1 1.00 1.04 1.07 .080 1.00 1.05 (1.00-­1.11) 1.09 (1.04-­1.14) .001 1.00 0.98 (0.93-­1.03) 0.96 (0.91-­1.01) .045
(0.98-­1.09) (1.01-­1.12)
Model 2 1.00 1.02 1.08 .012 1.00 1.03 (0.98-­1.08) 1.08 (1.02-­1.14) .003 1.00 0.99 (0.94-­1.05) 0.99 (0.94-­1.05) .734
(0.97-­1.08) (1.02-­1.14)
3 y exclusion, cases 2540 2709 2701 2815 2538 2946 2816 2471 2755
Model 2 1.00 1.05 1.09 .009 1.00 1.04 (0.99-­1.10) 1.09 (1.03-­1.15) .002 1.00 0.99 (0.94-­1.05) 0.99 (0.93-­1.05) .544
(0.99-­1.11) (1.03-­1.15)
Gastric cancer, 569 500 448 621 457 516 554 434 476
cases
Model 1 1.00 0.88 0.83 .006 1.00 0.89 (0.79-­1.00) 0.93 (0.83-­1.05) .095 1.00 0.88 (0.77-­1.00) 0.86 (0.76-­0.97) .007
(0.78-­0.99) (0.73-­0.94)
Model 2 1.00 0.86 0.81 .006 1.00 0.86 (0.76-­0.97) 0.90 (0.79-­1.01) .033 1.00 0.88 (0.77-­1.00) 0.87 (0.77-­0.99) .031
(0.76-­0.97) (0.71-­0.93)
Model 3 1.00 0.84 0.79 .002 1.00 0.85 (0.75-­0.96) 0.89 (0.78-­1.00) .023 1.00 0.85 (0.74-­0.97) 0.82 (0.71-­0.95) .004
(0.74-­0.95) (0.69-­0.91)
Colorectal cancer, 550 549 551 604 518 627 583 545 574
cases
Model 1 1.00 1.00 1.02 .773 1.00 1.03 (0.92-­1.16) 1.10 (0.99-­1.24) .022 1.00 1.03 (0.92-­1.16) 0.95 (0.85-­1.07) .042
(0.89-­1.13) (0.91-­1.15)
Model 2 1.00 1.00 1.08 .176 1.00 0.99 (0.88-­1.12) 1.11 (0.98-­1.25) .018 1.00 1.08 (0.96-­1.21) 1.03 (0.91-­1.17) .701
(0.88-­1.13) (0.95-­1.22)
Colon cancer, cases 393 380 385 435 358 428 411 392 384
Model 1 1.00 0.98 1.01 .610 1.00 1.00 (0.87-­1.15) 1.06 (0.92-­1.21) .185 1.00 1.04 (0.91-­1.20) 0.89 (0.77-­1.03) .015
(0.85-­1.13) (0.87-­1.16)
Model 2 1.00 0.97 1.04 .815 1.00 0.96 (0.83-­1.11) 1.06 (0.92-­1.22) .170 1.00 1.08 (0.94-­1.24) 0.95 (0.82-­1.10) .184
(0.84-­1.12) (0.90-­1.21)
Rectal cancer, cases 157 169 166 169 160 199 172 153 190
Model 1 1.00 1.06 1.07 .188 1.00 1.11 (0.90-­1.38) 1.23 (1.00-­1.51) .030 1.00 1.00 (0.80-­1.25) 1.10 (0.89-­1.35) .992
|

(0.85-­1.32) (0.86-­1.33)
749
TA B L E 2 (Continued)
750
|

Overall LCD score Animal-­based LCD score Plant-­based LCD score

Cancer type Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda

Model 2 1.00 1.07 1.15 .034 1.00 1.08 (0.86-­1.34) 1.24 (1.00-­1.54) .025 1.00 1.08 (0.86-­1.34) 1.25 (1.00-­1.56) .185
(0.85-­1.33) (0.91-­1.46)
Liver cancer, cases 103 140 119 108 142 133 159 132 128
Model 1 1.00 1.41 1.25 .271 1.00 1.61 (1.25-­2.07) 1.36 (1.05-­1.76) .097 1.00 1.02 (0.81-­1.29) 0.90 (0.71-­1.14) .381
(1.09-­1.82) (0.95-­1.63)
Model 2 1.00 1.45 1.24 .264 1.00 1.64 (1.27-­2.12) 1.34 (1.03-­1.75) .134 1.00 1.06 (0.83-­1.34) 0.93 (0.72-­1.19) .652
(1.11-­1.88) (0.94-­1.64)
Pancreatic cancer, 116 111 109 125 110 115 105 93 127
cases
Model 1 1.00 0.97 0.94 .571 1.00 1.06 (0.82-­1.37) 0.96 (0.74-­1.24) .544 1.00 1.02 (0.77-­1.34) 1.21 (0.93-­1.58) .327
(0.75-­1.27) (0.73-­1.23)
Model 2 1.00 0.96 0.93 .544 1.00 1.03 (0.79-­1.34) 0.92 (0.70-­1.21) .389 1.00 1.05 (0.79-­1.39) 1.28 (0.98-­1.69) .161
(0.73-­1.25) (0.70-­1.23)
Lung cancer, cases 368 390 404 400 373 434 415 351 373
Model 1 1.00 1.05 1.09 .466 1.00 1.11 (0.97-­1.28) 1.13 (0.98-­1.29) .083 1.00 0.97 (0.84-­1.11) 0.88 (0.76-­1.01) .081
(0.91-­1.22) (0.94-­1.26)
Model 2 1.00 1.05 1.14 .170 1.00 1.11 (0.96-­1.28) 1.16 (1.00-­1.34) .042 1.00 0.99 (0.85-­1.14) 0.92 (0.80-­1.07) .379
(0.91-­1.22) (0.98-­1.33)
Esophageal cancer, 76 79 82 77 82 95 110 88 86
cases
Model 1 1.00 0.96 1.07 .195 1.00 1.21 (0.89-­1.66) 1.29 (0.95-­1.75) .983 1.00 0.86 (0.65-­1.15) 0.74 (0.56-­0.99) .008
(0.70-­1.32) (0.78-­1.46)
Model 2 1.00 1.00 1.39 .352 1.00 1.13 (0.82-­1.55) 1.39 (1.01-­1.90) .492 1.00 1.03 (0.78-­1.38) 1.07 (0.79-­1.44) .883
(0.73-­1.39) (1.00-­1.94)
Biliary tract cancer, 113 104 113 136 95 119 120 120 108
cases
Model 1 1.00 0.97 1.03 .948 1.00 0.88 (0.68-­1.15) 0.96 (0.75-­1.23) .984 1.00 1.11 (0.86-­1.43) 0.84 (0.65-­1.10) .415
(0.74-­1.26) (0.79-­1.34)
Model 2 1.00 0.93 0.97 .725 1.00 0.84 (0.64-­1.10) 0.89 (0.68-­1.16) .580 1.00 1.10 (0.85-­1.42) 0.82 (0.62-­1.08) .343
(0.71-­1.23) (0.73-­1.28)
Kidney cancer, 36 49 50 44 39 48 44 38 48
cases
Model 1 1.00 1.26 1.27 .710 1.00 1.00 (0.65-­1.54) 1.05 (0.69-­1.59) .843 1.00 0.90 (0.58-­1.39) 0.95 (0.63-­1.44) .975
(0.82-­1.94) (0.83-­1.96)
CAI et al.

(Continues)
 CAI et al.

TA B L E 2 (Continued)

Overall LCD score Animal-­based LCD score Plant-­based LCD score

Cancer type Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda Q1 Q3 Q5 P-­trenda

Model 2 1.00 1.24 1.23 .879 1.00 1.01 (0.65-­1.56) 1.07 (0.69-­1.64) .889 1.00 0.85 (0.55-­1.32) 0.85 (0.55-­1.31) .591
(0.80-­1.92) (0.78-­1.94)
Bladder cancer, 78 88 81 85 79 84 100 60 88
cases
Model 1 1.00 1.07 1.01 .842 1.00 1.09 (0.80-­1.48) 1.04 (0.77-­1.41) .888 1.00 0.65 (0.47-­0.90) 0.80 (0.60-­1.08) .162
(0.79-­1.45) (0.74-­1.39)
Model 2 1.00 1.03 1.00 .806 1.00 1.06 (0.77-­1.44) 1.03 (0.75-­1.41) .807 1.00 0.64 (0.46-­0.89) 0.78 (0.58-­1.07) .149
(0.76-­1.41) (0.72-­1.39)
Upper urinary tract 19 21 18 20 27 20 22 22 18
cancer, cases
Model 1 1.00 1.06 0.89 .608 1.00 1.54 (0.86-­2.76) 0.98 (0.52-­1.83) .701 1.00 1.08 (0.60-­1.96) 0.77 (0.41-­1.44) .218
(0.57-­1.98) (0.47-­1.72)
Model 2 1.00 0.99 0.88 .588 1.00 1.46 (0.81-­2.63) 0.97 (0.51-­1.86) .690 1.00 1.03 (0.56-­1.89) 0.72 (0.37-­1.39) .178
(0.52-­1.86) (0.44-­1.73)
Prostate cancer, 232 298 300 261 256 300 280 270 315
casesb
Model 1 1.00 1.14 1.18 .099 1.00 1.08 (0.91-­1.29) 1.12 (0.95-­1.32) .076 1.00 1.05 (0.89-­1.25) 1.04 (0.88-­1.23) .518
(0.96-­1.36) (0.99-­1.40)
Model 2 1.00 1.12 1.17 .164 1.00 1.07 (0.90-­1.28) 1.11 (0.93-­1.32) .111 1.00 1.04 (0.87-­1.23) 1.02 (0.86-­1.21) .763
(0.94-­1.34) (0.97-­1.40)
Breast cancer, 157 169 188 181 161 198 158 183 177
casesb
Model 1 1.00 1.09 1.14 .218 1.00 1.03 (0.83-­1.28) 1.04 (0.85-­1.27) .384 1.00 1.15 (0.93-­1.43) 1.01 (0.82-­1.26) .922
(0.88-­1.36) (0.92-­1.41)
Model 2 1.00 1.09 1.10 .353 1.00 1.02 (0.82-­1.26) 0.99 (0.80-­1.23) .658 1.00 1.14 (0.92-­1.42) 0.99 (0.79-­1.25) .980
(0.87-­1.36) (0.88-­1.38)

Abbreviations: LCD, low-­c arbohydrate diet.

Model 1 adjusted for age sex area.
Model 2 was further adjusted for smoking, drinking, BMI, total physical activity levels (MET-­h/d), history of diabetes, total energy intake, green tea consumption, and coffee consumption.
a
Linear trend across quintiles of LCD score was tested by entering the median values of each quintile into the Cox proportional hazards model.
b
Prostate cancer was conducted in men; breast cancer was conducted in women, and were further adjusted for menopausal status (yes, no, natural; no, artificial), use of exogenous hormone pills (yes or no).
c
Model 3 was further adjusted for sodium intake (quintile) for GC cancer based on Model 2.
|
751
752 | CAI et al.

TA B L E 3 Hazard ratio (95% confident interval) of overall cancer, GC, CRC, and LC when further adjustment for macronutrient according
to quintiles of overall LCD score

Overall LCD score

Cancer type Q1 (2-­5) Q2 (9-­11) Q3 (14-­16) Q4 (19-­22) Q5 (24-­28) P-­trenda

Overall cancer
Model 2 1.00 1.03 (0.97-­1.08) 1.02 (0.97-­1.08) 1.03 (0.97-­1.08) 1.08 (1.02-­1.14) .012
Adjusted for animal protein 1.00 1.02 (0.97-­1.08) 1.01 (0.94-­1.08) 1.00 (0.92-­1.08) 1.03 (0.95-­1.13) .604
Adjusted for animal fat 1.00 1.04 (0.98-­1.10) 1.03 (0.97-­1.10) 1.03 (0.96-­1.10) 1.07 (0.99-­1.16) .162
Adjusted for plant protein 1.00 1.02 (0.97-­1.08) 1.02 (0.97-­1.08) 1.02 (0.97-­1.08) 1.07 (1.01-­1.13) .058
Adjusted for plant fat 1.00 1.04 (0.98-­1.09) 1.04 (0.99-­1.10) 1.05 (0.99-­1.11) 1.11 (1.05-­1.18) .001
GC
Model 2 1.00 0.84 (0.75-­0.95) 0.86 (0.76-­0.97) 0.84 (0.74-­0.95) 0.81 (0.71-­0.93) .006
Adjusted for animal protein 1.00 0.82 (0.71-­0.94) 0.80 (0.68-­0.95) 0.78 (0.65-­0.95) 0.76 (0.61-­0.95) .034
Adjusted for animal fat 1.00 0.86 (0.76-­0.99) 0.89 (0.76-­1.03) 0.86 (0.72-­1.02) 0.80 (0.65-­0.97) .058
Adjusted for plant protein 1.00 0.84 (0.74-­0.95) 0.85 (0.75-­0.97) 0.84 (0.74-­0.95) 0.81 (0.70-­0.93) .007
Adjusted for plant fat 1.00 0.85 (0.76-­0.97) 0.88 (0.77-­1.00) 0.87 (0.76-­1.00) 0.85 (0.73-­0.98) .065
CRC
Model 2 1.00 1.00 (0.89-­1.13) 1.00 (0.88-­1.13) 1.06 (0.94-­1.20) 1.08 (0.95-­1.22) .176
Adjusted for animal protein 1.00 1.00 (0.88-­1.14) 0.99 (0.84-­1.16) 1.03 (0.87-­1.23) 1.02 (0.83-­1.25) .798
Adjusted for animal fat 1.00 1.02 (0.90-­1.16) 1.00 (0.86-­1.16) 1.04 (0.88-­1.22) 1.04 (0.86-­1.25) .716
Adjusted for plant protein 1.00 0.99 (0.88-­1.12) 0.98 (0.87-­1.11) 1.04 (0.92-­1.17) 1.04 (0.91-­1.18) .471
Adjusted for plant fat 1.00 1.02 (0.91-­1.15) 1.03 (0.91-­1.17) 1.11 (0.97-­1.26) 1.13 (0.99-­1.30) .040
LC
Model 2 1.00 0.99 (0.86-­1.15) 1.05 (0.91-­1.22) 0.97 (0.83-­1.12) 1.14 (0.98-­1.33) .170
Adjusted for animal protein 1.00 0.96 (0.81-­1.13) 0.97 (0.80-­1.18) 0.87 (0.70-­1.08) 1.00 (0.78-­1.29) .850
Adjusted for animal fat 1.00 0.94 (0.80-­1.10) 0.93 (0.78-­1.12) 0.82 (0.67-­1.00) 0.93 (0.74-­1.17) .386
Adjusted for plant protein 1.00 0.98 (0.85-­1.14) 1.03 (0.89-­1.20) 0.94 (0.80-­1.10) 1.08 (0.92-­1.27) .517
Adjusted for plant fat 1.00 1.01 (0.87-­1.17) 1.07 (0.92-­1.25) 1.00 (0.85-­1.17) 1.19 (1.01-­1.41) .055

Abbreviations: CRC, colorectal cancer; GC, gastric cancer; LC, lung cancer; LCD, low-­c arbohydrate diet.
a
Linear trend across quintiles of LCD score was tested by entering the median values of each quintile into the Cox proportional hazards model.

intakes. One reported null associations with overall cancer and site-­ risk. 22 The biomedical plausibility is considerable. Red meat and
16
specific cancer incidence ; the other suggested that an LCHP diet processed meat would produce and contain carcinogens such as
was linked to lower prostate cancer incidence.17 For mortality, a HCAs, PAHs, and NOCs during cooking or processing. These sub-
positive association has been found for animal-­based LCD score and stances might act as pro-­oxidants and, therefore, lead to carcino-
cancer mortality for pooling NHS and HPFS.4 In cohort studies of genesis. 23-­26 Vegetables, fruits, cereals, and legumes are the major
Swedish women or Japanese adults,5,18 neither showed a tendency sources of vitamins, dietary fibers, and carbohydrates. Vitamins
toward a linear association between LCD score and cancer mortality. have been proven to have anti-­oxidant and anti-­inflammatory
Taken together, the previous studies to date were not consistent in properties. 27 Similarly, dietary fiber has anti-­inflammatory prop-
terms of the long-­term effects of LCD on cancer risk. erties28; some types could attenuate postprandial rises in blood
In our study, a higher animal-­b ased LCD score was related to glucose and insulin by reducing the rate of glucose absorption. 29
higher overall cancer, CRC, RC, and LC risk. However, these asso- Therefore, an animal-­b ased LCD might restrict healthy food con-
ciations disappeared for the plant-­r ich LCD score. Consistent with sumption in the long run, causing the adverse effects of red meat
our findings, previous studies have noted that a higher incidence to some extent. In the colon and rectal cancer analysis, we found
of CRC is related to a westernized dietary pattern, which favors that the animal-­b ased LCD was strongly associated with increased
a higher intake of animal products.19,20 According to the World RC risk. This finding was in line with previous studies on the asso-
21
Cancer Research Fund's Cancer Report, there is convincing evi- ciation of red meat intake with CRC risk, which have also shown
dence that high red meat and processed meat consumption are as- that NOCs from red meat or processed meat are more carcino-
sociated with increased CRC risk. A previous study in JPHC found genic to the rectum than the colon. 30 Differences in rates of me-
an adverse association between red meat consumption and LC tabolism, fermentation, transit time, and expression of enzymes
CAI et al. | 753

and different morphology, are considered to be the reasons for the effect of LCD on the risk of GC is unknown. Compared with the as-
31
difference in the effect of a risk factor on the colon and rectum. sociations in the whole population, the protective effects of over-
Alternatively, it has been pointed out that an LCD with higher all and animal-­b ased LCD on GC were more pronounced in the H.
animal product consumption would increase the levels of cancer-­ pylori antibody-­p ositive population (Table S1). We speculated that
promoting metabolites. 32 A long-­term higher intake of animal pro- interactions between foods and H. pylori might exist. Previous
tein and fat is associated with increased insulin or IGF-­1 levels, studies have revealed that a diet pattern high in sweets and car-
which are important tumor promoters, resulting in accelerated bohydrates was positively associated with prevalence of H. pylori
tumor cell proliferation. 33,34 This hypothesis also supports our infection.43 The prevalence of H. pylori-­related gastric precan-
findings that adjustment for animal protein attenuated the adverse cerous lesions progressively increased with increased starchy
association between overall LCD and cancer risk. Conversely, al- vegetable intake and reduced fresh fruit intake.44 It is supposed
though the plant-­b ased LCD score was not associated with overall that a higher starchy food intake leads to an elevation in blood
cancer, CRC, or LC risk, the positive associations of overall LCD glucose level to reduce gastric acid secretion and subsequently
were aggravated when adjusting for plant fat intake. In addition, creates an environment favorable for the growth and prolifera-
the adverse associations of overall LCD for overall cancer and CRC tion of H. pylori and other microorganisms.45,46 Protein-­e nriched
risk were only observed in the low plant fat intake groups when foods are potent stimulants of gastric acid secretion. 38 Therefore,
stratifying plant fat intake (Table S2). Therefore, we supposed for the H. pylori antibody-­p ositive population, animal-­b ased LCD
that increased plant fat intake could offset the adverse effects of had a more notable protective effect on GC through regulating
consuming animal foods. A previous study has reported that plant the gastric acid secretion process to inhibit the growth and pro-
fat enriched with unsaturated fatty acids could improve insulin liferation of H. pylori. However, a similar protective association
sensitivity and, in turn, reduce circulating insulin and markers of for plant-­b ased LCD in the whole population was not observed
inflammation. 35 in the H. pylori antibody-­p ositive population. Considering that
The stomach is the main organ that digests proteins, therefore the H pylori infection status could not be adjusted in the whole
it has high acidity of gastric juice. Previous studies have noted that population analysis, residual confounding of H. pylori might exist,
gastric juice ascorbic acid has a role in preventing the formation of therefore the inverse association between plant-­b ased LCD and
NOCs, and, therefore, protects against GC.36 It has been noted that GC should be interpreted with caution. Further investigations be-
the effects of carbohydrate and protein on stimulating gastric juice tween LCD and GC risk in non-­H. pylori infection populations are
secretion are different; a low carbohydrate with moderate protein also warranted.
diet would prolong the gastric secretion duration, therefore, increas- Our study had several strengths. This is a large, population-­
ing the amount of gastric acid37,38; fresh fruits and vegetables are based, prospective study with a long follow-­up period. The pro-
sources of ascorbic acid, which are linked to a reduction in stom- spective design reduced recall bias and reverse causation. The
ach carcinogenesis.39 Our study showed that LCD score was associ- reliable FFQ and available data from the questionnaire enabled us
ated with reduced GC incidence. This finding is consistent with the to calculate LCD scores and carefully adjust for important poten-
JPHC study on dietary patterns, which suggested that the traditional tial factors. Some limitations of our study warrant mention. First,
Japanese dietary pattern with high rice consumption increased GC due to the low validity of carbohydrate, protein, and fat intake,
incidence.40 Previous studies in JPHC have suggested that a higher dietary information was assessed at a single time point, this caveat
salt content in food is positively associated with GC risk,41 especially might have led to misclassification of LCD score. However, such
when typically consuming rice with salted foods.40 However, in our misclassification tends to attenuate the association described in
study, the group with low-­carbohydrate intake (Q5) had a higher our study. Second, some participants in a subhealthy status might
sodium intake, and further adjustment for sodium intake did not have changed their dietary behavior when answering the ques-
change the results of the association between LCD score and GC tionnaire. This may have obscured the relationship between LCD
(Model 3). Our findings may support the mechanism that carbohy- score and cancer risk. However, there was no material change in
drate restriction with high-­protein intake could promote gastric acid the results when we excluded the first 3 y of cancer cases in the
secretion to prevent gastric carcinogenesis.37 As there was a lack sensitivity assessment. Third, as we could not adjust for some un-
of data on H. pylori infection status for each subject, residual con- measured covariables such as socioeconomic status and H. pylori
founding of H. pylori might exist for the association between LCD infection status for the whole population, potential residual con-
score and GC. founding might not have been ruled out completely.
H. pylori is an independent factor responsible for GC, and 65%–­ In conclusion, LCD enriched with animal products was associ-
80% of all GC cases were caused by H. pylori infection.42 In our sub- ated with increased overall cancer, CRC, and LC incidence, and these
population, 92.2% of GC cases were H. pylori positive. Therefore, adverse associations could be attenuated by plant fat consumption.
we could not assess the P-­value for interaction between LCD LCD reduces the risk of developing GC. Long-­term adherence to a
score and H. pylori infection because GC cases without H. pylori in- LCD without paying attention to the balance between animal and
fections were limited. Analysis for the H. pylori antibody-­n egative plant food source might cause adverse overall cancer incidence
population also failed to be conduct, which meant that the direct consequences. Because the evidence on the association between
754 | CAI et al.

diets, and the risk of postmenopausal breast cancer. Am J Epidemiol.

LCD score and risk of cancer incidence is limited, further studies are 2011;174:652-­660.
warranted. 4. Fung TT, van Dam RM, Hankinson SE, Stampfer M, Willett WC, Hu
FB. Low-­c arbohydrate diets and all-­c ause and cause-­specific mor-
tality: two cohort studies. Ann Intern Med. 2010;153:289-­298.
AC K N OW L E D G M E N T S
5. Akter S, Mizoue T, Nanri A, et al. Low carbohydrate diet and all
This study was supported by the National Cancer Center for
cause and cause-­specific mortality. Clin Nutr (Edinburgh, Scotland).
Research and Development Fund (since 2011) (23-­A-­31[toku], 26-­A-­ 2021;40:2016-­2024.
2, 29-­A-­4, and 2020-­J-­4), Grant-­in-­Aid for Cancer Research from the 6. Tsugane S, Sawada N. The JPHC study: design and some findings on
Ministry of Health, Labor and Welfare of Japan (from 1989 to 2010) the typical Japanese diet. Jpn J Clin Oncol. 2014;44:777-­782.
7. Sasaki S, Kobayashi M, Ishihara J, Tsugane S. Self-­administered food
(19shi-­2). We thank all members of the JPHC Prospective Study
frequency questionnaire used in the 5-­year follow-­up survey of the
Group for their valuable contributions. JPHC members (as of April JPHC Study: questionnaire structure, computation algorithms, and
2020) are listed at https://epi.ncc.go.jp/en/jphc/781/8510.html. area-­based mean intake. J Epidemiol. 2003;13:S13-­S22.
8. Science and Technology Agency. Standard Tables of Food
Composition in Japan. 7th revised and, enlarged ed. :Printing Bureau
C O N FL I C T O F I N T E R E S T
of the Ministry of Finance; 2015.
Authors declare no conflicts of interest for this article. 9. Ishihara J, Inoue M, Kobayashi M, et al. Impact of the revision of a
nutrient database on the validity of a self-­administered food fre-
AU T H O R C O N T R I B U T I O N S quency questionnaire (FFQ). J Epidemiol. 2006;16:107-­116.
10. Sasaki S, Ishihara J, Tsugane S. Reproducibility of a self-­administered
Cai: responsible for the data collection, statistical analysis, data in-
food frequency questionnaire used in the 5-­year follow-­up survey
terpretation, and manuscript drafting; Sobue, Kitamura, Ishihara, of the JPHC Study Cohort I to assess food and nutrient intake. J
Nanri, Mizoue, Iwasaki, Yamaji, Inoue, Tsugane, Sawada: reviewed Epidemiol. 2003;13:S115-­S124.
and edited the manuscript, data collection, and contributed to the 11. Ishihara J, Sobue T, Yamamoto S, et al. Validity and reproducibility
of a self-­administered food frequency questionnaire in the JPHC
discussion; Sawada: (principal investigator): obtained funding and
Study Cohort II: study design, participant profile and results in com-
designed, initiated, and organized the study, management of the parison with Cohort I. J Epidemiol. 2003;13:134-­147.
study. All authors had primary responsibility for final content. All au- 12. Nanri A, Mizoue T, Kurotani K, et al. Low-­c arbohydrate diet
thors read and approved the final manuscript. and type 2 diabetes risk in Japanese men and women: the
Japan Public Health Center-­B ased Prospective Study. PLoS One.
2015;10:e0118377.
E T H I C A L A P P R OVA L
13. World Health Organization. International Classification of Diseases
The Institutional Review Board of the National Cancer Center, Tokyo, for Oncology. 3rd ed. World Health Organization; 2000.
Japan approved the JPHC study. The present study was approved by 14. Berglund P, Heeringa S. Multiple Imputation of Missing Data Using
the Ethical Review Board of Osaka University, Osaka, Japan. SAS. Cary, NC, USA: SAS Institute Inc.; 2014.
15. Ratitch B, Lipkovich I, O’Kelly M. Combining analysis results from
multiply imputed categorical data. 2013; 1-­19.
DATA AVA I L A B I L I T Y S TAT E M E N T 16. Nilsson LM, Winkvist A, Johansson I, et al. Low-­c arbohydrate, high-­
For information on how to apply to gain access to JPHC data, protein diet score and risk of incident cancer; a prospective cohort
following the instructions at https://epi.ncc.go.jp/en/jphc/805/8155. study. Nutr J. 2013;12:58.
17. Ax E, Garmo H, Grundmark B, et al. Dietary patterns and prostate
html.
cancer risk: report from the population based ULSAM cohort study
of Swedish men. Nutr Cancer. 2014;66:77-­87.
ORCID 18. Lagiou P, Sandin S, Weiderpass E, et al. Low carbohydrate?high pro-
Honglin Cai https://orcid.org/0000-0002-7831-6764 tein diet and mortality in a cohort of Swedish women. J Intern Med.
2007;261:366-­374.
Tomotaka Sobue https://orcid.org/0000-0003-2817-3483
19. Fung T, Hu FB, Fuchs C, et al. Major dietary patterns and the risk of
Junko Ishihara https://orcid.org/0000-0002-9516-5646 colorectal cancer in women. Arch Intern Med. 2003;163:309-­314.
Tetsuya Mizoue https://orcid.org/0000-0002-1198-2228 20. Shin S, Saito E, Sawada N, et al. Dietary patterns and colorectal
Motoki Iwasaki https://orcid.org/0000-0003-3319-4131 cancer risk in middle-­aged adults: a large population-­based pro-
spective cohort study. Clin Nutr. 2018;37:1019-­1026.
Manami Inoue https://orcid.org/0000-0003-1276-2398
21. Stewart BW, Wild C. World Cancer Report 2014. Lyon, France;
Shoichiro Tsugane https://orcid.org/0000-0003-4105-2774 International Agency for Research on Cancer WHO Press; 2014; 630.
Norie Sawada https://orcid.org/0000-0002-9936-1476 22. Cai H, Sobue T, Kitamura T, et al. Association between meat and sat-
urated fatty acid intake and lung cancer risk: the Japan Public Health
Center-­based prospective study. Int J Cancer. 2020;147:3019-­3 028.
REFERENCES
23. Sinha R, Knize MG, Salmon CP, et al. Heterocyclic amine content of
1. Krauss RM, Deckelbaum RJ, Ernst N, et al. Dietary guidelines for
pork products cooked by different methods and to varying degrees
healthy American adults: a statement for health professionals from
of doneness. Food Chem Toxicol. 1998;36:289-­297.
the Nutrition Committee, American Heart Association. Circulation.
24. Felton JS, Malfatti MA, Knize MG, Salmon CP, Hopmans EC, Wu
1996;94:1795-­1800.
RW. Health risks of heterocyclic amines. Mutat Res-­Fund Mol M.
2. Halton TL, Willett WC, Liu S, et al. Low-­c arbohydrate-­diet score
1997;376:37-­41.
and the risk of coronary heart disease in women. New Engl J Med.
25. Abdel-­Shafy HI, Mansour MS. A review on polycyclic aromatic hy-
2006;355:1991-­2002.
drocarbons: source, environmental impact, effect on human health
3. Fung TT, Hu FB, Hankinson SE, Willett WC, Holmes MD. Low-­
and remediation. Egypt J Petrol. 2016;25:107-­123.
carbohydrate diets, dietary approaches to stop hypertension-­style
CAI et al. | 755

26. Mirvish SS. N-­nitroso compounds: their chemical and in vivo for- 40. Kim MK, Sasaki S, Sasazuki S, Tsugane S. Prospective study of
mation and possible importance as environmental carcinogens. J three major dietary patterns and risk of gastric cancer in Japan. Int
Toxicol Environ Health. 1977;2:1267-­1277. J Cancer. 2004;110:435-­4 42.
27. Mozafar A. Plant vitamins. Boca Raton, FL.: CRC Press; 2018. 41. Tsugane S, Sasazuki S, Kobayashi M, Sasaki S. Salt and salted food
28. Ma Y, Griffith JA, Chasan-­Taber L, et al. Association between intake and subsequent risk of gastric cancer among middle-­aged
dietary fiber and serum C-­reactive protein. Am J Clin Nutr. Japanese men and women. Br J Cancer. 2004;90:128-­134.
2006;83:760-­766. 42. De Martel C, Ferlay J, Franceschi S, et al. Global burden of cancers
29. Kay R. Dietary fiber. J Lipid Res. 1982;23:221-­242. attributable to infections in 2008: a review and synthetic analysis.
30. Le Marchand L, Donlon T, Seifried A, Wilkens LR, Biomarkers P. Red Lancet Oncol. 2012;13:607-­615.
meat intake, CYP2E1 genetic polymorphisms, and colorectal cancer 43. Xia Y, Meng G, Zhang Q, et al. Dietary patterns are associated with
risk. Cancer Epidemiol. 2002;11:1019-­1024. Helicobacter pylori infection in Chinese adults: a cross-­sectional
31. Iacopetta B. Are there two sides to colorectal cancer? Int J Cancer. study. Sci Rep. 2016;6:1-­8.
2002;101:403-­4 08. 44. Kato I, Vivas J, Plummer M, et al. Environmental factors in
32. Russell WR, Gratz SW, Duncan SH, et al. High-­protein, reduced-­ Helicobacter pylori-­related gastric precancerous lesions in
carbohydrate weight-­loss diets promote metabolite profiles likely to Venezuela. Cancer Epidemiol Biomarkers Prev. 2004;13:468-­476.
be detrimental to colonic health. Am J Clin Nutr. 2011;93:1062-­1072. 45. Stacher G, Bauer P, Schulze D, Pointner H, Landgraf M. Effect
33. Allen NE, Appleby PN, Davey GK, Kaaks R, Rinaldi S, Key TJ. The of alterations of blood glucose levels on gastric acid secre-
associations of diet with serum insulin-­like growth factor I and its tion, plasma gastrin, and plasma osmolality in man. Am J Dig Dis.
main binding proteins in 292 women meat-­eaters, vegetarians, and 1976;21:563-­568.
vegans. Cancer Epidemiol Biomarkers Prev. 2002;11:1441-­1448. 46. Axon A. Relationship between Helicobacter pylori gastritis, gastric
34. Holmes MD, Pollak MN, Willett WC, Hankinson SE. Dietary cor- cancer and gastric acid secretion. Adv Med Sci. 2007;52:55–­60.
relates of plasma insulin-­like growth factor I and insulin-­like
growth factor binding protein 3 concentrations. Cancer Epidemiol
Biomarkers Prev. 2002;11:852-­861. S U P P O R T I N G I N FO R M AT I O N
35. Martínez-­González MA, Salas-­Salvadó J, Estruch R, et al. Benefits Additional supporting information may be found in the online
of the Mediterranean diet: insights from the PREDIMED study. Prog
version of the article at the publisher’s website.
Cardiovasc Dis. 2015;58:50-­60.
36. Schorah CJ, Sobala GM, Sanderson M, Collis N, Primrose JN.
Gastric juice ascorbic acid: effects of disease and implications for
gastric carcinogenesis. Am J Clin Nutr. 1991;53:287S-­293S. How to cite this article: Cai H, Sobue T, Kitamura T, et al.
37. Lennard-­Jones J, Fletcher J, Shaw D. Effect of different foods on Low-­carbohydrate diet and risk of cancer incidence: The
the acidity of the gastric contents in patients with duodental ulcer. Japan Public Health Center-­based prospective study. Cancer
3. Effect of altering the proportions of protein and carbohydrate.
Sci. 2022;113:744–­755. doi:10.1111/cas.15215
Gut. 1968;9(2):177-­182.
38. Brooks FP. Effect of diet on gastric secretion. Am J Clin Nutr.
1985;42:1006-­1019.
39. Drake IM, Davies MJ, Mapstone NP, et al. Ascorbic acid may pro-
tect against human gastric cancer by scavenging mucosal oxygen
radicals. Carcinogenesis. 1996;17:559-­562.