Sugar intake and colorectal cancer risk: A prospective Japanese cohort study

Abstract The influence of sugar consumption on the risk of colorectal cancer (CRC) remains controversial. Prospective cohort studies focusing on total and specific types of sugar intake among the Asian population who have different patterns of sugar intake sources than American and European populations are scarce. We intended to examine the association of sugar intake with CRC risk among middle‐aged adults in a Japanese large‐scale population‐based cohort study. The participants (42,405 men and 48,600 women) who were 45–74 years old and answered the questionnaire in 1995–1999 in the Japan Public Health Center‐based Prospective Study were followed up until December 2013. Total sugars, total fructose, and specific types of sugar intake were estimated using a validated 147‐item food frequency questionnaire and divided into quintiles (Q1–Q5). We used Cox proportional hazard regression models adjusted for potential confounders to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). During the follow‐up, 2118 CRC cases (1226 men and 892 women) were identified. We did not observe any clear association between all types of sugar intake and an increased risk of CRC. Analyses by tumor sites yielded a positive association of total sugar consumption with rectal cancer in women (1.75 [1.07–2.87] for Q1 vs. Q5; p linear trend = 0.03), but no statistically significant trend was detected among men. Sugar intake was not associated with CRC risk in middle‐aged Japanese adults. However, for rectal cancer, the probability of an increased risk among women with a higher total sugar intake cannot be excluded.


| INTRODUC TI ON
The incidence of CRC was the third highest and CRC was the second leading cause of cancer mortality globally in 2020. 1 Processed meat and alcohol consumption, body fat, height, and insufficient physical activity are known CRC risk factors. 2 However, the impact of dietary carbohydrate or carbohydrate-containing food intake on the risk of CRC remains controversial, except for whole-grain intake. 2 Several studies among the American population have assessed the association of sugar consumption with the risk of CRC.
Nevertheless, evidence from prospective cohort studies is lacking for the Asian population whose dietary habits and prevalence of obesity (Western Pacific: 6.4%, South-East Asia: 4.7%, Americas: 28.6%, and Europe: 23.3% in 2016) 3 are distinct from those of American and European populations. Thus, sugar intake has not been established as a factor for estimating the burden of CRC in the Japanese population. 4 Previous studies on the American population have reported that total sugar, 5,6 glucose, 7 and sucrose [6][7][8][9][10] intake showed no association with CRC risk. Two previous studies on fructose intake reported a positive association, 8,10 whereas three studies reported no association. 6,7,9 Previous reports on Asian populations have been limited to case-control studies. [11][12][13] Moreover, the percentage of individuals who consume free sugar for over 10% of their daily energy intake, which is the upper limit of intake recommended by World Health Organization guidelines, 14 is lower among the Japanese population (approximately 10%) 15,16 than among populations in American (>70%) or European countries (>40%). [17][18][19][20] Meanwhile, Japan is the only Asian country among the top 10 countries with the highest incidence of CRC worldwide. 21 Accordingly, studies of populations with different patterns of sources of sugar intake and high rates of CRC incidence will provide evidence for the distinct distribution of sugar consumption among populations and may have significant implications for public health.
Comprehensive investigations of both total and specific types of simple sugars (glucose, fructose, galactose, sucrose, maltose, and lactose) are essential to clarify their health impact on CRC risk and its underlying mechanisms. In particular, fructose (both monosaccharides and components of sucrose) evokes hepatic de novo lipogenesis that is unregulated by energy intake, unlike glucose metabolism. 22, 23 Excessive de novo lipogenesis can induce insulin resistance which has been reported to be a risk factor for colorectal neoplasia. 24,25 Therefore, previous studies have focused on the potential role of fructose in health and chronic diseases. 23,[26][27][28] Here, we aimed to investigate the association of the intake of total sugars (total mono-and di-saccharides), total fructose (sum of fructose as monosaccharides and half of the sucrose), and each type of mono-and disaccharide with CRC risk in a large-scale prospective cohort study in Japan to address the current gaps in the literature.

| Study population
This study involved participants of the JPHC study, which was launched in 1990 (Cohort I) and 1993 (Cohort II). 29 At the baseline survey, individuals aged 40-69 years who lived in 11 public health center areas were recruited. Participants reported their medical and family histories and health-related lifestyles via self-administered questionnaire surveys. The questionnaire surveys were conducted three times every 5 years from the baseline survey.
As the second (5-year follow-up) questionnaire survey asked for more detailed information on nutrient and food intakes compared with the first survey, we considered data from the second survey in 1995-1999 as the baseline data in this analysis. Participants were informed of the objectives of the JPHC study and that completion of the second questionnaire was regarded as providing consent to participate. Of individuals who participated in the second survey (n = 98,469), we excluded those who answered that they were diagnosed with cancer before the second survey (n = 1614), those with missing values for nutrients (n = 1063), and those with an extremely high or low intake of total energy (above the 97.5th percentile or below the 2.5th percentile of total energy intake according to sex) (n = 4787). A final total of 91,005 participants (men: n = 42,405 and women: n = 48,600) were included in the current analysis.
This study was approved by the Institutional Review Board of the National Cancer Center of Japan (approval no. 2015-085) and was conducted following the ethical guidelines for medical research in Japan.

| Dietary assessment
In the second survey, an FFQ of 147 items was used to estimate the habitual consumption of food and beverages over the previous year. 30 Details of the questionnaire for dietary assessment are described elsewhere. 30,31 We calculated the average daily sugar consumption (each saccharide, total sugars, and total fructose) and other foods and nutrients using the 7th revised edition of the Standard Tables of Food Composition in Japan 32 and the tables for available carbohydrates. 33 We calculated the sum of sucrose, glucose, fructose, lactose, galactose, and maltose consumption as total sugar intake. Because sucrose is a disaccharide binding glucose and fructose with the same molecular weight, we used the following formula: (fructose intake) + ½(sucrose intake) 34 to calculate total fructose intake. Energy-adjusted nutrient intake was computed by the density method to express the nutrient composition of their diet and to coordinate with the measure in the World Health Organization guidelines for sugar intake. Sugars, protein, fat, carbohydrate, and starch were expressed as % energy. Other nutrients and foods were presented as "per 1000 kcal." Sugar intake estimates from the FFQ have been validated elsewhere. 16 In brief, Spearman's correlation coefficients (CCs) of total sugar intake (% energy/day) were 0.34-0.57 in Cohorts I and II according to sex, between dietary records and the FFQ. Regarding reproducibility, Spearman's CCs were 0.55-0.66 between the two FFQs with a yearly interval.

| Confirmation of colorectal cancer
Newly diagnosed CRC cases were confirmed using population-based cancer registries and hospital records in the study areas. Participants were followed up for cancer incidence, relocating out of the study areas, and death from 1995 (Cohort I) and 1998 (Cohort II) (the second survey) to December 2013. CRC cases (C18-C20) were defined according to the third edition of the International Classification of Diseases for Oncology, and were grouped by site, such as proximal colon (C18.0-C18.5), distal colon (C18.6 and C18.7), and rectal (C19 and C20) cancer. Proximal, distal, overlapping sites (C18.8), and unspecified (C18.9) colon cancer were combined and defined as colon cancer.

| Statistical analysis
Participants were divided into quintiles of each type of sugar (total sugar, total fructose, glucose, fructose, galactose, sucrose, maltose, or lactose) consumption according to sex. The mean, SD, and proportion of participant characteristics were calculated according to quintiles of total sugar intake. For the comparison of the characteristics among the quintiles, p-values were computed using analysis of variance for continuous variables and chi-squared statistics for categorical variables.
We performed Cox proportional hazards regression analysis to assess the HRs and 95% CIs of CRC risk for the quintiles of sugar consumption according to sex. The lowest category was used as the reference. We categorized sugars into total sugars, total fructose, and specific types of sugars. Potential confounders were included based on their clinical and biological plausibilities. Model 1 was adjusted for study area and age, and Model 2 was additionally adjusted for BMI (kg/m 2 , quintiles), alcohol drinking status (none, occasionally, <150, 150-299, 300-449, or ≥450 g ethanol/week in men; none, occasionally, <150 or ≥150 g ethanol/week in women), smoking status (never, past, <20 or ≥20 cigarettes/day), MET-h/day (quintiles), a history of diabetes (yes or no), family history of CRC (yes or no), CRC screening in the past year (fecal occult blood test, barium enema examination, and colonoscopy; yes or no), postmenopausal status (yes or no in women only), exogenous female hormone use (yes or no in women only), total energy intake (kcal/d, quintiles), and category of quintiles for energy-adjusted nutrient intake using the density method. The nutrients included saturated fatty acids (SFAs), n-3 polyunsaturated fatty acids (PUFAs), magnesium, vitamin D, vitamin B6, vitamin B12, calcium, folate, and dietary fiber. To examine the linear and quadratic trends across quintiles, we modeled the median sugar intake for each category as a continuous variable.
We applied stratification analysis for the association between total sugar or total fructose intake and CRC risk according to smoking status, alcohol drinking status, BMI, and a history of diabetes (without diabetes). For sensitivity analyses, participants who had cancer onset during the first 3 years of follow-up were excluded.
The overall proportion of participants with complete data for the analyses was 68.1% and 65.3% in men and women, respectively. To handle missing data on BMI, alcohol drinking status, smoking status, METs, CRC family history, postmenopausal status, and use of exogenous female hormones, 50 rounds of multiple imputations were computed using a fully conditional specification according to sex, using the MI in the SAS procedure. 35 To account for missing data, we included the total sugar intake for saccharides, CRC incidence, person-years, and the confounders described above. We then combined estimations of HRs from each imputed dataset using Rubin's rules and MIANALYZE in the SAS procedure. The level of statistical significance was set at a two-sided p-value < 0.05 level. All statistical analyses were performed using SAS (ver 9.4; SAS Institute).

| RE SULTS
We identified 2118 cases of CRC incidence (1226 for men and 892 for women) over 1,367,197.1 person-years and an average of approximately 15.0 years of follow-up. Regardless of sex, participants in the highest quintile of total sugar intake tended to comprise lower proportions of current alcohol drinkers and smokers and had a lower prevalence of a history of diabetes. In contrast, the proportion of participants who underwent CRC screening and the intake of total energy, carbohydrates, dietary fiber, calcium, magnesium, and folic acid were higher in the highest quintile than in the lowest quintile (Table 1). Additionally, both sexes had lower proportions of participants with complete data in the lowest quintile than in the other quintiles (Table 1). For free sugar intake, the proportion of participants who consumed over 10% of daily energy intake was 8.1% overall, 9.7% for men, and 6.8% for women (data not shown in the tables).
No clear association was detected between sugar intake and CRC risk. However, point estimates of multivariable adjusted HRs indicated an inverse association of total fructose with CRC risk among men; in particular, the fourth quintile showed a significant inverse association ( Table 2). In contrast with the results in men, point estimates of adjusted HRs indicated positive associations for both total sugars and total fructose in women. Although the association was significant for total sugars with CRC risk, this was limited to the third quintile (1.38 [1.10-1.74] for Q1 vs. Q3; p for quadratic trend = 0.03), and no significant linear trend was noted (p for linear trend = 0.47) ( Table 2). Similar results were indicated when we excluded the onset of cancer during the first 3 years of follow-up. When CRC was classified according to cancer site, no association was detected among TA B L E 1 Baseline characteristics of participants according to quintiles of total sugar intake.

| DISCUSS ION
We did not identify any evident associations of sugar intake with CRC risk in this large-scale population-based cohort study. Analysis according to CRC tumor site represented a potential positive association between total sugar intake and rectal cancer in women. This is the first prospective study among Asian adults to investigate the association of the total and specific types of sugar intake with CRC risk. Furthermore, we provide novel data from the perspective that the level of sugar intake is relatively low and food sources are different among the Japanese population compared with the American and European populations.
Our findings may contribute to the development of a sugar intake reference that considers the characteristics of diverse populations.
Our findings indicated the highest HRs in the third quintile of total sugar intake among women. This could be influenced by the higher health consciousness among participants in the fourth and highest quintiles. Among the Japanese, the contribution of fruits and vegetables to sugar intake is high. 16 Accordingly, participants who consumed more fruits and vegetables and had other healthy characteristics (never smoked, never drank alcohol, CRC screening, etc.) constituted a higher proportion of the fourth and highest quintiles.
Although these variables were adjusted in the models, our results may have been influenced by a healthy background, and the increase in HRs for the fourth and highest quintiles may be attenuated.
Our findings suggest the lack of an association of total sugar intake with CRC risk, in line with previous reports. 5,6 However, the increased risk in the middle level of total sugar intake among women in our study was inconsistent with past studies. 5, 6 We also observed that a moderate level of fructose intake was associated with an b "Total sugars" represents the sum of the consumption of the following saccharides: glucose, fructose, galactose, sucrose, maltose, and lactose. c "Free sugars" includes monosaccharides and disaccharides added to foods and beverages by the manufacturers, cooks, or consumers; and sugars naturally present in honey, syrups, fruit juices, and fruit juice concentrates. 14 d "Naturally occurring sugars" represents sugars from foods which do not contain added sweeteners, such as fruits, vegetables, cereals, potatoes, beans, nuts, and dairy products.

TA B L E 1 (Continued)
TA B L E 2 HRs and 95% CIs of colorectal cancer according to quintiles of total sugars and total fructose intake.

Quintiles of sugar intake
increased risk in women, which was inconsistent with previous results. [6][7][8][9][10] Findings from the Women's Health Study suggested a positive association with CRC risk in the middle level (the second, third, and fifth quintiles) of total fructose intake, 10 but the findings were not for fructose as a monosaccharide. Of note, these previous studies were conducted among the American population whose total sugar intake was approximately 2-fold higher, 6 and the contribution of sugar-sweetened beverages, sweeteners, and confectionaries were 1.5-fold higher 16,36 than that of the population in this study.
Additionally, only 8.1% of this study's participants consumed more free sugars than the upper limit defined by the WHO (10% of total energy intake), 14 compared to the studies from the USA (>70%) or the European countries (>40%). [17][18][19][20] The proportion was not much different from the literature among Japanese studies. 15,16 Nevertheless, in this study, we observed a partial positive association with CRC risk among women. Further examinations of the presence of a positive association in other Asian populations whose sugar intake is lower than that of the American population are warranted.
The disagreement in the findings between men and women may be influenced by the source of sugar intake. Overall, the directions of the associations between sugar intake and CRC risk were positive in women but were inverse in men, although compared with women, men tend to have more visceral and hepatic adipose tissue and higher insulin resistance. 37 This difference may be explained by the profile that the range of sugar intake and contribution of confectionaries was higher in women than in men according to dietary records among a subsample of the JPHC study. 16 Moreover, given that vegetables and fruits are the main food sources of sugar intake among Japanese men, 16 the preventive effects of antioxidant factors and fiber from vegetables and fruits may be greater. The effects may have at least partially contributed to the inverse association between total sugar and total fructose intake and CRC risk among individuals who had never smoked. Meanwhile, the effects may have been offset by tobacco carcinogens among smokers. Among women, the higher contribution of confectionaries to sugar intake than that observed in men may also weaken the effect, which may be why the inverse associations were not observed.
For proximal colon cancer, our findings indicated lower point estimates of HRs than those for distal colon and rectal cancer, and a possible positive association was observed with rectal cancer incrementally from the rectum to the ascending colon, 41 and fruit consumption was inversely associated with BRAF-mutated tumor risk. 42 Therefore, protection against BRAF mutations provided by fruits, which are the main source of sugar intake in the Japanese population, may modify the association of sugar intake with proximal and distal colon cancer. Moreover, the gut microbiome, which influences the association between dietary factors and colorectal carcinogenesis, 43 is differentially distributed across colorectal sites. 44 The heterogeneity in microbiome composition may have also contributed to differences in the results between rectal and colon cancer.
The strengths of this study include its large-scale and population-based prospective design, high follow-up rate, longterm follow-up, and the use of validated sugar intake estimates.
However, our study has limitations. First, the participants reported their dietary intake in the 1990s, which may not fully reflect that of the contemporary Japanese population. They may have consumed a more westernized diet with more sugars from sugar-sweetened beverages compared with that of the current study population; the contribution of sugar-sweetened beverages to total sugar intake was 11.5% in 2013, 15 and 7.1% in 1990. 16 This alteration in the composition of total sugar intake might limit the external validity of our findings. Second, although sugar intake estimates using the FFQ have been validated, some correlation coefficients with dietary records were under 0.50 for several types of sugar. 16 Moreover, correlation coefficients among the FFQ at a yearly interval were also moderate. Therefore, potential measurement errors due to self-reported intake and unmeasured changing diet were not zero; the errors may have caused misclassification, which may attenuate the associations. Third, we could not exclude potential confounding effects of unmeasured variables, such as nonsteroidal anti-inflammatory drug use, and residual confounding factors, although we attempted to adjust for potential confounders in the current analyses.
In conclusion, the results of this prospective cohort study in the Japanese population suggested that dietary sugar was not associated with CRC risk, although we cannot exclude the potential association of higher total sugar intake with an increased risk of rectal cancer in women. We provide evidence for the potential role of

ACK N OWLED G M ENTS
We appreciate all the participants and members of the survey. We are grateful to Dr. Sarah K. Abe of the National Cancer Center Institute for Cancer Control for her beneficial comments on this study.

DATA AVA I L A B I L I T Y S TAT E M E N T
For data access to JPHC Study data, please follow the instructions at https://epi.ncc.go.jp/en/jphc/805/8155.html.

E TH I C S S TATEM ENTS
Approval of the research protocol by an Institutional Reviewer