Body Mass Index and Risk of Colorectal Cancer Incidence and Mortality in Asia

Key Points Question What is the association between body mass index (BMI) and colorectal cancer (CRC) incidence and mortality in the Asian population? Findings This cohort study from The Asia Cohort Consortium that comprises data from multiple countries across Asia demonstrated positive associations between BMI and risk of incident CRC among 619 981 participants and between BMI and risk of death from CRC among 650 195 participants. Risk was greater among participants with colon cancer compared with rectal cancer and among men compared with women. Meaning The findings of this study suggest a positive association between BMI and risks of incident CRC and related mortality in the Asian population, which may offer greater understanding of the burden of obesity on CRC incidence and related deaths in this population.


Introduction
Colorectal cancer (CRC) is one of the leading causes of morbidity and mortality in the US and Europe and many other parts of the world. 1 Colorectal cancer constitutes 10% of global cancer incidence and about 9.4% of total cancer-related mortality. 2The annual incidence of CRC was estimated at 1.9 million cases in 2020, and it is projected that annual incident CRC cases will reach 3.2 million by 2040. 3 The general consensus is that CRC is a marker of socioeconomic development with incidence rates reflecting the human development index. 4The surge of CRC is mainly attributed to changes in lifestyle, environmental exposures, and dietary risk factors. 5In many Asian countries, the incidence of CRC has been rising, following the increased trend in Western countries. 6e global prevalence of obesity has shown an upward trajectory between 1975 and 2016. 7,8e adoption of a Westernized lifestyle may contribute to a high prevalence of obesity in Asian countries.For instance, in 2020, 34.3% of Chinese adults were categorized as overweight and 16.4% as obese according to the World Health Organization's definitions. 9Similarly, data from other Asian countries have indicated a steady increase in the prevalence of overweight and obesity in accordance with economic growth and changes in dietary habits. 10e association between body mass index (BMI) and incidence of CRC has been evaluated in Asia, including China, 11 Japan, 12,13 Korea, 14,15 Singapore, 16 and Taiwan. 17Accordingly, a pooled analysis of more than 300 000 Japanese individuals reported an increased risk of CRC incidence among study participants who were overweight or obese. 18Results of a prospective cohort study of more than 61 000 individuals from Singapore showed a U-shaped association between BMI and incidence of colon cancer or an increased risk of CRC in individuals who were both underweight (BMI <18.5) and overweight or obese (BMI Ն27.5) compared with individuals with what is considered a normal BMI (between 18.5 and 23.0) 16 (calculated as weight in kilograms divided by height in meters squared).The Asian Pacific Working Group on Colorectal Cancer consensus recommendations showed that obesity is a risk factor for CRC development, among other factors such as being male, having a family history of cancer, and smoking. 19forts have been made to determine the association between BMI and risk of CRC mortality in the US and Australia. 20,21For instance, the Cancer Prevention Study II, a prospective mortality study of 1 184 659 US adults, reported that BMI was associated with colon cancer death in men. 22The association between BMI and CRC mortality, however, has been relatively understudied in Asia.For example, the Asia-Pacific Cohort Studies Collaboration, which included only 668 CRC deaths, found an increased risk of CRC mortality among individuals with obesity. 23The Japan Collaborative Cohort Study for Evaluation of Cancer, which included 127 colon cancer deaths, reported a positive association between BMI and cancer-related mortality among women only. 24There is a gap of knowledge on the association between BMI and CRC incidence and related mortality across Asia.
In the current study, we examined comprehensively the association between BMI and risk of CRC incidence and mortality across Asia.We included pooled data from The Asia Cohort Consortium that included more than 1 million participants.

Study Populations
Data for this cohort study were derived from The Asia Cohort Consortium (ACC), an international collaborative effort including more than 1 million participants across Asia.6][27] All participants were required to sign an informed consent form.Each cohort received approval from its respective institutional review board to conduct research, and the pooled analysis was approved by the ACC Executive Committee and by the Ethical Committee of the National Cancer Center Japan (additional details are provided in the eMethods in Supplement 1).This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

JAMA Network Open | Oncology
Body Mass Index and Colorectal Cancer Incidence and Mortality in Asia In this study, a total of 17 cohorts from mainland China, Japan, South Korea, Taiwan, Singapore, and Iran were included.Cohort enrollment was conducted from January 1, 1984, to December 31, 2002.Median follow-up time was 15.2 years (IQR, 12.1-19.2years).Due to the limited data on the date of CRC diagnosis, 2 cohorts from Taiwan were excluded from CRC incidence analysis, while the CRC mortality analysis was performed using all cohorts.
For analysis of CRC incidence, the exclusion criteria included (1) a missing or an extreme BMI value (>80 or <10) at baseline, (2) the unknown status of CRC diagnosis during the follow-up period, (3) unavailable time to CRC diagnosis, (4) CRC not being the first primary cancer diagnosed, or (5)   missing data on potential confounding factors.To prevent reverse causality, we also excluded participants with CRC and person-years observed within the first 3 years of observation after the enrollment.
Similarly, the following exclusion criteria were used in the analysis of CRC mortality: (1) a missing or an extreme BMI value (>80 or <10) at baseline; (2) missing vital status; (3) missing the date of CRC death; and (4) missing data on potential confounding factors, including age at baseline, diabetes, being a current smoker, and ever using alcohol.Deaths from CRC and person-years observed within the first 3 years of observation after the enrollment were excluded to minimize reverse causality.codes C19, C20, and D37.5.For those who had a CRC diagnosis with a missing ICD-9 or ICD-10 code, we included those participants in the incidence analysis, but they were excluded from the subsite analyses.We constructed the time to event by computing the interval between the date of enrollment and the date of incident CRC or death from CRC.For participants not experiencing these 2 outcomes, they were censored at the date of last contact or the date of death from other causes, whichever came first.

Primary Outcomes, Exposures, and Other Covariates
Body mass index was obtained from either the measurement provided by cohorts (11 cohorts)   or calculated using self-reported height and weight (6 cohorts).Consistent with a previous analysis in the ACC, 28 we used the following cutoffs: less than 18.5, 18.5 to 23.0, greater than 23.0 to 25.0, greater than 25.0 to 27.5, greater than 27.5 to 30.0, greater than 30.0 to 35.0, and greater than 35.0, in which greater than 23.0 to 25.0 was considered the reference group.A previous study in the ACC revealed that the BMI category of 23.0 to 25.0 had the lowest risk of all-cause mortality. 28As there were few CRC cases or related deaths in the category of greater than 35.0 (<1% in all cohorts except the Golestan [Iran] cohort), we combined this category with the greater than 30.0 to 35.0 category.

Statistical Analysis
Data were analyzed from January 15, 2023, to January 15, 2024.Adjusted hazard ratios (AHRs) and 95% CIs were calculated using Cox proportional hazards regression models, in which the cohort indicator was included as a random effect term to account for between-study heterogeneity.For primary analyses, multivariable AHRs were obtained after adjusting for age, sex, diabetes, current smoking status, and ever alcohol use.A further stratified analysis was performed by sex, country, diabetes, smoking status, and alcohol use.We examined the statistical significance of the product term between BMI and the corresponding stratification factor in multivariable models to assess whether heterogeneity existed in any of these subgroups.Potential sex difference was also explored in each subgroup of country, diabetes, smoking, and alcohol use.The definition of a complete dataset with covariates included those with complete information of age at baseline, sex, educational level, marital status, diabetes at baseline, current smoker, ever alcohol use, and enrollment period.
Linear trends between BMI with CRC incidence and mortality were assessed by including a continuous variable that took the median BMI of each BMI subgroup using Wald tests.To assess the nonlinear associations between BMI and CRC incidence and mortality, a restricted cubic spline analysis was performed in which we set a BMI of 24.0 as the reference value, and 4 knots were placed at equally spaced percentiles given the distribution of BMI in the specific study population as previously recommended. 29l statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc) or Stata, version 17.0 (StataCorp LLC).All P values were 2-sided, and P < .05 was the threshold for statistical significance.Bonferroni correction was used for subgroup analyses.(3.2), and in the population, 66.3% had a BMI between 18.5 and 25.0, 25.4% were obese (BMI between >25.0 and 30.0), and 3.5% had a BMI greater than 30.0 (eTables 1 and 2 in Supplement 1).

BMI and Risk of CRC Incidence
We observed an increased risk of CRC incidence in individuals with a BMI greater than 25.0 compared with individuals with what is considered normal BMI (ie, >23.0-25.0).Specifically, the risk of developing CRC consistently increased from 9% to 32% in individuals with a BMI greater than 25.0 (>25.0-27.5:AHR, 1.09 [95% CI, 1.03-1.16];>27.5-30.0:AHR, 1.19 [95% CI, 1.11-1.29];and >30.0:AHR, 1.32 [95% CI, 1.19-1.46];P < .001for trend).These findings remained unchanged after further adjusting for age and marital status at baseline, educational level, diabetes, smoking, alcohol consumption, and enrollment period (eTable 3 in Supplement 1).Additionally, the positive linear trend identified using BMI categories was consistent with the spline curves in which the risk of incident CRC was analyzed over the spectrum of BMI as a continuous variable (Figure 1A and eFigure 1A in Supplement 1).The association pattern was similar between sexes (Table 1 and Figure 1B and eFigure 1B in Supplement 1).

BMI and Risk of CRC Mortality
We found positive associations between individuals with a BMI greater than 27.5 and risk of CRC-related mortality compared with participants with what is considered a normal BMI (ie, >23.0-25.0).Among individuals with a BMI of greater than 27.5, the risk of CRC-related death showed a consistent increase across BMI categories ranging from 18% to 38% (>27.5-30.0:AHR, 1.18 [95% CI, 1.04-1.34]and >30.0:AHR, 1.38 [95% CI, 1.18-1.62];P < .001for trend) (Table 2 and Figure 2A).A heterogeneity between men and women for the association between BMI and CRC-related deaths is suggested (P = .02for heterogeneity) (Table 2), which was also supported by sex-specific spline curves (Figure 2B and eFigure 2B in Supplement 1), and an association between BMI and CRC-related  .001for trend) but not among women (P = .15for trend).The association followed a J-shaped pattern for men, while a flat spline curve was observed for women over the spectrum of BMI as a continuous variable in both main analyses and sensitivity analyses.These findings remained unchanged after further adjusting for age and marital status at baseline, educational level, diabetes, smoking, alcohol consumption, and enrollment period (eTable 4 in Supplement 1).

Incidence and Mortality of CRC Subsites
For incidence and mortality analyses according to subsites, the findings were in line with what was described above when colon cancer, rectal cancer, and unknown CRC groups were combined as CRC.
Risk of CRC incidence and mortality associated with BMI was greater among individuals with colon cancer (eTables 5-8 in Supplement 1).

Stratified Analyses by Geographic Location
Regarding the association between BMI and CRC incidence, our analysis showed no difference across geographic locations (Table 3).No association was found between BMI categories and geographic subgroups, suggesting that the association between BMI and CRC incidence and mortality was homogeneous across subpopulations.
We found a positive association between BMI and CRC-related mortality in China, Japan, and Iran.On the other hand, this significant trend was not present in South Korea.However, the patterns of association between BMI and CRC-related mortality did not show significant difference across geographic locations (eTable 9 in Supplement 1).

Stratified Analysis by Diabetes, Smoking, and Alcohol Use
Overall, the positive associations between BMI and risk of incident CRC and risk of death from CRC were mostly found in stratified analyses by diabetes, current smoking status, and ever alcohol use (eTables 10 and 11 in Supplement 1).Given the potential heterogeneity between men and women, we also explored a possible sex difference for the examined associations in various subgroups (eTables 12-15 in Supplement 1).Specifically, in the analysis of participants with complete data on covariates available, the risk of CRC-related mortality appeared to be more pronounced for men than for women among current smokers and those who consume alcohol.It should be noted that smoking was far less common in women than in men (current smokers constituted 5.8% of female and 51.0% of male participants).
We further conducted a sensitivity analysis by comparing results between cohorts with an actual measurement of BMI and cohorts with self-reported BMI.There was no distinguishable difference in the association between BMI and CRC incidence and mortality using the 2 methods of BMI measurements (eTables 16-19 in Supplement 1).

Discussion
This cohort study of 619 981 Asian participants assessed for CRC incidence and 650 195 Asian participants assessed for CRC-related mortality showed a positive association between BMI and risks of CRC incidence and CRC-related mortality.Risk of CRC-related mortality associated with BMI was greater among men with a BMI of more than 30.0 than among women, and risk of CRC incidence and mortality associated with BMI was more prominent among individuals with colon cancer.
Obesity has been associated with increased risk and mortality of many types of cancers, including CRC. 30,31 The mutagenic capability of obesity using peroxidation pathways as well as a dysregulated metabolism may promote tumor development. 32Specifically, an altered metabolism of glucose and increased insulin secretion may be responsible for the increased risk of CRC in people with obesity. 33This effect is mainly modulated by insulin as a mitogen and tumor growth promoter. 34 addition, the low-grade chronic inflammatory status stimulates secretion of signaling cytokines that are involved in initiation, progression, and metastasis of CRC. 31 Moreover, as a state of nutrient excess, obesity promotes neoplastic transformation using activating cellular growth pathways. 35In the current study, we detected a positive association between BMI and CRC incidence among individuals who were both diabetic and nondiabetic.This finding may partly explain the association of BMI with CRC development, independent of an altered glucose metabolic pathway.Compared with North Americans and Europeans, East Asians are leaner and tend to develop diabetes and glucose intolerance at lower BMI levels. 36,37Therefore, to better understand the impact of obesity and the potential impact of the altered glucose metabolism among Asian individuals, BMI cutoff points that are consistent with those of the ACC should be applied.
One main finding in our study was an association between BMI and risk of CRC-related death among men with a BMI of more than 30.0.The extrapolation of BMI on survival of CRC cases and the disparity between men and women are subjects of ongoing debate.Similar to our findings, a systematic review reported that prediagnostic obesity (but not overweight) was associated with increased risk of mortality among men only. 38The differential distribution of adipose tissue may partly explain the higher risk to tumorigenesis and cancer mortality in men.Men have a predilection to develop central adiposity rather than generalized obesity, which is more common in women. 39ntral adiposity has been reported to be strongly correlated with hyperinsulinoma and more intense insulin resistance. 40This association is prominent among Asians, as central obesity is an established risk factor for type 2 diabetes in addition to BMI. 41 Furthermore, excessive visceral adipose tissue distribution is an established risk factor for comorbid states such as type 2 diabetes, which increases the risk of colon cancer.The comparative studies have been between European or American and Asian populations and reported a distinguished pattern in adipose tissue 42 and risk of insulin resistance in relation to a visceral tissue compartment. 43It is noteworthy that the sex differential association between higher BMI and CRC mortality may be partially explained by incomplete adjustment or residual confounding by smoking and alcohol consumption among women.Therefore, the differential confounder adjustment may be partly responsible for the finding regarding CRC mortality among men and women.
Another explanation for our primary finding is the potential association between hormone replacement therapy (HRT) and BMI and the risk of CRC-related death in women.A meta-analysis of 5 cohorts, including 10 013 survivors of CRC, found an inverse association between current use of HRT and CRC-specific mortality and overall mortality. 44Additionally, Asian women have a relatively low prevalence of HRT use, 45 which must be considered in relation to global trends.The other reason might be a higher adherence of women to screening. 46While the current study adjusted for smoking and alcohol use status, detailed quantifiable data on smoking or alcohol consumption may be crucial in assessing the risk of mortality.Men also smoke and consume more alcohol than women. 47e association between BMI and CRC-specific mortality has been reported in different studies worldwide. 16,31,48Moreover, insulin resistance is an established factor associated with decreased survival in CRC. 49In the current study, the positive association between BMI and risk of CRC-related mortality was present in subpopulations of China, Iran, and Japan.The absence of a significant trend in CRC-related mortality, specifically in South Korea, may be partly explained by the limited number of recorded deaths in the South Korean participating cohort.The other countries that had significantly positive trends were China and Iran, in which the Golestan cohort study, for instance, was associated with either being overweight or obese (BMI Ն27.5).However, the overall patterns of association between BMI and CRC-related mortality were comparable across geographic locations.
A study by Croft et al 50 reported that among 4 components of metabolic syndrome (ie, overweight or obesity, diabetes, hypertension, and dyslipidemia), only diabetes was associated with progression-free survival, while obesity, dyslipidemia, and hypertension did not show a similar association with the outcome.The lack of a significant difference in mortality among the diabetic and nondiabetic strata may be attributable to the lack of power.

Strengths and Limitations
Our study has several strengths.This study benefited from the prospective design that limited recall bias.Body mass index was stratified using ACC-based cutoffs, which made the clinical interpretations more relevant for Asian populations.Furthermore, the analysis of individual-level data from large multicenter cohorts allowed us to evaluate the associations between BMI and risk of CRC incidence and mortality with standardized categorization of exposure and confounding variables.We could separately investigate tumors located in the colon or the rectum.Furthermore, the large sample size allowed us to conduct sensitivity analyses to evaluate the reverse causality, thus strengthening the rigorousness of our study.
Our study also has several limitations.First, the anthropometric data acquisition in some cohorts was based on self-reported forms, although the validation of the self-reported BMI, height, and weight was high among these cohorts. 12,51Second, there was a lack of longitudinal measurement of BMI, as our analyses were based on a baseline measurement of BMI.Therefore, the current study could not evaluate the impact of subsequent changes in BMI.Third, our analyses might have effects of residual confounding, as we did not control for other important confounding factors in the multivariable models, such as physical activity, socioeconomic status, or family history of CRC.
Furthermore, the tumor stage and treatment regimen were not captured at the time of enrollment.
While overall the study enrollment was focused on the early stage of the disease, these factors may play a partial role in the outcome of participants.

Conclusions
This cohort study using data derived from the ACC demonstrated positive associations between BMI and risk of incident CRC and related deaths.These risks were more apparent in participants with colon cancer and were greater in men.Findings from this study may have substantial implications regarding our understanding of the burden of BMI on CRC incidence and deaths in Asian populations.

Figure 1 .
Figure 1.Association Between Body Mass Index and Risk of Incident Colorectal Cancer

Table 1 .
Associations Between BMI and CRC Incidence by Sex Abbreviations: AHR, adjusted hazard ratio; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CRC, colorectal cancer; NA, not applicable.aPvaluewas obtained from the interaction test between BMI and sex after model adjustment.bModeladjusted for age at baseline, sex, diabetes at baseline, current smoker, and ever alcohol use.cModel adjusted for age at baseline, diabetes at baseline, current smoker, and ever alcohol use.

Table 2 .
Associations Between BMI and CRC Mortality by Sex was only found among men with a BMI of more than 30.0 (AHR, 1.87 [95% CI, 1.49-2.34];P < Abbreviations: AHR, adjusted hazard ratio; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CRC, colorectal cancer; NA, not applicable.aPvalue was obtained from the interaction test between BMI and sex after model adjustment.bModel adjusted for age at baseline, sex, diabetes at baseline, current smoker, and ever alcohol use.c Model adjusted for age at baseline, diabetes at baseline, current smoker, and ever alcohol use.death

Table 3 .
Associations Between BMI and CRC Incidence in Subgroups Model adjusted for age at baseline, sex, current smoker, and ever alcohol use.
a P value was obtained from the interaction test between BMI and subgroup after model adjustment.bModel adjusted for age at baseline, sex, diabetes at baseline, current smoker, and ever alcohol use.c d Model adjusted for age at baseline, sex, diabetes at baseline, and ever alcohol use.e Model adjusted for age at baseline, sex, diabetes at baseline, and current smoker.
This is an open access article distributed under the terms of the CC-BY License.© 2024 Paragomi P et al.JAMA Network Open.Hung N. Luu, MD, PhD, UPMC Hillman Cancer Center, UPMC Cancer Pavilion, 5150 Centre Ave, Ste 4C, Room 466, Pittsburgh, PA 15232 (hnl11@pitt.edu).Division of Cancer Control and Population Sciences, University of Pittsburgh Medical Center Hillman Cancer Center, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania (Paragomi, Pham, Yuan, Wang, Luu); Department of Epidemiology, University of Pittsburgh School of Public Health, Pittsburgh, Pennsylvania (Zhang, Pham, Yuan, Luu); Division of Prevention, National Cancer Center Institute for Cancer Associations Between BMI and CRC Incidence eFigure 2. Associations Between BMI and CRC Mortality eTable 1. Characteristics of Participating Cohorts in The Asia Cohort Consortium for CRC Incidence Analysis eTable 2. Characteristics of Participating Cohorts in The Asia Cohort Consortium for CRC Mortality Analysis eTable 3. Associations Between BMI and CRC Incidence During Follow-up by Sex Among Those Having Complete Data on Covariates eTable 4. Associations Between BMI and CRC Mortality During Follow-up by Sex Among Those Having Complete Data on Covariates eTable 5. Associations Between BMI and CRC Histologic Subtypes Incidence During Follow-up by Sex eTable 6. Associations Between BMI and CRC Subtypes Incidence by Sex Among Those With Complete Data on Covariates eTable 7. Associations Between BMI and CRC Histologic Subtypes Mortality by Sex eTable 8. Associations Between BMI and CRC Histologic Subtypes Mortality by Sex Among Those With Complete Data on Covariates eTable 9. Associations Between BMI and CRC Mortality in Subgroups Among Those Having Complete Data on Covariates eTable 10.Associations Between BMI and CRC Incidence in Subgroups Among Those With Complete Data on Covariates eTable 11.Associations Between BMI and CRC Mortality in Subgroups eTable 12. Associations Between BMI and CRC Incidence by Sex in Subgroups eTable 13.Associations Between BMI and CRC Incidence by Sex in Subgroups Among Those With Complete Data on Covariates eTable 14.Associations Between BMI and CRC Mortality by Sex in Subgroups eTable 15.Associations Between BMI and CRC Mortality by Sex in Subgroups Among Those With Complete Data on Covariates eTable 16.Associations Between BMI (Actual Measurement) and CRC Incidence by Sex eTable 17.Associations Between BMI (Self-Report) and CRC Incidence by Sex eTable 18. Associations Between BMI (Actual Measurement) and CRC Mortality by Sex eTable 19.Associations Between BMI (Self-Report) and CRC Mortality by Sex