A Prospective Investigation of Body Size, Body Fat Composition and Colorectal Cancer Risk in the UK Biobank

Obesity has been consistently associated with a greater colorectal cancer risk, but this relationship is weaker among women. In the UK Biobank, we investigated the associations between body size (body mass index [BMI], height, waist circumference, and waist-to-hip ratio) and body fat composition (total body fat percentage and trunk fat percentage) measurements with colorectal cancer risk among 472,526 men and women followed for 5.6 years on average. Multivariable hazard ratios (HRs) and 95% confidence intervals (95%CI) for developing colorectal cancer (2,636 incident cases) were estimated using Cox proportional hazards models. Among men, when the highest and lowest fifths were compared, BMI (HR = 1.35, 95%CI: 1.13–1.61; Ptrend < 0.0001), waist circumference (HR = 1.66, 95%CI: 1.39–1.99; Ptrend < 0.0001), waist-to-hip ratio (HR = 1.58, 95%CI: 1.31–1.91; Ptrend < 0.0001), total body fat percentage (HR = 1.27, 95%CI: 1.06–1.53; Ptrend = 0.002), and trunk fat percentage (HR = 1.31, 95%CI: 1.09–1.58; Ptrend = 0.002) were associated with greater colorectal cancer risk. For women, only waist-to-hip ratio (HR for highest versus lowest fifth = 1.33, 95%CI: 1.08–1.65; Ptrend = 0.005) was positively associated with colorectal cancer risk. Greater body size (overall and abdominal adiposity) was positively associated with colorectal cancer development in men. For women, abdominal adiposity, rather than overall body size, was associated with a greater colorectal cancer risk.


Statistical analysis.
Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards models. Age was the primary time variable in all models. Time at entry was age at recruitment. Exit time was age at whichever of the following came first: colorectal cancer diagnosis, death, or the last date at which follow-up was considered complete. Models were stratified by age at recruitment in 5-year categories, Townsend deprivation index fifths, and region of the recruitment assessment centre. Possible non-proportionality was assessed using an analysis of Schoenfeld residuals 11 , with no evidence of nonproportionality being detected. The associations between body size and body fat composition measurements with risks of colorectal cancer, colon cancer, proximal colon cancer, distal colon cancer, and rectal cancer were analysed separately for men and women. Heterogeneity of associations across anatomical cancer subsites was assessed by calculating X 2 statistics.
Participants were grouped into sex-specific fifths of BMI, body weight, height, waist circumference, waist-to-hip ratio, total body fat percentage, and trunk fat percentage. In additional analyses, each body size and composition measurement was modelled on the continuous scales (BMI [per 2.5 kg/m 2  other professional qualifications; college/university degree; unknown), ever use of HRT (no, yes, unknown), and frequency of red and processed meat consumption (<2 times/week; 2-3 times/week; 3-4 times/week; ≥4 times/ week; unknown). Trend tests across body size and composition measurement fifths were calculated by assigning the median value of each category and modelling as continuous terms in Cox regression models. The body size and colorectal cancer associations were further assessed across subgroups of ever HRT use, menopausal status (pre-menopausal and post-menopausal), prevalent diabetes (no or yes), and physical activity levels (MET hours/week; thirds). Interaction terms (multiplicative scale) between these variables and body size were included in separate models; the statistical significance of the cross-product terms were evaluated using the likelihood ratio test. To evaluate possible reverse causality, cases diagnosed within the first two years of follow-up were excluded in sensitivity analyses. Statistical tests used in the analysis were all two-sided and a P value of <0.05 was considered statistically significant. Analyses were conducted using Stata version 13.1.

Results
After a median follow-up time of 5.6 years, 2,636 incident cases of colorectal cancer (1,516 were in men and 1,120 were in women) were diagnosed. Of these cases, 1,731 were colon tumours (874 proximal colon, 769 distal colon, and 88 overlapping or unspecified) and 889 were rectal tumours. Compared with those in the lower fifth, men and women in the highest BMI fifth were less physically active, less likely to be never smokers and frequent alcohol consumers, less likely to have attained college education or university degree, and more likely to be regular aspirin/ibuprofen users and more frequent consumers of red and processed meat ( Table 1). In addition, ever HRT use was higher for women in the highest BMI fifth compared to those in the lowest fifth (Table 1).
Weight was associated with a higher risk of colorectal cancer among men (HR for highest versus lowest fifth = 1.57, 95%CI: 1.29-1.91; P trend < 0.0001), but not women (HR for highest versus lowest fifth = 1.11, 95%CI: 0.89-1.38; P trend = 0.56) ( Tables 2 and 3). For men, this positive relationship was slightly stronger for colon cancer versus rectal cancer (P heterogeneity = 0.09) and for proximal colon versus distal colon cancer (P heterogeneity = 0.08), although these differences did not reach statistical significance. Among women, no differences in the weight and colorectal cancer relationship was found across anatomical subsites (colon versus rectal P heterogeneity = 0.18; proximal colon versus distal colon P heterogeneity = 0.12) ( Table 3). Height was not associated with colorectal cancer risk among men (HR for highest versus lowest fifth = 1.01, 95%CI: 0.85-1.19; P trend = 0.67) and women (HR for highest versus lowest fifth = 1.11, 95%CI: 0.90-1.37; P trend = 0.30), with similar relationships found across anatomical subsites (all P heterogeneities > 0.20) (Tables 2 and 3).
Among men, waist circumference and waist-to-hip ratio were associated with a greater risk of colorectal cancer (waist circumference: HR for highest versus lowest fifth = 1.66, 95%CI: 1.39-1.99; P trend < 0.0001, waist-to-hip ratio: HR for highest versus lowest fifth = 1.58, 95%CI: 1.31-1.91; P trend < 0.0001) ( Table 2). Similar strength positive relationships were observed across anatomical subsites (colon versus rectal: waist circumference P heterogeneity = 0.27, waist-to-hip ratio P heterogeneity = 0.42; proximal colon versus distal colon: waist circumference P heterogeneity = 0.27, waist-to-hip ratio P heterogeneity = 0.61) ( Table 2). For women, greater waist-to hip ratio was also associated with an elevated colorectal cancer risk (HR for highest versus lowest fifth = 1.33, 95%CI: 1.08-1.65; P trend = 0.005) ( Table 3). This positive relationship for waist-to-hip ratio was of similar magnitude for colon cancer and rectal cancer (P heterogeneity = 0.84), and for proximal colon and distal colon cancer (P heterogeneity = 0.24) ( Table 3). Among women, waist circumference was positively related to colorectal cancer risk, although this relationship did not quite reach statistical significance (Table 3). Although no relationship was observed for waist circumference and rectal cancer in women, no heterogeneity was observed when compared with tumours in the colon (P heterogeneity = 0.35). Waist circumference was positively associated with proximal colon cancer risk in women (HR for highest versus lowest fifth = 1.49, 95%CI: 1.04-2.14; P trend = 0.04), but not distal colon cancer risk (P heterogeneity = 0.15).
Body fat composition and colorectal cancer risk. Among men, higher total body fat percentage and trunk fat percentage were associated with greater colorectal cancer risk (total body fat percentage: HR for highest versus lowest fifth = 1.27, 95%CI: 1.06-1.53; P trend = 0.002, trunk fat percentage: HR for highest versus lowest fifth = 1.31, 95%CI: 1.09-1.58; P trend = 0.002) ( Table 2). These positive relationships were stronger and reached statistical significance for colon cancer, but not for rectal cancer (total body fat percentage P heterogeneity = 0.21; trunk fat percentage P heterogeneity = 0.24), and for proximal colon cancer, but not distal colon cancer (total body fat percentage P heterogeneity = 0.90; trunk fat percentage P heterogeneity = 0.64) ( Table 2). For women, total body fat percentage and trunk fat percentage were not associated with colorectal cancer risk (Table 3).
Sensitivity and sub-group analyses. The positive association observed between waist-to-hip ratio and colorectal cancer among women was evident among never users of HRT (HR per 0.25 increase = 1.42, 95%CI: 1.02-1.95) and ever users of HRT (HR per 0.25 increase = 1.60, 95%CI: 1.14-2.23) (P interaction = 0.27) ( Similarly, the association of BMI and waist circumference with colorectal cancer did not differ across HRT use groups (P interactions > 0.12). The associations between BMI, waist circumference, and waist-to-hip ratio with colorectal cancer risk were similar across subgroups of physical activity levels, prevalent diabetes status, and menopausal status (all P interactions > 0.10) ( Table 4). Similar relationships were also observed when colorectal cancer cases diagnosed during the first two years of follow-up were excluded from the analyses (data not shown).

Discussion
In this prospective analysis of UK Biobank participants, overall and abdominal adiposity body size and composition measurements were positively related to colorectal cancer risk among men. For women, only greater abdominal adiposity was associated with elevated colorectal cancer risk, and this relationship did not differ by HRT use. Our finding that BMI was positively related to colorectal cancer risk for men is in accordance with a large body of epidemiological evidence 1,12 . For women, we observed no association between BMI and colorectal cancer development, which is inconsistent with a recent analysis from the Million Women's Study (n = 18,518 cases) 13 and two meta-analyses of >30 individual prospective studies 1,12 . This suggests that the null finding we observed between BMI and colorectal cancer among women may be the consequence of insufficient follow-up time and recorded case numbers that are necessary to detect the weak-to-moderate strength association which has previously been reported.
We found that greater abdominal adiposity measurements were positively related to colorectal cancer risk in men and women, findings consistent with the EPIC study and meta-analyses of prospective studies 1,4 , but inconsistent with the findings from a National Institutes of Health (NIH)-AARP Diet and Health Study analysis that reported positive associations for waist circumference and waist-to-hip ratio for men, but null relationships for women 6 . This was the first large-scale study to investigate how body fat composition measurements relate to colorectal cancer risk. Analyses of total body fat percentage and colorectal cancer revealed similar relationships to those observed for BMI. For men, greater total body fat percentage was associated with elevated risks of developing colorectal cancer and colon cancer, a finding in that is in accordance with results from a Swedish cohort study which reported a RR of 1.54 (95%CI: 1.06-2.23; P-trend = 0.012) for colorectal cancer (n = 584 cases) when the highest and lowest fourths were compared 14 . For women, we observed a null relationship between colorectal cancer risk and total body fat percentage, a result consistent with a sub-study of 11,124 post-menopausal women (n = 169 cases) from the Women's Health Initiative 15 . For trunk fat percentage, we found a positive relationship with colorectal cancer and colon cancer risk for men, and null associations for women. In general, our results for body fat composition were consistent with our findings for body size measurements.
It is not known why the positive relationships between adiposity and colorectal cancer are weaker and more inconsistent for women. One proposed explanation, is that greater exposure to estrogens in women may mitigate the potential tumorigenic effects of excess adiposity on the colorectum 3 . Experimental and epidemiological studies suggest that estrogens are potentially protective against the development of colorectal cancer [16][17][18][19][20] , while in observational studies, ever use of HRT has been consistently associated with lower colorectal cancer risk 2 . In the EPIC study, abdominal adiposity was only associated with greater colon cancer risk among non-HRT users, with no relationship observed for HRT users 4 . However, in the current UK Biobank study, we found no evidence that HRT use modified the abdominal adiposity and colorectal cancer relationship, as we observed consistent positive relationships among users and never users of HRT. There is evidence that the different formulations of HRT (e.g. estrogen only or estrogen plus progestin) may have distinct associations with colorectal cancer risk 21,22 .  We did not have information on the formulation of HRT used which may explain why we did not observe effect modification of the abdominal adiposity and colorectal cancer relationship by HRT use. Nonetheless, it remains possible that higher endogenous circulating levels of estrogens found in women may lessen the impact of adiposity on colorectal cancer development. Most recently, a case-control analysis nested within the Women's Health Initiative found that higher levels of endogenous circulating estradiol and estrone were associated with a lower colorectal cancer risk in post-menopausal women who were non-HRT users 23 . In the same study, the positive waist circumference and colorectal cancer relationship was strengthened, and became statistically significant, after the multivariable models were additionally adjusted for endogenous circulating estrogen concentrations 23 . This suggests that higher endogenous circulating estrogen concentrations may mask the body size and colorectal cancer relationship in post-menopausal women, and that future studies investigating how adiposity relates to colorectal cancer development in post-menopausal women should incorporate estrogen measurements to limit the effects of this confounding bias 23 .
The biological mechanisms that underlie the relationship between adiposity and colorectal cancer development are uncertain. Adiposity is associated with metabolic and endocrinologic abnormalities, such as alterations in insulin and insulin-like growth factor signalling. Experimental and epidemiological evidence indicate that hyperinsulinemia and elevated levels of insulin-like growth factor-1 (IGF-1) may promote colorectal tumorigenesis by stimulating mitogenesis and inhibiting apoptosis [24][25][26][27][28][29][30] . Excess body weight is considered a chronic inflammatory state characterised by increased adipose secretion of proinflammatory cytokines, such as interleukin-6 (IL-6), which has been shown to promote tumour initiation in experimental models 31 . Higher adiposity is also inversely related to the secretion of adiponectin, an adipose tissue secreted adipokine which has been shown to lower secretion of inflammatory cytokines, improve insulin sensitivity, and inhibit cell growth and angiogenesis 32,33 . In epidemiological studies, higher adiponectin levels have, in general, been associated with lower colorectal cancer risk 34,35 .
The current analysis represents one of the largest single studies investigating the relationships between body size and composition with colorectal cancer risk to date. This was the first time that bioelectrical impedance measurements, that distinguish between adipose and non-adipose mass, have been used on such a large-scale to assess how total body fat percentage and trunk fat percentage relate to colorectal cancer development. The large number of incident colorectal cancer cases allowed analyses by HRT use, physical activity levels, diabetes status,  Table 4. Subgroup analysis of the associations of body size measurements and colorectal cancer risk among men and women. All HRs per specified unit increase. Multivariable models only -Cox regression using physical activity (MET hours/week; fifths), smoking status and intensity (never; former; current -<15 cigarettes/day; current -≥15 cigarettes/day; current -intensity unknown; unknown), alcohol consumption frequency (never; special occasions only; 1-3 times/month; 1-2 times/week; 3-4 times/week; daily or almost daily; unknown), family history of colorectal cancer (no; yes; unknown), prevalent diabetes (no; yes; unknown), regular aspirin/ ibuprofen use (no; yes; unknown), qualifications (none; CSEs/O-levels/GCSEs or equivalent; NVQ/HND/HNC/ A-levels/AS-levels or equivalent; other professional qualifications; college/university degree; unknown), ever use of hormone replacement therapy (no, yes, unknown), frequency of red and processed meat consumption (<2 times/week; 2-3 times/week; 3-4 times/week;≥4 times/week; unknown), and stratified by age (5-year categories), Townsend deprivation index fifths, and region of the recruitment assessment centre. † Physical activity (MET hours/week) thirds used to categorise participants into low (lower third), medium (middlethird), and high (upper-third) levels of activity.
Scientific REPORtS | (2017) 7:17807 | DOI:10.1038/s41598-017-17997-5 menopausal status, and tumour anatomical site, and the detailed phenotypic information collected from UK Biobank participants enabled us to carefully adjust for known colorectal cancer risk factors. A limitation of our analysis is that body size and composition measurements were collected from all participants on one occasion; however, with an average follow-up of 5.6 years it is unlikely that body measurements would have changed markedly during this time to substantially alter our results.
In conclusion, in this prospective analysis of UK Biobank participants, which used the most comprehensive panel of anthropometric measurements of any study to date, greater body size (overall and abdominal adiposity) was positively associated with colorectal cancer development in men. For women, abdominal adiposity, rather than overall body size, was associated with a greater colorectal cancer risk, and this relationship was unaffected by HRT use. The null result observed for the BMI and colorectal cancer among women may have been a consequence of insufficient follow-up time and recorded case numbers to detect the weak-to-moderate strength association which has previously been reported. Our findings add to the large body of evidence which supports the promotion of weight control in population-wide cancer prevention programmes.