Background:

Dyslipidemia is closely associated with metabolic syndrome, a known risk factor for colorectal cancer. However, the association of dyslipidemia with colorectal cancer risk is controversial. Most previous studies did not consider cholesterol-lowering medication use at the time of lipid measurements, which could bias findings.

Methods:

We analyzed data from 384,862 UK Biobank participants to disentangle the associations between blood lipids and colorectal cancer risk. Serum levels of total cholesterol, high- and low-density lipoprotein cholesterol (HDL-C, LDL-C), and triglyceride were measured at study baseline. Multivariable-adjusted Cox models were used to estimate HRs and 95% confidence intervals (CI).

Results:

During a median follow-up time of 8.2 years, 3,150 incident primary colorectal cancer cases were identified. Triglyceride levels were positively, while HDL-C levels were inversely associated with colorectal cancer risk (both Ptrend < 0.005). No significant associations were found for total cholesterol and LDL-C. However, among nonusers of cholesterol-lowering medications, a high total cholesterol level (> 6.7 mmol/L, HR = 1.11; 95% CI, 1.00–1.24) and LDL-C level (>4.1 mmol/L, HR = 1.11; 95% CI, 0.99–1.23) was associated with an increased colorectal cancer risk compared with the referent group (5.2–6.2 mmol/L and 2.6–3.4 mmol/L for total and LDL cholesterol, respectively). Compared with nonusers, cholesterol-lowering medication users had 15% increased colorectal cancer risk (HR = 1.15; 95% CI, 1.04–1.26).

Conclusions:

Circulating total cholesterol, LDL-C, HDL-C and triglyceride were modestly associated with colorectal cancer risk.

Impact:

Our findings call for careful consideration of cholesterol-lowering medication use in future studies of blood lipid–colorectal cancer associations.

This article is featured in Highlights of This Issue, p. 463

Globally, colorectal cancer is the third most commonly diagnosed cancer and the second leading cause of cancer-related deaths (1). One known risk factor for colorectal cancer is metabolic syndrome (2), which encompasses several metabolic disorders including obesity, hypertension, hyperglycemia, and dyslipidemia. Dyslipidemia is characterized by decreased level of high-density lipoprotein cholesterol (HDL-C), and elevated levels of low-density lipoprotein cholesterol (LDL-C) and triglyceride (2). LDL-C has been found to stimulate inflammation and tumor progression in human colorectal cancer cells via activation of reactive oxygen species (ROS) and MAPK signaling pathway (3). A recent study has shown that excess cholesterol drives proliferation of intestinal stem cells both in vitro and in vivo, and upregulated cholesterol biosynthesis promotes intestinal tumorigenesis in murine models (4). Injection of HDL mimetic peptides has been shown to inhibit tumor development in both induced and spontaneous mouse models of colon cancer (5). Finally, experimentally reducing serum triglyceride levels via upregulating lipoprotein lipase activity has been found to concurrently suppress intestinal polyp formation in murine models (6, 7), implying a carcinogenic effect of triglyceride.

Despite compelling experimental evidence, previous epidemiologic studies have reported conflicting findings on associations of blood total cholesterol, LDL-C, HDL-C and triglyceride with colorectal cancer risk (8–15). Several factors may contribute to the inconsistencies, including relatively small study sizes and reverse causation (i.e., change in blood lipid levels induced by the colorectal cancer development). The number of colorectal cancer cases in prior studies ranged from 102 to 4,984, with most including fewer than 500 cases (8, 9). Besides, most of previous studies did not consider the effect of regular use of cholesterol-lowering medications, which lowers blood total cholesterol and LDC-C. Including users of cholesterol-lowering medications in the analyses of lipids–colorectal cancer associations could introduce bias as lipid levels measured at the time of blood collection do not reflect long-term exposure to these lipids. This bias could be substantial in studies conducted in the United States (US) and other developed countries where prevalence of cholesterol-lowering medication use is high. To disentangle the role of blood lipids in the etiology of colorectal cancer, we investigated the associations between serum lipids and colorectal cancer risk in UK Biobank, a prospective cohort study, after addressing the aforementioned issues.

Study design and population

The UK Biobank is a population-based, prospective cohort study in which over 502,000 adults aged 40 to 70 years were recruited in 22 assessment centers across England, Scotland, and Wales between 2006 and 2010. The study design and methods have been previously described (16). Briefly, at study enrollment, information on sociodemographic characteristics, health and medical history, and lifestyle factors was collected from participants using both a self-administered touchscreen questionnaires and interviews. At the baseline visit, each participant was asked a question “Do you regularly take any of the following medications” via a touchscreen questionnaire, one of the choices being cholesterol-lowering medications. Cholesterol-lowering medication use was categorized as regular users versus non-regular users at baseline. Participants also provided blood samples and underwent physical measurements at baseline visits. For the current study (Supplementary Fig. S1), we excluded participants with prevalent cancers except nonmelanoma skin cancer (n = 41,071), and with missing information on any lipid measurements at baseline (n = 67,131) and/or on cholesterol-lowering medication use at baseline (n = 3,497). We also excluded participants with fewer than 2 years of follow-up time (n = 5,852) to reduce potential influence of reverse causation on our study results. All participants provided written informed consent and the study was approved by the North West Multi-Center Research Ethics Committee. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Assessment of serum lipid profiles

At baseline, blood samples were collected from UK Biobank participants and were transported overnight by commercial courier to a central laboratory, where they were processed and stored at ultralow temperature (17). Details about serum biomarker measurement and assay performance have been described elsewhere (18). Briefly, serum lipid levels, including total cholesterol, LDL-C, HDL-C, and triglyceride were measured on the Beckman Coulter AU5800 analytical platform (Beckman Coulter Inc., Brea, CA) by enzymatic method for total cholesterol and triglyceride, by enzymatic selective protection method for LDL-C, and by enzyme immune-inhibition method for HDL-C. The measurement unit for the four lipid levels was millimoles per liter (mmol/L). To assess assay performance, third-party internal quality control samples of known high, medium and low concentrations were used to compute the coefficients of variation for each lipid. The average within-laboratory (total) coefficients of variation (%) across the three concentrations for total cholesterol, LDL-C, HDL-C, and triglyceride were 1.41–1.78, 1.57–1.71, 1.72–1.81, 2.05–2.27, respectively.

Ascertainment of outcome

Cancer cases in the UK Biobank were identified via linkage to national cancer registries. Information on incident cancer diagnoses were obtained from the National Health Service (NHS) Information Center for participants from England and Wales (follow-up through July 31, 2019) and from the NHS Central Register Scotland for participants from Scotland (follow-up through October 31, 2015). The primary outcome was defined as the first diagnosis of colorectal cancer based on International Classification of Disease, 10th Revision (ICD-10) codes C18 (colon cancer) and C20 (rectal cancer).

Statistical analysis

Baseline characteristics were described using median and interquartile range (IQR) for continuous variables and proportions for categorical variables. We employed Wilcoxon rank-sum test to assess differences between cases and non-cases for continuous variables, and Pearson χ2 test for categorical variables. Serum lipid levels were similarly described using median and IQR, and difference in lipid levels between users and nonusers of cholesterol-lowering medications were assessed by Wilcoxon rank-sum test. Participants contributed person-time to the study from 2 years after the date of blood draw until the date of first colorectal cancer diagnosis, diagnosis of primary cancers of other sites, death, loss to follow-up, or the end of the study (July 31, 2019 or October 31, 2015), whichever occurred first.

HRs and 95% confidence intervals (CI) for the associations of the four lipids with colorectal cancer risk were estimated with the Cox proportional hazard models using age as the time scale. The Cox models were stratified by educational attainment (college/university degree, some professional qualifications, secondary education, none of the above) and tobacco smoking [never, former/current (light), current (heavy)] because they failed to meet proportional hazards assumption. We used Schoenfeld residuals to examine the proportional hazards assumption. The association was examined in all subjects, and separately in nonusers and users of cholesterol-lowering medications. The four lipids were categorized into four groups based on clinically meaningful cut-off points in the regression analyses. These cut-off points were chosen based on how they were classified by the National Cholesterol Education Program Adult Treatment Panel III (19). Covariates included in the regression models were age at enrollment (continuous), sex (female, male), race (White, non-White), and fasting time (<3 hours, 3–5 hours, >5 hours), alcohol consumption (never/seldom, 1–4 times per week, > 4 times per week), fruit and vegetables intake (≤1 serving/day, 2–4 servings/day, ≥5 servings/day), processed and red meat intake (≤4 times/week, 5–7 times/week, ≥8 times/week), dietary fiber intake (< 11.3 g/day, 11.3–16.1 g/day, >16.1 g/day), baseline self-reported aspirin use (yes, no), family history of colorectal cancer (yes, no), and screening history of colorectal cancer (ever, never). Dietary variables included in the analyses were ascertained via the touchscreen questionnaire administrated at study baseline. Dietary fiber intake was calculated from questions on fruit, vegetables, bread and breakfast cereal described previously (20). We did not adjust for adiposity-related traits [i.e., body mass index (BMI), waist-to-hip ratio, physical activity] as they likely contribute to colorectal cancer development via perturbing blood lipid homeostasis. The linear trend test was performed by using the median of each lipid group as an ordinal variable. A small proportion of participants (<5%) had missing values for certain covariates, which were replaced with either the median for continuous variables or the mode for categorical variables. Given the observed linear trend of triglyceride levels with colorectal cancer risk, restricted cubic splines with 5 knots were incorporated into the Cox models to explore the relationship of triglyceride with colorectal cancer risk among the whole cohort and non–cholesterol-lowering medication users only (Supplementary Fig. S2), using the R package rms (21). We also modeled triglyceride levels continuously (per SD increment).

We conducted stratified analyses by sex, tumor site (colon, rectum), and age at colorectal cancer diagnosis (<60 years, ≥60 years). We chose 60 years as the age cut-off point because very few cases were diagnosed under 50 years of age in this population. Likelihood ratio test was used to examine the potential heterogeneous effect between males and females. The heterogeneity by tumor site and age at colorectal cancer diagnosis were evaluated by testing the statistical significance of an interaction term between the exposure and the indicators of tumor site or diagnosis age category in the joint Cox model using a global Wald test (22, 23). All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. All analyses were performed in Stata version 14 (StataCorp, RRID:SCR_012763) and R version 4.2.1 (https://www.r-project.org/, RRID:SCR_001905).

Data availability

Data used in this project can be obtained directly from the UK Biobank by submitting a data request proposal.

A total of 3,150 incident primary colorectal cancer cases among 384,862 UK Biobank participants were identified during the follow-up with the median of 8.2 years. Compared with non-cases, colorectal cancer cases were older, had a higher BMI and waist-to-hip ratio, consume more processed and red meat, drink alcohol more frequently, had longer fasting time before blood draw, more likely to be male, white, have a family history of colorectal cancer, have diabetes at the baseline, have received colorectal cancer screening, and use aspirin and cholesterol-lowering mediations. They were less likely to be never-smokers and attain college/university degrees (Table 1). colorectal cancer cases and non-cases did not differ significantly by the amount of physical activity, fruit/vegetables intake, and dietary fiber intake.

Table 1.

Selected baseline characteristics of study participants by study groups in the UK Biobank.

Colorectal cancer
CharacteristicsAll subjects (N = 384,862)Cases (N = 3,150)Non-cases (N = 381,712)Pa
Age at enrollment (year), median (IQR) 57 (50–63) 62 (56–66) 57 (50–63) <0.001 
 Male sex, % 46.9 56.5 46.8 <0.001 
 White race, % 94.2 96.3 94.2 <0.001 
 Education, %    <0.001 
 College/University degree 32.6 29.0 32.7  
 Some professional qualifications 11.8 13.3 11.7  
 Secondary education 39.0 37.4 39.0  
 None of the above 16.6 20.3 16.6  
 BMI (kg/m2), median (IQR) 26.8 (24.2–29.9) 27.4 (24.9–30.4) 26.8 (24.2–29.9) <0.001 
 Waist-to-hip ratio, median (IQR) 0.87 (0.80–0.94) 0.91 (0.83–0.96) 0.87 (0.80–0.94) <0.001 
Physical activityb, %    0.15 
 None 17.1 18.2 17.1  
 Intermediate 69.6 69.3 69.6  
 Intense 13.3 12.5 13.3  
Processed/red meat intake, %    <0.001 
 ≤4 times per week 29.6 25.0 29.6  
 5–7 times per week 51.5 52.7 51.6  
 ≥8 times per week 18.9 22.3 18.8  
Fruit/vegetables intake, %    0.76 
 ≤1 serving per day 4.1 4.3 4.1  
 2–4 servings per day 54.3 54.7 54.3  
 ≥5 servings per day 41.6 41.0 41.6  
Dietary fiber, %    0.85 
 Low (<11.3 g/day) 33.3 32.9 33.4  
 Intermediate (11.3–16.1 g/day) 33.4 33.8 33.3  
 High (>16.1 g/day) 33.3 33.3 33.3  
Alcohol consumptionc, %    <0.001 
 Never or seldom 30.4 27.2 30.4  
 1–4 times per week 49.3 46.9 49.3  
 Almost daily 20.3 25.9 20.3  
Tobacco smokingd, %    <0.001 
 Never 55.3 47.4 55.4  
 Former or current, light 37.4 45.3 37.4  
 Current, heavy 7.3 7.3 7.2  
 Family history of colorectal cancer, % 10.7 15.1 10.7 <0.001 
 Ever received colorectal cancer screening, % 29.7 35.6 29.6 <0.001 
 Diabetes diagnosis, % 5.2 7.8 5.1 <0.001 
 Self-reported aspirin use, % 13.8 17.9 13.8 <0.001 
 Self-reported regular use of cholesterol-lowering medications, % 17.3 25.2 17.2 <0.001 
 Fasting time (hour), %    0.003 
 <3 hours 25.7 23.3 25.7  
 3–5 hours 63.0 64.2 63.0  
 >5 hours 11.3 12.5 11.3  
Colorectal cancer
CharacteristicsAll subjects (N = 384,862)Cases (N = 3,150)Non-cases (N = 381,712)Pa
Age at enrollment (year), median (IQR) 57 (50–63) 62 (56–66) 57 (50–63) <0.001 
 Male sex, % 46.9 56.5 46.8 <0.001 
 White race, % 94.2 96.3 94.2 <0.001 
 Education, %    <0.001 
 College/University degree 32.6 29.0 32.7  
 Some professional qualifications 11.8 13.3 11.7  
 Secondary education 39.0 37.4 39.0  
 None of the above 16.6 20.3 16.6  
 BMI (kg/m2), median (IQR) 26.8 (24.2–29.9) 27.4 (24.9–30.4) 26.8 (24.2–29.9) <0.001 
 Waist-to-hip ratio, median (IQR) 0.87 (0.80–0.94) 0.91 (0.83–0.96) 0.87 (0.80–0.94) <0.001 
Physical activityb, %    0.15 
 None 17.1 18.2 17.1  
 Intermediate 69.6 69.3 69.6  
 Intense 13.3 12.5 13.3  
Processed/red meat intake, %    <0.001 
 ≤4 times per week 29.6 25.0 29.6  
 5–7 times per week 51.5 52.7 51.6  
 ≥8 times per week 18.9 22.3 18.8  
Fruit/vegetables intake, %    0.76 
 ≤1 serving per day 4.1 4.3 4.1  
 2–4 servings per day 54.3 54.7 54.3  
 ≥5 servings per day 41.6 41.0 41.6  
Dietary fiber, %    0.85 
 Low (<11.3 g/day) 33.3 32.9 33.4  
 Intermediate (11.3–16.1 g/day) 33.4 33.8 33.3  
 High (>16.1 g/day) 33.3 33.3 33.3  
Alcohol consumptionc, %    <0.001 
 Never or seldom 30.4 27.2 30.4  
 1–4 times per week 49.3 46.9 49.3  
 Almost daily 20.3 25.9 20.3  
Tobacco smokingd, %    <0.001 
 Never 55.3 47.4 55.4  
 Former or current, light 37.4 45.3 37.4  
 Current, heavy 7.3 7.3 7.2  
 Family history of colorectal cancer, % 10.7 15.1 10.7 <0.001 
 Ever received colorectal cancer screening, % 29.7 35.6 29.6 <0.001 
 Diabetes diagnosis, % 5.2 7.8 5.1 <0.001 
 Self-reported aspirin use, % 13.8 17.9 13.8 <0.001 
 Self-reported regular use of cholesterol-lowering medications, % 17.3 25.2 17.2 <0.001 
 Fasting time (hour), %    0.003 
 <3 hours 25.7 23.3 25.7  
 3–5 hours 63.0 64.2 63.0  
 >5 hours 11.3 12.5 11.3  

aP values were determined from Wilcoxon rank-sum test to compare differences for continuous variables and Pearson χ2 test for categorical variables between cases and non-cases.

bIntense physical activity is defined as having 6 to 7 days/week of moderate activity and 3 to 5 days/week of vigorous activity, or having 6 to 7 days/week of vigorous activity; intermediate physical activity is defined as having 1 to 5 days/week of moderate activity, or 1 to 2 days/week of vigorous activity.

cSeldom alcohol consumption is defined as special occasion only, or 1 to 3 times a month.

dLight current smoking is defined as smoking <20 pack-years; heavy current smoking is defined as smoking ≥20 pack-year.

The levels of total cholesterol, LDL-C, and HDL-C were lower in cholesterol-lowering medication users than nonusers, whereas that of triglyceride was higher in the users (Table 2). Compared with the nonusers, cholesterol-lowering medication users have 15% increased colorectal cancer risk (HR = 1.15; 95% CI, 1.04–1.26, Table 3). We observed a similar association among men and women, and among cancer cases of the colon and rectum. A stronger association was seen among those who were diagnosed with colorectal cancer at younger than 60 years of age (HR = 1.45; 95% CI, 1.07–1.96; Pheterogeneity < 0.001).

Table 2.

Serum lipid levels of study participants and by cholesterol-lowering medication use status in UK Biobank.

All subjectsNonusersMedication users
Serum lipida (mmol/L)N = 384,862N = 318,456N = 66,406% DifferencebPc
Total cholesterol 5.65 (4.91–6.42) 5.83 (5.16–6.55) 4.65 (4.06–5.30) −20.2% <0.001 
LDL cholesterol 3.52 (2.95–4.12) 3.67 (3.15–4.22) 2.74 (2.33–3.22) −25.3% <0.001 
HDL cholesterol 1.40 (1.17–1.67) 1.43 (1.20–1.70) 1.25 (1.06–1.50) −12.6% <0.001 
Triglyceride 1.48 (1.04–2.14) 1.44 (1.02–2.08) 1.70 (1.20–2.42) 18.1% <0.001 
All subjectsNonusersMedication users
Serum lipida (mmol/L)N = 384,862N = 318,456N = 66,406% DifferencebPc
Total cholesterol 5.65 (4.91–6.42) 5.83 (5.16–6.55) 4.65 (4.06–5.30) −20.2% <0.001 
LDL cholesterol 3.52 (2.95–4.12) 3.67 (3.15–4.22) 2.74 (2.33–3.22) −25.3% <0.001 
HDL cholesterol 1.40 (1.17–1.67) 1.43 (1.20–1.70) 1.25 (1.06–1.50) −12.6% <0.001 
Triglyceride 1.48 (1.04–2.14) 1.44 (1.02–2.08) 1.70 (1.20–2.42) 18.1% <0.001 

aLipid levels were presented as median (IQR).

b% difference for each lipid was calculated using the formula: 100% × (median levelmedication users – median levelnonusers)/median levelnonusers.

cP values were determined from Wilcoxon rank-sum test comparing differences in lipid levels between medication users and nonusers.

Table 3.

Associations of cholesterol-lowering medication use with colorectal cancer risk in UK Biobank, stratified by sex, cancer subsite, and age at diagnosis.

Age-adjustedMultivariable-adjusted
EventsParticipantsHR (95% CI)HRa (95% CI)Pheterogeneityb
All subjects     – 
 Nonusers 2,355 318,456 1 (ref.) 1 (ref.)  
 Medication users 795 66,406 1.18 (1.08–1.28)*** 1.15 (1.04–1.26)**  
Sex     0.79 
 Female      
 Nonusers 1,127 178,852 1 (ref.) 1 (ref.)  
 Medication users 244 25,579 1.10 (0.95–1.26) 1.12 (0.97–1.31)  
 Male      
 Nonusers 1,228 139,604 1 (ref.) 1 (ref.)  
 Medication users 551 40,827 1.10 (0.99–1.22) 1.15 (1.03–1.29)*  
Cancer subsite     0.96 
 Colon      
 Nonusers 1,655 318,456 1 (ref.) 1 (ref.)  
 Medication users 558 66,406 1.16 (1.05–1.28)** 1.12 (1.01–1.25)*  
 Rectum      
 Nonusers 700 318,456 1 (ref.) 1 (ref.)  
 Medication users 237 66,406 1.22 (1.05–1.43)* 1.20 (1.01–1.42)*  
Age at diagnosis      
 <60 years     <0.001 
 Nonusers 473 318,456 1 (ref.) 1 (ref.)  
 Medication users 54 66,406 1.41 (1.06–1.87)* 1.45 (1.07–1.96)*  
 ≥60 years      
 Nonusers 1,882 318,456 1 (ref.) 1 (ref.)  
 Medication users 741 66,406 1.16 (1.06–1.26)** 1.12 (1.02–1.23)*  
Age-adjustedMultivariable-adjusted
EventsParticipantsHR (95% CI)HRa (95% CI)Pheterogeneityb
All subjects     – 
 Nonusers 2,355 318,456 1 (ref.) 1 (ref.)  
 Medication users 795 66,406 1.18 (1.08–1.28)*** 1.15 (1.04–1.26)**  
Sex     0.79 
 Female      
 Nonusers 1,127 178,852 1 (ref.) 1 (ref.)  
 Medication users 244 25,579 1.10 (0.95–1.26) 1.12 (0.97–1.31)  
 Male      
 Nonusers 1,228 139,604 1 (ref.) 1 (ref.)  
 Medication users 551 40,827 1.10 (0.99–1.22) 1.15 (1.03–1.29)*  
Cancer subsite     0.96 
 Colon      
 Nonusers 1,655 318,456 1 (ref.) 1 (ref.)  
 Medication users 558 66,406 1.16 (1.05–1.28)** 1.12 (1.01–1.25)*  
 Rectum      
 Nonusers 700 318,456 1 (ref.) 1 (ref.)  
 Medication users 237 66,406 1.22 (1.05–1.43)* 1.20 (1.01–1.42)*  
Age at diagnosis      
 <60 years     <0.001 
 Nonusers 473 318,456 1 (ref.) 1 (ref.)  
 Medication users 54 66,406 1.41 (1.06–1.87)* 1.45 (1.07–1.96)*  
 ≥60 years      
 Nonusers 1,882 318,456 1 (ref.) 1 (ref.)  
 Medication users 741 66,406 1.16 (1.06–1.26)** 1.12 (1.02–1.23)*  

aCox proportional hazard model using age as the time scale and stratified by educational attainment and tobacco smoking; covariates including age at study enrollment, sex, race, fasting time, alcohol consumption, fruit/vegetables intake, processed/red meat intake, dietary fiber intake, self-reported aspirin use, family history of colorectal cancer, and colorectal cancer screening history.

bPheterogeneity was derived from multivariable-adjusted Cox models. Pheterogeneity for sex was calculated from the likelihood ratio test comparing models with and without an interaction term between the medication use and sex; Pheterogeneity for cancer subsite and age at diagnosis was calculated from a global Wald test for an interaction term between the medication use and the indicators of tumor site or diagnosis age category in the joint Cox model.

*P value < 0.05.

**P value < 0.01.

***P value < 0.001.

A high triglyceride level was associated with an increased risk of colorectal cancer (HR = 1.07; 95% CI, 1.04–1.11 per SD increment; Ptrend < 0.001), while A high HDL-C level was inversely associated with colorectal cancer risk (Ptrend = 0.004; Table 4). No significant association of colorectal cancer risk was found with high levels of either total cholesterol or LDL-C. Among nonusers of cholesterol-lowering medications, however, participants with a total cholesterol level > 6.7 mmol/L were at an elevated risk of colorectal cancer (HR = 1.11; 95% CI, 1.00–1.24) than those with a total cholesterol level 5.2 to 6.2 mmol/L. Similarly, participants with a LDL-C level > 4.1 mmol/L had an elevated risk of colorectal cancer (HR = 1.11; 95% CI, 0.99–1.23) compared with those with a LDL-C level 2.6–3.4 mmol/L. A closer examination of this group showed that participants with a LDL-C level 4.1 to 4.9 mmol/L (HR = 1.12; 95% CI, 0.99–1.25) or > 4.9 mmol/L (HR = 1.08; 95% CI, 0.92–1.26) showed a nonsignificant elevated risk of colorectal cancer than the referent. Similar associations were found for total cholesterol and LDL-C in stratified analyses by sex and cancer sites (Supplementary Table S1), although most of the associations became nonsignificant due to smaller sample sizes. The associations of both lipids with colorectal cancer risk differed by age at diagnosis (Pheterogeneity < 0.001). Among users of cholesterol-lowering medications, both total cholesterol and LDL-C levels were inversely associated with colorectal cancer risk, although the trend test did not reach statistical significance after adjusting for confounders (Supplementary Table S2). Results for HDL-C and triglyceride among medication users were largely consistent with those observed among nonusers of cholesterol-lowering medications.

Table 4.

Associations of serum lipids with colorectal cancer risk in all subjects and nonusers of cholesterol-lowering medications in UK Biobank.

All subjectsSubjects, without use of cholesterol-lowering medications
Age-adjustedMultivariable-adjustedbAge-adjustedMultivariable-adjustedb
Serum lipidsaNo. of events (%)HR (95% CI)HR (95% CI)No. of events (%)HR (95% CI)HR (95% CI)
Total cholesterol 
 1 (low) 1,120 (35.6) 1.13 (1.04–1.24)** 1.10 (1.01–1.20)* 525 (22.3) 1.06 (0.95–1.18) 1.04 (0.93–1.16) 
 2 1,002 (31.8) 1 (ref.) 1 (ref.) 845 (35.9) 1 (ref.) 1 (ref.) 
 3 436 (13.8) 0.98 (0.88–1.10) 1.00 (0.89–1.12) 413 (17.5) 1.00 (0.89–1.13) 1.02 (0.91–1.15) 
 4 (high) 592 (18.8) 1.04 (0.94–1.15) 1.08 (0.98–1.20) 572 (24.3) 1.07 (0.96–1.19) 1.11 (1.00–1.24) 
Ptrendc  0.03 0.52  0.82 0.24 
LDL cholesterol 
 1 (low) 504 (16.0) 1.19 (1.07–1.33)** 1.17 (1.05–1.31)** 159 (6.7) 1.24 (1.04–1.47)* 1.24 (1.04–1.48)* 
 2 917 (29.1) 1 (ref.) 1 (ref.) 590 (25.1) 1 (ref.) 1 (ref.) 
 3 900 (28.6) 0.99 (0.90–1.09) 1.00 (0.91–1.10) 805 (34.2) 1.05 (0.94–1.16) 1.05 (0.94–1.17) 
 4 (high) 829 (26.3) 1.01 (0.91–1.10) 1.03 (0.94–1.13) 801 (34.0) 1.09 (0.98–1.21) 1.11 (0.99–1.23) 
Ptrendc  0.02 0.17  0.78 0.58 
HDL cholesterol 
 1 (low) 650 (20.6) 1 (ref.) 1 (ref.) 408 (17.3) 1 (ref.) 1 (ref.) 
 2 1,375 (43.7) 1.01 (0.92–1.11) 0.89 (0.80–0.98)* 978 (41.5) 1.03 (0.92–1.16) 0.89 (0.79–1.01) 
 3 667 (21.2) 0.86 (0.77–0.95)** 0.86 (0.77–0.96)** 559 (23.8) 0.91 (0.80–1.03) 0.90 (0.79–1.02) 
 4 (high) 458 (14.5) 0.80 (0.71–0.90)*** 0.83 (0.73–0.94)** 410 (17.4) 0.87 (0.76–1.00)* 0.88 (0.77–1.02) 
Ptrendc  <0.001 0.004  0.003 0.17 
Triglyceride 
 1 (low) 772 (24.5) 1 (ref.) 1 (ref.) 624 (26.5) 1 (ref.) 1 (ref.) 
 2 927 (29.4) 1.08 (0.98–1.19) 1.06 (0.96–1.16) 714 (30.3) 1.08 (0.97–1.20) 1.06 (0.96–1.19) 
 3 575 (18.3) 1.17 (1.05–1.31)** 1.13 (1.01–1.25)* 404 (17.2) 1.11 (0.98–1.25) 1.06 (0.94–1.21) 
 4 (high) 876 (27.8) 1.28 (1.17–1.42)*** 1.19 (1.07–1.31)** 613 (26.0) 1.26 (1.13–1.41)*** 1.17 (1.05–1.32)** 
Ptrendc  <0.001 <0.001  < 0.001 0.007 
 Per SD 3,150 (100) 1.11 (1.07–1.14)*** 1.07 (1.04–1.11)*** 2,355 (100) 1.11 (1.06–1.15)*** 1.07 (1.03–1.12)*** 
All subjectsSubjects, without use of cholesterol-lowering medications
Age-adjustedMultivariable-adjustedbAge-adjustedMultivariable-adjustedb
Serum lipidsaNo. of events (%)HR (95% CI)HR (95% CI)No. of events (%)HR (95% CI)HR (95% CI)
Total cholesterol 
 1 (low) 1,120 (35.6) 1.13 (1.04–1.24)** 1.10 (1.01–1.20)* 525 (22.3) 1.06 (0.95–1.18) 1.04 (0.93–1.16) 
 2 1,002 (31.8) 1 (ref.) 1 (ref.) 845 (35.9) 1 (ref.) 1 (ref.) 
 3 436 (13.8) 0.98 (0.88–1.10) 1.00 (0.89–1.12) 413 (17.5) 1.00 (0.89–1.13) 1.02 (0.91–1.15) 
 4 (high) 592 (18.8) 1.04 (0.94–1.15) 1.08 (0.98–1.20) 572 (24.3) 1.07 (0.96–1.19) 1.11 (1.00–1.24) 
Ptrendc  0.03 0.52  0.82 0.24 
LDL cholesterol 
 1 (low) 504 (16.0) 1.19 (1.07–1.33)** 1.17 (1.05–1.31)** 159 (6.7) 1.24 (1.04–1.47)* 1.24 (1.04–1.48)* 
 2 917 (29.1) 1 (ref.) 1 (ref.) 590 (25.1) 1 (ref.) 1 (ref.) 
 3 900 (28.6) 0.99 (0.90–1.09) 1.00 (0.91–1.10) 805 (34.2) 1.05 (0.94–1.16) 1.05 (0.94–1.17) 
 4 (high) 829 (26.3) 1.01 (0.91–1.10) 1.03 (0.94–1.13) 801 (34.0) 1.09 (0.98–1.21) 1.11 (0.99–1.23) 
Ptrendc  0.02 0.17  0.78 0.58 
HDL cholesterol 
 1 (low) 650 (20.6) 1 (ref.) 1 (ref.) 408 (17.3) 1 (ref.) 1 (ref.) 
 2 1,375 (43.7) 1.01 (0.92–1.11) 0.89 (0.80–0.98)* 978 (41.5) 1.03 (0.92–1.16) 0.89 (0.79–1.01) 
 3 667 (21.2) 0.86 (0.77–0.95)** 0.86 (0.77–0.96)** 559 (23.8) 0.91 (0.80–1.03) 0.90 (0.79–1.02) 
 4 (high) 458 (14.5) 0.80 (0.71–0.90)*** 0.83 (0.73–0.94)** 410 (17.4) 0.87 (0.76–1.00)* 0.88 (0.77–1.02) 
Ptrendc  <0.001 0.004  0.003 0.17 
Triglyceride 
 1 (low) 772 (24.5) 1 (ref.) 1 (ref.) 624 (26.5) 1 (ref.) 1 (ref.) 
 2 927 (29.4) 1.08 (0.98–1.19) 1.06 (0.96–1.16) 714 (30.3) 1.08 (0.97–1.20) 1.06 (0.96–1.19) 
 3 575 (18.3) 1.17 (1.05–1.31)** 1.13 (1.01–1.25)* 404 (17.2) 1.11 (0.98–1.25) 1.06 (0.94–1.21) 
 4 (high) 876 (27.8) 1.28 (1.17–1.42)*** 1.19 (1.07–1.31)** 613 (26.0) 1.26 (1.13–1.41)*** 1.17 (1.05–1.32)** 
Ptrendc  <0.001 <0.001  < 0.001 0.007 
 Per SD 3,150 (100) 1.11 (1.07–1.14)*** 1.07 (1.04–1.11)*** 2,355 (100) 1.11 (1.06–1.15)*** 1.07 (1.03–1.12)*** 

aCategories for serum lipids are defined as the following:

total cholesterol: category 1: <5.2 mmol/L (200 mg/dL); category 2: 5.2–6.2 mmol/L (200–240 mg/dL); category 3: 6.2–6.7 mmol/L (240–260 mg/dL); category 4: >6.7 mmol/L (260 mg/dL).

LDL cholesterol: category 1: <2.6 mmol/L (100 mg/dL); category 2: 2.6–3.4 mmol/L (100–130 mg/dL); category 3: 3.4–4.1 mmol/L (130–160 mg/dL); category 4: >4.1 mmol/L (160 mg/dL).

HDL cholesterol: category 1: <1.0 mmol/L (40 mg/dL) for males or 1.3 mmol/L (50 mg/dL) for females; category 2: 1.0–1.5 mmol/L (40–60 mg/dL) for males or 1.3–1.5 mmol/L (50–60 mg/dL) for females; category 3: 1.5–1.8 mmol/L (60–70 mg/dL); category 4: >1.8 mmol/L (70 mg/dL).

Triglyceride: category 1: <1.1 mmol/L (100 mg/dL); category 2: 1.1–1.7 mmol/L (100–150 mg/dL); category 3: 1.7–2.2 mmol/L (150–200 mg/dL); category 4: >2.2 mmol/L (>200 mg/dL).

bCox proportional hazard model using age as the time scale and stratified by educational attainment and tobacco smoking; covariates including age at study enrollment, sex, race, fasting time, alcohol consumption, fruit/vegetables intake, processed/red meat intake, dietary fiber intake, self-reported aspirin use, family history of colorectal cancer, and colorectal cancer screening history.

cThe linear trend test was performed by using the median of each lipid category as an ordinal variable.

*P value < 0.05.

**P value < 0.01.

***P value < 0.001.

In this large prospective study of 384,862 UK Biobank participants, cholesterol-lowering medication users showed an elevated colorectal cancer risk than nonusers. Among nonusers of cholesterol-lowering medications, high levels of total cholesterol and LDL-C were modestly associated with an increased colorectal cancer risk. Serum levels of triglycerides and HDL-C were positively and inversely associated with colorectal cancer risk, respectively. These findings are supported by previous animal experiments and several epidemiologic studies, suggesting a possible role of blood lipids in the etiology of colorectal cancer.

Current evidence on the association between blood lipid levels and colorectal cancer risk is inconsistent. An early meta-analysis of 17 prospective cohort studies have reported 1.11- and 1.18-fold increased risk for serum total cholesterol and triglyceride levels, whereas the association of HDL-C with colorectal cancer risk was marginally inverse (8). Results from several recent cohort studies have been published, but the findings remain controversial. One Swedish population-based prospective cohort study including 785 colorectal cancer cases reported null associations of serum HDL-C and LDL-C with colorectal cancer risk (10). In an earlier analysis of data from the UK Biobank cohort which did not consider cholesterol-lowering medication use, serum level of triglyceride was found to be positively associated with colorectal cancer risk, whereas total cholesterol and LDL-C levels were not associated with the risk of colorectal cancer (13). The finding for triglyceride is consistent with ours. A recent population-based cohort study leveraging insurance claim data in South Korea and including 1,552 colorectal cancer cases showed that a low serum HDL-C level (< 1.0 mmol/L) was associated with a 1.20-fold higher risk than a HDL-C level of normal range (1.0–1.5 mmol/L), whereas a high LDL-C level (> 4.1 mmol/L) was associated with an nonsignificant 1.22-fold elevated risk than a LDL-C level below 3.4 mmol/L (14). None of these studies, however, considered cholesterol-lowering medication use in their analyses. Three cohort studies in the US, Germany and Sweden did consider cholesterol-lowering medication use by either excluding users from the analyses (11, 15) or adding estimated constants to lipid levels among users (12). In the US prospective study of 198 colorectal cancer cases from Women's Health Study, participants in the highest quartile of HDL-C and triglyceride levels showed significant 0.59-fold decreased and 1.99-fold increased risk than those in the lowest quartile (15). High levels of total cholesterol (HR = 1.23; 95% CI, 0.81–1.86) and LDL-C (HR = 1.17; 95% CI, 0.79–1.73) were associated with increased colorectal cancer risk. In the German case-cohort study based on 256 cases from the EPIC-Heidelberg Study, doubling in concentrations of serum total cholesterol (HR = 1.73; 95% CI, 0.99–3.03) and triglyceride (HR = 1.08; 95% CI, 0.88–1.33) was associated with elevated risk of colorectal cancer (11). The Swedish prospective cohort studies of 1,250 colorectal cancer cases reported that 1-SD increase in total cholesterol and triglyceride was associated with significant 1.08- and 1.12-fold elevated risk (12). Total cholesterol and triglyceride findings in these two studies are largely consistent with our results.

We found a positive association of LDL-C with colorectal cancer risk among nonusers of cholesterol-lowering medications. Interestingly, a significantly elevated risk of colorectal cancer was found among cholesterol-lowering medication users who had a lower LDL-C level than nonusers. Evidently, cholesterol-lowering medication users had an elevated LDL-C level prior to the use of cholesterol-lowering medications. Thus, the positive association between cholesterol-lowering medication use and colorectal cancer risk observed in our study may be due to exposure to a high level of LDL-C prior to the use of cholesterol-lowering medications. If LDL-C is truly positively associated with colorectal cancer risk, including medication users with an abnormally low LDL-C level could bias the association towards the null or even an inverse association. The inconsistencies in the lipid–colorectal cancer associations may be partly explained by the fact that few prior studies have considered cholesterol-lowering medication use in the analyses.

Our results showed that serum triglyceride level was positively associated with colorectal cancer risk among all subjects. The effects of cholesterol-lowering medications on blood triglyceride levels are varied. Fibrates, statins, and ezetimibe have been shown to reduce triglyceride levels, whereas bile acid sequestrants may raise triglyceride levels (24). In our analysis, we found use of cholesterol-lowering medications was associated with increased serum triglyceride levels. Elevated triglyceride levels are not only common characteristics of dyslipidemia related to insulin resistance, but also the central pathophysiologic feature of abnormal lipid profile (25). Insulin could promote tumorigenesis of the colon via its mitogenic effect or hyperinsulinemia-induced inflammation (26). Alternatively, the positive association of serum triglyceride with colorectal cancer risk may be explained by the gut microbiota, which both influences host circulating triglyceride levels (27) and contributes to colorectal cancer development via modulating the inflammatory process, metabolizing dietary components to carcinogens, and producing genotoxins (28).

This study has several unique strengths. To our knowledge, our study is the first to demonstrate that cholesterol-lowering medication use may bias the associations of blood total cholesterol and LDL-C with colorectal cancer risk. Other strengths include a prospective design and long follow-up period. However, this study also has several limitations. Lipid levels were measured only once at baseline and therefore it is impossible to examine time-dependent effects of lipids on colorectal cancer risk. Information on cholesterol-lowering medication use is self-reported and may be subject to response bias. In addition, because only data on cholesterol-lowering medication use status at baseline is available, we were unable to explore how duration, type, and dose of medication use, as well as nonadherence may affect colorectal cancer risk among users. Future investigations on these medication-related effects may help inform chemoprevention strategies for colorectal cancer. Furthermore, because the study population lacks racial diversity, we were unable to examine possible racial differences in colorectal cancer risk in relation to blood lipids. Finally, due to a relatively small number of colorectal cancer cases accrued over the follow-up period, our study may have limited power to detect associations in subgroup analyses.

In conclusion, results from this large, prospective cohort study in the UK Biobank suggest positive associations of serum total cholesterol and LDL-C with colorectal cancer risk. Studies on the lipid–colorectal cancer associations that fail to account for cholesterol-lowering medication use may produce biased results. Future investigations are warranted to replicate our findings in other populations and elucidate the role of blood lipids in the carcinogenesis of colorectum.

X. Shu reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

F. Yuan: Formal analysis, writing–original draft, writing–review and editing. W. Wen: Writing–review and editing. G. Jia: Writing–review and editing. J. Long: Writing–review and editing. X.O. Shu: Writing–review and editing. W. Zheng: Conceptualization, supervision, writing–original draft, writing–review and editing.

This work was supported in part by the Anne Potter Wilson chair endowment at Vanderbilt University (to W. Zheng). We acknowledge all the families and clinicians who contributed to the study.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).

1.
Keum
N
,
Giovannucci
E
.
Global burden of colorectal cancer: emerging trends, risk factors, and prevention strategies
.
Nat Rev Gastroenterol Hepatol
2019
;
16
:
713
32
.
2.
Giovannucci
E
.
Metabolic syndrome, hyperinsulinemia, and colon cancer: a review
.
Am J Clin Nutr
2007
;
86
:
s836
42
.
3.
Wang
C
,
Li
P
,
Xuan
J
,
Zhu
C
,
Liu
J
,
Shan
L
, et al
.
Cholesterol enhances colorectal cancer progression via ROS elevation and MAPK signaling pathway activation
.
Cell Physiol Biochem
2017
;
42
:
729
42
.
4.
Wang
B
,
Rong
X
,
Palladino
END
,
Wang
J
,
Fogelman
AM
,
Martin
MG
, et al
.
Phospholipid remodeling and cholesterol availability regulate intestinal stemness and tumorigenesis
.
Cell Stem Cell
2018
;
22
:
206
20
.
5.
Su
F
,
Grijalva
V
,
Navab
K
,
Ganapathy
E
,
Meriwether
D
,
Imaizumi
S
, et al
.
HDL mimetics inhibit tumor development in both induced and spontaneous mouse models of colon cancer
.
Mol Cancer Ther
2012
;
11
:
1311
9
.
6.
Niho
N
,
Takahashi
M
,
Kitamura
T
,
Shoji
Y
,
Itoh
M
,
Noda
T
, et al
.
Concomitant suppression of hyperlipidemia and intestinal polyp formation in Apc-deficient mice by peroxisome proliferator-activated receptor ligands
.
Cancer Res
2003
;
63
:
6090
5
.
7.
Niho
N
,
Mutoh
M
,
Takahashi
M
,
Tsutsumi
K
,
Sugimura
T
,
Wakabayashi
K
.
Concurrent suppression of hyperlipidemia and intestinal polyp formation by NO-1886, increasing lipoprotein lipase activity in Min mice
.
Proc Natl Acad Sci USA
2005
;
102
:
2970
4
.
8.
Yao
X
,
Tian
Z
.
Dyslipidemia and colorectal cancer risk: a meta-analysis of prospective studies
.
Cancer Causes Control
2015
;
26
:
257
68
.
9.
Tian
Y
,
Wang
K
,
Li
J
,
Wang
J
,
Wang
Z
,
Fan
Y
, et al
.
The association between serum lipids and colorectal neoplasm: a systemic review and meta-analysis
.
Public Health Nutr
2015
;
18
:
3355
70
.
10.
Borgquist
S
,
Butt
T
,
Almgren
P
,
Shiffman
D
,
Stocks
T
,
Orho-Melander
M
, et al
.
Apolipoproteins, lipids and risk of cancer
.
Int J Cancer
2016
;
138
:
2648
56
.
11.
Katzke
VA
,
Sookthai
D
,
Johnson
T
,
Kuhn
T
,
Kaaks
R
.
Blood lipids and lipoproteins in relation to incidence and mortality risks for CVD and cancer in the prospective EPIC-Heidelberg cohort
.
BMC Med
2017
;
15
:
218
.
12.
Myte
R
,
Gylling
B
,
Haggstrom
J
,
Haggstrom
C
,
Zingmark
C
,
Lofgren Burstrom
A
, et al
.
Metabolic factors and the risk of colorectal cancer by KRAS and BRAF mutation status
.
Int J Cancer
2019
;
145
:
327
37
.
13.
Fang
Z
,
He
M
,
Song
M
.
Serum lipid profiles and risk of colorectal cancer: a prospective cohort study in the UK Biobank
.
Br J Cancer
2021
;
124
:
663
70
.
14.
Han
KT
,
Kim
S
.
Do cholesterol levels and continuity of statin use affect colorectal cancer incidence in older adults under 75 years of age?
PLoS One
2021
;
16
:
e0250716
.
15.
Chandler
PD
,
Song
Y
,
Lin
J
,
Zhang
S
,
Sesso
HD
,
Mora
S
, et al
.
Lipid biomarkers and long-term risk of cancer in the Women's Health Study
.
Am J Clin Nutr
2016
;
103
:
1397
407
.
16.
Sudlow
C
,
Gallacher
J
,
Allen
N
,
Beral
V
,
Burton
P
,
Danesh
J
, et al
.
UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age
.
PLoS Med
2015
;
12
:
e1001779
.
17.
Elliott
P
,
Peakman
TC
,
Biobank
UK
.
The UK Biobank sample handling and storage protocol for the collection, processing, and archiving of human blood and urine
.
Int J Epidemiol
2008
;
37
:
234
44
.
18.
UK Biobank biomarker project: companion document to accompany serum biomarker data
.
Available from
: https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf.
19.
Expert Panel on Detection, Evaluation, and Treatment of high blood cholesterol in Adults
.
Executive summary of the third report of the national cholesterol education program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel III)
.
JAMA
2001
;
285
:
2486
97
.
20.
Bradbury
KE
,
Young
HJ
,
Guo
W
,
Key
TJ
.
Dietary assessment in UK Biobank: an evaluation of the performance of the touchscreen dietary questionnaire
.
J Nutr Sci
2018
;
7
:
e6
.
21.
22.
Xue
X
,
Kim
MY
,
Gaudet
MM
,
Park
Y
,
Heo
M
,
Hollenbeck
AR
, et al
.
A comparison of the polytomous logistic regression and joint cox proportional hazards models for evaluating multiple disease subtypes in prospective cohort studies
.
Cancer Epidemiol Biomarkers Prev
2013
;
22
:
275
85
.
23.
Murphy
N
,
Ward
HA
,
Jenab
M
,
Rothwell
JA
,
Boutron-Ruault
MC
,
Carbonnel
F
, et al
.
Heterogeneity of colorectal cancer risk factors by anatomical subsite in 10 European countries: a multinational cohort study
.
Clin Gastroenterol Hepatol
2019
;
17
:
1323
31
.
24.
Miller
M
,
Stone
NJ
,
Ballantyne
C
,
Bittner
V
,
Criqui
MH
,
Ginsberg
HN
, et al
.
Triglycerides and cardiovascular disease: a scientific statement from the American Heart Association
.
Circulation
2011
;
123
:
2292
333
.
25.
Ginsberg
HN
,
Zhang
YL
,
Hernandez-Ono
A
.
Regulation of plasma triglycerides in insulin resistance and diabetes
.
Arch Med Res
2005
;
36
:
232
40
.
26.
Zhang
AMY
,
Wellberg
EA
,
Kopp
JL
,
Johnson
JD
.
Hyperinsulinemia in obesity, inflammation, and cancer
.
Diabetes Metab J
2021
;
45
:
285
311
.
27.
Schoeler
M
,
Caesar
R
.
Dietary lipids, gut microbiota and lipid metabolism
.
Rev Endocr Metab Disord
2019
;
20
:
461
72
.
28.
Wong
SH
,
Yu
J
.
Gut microbiota in colorectal cancer: mechanisms of action and clinical applications
.
Nat Rev Gastroenterol Hepatol
2019
;
16
:
690
704
.