Back to search

The University of Edinburgh

Application of instrumental variables methods for the investigation of non-genetic biomarkers and other risk factors for colorectal cancer risk

Abstract

dc:description.abstract

INTRODUCTION: As the third most common cancer and the second leading cause of cancer death, colorectal cancer is a major public health problem in both sexes. Over 95% of colorectal cancer is colorectal adenocarcinoma, which mainly arises from pre-existing adenomatous polyps. The epidemiological characteristics of colorectal cancer, including high incidence and mortality rate, later stages at diagnosis, and remarkable decreased survival rate by stage, emphasise the importance of understanding the aetiology of colorectal cancer. Based on our knowledge of risk factors for colorectal cancer, early detection and prevention could be achieved through risk prediction, control of risk factors, and detection of pre-malignant adenomas. The current established risk factors for colorectal cancer risk include personal or family history of colorectal cancer or adenomatous polyps, the presence of premalignant polyps, inflammatory bowel disease, and the presence of inherited colorectal cancer syndromes. Factors associated with increased colorectal cancer risk based on existing convincing evidence include high intake of processed meat, alcohol drinking, smoking, physical inactivity, and body fatness. No convincing evidence has been established for associations between any biomarker and colorectal cancer risk. The overall aim of this thesis is to explore the association and causality between nongenetic biomarkers, other putative risk factors, and colorectal cancer risk by applying instrumental variable methods. METHODS: First, I conducted an umbrella review to summarise and classify the existing evidence of reported associations between non-genetic biomarkers and colorectal cancer risk. The umbrella review covered evidence from meta-analyses of observational studies, Mendelian randomization (MR) studies of non‐genetic biomarkers, and meta-analyses of randomized controlled trials (RCTs) on supplementation of micronutrients. A wide range of non-genetic biomarkers related to colorectal cancer were included, and the magnitude, direction, and strength of their effect was assessed and presented. The credibility of the evidence was evaluated and classified into four categories for metaanalyses (i.e., ‘convincing’, ‘highly suggestive’, ‘suggestive’, and ‘weak’), and three categories for MR studies (i.e., ‘evidence of causality’, ‘likely noncausal’, and ‘unknown’). In the next step, an MR study was conducted to establish or refute causality between physical activity and colorectal cancer. Common genetic variants associated with selfreported moderate-to-vigorous physical activity (MVPA), acceleration vector magnitude physical activity (AMPA), and sedentary time were used as instrumental variables. To investigate whether the effect of physical activity on CRC risk is fully or partly mediated by obesity, I performed a multivariable MR where I included instrumental variables for body mass index, body fat percentage, waist circumference, and arm, trunk and leg fat ratios in addition to IVs for physical activity. The effect of these instrumental variables was analysed in a colorectal cancer genome-wide association study comprising 31,197 cases and 61,770 controls from European ancestry, by applying two-sample and multivariable MR study designs. Finally, two phenome-wide association studies (PheWASs) were conducted to explore phenotypes that were associated with colorectal cancer genetic predisposition, leveraging a polygenic risk score (PRS) for colorectal cancer risk. The polygenic risk score was created by adding the weighted dosages of colorectal cancer risk alleles. The targeted phenotypes included all ICD codes based on health outcomes other than colorectal cancer itself, and non-genetic biomarkers such as blood count, blood biochemistry, infectious diseases antigens, and urine biomarkers. In addition to the PheWAS method, I also applied the tree-structured phenotypic model (TreeWAS) and PHEnome Scan ANalysis Tool (PHESANT) methods. The analyses were performed in 334,385 unrelated White British individuals from the UK Biobank cohort, and additional effect estimates were calculated by comparing the case frequency for health outcomes and the mean for biomarkers between the top and bottom quartiles of the CRC PRS. RESULTS: The umbrella review included a total of 72 meta‐analyses of observational studies and 18 MR studies on non‐genetic biomarkers: no meta-analyses of RCTs on blood level biomarkers were found. The umbrella review additionally included six meta‐analyses of RCTs on micronutrient intake. None of the associations was classified as convincing or highly suggestive, three were classified as suggestive (homeostatic model assessment‐insulin resistance [HOMA‐IR; risk ratio (95% confidence interval): RR (95% CI); 1.56(1.22, 1.98)], 25‐hydroxyvitamin D [RR (95% CI): 0.67(0.54, 0.83)], and human papillomavirus [RR (95% CI): 3.52(1.77, 7.00)]), and 26 were classified as weak. Three biomarkers explored in MR studies were classified as causal (linoleic acid [odds ratio (95%CI): OR (95% CI); 0.95 (0.93, 0.97)], arachidonic acid [OR (95% CI): 1.05 (1.03, 1.07)], and oleic acid [OR (95% CI): 0.77 (0.65, 0.92)]) and seven were classified as likely noncausal (low-density lipoproteins cholesterol, omega‐6 polyunsaturated fatty acids, total cholesterol, selenium, vitamin B12, telomere length, and adiponectin). The MR study supported that increased MVPA and AMPA were causally associated with decreased colorectal cancer risk with a causal effect mediated through obesity and a causal effect independent of obesity. The OR (95% CI) per standard deviation increase in MVPA and AMPA was 0.56 (0.31, 1.01) and 0.60 (0.41, 0.88), respectively. No association was found between sedentary time and colorectal cancer risk. The proportion of the effect mediated through body mass index was 2% (95% CI: 0%, 14%) and 32% (95% CI: 12%, 46%) for MVPA and AMPA, respectively. The PheWAS targeting health outcomes detected eight PheWAS phenotypes and 21 TreeWAS nodes to be associated with colorectal cancer predisposition polymorphisms. The associations were detected for neoplasms and diseases in the digestive system. The OR (95% CI) by comparing case frequency between the top and bottom risk quartiles of colorectal cancer PRS were: 1.93 (1.85, 2.01) for benign neoplasm of colon, 2.61 (2.26, 3.02) for benign neoplasm of caecum, 2.78 (2.41, 3.20) for benign neoplasm of ascending colon, 2.72 (2.40, 3.08) for benign neoplasm of transverse colon, 2.68 (2.27, 3.16) for benign neoplasm of descending colon, 2.31 (2.12, 2.53) for benign neoplasm of sigmoid colon, 2.19 (1.91, 2.51) for benign neoplasm of colon unspecified, 2.68 (1.90, 3.80) for benign neoplasm of rectosigmoid junction, 2.55 (2.27, 2.86) for benign neoplasm of rectum, 1.29 (1.25, 1.33) for benign neoplasm of unspecified sites, 1.13 (1.11, 1.16) for other malignant neoplasm, 1.13 (1.10, 1.16) for other disorders of intestine, 1.18 (1.14, 1.22) for diverticular disease, 1.19 (1.14, 1.23) for diverticular disease of large intestine without perforation or abscess, 1.14 (1.06, 1.22) for diverticular disease of intestine part unspecified without perforation or abscess, 1.66 (1.54, 1.78) for anal and rectal polyp, 1.78 (1.65, 1.92) for rectal polyp, and 1.85 (1.75, 1.96) for polyp of colon. The PheWAS targeting biomarkers detected 11 biomarkers associated with colorectal cancer predisposition polymorphisms at false discovery rate q<0.05. The estimated mean differences and the 95% CI between the top risk quartile and the bottom risk quartile of the colorectal cancer PRS were -0.001 (-0.002, -0.001) for platelet crit, - 0.844 (-1.104, -0.584) for alkaline phosphatase, -0.016 (-0.027, -0.004) for lymphocyte count, -0.012 (-0.016, -0.009) for red blood cell (erythrocyte) count, -0.004 (-0.006, - 0.002) for monocyte count, -1.170 (-1.753, -0.587) for platelet count, -1.653 (-2.442, -0.864) for urate, -0.029 (-0.041, -0.017) for hemoglobin concentration, -0.076 (- 0.110, -0.042) for hematocrit percentage, 0.127 (0.046, 0.209) for neutrophill percentage, and -0.038 (-0.058, -0.018) for white blood cell (leukocyte) count. CONCLUSION: This thesis presents a comprehensive exploration of the risk factors of colorectal cancer. The umbrella review summarised the evidence for 87 biomarkers related to colorectal cancer risk and found a gap in the literature with respect to meta‐analyses of RCTs or even individual RCTs on biomarkers of colorectal cancer risk. These biomarkers were only examined in observational studies. The umbrella review only found suggestive associations between three biomarkers and colorectal cancer risk (i.e., HOMA‐IR, 25‐hydroxyvitamin D, and human papillomavirus). The association between 25‐hydroxyvitamin D and colorectal cancer, as well as the association between human papillomavirus and colorectal cancer, have also been tested in the PheWAS analyses, but none of these biomarkers was found to be associated with colorectal cancer risk. Several biomarkers have been classified by the umbrella review under weak or unknown evidence of association with colorectal cancer risk, and no evidence has been found for those biomarkers in the follow-up PheWAS (such as serum calcium, haemoglobin A1C, fasting glucose, insulin-like growth factor 1, Creactive protein, H.pylori, human cytomegalovirus infection, high-density lipoproteins-cholesterol, low-density lipoproteins-cholesterol, total cholesterol, and triglyceride). In summary, the umbrella review did not find biomarkers that are convincingly associated with colorectal cancer. The MR study established a causal effect of both subjectively and objectively measured physical activity on colorectal cancer risk, independent of obesity and body fat distribution. This conclusion emphasised the importance of promoting physical activity in the general population for colorectal cancer prevention. In the PheWAS on other health outcomes, the only convincing associations were those observed between colorectal cancer predisposition polymorphisms and benign or premalignant colorectal neoplasms, as well as neoplasms of unspecified sites. These findings indicate the potential beneficial effects of polypectomy and family history of pre-malignant colorectal neoplasms. An association with diverticular disease was found by the PheWAS analysis targeting health outcomes. To verify the association, a follow-up bi-directional MR study was performed, which supported a causal effect of colorectal cancer on diverticular disease, but not the reverse. However, this association may be due to shared aetiology or biased ascertainment through investigation in people with a higher environmental risk factor loading for both conditions. In the PheWAS on biomarkers, the colorectal cancer polymorphisms were associated with decreased levels of red blood cell count, alkaline phosphatase, platelet crit, lymphocyte count, haematocrit percentage, haemoglobin concentration, monocyte count, urate, white blood cell count, platelet count, and with increased blood levels of neutrophil percentage of the white blood cell. However, combined with existing evidence, no convincing conclusion can be drawn about these biomarkers.

Degree

thesis:*
Grantor dc:publisher
The University of Edinburgh
Year dc:date.issued
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Zhang, Xiaomeng
Advisors dc:contributor.advisor
  • Theodoratou, Evropi
  • Timofeeva, Maria

Subjects

dc:subject × 5

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Edinburgh
Base URL
era.ed.ac.uk/server/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Zhang, Xiaomeng. Application of instrumental variables methods for the investigation of non-genetic biomarkers and other risk factors for colorectal cancer risk. The University of Edinburgh, 2022. https://hdl.handle.net/1842/38703