Melbourne School of Population and Global Health - Theses
Permanent URI for this collection
Search Results
1 - 3 of 3
-
ItemCancer risks for people with colorectal cancer-predisposing genetic mutations( 2013)Background: Colorectal cancer is one of the most frequently diagnosed cancers in the world. On average, 1 in 20 people will be diagnosed with colorectal cancer in their lifetime. However, an individual’s risk of colorectal cancer varies greatly depending on their genetic and environmental factors. One of the strongest risk factors for colorectal cancer is inherited mutations in predisposing genes such as DNA mismatch repair genes (Lynch syndrome), and DNA base excision repair gene, MUTYH. Given the availability and effectiveness of colorectal cancer screening and interventions, we need precise and accurate risks of cancers to reduce the burden of disease by targeting preventive measures to those at highest risk. The overall aim of the thesis was to estimate cancer risks for people with colorectal cancer-predisposing genetic mutations and their relatives. Methods: For the systematic reviews (Chapters 3, 4 and 8), PubMed was used to search for relevant studies. To address the different research questions (Chapters 5, 6, 7, 9, 10 and 11), appropriate study samples were selected from the Colon Cancer Family Registry, an international consortium of 37,041 participants from 10,375 colorectal cancer families recruited from the USA, Canada, Australia and New Zealand. Prospective cohort analysis (Chapter 5), retrospective cohort analyses (Chapter 6, 7 and 11), weighted cohort analysis (Chapter 9) and modified segregation analysis (Chapter 10) were conducted. Results: There was no model that sufficiently covers the known risk factors for colorectal cancer that is suitable for assessment of people from across the full range of risk (Chapter 3). The existing statistical models (MMRpro, PREMM, MMRpredict, and Leiden) had similar discrimination and were quite sensitive and specific in predicting mismatch repair gene mutations (Chapter 4). People with mismatch repair gene mutations without a prior diagnosis of cancer had an increased risk of cancers that are previously known to be associated with Lynch syndrome as well as pancreatic and breast cancers (Chapter 5). There was no evidence of their non-carrier relatives having an increased risk of any cancer, including colorectal cancer (Chapter 5). People with mismatch repair gene mutations with a prior diagnosis of colorectal cancer (Chapter 6) or endometrial cancer (Chapter 7) had substantially elevated risk for a greater range of cancers including breast and prostate cancers. About half of breast cancer tumours in mismatch repair gene mutation carriers showed loss of mismatch repair proteins, indicating a significant likelihood of an underlying germline mutation (Chapter 8). For all mismatch repair gene mutation carriers combined, there was no evidence of an association of colorectal cancer with the 11 genetic variants that are known to be associated with colorectal cancer for the general population (Chapter 9). There was a wide variation of colorectal cancer risk for monoallelic MUTYH mutation carriers depending on their family history, and their risk can be sufficiently high to warrant consideration of more intensive colorectal cancer screening than for the general population (Chapter 10). Risk of colorectal cancer for first-degree relatives of colorectal cancer cases with suspected Lynch syndrome was higher than for first-degree relatives of mismatch repair-proficient colorectal cancer cases (Chapter 11). A greater risk of colorectal cancer was estimated for first-degree relatives if colorectal cancer cases were diagnosed before age 50 years, had proximal colon cancer, or if their tumours had expanding tumour margin, peritumoural lymphocytes, tumour-infiltrating lymphocytes, or synchronous tumour (Chapter 11). Conclusions: These accurate and precise cancer risk estimates are critically important to guide and inform public health guidelines, such as for cancer screening; and clinical management of cancer families. The studies from this thesis are fundamental for the development of a new comprehensive model for colorectal cancer risk prediction that takes into account all risk factors for colorectal cancer and is suitable for assessment of people from across the full range of risk.
-
ItemThe impact of missing data in a longitudinal cohort study on the risks of colorectal cancer and mortality associated with change in body size( 2013)Background: Obesity is considered a global epidemic. Overweight and obesity, measured by attained weight and Waist Circumference (WC), is associated with an increased risk of Colorectal Cancer (CRC) and mortality. Measuring change in weight or WC accounts for changes in body composition (i.e. increased abdominal obesity and decreased lean muscle mass) associated with ageing, and may be a better predictor of the risks of CRC and/or mortality. Body weight, WC and hips circumference were measured in all participants of the Melbourne Collaborative Cohort Study (MCCS) at baseline, but missing data at follow-up (wave 2) complicates the analysis of changing body composition. This thesis will determine the best statistical approach for dealing with the missing WC data in the MCCS and estimates the risks of CRC and mortality associated with change in body composition. Methods: The literature was reviewed to determine how other large cohort studies have reported and dealt with missing data in the analysis of repeated exposures. The two most common statistical methods for handling missing data identified in the literature review were compared in a comprehensive simulation study. Next, I estimated the associations between change in body composition and the risks of CRC and mortality using the MCCS data and adopting the method for handling missing data that was the least biased in the simulation study. Results: Complete-case analysis was the most common method used to handle missing data in analyses with repeated exposures. Of the more complex approaches, multiple imputation was the most frequently used method. Comparing multiple imputation with complete-case analysis in an extensive simulation study showed that the estimates obtained from both complete-case analysis and multiple imputation were similar to the “true” values (i.e. very little bias was observed). As well, multiple imputation only showed gains in the precision of the estimates when an auxiliary variable was included in the imputation model that was not already included in the analysis model of interest. I used complete-case analysis to handle the missing anthropometric data at wave 2 for the MCCS participants. I found that a 5 cm increase in WC was associated with a slightly increased risk of CRC (Hazard Ratio (HR) = 1.05; 95% Confidence Interval (CI): 0.96, 1.14). For mortality, there was a non-linear association with change in body composition; weight loss (quintile 1: <-2.3 kg) and decreased WC (quintile 1: <1 cm) were associated with an increased risk of all-cause mortality (for weight, HR = 1.80; 1.54, 2.11; for WC, HR = 1.26; 1.09, 1.47) and Cardiovascular Disease mortality (for weight, HR = 2.40; 1.57, 3.65; for WC, HR = 1.39; 0.99, 1.97). This increased risk of all-cause mortality associated with weight loss was of greater magnitude for those older than 55 years at baseline (for age <55 years, HR = 1.30; 0.88, 1.92; for age 55 years, HR = 1.92; 1.61, 2.30). No association was observed for weight loss and death due to obesity-related cancer (for weight, HR = 1.20; 0.71, 2.02; for WC, HR = 1.27; 0.73, 2.23). As well, there was no association between weight gain or increased WC and all-cause or cause-specific mortality. Conclusion: Researchers need to consider the data that is both missing and available for their analyses and decide whether to use complete-case analysis or multiple imputation to handle the missing data. Increased WC increased the risk of CRC, however, weight loss and a decreased WC increased the risk of mortality, especially in older adults. From the results of this observational cohort study of middle-aged adults, weight stability seems to be the recommended option. Of note, the increased risk of mortality associated with weight loss and decreasing WC may be affected by the problem of “reverse-causation”; where an illness on the causal pathway to death is inducing the association between weight loss and mortality.
-
ItemColorectal cancer screening in Australia: what is the role of family history?( 2012)Colorectal cancer (CRC) is the third most common cancer suffered throughout the world, with nearly one million new cases diagnosed world-wide each year and half a million deaths. Due to demographic trends, significantly more individuals will be confronted with CRC in the future. Therefore, its control has become a major public health issue for industrialised countries. Australia has one of the highest CRC incidence rate in the world. It is the second most frequently diagnosed cancer and the second largest cause of cancer deaths in Australia (14,234 new cases and 4,047 deaths in 2007). Approximately 15% to 30% of these cases occur in just the 2% of the population with a strong family history of the disease. The characteristics of colorectal cancer fulfill the conditions that make mass screening useful and beneficial for public health: strong incidence, poor prognosis, a typically long pre-symptomatic latency period, effective treatment in early stages of the disease, it is preceded by a precancerous lesion in 95% of cases and there are testing methods which are scientifically proven to be effective at reducing mortality. Screening for colorectal cancer has been shown to be an effective method to reduce both incidence and mortality, and for several years now, international scientific consensus has recognised the need for setting up mass screening initiatives for colorectal cancer. In 2006 the Australian government introduced a national bowel cancer screening programme (NBCSP) using faecal occult blood test but only for people tuning 50, 55 or 65 years of age. The latest data from the programme published in 2010 showed that only 40% of those invited to screen actually performed the test. The primary aim of my thesis was to investigate the CRC screening practices taking place outside of the current national programme based on the current screening guidelines recommendations. I conducted two systematic reviews of the literature to determine the predictors of screening for people at familial-risk of CRC and the level of screening uptake in that population. I identified a low level of screening participation among people at increased risk of CRC and that family history of CRC and physicians’ recommendation to screen were the most consistent predictors of CRC screening uptake. Overall, I found only a small number of studies investigating the screening practices of those most at risk of CRC, with important methodological shortcomings in their analyses. I investigated CRC screening practices in Australia based on the Colorectal Cancer Family Registry (ACCFR) study. 3845 participants were classified into four risk categories according to the strength of their family history of CRC based on current Australian guidelines. Among participants categorised “at or slightly above” average risk of CRC (represent 98% of the general population) eligible for screening, 90% reported never having been screened and 8% reported over-screening. Middle- aged people, those with a family history of CRC and those with a university degree were more likely to be over-screened. For participants categorised “at moderately increased-risk” or “potentially high-risk” of CRC (represents the 2% of the general population in which 15-30% of all CRC occur), 95% were underscreeners and only 5% reported guideline-defined “appropriate” screening. People of middle-age, higher education and who had resided in Australia longer were more likely to have had screening for CRC in this risk category. In a cost-effectiveness analysis of screening strategies addressed to people at increased risk of CRC due to family history, I found that biennial screening with immunochemical faecal occult blood test, colonoscopy every ten years or colonoscopy every five years reduced the number of CRC cases by 11%, 22% and 34% respectively. All three strategies had an incremental cost-effectiveness ratio (ICER) under $50,000 per life-year gained (LYG), which is regarded as the upper limit of acceptable cost-effectiveness for screening technologies in the Australian health system. At $16,924 per LYG, colonoscopy screening every five-year appears to be the most cost-effective strategy. Overall, my results show that current guidelines for CRC screening are not being implemented in the population. CRC screening participation is low across all risk categories and the vast majority of screening tests undertaken were inappropriate in terms of timing, modality or frequency. Finally, family-history-based CRC screening is a cost-effective strategy and should be considered as an option to increase participation among those most at risk of developing CRC.