Investigation of causal associations between risk factors and cancer incidence

Abstract

The main goal of this thesis is the investigation of causal associations between risk factors and cancer incidence and it was achieved by following two approaches. Firstly, the burden of cancer was estimated as well as what part of it could be attributed to some established cancer risk factors. The second approach involved the investigation of nutritional epidemiology of cancer through the application of some modern techniques. In total four different studies were conducted to answer those questions. Cancer is a leading cause of morbidity and mortality at a global level with 18.1 million cancer cases estimated for 2018. In the last couple of years there is a a growing bodyof literature proposes the use of summary measures of population health such as disability-adjusted life years (DALYs), which reflect both healthy life-years lost because of premature death (YLLs) as well as years of life spent with disability from a disease (YLDs). However, there is a lack of studies that try to estimate the burden of cancer using cohort data. Therefore, a study was conducted in CHANCES consortium of prospective cohorts of elderly people to calculate DALYs for cancer, overall and by cancer site and to estimate population-attributable fractions (PAFs) for selected cancerrisk factors (smoking, adiposity, alcohol, physical inactivity, and type II diabetes) based on the calculated DALYs. Overall, 10,455 individuals developed a new cancer, 4,399 died from cancer and 34,474 DALYs were lost in a cohort of 90,199 participants that were followed for a period of 12 years. YLLs were responsible for the largest part of lost DALYs accounting for almost 82% of the total number while YLDs were responsible for the remaining 18%. Out of the 21 cancer types evaluated in a cohort of, lung cancer was responsible for the largest number of DALYs lost (22.9%), followed by colorectal (15.3%), prostate (10.2%), and breast cancer (8.7%). For 17 cancers, theYLLs were much larger than the YLDs. Pancreatic and liver cancers registered an extreme YLL to YLD ratio of more than 20 while breast and endometrial cancer had the smallest YLL to YLD ratios of 1.1 and 1.4, respectively. Smoking status was the risk factor responsible for the greatest cancer burden. Ever smoking could explain 24.0% (95% CI=22.2% to 26.0%) of total cancer burden followed by physical inactivity (4.9%, 95% CI= 0.8% to 8.1%) and adiposity (1.8%, 95% CI=0.2% to 2.8%). The findings of the current study suggest that there are considerable opportunities forimproving the overall health status of middle-aged and older adults in Europe related to malignancies both in terms of premature mortality but also in terms of living with disability especially after taking into account the increase in the survival rates for many cancer types in the last decades. Environmental factors such as diet, obesity and physical inactivity are considered to berisk factors for several cancers. However, some of the reported results might arise due to biases that can cause inflation in the results and a subsequent dilution of the scientific evidence.Therefore, an umbrella review of the evidence was conducted to systematically evaluate the robustness of the existing literature across many foods and nutrients, as well as, anthropometric measures and physical activity and the risk of cancer at different anatomical sites using the updated data from third expert report published from WCRF. The associations that were included in meta-analyses were categorized into four groups: strong, highly suggestive, suggestive, and weak based onpredefined statistical criteria that take into account the strength of the association as well as the presence of bias in the results. Out of the 1,123 meta-analytic comparisons were included in the UR 19 associations (1.7%) presented strong evidence and they included risk factors such as BMI (n=7), height (n=2) alcohol (n=3), wine (n=1), beer (n=1), dietary calcium (n=1), dairy products (n=2), milk (n=1), and wholegrain products (n=1). Elevated BMI was strongly associated with increased risk of post menopausal breast cancer among never or former users of MHT, post menopausalbreast cancer with a positive progestine receptor esophageal adenocarcinoma, kidney cancer in men while a protective effect was observed in the case of basal cell carcinoma. Furthermore, gained BMI during adulthood increased the risk of post menopausal cancer but a protective effect was observed for high BMI at early adulthood. Regarding the dietary factors, an increased alcohol consumption was associated with a higher risk of post menopausal breast cancer among current users of menopausal hormone therapy (MHT), colorectal and colon cancer. Additionally, an increased consumption of beer and wine was associated with an increased risk of colorectal and post-menopausal breast cancer respectively. On the other hand, elevated consumption of calcium, dietary products, milk and wholegrains showed a decreased risk for colorectal cancer. Fortynine associations (4.4%) were graded as highly suggestive. Most of them (n=36) included anthropometric factors like BMI, height, weight, and waist circumference. The remaining ones refered to alcohol (n=9), coffee (n=2), fruits (n=1) and vegetables (n=1). Finally, for 76 (6.8%) associations the evidence was deemed as suggestive and for 182 (22.7%) as weak. The current study concluded that only a small part of the literature was supported by strong evidence based on subjective statistical criteria. In several cases the evidence was more inconsistent, showing indication of bias and/or uncertainty. Prostate cancer is the most commonly diagnosed cancer in men residing in developed countries. Among the environmental factors, obesity has been suggested to affect prostate cancer risk but with different associations according to disease stage and grade.However, the evidence for specific foods or nutrients is inconsistent, and no specific dietary factors have been consistently associated with prostate cancer risk. Therefore, a NWAS was implemented to systematically and comprehensively evaluate the association between 92 foods or nutrients and risk of prostate cancer in EPIC cohort using proper adjustment for multiplicity of comparisons and replication of emerging findings in an independent population, in this case NLCS, following the paradigm of GWAS studies. In total, 122,998 individuals from EPIC were included in the currentanalysis and after a mean follow-up time of 14 years 5,916 incident cases of total prostate cancer were identified. Of the 92 dietary factors that were evaluated in EPIC, three (dry cakes-biscuits, sauces, confectionary [non-chocolate]) were positively associated with total prostate cancer risk, while seven (total fruits, citrus fruits, vitamin B6, stone fruits, beer-cider, vitamin C, mushrooms) showed a protective association at the nominal statistical significance level. However, no dietary factor retained an association after correcting for multiple comparisons. Additional analyses by subtype of prostate cancer revealed two significant associations. Intakes of dry cakes/biscuits and butter were associated with an increased risk of low-grade (HR, 1.07; 95% CI, 1.03 – 1.11) and aggressive prostate cancer (HR, 1.08; 95% CI, 1.04 – 1.13), respectively. The two suggested associations were further evaluated in NLCS where, the association for dry cakes/biscuits and risk of low-grade prostate cancer was replicated whereas the association for butter and aggressive prostate cancer risk wasn’t. However, the association estimates were similar in both cohorts, and we conducted therefore a metaanalysis, where positive associations were observed for both foods (dry cakes/biscuits: HR, 1.08; 95% CI, 1.04 – 1.11; butter: HR, 1.07; 95% CI, 1.02 – 1.11). In overall our analysis provided little support for an association between diet and risk of prostate cancer. Further replication is needed for the two significant results given the scarcity in the literature and the lack of clear mechanistic pathways. Breast cancer is the most common type of cancer in women at a global level. Regarding the association of diet and nutrition with risk of breast cancer, the literature suggets that the evidence is strong only for alcohol consumption and a higher risk of pre- and postmenopausal breast cancer, whereas it is deemed suggestive for the protective effect of non-starchy vegetables and foods containing carotenoids and calcium with risk of preand post-menopausal breast cancer. Additionally, the literature of biomarkers of nutritional consumption and risk of breast cancer is very limited. The inconsistency in the observed associations of diet and risk of breast cancer can be attributed, at least partially, to the limitations of investigations in nutritional epidemiology like measurement error and recall bias. The aim of the current study was to investigate potential causal associations between genetically determined concentrations of 11minerals and vitamins (beta-carotene, calcium, magnesium, phosphorus, zinc, selenium, copper, folate, vitamin B6, iron, and vitamin B12) and risk of total breast cancer and cancer by estrogen receptor status using the MR methodology in order to bypass the limitations of observational epidemiology. The SNPs associated with the dietary exposures of interest were selected from published GWAS. In total summary association data for 47 common genetic variants associated with the 11 nutrients of interest was obtained. The genetic effects of the selected instruments on risk of breast cancer were obtained from a recently published large GWAS of almost 230,000 participants of European ancestry in BCAC consortium. Across the 11 nutrients evaluated, statistically significant associations emerged for magnesium and phosphorus. A standard deviation (0.08 mmol/L) genetically elevated concentration of magnesium was associated with a 17% and 20% higher risk of total and ER+ve breast cancer, respectively. A protective effect of 16% was observed for a SD (0.5 mg/dL) higher genetically determined phosphorus concentration and risk of ER-ve breast cancer.Additionally, the initial results for calcium denoted a protective although not significant effect on overall breast cancer however, there was an evidence of potential pleiotropy. As a sensitivity analysis, after excluding the pleiotropic SNP a significant protective effect was observed as well. This is the first MR study to identify potential causal associations between 11 nutrients and total and by ER status breast cancer. Regarding the significant results observed in it, the exact mechanisms though which they affect breast cancer risk are not entirely clear and further studies should focus on disentangling the underlying pathways.

Ο κεντρικός σκοπός της συγκεκριμένης διατριβής ήταν ο προσδιορισμός αιτιακών συσχετίσεων μεταξύ παραγόντων κινδύνου και την εμφάνιση καρκίνου. Ο σκοπός αυτός πραγματοποιήθηκε μέσω δύο προσεγγίσεων. Αρχικά αποτιμήθηκε το φορτίο νοσηρότητας του καρκίνου και τι ποσοστό αυτού θα μπορούσε να αποδοθεί σε κάποιους γνωστούς παράγοντες κινδύνου. Στην συνέχεια διερευνήθηκε η διατροφική επιδημιολογία του καρκίνου με μια σειρά μοντέρνων τεχνικών. Τα συγκεκριμέναερωτήματα διερευνήθηκαν μέσα από τέσσερις ξεχωριστές μελέτες. Στην πρώτη μελέτη χρησιμοποιήθηκαν δεδομένα από τον συνασπισμό προοπτικών μελετών CHANCES και εκτιμήθηκε το φορτίο νοσηρότητας 21 διαφορετικών τύπων καρκίνου υπολογίζοντας τον δείκτη των προσδοκώμενων ετών ζωής με συνεκτίμηση της ανικανότητας (disability-adjusted life years, DALYs). Επιπλέον, υπολογίστηκε το φορτίο νοσηρότητας για τους διαφόρους τύπους καρκίνου που οφείλονται σε συγκεκριμένους παράγοντες κινδύνου (κάπνισμα, αλκοόλ, σακχαρώδης διαβήτης,παχυσαρκία και έλλειψη σωματικής άσκησης) μέσω της εκτίμησης των αποδοτέων κλασμάτων τους στον πληθυσμό (Population Attributable Fractions, PAFs). Συνολικά 34,474 DALYs χάθηκαν λόγω του καρκίνου με το μεγαλύτερο μέρος αυτών (82%) να οφείλεται στην πρώιμη θνησιμότητα. Ο καρκίνος του πνεύμονα ήταν υπεύθυνος για το μεγαλύτερο αριθμό χαμένων DALYs, ακολουθούμενος από τους καρκίνους του παχέος εντέρου, του προστάτη και του μαστού. Επιπλέον, το κάπνισμα ήταν ο κύριος παράγοντας κινδύνου υπεύθυνος για το 24% του φορτίου νοσηρότητας ενώ οι υπόλοιποι παράγοντες κινδύνου όπως η παχυσαρκία και η έλλειψη φυσικής άσκησηςακολούθησαν με μικρότερα ποσοστά. Στην δεύτερη μελέτη αξιολογήθηκε η ισχύς της βιβλιογραφίας αναφορικά με την διατροφή, την παχυσαρκία και την φυσική δραστηριότητα και τον κίνδυνο εμφάνισης 12 καρκίνων μέσω μιας ανασκόπησης ομπρέλας. Η παχυσαρκία ήταν ο παράγονταςκινδύνου με τις περισσότερες ισχυρές συσχετίσεις αναφορικά με την εμφάνιση καρκίνου με βάση κάποια συγκεκριμένα κριτήρια αξιολόγησης ενώ στην δεύτερη θέση ακολούθησε η κατανάλωση αλκοόλ. Οι περισσότερες από αυτές τις συσχετίσεις αφορούσαν τον κίνδυνο εμφάνισης καρκίνου του παχέος εντέρου και του μαστού. Επιπλέον ισχυρές προστατευτικές συσχετίσεις παρατηρήθηκαν μεταξύ της κατανάλωσης προϊόντων ολικής άλεσης όπως επίσης και γαλακτοκομικών και τονκίνδυνο καρκίνου του παχέος εντέρου. Χαρακτηριστικές ήταν όμως και περιπτώσεις εμφάνισης πιθανής μεροληψίας στην βιβλιογραφία μεταξύ συσχετίσεων όπως μεταξύ της παχυσαρκίας ή της κατανάλωσης αλκοόλ και τον κίνδυνο εμφάνισης μετάεμμηνοπαυσιακού καρκίνου του μαστού ή μεταξύ της κατανάλωσης κόκκινου και επεξεργασμένου κρέατος και τον κίνδυο εμφάνισης καρκίνου του παχέος εντέρου οι οποίες αν και είναι επιστημονικά εμπεριστατωμένες εντούτοις, είτε εμφάνισαν στοιχεία πιθανής μεροληψίας στα αποτελέσματα τους είτε η ισχύς των στατιστικών αποτελεσμάτων δεν τόσο ισχυρή. Στην τρίτη μελέτη μελετήθηκαν με μια συστηματική προσέγγιση μέσω μιας μελέτης συσχέτισης ευρείας κλίμακας διατροφικών παραγόντων (Nutrient - Wide AssociationStudy, NWAS) οι συσχετίσεις μεταξύ 92 τροφών και διατροφικών συστατικών και τον κίνδυνο εμφάνισης καρκίνου του προστάτη χρησιμοποιώντας δεδομένα από την προοπτική μελέτη EPIC με στόχο την ανακάλυψη νέων αιτιακών συσχετήσεων και επαλήθευση συσχετήσεων που υπάρχουν στην βιβλιογραφία. Ύστερα από την διόρθωση για πολλαπλές συγκρίσεις δεν βρέθηκε κάποιος παράγοντας που να σχετίζεται με την εμφάνιση καρκίνου του προστάτη. Οι επιμέρους αναλύσεις όμως με βάση τον σταδιοποίησης και διαφοροποίησης εμφάνισαν δύο στατιστικά σημαντικάαποτελέσματα. Αρχικά βρέθηκε ότι η κατανάλωση μπισκότων και κέικ αυξάνει τον κίνδυνο εμφάνισης καρκίνου χαμηλού βαθμού διαφοροποίησης, ένα αποτέλεσμα το οποίο επαληθεύτηκε και σε δεύτερη φάση στην ανεξάρτητη κοόρτη NLCS. Η δεύτερη συσχέτηση αφορούσε την κατανάλωση βουτύρου και τον αυξημένο κίνδυνο επιθετικού (υψηλός βαθμός σταδιοποίησης ή διαφοροποίησης) καρκίνου του προστάτη. Η συγκεκριμένη συσχέτηση δεν επαληθεύτηκε στην κοόρτη NLCS αν και το αποτέλεσμα της μέτα-ανάλυσης των δύο κοορτών ήταν επίσης στατιστικά σημαντικό. Στην τελευταία μελέτη ερευνήθηκαν πιθανές αιτιακές συσχετήσεις μεταξύ 11ιχνοστοιχείων και βιταμινών και τον κίνδυνο εμφάνισης καρκίνου του μαστού χρησιμοποιώντας την σύγχρονη τεχνική της μεντελιανής τυχαιοποίησης. Στην ανάλυση αυτή χρησιμοποιήθηκαν δεδομένα από δημοσιευμένες ευρυγονιδιωματικές μελέτες. Στις περισσότερες συγκρίσεις δεν βρέθηκαν κάποια στατιστικά σημαντικά αποτελέσματα με εξαίρεση δύο περιπτώσεις. Πιο συγκεκριμένα, γενετικά αυξημένες συγκεντρώσεις μαγνησίου οδήγησαν σε αυξημένο κίνδυνο εμφάνισης καρκίνου τουμαστού τόσο συνολικά όσο και στην περίπτωση καρκίνου με υποδοχέα οιστρογόνων. Αντίθετα, προστατευτική επίδραση παρατηρήθηκε μεταξύ γενετικά αυξημένων συγκεντρώσεων φωσφόρου και τον κίνδυνο εμφάνισης καρκίνου με χωρίς υποδοχέα οιστρογόνων.