Inflammation and risk of chronic diseases with a focus on colorectal cancer and the impact of dietary patterns

(1)

Inflammation and risk of chronic diseases

with a focus on colorectal cancer

and the impact of dietary patterns

Stina Bodén

Department of Radiation Sciences, Oncology

Department of Public Health and Clinical Medicine, Sustainable Health Umeå 2020

(2)

This work is protected by the Swedish Copyright Legislation (Act 1960:729) Dissertation for PhD

ISBN: 978-91-7855-366-2 ISSN: 0346-6612

Umeå University Medical Dissertations, New Series No 2099

Cover photo by Amber Locke, design by Ida Åberg, Inhousebyrån, Umeå University Electronic version available at: http://umu.diva-portal.org/

Printed by: Cityprint i Norr AB Umeå, Sweden, 2020

(3)

Have no fear of perfection; you’ll never reach it

-Salvador Dali

(4)

Abstract

Background: Inflammation is implicated in the development of several chronic

diseases including cardiovascular disease and cancer. Prevention of these would be greatly facilitated if the influence of lifestyle factors, such as dietary prefer-ences, on systemic chronic low-grade inflammation could be more clearly de-fined. Using biomarkers, including markers of inflammation, may help determine potential mechanistic links between diet and disease.

Aim: The aim of this thesis was to investigate the role of systemic, chronic,

low-grade inflammation in the risk of cardiovascular disease and cancer and in par-ticular colorectal cancer. A special focus was placed on the impact of dietary pat-terns and biological markers of inflammation.

Methods: The studies comprising this thesis were based on prospectively

col-lected data and biological samples from two population-based cohorts within the Northern Sweden Health and Disease Study. Paper I was a nested case-control study of the Dietary inflammatory index and risk of myocardial infarction, includ-ing 1389 cases and 5555 matched controls. A subset (605) of the control partici-pants was used to validate the index via the circulating inflammatory biomarkers C-reactive protein (CRP) and Interleukin-6 (IL-6). Paper II was a large cohort study of the association of diet with cancer risk using the Dietary inflammatory index and the Mediterranean diet score. The study included 100 881 participants with dietary data, of which 9 250 developed cancer after joining the cohort. Lon-gitudinal changes in the dietary patterns were studied in 35 393 participants with repeated measures. Papers III and IV were nested case-control studies, including 1010 and 735 colorectal cancer cases, respectively, with individually matched con-trols. Paper III investigated baseline and longitudinal changes in plasma CRP concentrations in relation to future risk of clinical and molecular subtypes of col-orectal cancer. In Paper IV, data-driven dietary patterns derived from factor anal-ysis were investigated in relation to colorectal cancer risk and to untargeted me-tabolite profiles measured in plasma.

Results: In Paper I, higher Dietary inflammatory index scores, indicative of a

more “pro-inflammatory diet”, were associated with a higher risk of myocardial infarction in men but not in women. Higher levels of the inflammatory bi-omarkers CRP and IL-6 were modestly associated with higher Dietary inflamma-tory index scores. In Paper II, lower Dietary inflammainflamma-tory index scores and higher Mediterranean diet scores, indicative of an “anti-inflammatory” or “heath-ier” diet, respectively, showed similar, weak inverse associations with overall can-cer risk, and primarily for lung and gastric cancan-cer in men. No significant associa-tions were seen for women or for changes in dietary patterns in Paper II. In Paper

(7)

ii

III, there was no association between plasma concentrations of CRP and the risk of future colorectal cancer, regardless of clinical or molecular tumor subtype, ex-cept for a possible increased risk of advanced colorectal cancer risk in participants with <5 years between sampling and diagnosis. In Paper IV, a data-driven break-fast food pattern, including fermented dairy products, firich cereals and ber-ries, was inversely associated with the risk of distal colon cancer, especially in women. Wholegrain, fiber, fruits and vegetables, and alcohol consumption were associated with metabolite profiles, which warrants further investigation to iden-tify individual metabolites involved.

Conclusions: Results from these population-based studies, using prospectively

collected data and blood samples, support a modest contribution of dietary pat-terns to the development of chronic diseases. Theoretical sex differences ob-served warrants further investigation. Data-driven methods to derive dietary pat-terns may complement hypothesis-driven methods in future work, and together with the use of metabolomics may provide more insight into the complexity of diet in the etiology of chronic diseases. In summary, although the findings in this thesis were somewhat consistent with a weak inflammatory effect of diet in the link between diet and cardiovascular disease and cancer, it appears unlikely to be a major causal mechanism.

(8)

iii

Abbreviations

BMI Body mass index

BRAF B-Raf proto-oncogene serine/threonine kinase

CI Confidence interval

CRP C-reactive protein

FFQ Food frequency questionnaire

HR Hazard ratio

KRAS Kirsten rat sarcoma viral oncogene homolog gene LC-MS Liquid chromatography mass spectrometry

IL-6 Interleukin-6

MI Myocardial infarction

MONICA Multinational MONitoring of Trends and Determi-nants of Cardiovascular Disease

MSI Microsatellite instability

MSS Microsatellite stability

MUFA Monounsaturated fatty acids

NSHDS Northern Sweden Health and Disease Study

OR Odds ratio

PUFA Polyunsaturated fatty acids

SD Standard deviation

SFA Saturated fatty acids

TFA Trans fatty acids

(9)

iv

Original papers

I. Bodén S*, Wennberg M, Van Guelpen B, Johansson I, Lindahl

B, Andersson J, Shivappa N, Hébert JR, Nilsson LM. Dietary in-flammatory index and risk of first myocardial infarction; a pro-spective population-based study. Nutr J. 2017;16(1):21.

II. Bodén S*, Myte R, Wennberg M, Harlid S, Johansson I,

Shivappa N, Hébert JR, Van Guelpen B, Nilsson LM. The in-flammatory potential of diet in determining cancer risk; a pro-spective investigation of two dietary pattern scores. PLoS One.

2019;14(4):e0214551.

III. Bodén S*, Myte R, Harbs J, Sundkvist A, Zingmark C, Löfgren

Burström A, Palmqvist R, Harlid S, Van Guelpen B. C-reactive protein and future risk of clinical and molecular subtypes of col-orectal cancer. Cancer Epidemiology Biomarkers & Prevention. 2020;29(7):1482-91.

IV. Bodén S*, Zheng R, Landberg R, Hanhinieva K, Harlid S,

Vid-man L, Johansson I, Winkvist A, Gunter M J, Van Guelpen B, Brunius C. Data-driven dietary patterns and their association with colorectal cancer risk and untargeted plasma metabolite profiles. Manuscript.

The original papers were reprinted with the permission of the publishers. Paper I-III were all open source published.

(10)

v

Overview of papers

Study

design Main exposure Main outcome Cases Main findings

I Nested case- control Dietary in-flammatory index

MI risk 1389 MI cases “Pro-inflammatory diet”

as-sociated with increased risk of MI in men but not in women. Moderate association between higher CRP/IL-6 and higher Dietary inflamma-tory index.

II Cohort Dietary

in-flammatory index, Mediterra-nean diet score All cancer risk, includ-ing common cancers 9250 cancer cases; 2109 prostate, 1546 breast, 1655 GI, 1036 CRC, 442 lung, 223 pancreas, 163 gastric cancer

Weak consistent associations between “anti-inflamma-tory”/”healthier” diet and re-duced risk of cancer, predom-inantly lung and gastric can-cer in men. No longitudinal association between 10-year change in diet and cancer risk. III Nested case- control CRP CRC risk, by BRAF and KRAS muta-tion, MSI, and clinical subtypes 1010 CRC cases (≥704) with molecular tumor data) No association between CRP and overall, clinical or molec-ular subtype CRC risk. Possi-ble direct association be-tween advanced tumors and short follow-up time.

IV* Nested case- control Data-driven dietary pat-terns deter-mined by factor analy-sis CRC risk, and metabo-lite profiles 735 CRC cases

A pattern of breakfast food associated with distal colon cancer, particularly in women. Pattern of alcohol and total intake of alcohol, wholegrain, fiber, and fruit and vegetables associated to pre-diagnostic plasma metab-olite profiles.

BRAF: B-Raf proto-oncogene serine/threonine kinase; CRC: Colorectal cancer; CRP: C-reactive protein; GI: gastrointestinal; IL-6: Interleukin-6; KRAS: Kirsten rat sarcoma viral oncogene homolog gene MI: Myocardial infarction; MSI: Microsatellite instability

(12)

vii

Populärvetenskaplig sammanfattning

Inflammation kan bidra till utvecklingen av flera kroniska sjukdomar så som hjärtkärlsjukdom och cancer. Inflammation som av karaktären är låggradig och ihållande tros ha ett samband med vissa livsstilsfaktorer, inklusive typ av kost. Det finns en rad etablerade samband mellan mat och risk för sjukdom. Bland an-nat vet vi att energirik mat eller mat i mängder som överskrider vad kroppen be-höver kan orsaka övervikt och fetma. Övervikt och fetma ger i sin tur ökad risk för en rad olika sjukdomar, däribland hjärtinfarkt och flera typer av cancer, in-klusive tjocktarmscancer. Det finns även mycket som tyder på att det är positivt att äta en kost som innehåller mycket frukt, grönsaker, fisk och fibrer men mindre av mättat fett och socker. Det är dock inte helt klarlagt vilka mekanismerna är bakom dessa samband eller i vilken utsträckning risken för olika sjukdomar ökar eller minskar. För att kunna bedöma sambandet mellan kosten i sin helhet och dess inverkan på risk för sjukdom är det lämpligt att studera individers kost-mönster. Genom att mäta ämnen i blodprover från individer som senare utvecklat olika sjukdomar kan man även ta reda på vissa av de underliggande mekan-ismerna. Man kan t.ex. mäta inflammationsmarkörer i blodprover för att studera hur inflammation är relaterad till en viss kost men också för att se om inflammat-ion är relaterad till sjukdomar som hjärtinfarkt eller cancer. Livsstilsfaktorer som övervikt, rökning, inaktivitet, hög konsumtion av rött kött och charkuterier, och låg konsumtion av kostfiber har visat sig vara särskilt förknippade med risken att utveckla tjock- och ändtarmscancer. Tjock- och ändtarmscancer är cancerformer som tar många år att utveckla och därför borde det finnas goda förutsättningar för att förhindra eller skjuta upp sjukdomen med hjälp av långsiktiga livsstilsför-ändringar. Tumörens egenskaper, både var i tarmen den sitter och dess mole-kylära egenskaper, har betydelse för behandling och prognos. Om dessa olika egenskaper hos tumörer påverkas av olika riskfaktorer är dock inte klarlagt. Syftet med denna avhandling var att undersöka hur kronisk låggradig inflammat-ion påverkar risken för kronisk sjukdom, med särskilt fokus på olika kostmöns-ters inverkan. I de första två arbetena studerades ett kostinflammationsindex, framtaget utifrån tidigare kunskap om sambandet mellan intag av vissa närings-ämnen och inflammation i kroppen. Kostinflammationsindex relaterades både till risk för hjärtinfarkt och till risk för cancer. I studien om hjärtinfarkt användes två blodprovsmått på inflammation. Syftet med det var att uppskatta hur väl kostinflammationsindex kunde fånga den inflammatoriska effekten av kosten. I studien om cancer där alla de vanligaste cancerformerna studerades, användes även ett annat kostindex för att mäta studiedeltagarnas följsamhet med en me-delhavsliknande kost. De två senare arbetena i avhandlingen fokuserade på ris-ken för tjock- och ändtarmscancer. Här studerades nivåer av C-reaktivt protein

(13)

viii

(CRP), ett mått på inflammation som även kallas för ”snabbsänka”, i blodprover och risken för olika undergrupper av tjock- och ändtarmscancer. Tidigare studier har gett olika resultat för sambandet mellan CRP och risk för tjock- och änd-tarmscancer men ingen har studerat om särskilda molekylära undergrupper har starkare samband än andra. Slutligen studerades också om icke hypotesdrivna (eller data-drivna) kostmönster kunde kopplas till framtida risk för tjock- och ändtarmscancer, liksom om dessa kostmönster avspeglade olika metabola profi-ler i blodet.

Deltagarna i denna avhandling var från två stora befolkningsstudier i norra Sve-rige; Västerbottensprojektet och MONICA-studien. I Västerbottenprojektet bju-der man in deltagare i hela Västerbotten det år de fyller 40, 50 och 60 år och i MONICA-studien bjuder man ca var 5:e år in slumpvalda invånare i Västerbotten och Norrbotten i åldrarna 25-74 år. Från deltagarna i dessa befolkningsstudier tas blodprover som sparas för att kunna användas i forskning så som i denna av-handling. Utifrån enkätsvar från deltagarna har olika kostmönster studerats mel-lan deltagare som senare utvecklat hjärtinfarkt eller cancer och deltagare som inte gjorde det. Enkätsvar gav också information om andra egenskaper och livsstilsvanor. Nivåer av inflammationsmarkörer liksom metaboliter har också mätts i deltagarnas sparade blodprover. Enkäterna och blodproverna var insam-lade flera år innan sjukdomsdiagnos.

I den första studien var risken för en framtida hjärtinfarkt högre hos män som rapporterade en kost som enligt hypotesen kan ge inflammation jämfört med de män som rapporterat en kost som enligt hypotesen kan ge lägre inflammation. Detta samband sågs dock inte för kvinnorna i studien. I studien om cancer med över 100 000 deltagare sågs svaga samband mellan en kost som enligt hypotesen kan ge lägre inflammation, liksom bättre följsamhet till en medelhavsliknande kost, och minskad risk för cancer, särskilt för cancer i lunga och magsäck hos männen i studien. Förändringar i kostmönster över en tioårsperiod var inte rela-terat till cancerrisk.

I studien om undergrupper av tjock- och ändtarmscancer kunde de olika egen-skaperna hos dessa undergrupper inte förklara tidigare varierande resultat för sambandet till CRP. Hos individer med mer avancerad sjukdom var högre nivåer av CRP i blodprover tagna närmare tidpunkten för diagnos associerade till ökad risk av tjock- och ändtarmscancer. Det berodde mest troligt på att tumören redan hade börjat växa då blodprovet togs och då kan tumören i sig ge ökad inflammat-ion.

I det fjärde delarbetet togs kostmönster fram med hjälp av data-drivna statistiska metoder, baserat helt på hur studiedeltagarna hade rapporterat. Ett sådant data-drivet kostmönster, förknippat med typisk frukostmat såsom filmjölk, yoghurt,

(14)

ix

fiberrika flingor och bär, var associerat med en lägre risk för cancer i vänster sida av tjocktarmen, särskilt hos kvinnorna i studien. Fiber, fullkorn, frukt och grön-saker samt alkohol var associerade till metabolitprofiler i blodet. Vi kommer att gå vidare med att försöka identifiera vilka specifika metaboliter som ingår i dessa metabolitprofiler, för att förhoppningsvis få ledtrådar till mekanismerna mellan kostfaktorerna och uppkomsten av cancer i tjock- och ändtarmen.

Sammantaget visade dessa observationsstudier att kostmönster i viss mån kan bidra till utvecklingen av hjärtinfarkt och vissa typer av cancer. Ett kostmönster särskilt utvecklat för att fånga den inflammatoriska effekten av kost visade lik-nande svaga samband till cancer som ett kostmönster baserat på Medelhavskost. Data-drivna metoder för att studera kostmönster liksom analyser av metaboliter från biologiska prover kan i framtiden vara av betydelse för att förstå mer av det komplexa förhållandet mellan kost och sjukdom. Potentiella skillnader mellan kvinnor och mäns kostvanor i relation till kroniska sjukdomar, och de bakomlig-gande orsakerna till dessa eventuella skillnader, är av intresse att studera i fram-tiden. I vilken grad kronisk låggradig inflammation bidrar till sambandet mellan kost och utvecklingen av kroniska sjukdomar är svårt att fastställa och antagligen är det flera mekanismer involverade. Att främja en hälsosam och hållbar livsstil som innefattar aspekter av kosten utifrån det vi känner till idag, oavsett bakom-liggande mekanismer, bör fortsatt vara den primära målsättningen för att för-bättra folkhälsan och förebygga kroniska sjukdomar.

(15)

(16)

1

Background

Inflammation

The immune system’s inflammatory response in the human body is critical for recovery and survival during acute physical injury or infection (1). This type of short-term inflammatory response is characterized by restricted upregulation of high-grade inflammatory activity and resolves when the infection or injury has passed. Systemic, chronic, low-grade inflammation (hereinafter referred to as low-grade inflammation) is instead characterized by collateral damage and is non-resolving. Both acute and chronic inflammation do share common mecha-nisms and a shift from short- to long-term inflammation can have significant neg-ative effects on at the tissue, organ, and cellular levels; local inflammation in the intestine in individuals with chronic inflammatory bowel disease is an example of this (2). Inflammation can thus act as both friend and foe, with more persistent, chronic low-grade inflammation leading to a pathological state. This is the case in several chronic diseases including cardiovascular disease (Figure 1), and hy-pothetically also in cancer (1, 3, 4).

Figure 1. Causes and consequences of systemic, chronic, low-grade inflammation.

Causes or triggers of low-grade inflammation that have been identified are illustrated to the left. To the right are some suggested consequences of this type of inflammation. Figure modified from Fur-man et al., 2019 (1).

Aging is strongly associated with chronic low-grade inflammation, related to the process called immunosenescence - a dysregulation of the immune system caused

Obesity Diet Microbiota dysbiosis Sedentary lifestyle Chronic infections

Disrupted circadian rythm Xenobiotics Immunosenescence Alzheimer Cancer Cardiovascular disease Metabolic syndrome including diabetes Autoimmune diseases Sarcopenia Osteoporosis Causes Systemic chronic low-grade inflammation Consequences

(17)

2

by natural aging (5). The incidence of many types of cancer increases rapidly in later mid-life and older ages, and age-related chronic conditions including in-flammation and cancer development may be prevented to a certain extent by maintaining a healthy lifestyle in young and mid-life individuals (6). It is there-fore important to investigate if and how lifestyle factors can influence low-grade inflammation and the subsequent risk of cancer.

This thesis focuses on low-grade inflammation and the potential contribution of dietary patterns towards the risk of cardiovascular disease (7) and cancer - and in particular colorectal cancer (4).

Nutritional epidemiology

Diet and the development of chronic diseases

The link between food and health has been documented since the antiquity. Re-gardless of whether the quote “Let food be thy medicine and medicine thy food” is correctly attributed to Hippocrates, there is massive evidence that food has di-rect effects on health. Early dietary studies from the 18th_{century identified} defi-ciencies that could lead to disease, such as the controlled trial-type investigations among sailors showing vitamin C deficiency leads to scurvy (8). In the late 19th century, thiamin deficiency as a result of very restricted diet – common during lengthy periods at sea – led to the unusual disease beriberi.

Nutritional epidemiologists have later shifted their research focus from deficien-cies in essential nutrients to the chronic diseases of Western civilization, includ-ing cardiovascular disease and cancer. Unlike nutritional deficiencies, these dis-eases most often have multiple causes, in which diet combines with other lifestyle factors such as lack of physical activity and tobacco consumption. Genetic, occu-pational, psychosocial, and infectious factors often play a role as well. Cardiovas-cular disease, cancer and other chronic diseases can have long latent periods, sug-gesting an importance of the cumulative exposure to several contributing factors over a lengthy duration. Distinguishing between various potentially contributing factors and their respective exposure periods is a veritable challenge when de-signing and interpreting nutritional epidemiology studies. Diet, in comparison to other risk factors such as tobacco consumption, represents a complex set of ex-posures that are strongly intercorrelated. And while tobacco exposure is often ac-curately reported, and thus easier to assess as a risk factor (9), diet is more diffi-cult to measure due to a greater number of sources of variation in an individual’s dietary intake, and in the reporting thereof (10).

(18)

3

Dietary assessment methods

In nutritional research, there are several methods for assessing an individual’s dietary intake (11). Prospective methods include estimated or weighed food rec-ords, typically collected over a period of 3-7 days. In the latter, the individual needs to weigh all the foods eaten throughout the whole assessment period. This method requires well-motivated participants as it places considerable demands on the individual: they need constant access to a measurement scale, and must report details on food preparation, descriptions, and possibly brand names. The risk of selection bias is therefore high. Another common bias in both estimated and weighed food records is that the methods can affect the actual amount of food intake for the individual, whereby the total dietary and energy intake decreases resulting in concomitant weight loss. Prospective dietary assessments are there-fore usually conducted in a limited number of study participants, and the drop-out risk is greater.

Retrospective methods to assess dietary intake include dietary history interviews that generally encompass the last 24 hours (known as 24-hour recalls), but also longer periods (11). The former can be performed in a larger number of individu-als but recall bias and interviewer bias are two disadvantages with this method. It can also be difficult for the individual to recall quantities of each food eaten. In order to collect dietary data from many individuals – which is necessary when investigating correlations with rare diseases or outcomes in which dietary expo-sure plays a minor role – Food Frequency Questionnaires (FFQs) are often used, due to their cost and time effectiveness. As for 24-hour recalls, FFQs are limited by the individual’s ability to recall intake. It can also be even more challenging to estimate both frequency and amounts for foods eaten over the previous months or year than the previous 24 hours. However, the same FFQ can be used over considerable durations, facilitating data collection in prospectively designed studies. Despite its limitations, the FFQ is thus the most commonly used dietary assessment method in nutritional epidemiology studies.

Measurement error in nutritional epidemiology

The complexity of food and diet can be challenging given the many aspects to consider: availability, price, quality and safety, knowledge and interest in healthy eating, religious and personal interests (e.g. animal welfare and environmental impact), socioeconomic status, personal taste preferences, preferences of related individuals, allergies and intolerances, time available, and cooking experience. The complexity of diet must also be considered when planning and conducting nutritional epidemiology studies. In such studies, investigators are generally in-terested in the long-term dietary intake and must therefore assess day-to-day or week-to-week variation. While day-to-day variation may constitute random measurement error, others may be more systematic, such as day-of-the-week or

(19)

4

seasonal variation, leading to bias. The degree of random variation generally dif-fers among measured parameters: energy intake has relatively low variation, macronutrient intake slightly higher, and micronutrient intake the highest (11). It is typically not possible to distinguish between within-person random meas-urement error and truly random day-to-day variation. Increasing the number of days dietary records are performed, or conducting repeated measurements of 24-hour recalls and FFQs may limit or correct for measurement errors (11), and help validate dietary measurements from FFQs in the same individuals. Nonetheless, no reference method can likely provide a true measurement of the dietary intake, so validation studies should be regarded as relative validations. Measurement er-ror likely accounts for the most substantial sources of bias in nutritional epide-miology and can have implications for etiological investigations of lifestyle-re-lated diseases. For example, foods perceived as unhealthy are generally underre-ported more than foods perceived as healthy (12), especially by women and by people who are overweight or obese (13).

Energy adjustment

There are strong arguments for adjusting total energy intake when controlling for confounding factors in diet-disease investigations (14). Firstly, energy intake is positively correlated to the intake of most nutrients. Secondly, energy intake may be a primary determinant of the disease in question, if it is in any way related to body size, physical activity, or metabolic efficiency as is the case in cardiovascular disease and some cancers. There are several approaches to address the influence of total energy intake on the analysis of nutrients, described by Walter Willett (14). A standard statistical approach, called standard multivariate model, in-cludes energy intake in a multivariable model together with the nutrient of inter-est: y=b1 nutrient + a (total energy intake). In this situation, the beta coefficient (b1) of the nutrient can be interpreted as the effect of increasing the nutrient by one unit while energy intake (a, alpha) is held constant. In order to remove ex-traneous variation due to energy intake, however, the residual method can be used instead, whereby regression analysis computes residuals of nutrient intake, representing the difference between each individual’s actual intake and the intake predicted by their total energy intake: y=b1 nutrient residual + b2 total energy in-take. The nutrient density method is another common approach, whereby the nu-trient of interest is divided by total energy intake: y=b1 nutrient/total energy in-take + b2 energy intake. This is often expressed as percentage of energy or intake per 1000 kcal. The nutrient density method is simple and does not involve statis-tical models as with the residual method. Regardless of method, when association of the total energy intake with the outcome is suspected, it is recommended to add it to the models as a confounder, in addition to the energy adjustment of the actual nutrient.

(20)

5

Dietary patterns

Traditionally, diet has been investigated and presented in terms of nutrients, foods, or food groups. But in the recent decades more emphasis has been placed on studying dietary patterns in an individual’s whole diet. Although studies of single foods or nutrients have the potential to identify some biological mecha-nisms, they cannot account for the complexity of the diet as whole (15). Foods incorporated in a person’s diet can provide thousands of bioactive substances that can act individually but also synergistically, influencing processes such as cell differentiation, apoptosis, and hormonal regulation (16). The effect of a sin-gle nutrient might also be confounded by a dietary pattern (17). Studying dietary patterns can therefore better capture combinations, quantities, and interrela-tions of food in the individual’s typical diet.

A priori and a posteriori dietary patterns

Various approaches exist for deriving patterns from a set of dietary data, such as studying adherence to dietary guidelines. An a priori dietary pattern is one that is based on prior knowledge or hypotheses about the response variables linking patterns to the risk of a certain disease (17). In that sense, one can say that a priori patterns are hypothesis-driven. Examples of commonly used a priori patterns studied in relation to colorectal cancer risk are the Healthy Eating Index (18), the Dietary inflammatory index (19), and the Mediterranean diet score (20). The lat-ter two are described in grealat-ter detail in this thesis, as they have been employed in the included investigations.

Another approach to describe overall dietary intake is by a posteriori methods, which are entirely data-driven. A data-driven method allows for food groups, sin-gle foods or nutrients to be put together in a way that reflects the actual dietary intake patterns of the population that is being studied. More commonly used data-driven methods include cluster analysis, principal component analysis and reduced rank regression (17). In another data-driven method, exploratory factor analysis, potential explanatory factors can be distinguished by observing rela-tions among the various dietary variables. Exploratory factor analysis can also be combined with confirmatory factor analysis – a method similar to principal com-ponent analysis – to produce dietary patterns (21).

Both a priori and a posteriori methods have been employed when investigating the association between dietary patterns and diseases. Studies of dietary patterns in relation to colorectal cancer risk have detected some consistent associations to reduced risk, such as low intakes of red and processed meats and high intakes of fruits and vegetables (22). However, risk relationships are not always consistent and underlying mechanisms are still not fully established (17). It might therefore be preferable to use different methods to derive dietary patters, complementing

(21)

6

each other, in order to better explain the diet-disease relationship. To use bi-omarker validations of dietary patterns and to continuously improve dietary anal-ysis methods in terms of both theoretical relevance and innovative statistical techniques is desirable.

Dietary inflammatory index

The Dietary inflammatory index, launched in 2014, is based on the hypothesis that chronic low-grade inflammation explains part of the link between diet and the development of cancer as well as cardiovascular and other chronic diseases (19). There are popular claims that some foods have “anti-inflammatory” modu-lating effects, such as foods rich in fiber and wholegrains, phytochemicals and micronutrients like vegetables, fruit, and certain spices (23). “Pro-inflammatory foods” have similarly been suggested and include foods rich in saturated or trans fatty acids. With this logic, the diet as a whole is likely more relevant for any hy-pothetical inflammatory effects than are single constituents. The Dietary inflam-matory index therefore aims to classify an individual’s diet ranging from the ex-tremes of “pro-inflammatory” positive (+) scores to “anti-inflammatory” negative (-) scores, and to adapt it to diverse populations (19). An increasingly positive score indicates a putative increasing effect on the inflammatory biomarkers in-terleukin-6 (IL-6), interleukin-1B, tumor necrosis factor-a, or C-reactive protein (CRP), or a putative decreasing effect on interleukin-4 and interleukin-10, ac-cording to qualifying scientific articles on which the index was based.

The Dietary inflammatory index has been widely used in observational research, meta-analyses, and reviews of various cancer outcomes (24-31) including colo-rectal cancer (32-34), as well as cardiovascular outcomes (35-37). Studies using the Dietary inflammatory index have generally indicated an increased risk of can-cer for more positive, “pro-inflammatory” index scores, with the most consistent results for breast-, lung-, and colorectal cancers (24). The original index includes 45 different dietary factors including both macronutrients, like carbohydrates, and individual food components or seasonings, like flavanols and garlic. How-ever, published articles on the index have generally used fewer components than the original score; in the 40 studies included in one meta-analysis on the Dietary inflammatory index, between 25 and 36 components of the 45 possible compo-nents were used (27).

Mediterranean Diet Score

The traditional Mediterranean diet is characterized by high intakes of vegetables, legumes, fruits, nuts, cereals, olive oil, moderate intakes of fish (depending on the distance to the sea), low to moderate intakes of dairy products (dominated by fermented products like cheese and yoghurt), low intakes of meat and poultry, and a regular but moderate intake of ethanol (dominated by wine and generally

(22)

7

together with meals). Trichopoulou et al. constructed the Mediterranean Diet Score as a nine-point scale, according to which higher scores indicate greater ad-herence to the diet (20). They showed that greater adad-herence was associated with reduced mortality, including mortality due to cardiovascular disease. The cut-points are based on population-specific median intakes and thus dependent on the population being studied. This score and its various adaptations have been used in epidemiology research for decades, and a recent meta-analysis of its as-sociation to cancer provided clear support for a protective effect of the Mediter-ranean diet, in particular for colorectal cancer (38). Good adherence to the Med-iterranean diet has also been associated with lower levels of circulating inflam-matory biomarkers (39), although several pathways in addition to inflammation have been suggested to explain the positive health benefits of the diet (40). Obe-sity and adipose tissue are directly associated with chronic inflammation, but Mediterranean diet intervention studies have shown reduced levels of plasma in-flammatory biomarkers unrelated to weight loss (41).

Biomarkers in nutritional epidemiology

Self-reported dietary data and its potential measurement errors reduces validity and reliability in nutritional epidemiology and cannot reveal causal relationships between diet and disease. Blood- or urine-based biomarkers may more objec-tively evaluate diet (42). Advances in various omics techniques such as epigenet-ics, transcriptomepigenet-ics, proteomepigenet-ics, lipidomepigenet-ics, and metabolomics are enabling the detection of novel biomarkers. The use of biomarkers in measuring human bio-logical variation in epidemiology research, not restricted only to dietary bi-omarkers, is called molecular epidemiology.

Diet and metabolomics

Metabolomics is the study of small-molecule metabolites in a biological sample, including both endogenous and exogenous metabolites (43). A biological sample can be extracted from blood, urine, tissue, or feces, for example, and the complete set of metabolites in the sample is referred to as its metabolome. Metabolomics combines analytical strategies to identify metabolites with statistical methods for data extraction and interpretation. Together with other omics, metabolomics can help us understand how environmental exposures affect the biology of an organ-ism.

Untargeted metabolomics is a hypothesis-free method that provides a broad over-view of metabolites indicating shifts in diverse metabolic pathways. It can include more than 1000 so called ‘features’ and allows examination of the relation be-tween interconnected metabolites (43). It is still not possible to obtain all metab-olites in a biological sample simultaneously, and numerous human metabmetab-olites

(23)

8

are still unknown. However, there are a number of growing, freely available an-notated metabolite databases that can be used for identification, such as the Hu-man Metabolome Database and the Kyoto Encyclopedia of Genes and Genomes (44). There are also specific databases for biomarkers of exposures to dietary and environmental factors, such as the Exposome-Explorer and the Food Component Database (45).

Metabolomics in human nutrition is a growing research area. Traditionally, most dietary biomarkers have been identified through prior knowledge about food composition. As more than 25 000 compounds are known in a normal individ-ual’s food intake, the complex food metabolome provides a plethora of potential metabolic relationships (16). Examining the food metabolome can facilitate de-tection of metabolites associated with certain diseases (Figure 2 provides a broad overview of this). Urine- and blood-derived metabolite profiles can also identify habitual dietary patterns (46) including a priori dietary patterns (47, 48). The number of studies investigating the diet-cancer relationship using metabo-lomics are limited, but the field is growing. Challenges in such studies include selection and handling of biological material, preprocessing, deriving the dietary data, and conducting complex multivariate statistical analysis (45).

Figure 2. Overview of the study of the food metabolome using a case-control study de-sign. Food-related biomarkers associated with disease can be discovered using an untargeted

metabolomics approach. Several potential sources can be discovered with this approach, including biomarkers of drugs, food and microbiota. Biomarkers can be discriminated that differ in study par-ticipants with, or who later develop, a given disease, compared to healthy subjects, and these may then be linked to exposure biomarkers. Figure modified from Scalbert et al., 2014 (16).

Diseased vs. Healthy subjects Drug biomarkers Pollutant

biomarkers Endogenousbiomarkers

Dietary biomarkers Un ta rg ete_d Unta rget ed Metabolome including food metabolome Exposures Biomarkers of disease Biomarkers of exposure

(24)

9

Mechanisms linking diet, lifestyle and chronic diseases

Age is one of the major risk factors for chronic diseases, including cardiovascular disease and cancer. With a continuously rising global age level and life expec-tancy, incidences of these diseases are also expected to increase unless primary and secondary prevention can mitigate them. Lifestyle factors including diet play major roles in the incidence of these diseases. Type 2 diabetes, cardiovascular disease, and cancer have largely mutual risk factors (1). Diet can affect the devel-opment of chronic diseases both directly, through specific dietary elements, as well as indirectly. Continuous overeating can lead to obesity, which is indisputa-bly one of the strongest risk factors for cardiovascular disease and breast- and colorectal cancer (49). Obese individuals have higher levels of circulating inflam-matory biomarkers compared to lean individuals since the adipose tissue itself releases various inflammatory mediators (50). Underlying mechanisms linking adiposity to cancer are thought to involve insulin and insulin-like growth factor 1 signaling, altered sex hormone metabolism, altered microbiota, and adipose tis-sue-derived inflammation (51-53).

Figure 3. Potential mechanisms promoting associations between diet and chronic dis-eases, where effects of various dietary components may act synergistically and/or additively. Figure

modified from Steck et al., 2020 (17).

Beyond obesity-related inflammation, dietary habits may contribute to a low-grade inflammatory state associated with increased risk of several cancers (54).

Figure 3 illustrates the main suspected mechanisms for the role of diet and

high-lights the complex and potentially simultaneous biological actions that may con-tribute to the development of chronic diseases in humans.Dietary compounds like fiber can modulate the gut microbiota, for which a convincing association with chronic disease development has been shown, including both cardiovascular disease and colorectal cancer (55, 56). Several biological mechanisms are likely

Chronic disease risk Adverse epigenetic effects Gut microbiota dysbiosis Chronic low-grade inflammation Oxidative stress Hyper-insulinemia Insulin resistance Sex hormones

(25)

10

involved, whereby dietary components act synergistically or additively (17, 57). Inflammation and gut dysbiosis as effects of diet may indirectly affect cardiovas-cular disease and cancer development (57). Metabolic and hormonal responses such as hyperinsulinemia, insulin resistance, abnormal sex hormone metabolites, and oxidative stress are also thought to be involved (17, 57). In pathological con-ditions, inflammation and oxidative stress can occur simultaneously and initiate one another (58), which may complicate distinguishing between potential mech-anisms in observational studies (59).

Cardiovascular disease

Development of myocardial infarction

Acute myocardial infarction is caused by a blockage of the blood flow somewhere in the heart, leading to necrosis of cardiac muscle (60). Blood clots in the epicar-dial artery supplying the heart muscle are a typical culprit, often resulting from atherosclerosis (61). Atherosclerosis is in itself regarded as an inflammatory pro-cess with risk factors related to inflammation (62). In this propro-cess, plaques con-sisting of an accumulation of lipids (cholesterol) and inflammatory substances form on the arterial walls. The molecules encapsulated in the plaque may vary but have some inflammatory properties that increases the risk for plaque growth. This growth may eventually rupture the capsule surrounding the plaque, expos-ing its contents to blood in the blood vessel. A series of reactions may then result in blood clot formation on the plaque, which can grow undetected and lead to infarction. Not all infarctions are caused by blood clots, however; an imbalance in the ratio of blood supply to oxygen demands of the cardiac muscle can lead to ischemia that causes infarction (61).

Incidence and mortality of myocardial infarction

Incidence of first acute myocardial infarction has decreased in Sweden over the past decades, from 32 977 in 1987 to 21 798 in 2018 (63), which is largely at-tributed to primary prevention (64). It is nonetheless still the leading cause of death in Sweden and also globally (65). In the Västerbotten region, the incidence of acute myocardial infarction between 1987 and 2005 was between 900 and 1200 cases per year. Since 2005, there has been a stably decline in incidence down to 674 cases in 2018, although a somewhat smaller decline has been observed in the Norrbotten region (65). Myocardial infarctions are more common in men than in women, but the decline in incidence is observed in both sexes. Higher incidence rates still occur in groups with lower education level - a prevailing dif-ference observed for many years (66).

(26)

11

Lifestyle and risk of myocardial infarction

The large observational study INTERHEART, encompassing 52 countries, found that the same risk factors for myocardial infarction are generally shared in all re-gions of the world (67). These include smoking, dyslipidemia, hypertension, dia-betes, abdominal obesity, psychosocial factors (including depression, perceived stress at home or work, low locus of control, and major negative life events), low consumption of fruits and vegetables, alcohol consumption, and low levels of reg-ular physical activity (67). While men and women share the same risk factors for myocardial infarction, women generally experience their first myocardial infarc-tion later in life, in part probably because the prevalence of these risk factors at younger ages is greater in men compared to women (68). Women also have a protective effect from the female sex hormone estrogen, which is associated with a healthier fat distribution and less atherogenic lipid profile, an effect that is at-tenuated after menopause (69).

Low-grade inflammation and cardiovascular disease

The role of chronic low-grade inflammation in the development of atherosclerosis is well established (61, 62, 70, 71). The link between high cholesterol levels and atherosclerosis is also well-known, although the mechanisms underpinning this were only partially understood for some time. The lipid hypothesis does not seem to exclusively explain atherosclerosis as many individuals with coronary artery disease have demonstrated relatively normal levels of cholesterol (61). Other risk factors and mechanisms involved in the pathogenesis of atherosclerosis, with strokes and myocardial infarctions as consequences, must therefore also be ad-dressed. Several inflammatory biomarkers have been demonstrated to have pre-dictive value for future cardiovascular disease, including tumor necrosis factor-alfa, fibrinogen, serum amyloid A, IL-6, and CRP (62, 72). Observational studies have also suggested that some underlying processes related to low-grade inflam-mation might be relevant for cardiovascular disease, in addition to classic vascu-lar risk factors like smoking and obesity (73).

(27)

12

Cancer

Cancer development

Normal cells can achieve malignant potential in any body tissue, under acquired biological conditions proposed as the hallmarks of cancer (Figure 4) (74). The six original hallmarks are sustaining proliferative signaling, evading growth suppressors, resisting cell death, enabling replicative immortality, inducing an-giogenesis, and activating invasion and metastasis (75). Hanahan and Weinberg later proposed two emerging hallmarks – reprogramming of energy metabolism and evading immune destruction

(74). These eight conditions are thought to be enabled by two char-acteristics of which the development of genomic instability and mutations is the most prominent. An inflam-matory state is proposed as the sec-ond characteristics enabling multi-ple hallmarks and thus contributing to a tumor promoting microenviron-ment (74, 76, 77).

Figure 4. The hallmarks of cancer. Adapted and reprinted from Cell, Vol.144(5), Hanahan D &

Weinberg RA, Hallmarks of cancer. The Next Generation, Pages No 645-674, Copyright (2011), with permission from Elsevier.

The causes of cancer are multifactorial and can involve both inherited genetic susceptibility and environmental exposures including lifestyle factors. Chronic low-grade inflammation is one possible mechanism thought to increase the risk of cancer through several etiological pathways (4). In most cases though, the cause of cancer is unknown. Whatever triggers the cancer development, the final common pathway is the changing of the cell’s genome. Linking genes to the risk of diseases like cancer is complex and can involve several environmental factors, so-called gene-environment interactions. The two key genetic influences are ini-tiation of irreversible DNA damage, and promotion of proliferation and clonal expansion of initiated cells. In addition, epigenetics is a term used to describe heritable changes independent of alterations in the DNA sequence in a cellular phenotype, such as DNA methylation (78). Environmental factors, including life-style, can affect epigenetic alterations (79).

(28)

13

Colorectal cancer development

Most malignant colorectal tumors are adenocarcinomas, meaning that the cancer develops from the glandular epithelial cells of the colon or rectum. Based on origin and expression, one can distinguish between three types of colorectal can-cer. Type 1, or sporadic type, are tumors that appear in individuals not carrying a mutation that makes them susceptible to developing this type of cancer (80). The vast majority of colorectal cancers, 60-80%, are of sporadic type. Type 2, or fa-milial type, are attributed to a family history of colorectal cancer without any ob-vious genetic cancer syndrome and constitute between 20-40% of cases. Type 3, or hereditary type, are caused by inherited germline mutations in known suscep-tibility genes (81) and include hereditary nonpolyposis cancer syndrome (Lynch syndrome) and familial adenomatous polyposis. Although a substantial propor-tion of colorectal cancer cases have a hereditary aspect, lifestyle factors including diet may also contribute to carcinogenesis in these types of tumors.

Colorectal carcinoma has been described as occurring in four stages: initiation, promotion, progression and metastasis, whereby initiation results in irreversible genetic damage that predisposes affected cells to undergo neoplastic transfor-mation (81). The initiated cells further proliferate and induce abnormal growth in the promotion stage. They subsequently undergo genetic and epigenetic alter-ations that can lead to transformation from benign tumor cells into malignant ones with invasive and metastatic potential - that is, to spread to adjacent tissues and distant organs. Although it is difficult to estimate the duration of each phase, colorectal cancer is undoubtedly a disease that takes many years, even decades, to develop. This probably explains the strong association between modifiable life-style factors and risk.

Diagnosis and treatment of colorectal cancer

Symptoms and signs associated with colorectal cancer are blood in the stool, rec-tal bleeding, changes in bowel habits such as diarrhea, constipation, or narrowing of the stool that lasts for more than a few days, cramping or abdominal pain, weakness/fatigue, unintended weight loss and anemia (82). If malignancy is sus-pected, the patient will undergo colonoscopy with biopsy and histological diag-nostics of the biopsy specimen by a pathologist (83). The quality of colonoscopy is an important determining factor for diagnosing colorectal cancer. As popula-tion-level screening programs are introduced, it is expected that more patients will be diagnosed at a precancerous or presymptomatic stage, leading to reduc-tions in incidence rates and better prognosis. However, the incidence among younger people is increasing, and screening programs are typically introduced at approximately 50-60 years of age. Colonoscopy is also an invasive method, po-tentially causing discomfort for the patient and not without risk, and it is more expensive in relation to some other screening methods (82). Early detection

(29)

14

biomarkers or markers of preclinical stages of colorectal cancer are therefore much sought after.

Surgery is the cornerstone and curative treatment of colorectal cancer, usually carried out by laparotomy (84), but the treatment regimen depends on tumor site, tumor stage and molecular pathology testing. The latter is becoming more com-mon for treatment prediction, as is the case with microsatellite instability (MSI) status. MSI can be a marker for identification of Lynch syndrome cases but is even more important as a prognostic and predictive marker in sporadic colorectal can-cer. For tumors at stage I with no spreading beyond the muscularis propria, sur-gery is the single treatment. For tumors that have spread beyond the muscularis propria and sometimes metastasized to local lymph nodes (Stage II-III), adjuvant chemotherapy for three or six months is a common regimen. Preoperative or ne-oadjuvant radiotherapy is common for rectal cancer. For tumors that have spread to other organs and/or tissues such as the liver and lungs (Stage IV), chemother-apy is the cornerstone, whereas surgery of the primary tumor or metastasis is a secondary consideration. Targeted therapy is sometimes used.

Subtypes of colorectal cancer

Colorectal cancer is a heterogeneous disease, and its etiology and prognosis varies dependent on anatomical location (85), disease progress (86), and molecular characteristics (87, 88). Distinct pathways lead to these different tumor charac-teristics, and these pathways may have different risk factors (85, 88, 89). Two well-described pathways in research literature are the adenoma-carcinoma se-quence or chromosomal instability-pathway and the serrated/non- chromosomal instability pathway (90). A proposed inflammatory pathway, driven by chronic inflammation, has been associated with patients with inflammatory bowel dis-ease (81), but is not regarded as an established pathway.

One clinically used predictive marker for colorectal cancer is genetic disruption of the Kirsten rat sarcoma viral oncogene homolog gene (KRAS) gene. This is an essential step in the development of colorectal cancer, and mutation is present in 30-50% of cases (91). KRAS mutation is also a predictive marker for response to the anti-EGFR antibody-based therapies (92). Mutually exclusive mutations in KRAS and B-Raf proto-oncogene serine/threonine kinase (BRAF), occur early in the carcinogenic process, but have different molecular characteristics. BRAF mutations are generally present in 4-18% of colorectal cancer tumors and are associated with higher age, female gender, right-sided location in the colon, poor differentiation, mucinous histology, high degree of infiltrating lympho-cytes, advanced disease stage, and poor prognosis (93). BRAF mutations are also more frequent in microsatellite instability (MSI) tumors (91, 94). Microsat-ellites are repeating units of DNA sequences and MSI is defined as somatic

(30)

15

alterations in microsatellite sequences with insertion or deletion of those repeat units. MSI is therefore a marker for genomic instability and a condition showing hypermutation propensity (91). MSI tumors often acquire BRAF mutations and are also associated with the CpG Island Methylator phenotype (95). The Con-sensus Molecular Subtypes (CMS) proposed four subtypes based on distinct gene expression and molecular profiles: CMS1 (MSI immune), CMS2 (canoni-cal), CMS3 (metabolic), and CMS4 (mesenchymal) (96). Regardless of classifi-cation system, colorectal cancer arises via the aforementioned major pathways, which are reflected by molecular tumor markers. It is therefore important to ac-count for the heterogeneity of the disease when investigating risk, prognosis, and survival. Molecular pathological epidemiology studies on cancer, including subtypes of colorectal cancer, aim to provide insight into etiology and the rela-tionships between risk to disease outcomes (97).

Incidence and mortality of cancer

Of the 18.1 million estimated new cancer cases according to the GLOBOCAN (Global cancer observatory) 2018 database, almost a quarter (4.2 million) oc-curred in the four European regions, despite those regions comprising only 9% of the global population (98). The latest World Cancer Report in 2020 stated that cancer is the first or second leading cause of premature death (i.e. at ages 30-69 years) in a majority of the countries around the world, and third or fourth in most other regions (99). For some of the most common cancer types (lung-, breast-, prostate-, colorectal-, and gastric cancers), incidence trends and mortality are somewhat diversified. Rising breast cancer incidence is related to menarche at earlier ages, advanced maternal age at first birth, and lower parity. Both incidence and mortality are declining for prostate cancer in most countries, while for lung cancer - still the most common cancer type worldwide – there are large variations across countries, which are highly correlated with the prevalence of tobacco smoking. As women generally begin smoking later in life, and in some countries not at all, this is reflected in corresponding rates. Incidence and mortality rates are decreasing for non-cardia stomach cancer as well, though the majority are still attributable to Helicobacter pylori infection. Cardia stomach cancer, on the other hand, is increasing in some populations, most likely due to increasing rates of overweight and obesity.

Incidence and mortality of colorectal cancer

Colorectal cancer was the third most common cancer and fourth leading cause of death worldwide in 2018 (98). Approximately 1.85 million new cases and 881 000 deaths were reported that year, although age-standardized incidence rates vary globally (Figure 5). Incidence rates have increased in countries in transition, whereas rates have stabilized or decreased in highly developed countries. This is particularly notable in older age groups (55 years or older), and likely due to the

(31)

16

introduction of screening programs for earlier detection (99). Younger age groups (sometimes referred to as <40 years and sometimes as <50 years of age), have shown increases in highly developed countries (including Sweden), however, driven by lifestyle-related risk factors (100, 101). As disease stage at diagnosis is the main determinant of survival in colorectal cancer, high-quality screening pro-grams and well-functioning health care systems are critical factors to improve disease outcomes. Public health strategies promoting physical activity and healthy eating are fundamental for reducing the colorectal cancer global burden (99).

Figure 5. Age-standardized incidence rates (per 100 000) of colorectal cancer in both sexes. Data from

GLOBOCANÓ 2018, IARC (International Agency of Research on Cancer) (Assessed 200210). Countries in grey lack data.

Lifestyle and cancer risk

The current evidence for lifestyle-related cancer risk factors including dietary in-take, obesity, alcohol consumption, physical activity, and smoking habits varies substantially dependent upon cancer site. The World Cancer Research Fund, to-gether with the American Institute for Cancer Research, regularly update their expert report - the Continuous Update Project (CUP) – summarizing the global scientific evidence (102), presented in brief below.

The single most important risk factor for lung cancer is tobacco smoking, ac-counting for 80-85% of lung cancer cases globally (99). Consumption of red- and processed meat, as well as alcohol consumption, can also increase risk, while con-suming foods that contain retinol, beta-carotene or carotenoids, or being physi-cally active can decrease the risk (102). In current smokers, consuming fruit, veg-etables and other foods containing vitamin C might decrease the risk.

(32)

17

There is little convincing evidence that specific dietary factors or dietary patterns increase the risk of female breast cancer, except for alcohol consumption (102). Some evidence exists for decreased risk associated to consuming non-starchy vegetables, carotenoids, dairy products, diets high in calcium, and for being phys-ically active. The risk of early premenopausal breast cancer (about age 18-30 years) may be reduced in overweight or obese women, or through vigorous phys-ical activity. On the contrary, risk may increase for postmenopausal breast cancer if one is overweight/obese and/or having a higher attained height throughout adulthood (>30 years and henceforth). Alcohol consumption is also associated to increased risk of postmenopausal breast cancer.

Strong evidence is lacking for dietary factors in relation to the risk of prostate cancer, although being tall, overweight and/or obese increases the risk (102). Some protective effects of dairy products and diets high in calcium have been sug-gested for prostate cancer. For pancreas cancer, a rare but aggressive type of can-cer, excess body fat is believed to increase risk, and some evidence exists for red- and processed meat, saturated fat, alcohol, and fructose-containing foods and beverages. Alcohol and salt-preserved foods increase stomach cancer risk and be-ing overweight or obese increases the risk for cardia stomach cancer specifically, and weaker evidence exists for processed meat (in non-cardia stomach cancer), grilled/barbequed meat and fish, and low fruit intake. The decreasing trend in gastric cancer incidence is attributable to a decrease in Helicobacter pylori infec-tion rates, as well as lower consumpinfec-tion of salt-preserved foods due to more wide-spread use of refrigeration.

Lifestyle and colorectal cancer risk

The long preclinical stage of colorectal cancer corresponds to its high association to lifestyle factors, in particular diet, height, and body fatness. Summary evidence from the latest CUP report and corresponding updated meta-analysis and review is presented in Table 1 (102, 103).

In addition to the risk factors listed in Table 1, smoking is a preventable risk factor for colorectal cancer. It is estimated that smoking two packs (40 cigarettes) per day increases the risk of colorectal cancer by approximately 40% and nearly doubles the risk of colorectal cancer-specific death (102). Also, long-term use of non-steroidal anti-inflammatory drugs and Aspirin (at least 75 mg/day for five years or more) can reduce the risk (104) . At the individual level, a specific cause of colorectal cancer is generally not distinguishable.

(33)

18

Table 1. Risk factors and preventive lifestyle factors for colorectal cancer with strong evidence (Continuous Update Project, CUP 2018) (102, 103).

In cr e a se d r is k

Lifestyle factor Meta-analysis relative risk

Body fatness a _{1.05 (1.03-1.07) per 5 kg/m}2_{increase in BMI}(105)

Adult attained height 1.05 (1.02-1.07) per 5 cm increase (105)

Alcohol consumption 1.07 (1.05-1.08) for 10g/day increase (106)

Processed meat 1.16 (1.08-1.16) for each 50g/day increase (103)

Red meat 1.12 (1.00-1.25) for each 100g/day (103)

De cr e a se d r is

k Wholegrain intake 0.83 for each 90g/day increase of whole grains (103)

Dietary fiber intake 0.91 per 10g per day (103)

Dairy products intake

0.87 (0.83-0.90) for each 400g/day increase

(103)

Calcium supplement

0.91 (0.86-0.98) for each 300mg/day increase

(107)

Physical activity b _{0.84 (0.78-0.91) highest vs lowest recreational}

phys-ical activity (103)

a_{Measured as BMI (body mass index), waist circumference, or waist-to-hip ratio.}

b _{Mainly related to colon cancer risk}

Low-grade inflammation and cancer

The link between inflammation and cancer is supported by numerous lines of ev-idence, and chronic low-grade inflammation promoting cancer development can be caused by major lifestyle-related risk factors, including obesity, alcohol and diet (108). The etiological pathways for inflammation in cancer are not entirely understood, but transcription factors, cytokines, chemokines and infiltrating leu-kocytes appear to be key orchestrators in inflammation-mediated tumor progres-sion (77). Inflammation can, however, play dual roles in cancer progress as cyto-toxic T-lymphocytes and regulatory T-cells may regulate or suppress cancer cells and thereby contribute to an anti-tumorigenic response. These dual functions of inflammatory mechanisms provide a challenge in navigating complex molecular pathways, for both prevention and treatment goals (108). Identifying causal in-flammatory biomarkers or biomarker patterns would enable better identification of risk populations (109). Various classes of inflammation-related biomarkers have been investigated for potential links to cancer, including cytokines, im-mune-related effectors, acute-phase proteins, reactive oxygen species, nitrogen species, and prostaglandins (4). Specific biomarkers include CRP, IL-6, interleu-kin-10, albumin, haptoglobin, fibrinogen, and whole blood leukocyte count (4, 109, 110), but there are currently no standard individual biomarkers with robust evidence and without significant bias (1). While inflammation increases with age, not all investigated inflammatory markers are associated with aging. Frailty and down-regulation of immune system components may explain part of the equation and not just biological aging itself (1). It is desirable to identify specific bi-omarkers of inflammation associated to cancer risk for pre-diagnostic purposes.

(34)

19

Aims

The aim of this thesis was to investigate the role of systemic, chronic, low-grade inflammation in the risk of cardiovascular disease and cancer, in particular colo-rectal cancer. A special focus was placed on the impact of dietary patterns and biological markers of inflammation. All studies were based on population-based cohorts and used prospectively collected data and/or blood samples.

The theoretical hypothesis framework is illustrated in Figure 6. The specific aims for each paper were:

Paper I

To investigate the Dietary inflammatory index in relation to the risk of acute myocardial infarction and to validate the Dietary inflammatory index against plasma levels of inflammatory biomarkers.

Paper II

To investigate two dietary pattern scores - the Dietary inflammatory index and the Mediterranean diet score - in relation to the risk of developing cancer, in a large cohort study with repeated measures.

Paper III

To investigate circulating levels of CRP in relation to the risk of molecular and clinical subtypes of colorectal cancer.

Paper IV

To identify data-driven dietary patterns derived from exploratory and confirma-tory factor analysis, and to assess their association to pre-diagnostic plasma me-tabolite profiles and to colorectal cancer risk. Hypothesis-driven dietary compo-nents, i.e. the most well-established dietary risk factors, were included for com-parison.

Figure 6. Theoretical hypothesis framework for this thesis.

Hypothet-ical relations between die-tary patterns and disease risk through inflammatory processes that can be meas-ured with inflammatory bi-omarkers, as well as other potential mechanisms for which relevant biomarkers or metabolites can be ob-tained by using metabo-lomics.

CRP; C-reactive protein, Il-6; Interleukin-6.

Dietary patterns _diseasesChronic

CRP IL-6 Metabolites Colorectal cancer Cardiovascular disease Cancer Low-grade

Inflammation and risk of chronic diseases with a focus on colorectal cancer and the impact of dietary patterns