Cohort Profile: The Melbourne Collaborative Cohort Study (Health 2020)

Why was the cohort set up?

The Melbourne Collaborative Cohort Study (MCCS), also known as Health 2020, was planned in the late 1980s and established in the early 1990s as an omnibus cohort to investigate prospectively the roles of diet and lifestyle in causing cancer and other non-communicable diseases.¹ It was developed contemporaneously with the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort² at a time when diet and nutrition were considered important to cancer causation,³ but detailed information adequate to inform prevention was scant. The literature was dominated by inconsistent evidence generated by a large number of small case-control studies which had problems not only with statistical power and dietary assessment but also, more importantly, with forms of information bias to which such studies are prone. A prospective design was chosen to reduce these biases. We also considered that a contributing factor to the modest strengths of association between dietary factors and cancer observed by previous cohort studies, might be related to the limited range of dietary intakes within a given population. We sought ways in which we might address this and, as Melbourne had a relatively large number of migrants from Italy and Greece with distinct dietary and lifestyle differences from the majority of British descent, we deliberately targeted them for recruitment in order to broaden the range of observations of the measured lifestyle risk factors.

Our original research proposal encompassed a broad range of research questions, including the full gamut of questions concerning the influence of individual foods and food groups, macronutrients and micronutrients on the risk of cancer, principally of the colorectum, breast and prostate, at least in the short term. Early questions, for example, tested the conventional wisdom regarding: the possible protective effect of fruit and vegetables and the harmful effect of red and processed meats; the effect of different dietary fats (especially oleic acid); the role of dietary fibre; and the role of alcoholic beverages. We were also interested in the role of energy balance, physical activity, overweight and obesity. To this end we took direct anthropometric and clinical measurements of all participants, including bioimpedance to estimate fat and lean mass separately, and blood pressure. Using plasma samples collected at baseline, we also had questions concerning markers of diet such as antioxidants and concentrations of circulating hormones, not only steroid homones but also peptide hormones such as insulin-like growth factor 1 (IGF-1) and their binding proteins.

As we addressed these questions ourselves and contributed data to consortia to increase the range and precision of estimates, we adapted our data collection at follow-up waves to refine our approach and better answer remaining questions. We also repeated blood sampling and physical measurements to be able to address associations with changes over time. What we did not anticipate at the inception of the MCCS was the rapid growth in genomics and the value of the blood (collected to measure dietary and other markers) as a source of DNA to examine genetic risk factors and their potential interaction with lifestyle-related exposures, which has been our principal focus for the past decade or so.

Who is in the cohort?

Between 1990 and 1994 (baseline), 41 513 residents of the Melbourne metropolitan area (24 469 women and 17 044 men), with an average age of 55 years (range 27–76; 99% aged between 40 and 69) were recruited to the study. All participants were of White European origin; most (69%) were born in Australia or New Zealand, 13% were born in Italy, 11% in Greece and 6% in the UK.

The majority of participants were identified from the Victorian Electoral Enrolment Register (enrolment to vote is compulsory for all adult Australian citizens) and the Melbourne metropolitan phone directory, and were invited by letter to participate. The main recruitment clinic was organized at the premises of Cancer Council Victoria, proximal to the Melbourne central business district (CBD). Additional strategies were adopted to recruit migrants, bearing in mind that those who had not taken out Australian citizenship would not be on the electoral register: announcements were made on multicultural radio and in newspapers, clubs and churches; and a mobile recruitment unit was also driven to major Italian and Greek festivals in Melbourne and to the premises of manufacturing companies with a large number of migrant employees. Participants were invited to encourage friends and family to participate in the study. The mobile unit was also used to process Australian-born participants closer to their area of residence. The proportion of responders was not recorded for any of these recruitment strategies. Compared with the population of metropolitan Melbourne aged 40–69 years at the time of recruitment,⁴ MCCS participants were more likely to be older and female (59%). As expected, there was a greater proportion of Greek- and Italian-born participants and fewer Australian-born participants compared with the general populations of Victoria and Australia in the same age range (77% and 79% Australian-born, respectively).^5,6

How often have the participants been followed-up?

Active follow-up

The first wave of active follow-up (follow-up 1) occurred between 1995 and 1998, approximately 4 years after recruitment for each participant. A total of 36 335 (88%) of the original participants, 21 664 women and 14 671 men, participated either by mail or phone; 778 had died before they could be contacted. A second wave of active follow-up (follow-up 2) occurred between 2003 and 2007. A total of 28 240 (68%) of the original participants, 17 138 women and 11 102 men, attended clinics organized at the Cancer Council Victoria premises near the Melbourne CBD for face-to face interviews and physical measurements; 3007 had died before they could be contacted. Those who attended this second wave were younger and more likely to be born in an English-speaking country, and had higher socioeconomic position than those who did not. No further active data collection is planned.

Every year participants are sent a study newsletter, which includes a reminder to contact the study coordinator to update contact information. This has been the case since 2005; before that, newsletters were sent every 2 years.

Data and sample collection

Information was collected at baseline attendance [http://www.pedigree.org.au/pedigree-studies/health2020/pages/h2020databooks.aspx] by physical measurement, interviewer-administered questionnaires on lifestyle, personal medical history and medications taken, and a self-administered food frequency questionnaire.⁷ Blood samples were collected from 41 133 participants; plasma (N = 40 524 participants), peripheral blood mononuclear cells (N = 9832) and buffy coats (N = 635) were separated and stored in liquid nitrogen. From the second year of recruitment, dried blood spots (N = 30 663) were stored on Guthrie card diagnostic cellulose filter paper in lieu of blood mononuclear cells or buffy coats. At follow-up 1, information on lifestyle, medical history and diet was collected by self-administered or telephone interviewer-administered questionnaire; no physical measurements or blood were taken. At follow-up 2, physical measurements were taken, interviewer- or self-administered questionnaires were used to collect lifestyle and medical history information (including medications taken) and diet was measured by self-administered food frequency questionnaire. Blood was collected from 26 824 follow-up 2 participants (99%) and stored on Guthrie cards for 26 522. Plasma (N = 26 318), serum (N = 26 606) and buffy coats (N = 26 299) were aliquoted and stored in liquid nitrogen. Formalin-fixed paraffin-embedded tissue has been collected from 3070 tumours diagnosed in cohort participants, the vast majority being breast, colorectal and prostate cancers. Further details of the samples and information collected are provided in Table 1.

Table 1

Data and samples collected at baseline and follow-up for participants in the Melbourne Collaborative Cohort Study

Measurement/sample	Baseline (1990–94)	Follow-up 1 (1995–98)	Follow-up 2 (2003–07)
Self-reported by questionnaire
Smoking	✓	✓	✓
Alcohol	✓		✓
Physical activity	✓		✓
Socioeconomic statusa	✓		✓
Family history of cancerb	✓	✓	✓
Health conditionsc	✓	✓	✓
Medication used	✓	✓
Psychosocial healthe	✓	✓	✓
Diet (food frequency)	✓	✓	✓
Diet (derived nutrients)f	✓		✓
Reproductive history (women)	✓	✓	✓
Hormone use (women)	✓	✓	✓
Physical measurements
Anthropometric measuresg	✓	✓	✓
Blood pressure, pulse	✓		✓
Samples
DNA (blood)	✓		✓
Plasma	✓		✓
Serum			✓
Tumourh
Laboratory measurements
Total glucose	✓		✓
Total cholesterol	✓		✓
Antioxidantsi	✓
Fatty acidsi	✓
Hormonesi	✓
Binding proteinsi	✓
Insulini	✓
Triglyceridesi	✓
HDL cholesteroli	✓
Homocysteinei	✓
C-reactive proteini	✓
Vitamin Dj	✓
B vitaminsk	✓
Serum ferritinl	✓		✓
Other data
Retina imagesm			✓
Mammograms (MD sub-study)m
SNPs (GWAS)n	✓		✓
DNA methylation (HM450K)o	✓		✓

^aDeciles of the Index of Relative Socioeconomic Disadvantage (IRSD) based on the smallest geographical area of output (average population of 400 persons) defined by the population census, to which each participant’s usual residential address at baseline and follow-up 2 have been mapped.

^bAt baseline, type of cancer was not specified and only first-degree relatives were considered; similar information was collected at follow-up 1, but with type of cancer specified; at follow-up 2, first-degree family histories of colorectal, breast and prostate cancer, as well as heart attack and stroke, were reported.

^cWeight change, asthma, angina, hypertension, diabetes, arthritis, cancer, kidney stones, gallstones, heart attack, stroke.

^dBy drug name; aspirin, NSAIDS, paracetamol collected specifically at follow-up 1.

^eQuestions about feelings at baseline and follow-up 1; psychological distress or anxiety (Kessler 10), social support, quality of life (SF12) and ability to perform activities of daily living assessed at follow-up 2.

^fCollected via a food frequency questionnaire developed specifically for this study.²

^gHeight and bioimpedance (to determine lean and fat mass) were collected at baseline only; weight, waist circumference and hip circumference were collected at baseline and follow-up 2; weight and waist circumference were self-reported at follow-up 1; photocopies of hands were taken at follow-up 2, to measure the ratio of second to fourth digit length.

^hFormalin-fixed paraffin-embedded samples have been collected for incident cases of colorectal (N = 869), prostate (N = 1074), breast (N = 1005), kidney (N = 206) and endometrial cancer (N = 136).

ⁱMeasured in plasma for the main case-cohort sub-study (see description in Sub-studies).

^jMeasured from dried blood spots for participants (N = 7045) selected for purpose-designed case-cohort sub-study (based on only participants with dried blood spot samples).

^kMeasured in plasma for selected nested case-control studies (lung, urothelial cell carcinoma, kidney).

^lMeaured in plasma at baseline and serum at follow-up 2 as part of the HealthIron sub-study (see description in Sub-studies).

^mTaken as part of the CERA sub-study (see description in Sub-studies).

ⁿSee description in Genetic data.

^oSee description in Epigenetic data.

A food frequency questionnaire was developed specifically to measure usual dietary intakes by cohort participants, with the food list based on weighed food records completed by a group of 810 men and women representing the main country of birth groups within the MCCS.⁷ The most recent version of the dietary questionnaire (DQES v3.2)⁸ [http://www.cancervic.org.au/research/epidemiology/nutritional_assessment_services] is available for use on a fee-for-service basis. It estimates nutritional intake based on 144 foods and beverages grouped as grain-based foods, dairy foods and fats, meat, fish and seafood, fruit, vegetables, miscellaneous, tea/coffee and alcoholic beverages in electronic (online) format that can be self- or interviewer-administered.

Passive follow-up

Participant addresses are updated using information from the Victorian Electoral Enrolment Register, which is received every 6 months. Incident diagnosis of cancer and deaths are identified by at least annual record linkage to the Victorian Cancer Registry (VCR) and the Victorian Registry of Births, Deaths and Marriages. Both registries are considered complete, with notification to the VCR of all cancers diagnosed in Victoria mandated by legislation since 1981. To capture these events for participants who may have moved interstate, these Victorian linkages are complemented by at least 2-yearly record linkage to the National Death Index (NDI) and the Australian Cancer Database, compiled by the Australian Institute of Health and Welfare (to which all state and territory cancer and death registries in Australia contribute). Linked NDI data also include cause of death. Altogether, 11 222 incident cancers (1553 breast; 1262 colorectal; 1975 prostate; 761 lung) were diagnosed in MCCS participants to June 2015 and 10 304 participants had died by December 2015.

Sub-studies

Case-cohort study

A case-cohort sub-study was defined⁹ in 2003 to provide a more economical design within which to perform expensive assays to measure various molecules in the stored plasma and to analyse their associations with selected diseases. The sub-study includes all incident cases to 30 June 2002 of prostate cancer (n = 605), breast cancer (n = 576), colorectal cancer (n = 525), type 2 diabetes (n = 402)¹⁰ and cardiovascular death (n = 532), and a random sample of all MCCS participants, making a total of 6950 MCCS participants. Plasma-based assays included biomarkers of dietary intake (antioxidants^11,12 and fatty acids^10,13–16), steroid sex hormones,^17,18 homocysteine, insulin, triglycerides, high-density lipoprotein (HDL) cholesterol, insulin-like growth factor 1 (IGF-1) and insulin-like growth factor binding protein 3 (IGFBP3),^19,20 nuclear magnetic resonance (NMR)-determined lipoprotein subclass levels²¹ and C-reactive protein (CRP).

Centre for Eye Research Australia (CERA)

At follow-up 2, non-mydriatic retinal photography was performed by collaborators from the Centre for Eye Research Australia (CERA) in order to study age-related macular degeneration (AMD) and early markers of cardiovascular disease (CVD). A total of 22 405 follow-up 2 participants (80%) took part in this study, with digital retina images from 21 287 passing quality control.²²

HeathIron

HealthIron, a study of hereditary haemochromatosis, was designed to genotype 31 192 participants of northern European descent and subsequently to assess clinically a random sample of 1438 participants stratified according to high-iron Fe (HFE) genotypes C282Y and H63D. There were 203 C282Y homozygotes.²³ The strengths of this study were its size and the availability of stored plasma from baseline attendance, which permitted repeated measures of iron indices 12 years apart.^24,25

Mammographic density (MD)

In 2009, record linkage was performed between the MCCS and BreastScreen Victoria, which is part of Australia’s national breast cancer screening programme. BreastScreen was established in 1992 to offer free biennial mammographic screening targeted at women aged 50–69 years. Of the 24 469 women in the MCCS, 20 444 (84%) had attended BreastScreen Victoria at least once and, from these, a nested case-control study was designed including 800 incident breast cancer cases (680 invasive) diagnosed before 2008, each matched to four controls. Of women initially selected in this case-control study, 69% had had a mammogram between baseline and reference age (age at diagnosis of the case in each case-control matched set), and for each of these, the mammogram closest to baseline was digitized; from the image, total breast area and dense area were determined using Cumulus.⁹

Genetic data

Genome-wide genotyping has been conducted on germline DNA samples from a subset of MCCS participants using the Illumina OncoArray beadchip.²⁶ These include prevalent and incident cases of breast (N = 1700), prostate (N = 1870) and colorectal cancer (N = 1492), and incident cases of urothelial cell carcinoma (N = 460), as well as 3583 controls matched to a subset of these using incidence density sampling. OncoArray includes more than 500 000 genetic variants, including a 250K GWAS backbone, and using imputation to the 1000 genomes reference panel (Phase 3), has generated genotypes for ∼21M single nucleotide polymorphisms (SNPs).

Epigenetic data

Genome-wide DNA methylation has been determined using the Illumina HumanMethylation450 BeadChip array in peripheral blood DNA samples collected at baseline from participants in eight nested case-control studies. These include 498 breast cancer,²⁷ 877 prostate cancer, 842 colorectal cancer, 460 urothelial cell carcinoma,²⁸ 143 kidney cancer, 179 gastric cancer, 376 lung cancer²⁹ and 469 mature B-cell neoplasm³⁰ case-control pairs. All cases were unaffected at baseline. OncoArray genotype data are available for all participants in the first four of these studies. Methylation has also been measured in DNA from peripheral blood samples collected at follow-up 2 for a subset of controls from these studies so that we have repeated measures, on average 10 years apart for 1100 participants.

What has it found? Key findings and publications

MCCS publications are listed at [http://www.cancervic.org.au/research/epidemiology/cec-publications].

When the MCCS was being planned, it was thought that diet played an important role in cancer development.³ In 1997 the World Cancer Research Fund report concluded that there was convincing evidence that consumption of fruit and vegetables decreased the risk of five cancers, on the basis of data from case-control studies.³¹ Ten years later, the updated report found no convincing evidence for an association between intake of fruit and vegetables with cancer at any site, based on more recent prospective data.³² In 2014 the review was updated again, with the new evidence available supporting only probable associations between consumption of fruit and vegetables and reduced cancer risk.³³ The MCCS investigators chose not to repeat these analyses.

The MCCS has been used to investigate associations between diet and a wide range of health outcomes. Dietary patterns derived using both a posteriori (factor analysis) and a priori (Mediterranean Diet Score, Dietary Inflammatory Index and Alternative Healthy Eating Index) have been investigated in relation to type 2 diabetes,³⁴ lung,³⁵ urothelial cell³⁶ and breast cancers,³⁷ successful ageing,³⁸ psychological distress,³⁹ CVD mortality⁴⁰ and AMD;⁴¹ all findings support the health benefits of diets based on fruits, vegetables and limited meat intake.

B vitamins and other dietary components may play a role in carcinogenesis as they are involved in one-carbon (1-C) metabolism, which is necessary for DNA replication, DNA repair, and regulation of gene expression. The MCCS has been used to assess associations between intakes of these nutrients and risk of lung,⁴² breast,¹⁴ prostate,⁴³ colorectal⁴⁴ and gastric cancer and urothelial cell carcinoma. The results have been inconsistent and do not support increasing intakes of these nutrients as a means of reducing cancer risk.

Possible associations between fatty acids and health outcomes have been explored using both dietary intake and biomarker data. For breast,¹⁴ prostate¹⁵ and colorectal cancers,¹⁶ diabetes,¹⁰ lesions in knee cartilage and bone associated with osteoarthritis,⁴⁵ abdominal aortic calcification (a risk factor for CVD)⁴⁶ and AMD,⁴⁷ there is some evidence that omega-3 fatty acids (from fish and walnuts/flaxseeds) are preferable to omega-6 (seed oils and margarines made from these).

In several analyses the MCCS identified the importance of measures of body size to cancer risk, though the strength of association varied by cancer type^48–54 (Figure 1). Obesity was also related to mortality^55,56 and disability.^57,58 Further, more vigorous physical activity, a diet high in carbohydrate and fibre and relatively low in fat and protein, and limited alcohol consumption were associated with less weight gain after middle age.⁵⁹ However, weight loss after middle age was associated with increased mortality.⁶⁰ Having direct measurements of waist and hip circumferences was a particular strength of this work, as these measures are rarely available in large cohort studies.

Figure 1

Relative risks of cancer associated with waist circumference.

The MCCS, in addition to collecting information on consumption of alcoholic beverages at baseline and follow-up, also collected extensive data on historical drinking patterns. These data have been used to assess the associated risks of several organ-specific cancers and subtypes therein^61–63 (Figure 2 ), as well as all-cause mortality.⁶⁴

Figure 2

Relative risks of cancer associated with alcohol intake.

The most often cited papers have come from pooling data with other cohorts in international consortia; examples include: the finding that overweight and obesity (and possibly underweight) are associated with increased all-cause mortality;⁶⁵ and the identification of common genetic variants associated with risk of breast and prostate cancer.^66,67

The MCCS collaborates internationally with many other researchers and research consortia involved in cancer and non-cancer research. These include the National Cancer Institute (NCI) Cohort Consortium [http://epi.grants.cancer.gov/Consortia/cohort.html], the Pooling Project of Prospective Studies of Diet and Cancer [http://www.hsph.harvard.edu/pooling-project/] and the Asia Pacific Cohort Studies Collaboration [http://www.apcsc.net/], the latter focused primarily on CVD.

Through its participation in the Breast and Prostate Cancer Cohort Consortium (BPC3), the Breast Cancer Association Consortium (BCAC) and the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL), the MCCS has contributed to the identification of the vast majority of the ∼100 common genetic susceptibility variants for each of breast^67–72 and prostate cancer.^66,73–77 Further, analyses of gene-gene and gene-environment interactions have consistently shown that it seems reasonable to assume these established genetic risk factors combine independently, both with each other^78–80 and with other potential and established risk factors,^78,81–85 to influence cancer risk. MCCS participation in collaborations with the Ovarian Cancer Association Consortium (OCAC) and Pancreatic Cancer Cohort Consortium (PanScan) has contributed to the discovery of many novel susceptibility variants for these rarer cancers.^86–88

What are the main strengths and weaknesses?

The main strengths of the MCCS include its size, its virtually complete blood sampling, and direct physical measurements and validated questionnaires to capture lifestyle exposures. The inclusion of southern European migrants also widens the range of social and lifestyle measures. With substantive follow-up since recruitment and low loss to follow-up (only 96 participants are known to have left Australia), the MCCS provides a rich resource for the prospective investigation of the role of diet, other lifestyle factors, genetics and epigenetics in the aetiology of cancer. With over 41 000 participants, the MCCS has adequate power to conduct research over a spectrum of exposures and outcomes, but it has limited power for less common outcomes and weak associations. To increase available power to address these interests, the MCCS has contributed resources to international consortia and has in this way, for example, further enabled the global effort to study the dietary and genetic aetiology of disease. The MCCS has extensive epidemiological data at three time points and direct physical measurements and blood samples collected at two time points. DNA has also been extracted for most of the cohort for genetic analysis, with genome-wide genotyping and blood DNA methylation already determined for a substantial subset. It also has ongoing and complete passive follow-up for cancer diagnosis, death and cause of death.

Can I get hold of the data? Where can I find out more?

Researchers interested in collaborative projects using the MCCS can contact the access committee by e-mail at [[email protected]]. The MCCS has a formal process for permitting access to its data and materials for collaborative research proposals. If the proposal has been approved by an appropriate ethics committee, an expression of interest can be submitted for review by the MCCS access committee. If approved, a data request can then be submitted for costing. Data and biospecimens are sent to collaborators by secure file transfer under standard written agreements, with fees charged on a cost-recovery basis. The lead investigator is required to submit annual reports for the duration of the project. The MCCS also has formal publication registration processes to ensure that all collaborators are given the opportunity to contribute as authors or are otherwise appropriately acknowledged. These processes are described at [http://www.pedigree.org.au/].

MCCS (Health 2020) profile in a nutshell

• The MCCS is an omnibus cohort set up to investigate prospectively the roles of diet and lifestyle in causing cancer and other non-communicable diseases.

• Between 1990 and 1994 (baseline), 41 513 residents of the Melbourne metropolitan area (59% women) with an average age of 55 years (range 27–76; 99% aged between 40 and 69) were recruited.

• Follow-up has included two waves of data collection, in 1995–98 and 2003–07. Record linkage to the regional and national cancer and death registries is done annually for all living participants.

• The dataset comprises a wide range of health and lifestyle information, including diet, anthropometric and clinical measurements, blood samples, genetic and epigenetic information, biomarkers and tumour samples.

• Formal processes exist to access the MCCS data and materials for collaborative research [http://www.pedigree.org.au/].

Funding

Cohort recruitment was funded by Cancer Council Victoria [http://www.cancervic.org.au/] and VicHealth [https://www.vichealth.vic.gov.au/]. The MCCS was further supported by grants 209057, 251553 and 504711 from the Australian National Health and Medical Research Council (NHMRC) [http://www.nhmrc.gov.au/] and ongoing follow-up and data management has been funded by Cancer Council Victoria since 1995. Research related to mammographic density was supported by the Victorian Breast Cancer Research Consortium (VBCRC).

Acknowledgements

The MCCS was made possible by the contribution of many people, including the original investigators, the teams that recruited the participants and continue working on follow-up, and the many thousands of Melbourne residents who continue to participate in the study.

Conflict of interest: None declared.