June 29, 2018

Cancer Epidemiology Program: Pan Life-Span Population Studies Beyond Odds Ratios and P-Values

“No man is an island, unto itself; every man is a piece of the continent, a part of the main” – John Donne

The fundamental unit in cancer treatment is the individual patient, but the unseen hand that informs and guides the entire continuum of cancer care--from screening to risk assessment, treatment to survivorship--is knowledge gained from analysis of patient populations. Patients with cancer are not islands; they are indeed parts of the “main”, meticulously grouped, classified, and stratified by cancer epidemiologists.

In the age of personalized oncology, members of the Cancer Epidemiology (CE) program at DF/HCC are tackling the challenge of integrating molecular, genetic, and metabolomic parameters, along with traditional epidemiological, sociodemographic, and clinicopathologic features, to estimate risk, finesse interventions, untangle genetic underpinnings, and identify care inequities in cancer. The long history of trans-disciplinary and inter-programmatic collaborations is a strength of the CE program, according to Lorelei Mucci, MPH, ScD (HSPH), who leads the program along with Andrew Chan, MD, PMH (MGH), and Heather Eliassen, ScD (BWH).

Four overarching themes tie together recent work from CE program members: primary and secondary precision cancer prevention, next-generation genomic cancer epidemiology, metabolic signatures in cancer, and epidemiological studies of cancer survivorship. CE program members identified novel tissue biomarkers in colorectal, prostate, and breast cancers that are associated with molecular tumor subtypes, which can then be harnessed in precision cancer prevention trials. Data and specimens from four large DF/HCC cohorts collected for decades are also facilitating large-scale analyses of circulating biomarkers and metabolites in blood. One study led by Brian Wolpin, MD, MPH (DFCI) found that elevated plasma levels of branched-chain amino acids (BCAAs) were associated with a more than two-fold increased risk of pancreatic cancer.CE and Prostate Cancer program members collaborated to pinpoint metabolic parameters and circulating metabolic factors, such as insulin and adiponectin, associated with developing more aggressive types of prostate cancer.Ongoing studies are addressing whether the aberrant metabolic profiles are drivers or prognostic markers for prostate cancer. 

Mucci noted that the altered metabolism may eventually be useful for identifying individuals earlier, when they are not exhibiting signs and symptoms of disease. She also indicated that the studies may delineate the mechanism through which certain lifestyle factors, such as statin use, may impact cancer risk and survival, and the potential role of altered metabolism in mediating these effects. A recent study by Wolpin and members of the Gastrointestinal Malignancies and Cancer Data Sciences programs incorporated markers associated with cigarette smoking in their assessment of association between pancreatic cancer survival and cigarette smoking. They found that among patients with circulating levels of cotinine, a metabolite of nicotine that is proportional to tobacco smoke exposure measured within 5 years before diagnosis of pancreatic cancer, heavy smokers had a 2.5-times increased risk of pancreatic cancer death.“This was one of the first prospective studies to look at survival in pancreatic cancer as a function of cigarette smoking,” Mucci noted.

“There has been an increasing understanding about the role of genetic predisposition in developing certain types of cancer--not family syndromes, but genetic factors or polymorphisms that are associated with increased cancer risk,” stated Mucci, speaking to the recent expansion of genomic epidemiology studies. CE program members have led or are leading international consortia in the genetic epidemiology of cancer, including the NCI Breast and Prostate Cancer Cohort Consortium (BPC3Peter Kraft, PhD [HSPH]), the Genetic Associations and Mechanisms in Oncology Consortium (GAME-On; Peter Kraft), Markers of Density (MODE; Rulla Tamimi, ScD [BWH]) and the Endometrial Cancer Consortium (E2C2Immaculata De Vivo, MPH, PhD [BWH]). 

Portrait of Cancer Risk in a Palette of Genetic Variants 

“There is a lot of enthusiasm about using the knowledge gathered regarding inherited genetic information to stratify people according to their risk, and on the basis of predicted risk, make different choices in terms of intervention, screening, and treatment strategies,” Kraft said. Kraft is a co-senior author on two of the largest genome-wide association studies (GWAS) to date on the inherited genetic contribution to breast cancer risk.4,5 His study identified 65 new loci associated with overall breast cancer risk,“nearly doubling the known breast cancer risk loci”. This study, involving 550 researchers from 300 institutions across six continents, included data from 10 million sites across the genome in a large cohort of over 250,000 women, about half of whom had breast cancer. “Based on those data we can now identify a proportion of women who have notably increased risk of breast cancer. The analogy is [that] it’s as if we discovered a new BRCA gene, although it is made up of many, many alleles across the genome, rather than just one allele with high effect,” Kraft explained. Women who inherit specific mutations in the BRCA1/2 genes have an increased breast cancer risk, with a cumulative lifetime risk of 60-80%. While each of the single-nucleotide polymorphisms (SNPs) identified in the GWAS study had low penetrance, women carrying subsets or combinations of the SNPs had a significantly higher risk. Kraft noted that these numbers are a conservative estimate of new breast cancer risk loci, and he expects the number of genetic variants that contribute to inherited breast cancer risk will approach 1000. 

Kraft and Wolpin are collaborating to develop risk prediction models for pancreatic cancer that incorporate relevant epidemiological risk factors, SNPs, metabolomics, and circulating biomarkers. In a recent Nature Communications article, Kraft, Wolpin, and colleagues identified 5 new pancreatic cancer susceptibility loci in the largest GWAS study to date.

Improved risk-prediction models can help drive screening and prevention interventions for those with high cancer risk. The precise and optimal incorporation of stratified risk estimation in clinical practice and intervention delivery is an “interesting and difficult problem,” Kraft noted. In an incipient study, Kraft, along with co-investigator Alexander Gusev, PhD (DFCI), is also evaluating the link between inherited germline genetic variations and tumor somatic profiles across a wide range of tumor types. 

Transgenerational Cancer Risk Tracking

It takes GUTS to follow cancer risks across generations, but the GUTS that Tamimi, who is interested in molecular and radiologic features linked to breast cancer development, is working with stands for the Growing UToday Study (GUTS). GUTS enrollees are children of the Nurses Health Study (NHS) II cohort. The NHS cohorts are among the largest prospective investigations of risk factors for major diseases in women. With the information from biospecimens and epidemiological data gathered from NHS cohorts for over 40 years, researchers “are able to, in a reliable and prospective way, determine how certain circulating markers or common genetic variants and most recently, features on a mammogram, relate to later risk of developing breast cancer,” Tamimi said. Participants in the GUTS cohort were aged 5 to 10 at enrollment and are now between 26 to 32. “They are not at an age where they have breast cancer diagnoses, but many have had benign biopsies. So, we can look prospectively at lifestyle factors and how those factors affect benign breast lesion development,” Tamimi added. While data from GUTS is not currently mature enough for analysis of cancer risk-associated factors, Tamimi’s work with the NHS cohorts have provided important insights into breast cancer.  

In ongoing work, Tamimi is building on previous findings from her group and others on mammographic density, which is highly inheritable and one of the strongest breast cancer risk factors. However, the mechanisms are as-yet unknown. Seminal studies led by Tamimi were the first to demonstrate that circulating hormones and mammographic density were independent predictors of breast cancer risk. The Addition of breast density to current breast cancer risk prediction models improves their performance, and the risk is strongly associated with both estrogen receptor (ER)-positive and ER−negative breast cancers.Tamimi and collaborators are carrying these observations forward, analyzing mammographic images with machine learning techniques to identify and extract unique features that may help identify women at heightened risk of developing breast cancer. They are also actively assessing the link between breast density and gene expression to shed light on how even subtle differences in breast density (as low as 5%) can have a significant effect on breast cancer risk.

Other Parts of the “Main”: Local and Global Epidemiological Perspectives 

Epidemiologists have a panoramic view of disease, and work from CE program members in Boston and Botswana exemplifies this socio-ethnically-mindful perspective. CE program members are involved in assessing risk of cancer and associated deaths in patients with HIV infection. “These studies have been an interesting international project that works across disciplines of infectious diseases, medicine, and CE,” Mucci remarked. 

Locally, Erica Warner, ScD, MPH (HMS) whose work focuses on breast cancer and health disparities, is spearheading a GWAS study of mammographic density in African-American and Hispanic women enrolled in the Partners BioBank, aimed at analyzing racial/ethnic differences in breast density as well as the genetic, behavioral, and lifestyle factors underlying these differences. In earlier work, Warner and colleagues found that African-American women are more likely to die from breast cancer than white women and that, surprisingly, the survival disparity was noted in ER-positive breast cancers,8for which more targeted therapeutic agents exist.  

Tamimi, Warner, and others have also shown that early life adiposity is inversely associated with breast cancer risk.9“In postmenopausal women, obesity is associated with increased risk of breast cancer, but before that, and especially in childhood and between first menarche and first pregnancy, being heavier is protective and is associated with decreased breast cancer risk later in life, regardless of your weight later in life,” Warner stated.

According to Warner, one of the challenges in conducting large-scale studies on cancer risk and survival disparities is the limited access to data from minority populations. “Locally, there is the Black Women’s health study with Boston University, but many other cohorts are not particularly diverse.” She has tried to address this deficit by creating partnerships with Boston Medical Center, which treats a disproportionately large fraction of African-American Bostonians, and through the Boston Breast Cancer Equity Coalition.

The Next Generation of Cancer Epidemiologists

One of the core missions of the CE program is to prepare the next generation of cancer epidemiologists. “The CE program along with the Cancer Data Sciences program launched the DF/HCC Celebration of Junior Investigators in Cancer Research symposium in 2013 and the programs continue to co-organize and be a driving force behind its yearly recurrence.” According to Mucci, the Junior Investigators program is meant to showcase the young talents in all areas of population science. In addition, the “Hot Topics”/”Brief Updates” seminar series, moderated by junior investigators, provides another opportunity whereby faculty, fellows, and students can discuss newly-published CE-relevant research.

No cancer patient is an island; neither is a cancer researcher

The meaningful and close interactions among the 78 CE program members and those of DF/HCC’s other clinical and basic science programs pave the way for more exciting new discoveries. “We are open to collaborating on any question that would benefit from population-based research,” Tamimi said.

Mucci concurs. “It is always incredible how successful the collaborations are across these 7 institutions.” 

Written by Krithika Subramanian


  1. Mayers JR, Wu C, Clish CB, et al. Elevation of circulating branched-chain amino acids is an early event in human pancreatic adenocarcinoma development. Nat Med. 2014;20(10):1193-1198.
  2. Dickerman BA, Torfadottir JE, Valdimarsdottir UA, et al. Midlife metabolic factors and prostate cancer risk in later life. Int J Cancer. 2018;142(6):1166-1173.
  3. Yuan C, Morales-Oyarvide V, Babic A, et al. Cigarette smoking and pancreatic cancer survival. J Clin Oncol.2017;35(16):1822-1828.
  4. Michailidou K, Lindström S, Dennis J, et al. Association analysis identifies 65 new breast cancer risk loci. Nature. 2017 Nov 2;551(7678):92-94.
  5. Milne RL, Kuchenbaecker KB, Michailidou K, et al. Identification of ten variants associated with risk of estrogen-receptor-negative breast cancer. Nat Genet. 2017;49(12):1767-1778.
  6. Klein AP, Wolpin BM, Risch HA, et al. Genome-wide meta-analysis identifies five new susceptibility loci for pancreatic cancer. Nature Communications. 2018;9:556. 
  7. Rice MS, Tamimi RM, Bertrand KA, et al. Does mammographic density mediate risk factor associations with breast cancer? An analysis by tumor characteristics. Breast Cancer Res Treat. 2018;170(1):129-141.
  8. Warner ET, Tamimi RM, Hughes ME, et al. Racial and ethnic differences in breast cancer survival: Mediating effect of tumor characteristics and sociodemographic and treatment factors. J Clin Oncol. 2015;33(20):2254-2261.
  9. Warner ET, Hu R, Collins LC, et al. Height and body size in childhood, adolescence, and young adulthood and breast cancer risk according to molecular subtype in the Nurses' Health Studies. Cancer Prev Res (Phila).2016;9(9):732-738.