Epidemiology pinpoints cancer risk factors and survival strategies
Some cancers are on the rise, but the overall U.S. rates of newly diagnosed cancers and of cancer deaths have been falling for 10 years. This year, just over one million people will learn about their cancers for the first time. Within these numbers and trends are large groups of people whose diets, activities, health histories, genes, tumors, blood, and tissue samples hold powerful lessons for preventing and treating cancer.
“Our overriding goal is to find ways to reduce cancer risk and to improve survival of patients with cancer,” says cancer epidemiologist Susan Hankinson, MPH, ScD (BWH), co-leader of the DF/HCC Cancer Epidemiology Program with Walter Willett, MD, MPH (HSPH). The vigorous program boasts nearly 60 faculty members at Harvard and affiliated hospitals.
To determine risk factors for developing—and increasingly, for surviving—cancer, many researchers in the program work with some of public health’s most famous epidemiological datasets: The Nurses’ Health Study and Nurses’ Health Study II, the Women’s Health Study, the Physicians Health Study, and the Health Professionals Follow-up Study. Further research uses several large hospital and population-based case-control studies.
Thanks to new tools and collaborations with DF/HCC pathologists, cancer biologists, and clinicians, such studies are expanding beyond the lifestyle factors associated with healthy living to probe the underlying biology. “We used to use questionnaires and interviews only,” Hankinson said. “Now we’re getting to the tissue level. We used to ask about family history. Now we do whole genome scans. We’re looking at population-based questions with a tremendous depth of information.”
The three researchers featured in this story reflect the new directions, the diversity of approaches, and the range of technologies being deployed at the population level against many cancers to understand their causes and to devise new strategies for prevention and treatment.
Surviving lung cancer
In lung cancer, the biggest non-occupational risk factor in this country is no mystery. Half of the people who walk into the clinic of pulmonary physician and molecular epidemiologist David Christiani, MD, MPH (HSPH) are former smokers.
Nationwide, “30 million people have taken our advice and have stopped smoking for more than one year,” Christiani says. “Now what can we do for them?” Every year after quitting, the lung-cancer risk goes down, bottoming out after 10 to 13 years to a non-smoker’s risk for light smokers and two to four times that for heavy smokers.
Christiani has led a case-control study for 20 years looking at the genetic factors that make people susceptible to smoking-related lung cancer and help predict the outcomes of their treatments. After a decade of studying the risks of disease, Christiani became more interested in the molecular and genetic determinants of survival, an uncommon epidemiology perspective at the time.
“From a public health point of view, it’s interesting to see how genes modify risk,” he says. “But clinicians don’t see controls, they see cases.” All new cases of lung cancer that present at the surgery or medicine departments at MGH are eligible to participate in the study, which pairs the clinical information with blood samples and with frozen tissue from surgery for what Christiani calls “deep biological epidemiology.”
Tumor staging has been the best, if imprecise, predictor of patient outcomes, but Christiani and his colleagues have further parsed those estimates within each stage with additional influential genetic and non-genetic factors.
A recent analysis uncovered five genetic markers in tumor tissue that seem to be prognostic of relapse-free and overall survival in early stage non-small-cell lung cancer (NSCLC). Collectively, the markers add up to a greater risk, making the cumulative dosage a promising prognostic marker. The team also found amplifications known as copy number gains on three chromosomes, whose meaning they are still analyzing. Christiani and his colleagues validated the findings in a similar set of Norwegian patients.
The latest studies build on earlier findings that high levels of vitamin D and its genetic variants were associated with greater survival in early stages of NSCLC, but not in advanced stages.
While many of these results need to be assessed in other populations and further validated before widespread clinical use, another finding has had immediate clinical application: second-hand smoke bodes poorly for the lung cancer patient. “It’s as simple as telling the family members, ‘Here’s what you can do: All of you can stop smoking,’” Christiani said.
In the works are plans for prospective pharmacogenomics trials to identify specific drug pathways and toxicities. For example, persons with genetic variants in DNA repair who have late-stage disease have worse survival and more side effects, whereas the same variants in early stage disease are associated with better survival. “This kind of information on common variants, combined with tumor mutation status (like EGFR, K-ras, etc.), will lead to better, more precise individualized treatment regimens,” he says.
Christiani also studies gene-environment interactions for esophageal cancer and has a wide network of collaborators in Asia, Africa, and North American to investigate the effects of exposure to environmental pollutants and occupational toxins on people’s health, including arsenic exposure and bladder and skin cancer in Taiwan and Bangladesh, and petrochemical exposures, brain tumors, and leukemia in Taiwan.
Chemoprevention for colon cancer
The unfolding saga of aspirin and other anti-inflammatory agents in preventing colon cancer is “a real epidemiological success story,” says gastroenterologist Andrew Chan, MD, MPH (MGH). Chan spends most of his week in the clinic and wants to translate the epidemiology of risk factors into better preventive tools to help patients.
Chan and his colleagues have repeatedly leveraged the extensive information in the Nurses’ Health Study and Health Professionals Follow-up Study to try to define the circumstances in which the compelling benefits of aspirin outweigh the toxicities. In those cohorts, men and women who regularly took aspirin (two standard tablets a week or aspirin use twice a week) had a lower risk of death from colon cancer. Compelling evidence from other human and molecular studies had implicated the molecular target, COX2, which aspirin blocks to reduce pain, fever, and inflammation.
With the help of collaborator and molecular pathologist Shuji Ogino, MD, PhD (DFCI, BWH), Chan pursued an idea about how aspirin protects against colon cancer. They assessed COX2 activity in tiny samples of preserved colorectal-cancer specimens and compared them to aspirin use data gleaned from questionnaires submitted by participants in the two large cohorts every two years. As Chan had hypothesized, aspirin’s protective benefits were almost entirely confined to a reduction in the number of COX2-dependent tumors among with no apparent affect on tumors without the COX2 signature.
“There are further things we can do to stratify risk and to apply this information in a more wide-scale way to optimize treatment,” said Chan. In one follow-up avenue of study, for example, Chan and his colleagues are seeking patterns in genetic markers or inflammatory factors in the blood, especially those variations in aspirin metabolism that influence the subsequent colon cancer risk. He would also like to understand how the risk factors interact—is the extra risk from obesity and red meat modifiable by aspirin? (Chan notes that routine use of aspirin and other non-selective or selective COX2 inhibitors is not recommended for colon-cancer prevention in the general population because of associated risks, such as gastrointestinal bleeding.)
Chan’s work is supplementing the landmark datasets with new analyses on the existing collection of tumor, blood, and DNA samples. For a new project on gut microbes and their relationship over time to health, he is exploring the possibility of augmenting one cohort’s biospecimen archive by collecting stool samples to establish what may be the first large-scale population-based study of its kind.
His immediate challenges: How feasible is it to collect, ship, process, and store stool samples? “There’s a level of yuck factor,” he admits, “but health professionals understand the value of the study and we have a good response rate of people willing to submit samples.”
There is a long list of scientific questions that can best be answered with a real-world, population-based cohort, he says. How consistent is microflora over time and how do factors such as diet and aging influence any changes? What aspects of the microflora are important? Are there patterns in the species or their interactions that predispose people to certain diseases? What is the interaction between microflora and the diet? The list goes on and new questions will come from researchers and technologies in the future.
Earlier biomarkers of breast cancer risk
Mammography has been in the news lately for the debate about its relative benefits and timing as a screening tool to lower a woman’s risk of death from breast cancer. Experts continue to disagree about the use of mammograms in preventive health care, but there is no quarrel about the value of aggregate mammograms in rooting out the biological correlates of breast cancer risk.
Cancer epidemiologist Rulla Tamimi, ScD (BWH), works with mammograms, tissue and blood samples, and clinical and lifestyle data from the Nurses’ Health Study and Nurses’ Health Study II to reexamine risk factors for breast cancer, especially seeking intermediate markers and ways to modify the risk.
Women with denser breasts are at four to six times greater risk of breast cancer than women with fatty breasts. “It’s one of the strongest risk factors we have,” she says. The highest risk comes from breasts composed of more than 75 percent of dense tissue containing epithelial cells and stroma (which appears light on screening mammograms, while fat appears dark). Density is highly heritable, says Tamimi, who is working on determining the genes that influence density.
In a recent attempt to test the connection between endogenous sex hormones and breast density, Tamimi and her colleagues were surprised by the findings. Among postmenopausal women, circulating estradiol had no correlation with density, but both features were strong and independent risk factors for breast cancer. Tamimi is repeating the study in premenopausal women in collaboration with Hankinson and Willett. To help elucidate the molecular and cellular basis for differences in breast density, Tamimi has teamed up with Kornelia Polyak, MD, PhD (DFCI).
In other work, Tamimi and pathology colleagues from BIDMC have begun parsing the multiple breast changes grouped together under the term "benign breast disease." The changes are regarded as generalized markers (not precursor lesions) of a variably increased cancer risk ranging from 50 percent to four-fold. A pending publication from Tamimi and collaborators in the lab of Joan Brugge, PhD (HMS), will report a high-risk molecular pattern in breast biopsies that predicts later risk of breast cancer. Moving forward, Tamimi will fall back on more traditional epidemiological studies to uncover lifestyle influences on the risky biomarker.
“There are so many new molecular tools out there that we can apply to biologic specimens and to mammograms to give us a lot more information,” Tamimi says. “The biomarkers and other measurements help us make better causal inferences.”
With a new assay optimized for paraffin-embedded tissue samples, Tamimi and her colleagues are planning a study to correlate genome scans with the tumor gene expression patterns to link hits from genome-wide association studies to gene expression. She will also be examining 5,000 tumor samples from the Nurses Health Study to identify new biological markers that may predict differences in survival of women with breast cancer and then look for influential lifestyle patterns.
Collaborations key to future epidemiology
“The reason this all works is because of the interconnections of our program members with clinicians and cancer biologists,” Hankinson says. “The tremendous collegiality among people with different expertise at the DF/HCC helps get people talking to each other to form new scientific marriages that result in hybrid projects. These projects provide much greater insight into disease etiology and prevention.”
The new projects, in turn, generate fresh challenges requiring further expert teamwork, such as how to analyze and interpret the extensive data. In the genome correlation project, for example, Tamimi is partnering with John Quackenbush, PhD (DFCI) for bioinformatics; Stuart Schnitt, MD (BIDMC) for breast pathology; David Hunter, ScD, MPH (HSPH) for genetic epidemiology; and more.
Seminars, retreats, and workshops further serve to share findings from one cancer that might be relevant to another. For example, Hankinson cites a puzzling childhood exception she and her colleagues found to the strong link between obesity and breast cancer. “Children overweight at ages 5 and 10 have a lower risk of breast cancer,” she says. “The association is strong, and we don’t know what it means, but it leads us to think in more depth about the risk factors over the life course not only with breast but also with other cancers.”
— Carol Cruzan Morton