Pooling Project Harnesses Data from Studies Around the World to Investigate Links Between Diet and Cancer, September 14, 2007, Harvard Public Health NOW

The link between decreased colon cancer risk and fiber found in foods such as cereal and fruit was examined by the Pooling Project.

The story is fairly well known among nutritionists. In the 1950s, two Western doctors traveled to Uganda to visit health clinics and observed that the patients did not have many of the diseases so common among their patients at home: heart disease, diabetes, obesity, colon cancer, and others. Casting for an explanation, the doctors suggested that high levels of roughage, specifically fiber, in the diets of Ugandans may have helped protect them from these diseases. Western researchers began searching for links between eating fiber and cancer risk, particularly colorectal cancer.

For decades, the possible association was examined in different study populations, but time and again, study results were inconsistent. Many studies of different types found that higher levels of dietary fiber consumption were associated with lower risk of colorectal cancer. Yet, conversely, several nutrition prospective studies - which assess diet and other lifestyle factors before the disease emerges and then follows people over time - found no association.

This scientific conundrum is just the type of challenge undertaken by researchers at the Pooling Project of Prospective Studies of Diet and Cancer, an international consortium based in the Department of Nutrition at HSPH. The project, started in 1991, involves a team of investigators who analyze primary data from nearly 30 cohort studies around the world.

The Pooling Project offers benefits that other study approaches may not. For one thing, the project harnesses the statistical power of many cohorts to investigate a scientific question. For another, the project uses primary data from these studies and then reanalyzes the collective data using standardized criteria across all of the studies - sometimes providing whole new insights into links between cancer and diet. This approach is different - and arguably more powerful - than the common method of reviewing scientific literature and pulling out summarized statistics.

In the case of dietary fiber and colorectal cancer, Pooling Project researchers published a paper in the December 14, 2005, issue of the Journal of the American Medical Association. They used data gathered in 13 prospective cohort studies of more than 725,000 men and women who were followed anywhere from six to 20 years, depending on the individual study. Among the study subjects, more than 8,000 people developed colorectal cancer.

By reanalyzing the data, Pooling Project researchers identified two key findings: people with very low amounts of dietary fiber (less than 10 grams a day) had a higher risk of colorectal cancer than people who ate more fiber each day. But risk did not change for people who ate lots of fiber (10 to 30+ grams a day). In other words - after accounting for other dietary risk factors such as alcohol consumption, exercise, and folate intake - the researchers did not find an association between high dietary fiber consumption and reduced risk of colorectal cancer.

"This was the largest study to date of the association between dietary fiber and colorectal cancer risk, but the debate does continue," said Stephanie Smith-Warner, who heads the Pooling Project. "Before our study was published, a large European-based study called EPIC had found a linear decrease in risk of colorectal cancer associated with high levels of dietary fiber consumption. We haven't closed the book on the subject, but we have helped to investigate the question of whether the benefits of eating dietary fiber level off when it comes to colorectal cancer." Smith-Warner is assistant professor of nutritional epidemiology in the HSPH Departments of Nutrition and Epidemiology.

The Pooling Project was initiated 16 years ago by David Hunter, Vincent L. Gregory Professor of Cancer Prevention in the Departments of Epidemiology and Nutrition at HSPH. The effort initially had six participating studies. Said Hunter, "We realized that the literature on diet and cancer relations was riddled with results that were not replicated or for which no replication seemed to have been attempted, or at least published."

He recalled, "Many told us we would not be able to get leaders in the field to contribute data, let alone agree to a consensus on the results." Today there are 28 participating cohorts based in places as diverse as Iowa, Japan, Sweden, China, and Boston.

Smith-Warner took over as project leader in 2002 and manages the study day to day.

The Pooling Project is unusual in that it receives primary data from each participating study, she said. From there, the data can be analyzed in many different ways, factoring in various influences. This flexibility allows the project's researchers to look at subgroups in populations - such as pre- or postmenopausal women, former or current smokers, people who take a daily multivitamin or those who do not - to examine disease associations, while retaining enough numbers of people in the subgroups to make the data meaningful.

Investigators involved in individual cohorts retain control of their data and are included in the publication of Pooling Project papers that involve their cohorts. The Pooling Project doesn't collect initial data but instead reanalyzes existing data made available by research teams to calculate summary estimates of specific diet and disease associations.

So far, Pooling Project teams have published papers on breast, lung, colorectal, ovarian, and renal cell cancer. At the end of this month, the Journal of the National Cancer Institute will publish another Pooling Project paper on fruits, vegetables, and colon cancer risk. Smith-Warner is senior author.

The project has been continuously funded by the National Cancer Institute, which recently provided a grant to look at less common, but still deadly, cancers including advanced prostate cancer, brain tumors called gliomas, and estrogen-receptor negative breast cancer. Indeed, research into these rarer kinds of cancer may benefit from the Pooling Project even more than do common cancers. Why? If a cancer is rare, it may be difficult to find enough subjects to study the illness with any kind of statistical assurance. The Pooling Project overcomes that challenge by reanalyzing data from numerous studies.

"The American Cancer Society estimates that women have a one in three chance of developing cancer at some point in their lifetimes and that men have a one in two chance," noted Smith-Warner. "Hopefully, studies like the Pooling Project can improve our understanding of how cancer is associated with diet to gain an edge on this pervasive disease."

To learn more, visit www.hsph.harvard.edu/poolingproject.

—MSS and CR