ASA Leaders Reminisce: Mitchell H. Gail
Jim Cochran
Mitch Gail
Gail is a Fellow of the American Statistical Association and member of the Institute of Medicine of the National Academy of Sciences. He has received several prestigious awards, including the Spiegelman Award from the American Public Health Association, the Snedecor Award from the American Statistical Association, the Marvin Zelen Leadership Award in Statistical Science from the Harvard T.H. Chan School of Public Health, the Nathan Mantel Lifetime Achievement Award from the American Statistical Association, and the Award for Excellence in Cancer Epidemiology and Prevention from the American Association for Cancer Research and American Cancer Society.
Q: Mitch, thank you for taking time to talk with me. You have what I suspect is a unique distinction among ASA presidents. Several years before you earned your PhD in statistics from The George Washington University, you earned an MD from Harvard Medical School. What motivated you to return to school to work on a PhD in statistics several years after you earned your MD?
A: After working as a medical intern in at the Peter Bent Brigham Hospital in Boston for one year, I joined the U.S. Public Health Service (USPHS) as a researcher at the National Cancer Institute (NCI). My first project was studying the motility of cells in tissue culture. They moved in a two-dimensional random walk, and I needed statistical methods to characterize their rates of diffusion. John J. Gart at NIH showed me how to perform certain regressions, plotting mean squared cell displacement against time. Three years later, David P. Byar invited me to join his clinical and diagnostic trials section at NCI to work on lung cancer clinical trials and problems in the immune-diagnosis of cancer. These projects required statistical knowledge, and the USPHS paid for some night courses at The George Washington University in DC. At first, I enrolled as a master’s student, but before I knew it, I was working with Joseph L. Gastwirth on my PhD thesis. I was lucky to be working in an environment that supported continuing education and where my increasing statistical skills were useful and appreciated.
Q: This spring, you received the 24th annual AACR-American Cancer Society Award for Excellence in Cancer Epidemiology and Prevention from the American Association for Cancer Research (AACR) and American Cancer Society. The AACR press release cited a statistical model you developed in the late 1980s for estimating the absolute risk for a white woman of a specific age with specific risk factors to develop breast cancer. What led you to develop this model, which is commonly known as the “Gail model,” and what has been the impact of this model?
A: One day, I went to lunch with an NCI collaborator, John J. Mulvihill, who was advising women in a clinic for women at high risk of breast cancer. Many such women thought they had unrealistically high risks of cancer because of strong family histories, and John wanted to be able to give them realistic estimates. I was fortunate to be working with other collaborators (Louise A. Brinton and Catherine Schairer) who were familiar with data on breast cancer risk factors from the Breast Cancer Detection and Demonstration Project and with statisticians (Dave P. Byar, Donald K. Corle, and Sylvan B. Green) who knew a lot about relative risk modeling and attributable risk estimation. My contribution was to bring all these tools together to compute absolute, not relative, risk. Absolute risk is the chance where to buy ativan online that a woman with specific risk factors will develop breast cancer over a defined time interval. Absolute risk has many applications in individual counseling and public health planning.
One of the first uses of the model was to help design a large randomized intervention trial to determine whether tamoxifen could prevent breast cancer. Statisticians at the National Surgical Breast and Bowel Project (NSABP) combined national breast cancer incidence data with the relative and attributable risks from the original model to produce what is now known as the “Gail model,” or the National Cancer Institute’s Breast Cancer Risk Assessment Tool. This model correctly predicted the number of breast cancers that would develop over five years during the Breast Cancer Prevention Trial, which showed that tamoxifen cut breast cancer incidence by half. Unfortunately, tamoxifen also caused some serious side effects, such as strokes and endometrial cancer.
The absolute risks associated with each of the risks and benefits of tamoxifen are needed to help a woman decide whether she stands to benefit from such an intervention. Typically, young women with high breast cancer risks benefit most. Today, there are several good risk models for breast cancer with various strengths and weaknesses, and they are widely used in counseling and for making public health recommendations. NCI’s website for breast cancer risk assessment is visited more than 3 million times per year.
Q:You have developed many breast cancer risk assessment tools, including several SAS macros for breast cancer risk assessment that have been designed for different demographic/ethnic groups. What led you and your colleagues to develop different breast cancer risk assessment tools for different demographic groups?
A: The National Cancer Institute’s Surveillance, Epidemiology, and End Results Program (SEER) gathers breast cancer incidence data for racial and ethnic subgroups, and there is considerable variation in the age-specific incidence rates. Therefore, specialized models are needed for various racial and ethnic groups to provide realistic estimates of absolute risk. Our initial attempt was to combine race/ethnicity-specific SEER rates with the relative and attributable risks from the original Gail model, which was developed from data on white women, to compute race/ethnicity-specific absolute risks. More recently, we have been able to obtain data on relative and attributable risks that are also specific to various racial/ethnicity groups. Fully specific models are available for African-American and Asian-American women, and we are obtaining such data for Hispanic women.
Q: In 2012, you served as chair of the Section on Statistics of the American Association for the Advancement of Science (AAAS). What issues did this AAAS section face during your term as its chair?
A: The Statistics Section (Section U) of AAAS was formed in 1962. Jerzy Neyman described the role of the section and stressed the opportunity for interdisciplinary exchange (Science, 1962; 138:1080-1083). Several issues related to this tradition come to mind. The first was developing an exciting set of symposia from Section U with broad appeal to the scientific community on topics ranging from “Understanding and Communicating Uncertainty in Climate Change Science” to “Benefits of Randomized Experiments for Science and Society.” A second was nominating and recognizing AAAS fellows for their contributions in statistics and its applications. A third was increasing visibility of Section U and the role of statistics. In coordination with the ASA (and with a big boost from Marie Davidian, former ASA president, and Ron Wasserstein, ASA executive director), we geared up to celebrate the 50th year of Section U and the International Year of Statistics (2013). This cooperation with the ASA has continued, and Section U has grown during a period when overall AAAS membership has not.
Q: What challenge that you faced during your ASA presidency stands out?
A:That would be the immense diversity of interests of ASA members. It seemed to me that one of the challenges of the ASA was to serve such a professionally diverse and talented membership. In my presidential address, I reviewed the contributions of Bradford Hill and Jerome Cornfield to the development of the case-control method and clinical trials. These great statisticians combined the development of innovative methods with deep involvement in collaborations and substantive scientific research and set an example that many of us could try to emulate, regardless of our particular interests or specializations.
(No Ratings Yet)
Loading...
