Thomas R. Fears, Ph.D.
|Organization:||National Cancer InstituteDivision of Cancer Epidemiology & Genetics, Genetic Epidemiology Branch|
Dr. Fears received a Ph.D. in statistics from Iowa State University in 1972. He joined the NCI in 1973 as a Senior Staff Fellow, and has been a mathematical statistician in the Epidemiology and Biostatistics Program since 1979. He is a recipient of the NIH Merit Award, an Equal Employment Opportunity Special Achievement Award, and the Snedecor Award from the American Statistical Association.
As part of an ongoing collaboration with epidemiologists in DCEG and elsewhere, quantitative methods are introduced into study designs and analyses to help investigators achieve more efficiently their research objectives. Issues in design and analysis frequently arise which motivate new methodologic research.
DCEG participated in two studies of the long-term impact of treatment for childhood cancer. In the Five Center Study, cancer registries identified children who were diagnosed with cancer between 1945 and 1974, and interviewed all adult survivors and their siblings between 1980 and 1983. This open historical cohort provided opportunities to study outcomes such as early menopause, reproductive problems and birth defects, marriage and divorce, and death from non-cancer causes. Since the Five Center Study, survival has dramatically improved for children diagnosed with acute lymphoblastic leukemia. To study long-term effects of cancer treatment among this group, we identified children participating in clinical trials sponsored by the NCI-funded Children's Cancer Group, and interviewed long-term survivors and their siblings. The unusual study design allowed entry into the study at different times and required development of special computer programs for data analysis. In this large cohort, younger female patients who received 1800 cGy cranial radiation and those who received 2400 cGy below the diaphragm experienced early or delayed menarche. Patients treated with 2400 cGy cranial radiation experienced educational difficulties. Other long-term effects studied in this cohort include loss of fertility, birth defects, psychosocial outcomes, and general health problems.
Nonmelanoma skin cancer is the most common malignancy in Caucasians. Although it is generally accepted that ultraviolet radiation from the sun is the dominant risk factor, epidemiologic study of these cancers has been limited because most patients are not hospitalized. In 1977-1980, we conducted nonmelanoma skin cancer incidence surveys and interview studies at ten U.S. locations where it was feasible to estimate the dose-response relation between ultraviolet radiation exposure and skin cancer incidence. Interview studies conducted at each location, employing complex samples of cases and controls, served as the basis for methodologic research. We have prepared review articles, one on the epidemiology of nonmelanoma skin cancer and the second on solar radiation as a carcinogen.
Some reports have suggested that non-Hodgkin lymphoma (NHL) may be related to sunlight exposures. In a study of geographic patterns of mortality rates, however, we found that ultraviolet radiation does not have the same clear relation to NHL that is evident for melanoma and nonmelanoma skin cancer. The etiology of melanoma remains unclear, although an association with sunlight exposure has been found, particularly when it occurs intermittently at young ages and in relation to sunburns. We also studied the role of melanocytic and dysplastic nevi in the development of familial and sporadic melanoma. We are now investigating a measure of exposure to solar radiation, based on residential history, in an effort to better understand the relation between solar ultraviolet radiation and risk of melanoma.
Plasma and urine samples were collected as part of a case-control study of breast cancer among Asian-American women. We plan to evaluate the association between the risk of breast cancer and hormone levels in these biologic samples. To assure the reliability and validity of hormone assays, we studied their reproducibility in several laboratories. Our findings suggest that estrogen assays performed over several months are sufficiently reliable for epidemiologic studies. An evaluation is underway of the results for androgen assays. A careful examination of laboratory procedures on biologic specimens is underway, including the effects of repeated freezing and thawing, use of borate as a preservative, and length of time in storage. An example of the fallibility of the Wald statistic was observed in these data.
At each location in the nonmelanoma skin cancer studies, interviews with control subjects were obtained using a population-based, two-stage cluster sampling procedure known as random digit dialing. A pseudo-likelihood approach for analyzing these data was developed to allow the effects of ultraviolet radiation exposure to be evaluated while adjusting for other risk factors. The methods yield estimates of absolute risk, as well as relative risk with standard errors.