A person’s overall physical health and well being is often dictated more so by the experiences of their past, than by their present. The health behavior choices we make and exposures we suffer can have a significant impact on our health later in life.
[Photo: Dr. Carmen D. Tekwe]
Dr. Carmen D. Tekwe, assistant professor with the Texas A&M Health Science Center School of Public Health, recently published a study detailing the development of a new statistical model in Statistics in Medicine. Using this model, she examined the effects of radiation exposure on cardio-metabolic risk factors associated with dyslipidemia, an abnormal amount of lipids (e.g. cholesterol and/or fat) in the blood. The study was based on data from the Adult Health Study cohort of survivors of the atomic bombings of Hiroshima and Nagasaki, Japan.
With this study, entitled “Multiple indicators, multiple causes measurement error (MIMIC ME) models”, Dr. Tekwe and her co-authors participated in an international collaboration of Japanese and American scientists and statisticians to extend a standard latent variable model known as the MIMIC model to study the effects of true radiation dose at exposure on triglyceride, HDL and LDL cholesterol levels in the blood. Their MIMIC ME model allowed adjustment for measurement error associated with the manner in which the amount of radiation exposure was determined.
“The calculations used to estimate the exact amount of radiation doses received by the survivors was based primarily on self-reported measures of distance and shielding at the time of exposure,” said Dr. Tekwe. “Therefore, they are prone to classical measurement error.”
In addition to typical measurement errors, this study adjusted for a second source of error known as Berkson error. This error occurred primarily due to all individuals sharing a common characteristic (i.e., close proximity) being assigned the radiation dose calculated for a hypothetical representative of their group, rather than individually estimated doses.
As a result, this study was the first to assess the impact of radiation dose on dyslipidemia among the survivors adjusting for both types of errors. Results indicated that radiation exposure increases cardio-metabolic risk.
According to Dr. Tekwe, failure to properly account for measurement error(s) in assessing the impact of an exposure agent on various health outcomes such as dyslipidemia or cancer can lead to either under- or over-estimation of true effects, depending on the presence or absence of measurement error. MIMIC ME model analysis provides corrections for both types of errors and thus, an unbiased estimate of the effect of exposure.
“In order to make recommendations to appropriate regulatory or public health agencies on the effects of exposure to a particular environmental agent on the public, it is essential that the exposure to the environmental agent be approximated as reliably as possible and that analysis methods such as the MIMIC ME model be employed when errors remain. ”
This method has the potential to be applied to a wide array of fields such as cancer prevention, tobacco exposure, and behavioral health patterns research. With the help of this model researchers will not only improve study protocol and reliability, but also enhance the ability to accurately target preventative patient care.
Additional authors are Dr. Randy Carter of the University of Buffalo in New York, Dr. Harry Cullings of the Radiation Effects Research Foundation in Hiroshima, Japan, and Dr. Raymond Carroll of Texas A&M University.