Page 390 - 2014 Printable Abstract Book
P. 390
dose response. The ERR significantly increased with decreasing age at exposure; age at exposure effect
was not significant after adjusting for attained age. The EAR increased significantly with decreasing age at
exposure and increasing attained age. To improve the understanding of breast cancer etiology and to
evaluate interaction with radiation, we are currently conducting analyses of the joint effects between
radiation and reproductive factors, which are available for about 50% of women. This analysis represents
the first time that comprehensive information on reproductive factors has been included in the
characterization of breast cancer risk in the full LSS cohort. The extended follow up of the LSS cohort
together with improved dose estimates and data on reproductive factors allows for more detailed dose-
response analyses, particularly for those who were younger than 20 years at the time of the bombings.
(S503) Stomach cancer radiation risks in the LSS. Ritsu Sakata, Radiation Effects Research Foundation,
Hiroshima, Japan
Stomach cancer has the highest incidence rate among solid cancers in Japan and accounts for 25%
of all solid cancers observed in the Life Span Study (LSS). During 1958-2009, 6,026 first primary stomach
cancers were diagnosed among the 111,917 LSS members eligible for this study. The excess relative risk
(ERR) of developing stomach cancer following radiation exposure was significant with the sex-dependent
ERR decreasing with attained age and no significant effect of age at exposure, as reported previously.
Smoking, alcohol use and socio-economic status have been reported as possible risk factors for stomach
cancer. We investigated the radiation effects on stomach cancer incidence adjusted for these factors.
Smoking was significantly associated with stomach cancer incidence in the LSS. The interaction between
radiation and smoking was better described as additive rather than multiplicative. The ERR/Gy based on
the smoking-adjusted additive model was slightly larger than the ERR/Gy based on the model that did not
take smoking into account.
(S504) Non-parametric smoothing for radiation dose-response estimation. Akira Furukawa,
Nat'l Inst. of Radiological Sciences, Chiba, Japan
Department of Statistics, Radiation Effects Research Foundation, Japan Characterizing the dose-
response relationship and estimating acceptable exposure levels are the primary goal of risk assessments.
In analysis of health risks associated with exposure to ionizing radiation, while there is a clear agreement
that moderate to high radiation doses cause harmful effects in humans, information is often limited to
understand the possible biological effects at low doses, e.g., below 100 mGy. A conventional approach to
the dose-response analysis considers models of relatively simple parametric forms and chooses one based
on the goodness of fit to the observed data for use in subsequent risk characterizations. While a simple
parametric model, such as the linear non-threshold model, is easily interpretable and convenient in risk
communications, it is often overlooked that risk analysis based on such a simple model can be misleading
in evaluating the risk and its uncertainty associated with a low dose exposure. The risk at low doses may
be small but the implication of inappropriately evaluated uncertainty of the risk may not be small in risk
projections since the number of exposed subjects at this dose range is large in many exposed populations.
As an alternative approach, this study considers a Bayesian semi-parametric model that has a piecewise-
linear dose-response function to be smoothed through an auto-regressive structure among the random
slope coefficients defined over closely spaced dose categories. Simulation studies and application to Life
was not significant after adjusting for attained age. The EAR increased significantly with decreasing age at
exposure and increasing attained age. To improve the understanding of breast cancer etiology and to
evaluate interaction with radiation, we are currently conducting analyses of the joint effects between
radiation and reproductive factors, which are available for about 50% of women. This analysis represents
the first time that comprehensive information on reproductive factors has been included in the
characterization of breast cancer risk in the full LSS cohort. The extended follow up of the LSS cohort
together with improved dose estimates and data on reproductive factors allows for more detailed dose-
response analyses, particularly for those who were younger than 20 years at the time of the bombings.
(S503) Stomach cancer radiation risks in the LSS. Ritsu Sakata, Radiation Effects Research Foundation,
Hiroshima, Japan
Stomach cancer has the highest incidence rate among solid cancers in Japan and accounts for 25%
of all solid cancers observed in the Life Span Study (LSS). During 1958-2009, 6,026 first primary stomach
cancers were diagnosed among the 111,917 LSS members eligible for this study. The excess relative risk
(ERR) of developing stomach cancer following radiation exposure was significant with the sex-dependent
ERR decreasing with attained age and no significant effect of age at exposure, as reported previously.
Smoking, alcohol use and socio-economic status have been reported as possible risk factors for stomach
cancer. We investigated the radiation effects on stomach cancer incidence adjusted for these factors.
Smoking was significantly associated with stomach cancer incidence in the LSS. The interaction between
radiation and smoking was better described as additive rather than multiplicative. The ERR/Gy based on
the smoking-adjusted additive model was slightly larger than the ERR/Gy based on the model that did not
take smoking into account.
(S504) Non-parametric smoothing for radiation dose-response estimation. Akira Furukawa,
Nat'l Inst. of Radiological Sciences, Chiba, Japan
Department of Statistics, Radiation Effects Research Foundation, Japan Characterizing the dose-
response relationship and estimating acceptable exposure levels are the primary goal of risk assessments.
In analysis of health risks associated with exposure to ionizing radiation, while there is a clear agreement
that moderate to high radiation doses cause harmful effects in humans, information is often limited to
understand the possible biological effects at low doses, e.g., below 100 mGy. A conventional approach to
the dose-response analysis considers models of relatively simple parametric forms and chooses one based
on the goodness of fit to the observed data for use in subsequent risk characterizations. While a simple
parametric model, such as the linear non-threshold model, is easily interpretable and convenient in risk
communications, it is often overlooked that risk analysis based on such a simple model can be misleading
in evaluating the risk and its uncertainty associated with a low dose exposure. The risk at low doses may
be small but the implication of inappropriately evaluated uncertainty of the risk may not be small in risk
projections since the number of exposed subjects at this dose range is large in many exposed populations.
As an alternative approach, this study considers a Bayesian semi-parametric model that has a piecewise-
linear dose-response function to be smoothed through an auto-regressive structure among the random
slope coefficients defined over closely spaced dose categories. Simulation studies and application to Life
