中国乳腺癌自然史参数选择的优化方法

Breast cancer is one of the most common cancers for females. In order to propose efficient screening strategies and evaluate the effect, one fundamental and important step is to choose proper parameters, i.e., transition rates, in the natural history model of breast cancer in China. There are two challenges to choosing reasonable transition rates. Firstly, the transition rates used in other countries may be not applicable to China, due to the cancer epidemiology. Secondly, little screening sample data are available, making the traditional statistical based approaches such as maximum likelihood methods fail. In this paper, we propose an optimization approach for parameter selection in the natural history of breast cancer in China. A mathematical optimization model is established based on the published statistical data. A coordinate descent algorithm is proposed to solve the resulting box-constrained and highly nonlinear problem, whose objective function does not have an explicit form. Finally, the best matching method is proposed to estimate the distributions of each transition rate. Our approach provides solid foundations to further propose reasonable screening strategies for breast cancer in China and analyze the economic burden of breast cancer.