The important traditional indicators for cancer surveillance and prevention and control are incidence and mortality, but these two indicators only reflect the degree of harm caused by the disease, while the degree and duration of disability caused by the disease is not reflected. DALYs is a comprehensive measure of population health and is classically defined as the total number of healthy life years lost from onset to death, including Years of Life Lost, (YLL) and Years Lived with Disability, (YLD)(18). Since diseases reduce human life expectancy by causing premature death and disability, DALY can be used as a scientific and comprehensive assessment of disease burden in different diseases and regions, and has become an important indicator commonly used in the international cancer disease burden field in response to the needs of today's biopsychosocial medical paradigm shift. Therefore, we used DALY to model the attributed burden of cause and risk in this study. The secondary analysis of data in this study did not require ethical approval and consent from an institutional review board or ethics committee.
Data Sources
The GBD 2019 provided a comprehensive annual estimate of incidence, prevalence, death, and risk factors for 204 countries and territories globally, regionally, and nationally from 1990 to 2019(19). In this study, data from the IHME website (http://www.healthdata.org/), which platform regularly publishes disease and injury incidence, prevalence, mortality, YLL, YLD and DALY indicators, with refinement to differentiate by country, year, sex, and age. The socio-demographic index (SDI) was divided countries into five SDI quintiles(20). The GBD study divides neurological disorders into the following five specific types of data: Alzheimer's disease and other dementias, Parkinson's disease, Idiopathic epilepsy, Multiple sclerosis, Motor neuron disease. Alzheimer's disease and other dementias mortality data from vital registrations, oral autopsies, surveillance systems, surveys, censuses or police reports. For GBD 2019, we modeled the burden of nonfatal disease using DisMod-MR 2.1, which is a meta-regression-Bayesian modeling tool with three steps. Standardization of modeling using the Causes of Death Ensemble Model (CODEm), Spatio-Temporal Gaussian Process Regression (ST-GPR), and DisMod-MR. Furthermore, we estimated smoking, high fasting glucose, and high BMI as risk factors for dementia. For each of these risk factors, we set a theoretical minimum exposure level at which the risk of health outcomes was lowest. Smoking was set at zero; high fasting glucose was greater than 4.5 and less than 5.4 mmol/L; and BMI was set at greater than 20 and less than 25 kg/m 2.
Statistical Analyses
Join-point regression analysis models are often used to identify recent trends in mortality and morbidity data [30]. This study used a Join-point regression analysis model to assess global changes in morbidity, mortality and DALY rates for each SDI quintile based on Annual Percent Change (APC) and Average Annual Percent Change (AAPC).The connection function used in the model is a logarithmic function, and the year is used as the independent variable. Regression fitting was performed on the natural logarithm of age-standardized morbidity and mortality and to calculate the trend changes of morbidity and mortality, as well as the corresponding APC value for each trend segment. Finally, according to the Monte Carlo Permutation method was used to verify that the APC values of each trend segment and the total AAPC values were statistically significant [29]. The above operations were implemented with the Join-point regression program version 4.7.0.0 provided by the National Cancer Institute.
We also demonstrated the burden of Alzheimer's disease in all 21 regions from 1990 to 2019, flexibly modeling the association of morbidity, mortality, and DALY rates with SDI using restricted cubic splines. The above statistical descriptions and analyses all were performed using the R program (version 3.6.0, R Core Team).
Furthermore, since Mason first proposed the APC model in 1973, the application of the APC model has been plagued by the problem of multi-collinearity. The APC model can be essentially considered as a multiple regression model with the following expressions:
Y = log(M) = µ + αagei + βperiodj + γcohortk + ε.
Where M represents the corresponding Alzheimer's disease and other dementias mortality rate, and α, β, and γ denote the effect values of age, period, and cohort estimated by the APC model, respectively. Therefore, we calculated relative risk values (RR values) to help explain the independent effects of age, period, and cohort on Alzheimer's disease and other dementias mortality. In this study, APC analysis was performed using Stata 12.0 software (StataCorp, College Station, TX, USA).