Predicted Cancer Risk Probability using Linear and Nonlinear Regression Models with Medical Condition Inputs from Chronic Diseases and Lifestyle Details Collected Data of a Type 2 Diabetes Patient Based on GH-Method: Math- Physical Medicine (No. 550)

Abstract

The author has spent approximately 40,000 hours over the past 12 years self- studying and researching internal medicine branches, with a focus on endocrinology and diabetes. Since 2018, he has expanded his interest, learning and research work into other medical branches related to lifestyle, metabolism and immunity. Currently, his type 2 diabetes (T2D) is well under control, where the HbA1C level decreased from 10% in 2010 down to 5.8% in 2021 without medication intervention. Naturally, he is concerned about other life-threatening diseases of the elderly population, specifically cancers and dementia. Over the past decade, he has written and published more than 540 medical papers in various medical journals. In total, he applied about 30 different research methodologies based on his developed GH-method: math-physical medicine, including physics theories, engineering modeling, mathematical equations, computer science tools of big data analytics and artificial intelligence (AI), as well as some traditional statistical approaches to explore and interpret various biomarkers and their biophysical phenomena. However, the majority of published medical research papers he has read to date are primarily based on statistics (~90% of his total reading volume of ~2,000 papers). In this particular article, he decides to follow the majority of other medical scientists’ footsteps, to use the traditional statistical regression model with linear and various nonlinear formulas involving multiple independent variables to investigate his overall risk probability of developing cancer versus 4 categories of metabolic disorder induced chronic diseases (obesity, diabetes, hypertension, and hyperlipidemia) and 6 categories of lifestyle details (food, water, exercise, sleep, stress, and daily life routines). In this paper, he will not repeat the detailed introduction of the regression analysis models in the Methods section because it is available in many statistics text books. It should be noted that in regression analysis, the correlation coefficient R should be > 0.5 or 50% to indicate a strong inter-connectivity and the p-value should be < 0.05 or 5% to be considered as statistically significant. The author recently studied a consensus report published jointly by the American Diabetes Association (ADA) and the American Cancer Society (ACS) in 2010 regarding relationships between cancers and diabetes. Based on the information from this report plus ~3 million collected data of his own overall metabolism situation, including medical conditions and lifestyle details, he decided to conduct a research study regarding the estimation of his overall and relative risk probability of having cancer over two time periods: the longer 12-year period from 2010 to 2021 and the shorter 9-year period from 2013 to 2021. The reason for using two time-periods is due to his insufficient data gathering and guesstimateddata from 2010 and 2012. He would like to investigate the prediction difference. In summary, from the time-domain analysis, the two cancer risk waveforms, metabolism index (MI) based cancer risk curve and regression predicted cancer risk curve, are similar to each other in terms of shape similarity. They have an extremely high of 92% - 98% correlation between the curves and with a 100% prediction accuracy of averaged cancer risks.