Predicting the Fasting Plasma Glucose Level Using Two Linear Regression Equations from Body Temperature and Weight in the Early Morning as Inputs over a 14-Month Period for a type 2 Diabetes Patient Based on GH-Method: Math-Physical Medicine (No. 558)

Abstract

The author is a 27-year type 2 diabetes (T2D) patient, who has self-studied and researched diabetes, endocrinology, and chronic disease induced complications since 2010. He is a mathematician and engineer but not a medical doctor; therefore, he does his best to derive some mathematical equations or formulas with sufficient accuracy to describe the observed biomedical or biophysical phenomena. His medical research work started with the task of collecting big data on his own biomarker values and lifestyle details. To date, he has collected and processed nearly 3 million data related to his health. The data in this article covers a few categories. Since 1/1/2012, he has accumulated data on his body weight in the early morning. Beginning on 1/2/2013, he measures his finger-piercing fasting plasma glucose (FPG) at the wakeup moment in the morning. In addition, starting on 5/8/2018, he measures his FPG using a continuous glucose monitoring (CGM) sensor device at 15-minute time intervals. His sensor FPG uses the average glucose value between 12:00 midnight and 07:00 AM for a total of 29 glucose values. Incidentally, the difference between his average finger FPG (104.6 mg/dL) and average sensor FPG (106.8 mg/dL) over the 3.5-year period from 5/8/2018 to 11/27/2018 is a mere 2%. In addition, since 10/1/2020, he measures his body temperature (BT) and finger blood oxygen levels at the wake-up moment in the morning as biomarkers to monitor for possible COVID-19 infection. Currently, he has over one year’s worth of data on his BT and wondered which primary biomarkers would have a connection. Through a quick and easy time-domain analysis, he identified that his FPG has an extremely high correlation with BT, using the 90-days moving average data, finger FPG vs. BT at 73%, and using the 90-days moving average data, sensor FPG vs. BT at 85%, over the one-year period from 11/21/2020 to 11/21/2021. Several years ago, he identified a strong correlation (≥ 90%) existing between his finger FPG and body weight (Weight). In this article, he decided to use his CGM sensor FPG as the dependent variable Y and BT along with body weight as the independent variables X to conduct a space-domain regression analysis. The purpose is to develop a series of linear equations using BT and Weight as the inputs to quickly determine the desired output, his guesstimate sensor FPG value having a high prediction accuracy percentage without applying either finger-piercing device or CGM sensor device to measure his FPG level in the morning. In summary, there are 3 observed conclusions as follows: (1) There are no observed high correlation (-43%) existing between his measured BT and Weight. Furthermore, from the two-regression predicted FPG via BT and Weight as individual inputs, both have different and are not highly correlated (69% for FPG via BT and 12% for FPG via Weight) with his measured FPG over this 14-month period. However, the average regression predicted FPG, i.e. (FPG via BT + FPG via weight) / 2, has a high correlation of 82% (variance 67%) with his measured FPG over the same period.(2) The regression model derived three average predicted FPG data: FPG via BT, FPG via Weight, and average predicted FPG. They have extremely high prediction accuracies (96%-99%) in comparison with the average measured FPG value. This means that the 3 predicted FPG equations are accurate enough for the time period. The regression predicted FPG equations are listed as follows: Body Weight Case Predicted finger FPG (Y) = 4.314 * Weight (X) - 629.34 Body Temperature Case Predicted sensor FPG (Y) = 51.85 * BT (X) - 4970.28 (3) In conclusion, the average predicted FPG equation is the most suitable for predicting his FPG value (either sensor FPG or finger FPG values). Therefore, he can guesstimate an accurate FPG value using his BT and Weight as inputs, without applying either finger-piercing or CGM sensor devices for measuring his FPG value in the morning.