Using Regression Analysis Model to Explore the Hidden Linkage between the CGM Sensor Measured Fasting Plasma Glucose from Sleeping Hours and Body Temperature in Early Morning (both are 90-Days Moving Average Values) over a One-Year Period from 11/21/2020 to 11/21/2021 for a type 2 Diabetes Patient Based on GH-Method: Math-Physical Medicine (No. 553)

Abstract

The author started to measure his finger-piercing fasting plasma glucose (FPG) at his wakeup moment in the morning starting on 1/1/2012. In addition, he began to measure his FPG using a continuous glucose monitoring (CGM) device at each 15-minute time interval since 5/8/2018. His sensor 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 and average sensor FPG is a mere 1%. During the period of 2015-2017, he investigated the correlation between his postprandial plasma glucose (PPG) and its 19 influential factors. He identified that both warmer or colder ambient weather temperature affects PPG level. However, the role of the ambient weather temperature is not like the carbs/sugar intake amount or the post-meal exercise that serve as the primary influential factors of PPG. The temperature only provides a secondary and weaker influential factor of the PPG formation. Starting from 10/1/2020, he set out to measure his daily body temperature (BT) at his wakeup moment in the early morning as an additional daily biomarker to monitor for COVID-19 infection. Currently, he has collected more than one-year data of his body temperature. He wondered which primary biomarkers would have a connection with his BT. Through a quick and easier time-domain analysis, he identified that his FPG has a very high correlation with BT, finger FPG vs. BT at 73% and sensor FPG vs. BT at 85%, using 90-days moving average values during the one-year period from 11/21/2020 to 11/21/2021. Therefore, he decided to use his CGM sensor FPG as the dependent variable Y and his BT as the independent variable X to conduct a space-domain regression analysis. In summary, his CGM sensor FPG and BT have a high correlation of 85% and variance of 72% using data from 366 observation days. His predicted CGM sensor FPG using his BT as input to achieve a prediction accuracy of 100% is shown in the time-domain diagram with highly comparable waveforms. The predicted FPG equation has a linear variance (R^2) of 1.0 or 100%: Predicted sensor FPG (Y) = 51.846 * BT (X) - 4970.3 The significance F and p-values are extremely small, near zero, which indicates that the data in the dataset are statistically significant. Simply put, based on the data from 11/21/2020 to 11/21/2021, his average FPG value during sleep has an extremely tight relationship with BT in the early morning. As a result, he can guess his FPG value once he knows the BT.