Since 1/1/2012, the author has been collecting various biomedical and lifestyle of ~3 million data related to his health conditions. This includes the medical categories for 4 chronic diseases of obesity, diabetes, hypertension, and hyperlipidemia (m1 through m4), along with 6 categories of lifestyle details, including exercise, water intake, sleep, stress, food, and daily life routines (m6 through m10). In early 2018, he studied, researched, and published many articles regarding the risks of having CVD/Stroke based on his developed metabolism index (MI) model. In this paper, he will compare the calculated CVD risks based on the MI model through his developed GH-method: math-physical medicine methodology against the recently calculated 3 CVD risk probabilities based on a traditional statistical regression model but using 3 different input datasets. In this article, he will not repeat the detailed introduction of the regression analysis in the Method section because it is available in many statistics textbook. 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 to be considered as statistically significant. The main purpose is to distinguish the degree of influences from “diabetes alone” and “combined 4 chronic diseases” on his risk probability of having an episode of CVD or Stroke during the same time period. Within the category of diabetes alone or HbA1C, he also differentiates the results based on the different time periods of the collected data: one 14-month period along with an 8-year period. In conclusion, his risk of having a CVD/Stroke is highly connected to the combined medical conditions, weight, glucose, blood lipids, blood pressure and heart rate, during the recent 14-month period. However, his risk of having a CVD/Stroke is not strongly related to the main diabetes measuring biomarker of HbA1C during the same period. Of course, any analysis work using various statistical tools must pay attention on the section of the dataset and time period. Within two different time windows, the data distribution pattern and the data variability may differ according to the selected time-window; therefore, the analysis results and conclusions can vary. The above observed two different conclusions based on the same 14-month period is due to the fact that his glucoses have been under stringent control (with an average HbA1C of 6.1%). Therefore, the variance during the 14-months using A1C as the input is a miniscule 0.4% compared to the variance using medical condition as input is an exceptionally high 89%. If we examine the space-domain diagrams closely, using HbA1C as the input, the data results in the scattered map are spread out all over while a straight trend-line has a high difficulty to represent or simulate the majority of his CVD risk data. That is why its variance (R^2) is a mere 0.4%; therefore, the predicted CVD curve using A1C as the input is almost a horizontal line (similar to his A1C curve) and completely out-of-synch with the MI-model calculated CVD risk curve. On the contrary, by using the medical condition as the input, the scattered results are located within a narrow data bend from the lower left corner to the upper right corner, whereas the straight trend-line represents 89% of the total CVD risk data. As a result, the predicted CVD curve using the medical condition as input is almost identical with the MI-model calculated CVD risk curve in the time-domain chart. Furthermore, if he uses his finger A1C from the past 8 years (2014 to 2021), the scattered results are also located within a narrow data bend from the lower left corner to the upper right corner, while the straight trend-line represents 82% of the total CVD risk data. This conclusion is a result of the A1C curve trend which fluctuates and matches his CVD risk waveform from the past 8 years. The findings from this article have demonstrated the importance of selection of both data and time-window which directly influence the final results and conclusions, if we uses statistical methods as our biomedical research tools.
Read full abstract