台灣成人飲食型態、生活型態危險因子、血鐵生化指標及肝功能異常進展之相關性﹕由傳統方法至機器學習綜整分析

TMUIR > College of Nutrition > School of Nutrition and Health Sciences > Dissertations/Theses > Item 987654321/64616

Please use this identifier to cite or link to this item: http://libir.tmu.edu.tw/handle/987654321/64616

Title:	台灣成人飲食型態、生活型態危險因子、血鐵生化指標及肝功能異常進展之相關性﹕由傳統方法至機器學習綜整分析 Associations of dietary patterns, lifestyle risk factors, and iron biomarkers with the progression of liver abnormalities among Taiwanese adults: A comprehensive analysis from traditional methods to machine learning approaches
Authors:	PARAMASTRI, RATHI
Contributors:	保健營養學系博士班趙振瑞
Keywords:	Liver disease;Iron;Dietary patterns;Risk factors;Anthropometric status;Machine learning
Date:	2024-07-12
Issue Date:	2024-11-06 14:57:55 (UTC+8)
Abstract:	Objectives: Liver abnormalities is global health concern ranging from simple disturbance to severe conditions like NASH. We hypothesize the independent or interplay of such modifiable risk factors, such as diet, smoking, sleep patterns, alcohol status, physical activity, anthropometric status, and iron biomarkers, may have influence in development of liver abnormalities. Specifically, our aims are (1) to explore the interaction between dietary pattern associated with liver biomarkers and lifestyle factors with abnormal serum liver enzymes among Taiwanese adults in order to take a step for preventing liver disease. Moreover, to the best of our knowledge, our study is the first one to investigate the synergistic effect of dietary pattern and lifestyle habits on liver dysfunction in Taiwan; (2) to investigate the mediation effect of dietary pattern in the developing of elevation serum liver; (3) to identify the risk factors associated with liver abnormalities employing 5 ML approaches and SHAP value to determine variable importance. Methods: We included adult participants aged 20-45 years of age from MJ database from 2001 to 2015. The total participants enrolled in the database is 190,200, and we selected the participants specifically for each subs of the study. Statistical analysis was performed following the aims of the specific sub studies. Results: Individuals who had the highest levels of both liver-associated dietary patterns and unhealthy lifestyle habits were more likely to develop liver function abnormalities compared to those with the lowest levels of both factors. This association was significant, with an odds ratio of 2.14 (95% CI: 2.02, 2.26), indicating a substantial increase in the likelihood of liver function abnormalities among those with the worst combination of dietary and lifestyle habits. Moreover, the study employed a hierarchical linear regression model to identify predictors of serum AST, ALT, and GGT. The final model revealed that the dairy-meats dietary pattern significantly contributed to the association between serum ferritin and serum ALT, accounting for 38.1% of the overall model. Mediation analysis showed a partially mediated trend in this model, with a Sobel test value of 2.01 (????<0.005p<0.005). Additionally, machine learning models were used to confirm the importance of modifiable risk factors for abnormal LF progression. The top 5 determinants identified were the TG/HDL-C ratio, BMI, body fat percentage, liver/obesity dietary pattern, and LDH. These findings highlight the significant role of dietary patterns and lifestyle factors in influencing liver function and disease progression. Conclusion: Our findings indicate that regular consumption of unhealthy dietary patterns, primarily consisting of processed foods and SSB, is linearly linked to the development of abnormal LF, especially when combined with other unhealthy lifestyle factors. Additionally, our study demonstrates the advantage of employing machine learning approaches, such as XGBoost, to create more accurate prediction models.
Description:	博士指導教授：趙振瑞口試委員：劉沁瑜口試委員：徐建業口試委員：白其卉口試委員：楊素卿口試委員：趙振瑞
Note:	論文公開日期：9999-12-31
Data Type:	thesis
Appears in Collections:	[School of Nutrition and Health Sciences] Dissertations/Theses

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	0	View/Open

Loading...