The first article of ‘‘AI Medical Scientists” analyzed the related comorbidities of ALS
Background and purpose: It is helpful to study the causes of diseases by discussing their cormorbidities. The cause of Amyotrophic Lateral Sclerosis (ALS), a deadly disease, remains unknown. The study used a medical database to explore the diseases associated with ALS.
Method: This study included 705 newly diagnosed ALS patients over 15 years old between January 1, 2007 and December 31, 2013; 14,100 controls matching in sex, age, residence, and insurance premium. The research data were drawn from the National Health Insurance Research Database and Serious Disabling Diseases database, which were designed as case-control studies for the entire population. The comorbidities diagnosed by physicians prior to the date of initial ALS diagnosis were analyzed during the observation period 1, 3, 5, 7, or 9 years before the first ALS diagnosis, respectively. Chi-square test or t test was used to test the demographic differences between ALS patients and the control group. The relationship between these diseases and ALS risk was studied by using the conditional logistic regression model and the stepwise selection method. The path analysis was used to assess the pathway relationship between the comorbidities and ALS before the initial diagnosis of ALS.
Results: ALS was associated with 28 comorbidities prior to their initial diagnosis, including 17 positive and 11 negative comorbidities. Path analysis showed that these 11 negative associated diseases could be classified as diabetes and its complications. Seventeen positive associated diseases could be classified as metabolic syndrome, neuroinflammation, head trauma, motor injury, infection, and complications.
Conclusion: The results of this study supported the hypothesis that metabolic abnormalities occurred prior to ALS diagnosis. Metabolic disturbance might have an influence on the incidence of ALS, and energy metabolic defects might play a role in the pathogenesis of ALS.
Why do people get a major disease? This question has puzzled human beings for thousands of years. Many medical scientists have spent their lives searching for the answer to it, and their efforts have resulted in an accumulation of medical knowledge. However, the causes of many major and deadly diseases remain unknown, and there is no cure. The research team used a big data database and a self-developed statistical analysis program with the aim of providing information regarding the causes of the disease by means of machine learning, hoping to give the medical scientists a reference for a new research direction, to accelerate the understanding of the causes of specific diseases and the development of new treatments.
How can we prove that the analysis program in this study is correct and effective? In addition to the theoretical and methodological exploration, the application of our analytical program (procedure) in the exploration of the causes of disease can preliminarily verify the effectiveness of the analytical program. The automatic results of the new machine learning analysis program must be largely in line with existing medical knowledge, and the newly discovered information must conform to basic medical theory. This study was the first in a series of 'AI Medical Scientists' verification papers: Amyotrophic Lateral Sclerosis (ALS), also known as motor neuro disease (MND) or Lou Gehrig's disease, associated network analysis of comorbidities in those suffering from the disease. (1) A similar analysis program (procedure) has been successfully confirmed in late-onset Alzheimer's disease. (2) The verification results of such diseases as colorectal cancer, pancreatic cancer, systemic lupus erythematosus and rheumatoid arthritis have been submitted for review.
This study included 705 newly diagnosed ALS patients over 15 years old between January 1, 2007 and December 31, 2013; the control group comprised 14,100 people who were categorized according to gender, age, the insured unit by urban and rural areas and insured premium matching. The research data were drawn from the National Health Insurance Research Database and Serious Disabling Diseases database, which were designed as case-control studies for the entire population. The comorbidities diagnosed by physicians prior to the date of initial ALS diagnosis were analyzed during the observation period 1, 3, 5, 7, or 9 years before the first ALS diagnosis, respectively. Chi-square test or t test was used to test the demographic differences between ALS patients and the control group. The relationship between these diseases and ALS risk was studied by using the conditional logistic regression model and the stepwise selection method. The path analysis was used to assess the pathway relationship between the comorbidities and ALS before the first diagnosis of ALS.
ALS was associated with 28 comorbidities prior to their first ALS diagnosis, including 17 positive and 11 negative comorbidities. Path analysis showed that these 11 negative associated diseases could be classified as diabetes and its complications. Seventeen positive associated diseases could be classified as metabolic syndrome, neuroinflammation, head trauma, motor injury, infection, and complications (Figure 1).
The results of this study support the hypothesis that metabolic abnormalities occurred prior to ALS diagnosis. Metabolic disturbance might have an influence on the incidence of ALS, and energy metabolic defects might play a role in the pathogenesis of ALS.
Figure I: Path analysis model for prior diseases associated with amyotrophic lateral sclerosis (ALS).The red and blue lines indicated a positive (728, 354, 783, 333, 353, 920, 434, 722, 357, 355, 356, 682 and 717) and negative (250, 709, 521, 532, 536, 414, 382, V67 and 924) direct links of prior diseases with ALS, respectively.. The International Classification of Diseases codes, the ninth revision, are displayed in the boxes.
Source：https://rh.acad.ntnu.edu.tw/en/article/content/49
Tzu-Chi Lee Professor | Department of Health Promotion & Health Education
Professor Lee does research on large health databases. There have been more than 50 published SCI/SSCI papers on statistical analysis of big data as the first or corresponding author, some of which were published in JAMA network, British Journal of Psychiatry and other well-known academic journals. Professor Lee's interest lies in the development of big data statistical analysis programs, combined with self-developed high-efficiency analysis programs, epidemiology and genomics research methods; with the validation of actual database to proceed analysis program and empirical medical database under the rigorous methodological framework of the clinical medical research potential.