The prediction model of fatty liver was established by blood biochemical indexes

Background The main cause of chronic liver disease is fatty liver, which includes alcoholic fatty liver and nonalcoholic fatty liver disease. This study is aimed to establish the prediction model of fatty liver, and provide help for the prevention and treatment of fatty liver, especially NAFLD in the future. Methods Datasets from 2017 to March 2020 NHANES required for the analysis were downloaded from the NHANES web site and R 4.1.1. Software was used for data analysis. A total of 3762 subjects were enrolled in this study, which were divided into model construction group and model validation group in a 2:1 ratio. Results The study selected 6 indicators to build the prediction model, which are as follows: ALT, Platelet count, Creatinine, LDH, HS C-Reactive Protein, Glucose. Then the prediction model was constructed. The area under ROC curve of the model was 0.7471. In the validation population, the area under the ROC curve of the model was 0.7816.


Introduction
The main cause of chronic liver disease is fatty liver, which includes alcoholic fatty liver and nonalcoholic fatty liver disease (NAFLD). It includes complex pathological manifestations, from simple fat accumulation to non-alcoholic steatohepatitis (NASH), which is an important factor in other metabolic disorders, and also an important cause of type 2 diabetes and cardiovascular disease (1). In the United States, NAFLD has affected the lives of a quarter of people and has become the second leading cause of liver transplantation (2). Alcoholic liver disease is a related disease state characterized by early fatty liver (alcoholic steatosis). NAFLD and alcoholic steatohepatitis are both important causes of serious complications such as HCC and liver cirrhosis (3)(4)(5)(6).
The pathophysiological feature of fatty liver is the accumulation of large amounts of fat in liver parenchymal cells (liver cells). Under normal circumstances, these cells are particularly adept at storage and excretion of fat to promote the body's metabolic needs. However, in disease states, they are prone to accumulate fat at a super-pathological level (1). NAFLD is characterized by excessive accumulation of triglycerides in the liver, which is a common complication of obesity and is related to insulin resistance and metabolic syndrome (7). According to MRI evaluation, more than 5% of the fat in the liver is considered to be early liver steatosis(8). It is characterized by lipid apoptosis, in ammation and brosis of liver cells (9).
The diagnosis of fatty liver requires imaging and histological examination, while the diagnosis of NAFLD is more complicated (10,11). The initial challenge was when to suspect fatty liver, because most patients are asymptomatic. Due to the lack of cost-effective, non-invasive tests with high predictive value, and the lack of effective treatments, comprehensive NAFLD screening is not currently recommended (12).
Therefore, it is necessary to explore a new effective clinical prediction model for fatty liver, so that the high-risk population can change their lifestyle earlier, better prevent the occurrence of fatty liver and reduce the incidence of various complications. In this study, the clinical prediction model of fatty liver was established by using the database of National Health and Nutrition Examination Surveys (NHANES), and its good predictive effect was veri ed, which could become an important means for the prevention and treatment of fatty liver in the future.

Study Sample
Datasets from 2017 to March 2020 NHANES required for the analysis were downloaded from the NHANES web site. After excluding the information de cit group, a total of 3762 subjects were enrolled in this study, which were divided into model construction group and model validation group in a 2:1 ratio. The characteristics of the study population are shown in Table 1.

Prediction model Construction
The model was constructed with the included indicators in 2508 subjects.  It can be seen that F score has a good predictive effect on fatty liver in the validated population.

Discussion
Liver function enzymes are important markers of the severity of many liver diseases, especially aspartate aminotransferase (AST) and alanine aminotransferase (ALT) (13). The study by Helene Gellert-Kristensen et al. found that the concentration of ALT has an important correlation with the risk of fatty liver, cirrhosis and HCC (14). And ALT is an independent related factor of metabolic syndrome, obesity, diabetes and other diseases (15,16). Alcohol intake is an important factor leading to fatty liver and one of the important reasons leading to elevated ALT. Moderate alcohol consumption does not cause a signi cant increase in ALT, but long-term alcohol intake can cause an increase in ALT and liver changes (17)(18)(19).
NAFLD is closely related to ALT activity and is a common cause of unexplained ALT elevation (20,21). Unexplained elevated ALT should consider the possibility of NAFLD(16). This is also consistent with our experimental results, ALT is one of the most important factors related to fatty liver.
Studies by Muhammad et al. have shown that creatinine is signi cantly elevated in patients with fatty liver, which is obviously correlated with fatty liver (22). Creatinine is an important indicator of model for end-stage liver disease (MELD) score, which measures liver function and predicts survival in patients with liver disease (23)(24)(25)(26).
NAFLD is closely linked with hepatic insulin resistance (27,28). It is the liver manifestation of metabolic syndrome, and increased glucose is one of the important manifestations (29). Insulin resistance is a key pathogenic factor of metabolic syndrome and the most common risk factor for NAFLD (30)(31)(32). In patients with insulin resistance, insulin cannot inhibit liver glucose production, but it continues to stimulate adipogenesis, leading to hyperglycemia, hyperlipidemia, liver steatosis and type 2 diabetes (33,34). However, in our study, insulin content was not an important correlation factor of fatty liver. This may be because the level of insulin does not directly re ect the liver insulin resistance like glucose (35,36).
Researches showed that platelet count is related to NALFD and liver cirrhosis (37)(38)(39), and can re ect the degree of liver injury (40). Studies have also shown that platelet count is closely related to insulin resistance, and its severity and complications (41,42). This may have the following reasons: (1) the in uence of portal hypertension; (2) splenic sequestration of platelets; (3) liver damage may also cause TPO release defects and reduce platelet production in the bone marrow (43)(44)(45).
There are also some shortcomings in this study, for example, the included population is from the NHANES database, and the model still needs prospective research veri cation. Whether there are other indicators that can improve the accuracy of the model also needs to be explored in the future, but at present, the model can predict the occurrence of fatty liver with satisfactory accuracy by using blood biochemical indicators.

Conclusion
F score has a good predictive effect on fatty liver, which can be used as an important means for the prevention and treatment of fatty liver, especially NAFLD in the future. The data were obtained from NHANES database and no ethical review was required.

Consent for publication
Not applicable.

Availability of data and materials
All data for the study can be obtained from the corresponding author and author

Competing interests
The authors declare that they have no competing interests.   The ROC curve of the prediction model in the model building population.