Analysis of risk factors and establishment of predictive model for elevated ALT in adult patients with nonalcoholic fatty liver disease

Shao-Jie Duan, Zun-Jing Liu, Jia-Liang Chen, Shu-Kun Yao

1. Graduate School of Beijing University of Chinese Medicine,Beijing 100029,China

2. Department of Neurology,China-Japan Friendship Hospital,Beijing 100029,China

3. Center of Integrative Medicine,Beijing Ditan Hospital,Capital Medical University,Beijing 100015,China

4. Department of Gastroenterology,China-Japan Friendship Hospital,Beijing 100029,China

Keywords:Non-alcoholic fatty liver disease Alanine aminotransferase Tongue characteristics Risk factors Prediction model

ABSTRACT Objective: To explore the risk factors of elevated alanine aminotransferase (ALT) in patients with nonalcoholic fatty liver disease (NAFLD), and to establish a risk prediction model of elevated ALT in patients with NAFLD. Methods: A total of 200 NAFLD subjects were enrolled in a third-class hospital in Beijing. The relevant clinical indexes and TCM tongue picture data were collected standardized. According to the elevation of ALT, the patients were divided into ALT elevation group and ALT normal group. The independent risk factors of ALT elevation were obtained by logistic regression analysis. Based on this, the logistic regression prediction model of ALT elevation in NAFLD patients was established, and the calibration of the model was evaluated by Hosmer-Lemshow goodness-of-fit test. The area under the subject's working characteristic curve (AUROC) was used to test the discrimination of the model. Results: The multivariate logistic regression analysis showed that the OR value of male, obesity, elevated total cholesterol (TC), elevated triglyceride (TG) and prickly tongue were 6.059, 2.216, 2.649, 2.106, 3.646, respectively, and the P-values were all ＜ 0.05. The AUROC of logistic regression prediction Model 1 (without prickly tongue) and Model 2(include prickly tongue ) were 0.771 (95%CI:0.703-0.840) and 0.801 (95%CI:0.736-0.866),respectively, and the maximum Youden index, sensitivity and specificity were 0.414, 0.829,0.585 and 0.478, 0.686, 0.792, respectively. Conclusion: Male, obesity, elevated TC, elevated TG and prickly tongue were independent risk factors for elevated ALT in NAFLD patients;This study established an integrated traditional Chinese and Western medicine model that includes the tongue characteristics, which have certain clinical value in predicting the risk of elevated ALT in patients with NAFLD, and are worth popularizing and applying.

1. Introduction

Nonalcoholic fatty liver disease (NAFLD) is a pathological syndrome with over 5% hepatocyte steatosis, excluding the condition of long-term heavy drinking and other definite liver diseases[1]. The disease spectrum ranges over nonalcoholic steatosis, nonalcoholic steatohepatitis (NASH), liver cirrhosis, and hepatocellular carcinoma (HCC) [2]. In recent years, the prevalence rate of NAFLD is rising rapidly and NAFLD patients tend to be younger, which has become the largest number of chronic liver disease and the primary cause of abnormal liver biochemical indexes in a health examination in China[3-6]. Alanine aminotransfease(ALT), as an important biochemical indicator for evaluating liver damage, is closely related to liver steatosis, inflammation,oxidative stress, insulin resistance, etc. If ALT continues to rise,it will not only accelerate the continuous progress of NAFLD,NASH, liver cirrhosis and liver cancer as well as increase the risk of metabolic syndrome and its components (such as hypertension,hyperglycemia, hyperlipidemia, etc.) [7,8]. Therefore, it is of great clinical significance to strengthen the early prediction and timely intervention of elevated ALT in NAFLD patients.

As an important part of “ Wang ” (observation of the patient’s complexion, tongue, expression, etc. ), one of the four diagnosis practices of traditional Chinese medicine(TCM), the tongue diagnosis is a non-invasive and effective means, which has an important application value in the diagnosis and treatment of liver diseases [9].The purpose of this study is to explore the risk factors of elevated ALT in NAFLD patients by observing and analyzing the tongue image characteristics and related clinical indexes of NAFLD patients and try to establish a related risk prediction model of elevated ALT in NAFLD patients by integrated traditional Chinese and western medicine based on tongue diagnosis, which includes tongue image characteristics and related clinical indexes, to provide new ideas for integrated traditional Chinese and western medicine prevention and treatment of this disease.

2. Materials and methods

2.1 Sample sources

The samples of this study were all collected from the physical examination center of China-Japan Friendship Hospital in Beijing.Through the standardized process, we collected the subjects’demographic data, anthropometric indicators (height, weight, etc.),and laboratory indicators (blood lipids, blood sugar, uric acid, liver function, etc.), as well as standardized tongue image characteristics.According to the diagnostic criteria, inclusion, and exclusion criteria,a total of 200 subjects diagnosed as NAFLD were included in this study, including 136 males and 64 females, aged (40.22±10.57)years old. According to the ALT scale, they were divided into the ALT elevated group (70 cases, accounting for 35%) and ALT normal group (130 cases, accounting for 65%).

This study had been approved by the Clinical Research Ethics Committee of China-Japan Friendship Hospital (2018-110-K79-1),and all the subjects had signed the informed consent form.

2.2 Diagnostic criteria

The diagnostic criteria of NAFLD refer to the “Guidelines For the Prevention and Treatment of Non-alcoholic Fatty Liver Disease(updated in 2018)” revised by the Fatty liver and alcoholic liver Disease Group of the Hepatology Branch of the Chinese Medical Association and the Fatty liver Disease expert Committee of the Chinese Medical Doctor Association[13]. The diagnosis of fatty liver in this study is based on the ultrasonic diagnosis. Fatty liver was diagnosed according to the characteristics of hepatic anterior echo enhancement (“bright liver”), far-field echo attenuation, and unclear display of intrahepatic duct structure.In addition, ALT ＞ 40u/L was diagnosed as elevated ALT; body mass index (BMI) ≥ 28kg/m2 was defined as obesity; BMI ≥ 24kg/m2 was defined as overweight; waist circumference (WC) male 90cm or female 85cm is defined as abdominal obesity; systolic blood pressure (SBP) ≥140mmHg and/or diastolic blood pressure (DBP)90mmHg or those who have taken antihypertensive drugs within the past 2 weeks are considered to have elevated blood pressure ; Fasting blood glucose (FBG) ≥6.1mmol/L means elevated blood glucose;total cholesterol (TC) ≥ 5.18mmol/L as elevated TC; triglyceride(TG) ≥ 1.70mmol/L as elevated TG. Uric acid (UA) ≥ 420 umol /L (male) or ≥ 360 umol / L (female) as elevated UA.The classification standard of tongue image characteristics refers to the content of “looking at tongue” in the textbook Diagnostics of Traditional Chinese Medicine [11] and the formulation of standard tongue diagnosis atlas[12,13].

2.3 Inclusion criteria

(1) Diagnostic criteria by following per under NAFLD.

(2) 18 to 65 years old.

(3) Complete data collection as required and voluntarily sign informed consent.

2.4 Exclusion criteria

(1) Long-term history of smoking and drinking.

(2) Pregnant or lactating women.

(3) Patients with severe brain, heart, lung, kidney, blood system diseases, mental diseases, infectious diseases, and malignant tumors.

(4) Fatty liver or liver damage caused by drugs, alcohol,autoimmunity, and other factors.

2.5 Collection and analysis of tongue images

Under the premise of excluding all food and drug contaminate tongue coating, patients are required to sit upright in the natural light environment, naturally, extend the tongue body out of the mouth, relax the tongue body, spread the tongue surface, the tip of the tongue slightly downward, fully expose the tongue body. We use the intelligent tongue diagnosis instrument of Xinyi Health Technology (parameters: screen resolution 800 × R.G.B × 1280,image resolution 4224 × 3136) to collect tongue images and carry out artificial intelligence color correction. Then, two trained Chinese medicine practitioners with the title of deputy senior high and above will label each tongue image according to the standardized rules, and a unified opinion should be made after consultation with the other chief Chinese medicine practitioner when they have inconsistent results.

2.6 Statistical analysis

All statistical analyses were performed using SPSS software version 20.0 (IBM Corp., Armonk, New York, USA). All statistical tests were two-tailed, and comparisons where P ＜ 0.05 were considered statistically significant. Categorical variables were expressed as frequency (percentage) and analyzed by Chi-squared tests. Normally distributed continuous variables were expressed as the mean ± SD and differences were analyzed with an independent sample t-test.Skewed variables were expressed as median with the interquartile range (IQR) and differences were analyzed with Mann-Whitney U test.

Through univariate logistic regression analysis, the covariates of P ＜ 0.05 were further incorporated into a multivariate logistic regression analysis to screen independent risk factors, and then the risk prediction model of elevated ALT in NAFLD patients was established. The Hosmer-Lemeshow goodness-of-fit test was used to evaluate the calibration of the model. If P ＞ 0.05, it indicates that the goodness-of-fit of the model is higher and the prediction model has good calibration ability. The area under the subject working characteristic curve (AUROC) and its 95% confidence interval(95%CI) were used to evaluate the discrimination of the model,and the corresponding maximum Youden index, sensitivity, and specificity were calculated. If 0.9 ≥ AUROC ＞ 0.7, it suggests a certain accuracy, and if AUROC ＞ 0.9, the accuracy is very high.All statistical results are expressed as statistically significant with P＜0.05.

3. Result

3.1 Analysis of baseline characteristics of subjects

The clinical baseline characteristics of all participants are presented in Table 1. This study found that the prevalence rate of the male in the ALT elevated group was significantly higher than that in the ALT normal group (88.6% VS 56.9%, P ＜ 0.001). The average age of the ALT elevated group (37.81 ±10.36 ) years was significantly lower than that of the ALT normal group (41.51±10.50) years (P ＜ 0.05).Inspect of anthropometrics, the levels of BMI, Waist circumference(WC), and Waist-to-hip ratio in the ALT elevated group were significantly higher than those in the ALT normal group. Inspect of laboratory tests, the levels of UA, TC, and TG in the ALT elevated group were significantly higher than those in the ALT normal group,but the level of HDL-C in the ALT elevated group was significantly lower than that in the ALT normal group. In terms of tongue characteristics, the incidence of tongue pricks in the elevated ALT group was significantly higher than that in the normal ALT group(75.7% VS 49.2%,) P＜0.001. Besides, there was no significant difference in SBP, DBP, and GLU between the two groups.

3.2 Univariate logistic regression analysis

The above statistically significant indicators were used as independent variables and assigned values, as follows: gender(male=1, female=0), age＜40 years old (yes=1, no=0), obesity (yes=1,no= 0), TC rises (yes=1, no=0), TG rises (yes=1, no=0), SUA rises (yes=1, no=0), prick the tongue (yes=1, No=0); According towhether ALT is elevated as a dependent variable (ALT elevation=1,ALT normal=0). The univariate logistic regression analysis showed that seven variables including male, age＜40 years old, obesity,elevated TC, elevated TG, elevated SUA, and tongue prick were all statistically significant (all P values ＜0.05)(Table 2).

Table 1 Baseline characteristics of study participants

Table 2 Univariate logistic regression analysis

3.3 Multivariate logistic regression analysis

According to the screening results of the above univariate analysis,seven variables including age ＜ 40 years old, male, obesity, elevated UA, elevated TC, elevated TG, and prickly tongue prickly were included in multivariate logistic regression analysis. The results showed that male, obesity, elevated TC, elevated TG and tongue prickly were independent risk factors of elevated ALT in NAFLD patients. Among them, the risk of elevated ALT in NAFLD patients with tongue prick was 3.646 (95%CI: 1.747-7.610) times that of non-tongue prick patients.

3.4 Establishment of prediction model of elevated ALT in NAFLD patients

As shown in Table 4 and Table 5, based on the results of multivariate logistic regression analysis, two prediction models were established in this study. Model 1 (without prickly tongue) was a simple western medicine prediction model. The logistic regression equation is: logit(P1)=-4.607+1.698*X1+0.788*X2+ 0.768*X3+0.917*X4;Model 2 (include prickly tongue) was a prediction model of integrated traditional Chinese and western medicine. The logistic regression equation is: logit(P2)=-5.381+1.802*X1+ 0.754*X2+0.974*X3+ 0.745*X4+ 1.294*X5.

Table 3 Multivariate logistic regression of model 1.

Table 4 Multivariate logistic regression of model 2.

3.5 Hosmer-Lemeshow goodness-of-fit test the calibration of the prediction model

The goodness of fit test was performed on the above two models respectively. Model 1: Hosmer-Lemshow χ2 =4.498, P =0.721＞0.05; Model 2: Hosmer-Lemshow χ2 =5.331, P =0.722＞0.05.The results show that both of the two prediction models have good calibration.

3.6 AUROC test the discrimination of prediction model

As shown in Figure 1, the AUROC of ,Model 1 and Model 2 were 0.771 (95%CI:0.703-0.840) and 0.801 (95%CI:0.736-0.86), respectively, and the maximum Youden index, sensitivity and specificity were 0.414, 0.829, 0.585 and 0.478, 0.686, 0.792,respectively. At the same time, the accuracy of Model 1 and Model 2 in predicting the increased risk of ALT in patients with NAFLD were both better than that of a single independent risk factor( including age ＜ 40 years old, male, obesity, elevated UA, elevated TC, elevated TG and prickly tongue prickly).

Figure1 ROC curve of risk prediction models for elevated ALT in patients with NAFLD

4. Discussion

Clinically, NAFLD is the primary cause of asymptomatic elevated ALT. Because there are no symptoms, it is not enough to draw people’s attention, which often leads to missed diagnosis and delayed treatment. As an important biochemical index to evaluate liver function damage, ALT is closely related to liver steatosis,inflammation, oxidative stress, and insulin resistance. If ALT continues to increase, it will not only accelerate the progression of NAFLD to NASH, liver cirrhosis, and liver cancer but also increase the risk of metabolic syndrome and its components (such as hypertension, hyperglycemia, hyperlipidemia, etc.). Therefore,it is particularly important to identify the risk factors related to the elevated ALT in NAFLD patients and to achieve an early predictive diagnosis and targeted early intervention. The results of univariate logistic regression analysis show that the elevated ALT in NAFLD patients is closely related to age, sex, obesity, UA, TC, TG, and prickly tongue in TCM. After correcting confounding factors by multivariate logistic regression analysis, male, obesity, elevated TC,elevated TG and prickly tongue were still statistically significant,which are independent risk factors for elevated ALT in patients with NAFLD.

This study showed that the risk of increased ALT in male NAFLD patients was significantly higher than that in female patients (P ＜0.001), which was 5.421-6.059 times. Hu SJ’s study on the level of serum ALT in 11553 middle-aged and elderly residents showed that the level of ALT in males was significantly higher than that in females, which was consistent with the results of this study[14]. It may cause by the difference in estrogen levels between men and women. Previous studies have shown that estrogen can protect hepatocytes by anti-oxidative stress, inhibiting the activation of hepatic stellate cells, protecting the structure and function of mitochondria, and regulating lipid metabolism[15]. Therefore,compared with premenopausal women, male NAFLD patients may be more likely to elevated ALT due to the relative lack of hepatoprotective effect of estrogen, so it should be paid more attention to care about.

Besides, our study shows that NAFLD patients with obesity,elevated TC and elevated TG have an increased risk of elevated ALT by 2.198-2.216 times, 2.195-2.649 times, and 2.106-2.502 times,respectively. Shu Yang’s study on the correlation between obesity and hypersensitive C-reactive protein and liver function index showed that the levels of hypersensitive C-reactive protein and ALT in the obese group were significantly higher than those in the normal weight group (P ＜ 0.001)[16]. It was pointed out that may be related to the chronic low-grade inflammation caused by many bioactive substances from the adipose tissue expansion in obese patients. At the same time, obese patients with abnormal lipid metabolism are usually accompanied by abnormal total cholesterol and triglycerides,which significantly increases the risk of elevated ALT in patients with NAFLD because so many factors interact with each other. In the study of the effect of exercise on the improvement of fatty liver in obese adults, Zhang HJ found that when the bodyweight decreased by 7% to 10%, it could significantly reduce the level of serum ALT and improve NASH[17]. Therefore, strengthening the weight control of NAFLD patients and monitoring the blood lipid levels has important clinical significance for the prevention and treatment of elevated ALT. Of course, in addition to exercise, the most important thing is to improve the dietary structure and reduce dietary calorie intake.Tongue diagnosis, as the main way to judge the inside condition from the observation of outside in traditional Chinese medicine, is one of the main bases of syndrome differentiation and treatment. As the Tongue Identification Guide says: “ By distinguishing the quality of the tongue, you can determine the deficiency and reality of the five internal organs; Looking at the tongue coating, you can know the shallow and deep of the six external factors that cause diseases.”Tongue image characteristics can objectively reflect the deficiency or excess of Zang-fu organs and meridians, as well as the abundance or lack of qi and blood, and has a good prediction of nature, the depth,and the progress and retreat of the disease. Based on a systematic review of tongue diagnosis in the field of chronic liver disease, we found that different tongue image characteristics were closely related to the grading, staging, and prognosis of liver disease. The prickly tongue is a common tongue image characteristic in tongue diagnosis of TCM, which mainly indicates the extreme heat of Zang-fu organs or the heat of blood. From the perspective of view of modern tongue anatomy, it is mainly caused by tongue mushroom nipple congestion, swollen or even hyperplasia, and protuberance, which is often seen in hypermetabolism or a variety of acute and chronic inflammation. The results of this study show that prickly tongue is an independent risk factor for elevated ALT in NAFLD patients,and the risk of elevated ALT in NAFLD patients with the prickly tongue is 3.646 times higher than that in non-prickly tongue NAFLD patients, indicating that NAFLD patients with prickly tongue may be more prone to liver function damage, so clinically, more attention should be paid to monitor the liver function in this kind of NAFLD patients. .According to the TCM connotation of the prickly tongue,clearing away heat and toxic materials, cooling blood, and activating blood circulation should be the treatment principles, which broaden the thoughts of TCM differentiation and treatment of elevated ALT in patients with NAFLD.

Based on the above independent risk factors of elevated ALT in NAFLD patients, we established two logistic regression prediction models, Model 1 is a western medicine prediction model without prickly tongue, and Model 2 is a integrated traditional Chinese and western medicine prediction Model with the prickly tongue. In comparison, the prediction accuracy (0.801 VS 0.771) and specificity(0.792 VS 0.585) of model 2 are slightly higher than that of Model 1, while the sensitivity (0.680 VS 0.829) was slightly lower than that of model 1. According to the analysis of clinical practice, Model 2 with high specificity may better identify NAFLD patients with nonelevated ALT. To some extent, it can be used to reduce misdiagnosis rate and waste of medical resources; while Model 1 with high sensitivity may better identify NAFLD patients with elevated ALT and reduce missed diagnosis rate to a certain extent. On the whole,the two models have good calibration and discrimination, and their ability to predict the risk of elevated ALT in NAFLD patients is significantly better than that of a single independent risk factor, and they can complement each other in different applicable scenarios.For the prediction model of NAFLD, there have been many previous studies[18,19], but there are few reports on the prediction model of ALT elevation in NAFLD patients. This study established a risk prediction model for ALT elevation in NAFLD patients with an integrated traditional Chinese and Western medicine model that includes the characteristics of the tongue for the first time .

In summary, this study found that men, obesity, elevated total cholesterol, elevated triglycerides, and tongue piercing were independent risk factors for elevated ALT in patients with NAFLD.Based on this, the two logistic regression prediction models established based on this have good calibration and discrimination,which can be used as an effective auxiliary diagnostic and predictive tool for NAFLD patients with elevated ALT, and it is worthy of further exploration and research in the future. Of course, this study also has limitations such as a small sample size. In the future, it is still necessary to further include more sample sizes from multiple centers to further verify and improve the model.