The value of atherogenic index of plasma in nonobese patients with non-alcoholic fatty liver disease: a secondary analysis based on a cross-sectional study. CURRENT

Background Atherogenic index of plasma (AIP) is elevated in fatty liver, but its value in non-obese patients with non-alcoholic fatty liver disease (NAFLD) is unclear. Methods The present study involved non-obese Chinese and Japanese participants. We used univariate and multivariate logistic regression model to determine risk factors. The performances of risk factors were compared by calculation of the area under the receiver operating characteristic curve. Results In the unadjusted model, we observed that odds ratio (OR) for every one standard deviation (SD) increase in AIP was 52.30. In the adjusted model I and model II, the OR for every one SD increase in AIP was 36.57 and 50.84, respectively. The area under the receiver operating characteristic curve for AIP was 0.803 and 0.802 in the development group and in the validation group, respectively. The best cut-off value of AIP in the discrimination between NAFLD and non-NAFLD was 0.005 in the Chinese group and was − 0.220 in the Japanese group, respectively. Our results indicate that AIP and NAFLD have a positive correlation in Chinese and Japanese populations. AIP is an independent correlative factor of non-obese patients with NAFLD, which can be used as a new indicator of health examination.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is a common chronic liver disease in the world. The disease may progress to liver cirrhosis and liver cancer [1,2]. It's well known that obesity is an important risk factor for NAFLD [3]. Liver histology and non-invasive fibrosis tests suggest that some non-obese patients may also have non-alcoholic steatohepatitis and advanced fibrosis, whose body mass index (BMI) is within the normal range [4][5][6].
The atherogenic index of plasma (AIP) is a quantitative indicator to determine blood lipid levels. AIP has good predictive value for dyslipidemia diseases such as atherosclerosis, diabetics, heart disease, etc [7][8][9]. AIP not only has significant value in cardiovascular disease but also has good predictive value for patients with hyperuricemia [10]. Several previous studies reported that there is a positive correlation between high levels of AIP and obesity [11], and AIP has better discrimination for NAFLD in obese people [12]. However, there is not much research done on non-obese people. Non-obese patients with NAFLD should be paid more attention because they often think that fatty liver is not easy without obesity.
In this study, we investigated the relationship between AIP and NAFLD in non-obese patients. We demonstrated that AIP is an independent risk factor of NAFLD in non-obese patients.

Materials And Methods Data sources
We downloaded data from the 'Dryad Digital Repository' website (www.Datadryad.org). This website allows readers to download raw data for free. These data are anonymous. According to the Dryad Terms of Service, we could apply these data for secondary analysis without infringing on the authors' rights. When we employed these data, we cited Chinese data as follows [13]: Association of lowdensity lipoprotein cholesterol within the normal range and NAFLD in the non-obese Chinese population: a cross-sectional and longitudinal study. Dataset website: https://doi.org/10.5061/dryad.1n6c4. Japanese data as follows [14]: Ectopic fat obesity presents the greatest risk for incident type 2 diabetes: a population-based longitudinal study. Dataset website: https://doi.org/10.5061/dryad.8q0p192. Variables included in the Chinese database file were as follows: age, sex, γ-glutamyltranspeptidase (GGT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total protein (TP), albumin (ALB), globulin (GLB), total bilirubin (TB), direct bilirubin (DBIL), blood urea nitrogen (BUN), creatinine (Cr), Estimated Glomerular Filtration Rate

Study Design And Participants
The Chinese study population was from January 2010 to December 2014. Participants took part in the health examination at the First Affiliated Hospital of Wenzhou Medical University. A total of 78304 participants were recruited and selected on the basis of exclusion criteria. Exclusion criteria: (1) participants lacking in required data; (2) excess alcohol consumption (men daily alcohol intake of more than 20 grams, women's consumption more than 10 grams a day); (3) known liver disease; (4) BMI ≥ 25 kg/m 2 ; (5) LDL-c > 3.12 mmol/L; (6) participants taking antihypertensive agents, antidiabetic agents or lipid-lowing agents. Diagnosis of fatty liver was in accordance with the ultrasound diagnostic criteria of the Chinese Liver Disease Association [15]. AIP is meant by the base-10 logarithm of the ratio of the concentration of TG to HDL-c. The formula is AIP = log (TG / HDL-c) [16].
The unit of concentration is mmol/L. BMI is calculated as weight in kilograms divided by height in m 2 , which is the body fat index. All biochemical values were analyzed by an automatic measurement analyzer (Abbott) using standard methods. Among the research population in Japan, a total of 12,932 participants were recruited and selected according to exclusion criteria. Exclusion criteria: (1) participants missing important data; (2) known liver disease; (3) men daily alcohol intake of more than 60 grams, women's consumption more than 40 grams a day; (4) used drugs; (5) fasting blood glucose ≥ 6.1 mmol/L; (6) BMI ≥ 25 kg/m 2 . Because this study was a secondary study and the data was anonymous, no informed consent was required. More specific details were presented in the original report [14] .

Statistical analysis
The overall process of statistical analysis in this study consisted of five steps. Firstly, we divided the Chinese study population into the development group and the validation group according to 7:3.
Continuous variables were expressed as the means ± standard deviations (SD) (normal distribution) or medians (quartiles) (skewed distribution), and categorical variables were expressed as the frequency or percentages. One-way ANOVA (normal distribution), Kruscal Whallis H (skewed distribution) test and chi-square test (categorical variable) were used to identify differences between different groups.
Secondly, we used univariate and multivariate regression analysis to identify risk factors in the development group. Independent variables were tested for collinearity and excluded with variance inflation factor (VIF) > = 10. Collinear VIF = 1 / (1-Rsquared) [17]. The subgroups were grouped using a linear regression model. Thirdly, according to the recommendation of the STROBE statement [18], we showed the results of the unadjusted, minimally adjusted analysis, and fully adjusted analysis.
Fourthly, the area under the receiver operating characteristic (AUROC) curve of each predictor was used to compare the predictive utility. Fifthly, we used box plots to intuitively reflect the predictive value of cut-off. All tests were two-sided. A p-value < 0.05 was considered statistically significant.
Statistical packages R (version 3.4.3, The R Foundation; http://www.r-project.org) and GraphPad Prism (version 8.0; GraphPad Software) were used for statistical analysis.

Baseline Characteristics of participants
As shown in Table 1, there were a total of 78,304 Chinese participants in our study. In the development group, there were 23,265 women and 31,608 men with an average age of 44.6 years.
The average values of age, TP, ALB, GLB, TB, BUN, Cr, eGFR, UA, FPG, TC, LDL-c and BMI were larger in patients with NAFLD than in patients with non-NAFLD. The median values of GGT, ALT, AST, TG and AIP were greater in patients with NAFLD than in patients with non-NAFLD. There was the same trend in the validation group. In the Japanese participants, we included 11,598 NAFLD and 1334 non-NAFLD.
The values of age, GGT, ALT, BMI, AIP, TG and FPG were greater in patients with NAFLD than in patients with non-NAFLD (Table S1).    The results of univariate and multivariate regression analysis in the development group Univariate and multivariate regression analysis results were shown in Table 2  positively correlated with NAFLD and the results were stable (Fig.S1). For further sensitivity analysis, we converted AIP to categorical variable processing, and the results obtained were consistent (Table 3). The similar results were seen in the Japanese population (Table S2). Considering that AIP, age, GGT, ALT, ALB, eGFR, DBIL, UA, FPG, LDL-c and BMI were independent risk factors for NAFLD, we evaluated their diagnostic performance for NAFLD. AIP, the highest AUROC in these indicators, had the better discrimination capacity (AUROC: 0.803, 95% CI: 0.798-0.808) in the development group ( Fig. 1 and Table 4). DBIL gave the worst performance (AUROC: 0.516, 95% CI: 0.509-0.523). In the validation group, BMI performed the best (AUROC: 0.808, 95% CI: 0.801-0.814). AIP ranked second (AUROC: 0.802, 95% CI: 0.795-0.810). AIP had also the better discrimination capacity (AUROC: 0.798, 95% CI: 0.787-0.810) in the Japanese group ( Fig.S2 and Table S3). Then we determined the best cutoff value based on the maximum Youden index of the AUROC curve. As shown in Fig. 2, the best cutoff value of AIP in the discrimination between NAFLD and non-NAFLD was 0.005 in the Chinese group and was − 0.220 in the Japanese group, respectively.  The unit is mmol/L: HDL-c, LDL-c, GGT, DBIL, FPG and UA; The unit is U/L: ALT and ALB.

The Results Of Subgroup Analysis
Subgroup analysis was shown in Table 5. The test for interactions was significant for sex, Age, AST, TB, BUN, UA, TC, TG, HDL-c, GGT, ALT, eGFR, FPG and BMI (p < 0.01), while the test for interactions was not statistically significant for ALB, GLB, DBIL and LDL-c (p > 0.05). Although all variables were risk factors, they did not destroy the correlation between AIP and NAFLD. Compared with patients more than 60 years old, AIP in patients under 60 years old was associated with higher risks of NAFLD

Discussion
The purpose of this study is to analyze the relationship between AIP and NAFLD as well as to verify the diagnostic value of AIP in non-obese patients with NAFLD. We found that AIP is an independent risk factor for NAFLD in non-obese patients through univariate and multivariate regression analysis, which is consistent with the results in obese-patients [12]. We further found that AIP is positively correlated with NAFLD in non-obese Chinese and Japanese patients. Subgroup analysis confirmed that many variables do not affect the positive correlation between AIP and NAFLD. In addition, our results showed that AIP has the better diagnostic value than the other variables in Chinese and Japanese patients, which implies that AIP as a diagnostic indicator is applicable to different regions and ethnicities. However, the best cut-off value of AIP in the discrimination between NAFLD and non-NAFLD was different in Chinese and Japanese patients, demonstrating that AIP has different standards for different regions and ethnicities. Our research provides a reference value for health examination.
Previous studies focused mainly on obese-patients with BMI greater than 25, causing people with a normal range of BMI to often ignore their eating habits [19]. Therefore, our study is alarming for people with BMI in the normal range. Because some studies have indicated that the prevalence of NAFLD in non-obese patients varies greatly in different regions [20,21], our study included populations from Chinese and Japanese patients to exclude regional disparity. For the positive correlation between AIP and NAFLD, we have the following assumptions: 1. Dysfunctional, expanded and inflamed adipose tissue may be related to these NAFLD patients with normal-range BMI [22]. AIP is from traditional lipid profiles, but it is better than traditional pro-atherogenic lipid profiles [23]. We assume that AIP as an excellent combination factor has good performance in the diagnosis of NAFLD.
2. Insulin resistance is just an independent risk factor for NAFLD in non-obese patients [24]. AIP is an independent predictor of insulin resistance [25][26][27][28]. Therefore, we can assume that there is a correlation between AIP and NAFLD. As a matter of fact, our study supports this hypothesis.
Previous research has indicated that TG/HDL-c can be used as predictors of NAFLD in non-obese patients [29]. This study confirmed the previous findings and provided new references for clinical medicine. Through subgroup analysis, we found that AIP and NAFLD are more closely related in the female group than in the male group, consistent with the previous results [29]. Unexpectedly, we found that AIP is more closely related to NAFLD in non-obese people under 60 years old, which has not been reported in the previous literature. We speculate that this may be related to a slower metabolism in older people. However, the specific mechanism needs to be explored in the future.
Our study has a number of strengths. First, our research was a large sample study with multi-region and multi-ethnic. Second, the study was a retrospective study that contained many confounding factors. We used strict statistics to minimize residual confounding. Third, we found the effective cutoff points for AIP in Chinese and Japanese populations, which provided data support for our clinical diagnosis.
There are several limitations to our study. First, because it was a secondary study, the lifestyles and eating habits of the participating population were not collected. Second, this study focused on the relationship between AIP and NAFLD. There are no further studied on other variables.

Conclusion
In summary, our results indicate that AIP and NAFLD have a positive correlation in Chinese and Japanese populations. AIP is an independent risk factor of non-obese patients with NAFLD, which can be used as a new indicator of health examination.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.