Ferritin as a key risk factor for nonalcoholic fatty liver disease in children with obesity

Abstract Background The association between serum ferritin and nonalcoholic fatty liver disease (NAFLD) in children with obesity is not clear. This study was designed to investigate whether serum ferritin can be an independent predictor for NAFLD. Methods According to the hepatic ultrasound results, a total of 347 children with obesity were enrolled in this study. Among them, 95 patients with NAFLD and 95 without NAFLD were matched for gender, age, blood pressure and body mass index, the odds ratios (OR) and 95% confidence intervals (CI) for the association of ferritin and the risk of NAFLD were analyzed. Results After propensity score matching, ferritin values of the patients with NAFLD were significantly higher than those without NAFLD group. Alanine aminotransferase and ferritin were strongly associated with NAFLD in multivariate stepwise logistic regression analysis. The medium and high levels of ferritin increased risk of NAFLD, and the adjusted ORs were 3.298 (95% CI:1.326‐8.204), 7.322 (95% CI:2.725‐19.574) across the ferritin concentration tertiles after adjustment for confounders. Ferritin was shown to be the best predictor for NAFLD with sensitivity and specificity of 60.0% and 77.9%, respectively, area under the curve was 0.733. Conclusion The results show that serum ferritin can usefully be considered as a predictor of NAFLD in children with obesity.

worldwide, due to the westernized diet and lifestyle. Therefore, the early diagnosis and proper treatment at appropriate time of NAFLD in children is becoming more and more important.
Liver biopsy is used for the diagnosis of patients with NAFLD as the gold standard method; however, it may be affected by other liver diseases during the image evaluation and limited by invasiveness. 6 So, it is not feasible to recommend it in all patients with NAFLD, especially for children. There are several scoring systems have been investigated to diagnose NASH as well as fibrosis stage without histological data; however, the controversy associated with accuracy is ongoing. 7 As a non-invasive method, Fibroscan still cannot discriminate the entity of fibrosis because of its gray area and the tackle for magnetic resonance imaging is not widely used due to its high price and the exact cut-off value has not been established. 8,9 Thus, researches are making efforts to seek cheap and non-invasive biological markers for diagnosis and prognosis of NAFLD.
Recently, some investigations stated that serum ferritin are correlated with dyslipidemia, 10 hypertension, 11 central obesity, 12 type 2 diabetes mellitus, 13 liver disease, 14 and stroke 15 in adults and also has a role in the development of metabolic syndrome in children and adolescents. 16 It has reported that the level of serum ferritin can be an independent predictor of fibrosis in NAFLD patients based on its related to hepatic iron storage and hepatic inflammation 17 and associated with insulin resistance and hepatocyte damage in NAFLD, 18 while other researches have not made this findings, 19 and there are few investigations assess the association between serum ferritin and NAFLD in children with obesity.
In this present study, we sought to evaluate the association between serum ferritin and NAFLD in children with obesity. It will be very helpful to provide medical treatment at an appropriate time if serum ferritin could be as a simple effective and less-invasive biological marker, which reflect hepatic inflammatory change in obesity.

| Study population
Data from the Zhejiang university children's hospital from September 2018 to September 2019 were analyzed in the present study, a total of 347 children with obesity that met inclusion criteria for our final investigation, and participants were divided into two groups: with or without NAFLD. Obesity was defined as BMI for age and sex ≥95th percentile. 20 All subjects has no iron deficiency or received iron food supplements or any drugs which could influence serum ferritin.
Nonalcoholic fatty liver disease was defined based on hepatic ultrasound which was conducted to assess the presence and extent of hepatic steatosis, according to the following guideline: (a) a diffuse hyperechoic texture (bright liver); (b) increased liver echo texture compared to kidney; (c) deep beam attenuation; (d) vascular blurring (absence of normal echogenic walls of the portal veins and hepatic veins).
The exclusion criteria included the following: (a) liver diseases such as alcohol consumption, drug induced hepatitis, viral hepatitis, autoimmune hepatitis, schistosomiasis, liver cirrhosis; (b) chronic diseases such as heart disease, kidney disease, epilepsy, tuberculosis, cancer, and other diseases which could affect liver function; (c) missing data on laboratory tests, ultrasound results, or physical examination.
All procedures performed in our study accordance with the guidelines of the Ethical Committee in our hospital and with the 1964 Helsinki Declaration and its later amendments.

| Measurements
Height and waist circumference were measured to the nearest 0.1 cm, body weight nearest 0.1 kg. Body mass index (BMI) was calculated as weight(kg)/square of height(m 2 ). A mercury sphygmomanometer was used to measure systolic blood pressure (SBP, mm Hg) and diastolic blood pressure (DBP, mm Hg) in the right upper arm after patients having rested for at least 10-minute rest period. Blood pressure was assessed three times, at 2-min intervals, and the mean of these three measurements was calculated.

| Laboratory tests
Blood samples were collected after fasting overnight for at least 8 hours. All obtained blood samples were processed, refrigerated, transported to our clinical laboratory, and analyzed within 8 hours.
All clinical analyses were performed by our hospital, a laboratory Insulin resistance (HOMA-IR) was determined by the homeostasis model (16) and calculated using the following equation:

| Statistical analyses
Normally distributed variables are presented as mean ± standard deviation (SD), while non-normally distributed variables are presented as medians and interquartile range (25th-75th percentiles). Normally Propensity analysis was carried out using logistic regression in order to create a propensity score for obese with NAFLD and without NAFLD. Propensity score matching analysis using a 1:1 greedy method without replacement. The caliper used in this study was 0.01, and the variables entered into the propensity model were gender, age, SBP, DBP, and BMI.
Analysis of stepwise logistic regression was conducted to determine the significant predictors after controlling for all variables.
Statistical analyses were performed using SPSS 22.0 software, and a two-sided P-value < .05 was considered to be statistically significant.

| Clinical characteristics of the study population
From 221 children with obesity were classified as NAFLD (mean age: 11.52 ± 2.39 years, 164 males and 57 females), 126 obese without NAFLD (mean age: 10.42 ± 2.62 years, 84 males and 42 females) based on ultrasound findings were selected for propensity matching. Finally, 95 patients were matched for gender, age, SBP, DBP, and BMI. The baseline clinical characteristics and laboratory data of the study population were described in Table 1. As expected, we found that ferritin, ALT, AST, GGT, ALP, total protein, LDL-cholesterol, and insulin levels were significantly higher in obese with NAFLD than obese without NAFLD. No significant differences were observed for other variables between the two groups.

| Logistic regression analysis for predictors of NAFLD
ALL the 29 variables were included in multivariate stepwise logistic regression analysis to determine independent predictors. As presented in Table 2 were found to be an independent markers for the prediction of NAFLD.

| Associations between serum ferritin and NAFLD
All subjects were divided into three gradients based on ferritin ter-  Table 3.
The receiver operating characteristic curve analysis revealed the relationship between serum ferritin and NAFLD, and the area under the curve was 0.733 (95% CI: 0.663-0.803; P < .001). Using the best cut-off value of ferritin predicted, the occurrence of NAFLD was ≥ 58.55 μg/L with a sensitivity of 60.0% and specificity of 77.9% ( Figure 1).

| D ISCUSS I ON
This study focused on the association between ferritin and NAFLD in children with obesity by using a propensity score method. We observed that NAFLD patients had higher ferritin levels than control subjects, and elevated serum ferritin level was associated with a significantly increased risk of NAFLD.
Hsiao TJ 21 reported that iron metabolism disorders can be re- Ferritin levels can be increased secondary due to viral hepatitis, steatohepatitis, chronic consumption of alcohol as well as obesity. 17 Hyperferritinemia observed in obesity-related chronic inflammation such as metabolic syndrome, diabetes mellitus, and NAFLD, 17 and obesity led to hyperferritinemia are considered as due to inflammation responsible for high levels of ferritin. 32 Several studies demonstrated that higher levels of ferritin are considered as marker of hepatic damage, because of inflammatory cytokine activation. 33,34 It has been reported that hepatic iron accumulation can stimulate production of inflammatory cytokines which is significantly related to hepatic fibrosis in NAFLD patients. 35 In conclusion, we concluded that elevated serum ferritin can usefully be considered as a predictor of NAFLD in children with obesity.

CO N FLI C T O F I NTE R E S T S
The authors have no conflict of interests for this paper.

I N FO R M E D CO N S E NT
Informed consent was obtained from the parents or legal guardians of all patients. F I G U R E 1 Receiver operating characteristic curve of serum ferritin to predict the occurrence of NAFLD in children with obesity