Sex Hormone Relations to Histologic Severity of Pediatric Nonalcoholic Fatty Liver Disease

Abstract

Context

Sex hormones have been linked with presence and severity of nonalcoholic fatty liver disease (NAFLD) in adults, but it is unknown if they affect severity of pediatric NAFLD.

Objective

To examine associations of circulating SHBG, estrogens, and androgens with key histologic features of pediatric, biopsy-confirmed NAFLD.

Design

Baseline assessment of longitudinal cohorts and randomized clinical trials.

Setting

Nonalcoholic Steatohepatitis Clinical Research Network.

Patients

Children and adolescents ≤18 years with liver biopsy-confirmed NAFLD in the United States.

Main Outcome Measures

We assayed SHBG, estrone, estradiol, dehydroepiandrosterone (DHEAS), androstenedione, and testosterone in relation to grade/stage of steatosis, portal inflammation, hepatic ballooning, fibrosis, and nonalcoholic steatohepatitis (NASH) severity using linear regression.

Results

Mean age of 573 children at the time of biopsy was 13.1 years (SD 2.8). Lower SHBG was inversely associated with steatosis severity in boys and girls (P = 0.001), and with portal inflammation in girls only (P for sex interaction <0.001). Higher testosterone was related to improved features of steatosis and fibrosis (P for sex interaction = 0.003 and 0.01, respectively) in boys, but detrimental in girls. In boys and girls, higher estrone, estradiol, and testosterone were associated with lower portal inflammation grade; higher estradiol was positively associated with hepatic ballooning severity; DHEAS was inversely associated with hepatic ballooning and NASH severity (all P < 0.05). Androstenedione was not associated with NAFLD features.

Conclusions

Largely consistent with findings in adults, sex hormones are associated with distinct histologic features of NAFLD in children and adolescents. These hormone levels relate to differences with gender and pubertal change.

Nonalcoholic fatty liver disease (NAFLD) is defined as the accumulation of macrovesicular fat in hepatocytes in the absence of excessive alcohol consumption or other known causes of fatty liver (1). It encompasses a wide spectrum of liver abnormalities ranging from isolated hepatocellular steatosis to steatohepatitis (nonalcoholic steatohepatitis [NASH]), that is characterized by coexistent lobular hepatic inflammation and hepatocyte injury, with variable degrees of fibrosis. This can result in progression to cirrhosis or hepatocellular carcinoma (2). Among U.S. children, it has been estimated that pediatric NAFLD prevalence has doubled over the past 2 decades and is now estimated to be nearing 10% (3). NAFLD is now recognized as the leading cause of chronic liver disease in both children and adolescents (4). Considering the paucity of available treatments for NAFLD and its potential for devastating long-term health consequences (5), understanding its etiopathogenesis is essential for mitigation.

Differences in NAFLD prevalence by sex (6), timing of puberty onset (7), Tanner staging (8), menopausal status, and synthetic hormone use in women (9) strongly suggests the involvement of certain sex hormones as mediators in NAFLD etiopathogenesis. Only adult studies have examined and found inverse associations of SHBG, estrogens, and androgens with NAFLD features, suggesting an etiologic role of sex hormones (10-15). In contrast, there have been no studies on such hormones in pediatric populations, in which NAFLD could be etiologically dissimilar from adult NAFLD (16). As well, there are no studies in pediatrics on levels of SHBG or sex steroid hormones relating to histologic severity or specific features of NAFLD, such as hepatocellular injury, inflammation, steatosis, and fibrosis.

The current study aims to examine associations of circulating levels of SHBG, estrogens (estrone and estradiol), and androgens (dehydroepiandrosterone [DHEAS], androstenedione, and testosterone) with graded or staged features of histologically confirmed NAFLD; including steatosis, portal inflammation, hepatocellular ballooning, NASH, and fibrosis. We hypothesized that SHBG and sex hormones, which are highly influenced by sex and puberty, are variably expressed in relation to histologic features of pediatric NAFLD.

Materials and Methods

Study design and population

We performed a cross-sectional analysis of data obtained from the National Institute of Diabetes and Digestive and Kidney Diseases Nonalcoholic Steatohepatitis Clinical Research Network (NASH-CRN) database. This includes baseline data from 12 participating pediatric clinical centers across the United States. Participants in this study were selected from children enrolled in the following NASH-CRN studies: the longitudinal cohort studies of the Pediatric Database and Pediatric Database 2 (NCT01061684) and the randomized clinical trial of Treatment of Nonalcoholic Fatty Liver Disease in Children (NCT00063635). The NAFLD Database began enrollment in September 2004, Treatment of Nonalcoholic Fatty Liver Disease in Children in August 2005, and Database 2 in October 2009. The NASH CRN and the institutional review board at each center approved these studies (17). Written consent for all participants was obtained from a parent or legal guardian, and assent from all children 7 years or older before participation. For this analysis, we included children younger than 18 years with liver biopsy-confirmed NAFLD, based on liver histology with ≥5% of hepatocytes containing macrovesicular fat, exclusion of other causes of chronic liver disease by clinical history, exclusion of those exposed to potentially hepatotoxic medications, pertinent laboratory determinations, and central histology scoring and grade consensus. We excluded children who ever consumed more than an equivalent of 2 4-oz glasses of wine per day per week, as well as girls who ever took oral contraceptives.

Measurement of NAFLD features

Percutaneous liver biopsies were obtained as standard of care by sampling the right liver lobe using a 15-gauge needle. Biopsies were routinely ≥15 mm in length. Biopsy specimens were stained with hematoxylin-eosin and Masson trichrome stains and centrally reviewed by the Pathology Committee of the NASH CRN according to the NASH-CRN scoring system, which has been validated in the pediatric population (18). The Pathology Committee was unaware of all demographic and clinical data.

Individually assessed histologic features included grades for steatosis, portal inflammation, hepatic ballooning, and stages of fibrosis (18). Biopsies were scored for the degree of steatosis present in hepatocytes as follows: grade 0 was macrovesicular fat in <5% of hepatocytes; grade 1 was 5% to 33%; grade 2 was 34% to 66%; and grade 3 was >66%. Portal inflammation was scored as none (grade 0), mild (grade 1), and more than mild (grade 2). Hepatic ballooning was scored as none (grade 0), few (grade 1), and many (grade 2). Fibrosis was scored as none (stage 0), mild zone 3 perisinusoidal (stage 1a), moderate zone 3 perisinusoidal (stage 1b), portal/periportal fibrosis only (stage 1c), zone 3 and periportal fibrosis (stage 2), bridging fibrosis (stage 3) and cirrhosis (stage 4). For the current analysis, some grades/stages of the histologic features were combined as follows to not have too few children in each category and assure statistical power: steatosis (grades 0-1 vs grade 2 vs grade 3); portal inflammation (grades 0-1 vs grade 2); hepatic ballooning (grade 0 vs grade 1 vs grade 2), and fibrosis (stage 0 vs stage 1 vs stage 2 vs stages 3-4).

Liver biopsies also were assigned to diagnostic categories based on the aggregate presence and degree of certain features of NASH. The diagnosis is independent of a particular NASH feature and is based on the pattern of injury as well as presence of specific histologic features. The minimum criteria to diagnose NASH are ≥5% steatosis, lobular inflammation, and hepatocyte injury as manifested by ballooning degeneration. The assignment of “NAFLD without NASH,” “borderline zone 3 NASH,” “borderline zone 1 NASH,” or “definite NASH” was made as a consensus agreement of the NASH CRN Pathology Committee at the time of central review of cases according to specific established criteria (19). Most pediatric fatty liver disease is categorized as borderline zone 1 NASH as a reflection of the particular zonal location of steatosis seen in many pediatric cases as well as the lack of ballooned hepatocytes in most pediatric cases. Borderline zone 3 NASH in most cases reflects absence or equivocal presence of a particular feature, for example, appropriate pattern fibrosis, steatosis, and inflammation but absence of ballooned hepatocytes.

Measurement of SHBG and sex hormones

All participants were fasted overnight for 12 hours before phlebotomy via venipuncture. The range in time between liver biopsy and blood sampling was from 1 day to 6 months with a mean of 3 months. SHBG and serum sex hormones, including estrone, estradiol, DHEAS, androstenedione, and testosterone were assayed in duplicate from stored frozen serum (–80 °C) using ELISA. ELISA kits were purchased from R&D Systems (Minneapolis, MN). These ELISA assays were performed in the Irving Clinical and Translational Research Institute at Columbia Vagelos College of Physicians and Surgeons. All samples were diluted to read within the standard curve range, which was designed and validated to be the linear binding range of the assay. The acceptable ranges for each immunoassay quality control set are within 2 SD from the control’s mean value. We considered the limit of quantification for each assay to be the lowest prepared standard. Two participants had DHEAS <0.1 μg/dL (the detectable limit of the ELISA) and were thus assigned a value of 0.1 μg/dL; the 36 with estrone <15 pg/mL were assigned 15 pg/mL; the 6 with androstenedione <0.1 ng/mL were assigned 0.1 ng/mL; and the 5 with SHBG < 6.8 pg/mL were assigned 6.8 pg/mL.

Measurement of covariates

A structured face-to-face interview was used to obtain demographic data on study participants. Weight and height measurements were performed in duplicate while the participant was wearing light clothing without shoes and was measured to the nearest 0.1 kg and 0.1 cm, respectively. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. BMI percentile was determined according to the age- and sex-specific data from the Centers for Disease Control and Prevention (20). Based on these percentiles, we calculated age- and-sex specific BMI z-scores to compare BMI adjusted for age and sex.

Pubertal development was classified into 3 stages: namely prepuberty, puberty, and mature on the basis of Tanner stages. For girls, Tanner stages for genitalia and breasts were used. Prepuberty was defined as both breast and genital Tanner stages being stage 1, whereas mature was defined as both Tanner stages being 5. The remaining cases were classified into puberty category. When only 1 component of Tanner stages (genitalia or breasts) was available, the case was classified on the basis of that component. For boys, Tanner stages for pubic hair and genitalia were used to classify the stages of pubertal development, also with a range of Tanner 1 through 5.

Statistical analyses

We tabulated demographic and clinical characteristics for the overall sample and separately for boys and girls. We tabulated continuous variables as means and SD if normally distributed, median and interquartile range if non-normally distributed, and tabulated categorical variables as counts and percentages. We tested for differences by using t tests for normally distributed variables, Kruskal-Wallis rank sum tests for non-normally distributed variables, and χ ² tests for categorical variables.

SHBG and sex hormones were non-normally distributed and thus were normalized by log-transformation to not avoid the normality assumption required by linear regression. We treated log-transformed SHBG and sex hormones as dependent variables and histologic features as independent variables. We used linear least-squares regression to assess whether the mean of each log-transformed hormone differs across grades of each histologic feature, compared with the grade with least severity. Considering that the estimate coefficients (i.e., βs) cannot be easily interpreted (i.e., the difference in the mean of log-transformed hormone after adjustment), we further performed a back transformation (i.e., exponentiation) to estimate and report the adjusted geometric mean and 95% confidence interval (CI) of each hormone across levels of each histologic feature. The geometric means reported are based on the mean values for each covariate included in the final multivariable model. Association analyses were performed first for boys and girls separately and then combined.

To address potential confounding bias, we began with an unadjusted model and then added covariates that represented potential confounders, as determined by previous literature linking them to SHBG, sex hormone levels, and NAFLD. The final multivariable adjusted linear models included age (continuous), sex (boys, girls), race/ethnicity (Hispanic, non-Hispanic white, other), BMI z-score (continuous), and Tanner stage (prepuberty, puberty, mature) as confounders. To test the overall trend of SHBG and hormones across levels of histologic features, we further treated each histologic NAFLD feature, except NASH, as a continuous (instead of categorical) variable in the multivariable adjusted model (21). For NASH, considering that it is a nominal category, we used ANOVA to test whether the adjusted geometric means of SHBG or hormones are different across groups of NASH, instead of testing for linear trend as for other histologic features.

Then, we tested effect measure modification by sex on the multiplicative scale by including an interaction term between sex and each NAFLD feature in the multivariable adjusted model. P values for interaction by sex were obtained using likelihood ratio test comparing models with and without the interaction term. If there was evidence of interaction by sex, we presented the corresponding results separately for boys and girls.

Results

Participants’ characteristics, overall and separately for boys and girls, are presented in Table 1. The mean age of all children in our study sample (n = 573) was 13.1 years (SD: 2.8). Overall, there were 71 (12.4%) children younger than age 10 years, 341 (59.5%) between 10 and 14 years old, and 161 (28.1%) older than 14 years of age. The numbers and percentages among boys were 47 (11.4%), 250 (60.5%), and 116 (28.1%); and among girls were 24 (15.0%), 91 (56.9%), and 45 (28.1%). The mean BMI z-score was 2.3 (SD: 0.5). There were 366 (63.9%) Hispanic children in total, and 265 were boys; 101 were girls. In total, 180 (33.1%) children were prepubertal, 273 (50.2%) were pubertal, and 91 (16.7%) were mature; the numbers and percentages in boys were 148 (37.2%), 199 (50.0%), and 51 (12.8%); in girls, they were 32 (21.9%), 74 (50.7%), and 40 (27.4%). Characteristics were largely similar by sex, although girls had slightly lower BMI z-score and were more likely to be further along in pubertal transition than their male counterparts. Further, girls had lower DHEAS and testosterone levels, whereas they had higher androstenedione levels than boys.

Table 1.

Demographic Characteristics, SHBG, and Sex Hormone Distributions in All Children from the NASH-CRN by Sex (N = 573)

Characteristic	Overall	Boys	Girls	P
	N = 573	n = 413	n = 160
Age (y), mean (SD)	13.1 (2.8)	13.2 (2.8)	13.0 (2.9)	0.61
Race/ethnicity
White (not Hispanic), n (%)	169 (29.5)	126 (30.5)	43 (26.9)	0.12
Other (not Hispanic), n (%)	38 (6.6)	22 (5.3)	16 (10.0)
Hispanic, n (%)	366 (63.9)	265 (64.2)	101 (63.1)
BMI (kg/m2)a, mean (SD)	32.4 (6.5)	32.5 (6.5)	32.3 (6.5)	0.83
BMI Z-scorea, mean (SD)	2.3 (0.5)	2.3 (0.5)	2.2 (0.4)	0.002
BMI percentilea, mean (SD)	97.9 (4.4)	98.0 (4.8)	97.6 (3.0)	0.40
Waist circumference (cm)a, mean (SD)	104.0 (16.1)	105.0 (16.4)	101.6 (15.2)	0.03
Tanner stagea
1, pre-puberty, n (%)	180 (33.1)	148 (37.2)	32 (21.9)	<0.001
2-4, puberty, n (%)	273 (50.2)	199 (50.0)	74 (50.7)
5, mature, n (%)	91 (16.7)	51 (12.8)	40 (27.4)
SHBGa (nmol/L), median (IQR)	24.6 (17.2–33.0)	24.8 (18.0–32.1)	23.4 (16.0–38.7)	0.93
Estronea (pg/mL), median (IQR)	49.7 (33.7–77.9)	48.7 (32.9–79.1)	51.7 (35.3–74.2)	0.91
Estradiola (pg/mL), median (IQR)	111.1 (64.1–154.4)	113.0 (68.0–155.9)	99.4 (60.0–150.5)	0.20
DHEASa (μg/dL), median (IQR)	1.5 (0.8–2.9)	1.6 (0.9–3.0)	1.3 (0.7–2.5)	0.03
Androstenedionea (ng/mL), median (IQR)	0.7 (0.4–1.2)	0.6 (0.4–1.0)	1.0 (0.6–1.7)	<0.001
Testosteronea (ng/mL), median (IQR)	3.4 (1.7–7.2)	4.0 (2.0–8.1)	2.4 (1.3–3.9)	<0.001

Abbreviations: BMI, body mass index; IQR, interquartile range; DHEAS, dehydroepiandrosterone; NASH-CRN, Nonalcoholic Steatohepatitis Clinical Research Network.

^a Missing values: BMI = 1, waist circumference = 8, Tanner stage = 29, SHBG = 6, estrone = 2, estradiol = 24, DHEAS = 2, androstenedione = 1, testosterone = 23.

There was little to no evidence of effect modification by sex on the majority of tested associations (17). In boys and girls combined (Table 2), we did not find that histologic features, other than NASH, were associated with androstenedione. However, progression of certain features among them were generally associated with lower levels of other hormones after multivariate adjustment. Specifically, children with higher SHBG were less likely to have severe steatosis (P = 0.001); children with higher estrone were less likely to have severe portal inflammation (P = 0.001) and fibrosis (P = 0.04); children with higher estradiol were less likely to have severe portal inflammation (P = 0.03), but were more likely to have severe hepatic ballooning (P = 0.02); children with higher DHEAS were less likely to have severe hepatic ballooning (P = 0.04) and NASH (P for trend = 0.01); and children with higher testosterone were less likely to have severe portal inflammation (P for trend = 0.04). As for NASH, we found that there were statistically significant differences in geometric means of all hormones, but estradiol, across NASH categories. In general, children without NASH tended to have higher sex hormones (Table 2).

Table 2.

Multivariable Adjusted Geometric Means (95% CI) of SHBG and Sex Hormones by Features of NAFLD in All Subjects

	SHBG (nmol/L)		Estrone (pg/mL)		Estradiol (pg/mL)		DHEAS (μg/dL)		Androstenedione (ng/mL)		Testosterone (ng/mL)
	N	Mean (95% CI)	N	Mean (95% CI)	N	Mean (95% CI)	N	Mean (95% CI)	N	Mean (95% CI)	NC	Mean (95% CI)
Steatosis
Grades 0-1	180	26.5 (25.7-27.3)	179	51.8 (50.6–53.1)	168	98.8 (97.2-100.0)	180	2.7 (2.5-2.9)	180	0.82 (0.77-0.88)	NAa
Grade 2	155	24.6 (23.8-25.5)	157	53.8 (52.3-55.3)	153	92.4 (91.0-93.9)	156	2.5 (2.2-2.7)	157	0.71 (0.66-0.77)
Grade 3	203	22.1 (21.4-22.7)	205	48.7 (47.5-49.9)	198	102.0 (101.3-103.5)	205	2.3 (2.1-2.5)	205	0.61 (0.57-0.65)
P value		0.001		0.89		0.37		0.53		0.10
Portal inflammation
Grades 0-1	N/-Aa		451	53.8 (53.0-54.6)	433	101.5 (100.4-102.9)	451	2.6 (2.4-2.7)	452	0.74 (0.71-0.78)	433	3.5 (3.4-3.7)
Grade 2			90	39.9 (38.7-41.0)	86	84.5 (83.2-85.9)	90	2.0 (1.9-2.3)	90	0.53 (0.49-0.59)	87	2.2 (2.0-2.4)
P value				0.001		0.03		0.97		0.38		0.04
Hepatic ballooning
Grade 0	316	24.8 (24.2-25.4)	318	50.9 (49.9-51.9)	304	94.4 (93.4-95.5)	319	2.6 (2.5-2.8)	319	0.70 (0.66-0.74)	308	3.2 (3.0-3.4)
Grade 1	154	22.5 (21.8-23.3)	155	52.2 (50.8-53.5)	151	98.9 (97.3-100.8)	155	2.4 (2.2-2.6)	155	0.72 (0.67-0.78)	148	3.2 (2.9-3.5)
Grade 2	68	25.3 (24.1-26.5)	68	50.4 (48.2-52.6)	64	115.2 (112.8-117.4)	67	2.0 (1.8-2.3)	68	0.67 (0.60-0.76)	64	3.7 (3.3-4.3)
P value		0.48		0.84		0.02		0.04		0.26		0.44
Fibrosis
Stage 0	182	23.7 (23.0-24.5)	181	58.1 (56.8-59.5)	177	99.0 (97.3-101.3)	182	3.0 (2.7-3.2)	182	0.86 (0.80-0.92)	NAa
Stage 1	206	24.0 (23.3-24.7)	210	50.1 (49.0-51.2)	204	95.1 (93.9-96.4)	210	2.5 (2.3-2.7)	210	0.63 (0.59-0.68)
Stage 2	76	24.4 (23.5-25.3)	76	44.8 (43.3-46.4)	71	106.7 (104.2-108.7)	76	1.6 (1.5-1.8)	76	0.67 (0.61-0.75)
Stages 3-4	72	26.2 (24.9-27.6)	72	46.2 (44.5-48.0)	65	95.8 (93.9-97.7)	71	2.2 (1.9-2.5)	72	0.59 (0.53-0.66)
P value		0.35		0.04		0.53		0.23		0.59
NASH
None	155	23.9 (23.0-24.8)	154	56.6 (55.1-58.1)	149	98.7 (96.9-101.3)	155	3.8 (3.4-4.1)	155	0.87 (0.80-0.93)	149	3.6 (3.3-3.9)
Zone 3	87	21.6 (20.8-22.5)	88	49.4 (47.9-51.1)	84	106.8 (103.2-108.5)	88	2.4 (1.2-2.7)	88	0.78 (0.71-0.86)	87	3.3 (2.9-3.8)
Zone 1	149	27.1 (26.4-27.8)	151	45.8 (44.7-46.9)	143	91.2 (90.0-92.4)	151	1.7 (1.6-1.9)	151	0.49 (0.46-0.53)	141	2.6 (2.4-2.8)
Definite	127	22.3 (21.6-23.0)	128	53.5 (52.0-55.0)	123	103.4 (101.6-105.8)	127	2.5 (2.2-2.7)	128	0.77 (0.71-0.83)	123	3.8 (3.5-4.2)
P value		0.01		0.02		0.18		0.003		<0.001		0.003

Multivariable model adjusted for age, sex, race/ethnicity, Tanner stage, and BMI z-score. The adjusted geometric means reported are based on the mean values for each covariate included in the multivariable model. Bold indicates statistically significant findings (P < 0.05) compared with the lowest stage/grade for each histologic feature of NAFLD. P value was obtained by testing for overall trend across levels of histologic NAFLD features other than NASH, or by testing any difference across groups of NASH.

Abbreviations: CI, confidence interval; DHEAS, dehydroepiandrosterone; NA, not available; NASH, nonalcoholic steatohepatitis; NAFLD, nonalcoholic fatty liver disease.

^a Pooled estimates not applicable because evidence of sex interaction. See Table 3.

Sex modified some tested associations (17). Multivariable adjusted geometric means of SHBG and sex hormones separately for boys and girls are shown in Fig. 1 (22, 23). There was evidence of effect modification by sex on the association between SHBG and portal inflammation (P for interaction <0.001, Table 3), such that higher SHBG was nonsignificantly associated with higher grades of portal inflammation in boys (22) but significantly associated with lower grades in girls (P = 0.03) (23). Sex-modified association between testosterone and steatosis (P for interaction = 0.003). In boys, testosterone was nonsignificantly negatively associated with steatosis severity, whereas in girls it was significantly positively associated with steatosis severity (P = 0.03). Sex also modified the effect of testosterone on fibrosis (P for interaction = 0.01). Higher testosterone was significantly associated with lower stages of fibrosis in boys (P = 0.02) but not in girls. Additionally, there was suggestive evidence that sex modified the SHBG-hepatic ballooning and DHEAS-steatosis relationships (Fig. 1), though the P values for interaction (0.08, 0.09, respectively) (17) did not reach the statistically significant level.

Figure 1.

Multivariable adjusted geometric means and 95% confidence intervals of SHBG and sex hormones by features of NAFLD, overall and separately in boys and girls from the NASH-CRN. *Statistically significant P value across levels of the NAFLD. feature (P < 0.05) in boys and girls combined. DHEAS, dehydroepiandrosterone; NAFLD: nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis.

Table 3.

Multivariable Adjusted Geometric Means (95% CI) of SHBG and Testosterone by Features of Steatosis, Fibrosis, and Portal Inflammation, Separately for Boys and Girls

	Boys		Girls
	SHBG (nmol/L)
	N	Mean (95% CI)	N	Mean (95% CI)
Portal Inflammation (P for interaction by sex <0.001)
Grades 0-1	323	23.4 (22.7-24.0)	126	26.3 (25.4-27.3)
Grade 2	69	27.3 (26.0-28.6)	20	17.0 (15.7-18.4)
P value		0.38		0.03
	Testosterone (ng/mL)
Steatosis  (P for interaction by sex = 0.003)
Grades 0-1	113	4.6 (4.1-5.2)	63	1.9 (1.8-2.0)
Grade 2	110	3.8 (3.4-4.2)	40	2.5 (2.3-2.7)
Grade 3	157	3.2 (2.9-3.5)	37	2.5 (2.2-2.8)
P value		0.17		0.03
Fibrosis  (P for interaction by sex = 0.01)
Stage 0	116	5.3 (4.8-5.9)	59	2.1 (2.0-2.2)
Stage 1	157	3.5 (3.2-3.8)	45	2.0 (1.9-2.1)
Stage 2	52	4.2 (3.6-4.8)	22	2.8 (2.6-3.0)
Stages 3-4	54	2.1 (1.8-2.3)	13	3.0 (2.7-3.3)
P value		0.02		0.09

Multivariable model adjusted for age, race/ethnicity, Tanner stage, and BMI z-score. The adjusted geometric means reported are based on the mean values for each covariate included in the multivariable model. Bold indicates statistically significant findings (P < 0.05) compared with the lowest stage/grade for each histologic feature of NAFLD. P value was obtained by testing for overall trend across levels of steatosis, fibrosis, and portal inflammation.

Abbreviations: CI, confidence interval; NAFLD, nonalcoholic fatty liver disease.

Discussion

In this multiethnic sample of children with liver biopsy-confirmed NAFLD, a population at higher risk for poor long-term outcomes, we observed several significant associations of SHBG and sex hormones with histologic features of NAFLD, independent of stages of pubertal development and adiposity. Certain associations, notably those for SHBG and testosterone, were sex-specific.

Our finding that SHBG was inversely associated with the histologic severity of steatosis among pediatric NAFLD population is consistent with prior findings in adults. Jaruvongvanich et al. (15) conducted a systematic review and meta-analysis on 12 observational studies that assessed SHBG levels in adult participants with and without NAFLD, comprising 7841 adult men and 4930 adult women, and concluded that in both sexes, SHBG was lower in NAFLD patients than normative pairs. Further, our study adds to the literature base by showing that in girls with NAFLD, SHBG was inversely associated with the degree of portal inflammation. The implications of this finding are considerable as portal inflammation is considered a marker of advanced fatty liver disease (24). However, because there was no association detected between SHBG and portal inflammation in boys, future research is needed to verify the observation that SHBG is associated with the severity of portal inflammation in pediatric NAFLD in a sex-specific manner.

Several epidemiologic lines of evidence have suggested a protective effect of estrogens on the severity of fibrosis in adults with NAFLD (11, 25, 26). Additionally, in 40 children with NAFLD, Elbel et al. (27) reported that nuclear estrogen receptor was increased in livers with more lobular inflammation, hepatic ballooning, fibrosis, and NASH diagnosis. Considering that the up-regulated receptors may be partially compensating for reduced levels of estrogens, our findings that estrone and estradiol levels decreased in participants with more severe portal inflammation and fibrosis are consistent with this prior study. However, it is noteworthy that our study detected a positive association between estradiol and grades of hepatic ballooning, which is contradictory to previously mentioned studies. Given that Lazo et al. (14) also reported that higher estradiol was associated with higher levels of liver fat measured by imaging in both adult men and women, the inconsistent effects of estradiol on different histologic features of NAFLD warrants further assessment in pediatric populations.

As for androgens, the inverse association between DHEAS and severity of NAFLD, as observed in our study, has been widely reported in adult populations (28-31). In our study, boys with higher testosterone had lower grades of steatosis and fibrosis. Conversely, girls with higher testosterone had higher grades of steatosis and fibrosis. Our observation of the opposing effects of testosterone with features of NAFLD in boys and girls is consistent with findings in adults. In the meta-analysis by Jaruvongvanich et al. (15), pooled analysis on 7 observational studies comprising 4715 adult men and 1581 adult women, showed a decreased odds of NAFLD (odds ratio: 0.56, 95% CI: 0.39-0.80) in men with higher serum total testosterone, whereas an increased odds in women (odds ratio: 1.40, 95% CI: 1.11-1.77).

Our study is the first and the largest to provide evidence suggesting that circulating SHBG and sex hormones are associated with histologic severity of NAFLD in a cross-sectional, multiethnic pediatric population in the United States. The ethnic diversity in our cohort improves the external validity of the conclusion. Moreover, liver biopsy followed by histology examination, as used in our study, remains the standard to diagnose NAFLD, to exclude other causes of liver diseases, to differentiate between simple steatosis and advanced features (e.g., NASH), and to perform disease staging (32). Along with the consensus and blind review of liver biopsies by expert NASH-CRN pathologists, we were able to eliminate misclassifications of histologic scoring and feature identification.

There are several limitations of our study. First, because liver biopsies were performed once, they represent a single-time snapshot of the liver histology. Although steatohepatitis is a diffuse disease, the degree of microheterogeneity, especially because it affects rare histologic findings such as ballooning degeneration of focal fibrosis, is not known (33). Therefore, estimated associations may be biased toward the null. Second, a limitation of ELISA is that their accuracy can be suboptimal when samples read at the low end of the standard curve, where the optical density is highest. This may have led to some measurement error. Third, the smaller sample size for girls may have limited power to detect statistically significant associations. The small sample size also precluded us from conducting analyses stratified by both sex and other potential modifiers, such as race/ethnicity. Fourth, for postmenarche girls we did not have information on whether the blood sample was obtained during the follicular or luteal phase. We also did not have information on polycystic ovary syndrome, and therefore were not able to adjust for this variable in the analyses. Finally, the cross-sectional nature of this analysis precluded examination of longitudinal associations between changes in sex hormones with pubertal development in individuals and its relation to changes in NAFLD histology.

In conclusion, our findings that low circulating levels of SHBG, estrogens, and androgens were associated with histologically worse NAFLD in pediatric populations are novel. Additionally, we found that SHBG and testosterone had sex-specific associations: SHBG was lower in boys with higher grade of portal inflammation and testosterone was higher in boys with higher grades of steatosis; whereas in girls, testosterone was decreased as grades of fibrosis increased. These associations were independent of puberty stage and adiposity. Our study provides new insight into contributors to etiopathogenesis of key histologic features of advanced NAFLD in children and adolescents. It remains of interest to investigate the mechanisms underlying these findings because it could lead to novel mitigation strategies for prevention or treatment in this age group.

Glossary

Abbreviations

BMI

body mass index

CI

confidence interval

DHEAS

dehydroepiandrosterone

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

NASH-CRN

Nonalcoholic Steatohepatitis Clinical Research Network

Additional Information

Disclosure Summary: J.E.L. is an ad hoc consultant for NovoNordisk, Intercept Pharmaceuticals, and TARGET Pharmaceuticals. He received grant funding from Genfit.

Data Availability

The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author.