Novel plasma biomarkers associated with liver disease severity in adults with nonalcoholic fatty liver disease
Download the PDF here
"In our study, the only biomarker that remained associated with the histological diagnosis of NASH after adjustment for clinical factors was aPAI1. ......Higher aPAI1, tPAI1, and IGFII and lower adiponectin were associated with significant steatosis; and increased IL-8 levels and decreased IGFII were associated with hepatocyte ballooning. In addition, among seven biomarkers associated with the presence of significant fibrosis, increased IL-8 and sIL-2Rα and decreased IGFII demonstrated the strongest associations......Seven biomarkers were strongly associated with significant fibrosis in the multivariable model: IL-8, MCP-1, resistin, sIL-1R1, sIL-2Rα, TNFα, and IGFII.....We found that increased IL-8 was associated with the presence of hepatocyte ballooning. IL-8 is a chemokine that serves as a chemoattractant for neutrophils and contributes to acute liver inflammation. Previous studies have shown an association between NASH and increased IL-8 levels; however, ours is the first to document a significant association specifically between IL-8 levels and hepatocyte ballooning."
from Editorial: Finally, the importance of biomarkers associated with liver disease severity should be tempered by perspective. Given that fibrosis progression in NASH is slow (mean increase of 0.09 fibrosis stages per year) and not universal (41% of patients), non-liver-related death remains far more common than liver-specific death or hepatocellular carcinoma combined. Is there any value in diagnosing NASH or even advanced fibrosis with a biomarker in a patient who will die from cardiovascular disease? The focus and validation of serum biomarkers currently falls in the realm of diagnosis, whereas a shift in focus to the prediction of clinical outcomes is needed. At the end of the day, the most important issue is the ability to predict whether a patient will develop hepatic complications (or even cardiovascular events) in the future. As such, we look forward to further longitudinal studies before serum biomarkers become prime time.
Jnl of Hepatology Jan 2017
Despite the high prevalence of nonalcoholic fatty liver disease (NAFLD), therapeutic options and noninvasive markers of disease activity and severity remain limited. We investigated the association between plasma biomarkers and liver histology in order to identify markers of disease activity and severity in patients with biopsy-proven NAFLD. Thirty-two plasma biomarkers chosen a priori as possible discriminators of NAFLD were measured in participants enrolled in the Nonalcoholic Steatohepatitis (NASH) Clinical Research Network. Dichotomized histologic outcomes were evaluated using centrally read biopsies. Biomarkers with statistically significant associations with NAFLD histology were evaluated in multivariable models adjusted for clinical factors. Of 648 participants (74.4% white, 61.7% female, mean age 47.7 years), 58.0% had definite NASH, 55.5% had mild/no fibrosis (stage 0-1), and 44.4% had significant fibrosis (stage 2-4). Increased activated plasminogen activator inhibitor 1 had a strong association with definite NASH compared to not NASH or borderline NASH in multivariable analysis (odds ratio = 1.20, 95% confidence interval 1.08-1.34, P < 0.001). Biomarkers associated with significant fibrosis (versus mild/no fibrosis) in multivariable analysis included higher levels of interleukin-8, monocyte chemoattractant protein-1, resistin, soluble interleukin-1 receptor I, soluble interleukin-2 receptor alpha, and tumor necrosis factor alpha and lower levels of insulin-like growth factor 2. Conclusions: Specific plasma biomarkers are significantly associated with disease activity and severity of fibrosis in NAFLD and are potentially valuable tools for noninvasive stratification of patients with NAFLD and identification of targets for therapeutic intervention.
ALT alanine aminotransferase
aPAI1 activated PAI1
AST aspartate aminotransferase
BMI body mass index
CI confidence interval
FDR false discovery rate
HOMA-IR homeostasis model assessment of insulin resistance
IGFII insulin-like growth factor 2
MCP-1 monocyte chemoattractant protein-1
NAFLD nonalcoholic fatty liver disease
NASH nonalcoholic steatohepatitis
NASH CRN NASH Clinical Research Network
OR odds ratio
PAI1 plasminogen activator inhibitor 1
PIVENS Pioglitazone or Vitamin E for Nonalcoholic Steatohepatitis
sIL-1R1 soluble IL-1 receptor 1
TNFα tumor necrosis factor alpha
tPAI1 total PAI1
Approximately one-third of the US adult population is estimated to have nonalcoholic fatty liver disease (NAFLD). NAFLD is characterized by a spectrum of histologically defined stages from steatosis to steatohepatitis, advanced fibrosis, cirrhosis, and hepatocellular carcinoma. However, only a subset of patients progress through each stage. An improved understanding of risk factors associated with disease severity could help in risk stratification of patients with NAFLD.
The diagnosis and staging of NAFLD rely on histologic evaluation of a liver biopsy, an invasive test that is associated with some risk of complications and inconvenient as a repeat measure of disease severity. Moreover, a liver biopsy is susceptible to significant sampling variability. A growing appreciation of the relationship of mediators of inflammation with the histologic features of NAFLD coupled with the ability to detect many mediators of inflammation in the peripheral blood has resulted in a growing interest in novel biomarkers that could improve the utility of or even replace liver biopsy.
Plasma levels of several candidate biomarkers that correlate with steatosis, steatohepatitis, or fibrosis have been identified in small clinical studies of patients with NAFLD. Interleukin (IL)-6 levels were higher in patients with NAFLD than controls; however, the authors did not discriminate between those with steatosis and those with steatohepatitis. IL-8 levels were higher in patients with nonalcoholic steatohepatitis (NASH) than steatosis; however, other studies showed no difference.[6, 7] Elevated monocyte chemoattractant protein 1 (MCP-1) levels and increased MCP1 gene expression were associated with NAFLD and NASH.[5, 8] Reduced adiponectin levels were independently associated with NASH, and increasing tumor necrosis factor alpha (TNF-α) levels correlated with insulin resistance, steatohepatitis, and fibrosis.[9, 10] Multiple other biomarkers (resistin, leptin, retinol binding protein-4, hyaluronic acid, procollagen type 3 N-terminal peptide, tissue inhibitor of metalloproteinases 1, transforming growth factor beta 1) have also been associated with NAFLD severity and fibrosis, but all of these studies were limited by a small to modest sample size or examination of only a limited number of biomarkers.[11-14]
The purpose of this study was to evaluate a broad set of candidate biomarkers of inflammation, fibrosis, angiogenesis, and insulin and glucose metabolism in a large, well-characterized cohort of adults with NAFLD to determine their association with the histologic features of NAFLD.
CHARACTERISTICS OF THE STUDY POPULATION
Six hundred and forty-eight patients with NAFLD from the NASH CRN were included in the analysis. Participants had a mean age of 47.7 years and were predominantly female (61.7%), white (74.4%), and obese (mean BMI = 34.6 kg/m2). Diabetes (22.1%), hypertension (46.1%), and hyperlipidemia (55.7%) were common comorbidities. Of the 648 adult participants, 376 were classified as definite NASH (58.0%), 143 as not NASH (22.1%), and 129 as "borderline" NASH (19.9%).
BIOMARKERS ASSOCIATED WITH THE DIAGNOSIS OF DEFINITE NASH
In univariable logistic regression analysis, participants with definite NASH had significantly higher levels of total plasminogen activator inhibitor 1 (tPAI1), activated plasminogen activator inhibitor 1 (aPAI1), IL-8, and soluble IL-1 receptor 1 (sIL-1R1) than those with no or borderline NASH (Tables 2 and 3). In multivariable analysis adjusting for clinical factors, only increased aPAI1 was independently associated with definite NASH (odds ratio [OR] = 1.20, 95% confidence interval (CI) 1.08-1.34, P = 0.001) (Table 4). aPAI1 was associated with BMI, HOMA-IR, high-density lipoprotein cholesterol, and triglycerides (Table 5); and levels strongly correlated with tPAI1, ρ = 0.65 (Supporting Table S3). In sensitivity analysis excluding patients with borderline NASH from the comparison group, only aPAI1 remained associated with definite NASH after multivariable adjustment (Supporting Table S4).
BIOMARKERS ASSOCIATED WITH MODERATE TO SEVERE STEATOSIS, HEPATOCYTE BALLOONING, AND LOBULAR INFLAMMATION
Twenty-five percent of participants had >66% steatosis (grade 3), 34% had 34%-66% (grade 2), 37% had 5%-33% (grade 1), and 5% had <5% steatosis (grade 0). Participants with moderate to severe steatosis (>33%) had lower levels of adiponectin and sIL-2Rα as well as elevated tPAI1, aPAI1, and insulin-like growth factor 2 (IGFII) (Table 3; Supporting Table S1) compared to those with no or mild steatosis (≤33%) in univariable logistic regression. sIL-2Rα did not meet the FDR adjusted threshold for statistical significance. In a multivariable analysis of biomarkers associated with steatosis >33% compared to steatosis ≤33%, decreased adiponectin (OR = 0.91, 95% CI 0.83-0.99, P = 0.029) and increased aPAI1 (OR = 1.20, 95% CI 1.09-1.32, P < 0.001), tPAI1 (OR = 1.31, 95% CI 1.18-1.45, P < 0.001) and IGFII (OR = 1.16, 95% CI 1.05-1.28, P = 0.005) were associated with higher grades of steatosis (Table 4). Adiponectin was associated with female sex and increased age and high-density lipoprotein cholesterol and inversely associated with ALT, HOMA-IR, and triglycerides (Table 5). IGFII was associated with male sex and increased ALT and triglycerides and inversely associated with age and BMI. tPAI1 was associated with increased age, BMI, AST, HOMA-IR, and triglycerides. IGFII was inversely correlated with IL-8 and sIL-2Rα ρ = -0.20 and ρ = -0.22, respectively (Supporting Table S3). Adiponectin was not strongly correlated with any other biomarker, and tPAI1 only correlated strongly with aPAI1.
Thirty-two percent of participants had no evidence of hepatocyte ballooning (grade 0), 28% had a few ballooned hepatocytes (grade 1), and 39% displayed many ballooned hepatocytes (grade 2). In univariable logistic regression, participants with hepatocellular ballooning had increased aPAI1, tPAI1, IL-8, MCP-1, sIL-1R1, and TNFα and decreased IGFII compared to those without ballooning (Table 3; Supporting Table S1). MCP-1 and tPAI1 did not meet the FDR adjusted threshold for statistical significance. In a multivariable logistic regression analysis, the presence of hepatocyte ballooning was associated with increased IL-8 (OR = 1.27, 95% CI 1.04-1.56, P = 0.021) and decreased IGFII (OR = 0.86, 95% CI 0.77-0.96, P = 0.006) (Table 4). IL-8 was associated with increased age, AST, and HOMA-IR (Table 5).
Lobular inflammation was seen in fewer than two foci per ×20 field (grade 1) in 52% of participants, two to four foci (grade 2) in 37% of participants, and more than four foci (grade 3) in 11% of participants. Compared to participants with grade 0-1 lobular inflammation, those with grade 2-3 lobular inflammation had decreased MCP-1 (Table 3; Supporting Table S2) in univariable logistic regression. MCP-1 did not meet the FDR adjusted threshold for statistical significance. In multivariable logistic regression analysis for the presence of grade 2-3 lobular inflammation compared to grade 0-1 lobular inflammation, decreased MCP-1 (OR = 0.90, 95% CI 0.82-0.99, P = 0.025) was associated with increased odds of lobular inflammation (Table 4). In addition, MCP-1 was associated with increased age, BMI, and HOMA-IR. Because advancing fibrosis in NAFLD is associated with a shift from lobular inflammation to more portal-based inflammation and MCP-1 is associated with fibrosis, we performed sensitivity analysis on the association between MCP-1 and lobular inflammation excluding patients with cirrhosis; and the association was no longer statistically significant (OR = 0.92, 95% CI 0.83-1.01, P = 0.073).
BIOMARKERS ASSOCIATED WITH FIBROSIS
Twenty-six percent of participants had no fibrosis on liver biopsy, 29% had stage 1, 20% had stage 2, 17% had stage 3, and 8% had stage 4 fibrosis. In univariable logistic regression participants with significant fibrosis (stage 2-4) had lower levels of IGFII and haptoglobin and higher levels of IL-8, resistin, sIL-1R1, sIL-2Rα, MCP-1, and TNFα compared with no or mild fibrosis (Table 3; Supporting Table S2). Haptoglobin did not meet the FDR adjusted threshold for statistical significance.
In a multivariable analysis of significant fibrosis compared to no or mild fibrosis, increased IL-8 (OR = 1.81, 95% CI 1.47-2.23, P < 0.001), MCP-1 (OR = 1.12, 95% CI 1.02-1.22, P = 0.020), resistin (OR = 1.10, 95% CI 1.00-1.20, P = 0.040), sIL-1R1 (OR = 1.16, 95% CI 1.04-1.29, P = 0.009), sIL-2Rα (OR = 1.27, 95% CI 1.13-1.43, P < 0.001), and TNFα (OR = 1.22, 95% CI 1.04-1.44, P = 0.018) and decreased IGFII (OR = 0.75, 95% CI 0.67-0.84, P < 0.001) were associated with significant fibrosis (Table 4). Resistin was associated with female sex and increased HOMA-IR and BMI and inversely associated with ALT. sIL-1R1 was associated with female sex and increased BMI, ALT, AST, and HOMA-IR. sIL-2Rα was associated with female sex and increased age, BMI, and AST and inversely associated with triglycerides. TNFα was associated with increased age (Table 5). sIL-2Rα correlated with IL-8, MCP-1, resistin, and sIL-1R1. IL-8 correlated with MCP-1 and TNFα as well. IGFII was inversely correlated with IL-8 and sIL-2Rα (Supporting Table S3).
In this study, using a large population with well-characterized histological NAFLD, we explored the relationship between putative biomarkers and the histologic disease spectrum associated with NAFLD. We identified biomarkers that remain strongly and independently associated with specific NAFLD phenotypes after adjustment for clinical factors. We found that higher aPAI1 levels were associated with NASH (versus non-NASH or borderline NASH). Higher aPAI1, tPAI1, and IGFII and lower adiponectin were associated with significant steatosis; and increased IL-8 levels and decreased IGFII were associated with hepatocyte ballooning. In addition, among seven biomarkers associated with the presence of significant fibrosis, increased IL-8 and sIL-2Rα and decreased IGFII demonstrated the strongest associations.
The diagnosis and staging of NASH have long relied on histologic evaluation of liver biopsy, which is invasive, potentially subject to sampling error, and inconvenient as a repeat measure of disease severity over time. Many studies have explored the ability of biomarkers to predict disease activity.[9, 11, 22-24] These studies have been plagued by small sample sizes, have often lacked liver biopsy for definitive diagnosis, and have measured very few candidate biomarkers. To date, the most promising biomarker for NAFLD has been cytokeratin 18. Wieckowska et al. initially examined the relationship between cytokeratin 18, a protein cleaved during apoptosis, and NASH in 44 consecutive patients with suspected NAFLD. While cytokeratin 18 levels yielded an area under the receiver operating curve of 0.93, only 21 patients had NASH. Moreover, attempts to validate these findings in a larger population of 318 patients yielded unacceptable performance characteristics for the diagnosis of NASH and provided little additional data beyond ALT levels, leaving the need for a biomarker(s) to predict disease activity in NAFLD unmet.
In our study, the only biomarker that remained associated with the histological diagnosis of NASH after adjustment for clinical factors was aPAI1. PAI1 is a serine protease inhibitor and a primary regulator of the fibrinolytic system but also has significant effects on cell adhesion, detachment, and migration. PAI1 gene expression in the liver is up-regulated by endotoxin and inflammatory mediators, and levels may reflect TNFα signal dysregulation.[28, 29] PAI1 gene expression is up-regulated in livers with NASH, and plasma levels are increased in patients with NASH.[30, 31] Previous studies have not evaluated differences in levels of aPAI1 and tPAI1 and their relationship to NAFLD. The active form is secreted by cells but is unstable and spontaneously converts into the latent form within 1-2 hours. The latent form can be converted to the active form, and the increased inflammatory milieu associated with NASH could drive this process. This finding warrants evaluation in further studies.
Our study found an association between increased aPAI1, tPAI1, and IGFII as well as decreased adiponectin and significant steatosis. PAI1 levels correlate more with liver fat content than visceral adipose content. Insulin induces PAI1 expression, and improvement in insulin resistance with weight loss or troglitazone is associated with a reduction in PAI1 levels.[34-36] Multiple studies have shown an association between PAI1 levels and NAFLD, which our study corroborates.[30, 37, 38] IGFII is part of the family of ligands that mediate growth, development, and differentiation and act primarily through IGF1 receptor. IGFII has significant structural homology with insulin, shares biological actions through the insulin receptor, and thereby could lead to decreased lipolytic activity and increased steatosis, as was seen in our study. Adiponectin is an adipokine with anti-inflammatory and insulin-sensitizing effects. Our study confirms previously documented inverse associations with hepatic steatosis but contrasts with previous studies suggesting an inverse association with NASH severity.
We found that increased IL-8 was associated with the presence of hepatocyte ballooning. IL-8 is a chemokine that serves as a chemoattractant for neutrophils and contributes to acute liver inflammation. Previous studies have shown an association between NASH and increased IL-8 levels; however, ours is the first to document a significant association specifically between IL-8 levels and hepatocyte ballooning.[46, 47] Increased lobular inflammation was associated with decreased MCP-1. MCP-1 is a chemoattractant that activates target cells including macrophages and has been associated with hepatic steatosis and NASH; however, prior studies had no or few participants with fibrosis.[5, 8, 48] Our finding of decreased MCP-1 among those with more lobular inflammation is likely confounded by the presence of cirrhosis. Patients with cirrhosis had higher MCP-1 and less lobular inflammation, and when excluded from the analysis the association between MCP-1 and lobular inflammation was no longer statistically significant. Furthermore, MCP-1 did not meet the FDR adjusted threshold for statistical significance on univariable analysis. Therefore, these results should be interpreted with caution.
Seven biomarkers were strongly associated with significant fibrosis in the multivariable model: IL-8, MCP-1, resistin, sIL-1R1, sIL-2Rα, TNFα, and IGFII. The strongest associations in terms of statistical significance and effect size were IL-8, sIL-2Rα, and IGFII. In addition to its relationship with hepatocyte ballooning, increased IL-8 was associated with significant fibrosis. In hepatitis C virus models of hepatic fibrosis, IL-8 was the strongest inducer of alpha-smooth muscle actin expression in primary hepatic stellate cells. IL-8 serum levels and increased gene expression were associated with fibrosis and advanced cirrhosis in a study of other causes of chronic liver disease, and our study supports these findings in patients with NAFLD. sIL-2Rα is formed by proteolytic cleavage of the IL-2 receptor from the cell surface of multiple immune cells and is proportional to its membrane-bound expression. An association between increased levels and advanced fibrosis has been documented in non-NAFLD liver disease.[52-54] Our study is the first to document this association in patients with NAFLD. Finally, our study found an association between decreased IGFII and significant fibrosis, in addition to the previously highlighted association between increased IGFII and steatosis. Prior studies of IGFII have revealed that levels are lower in patients with cirrhosis. In a small study of pediatric patients with NAFLD, lower IGFII levels correlated with the degree of fibrosis; and our study corroborates this finding in adults. IGFII was found to protect against the effects of caspases 3 and 9 in the liver, thereby decreasing free radical damage and apoptosis in rats; and this may underlie the association of IGFII with fibrosis. Further studies are warranted.
Our study has a number of strengths, including the large number of well-characterized participants and the number of biomarkers evaluated. All of the biomarkers were chosen based on a priori hypotheses regarding their relationship with NAFLD, and our findings are well supported by previous literature. We did evaluate multiple biomarkers and outcomes in this study; however, we have highlighted only those with strong associations in multivariable models in our discussion. We provided FDR adjusted P value thresholds in univariable analysis for readers to interpret. By applying the FDR adjusted thresholds to biomarker selection for multivariable models, only the association between MCP-1 and lobular inflammation would have been eliminated from our multivariable models.
Furthermore, a corresponding study among pediatric patients in the NASH-CRN corroborates many of our findings including the associations between IGFII and steatosis and between IL-8 and fibrosis. While we have demonstrated significant relationships between plasma levels of biomarkers and the pathological spectrum of NAFLD, we cannot evaluate causal relationships. Therefore, while certain biomarkers may be critical to the pathogenesis of NAFLD, their plasma levels may not correlate with disease activity or the relationship may have been diminished by a clinical factor included in the multivariable model. However, the goal of this study was to explore the most impactful biomarkers in NAFLD and evaluate the relationships between biomarkers and clinical factors. Therefore, we also evaluated the associations between significant biomarkers and clinical factors to provide context for the potential role of the biomarker in NAFLD. Our study was limited by the cross-sectional measurement of our outcome, which does not allow evaluation of the temporal relationship between specific biomarkers and disease activity or prediction of how changes in biomarkers will affect disease progression. However, the findings of this study should inform biomarker selection for future longitudinal studies and clinical trials. Finally, the patients drawn from the NASH CRN represent those with an a priori diagnosis of NAFLD. Inclusion of patients from the PIVENS trial decreased the prevalence of diabetic patients in our study sample compared to the NAFLD population at large and may affect generalizability. However, our multivariable models adjusted for HOMA-IR, and strong associations between biomarkers and histology remained present.
In conclusion, our study explored relationships between clinical factors, biomarkers, and histologic severity of disease in NAFLD. We found strong associations, after multivariable adjustment for clinical factors, of specific biomarkers with the histologic manifestations of NAFLD. Measurement of serum biomarkers will advance our ability to stratify disease severity in NAFLD and may identify additional pathways to target for therapeutic intervention.