Validation of terminal peptide of procollagen III for the detection and assessment of nonalcoholic steatohepatitis in patients with nonalcoholic fatty liver disease

Authors

  • Sudeep Tanwar,

    Corresponding author
    1. UCL Institute of Liver & Digestive Health, Centre for Hepatology, Division of Medicine, University College London, London, U.K
    • UCL Institute of Liver & Digestive Health, Centre for Hepatology, Division of Medicine, University College London, London, U.K
    Search for more papers by this author
  • Paul M. Trembling,

    1. UCL Institute of Liver & Digestive Health, Centre for Hepatology, Division of Medicine, University College London, London, U.K
    Search for more papers by this author
  • Indra N. Guha,

    1. National Institute for Health Research Biomedical Research Unit (Gastrointestinal and Liver), Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, U.K
    Search for more papers by this author
  • Julie Parkes,

    1. Public Health Sciences and Medical Statistics, University of Southampton, Southampton, U.K.
    Search for more papers by this author
  • Philip Kaye,

    1. National Institute for Health Research Biomedical Research Unit (Gastrointestinal and Liver), Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, U.K
    Search for more papers by this author
  • Alastair D. Burt,

    1. Institute of Cellular Medicine, Newcastle University, Newcastle, U.K
    Search for more papers by this author
  • Stephen D. Ryder,

    1. National Institute for Health Research Biomedical Research Unit (Gastrointestinal and Liver), Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, U.K
    Search for more papers by this author
  • Guruprasad P. Aithal,

    1. National Institute for Health Research Biomedical Research Unit (Gastrointestinal and Liver), Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, U.K
    Search for more papers by this author
  • Christopher P. Day,

    1. Institute of Cellular Medicine, Newcastle University, Newcastle, U.K
    Search for more papers by this author
  • William M. Rosenberg

    1. UCL Institute of Liver & Digestive Health, Centre for Hepatology, Division of Medicine, University College London, London, U.K
    Search for more papers by this author

  • Potential conflict of interest: Dr. Guha received grants from Pfizer and Conatus.

Abstract

Liver biopsy is the reference standard for the detection of nonalcoholic steatohepatitis (NASH) within nonalcoholic fatty liver disease (NAFLD). The aim of this study was to identify a biomarker of NASH in patients without significant fibrosis. In all, 172 patients from two centers with biopsy-proven NAFLD were included in this study. Eighty-four patients from a single center were included as a derivation cohort and 88 patients from a second center were included as a validation cohort. Serum samples were tested for candidate markers of fibrosis and inflammation alongside hematological and biochemical markers. Among patients without advanced fibrosis, terminal peptide of procollagen III (PIIINP) was the only marker found to be associated with a histological diagnosis of NASH in both cohorts. PIIINP also correlated with the total NAFLD activity score (NAS) and its constituent components (P < 0.001). Area under receiver operating characteristic curve (AUROC) for PIIINP in discriminating between NASH and simple steatosis (SS) was 0.77-0.82 in patients with F0-2 fibrosis and 0.82-0.84 in patients with F0-3 fibrosis. PIIINP was elevated in patients with advanced fibrosis, the overwhelming majority of whom had NASH. When incorporating patients with all degrees of fibrosis from both cohorts, PIIINP was able to discriminate between patients with SS and those with NASH or advanced fibrosis with AUROC 0.85-0.87. Conclusion: PIIINP discriminates between SS and NASH or advanced fibrosis. The use of a single biomarker in this context will be of clinical utility in detecting the minority of patients with NAFLD who have NASH or advanced fibrosis related to NASH. (HEPATOLOGY 2013)

Nonalcoholic fatty liver disease (NAFLD) is the hepatic manifestation of the metabolic syndrome and is now one of the leading causes of liver disease worldwide.1, 2 The clinical spectrum of NAFLD encompasses simple steatosis (SS), nonalcoholic steatohepatitis (NASH), and progressive liver fibrosis.3, 4 It is well recognized that the development of fibrotic liver disease in NAFLD is attributed to the progression of SS to NASH. Increasing literature also suggests that patients with histological NASH follow a progressive course that may result in cirrhosis, portal hypertension, liver failure, and hepatocellular carcinoma.5, 6 NASH is thus the most important prognostic feature of liver damage in NAFLD. The early identifiation of NASH is important for stratifying patients for therapeutic intervention prior to the development of significant hepatic fibrosis. Histological staging of a liver biopsy is the reference standard used for stratifying patients into those with benign disease (SS) and those at risk of progressive disease (NASH with developing fibrosis). With the rising prevalence of obesity, up to 30% of the population is at risk of NAFLD, of whom 3%-5% will have NASH and only 1%-2% will have progressive liver fibrosis.7 Liver biopsy cannot be used as the primary method to detect and quantify NASH and fibrosis in the wider population at risk of NAFLD due to practical and ethical reasons because it is invasive, hazardous, resource-intensive, and costly.8

The noninvasive assessment of patients with NAFLD at risk of progressive liver disease has focused on one of two approaches: either to identify fibrosis at an early stage or to identify NASH prior to the development of significant fibrosis. With regard to the detection of developing liver fibrosis, numerous noninvasive tests have been developed for the detection of fibrosis in NAFLD. However, whereas their performance in the detection of advanced fibrosis ranges from good to excellent, their performance in the detection of mild liver fibrosis is in general modest.9 For example, the enhanced liver fibrosis (ELF) test10 (hyularonic acid [HA], terminal peptide of procollagen III [PIIINP], tissue inhibitor of matrix metalloproteinase-1 [TIMP-1]) has been validated in a secondary care setting in the detection of fibrosis in NAFLD.11 Although the detection of severe fibrosis (>F3) in this study was excellent (area under the curve [AUC] 0.90), the performance in the detection of any fibrosis was more modest (AUC 0.76).

Noninvasive tests have also been developed for the diagnosis of NASH in NAFLD but these models have generally had less success than their fibrosis counterparts.12-14 The aim of this study was to identify a biomarker of NASH that would be able to detect NASH prior to the development of significant hepatic fibrosis. Candidate biomarkers of hepatic inflammation (YKL-40, TIMP-1), apoptosis (CK-18), and liver fibrosis (HA, PIIINP, Collagen IV) were selected on the basis of biological plausibility and previous association with NAFLD and analyzed alongside standard laboratory and clinical measurements, considered in part because of their ubiquity in clinical practice, low cost, and ease of use.15-21

Abbreviations: AUROC, area under receiver operating characteristic curve; ELF, enhanced liver fibrosis; HA, hyularonic acid; IQR, interquartile range; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score; NASH, nonalcoholic steatohepatitis; NPV, negative predictive value; PIIINP, terminal peptide of procollagen III; PPV, positive predictive value; SD, standard deviation; SS, simple steatosis; TIMP-1, tissue inhibitor of matrix metalloproteinase-1.

Patients and Methods

The patients included in this study were recruited consecutively from outpatient clinics in two hepatology centers in the United Kingdom. The derivation and validation cohorts were comprised of patients recruited from the Queen's Medical Centre, Nottingham, and the Freeman Hospital, Newcastle-Upon-Tyne, respectively. In both cohorts NAFLD was diagnosed on the basis of the following criteria: (1) elevated aminotransferases (aspartate aminotransferase [AST] or alanine aminotransferases [ALT]); (2) appropriate exclusion of “other” causes of liver disease including alcohol, drugs, autoimmune or viral hepatitis, or cholestatic or metabolic/genetic liver disease. The “other” causes of liver disease were excluded using specific clinical, biochemical, radiographic, and histological criteria. All patients had a weekly ethanol consumption of less than 140 g in women and 210 g in men. The patients included in this study all had liver biopsies at the centers between 1998 and 2006 and histology was consistent with NAFLD. The serum samples used in the study were taken within 3 months of biopsy. Height and weight were recorded for all patients and body mass index (BMI) calculated. Serum sample samples were obtained for routine biochemistry (including ALT, AST, gamma glutamyl transferase [GGT], bilirubin, albumin, and alkaline phosphatase), full blood count, measurements of insulin resistance (fasting glucose and insulin), ferritin, and cholesterol. Serum samples were analyzed for levels of TIMP-1, HA, PIIINP, Collagen IV, and YKL-40 at an independent reference laboratory (iQur, Southampton, UK). Stored sera taken from patients in the derivation cohort (Nottingham) were also tested for caspase-cleaved CK-18 (M30 Apoptosense enzyme-linked immunosorbent assay [ELISA] assay, Peviva, Sweden).

Liver Biopsy.

Liver biopsies were assessed by a single senior liver histopathologist at each center (Nottingham, P.K., Newcastle, A.B.) and scored for histological grade and fibrosis stage using the classification system described by Kleiner et al.22 In this system, fibrosis was scored using a 5-point score system (1 = mild/moderate zone 3 perisinusoidal fibrosis, or portal fibrosis only; 2 = zone 3 and portal/periportal fibrosis; 3 = bridging fibrosis; 4 = cirrhosis). The grade of NASH was assessed using the NAFLD activity score (NAS) scored from (0-8), which incorporates the scores of steatosis (0-3), ballooning (0-2), and lobular inflammation (0-3). In addition to using the Kleiner scoring system, both histopathologists classified the biopsies in a dichotomous manner as either “definite NASH” or “borderline NASH / simple steatosis” using agreed criteria.

Analysis.

Patients with F3 and F4 fibrosis in both cohorts were excluded from the initial analysis as the aim of the study was to assess the performance of tests in the detection of patients with NAFLD who have NASH prior to the development of significant liver fibrosis. Furthermore, it is well described that patients with NAFLD and advanced fibrosis may no longer exhibit the necroinflammatory activity or steatosis that was present earlier in their disease course.23 Nineteen and 17 patients (total 36) were excluded from the initial analyses of the derivation and validation cohort, respectively, due to the presence of advanced fibrosis.

Statistical Analysis.

Data were analyzed using SPSS v. 20.0 (SPSS, Chicago, IL) and STATA v. 11.0 (SAS, Cary, NC). Patients without advanced fibrosis (F0-2) were separated into two categories based on their final histological diagnosis of either “definite NASH” or “borderline NASH / simple steatosis” and clinical and laboratory variables were compared between categories. Categorical variables including sex and presence of diabetes were compared between the two groups and the chi-square test was applied to test significance. Parametric continuous variables and nonparametric continuous variables were compared between the two groups using two-sided Student's t test and the Mann-Whitney U test, respectively.

Binary logistic regression was used to assess the association of variables with a histological diagnosis of NASH. Correlation coefficients (Spearman's rho) were used to determine the strength of the relationship of the multiple variables to the NAS and its individual components in patients without advanced fibrosis. Receiver operating characteristic (ROC) curves together with 95% confidence intervals were plotted to determine the performance of the identified variable(s). Area under receiver operator characteristic (AUROC) curves were compared using the method of Delong et al.24 Sensitivity and specificity, predictive values, likelihood ratios (positive and negative), and diagnostic odds ratios were calculated at thresholds derived from ROC curves. The identified parameter(s) in the initial analyses were then compared with the NASH scores and fibrosis stages of patients in the entire cohorts.

Results

Derivation and Validation Cohorts.

The baseline characteristics of both cohorts are shown in Tables 1 and 2 and Fig. 1. Although the distribution of gender, age, and diabetes was similar within both cohorts, patients in the derivation cohort were found to have significantly lower BMI (P < 0.001), more severe fibrosis (P = 0.022), a greater prevalence of NASH, and higher NAS (P < 0.001). The majority of patients in the validation cohort had either minimal or no fibrosis. In contrast, the distribution of fibrosis in the derivation cohort was more analogous to a secondary care population with all stages of fibrosis well represented.

Figure 1.

Distribution of fibrosis stages in the derivation (A) and validation (B) cohorts. Distribution of NAS in patients without advanced fibrosis (F0-F2) in derivation (C) and validation (D) cohorts.

Table 1. Derivation Cohort: Demographic and Baseline Data of Patients Without Advanced Fibrosis (F0-2)
VariableSimple Steatosis/ Borderline NASH n = 11NASH n = 54P valueOverall n = 65Univariate Regression OR95% CI (P value)Multiple Regression OR95% CI (P value)
  1. Significant associations (P < 0.05) are displayed in bold.

  2. Parametric variables are presented as mean ± SD.

  3. Non-parametric variables are presented as median values and IQR.

Male (n,%)7 (64%)14 (26%)NS45 (69.2%)    
Age (years)48.8 ± 9.847.9 ± 11.5NS48.0 ± 11.1    
BMI (kg/m2)30.55 ± 6.1729.69 ± 4.50NS29.83 ± 4.56    
Diabetes7 (64%)32 (59%)NS39 (60%)    
Trig. (mmol/L)1.95 ± 0.761.99 ± 1.19NS1.98 ± 1.13    
Glucose (mmol/L)6.07 ± 2.005.67 ± 1.34NS5.74 ± 1.47    
HOMA-IR2.83 (IQR 2.11)2.86 (IQR 2.85)NS2.86 (IQR 2.53)    
ALT (IU/L)53 (IQR 31)64.0 (IQR 50.8)NS61.0 (IQR 38.5)    
GGT (IU/L)131 ± 111120 ± 111NS116 ± 82    
Ferritin (ng/mL)55 (IQR 63.5)179 (IQR 200)0.029147 (IQR 212)1.011.00-1.02 (0.04)1.021.00-1.03 (0.026)
Platelets (x109/L)235.0 ± 35.4251.4 ± 70.8NS248.8 ± 66.6    
CK18 (ng/mL)149.3 ± 60.3208.4 ± 162.4NS1561.010.99-1.01 (0.295)  
HA (ng/mL)30.9 (IQR 33.7)42.3 (IQR 26.8)NS21.4 (IQR 28.1)    
TIMP-1 (ng/mL)619 ± 54755 ± 1330.002733 ± 1341.011.00-1.03 (0.005)  
PIIINP (ng/mL)5.36 ± 2.058.92 ± 2.930.0047.03 ± 2.902.221.26-3.91 (0.006)3.001.40-6.50 (0.005)
YKL-40 (ng/mL)81.8 (IQR 67.6)92.0 (IQR 77.1)NS84.9 (IQR 82.5)    
Coll IV (ng/mL)145 ± 16177 ± 350.005172 ± 351.041.00-1.07 (0.009)  
NAS1.2 ± 1.14.3 ± 1.3 3.8 ± 1.7    
Table 2. Validation Cohort- Demographic and Baseline Data of Patients Without Advanced Fibrosis (F0-2)
VariableSimple Steatosis/ Borderline NASH n = 51NASH n = 20P valueOverall n = 71Univariate Regression OR95% CI (P value)Multiple Regression OR95% CI (P value)
  1. Significant associations (P < 0.05) are displayed in bold.

  2. Parametric variables are presented as mean ± SD.

  3. Non-parametric variables are presented as median values and IQR.

Male (n,%)40 (78%)12 (60%)NS52 (73%)    
Age (years)44.7 ± 11.343.8 ± 14.3NS44.3 ± 12.1    
BMI (kg/m2)33.3 ± 5.936.4 ± 5.1NS34.20 ± 5.85    
Diabetes28 (54%)10 (50%)NS33 (47%)    
Trig. (mmol/L)2.81 ± 1.992.62 ± 1.52NS2.76 ± 1.87    
Glucose (mmol/L)5.53 ± 1.095.67 ± 0.87NS5.58 ± 1.02    
HOMA-IR3.38 (IQR 4.91)5.36 (IQR 9.48)NS3.58 (IQR 4.83)    
ALT (IU/L)58.0 (IQR 47.0)69.0 (IQR 34.0)NS66.0 (IQR 46.0)    
GGT (IU/L)93.4 ± 53.568.8 ± 49.5NS86 ± 53  0.970.95-0.99 (0.02)
Ferritin (ng/mL)125 (IQR 150)180 (IQR 171)NS143 (IQR 152)    
Platelets (x109/L)264.1 ± 53.4252.9 ± 60.6NS261.0 ± 55.6    
CK18 (ng/mL)NANANANA    
HA (ng/mL)25.5 ± 19.930.5 ± 19.2NS26.9 ± 19.7    
TIMP-1 (ng/mL)709 ± 121730 ± 120NS715 ± 121    
PIIINP (ng/mL)6.4 ± 2.7310.2 ± 6.580.0017.52 ± 4.481.291.07-1.56 (0.009)1.441.14-1.82 (0.002)
YKL-40 (ng/mL)44.4 (IQR 56.1)73.2 (IQR 92.1)NS54.0 (IQR 57.1)    
Coll IV (ng/mL)147 (IQR 43)161 (IQR 39)NS149 (IQR 43)    
NAS1.82 ± 0.874.20 ± 1.20      

PIIINP Is the Only Marker Associated With a Histological Diagnosis of NASH in Both Cohorts After Multivariate Analysis.

Within the derivation cohort, univariate analysis identified a significant association between a histological diagnosis of NASH and serum ferritin, TIMP-1, PIIINP, and Collagen IV. However, only PIIINP and ferritin were significantly associated with a histological diagnosis of NASH after accounting for the remaining variables as potential confounders. Within the validation cohort only PIIIINP was significantly associated with a histological diagnosis of NASH on univariate analysis. However, multivariate analysis identified that both PIIINP and GGT were associated with a histological diagnosis of NASH.

PIIINP Correlates With NAS and Its Constituent Components.

Given the positive association of PIIINP with steatohepatitis in both cohorts, the relationship of PIIIINP with both the NAS and its constituent components was explored. Within both cohorts PIIINP was found to correlate significantly with the NAS, the degree of steatosis, ballooning, and lobular inflammation (Supporting Data).

PIIINP Performs Well in Discriminating Both Between NASH and SS and Also Between Differing Grades of NASH in Patients With F0-2 and F0-3 Fibrosis.

ROC curves were plotted for PIIINP in discriminating differing both between NASH and SS and also differing degrees of NASH (NAS, degree of ballooning, and degree of lobular inflammation) (Table 3). Among patients without advanced fibrosis (F0-2) the performance of PIIINP in discriminating NASH from SS was 0.83 and 0.77 in the derivation and validation cohorts, respectively. When considering patients without cirrhosis (F0-3), the performance of PIIINP in discriminating NASH from SS was 0.84 and 0.82 in the derivation and validation cohorts, respectively.

Table 3. PIIINP: Performance With Respect to Discriminating Both Histological NASH From ‘Non-NASH’ and Differing Grades of NASH in Patients With F0-2 in Both Cohorts
NASH GradeCohortNumbers in Each GroupAUROC95% CIP valueStandard Error
Histological Diagnosis: ‘Non-NASH’™ vs. NASHDerivationn = 11 vs. n = 540.830.71-0.940.0010.058
Validationn = 51 vs. n = 200.780.67-0.89<0.0010.057
NAS: 0-2 vs. 3-8Derivationn = 13 vs. n = 520.850.74-0.96<0.0010.056
Validationn = 41 vs. n = 300.800.70-0.90<0.0010.051
NAS: 0-3 vs. 4-8Derivationn = 28 vs. n = 370.860.77-0.95<0.0010.046
Validationn = 54 vs. n = 170.790.67-0.91<0.0010.060
NAS: 0-4 vs. 5-8Derivationn = 41 vs. n = 240.880.80-0.96<0.0010.041
Validationn = 65 vs. n = 60.800.61-0.980.0170.092
NAS: 0-5 vs. 6-8Derivationn = 55 vs. n = 100.830.70-0.960.0010.066
Validationn = 68 vs. n = 30.830.70-0.970.0520.052
Lobular Inflammation: 0 vs. 1-3Derivationn = 15 vs. n = 500.770.63-0.910.0020.072
Validationn = 46 vs. n = 250.850.76-0.94<0.0010.046
Lobular Inflammation: 0-1 vs. 2-3Derivationn = 59 vs. n = 60.890.74-1.000.0020.064
Validationn = 67 vs. n = 40.860.75-0.970.0550.055
Ballooning: 0-1 vs. 2Derivationn = 48 vs. n = 170.800.68-0.91<0.0010.058
Validationn = 69 vs. n = 20.760.64-0.880.2110.061

The performance of PIIINP in discriminating between differing degrees of NASH in both cohorts, both among patients with F0-2 and F0-3, was good and, in particular, the ability to discriminate lobular inflammation was very good (0.86-0.89) (Supporting Data).

PIIINP Levels in Patients With Advanced Fibrosis (F3-4).

In both cohorts, the majority of patients (90%-94%) with advanced fibrosis had a histological diagnosis of NASH. PIIINP levels reflected a hierarchy of liver disease severity ranging from SS with no/mild fibrosis, steatohepatitis with no/mild fibrosis to advanced fibrosis (Fig. 2). This observation was made regardless of which scoring system for steatohepatitis was employed (histological NASH, NAS ≥5, lobular inflammation ≥2).

Figure 2.

Boxplots of PIIINP concentration (ng/mL) in both the derivation and validation cohorts with respect to patients with SS, NASH, and advanced fibrosis. (A) Derivation cohort and (B) validation cohort: Patients stratified into F0-2 with simple steatosis, F0-2 with NASH, and advanced fibrosis (F3-4). (C) Derivation cohort and (D) validation cohort: Patients stratified into F0-2 with NAS 0-4, F0-2 with NAS 5-8, and advanced fibrosis (F3-4). (E) Derivation cohort and (F) validation cohort: Patients stratified into F0-2 without moderate or severe lobular inflammation, F0-2 with moderate or severe lobular inflammation, and advanced fibrosis (F3-4).

Clinical Utility of PIIINP in the Diagnosis of NASH and the Exclusion of Advanced Fibrosis.

The performance of PIIINP in the detection of NASH alone, advanced fibrosis alone, or NASH and advanced fibrosis is represented in Fig. 2 and presented numerically in the Supporting Data. Despite the two cohorts having a very different prevalence of steatohepatitis, the performance of PIIINP in discriminating SS from either NASH or advanced fibrosis was uniformly good (AUC 0.85-0.87) (Table 4).

Table 4. Overlapping Utility: the Ability of PIIINP to Discriminate Between Either Simple Steatosis and NASH or Advanced Fibrosis (F3-4) in Both Cohorts
NASH GradeCohortNumbers in Each GroupAUROC95% CIP valueStandard Error
Non-NASH and F0-2 vs. NASH or F3-4Derivationn = 11 vs. n = 730.860.77-0.95<0.0010.047
Validationn = 51 vs. n = 370.850.77-0.93<0.0010.041
NAS 0-4 and F0-2 vs. NAS 5-8 or F3-4Derivationn = 14 vs. n = 430.860.78-0.94<0.0010.041
Validationn = 65 vs. n = 230.870.78-0.96<0.0010.045
Lob. Inf.0-1 and F0-2 Vs. Lob. Inf. 2-3 or F3-4Derivationn = 59 vs. n = 250.860.77-0.96<0.0010.049
Validationn = 68 vs. n = 200.870.78-0.95<0.0010.044

As previously stated, PIIINP values associated with NASH in patients without advanced fibrosis were lower than those associated with advanced fibrosis. The clinical utility of this observation is that a clinically relevant diagnostic threshold selected for the detection of NASH (as derived from patients without advanced fibrosis) will have a higher sensitivity (and therefore higher negative predictive value [NPV]) for a detection of advanced fibrosis.

In this way, PIIINP levels above a threshold set for the diagnosis of NASH could be used as a screening tool in a primary care setting. Although a positive result could be used to “rule in” NASH, a negative result could be used to “rule out” NASH and/or advanced fibrosis. Diagnostic thresholds for PIIINP and its performance in the diagnosis of histological steatohepatitis and differing degrees of NASH (NAS, ballooning, and lobular inflammation) are shown in Table 5. The thresholds corresponding to 80% sensitivity and specificity are displayed together with a third threshold that maximizes the positive predictive value (PPV) for either histological NASH or the relevant grade of NASH. Also displayed are the NPVs for advanced fibrosis and cirrhosis when these thresholds were applied to the entire cohort (F0-4).

Table 5. Performance of PIIINP in Discriminating Either Histological NASH or Differing Grades of NASH and Utility in Excluding Advanced Fibrosis
Histological NASH or Grade of NASHPrev.CenterP3NP Threshold (ng/mL)Sens.Spec.PPVNPVLR (+ve)LR (-ve)DORNPVa in Entire Cohort for Fibrosis
<F3-4<F4
  1. Thresholds of PIIINP concentration (ng/mL) applied to patients with F0-2 fibrosis.

  2. a

    Prevalence of F3-F4 & F4 fibrosis in the derivation and validation cohorts is 23% & 8% and 19% & 8% respectively.

Histological NASH83%Derivation5.28073944330.2711.090100
6.0658094323.30.447.494100
11.01310010019N/A0.87N/A8999
28%Validation6.4805641881.80.365.196100
7.26080538430.506.0098100
11.030978078100.7213.99399
NAS ≥537%Derivation6.6806860852.50.298.695100
7.2778069863.90.2913.395100
11.02197806870.818.68999
9%Validation6.6805716971.90.355.397100
7.2678025963.40.418.298100
11.0509445958.30.5315.79399
Severe Ballooning26%Derivation6.7806444902.20.317.295100
7.7598050853.00.515.89298
11.0299777809.70.7313.28999
3%Validation7.210071161006.250N/A98100
7.950807982.50.634.098100
11.0N/AN/AN/AN/AN/AN/AN/A9399
Lob Inflamm. ≥123%Derivation5.2806037912.00.336.190100
7.0528044852.60.604.395100
11.0129910080N/A0.88N/A8999
35%Validation6.6807462873.10.2711.497100
7.0768067863.80.3012.798100
11.03210010073N/A0.68N/A9399
Lob. Inflamm. ≥29%Derivation7.2836720982.50.2510.195100
8.0838029984.20.2119.89298
11.050955095100.5315.78999
6%Validation7.2807316983.00.2711.098100
7.6758019983.80.3112.198100
11.0509126975.60.5510.19399

Clinically Relevant PIIINP Thresholds.

The application of PIIINP at a threshold of 11.0 ng/mL resulted in a PPV of between 80%-100% for a histological diagnosis of NASH, and 100% for the presence of lobular hepatitis in the F0-2 group of both cohorts. When applied to patients with all degrees of fibrosis in both cohorts, a threshold of 11.0 ng/mL resulted in a PPV of 74%-100% for a diagnosis of either steatohepatitis or advanced fibrosis. At this threshold the NPV for a diagnosis of advanced fibrosis alone ranged between 89%-93%.

The use of the PIIINP at a 6.6 ng/mL threshold resulted in an NPV for a severe grade of steatohepatitis (NAS 5-8, severe ballooning, severe lobular inflammation) in the F0-2 portion of both cohorts of 85%-100%. When applied to patients with all degrees of fibrosis in both cohorts, a PIIINP level of 6.6 ng/mL had an NPV for either NASH or advanced fibrosis of 80%-95% and an NPV for advanced fibrosis of 95%-97% and 100% for cirrhosis.

PIIINP as a Marker of Fibrosis or NASH?

PIIINP has previously been studied as a marker of fibrosis and is one of the three component proteins of the ELF test. Given that NASH is associated with the development of fibrosis, we assessed whether or not the performance of PIIINP as a marker of NASH was simply due to its ability to detect patients with NASH who were accruing fibrosis. The ability of PIIINP to discriminate differing degrees of fibrosis in patients without advanced fibrosis (F0-2) in the derivation cohort was poor (AUROC 0.63 F0 versus F1-2, AUROC 0.65 F0-1 versus F2) (Fig. 3). It is evident that in patients with Kleiner fibrosis stages 0-2, PIIINP discriminated poorly between fibrosis stages, and median PIIINP levels were similar between groups (Fig. 3A). However, PIIINP levels were significantly different between patients stratified by a histological diagnosis, NAS (0-4 and 5-8), and lobular inflammation at each fibrosis stage (Figs. 2, 3B). Furthermore, when analyzing patients with no and minimal fibrosis only (F0-1) in the derivation cohort, the ability of PIIINP to discriminate between NASH and SS was consistent (AUROC 0.83) and the ability of PIIINP to discriminate differing degrees of NASH was even better (NAS 0-4 versus 5-8 AUROC 0.90, ballooning 0-1 versus 2 AUROC 0.92, lobular inflammation 0-1 versus 2-3 AUROC 0.96) (Supporting Data).

Figure 3.

(A) Boxplots of serum PIIINP concentration (ng/mL) with respect to fibrosis stage in the derivation cohort. (B) Boxplots of serum PIIINP concentration (ng/mL) in patients without advanced fibrosis (F0-F2) with respect to fibrosis stage when stratified by NAS (NAS 0-4 and NAS 5-8).

Discussion

In this study we have shown that PIIINP, a biological component involved in fibrogenesis, is able to discriminate between patients with SS alone and those with NASH among a population of patients with NAFLD. Among NAFLD patients with fibrosis ranging from none to moderate (F0-2), PIIINP levels differed significantly between patients with or without a histological diagnosis of NASH (AUC 0.78-0.83), a NAS of 5-8 (0.80-0.88), and with or without severe lobular hepatitis (AUC 0.86-0.89), thus permitting the detection of patients with pathological lesions from those with less concerning manifestations of NAFLD. Considering patients with all degrees of fibrosis, PIIINP was able to discriminate between patients with SS and those with either NASH or advanced fibrosis (AUC 0.85-0.87). This is unsurprising, as progressive fibrosis in NAFLD is thought to be the consequence of longstanding NASH. Elevated serum PIIINP has been previously been shown to be associated with elevated proinflammatory cytokines in liver disease and this proinflammatory state is likely to be the driving force for progressive liver fibrosis.25 Furthermore, serum PIIINP has been extensively studied in inflammatory joint disease where elevated PIIINP has been associated with active inflammation representing increased tissue turnover.26 We hypothesize a similar process occurs in active steatohepatitis.

Hitherto, almost all published studies of markers for NASH have evaluated combinations of laboratory parameters and clinical scores. The only other single marker to be validated in the detection of NASH is CK-18, a marker of apoptosis. A validation study described good performance of caspase-cleaved CK-18 in discriminating NASH from SS with AUC of 0.83 in a cohort of patients with advanced fibrosis.27 Biopsies in this study were scored as “not NASH,” “borderline NASH,” and NASH. In the present study CK-18 was not identified as a discriminatory marker of NASH.

In addition to using a categorical diagnosis of NASH, we also used the NAS and its constituent components to grade the degree of steatohepatitis. Despite its criticisms, the use of NAS in the grading of NASH in contrast to a binary diagnosis of NASH (NASH or “not NASH”) permits the evaluation of the relative contributions of the components of the NAS to a specific variable or panel of tests. However, whereas many investigators have taken NAS ≥5 to be synonymous with a diagnosis of NASH, a recent study suggested that NAS at this score is associated with a histological diagnosis of NASH in ∼70% of cases.28 Higher NAS, while more specific for a histological diagnosis of NASH, is less sensitive, necessitating the use of a NAS of 5 in clinical practice. In addition, the NAS has been criticized because it places equal weighting on lobular hepatitis and less weight on ballooning than steatosis. In this study, we noted that patients in the more obese validation cohort had more steatosis and less lobular inflammation than patients with the same NAS in the derivation cohort. With regard to caspase-cleaved CK-18, in our study within the derivation cohort we found that the AUC for CK-18 for its ability to discriminate between patients with a histological diagnosis of steatohepatitis was 0.56, and in patients without severe lobular inflammation or ballooning were both 0.71. Another approach to the use of CK-18 as a biomarker of NASH has been to measure the total level of CK-18. In a recent study, total CK-18 levels performed better than caspase-cleaved CK-18 with AUC for discriminating NASH from SS of 0.81.14 In this study, the AUC for caspase-cleaved CK-18 in its ability to discriminate NASH from SS was 0.71, which is similar to that observed in the present study. In contrast, however, both total CK-18 and CK-18 fragments performed similarly in their ability to discriminate NASH from SS in another study of 101 patients with morbid obesity with AUC of 0.81 and 0.83, respectively.29

Despite both cohorts containing patients who had been recruited from specialist liver centers, we found a differing spectrum of NAFLD-related liver disease in each cohort. The derivation cohort is comprised of a patient population that is more analogous to what one might expect to see in secondary care with patients with minimal fibrosis and SS being underrepresented. In contrast, the validation cohort is comprised of patients similar to those encountered in primary care, with the majority of patients having no or mild fibrosis. As a result the two populations studied have provided a useful platform on which to derive and validate biomarker test performance. In addition, in order to simulate biomarker performance in a primary care setting where the advantage of noninvasive assessment of NAFLD is likely to be of greatest benefit, our initial analysis was restricted to patients who did not have advanced fibrosis. Critical evaluations of test performance in primary care are required in order to determine the utility of PIIINP in that setting.

Although we acknowledge that our results need to be reproduced in larger independent series, the relative consistency of the thresholds of PIIINP for detecting differing grades of steatohepatitis in each cohort is striking in view of their differing disease severity. PIIINP at thresholds of 6.6, 7.2, and 11.0 ng/mL were found in the F0-2 portion of both cohorts to have 80% sensitivity, 80% specificity, and the highest PPV (45%-80%), respectively, for a NAS of 5-8. When the same thresholds were applied to the entire cohort, their performance in the detection of NASH or fibrosis remained high. In theory, these thresholds could be applied in both primary and secondary care settings. The lower threshold of 6.6 ng/mL with 80% sensitivity could be used to “rule out” patients with NASH in primary care. Conversely, the higher thresholds of 7.6 or 11 ng/mL could be used to “rule in” NASH or advanced fibrosis, respectively, in secondary care where more advanced disease is more prevalent. Nevertheless, we are mindful that the performance of these proposed thresholds may vary significantly in other patient populations due to the relatively small sample sizes of our two cohorts.

What is also intriguing is the relatively linear relationship between PIIINP and the grade of NASH. As a result, there is potential for PIIINP to be used as a marker of NASH both in treatment studies and longitudinal studies. Also of interest will be whether PIIINP can also be used in the noninvasive assessment of inflammation in diseases such as alcoholic hepatitis and viral hepatitis.

In summary, we have made two important observations in our two cohorts that have differing severities of NAFLD-related liver disease that are relevant to both primary and secondary care settings. Within a cohort of patients with NAFLD without advanced fibrosis, PIIINP levels allowed discriminating the majority of patients with SS from those with steatohepatitis. Secondly, within a cohort of patients with NAFLD comprised of all stages of fibrosis, PIIINP levels allowed discriminating the majority of patients with NASH or advanced fibrosis from those with SS. We suggest further studies are required in both primary and secondary care to confirm our findings, including studies to assess the relationship between changes in steatohepatitis and PIIINP in response to interventions.

Author Contributions: W.M.R. and A.D.B. are inventors of the ELF test. S.T., P.T., J.P., and W.M.R. contributed to the study design, analysis, and article drafting. I.N.G., S.D.R., G.P.A., and C.D. were responsible for patient recruitment, sample collection, and review of the article. P.K. and A.D.B. scored the liver biopsy specimens used in the study.

Ancillary