Importance of hepatic fibrosis in cystic fibrosis and the predictive value of liver biopsy

Authors

  • Peter J. Lewindon,

    1. Gastroenterology, Royal Children's Hospital, Brisbane, Queensland, Australia
    2. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ross W. Shepherd,

    1. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    2. Department of Pediatrics, Washington University School of Medicine, St. Louis, MO
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Meagan J. Walsh,

    1. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ristan M. Greer,

    1. School of Veterinary Science, Queensland, Australia
    Search for more papers by this author
  • Richard Williamson,

    1. Department of Histopathology, Royal Brisbane Hospital, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Tamara N. Pereira,

    1. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Kieran Frawley,

    1. Medical Imaging, Royal Children's Hospital, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Scott C. Bell,

    1. Department of Thoracic Medicine, Prince Charles Hospital and School of Medicine, Queensland, Australia
    Search for more papers by this author
  • Jeffery L. Smith,

    1. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    2. School of Chemistry and Molecular Biology, University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Grant A. Ramm

    Corresponding author
    1. Hepatic Fibrosis Group, Queensland Institute of Medical Research, Brisbane, Queensland, Australia
    • Hepatic Fibrosis Group, Queensland Institute of Medical Research, P.O. Royal Brisbane and Women's Hospital, Brisbane, Queensland, Australia 4029
    Search for more papers by this author
    • fax: 61-7-3362 0191


  • Potential conflict of interest: Nothing to report.

Abstract

Cystic fibrosis liver disease (CFLD), which results from progressive hepatobiliary fibrosis, is an important cause of morbidity and mortality, but it is difficult to identify before portal hypertension (PHT) ensues. Clinical signs, serum alanine aminotransferase (ALT) levels, and ultrasound (US) are widely applied, but their value in predicting the presence of cirrhosis, the development of PHT, or adverse outcomes is undetermined. The potential gold standard, liver biopsy, is not standard practice and, notwithstanding sampling error considerations, has not been systematically evaluated. Forty patients with cystic fibrosis (median age = 10.6 years) with abnormal clinical, biochemical, and US findings were subjected to dual-pass percutaneous liver biopsy. Clinical outcomes were recorded over 12 years of follow-up (median = 9.5 years for survivors). Logistic regression and receiver operating characteristic analyses were applied to predict hepatic fibrosis (which was assessed by fibrosis staging and quantitative immunohistochemistry) and the occurrence of PHT. PHT occurred in 17 of 40 patients (42%), including 6 of 7 (17%) who died during follow-up. Clinical examination, serum ALT levels, and US findings failed to predict either the presence of liver fibrosis or the development of PHT. Fibrosis staging on liver biopsy, where the accuracy was improved by dual passes (P = 0.002, nonconcordance = 38%), predicted the development of PHT (P < 0.001), which occurred more frequently and at a younger age in those with severe fibrosis. Conclusion: Clinical modalities currently employed to evaluate suspected CFLD help to identify a cohort of children at risk for liver disease and adverse outcomes but do not predict an individual's risk of liver fibrosis or PHT development. Liver fibrosis on biopsy predicts the development of clinically significant liver disease. Dual passes help to address sampling concerns. Liver biopsy has a relevant role in the management of patients with suspected CFLD and deserves more widespread application. (HEPATOLOGY 2011)

Hepatobiliary fibrosis causes significant mortality and morbidity in patients with cystic fibrosis (CF).1 Liver cirrhosis has been reported in up to 10% of children2 and in less than 2% of adults with CF3; suggesting a survival disadvantage. Liver disease is reportedly the third leading direct cause of death.3 Complications such as portal hypertension (PHT), nutritional growth failure, and, in some cases, liver synthetic failure impair the quality of life in up to 20% of patients1, 4; this is also highlighted by improvements in the quality of life, nutritional status, and respiratory function after successful liver transplantation.1, 4 Cystic fibrosis liver disease (CFLD) has its origins early in life,1, 4 and its onset and progression to cirrhosis and PHT are unpredictable. Early diagnosis is elusive,1, 2, 4 and a rigorous comparison of current diagnostic modalities against a gold standard is lacking. Clinical signs or elevated serum aminotransferases may vary over time or may be absent in patients with advanced cirrhosis.2, 4 Ultrasound (US) is widely used5-8 but may not distinguish liver fibrosis from steatosis.8 Although these tests are widely employed in CFLD evaluation, their value in predicting significant liver disease has not been determined to date by a prospective study. Liver biopsy is not widely used and has not been systematically evaluated in this clinical context. There are perceived but poorly tested issues of sampling error in CFLD, and only limited studies have included histology in diagnosis, management, or the study of putative therapies.8-10

However, liver pathology is being characterized in CFLD, and the importance of hepatic fibrogenesis is generating interest in the role of liver biopsy in clinical practice. The CF transmembrane regulator protein is expressed in the cholangiocyte.11 Altered biliary transport12 appears to lead to focal obstruction of bile flow, retention of toxic bile acids,13 up-regulation of key chemokines,13 induction of hepatic stellate cell chemotaxis and proliferation, and peribiliary fibrogenesis,13, 14 which is the key event leading to the pathognomonic focal biliary fibrosis of CFLD.14 Some but not all cases progress to multilobular biliary cirrhosis via bile duct and hepatocyte injury and active fibrogenesis along the expanding scar interface13, 14; this is reflected also by the appearance of potential biomarkers in the serum.12, 13, 15, 16 Recent studies have suggested that the Z allele of the serpin peptidase inhibitor clade A member 1 gene is a risk factor for cirrhosis in CF, although the role of this and other potential genetic modifiers in CFLD requires further mechanistic evaluation.17

Here we evaluate dual-pass liver biopsy and the commonly used clinical tools available to clinicians when they are confronted with a patient with suspected CFLD. We look at the ability of the latter to predict hepatobiliary fibrosis on biopsy, and we compare the value of biopsy to the value of clinical modalities currently used to predict adverse outcomes (i.e., PHT and/or liver failure) and mortality over prolonged clinical follow-up (up to 12 years). We hypothesized that hepatic fibrosis on biopsy best predicts clinically significant CFLD and that the evaluation of dual-pass liver biopsy pairs improves diagnostic accuracy.

Abbreviations

α-SMA, α-smooth muscle actin; ALT, alanine aminotransferase; AUROC, area under the receiver operating characteristic curve; CF, cystic fibrosis; CFLD, cystic fibrosis liver disease; CFRD, cystic fibrosis–related diabetes mellitus; FEV, forced expiratory volume; FEV1, forced expiratory volume in 1 second; HM, hepatomegaly; HSM, hepatosplenomegaly; LFT, liver function test; NPV, negative predictive value; PHT, portal hypertension; PPV, positive predictive value; ROC, receiver operating characteristic; US, ultrasound.

Patients and Methods

Study Subjects and Design

This is a prospective cohort follow-up study of 40 patients with suspected CFLD: they were identified and referred by the CF clinic of the Royal Children's Hospital (Brisbane, Australia), were enrolled between 1999 and 2004, and were followed until death, transplantation, or survival as of March 2009. This clinic is a major CF referral center (over 250 patients) for Queensland, Australia. The details and progress of all patients were recorded prospectively via a detailed clinical database. Suspected CFLD was defined as two of the following: (1) hepatomegaly (HM) with or without splenomegaly, (2) a persistent (>6-month) elevation of serum alanine aminotransferase (ALT; level > 1.5 × upper limit of normal), and (3) abnormal liver US findings (abnormal echogenicity or a nodular edge). Those with liver synthetic dysfunction or a history of hepatobiliary surgery were excluded. The study conformed to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the ethics committees of the Royal Children's Hospital and the Queensland Institute of Medical Research. Informed consent was obtained from parents and, when appropriate, from patients.

At enrollment, the following were performed or determined for all patients: history, physical examination, Δf508 genotype, lung function, serum aminotransferases, liver synthetic function (international normalized ratio and albumin), and liver US as well as upper gastrointestinal endoscopy, serum draw for research, and dual-pass liver biopsy under general anesthesia. Specific note was made of the presence or absence of PHT, which was defined as the occurrence of any of the following: endoscopic esophageal varices and persistent clinical splenomegaly (palpable spleen below the left costal margin that was confirmed to span outside the normal range for the patient's age by US) with or without thrombocytopenia (platelet count <150,000). Portal vein thrombosis was excluded by Doppler imaging. When PHT was present, the age of onset was recorded by chart review.

During follow-up (up to 12 years), all patients received standard CF pulmonary and nutritional care, all patients with biopsy-confirmed fibrosis were prescribed ursodeoxycholic acid (15 mg/kg/day), and all patients were reviewed at least on a 6-month basis. For the purposes of this study, prospectively recorded follow-up data included clinical progress, occurrence of cystic fibrosis–related diabetes mellitus (CFRD; defined as insulin-dependent diabetes mellitus), survival, solid organ transplantation, forced expiratory volume in 1 second (FEV1), liver aminotransferases, liver synthetic function, and occurrence of PHT (as defined previously).

This study cohort, now followed for up to 12 years, was the subject of earlier extensive studies at enrollment; these studies included the US findings,8 the bile acid composition,12 the mechanisms of fibrogenesis,13, 14 the roles of chemokines,13 and the cross-sectional evaluation of serum fibrosis biomarkers in those with liver fibrosis versus those without liver fibrosis and in those with early fibrosis versus those with more severe fibrosis13, 15 in CFLD.

Liver Biopsy

Dual-pass liver biopsy samples (n = 80 from 40 patients) were obtained percutaneously with US guidance under general anesthesia (14-Fr Tru-Cut, throw length = 20 mm) from the right lobe via the same skin incision with different angles of insertion. The tissue was immediately fixed in 10% buffered formalin and embedded in paraffin. No serious complications of liver biopsy (bleeding, hospital admission, prolonged pain, or surgery) were encountered.

Histological Assessment of the Liver Biopsy Samples

Liver sections (n = 80) were evaluated by a hepatopathologist (Richard Williamson) blinded to the clinical data; more than 10 levels of tissue sections stained with hematoxylin and eosin or hematoxylin and Van Gieson's stain were used. For fibrosis scoring, the Scheuer F0-F4 staging system18, 19 was used (F0 = no fibrosis, F4 = cirrhosis). Only sections with at least five portal tracts were deemed adequate for assessment. Steatosis was noted to be absent or present. Fibrosis was also quantified by immunohistochemistry for α-smooth muscle actin (α-SMA), a marker of activated hepatic stellate cells and myofibroblasts (and thus fibrogenesis), as previously described in detail14 with 1:400 mouse anti–α-SMA (clone 1A4; Sigma) and by Aperio Spectrum imaging analysis of whole sections.

US Assessment

US images were obtained after fasting to induce gallbladder distension, using real-time scanners: Acuson Sequoia (Siemens Medical, Erlangen, Germany) with 2.5- to 4-MHz or 5.5- to 8.5-MHz probes or ATL HDI 5000 (Philips Medical Systems, Best, the Netherlands) with 2- to 5-MHz or 5- to 7-MHz probes. Sonographic images were reviewed, as previously described in detail,8 by a pediatric radiologist (Kieran Frawley) blinded to clinical and biopsy findings and previous interpretations. Briefly, liver images were recorded as nodular edge, nodular, heterogeneous, or normal echogenicity with or without splenomegaly. Normal US was defined as normal echogenicity with no splenomegaly. US evidence of PHT included a nodular liver with splenomegaly.

Statistical Analysis

Statistical analysis was conducted by Meagan J. Walsh and Ristan M. Greer with Stata IC 10 (StataCorp LP, College Station, TX). Normally distributed variables are depicted as means and standard errors of the mean, and non-Gaussian variables (e.g., the age and fibrosis stage) are presented as medians and ranges. Fisher's exact text or the chi-square goodness of fit was used to determine differences in fibrosis staging based on gender, forced expiratory volume (FEV), age, and steatosis. McNemar's test for paired samples was used to evaluate the effect of two biopsy passes in detecting fibrosis. Agreement between biopsy pairs was evaluated by weighted κ analysis: κ = 1 indicates perfect agreement, 1 〈κ〉 0.80 indicates almost perfect agreement, 0.80 〈κ〉 0.60 indicates substantial agreement, 0.60 〈κ〉 0.20 indicates moderate to fair agreement, 0.20 〈κ〉 0 indicates slight agreement, and κ = 0 indicates no agreement. Logistic regression and receiver operating characteristic (ROC) analysis were applied to each diagnostic modality individually and in combination to predict hepatic fibrosis and PHT; the positive predictive value (PPV) and the negative predictive value (NPV) were included. An area under the receiver operating characteristic curve (AUROC) >0.80 indicated potential diagnostic utility. Multivariate logistic regression, corrected for age, FEV at enrollment, treatment with Urso, the presence of steatosis, and the presence of diabetes mellitus, was performed to identify factors associated with PHT, the occurrence of which was evaluated with Kaplan-Meier statistics. A backward elimination approach was used to remove nonsignificant variables and to determine the most parsimonious model. A Cox proportional hazards model was used to determine factors independently associated with the time to the development of PHT. All statistical significance was taken at the 95% confidence interval.

Results

Patient Characteristics and Outcomes

The 40 children (24 females and 16 males) were 2.38 to 18.73 years old at enrollment (median age = 10.64 years). Most (96%) were Caucasian, 68% were Δf508 homozygotes, 20% had CFRD, 43% underwent gastrostomy for supplemental nutrition, and 35% had meconium ileus at birth. The median FEV1 value was 83.5%. At enrollment, 9 of 40 had evidence of PHT, as defined previously (Table 1). No patient was found to have or was suspected of having portal vein thrombosis.

Table 1. Patient Clinical and Demographic Variables at Study Enrollment and Follow-Up
  1. * The genotype was unavailable.

  2. † The patient also had end-stage liver disease.

inline image

During follow-up (up to 12 years; median = 9.5 years), seven (17.5%) died: five (12.5%) from respiratory failure (three also had end-stage liver disease), one (2.5%) from end-stage liver disease alone (on a transplant waiting list), and one from leukemia (2.5%). Three (7.5%) received a transplant (liver transplant, lung transplant, or heart and lung transplant). Another eight patients developed PHT, as defined previously, during follow-up (median age = 12.9 years). Seventeen of the 40 patients (including the 9 patients with PHT defined at enrollment) had PHT, which was present in the majority of those who died (6 of 7) or underwent transplantation (2 of 3).

Liver Histology by Dual-Pass Biopsy

Seventy-seven of the 80 biopsy specimens had at least 5 portal tracts allowing adequate assessment (range = 5-13). The 3 specimens deemed inadequate were from different patients, and the alternate core had F2 or F3; this allowed fibrosis staging to be reported in all 40 patients (Table 2): F0 (no fibrosis) in 9 (22.5%), F1 (mild fibrosis) in 10 (25%), F2 (moderate fibrosis) in 10 (25%), F3 (advanced fibrosis) in 9 (22.5%), and F4 (cirrhosis) in 2 (5%). Steatosis was evident in 28 of 40 (70%).

Table 2. Liver Histology and Detection of Fibrosis in Children With Suspected CFLD (n = 40) Subjected to Dual-Pass Liver Biopsy
 Final Fibrosis Stage [n (%)]Nonconcordance [n (%)]
  • The fibrosis staging was performed according to Scheuer et al.18 If each of the 77 adequate biopsy samples had been taken randomly, a diagnosis of fibrosis would have been missed in 22.5% of the cases. More stage F1-F4 patients were documented with two passes versus one pass (P = 0.002).

  • *

    Three cores with fewer than five portal tracts were not assessed.

Fibrosis stage
 09 (22.5)0 (0)
 110 (25)6 (60)
 210 (25)6 (60)
 39 (22.5)2 (22)
 42 (5)1 (50)
Total4015/37 (40)*

Dual-pass biopsy improved the detection of fibrosis (F1-F4): the first pass detected fibrosis in 26 patients, and the second detected fibrosis in another 5 patients (12.5%, P = 0.002). There was nonconcordance for the fibrosis stage in 14 of 37 biopsy pairs (38%; Table 2), although the agreement of stages F0 to F4 between the first and second passes was substantial (weighted κ = 0.61). Six children scored F0 and F1, three children scored F0 and F2, three children scored F1 and F2, one child scored F2 and F3, and one child scored F2 and F4. If each of the 77 adequate biopsy samples had been taken randomly, a diagnosis of fibrosis (F1-F4) would have been missed in 22.5% of the cases. In addition, if biopsy samples had been classified as F0-F1 (none/mild) or F2-F4 (significant fibrosis), there would have been six patients (16%) for whom a diagnosis of significant fibrosis would have been missed without the use of dual passes (P = 0.01).

With quantitative α-SMA immunohistochemistry, some level of immunoreactivity was observed in all 77 adequate biopsy samples, even in those with F0 fibrosis; the mean α-SMA level was 7.8% ± 0.9% of the total area. An increase in the fibrosis stage was associated with increasing α-SMA immunoreactivity in a nonlinear fashion (P < 0.001; Supporting Fig. 1).

Figure 1.

Common clinical modalities do not reliably predict the presence of hepatic fibrosis. Liver fibrosis in children with suspected CFLD was predicted with commonly used clinical modalities: HSM on clinical examination, elevated ALT levels, and abnormal liver US findings. Neither US alone (AUROC = 0.63, P = 0.14, sensitivity = 100%, specificity = 0%, PPV = 0.77, NPV = 0) nor US in combination with HSM and ALT (AUROC = 0.69, P = 0.19, sensitivity = 100%, specificity = 11%, PPV = 0.79, NPV = 1) significantly predicted the presence of fibrosis. Abbreviations: HSM, hepatosplenomegaly; LFT, liver function test.

Clinical Examination Findings, ALT, and US as Predictors of Hepatic Fibrosis

None of these routinely used clinical modalities, individually or in combination, significantly predicted liver fibrosis (Fig. 1). HM with or without splenomegaly, which was present at enrollment in 27 (67.5%), had a PPV of 0.77 and an NPV of 0.0 in detecting liver fibrosis (AUROC = 0.51, P = 0.77, sensitivity = 100%, specificity = 0%). One-third of patients with F0 fibrosis had HM (possibly due to fatty liver). Seven of 11 patients with splenomegaly had F3 or F4 fibrosis on biopsy. Interestingly, 4 of 11 with F3 or F4 fibrosis did not have splenomegaly. Elevated serum ALT measurements had very good specificity (100%) but poor sensitivity (0%) in detecting liver fibrosis (AUROC = 0.59, P = 0.3, PPV = 0.77, NPV = 0). Twenty-five percent of those with F3 or F4 fibrosis had a normal ALT level. US abnormalities were not predictive (AUROC = 0.63, P = 0.14); they had good sensitivity (100%) and PPV (0.77) but very poor specificity (0%) and NPV (0). Twenty percent of those with F1-F2 fibrosis had normal US findings, and five of nine with F0 fibrosis had heterogeneous echogenicity, which was interpreted as indeterminate (either steatosis or fibrosis). All except one patient with F3 or F4 fibrosis had some abnormality (eight with a nodular edge and three with heterogeneous echogenicity; data not shown). When they were combined by multivariate logistic regression, these three clinical tests showed marginally improved clinical utility in the detection of fibrosis (F1-F4; AUROC = 0.69, P = 0.19, sensitivity = 100%, PPV = 0.79) with very poor specificity (11%, NPV = 1; Fig. 1). A similar result was demonstrated for the detection of significant fibrosis (F2-F4; AUROC = 0.68, P = 0.2, sensitivity = 81%, PPV = 0.63) with some improvement in specificity (47%, NPV = 0.69; figure not shown).

Liver Fibrosis and Other Clinical Findings as Predictors of Outcome

Death and Transplantation.

Nine of 31 patients (29%) with fibrosis (F1-F4) had a serious clinical outcome (6 deaths and 3 transplants). No patients with F0 fibrosis died or received transplants (Table 1). Both children with cirrhosis (F4) on biopsy died, and so did 2 of 9 children with F3 fibrosis, 1 of 10 children with F2 fibrosis (3 transplants), and 1 of 10 with F1 fibrosis. Neither the presence of the Δf508 homozygous genotype nor the presence of CFRD at the time of enrollment was a significant predictor of mortality. Interestingly, 3 of 23 Δf508 homozygotes, 4 of 11 Δf508 heterozygotes, and 0 of 6 ungenotyped children died during the follow-up period. All three transplant patients were Δf508 heterozygotes. Three of the eight patients with CFRD at presentation died during follow-up.

PHT.

During this long-term follow-up study, 17 of 40 patients were diagnosed with or subsequently developed PHT, as defined in the Patients and Methods section: 9 at enrollment and 8 more during the study. The median age of onset was 13.3 years (range = 4.4-17.4 years). According to binary logistic regression, the only factor independently associated with PHT was the fibrosis stage (P < 0.001, odds ratio = 7.16). Figure 2A depicts the occurrence of PHT with respect to the age of onset and fibrosis stage on biopsy. Those children who developed PHT earlier were more likely to have more severe liver fibrosis (P < 0.001, hazard ratio = 3.9). Among those with stage F2-F4 fibrosis on biopsy at enrollment, 15 of 21 (71%) had or later developed PHT, whereas only 2 of 19 (10.5%) with stage F0-F1 fibrosis did. Only 1 of 9 patients with F0 fibrosis and only 1 of 10 patients with F1 fibrosis developed PHT (3.3 and 2.8 years after enrollment, respectively). Figure 2B depicts the development of PHT with respect to the fibrosis stage in those who did not have PHT at enrollment (i.e., the nine patients who already had PHT were excluded). Again, those with more severe fibrosis (F3-F4) at enrollment developed PHT more frequently (P < 0.002, hazard ratio = 3.4) with a trend toward an earlier age of development in comparison with those with F0 or F1 fibrosis (P < 0.14).

Figure 2.

Predicting the development of PHT by (A) age and (B) the time from liver biopsy in children referred for suspected CFLD. (A) Survival without PHT from birth was stratified by fibrosis staging on liver biopsy at enrollment in a cohort of 40 patients with suspected CFLD: stage F0 (no fibrosis) in 9 (22.5%), stage F1 (mild fibrosis) in 10 (25%), stage F2 (moderate fibrosis) in 10 (25%), stage F3 (advanced fibrosis) in 9 (22.5%), and stage F4 (cirrhosis) in 2 (5%). According to Cox proportional hazards analysis, an increase in the fibrosis stage on biopsy was significantly associated with an increase in the risk of developing PHT (P < 0.001, hazard ratio = 3.9) and with a trend toward a younger age of onset (P = 0.07). (B) Survival without PHT from the date of liver biopsy was stratified by fibrosis staging on dual-pass liver biopsy in a cohort of 31 patients who did not have PHT at study enrollment: stage F0 (no fibrosis) in 9 (29%), stage F1 (mild fibrosis) in 10 (32%), stage F2 (moderate fibrosis) in 8 (26%), stage F3 (advanced fibrosis) in 4 (13%), and stage F4 (cirrhosis) in 0 (0%). According to Cox proportional hazards analysis, an increase in the fibrosis stage on liver biopsy was associated with an increase in the risk of developing PHT (P = 0.002, hazard ratio = 3.4) and with a trend toward a younger age of onset (P = 0.14, hazard ratio = 1.16). The fibrosis staging was performed according to Scheuer et al.18

According to ROC analyses (Fig. 3), the degree of liver fibrosis on biopsy by fibrosis staging, α-SMA immunoreactivity, or their combination was significantly predictive of the development of PHT (for fibrosis staging, AUROC = 0.81, P = 0.0021, sensitivity = 50%, specificity = 100%, PPV = 1, NPV = 0.85; for quantitative α-SMA immunoreactivity, AUROC = 0.73, P = 0.024, sensitivity = 50%, specificity = 95.65%, PPV = 0.80, NPV = 0.85; for their combination, AUROC = 0.802, P = 0.0081, sensitivity = 50%, specificity = 95.65%, PPV = 0.8, NPV = 0.85).

Figure 3.

Hepatic fibrosis predicts PHT in patients with suspected CFLD (n = 40). According to logistic models, both the stage of hepatic fibrosis (AUROC = 0.81, P = 0.0021, sensitivity = 50%, specificity = 100%, PPV = 1, NPV = 0.85) and the quantitative α-SMA immunoreactivity (AUROC = 0.73, P = 0.024, sensitivity = 50%, specificity = 95.65%, PPV = 0.8, NPV = 0.85) according to liver biopsy at the time of diagnosis showed significant clinical utility in predicting the development of PHT.

No noninvasive clinical modality (HM, ALT, or US), either individually or in combination, was significantly predictive of the development of PHT (results not shown); splenomegaly, which is included in the definition of PHT, was excluded from the analysis (for HM, AUROC = 0.53, P = 0.76; for elevated ALT, AUROC = 0.54, P = 0.6; for abnormal US, AUROC = 0.59, P = 0.29; for their combination, AUROC = 0.66, P = 0.6). According to multivariate analyses, there was no significant relationship between Urso therapy, the Δf508 genotype, the presence of CFRD at enrollment, meconium ileus, or FEV1 and the occurrence of PHT (data not shown). Interestingly, 8 of 23 Δf508 homozygotes, 7 of 11 Δf508 heterozygotes, and 2 of 6 ungenotyped children had or developed PHT during the course of the study. Three of eight children with CFRD at enrollment had PHT, and another two children developed PHT during follow-up.

Discussion

In this well-characterized, prospectively followed cohort of children with CF presenting with common clinical test findings indicating possible liver disease, we have demonstrated the importance of detecting liver fibrosis by liver biopsy to predict the occurrence of PHT, the poor performance of nonbiopsy tests currently employed to detect or predict the development of clinically significant CFLD, and a high rate of future adverse outcomes. Although CFLD is the third leading cause of mortality in CF3 and accounts for less than 2% of directly caused deaths, the 17.5% mortality rate in this cohort, in which six of seven patients who died had PHT, might explain why significant liver diseases such as cirrhosis have a lower reported prevalence in adults versus children (2% versus 5%-10%).1-4 Such outcomes lend weight to the importance of detecting and managing clinically significant CFLD in childhood.

These data confirm and extend the results of previous studies,9 which found that clinical evaluation, serum aminotransferases, and US, despite continued widespread clinical practice,1, 5-7 are nonspecific for the detection of liver fibrosis and are imprecise for predicting the presence of or progression to PHT. Nevertheless, this study does gives support to the practice of performing these standard tests to screen for clinically significant liver disease in patients with CF because they identify a cohort of children with a 42% chance of developing significant liver disease (i.e., PHT) versus the expected clinical prevalence of 4%,20 which is comparable to our clinical prevalence of 7% (17 of 240 patients; unpublished data, 2010). It is also noted that US abnormalities become more specific for higher stages of fibrosis and cirrhosis.8 Importantly, however, only fibrosis staging on liver biopsy predicted the development of clinically significant CFLD (defined by the occurrence of PHT) and the serious outcomes of transplantation or mortality; this indicates that biopsy has a relevant predictive role in the care and evaluation of patients with suspected CFLD. If liver biopsy is the gold standard for the diagnosis of hepatobiliary fibrosis in CF, sampling error concerns may be assuaged by dual-pass biopsy.

Even though nine patients had clinical evidence of PHT at enrollment, only two had Scheuer stage 4 fibrosis, the histological criterion for established cirrhosis. All the remaining seven patients had evidence of significant fibrosis: five had advanced fibrosis with architectural collapse (Scheuer stage 3), and two had portal-portal linkage (Scheuer stage 2). PHT is usually a result of advanced fibrosis, but the need for established cirrhosis in every case remains contentious. Sampling error remains a significant issue for all biopsy procedures in any chronic liver disease with advanced fibrosis. On the basis of our findings, it is likely that the finding of Scheuer stage 2 or 3 fibrosis in patients with clinical PHT reflects a combination of both principles: established cirrhosis is not always required for the development of PHT, and some patients with established cirrhosis may have a biopsy sample interpreted as falling short of this staging. Thus, our conclusion that dual-pass liver biopsy improves the detection of significant fibrosis (F2-F4), with another 16% detected with two passes (P = 0.01), is an important observation from this study.

It is not surprising that dual-pass biopsy improved the sensitivity of fibrosis detection. Sample size and focal histological lesions have commonly presented challenges in many liver diseases. Various techniques, including dual passes, rejection of sections with fewer than five portal tracts, and quantitative histochemistry (all portrayed here), help to overcome these limitations.19 In the current study, a single biopsy core might have missed a diagnosis of fibrosis in 22% of the patients. Gaskin et al.21 reported discordance between percutaneous biopsy and open liver biopsy in 11 CFLD patients. Routinely obtaining two cores for the evaluation of suspected CFLD is therefore advised for clinical purposes because the second pass in our study detected additional patients (12.5%) whose fibrosis was missed by the first pass. However, the agreement of fibrosis stages between the first and second passes was substantial (weighted κ = 0.61), and this suggests that one pass is better than no biopsy at all. Some of these concerns may be overcome by the application of alternative and more quantitative histological methods, such as confirmatory α-SMA immunoreactivity, as shown in this study. Certainly, dual biopsy would have even more relevance in research studies for which a gold-standard point of reference is necessary (e.g., for the evaluation of noninvasive diagnostic modalities) or in therapeutic trials.

These data, in conjunction with our earlier comparison of US with liver histology,8 suggest that caution is warranted in interpreting US findings in patients with suspected CFLD, particularly in the absence of liver nodularity and splenomegaly. This is contrary to the conclusions of Lenaerts et al.,6 who did not evaluate liver histology or clinically significant outcomes such as PHT. It is widely recognized that US poorly differentiates between liver steatosis and fibrosis; this is evidenced by the finding of heterogeneous echogenicity on scans in patients with steatosis but no fibrosis. However, US may value add to monitoring for PHT once the presence of hepatic fibrosis is confirmed by liver biopsy or novel noninvasive means. New imaging methods that characterize hepatic fibrosis, such as diffusion-weighted magnetic resonance imaging and elastography,22-24 require further study in this regard. Serum measurements of fibrogenesis biomarkers may also be proven to be useful noninvasive adjuncts in predicting liver fibrosis or the development of PHT. Previously, using sera taken from this cohort of patients at enrollment, we found that collagen IV, prolyl hydroxylase, and tissue inhibitor of metalloproteinase 1 levels distinguished patients with biopsy-proven CFLD from those patients with CF but no liver disease and from healthy controls,15 and increased levels of prolyl hydroxylase, tissue inhibitor of metalloproteinase 1,15 and monocyte chemoattractant protein 113 distinguished earlier stage fibrosis from later stage fibrosis in children with CFLD. The evaluation of serial changes in serum marker patterns over time may be even more useful. The application of similar biomarker analyses to larger cohorts of patients with CFLD is warranted.

Importantly, liver histology findings are highly predictive of the occurrence of PHT. Although this was expected, this is the first study clearly demonstrating that biopsy-proven liver fibrosis leads to cirrhosis and PHT in patients with CFLD. There are several interesting observations in this regard. First, this study found that the subsequent development of PHT is associated with a higher stage of fibrosis on initial biopsy and a young age of onset (median age = 13 years). Second, other studies have suggested that progressive liver disease is uncommon in adult CF patients.25, 26 Finally, it is generally recognized that the predominant liver outcome of CFLD is PHT, and liver synthetic dysfunction is uncommon.1 Although this study did not set out to follow the epidemiology of hepatobiliary fibrosis in a CF population (there was inevitable enrollment bias), it does plot the progress of a group of CF patients referred with suspected CFLD who had variable degrees of liver fibrosis and places the value of standard investigational modalities and specifically liver biopsy in a clinical context. This study also highlights that those with liver fibrosis are in a high-risk group in terms of the development of PHT, the need for transplantation, or mortality from CF-related causes. These patients with suspected CFLD carried a cohort chance of a serious clinical endpoint of 25%, and the severity of histological fibrosis at the time of diagnosis placed these children in an ascending risk category.

We conclude from this rigorous prospective cohort study that CF patients with liver fibrosis have a significant risk of future morbidity and mortality, and clinical, biochemical, and US evaluations for CFLD without dual-pass biopsy are imprecise for the early diagnosis of liver fibrosis in CF. The early diagnosis of CFLD requires liver biopsy enhanced by dual-pass biopsy paired evaluation and aided by immunohistochemical analysis of α-SMA expression. Liver histology at the time of diagnosis strongly predicts clinically significant CFLD. Although doctors widely rely on US, US alone lacks specificity, particularly for early fibrosis. The use of histopathology and the evaluation of promising serum fibrosis marker panels deserve wider application and prospective systematic study (in conjunction with newer noninvasive imaging modalities) in larger cohorts of patients with CFLD.

Ancillary