Accuracy of Liver Surface Nodularity Quantification on MDCT as a Noninvasive Biomarker for Staging Hepatic Fibrosis

Transcription

1 Gastrointestinal Imaging Original Research Pickhardt et al. MDCT of Hepatic Fibrosis Gastrointestinal Imaging Original Research Perry J. Pickhardt 1 Kyle Malecki John Kloke Meghan G. Lubner Pickhardt PJ, Malecki K, Kloke J, Lubner MG Keywords: biomarker, cirrhosis, CT, elastography, hepatic fibrosis DOI: /AJR Received April 1, 2016; accepted after revision May 26, Supported in part by the Clinical and Translational Science Award program through the National Institutes of Health (NIH) National Center for Advancing Translational Sciences (grant UL1TR000427). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. 1 All authors: Department of Radiology, University of Wisconsin School of Medicine and Public Health, E3/311 Clinical Science Center, 600 Highland Ave, Madison, WI Address correspondence to P. J. Pickhardt (ppickhardt2@uwhealth.org). AJR 2016; 207: X/16/ American Roentgen Ray Society Accuracy of Liver Surface Nodularity Quantification on MDCT as a Noninvasive Biomarker for Staging Hepatic Fibrosis OBJECTIVE. The purpose of this study was to investigate objective semiautomated measurement of liver surface nodularity on MDCT for prediction of underlying hepatic fibrosis (stages F0 F4). MATERIALS AND METHODS. Contrast-enhanced abdominal MDCT scans were assessed with an independently validated semiautomated surface nodularity tool. A series of 10 or more consecutive ROI measurements along the anterior aspect of the liver totaling a length of 80 cm or more were made with the left lateral segment as the default. All intermediate stages of fibrosis (F1 F3) were based on liver biopsy results within 1 year of MDCT. RESULTS. The study participants were 367 patients (191 men, 176 women; mean age, 51.1 years) divided into a healthy (F0) control group (n = 118) and patients with fibrosis in stages F1 (n = 47), F2 (n = 38), F3 (n = 67), and F4, which constituted cirrhosis (n = 97). MDCT-based liver surface nodularity scores increased with stage of fibrosis: F0, 2.01 ± 0.28; F1, 2.34 ± 0.39; F2, 2.37 ± 0.39; F3, 2.88 ± 0.68; and F4, 4.11 ± For discriminating significant fibrosis ( F2), advanced fibrosis ( F3), and cirrhosis ( F4), the ROC AUCs were 0.902, 0.932, and 0.959, respectively. The sensitivity and specificity for significant fibrosis ( F2; liver surface nodularity threshold, 2.38) were 80.2% and 80.0%, for advanced fibrosis ( F3; liver surface nodularity threshold, 2.53) were 89.0% and 84.2%, and for cirrhosis ( F4; liver surface nodularity threshold, 2.81) were 97.9% and 84.8%. CONCLUSION. Objective quantification of liver surface nodularity at MDCT allows accurate discrimination between stages of hepatic fibrosis, especially at more advanced levels. Although the results are comparable to those of elastography, this simple semiautomated biomarker can be measured retrospectively without additional equipment or patient time. C irrhosis can have a variety of underlying causes and is generally irreversible in the end stage. Earlier stages of hepatic fibrosis, however, may be reversible, making timely diagnosis and intervention extremely important. Although liver biopsy can be used to confirm the diagnosis of and to stage hepatic fibrosis, it has a number of limitations (e.g., invasiveness, high cost, subjectivity, and sampling error) [1, 2]. Thus, there has been great interest in noninvasive techniques for detection and staging of hepatic fibrosis [2, 3]. Noninvasive approaches to detecting hepatic fibrosis include laboratory and imaging tests. Serum-based markers are widely available but are not sufficiently accurate to replace histologic assessment [2 4]. As for imaging, elastography with ultrasound or MRI techniques for measuring liver stiffness as a proxy for fi- brosis have received the most attention [5 14]. Although elastography has proven to be effective, dedicated equipment is required and must be prospectively applied. In addition, the technical failure rate exceeds 5% [13], and overlap in stiffness exists between early fibrosis and inflammation (steatohepatitis) [15, 16]. Beyond invasive and prospective approaches, intrinsic cross-sectional imaging features that may reflect changes of underlying hepatic fibrosis have the advantage of being available retrospectively. For example, a variety of morphologic and volumetric changes in the liver and spleen can be assessed retrospectively on MDCT [17 22]. Most CT-based imaging biomarkers have been compared in cohorts with cirrhosis and healthy cohorts, making changes most evident, or have been focused primarily on viral causes. There is a clinical need for more robust noninvasive imaging biomarkers that can be 1194 AJR:207, December 2016

2 MDCT of Hepatic Fibrosis used to detect and stage hepatic fibrosis without the need for dedicated equipment or prospective evaluation. Increased liver surface nodularity associated with cirrhosis has been found to be a useful feature at cross-sectional imaging [23, 24], but subjective assessment and interobserver variability limit its utility. A semiautomated quantitative tool for deriving objective scores of liver surface nodularity on MDCT has been developed and validated [25]. The initial investigation showed this measurement to be highly reproducible and capable of differentiating cirrhotic from normal livers, but precirrhotic fibrosis was not evaluated. We hypothesized that the degree of liver surface nodularity may also allow one to differentiate intermediate stages of hepatic fibrosis. The purpose of this study was to investigate whether the CT tool for quantifying liver surface nodularity is accurate for staging hepatic fibrosis (stages F0 F4), including intermediate (precirrhotic) stages of fibrosis. Materials and Methods This HIPAA-compliant retrospective study was approved by our institutional review board. The requirement for signed informed consent was waived. Patient Population Patients were primarily categorized according to stage of hepatic fibrosis (METAVIR stages F0 F4) [26], ranging from no fibrosis (F0) through intermediate stages of hepatic fibrosis (F1 F3) to end-stage cirrhosis (F4). The primary inclusion criterion for this convenience cohort was a contrast-enhanced abdominal CT study available within our PACS. The patient population consisted of the following three discrete subcohorts: patients with chronic end-stage liver disease (cirrhosis) undergoing evaluation for liver transplant (F4); patients with varying degrees of precirrhotic hepatic fibrosis, including early (F1), intermediate (F2), and advanced (F3) disease; and a group without symptoms who had no known or suspected liver disease and were undergoing CT as potential kidney donors (F0). This patient cohort participated in a separate investigation of hepatosplenic volume changes [22]. Liver biopsy within 1 year of CT was required for all patients in the early (F1), intermediate (F2), and advanced (F3) fibrosis cohorts. Within the end-stage cirrhosis cohort (F4), A B C Fig. 1 Examples of liver surface nodularity quantification. Top row shows contrast-enhanced CT images; middle row, placement of liver surface nodularity ROI along anterior surface of left hepatic lobe; bottom row, magnified images. Liver surface nodularity scores (averaged from 10 or more measurements with cumulative length of at least 80 cm) increased for each fibrosis stage. A, 54-year-old woman with biopsy-proven stage F2 hepatic fibrosis (liver surface nodularity score, 2.37). B, 57-year-old woman with biopsy-proven stage F3 hepatic fibrosis (liver surface nodularity score, 3.03). C, 67-year-old woman with biopsy-proven stage F4 cirrhosis (liver surface nodularity score, 5.31). AJR:207, December

3 Pickhardt et al. TABLE 1: Summary Statistics for Liver Surface Nodularity Score According to Pathologic Stage of Hepatic Fibrosis Value the requirement for liver histologic analysis was waived by our hepatologists when the following conditions were met: clear cause of cirrhosis, clinical evidence of chronic end-stage liver disease or complications of portal hypertension, and clearcut imaging evidence of cirrhosis. Chart and image review confirmed that these conditions were met without biopsy for all patients with cirrhosis (n = 51). For the healthy control subjects (F0) without evidence of underlying liver disease, biopsy was generally not performed and not required. Among patients with hepatic fibrosis or cirrhosis, the most common causes of the underlying liver disease were chronic hepatitis C, alcoholism, and nonalcoholic fatty liver disease. Other conditions included primary sclerosing cholangitis, primary biliary cirrhosis, cryptogenic liver disease, hepatitis B virus infection, autoimmune hepatitis, and α 1 -antitrypsin deficiency. Some patients had more than one etiologic factor, such as both hepatitis C and alcoholism. MDCT Technique All CT scans were acquired with 16- or 64-MDCT scanners. Specific CT technique varied somewhat with patient cohort, but multiphasic protocols were used in most cases (i.e., triphasic for pretransplant evaluation, biphasic for precirrhosis liver evaluation, and multiphasic for renal donor evaluation). For liver surface nodularity measurement, the portal venous phase was used, and images were reconstructed with 5-mm slice thickness at 3-mm intervals. Specific tube voltage and current settings were based on patient size and study indication. Liver Surface Nodularity Quantification Liver surface nodularity scores were obtained with a dedicated semiautomated CT software tool that was independently validated for discriminating cirrhotic from normal livers [25]. The tool was provided by the inventor of this technology (Smith AD, University of Mississippi Medical Center) without industry support. This Microsoft Windows PCbased software tool loads and displays DICOM-format CT images in a viewer that allows the user to scroll, pan, zoom, and window the images as needed. Pathologic Fibrosis Stage F0 (n = 118) F1 (n = 47) F2 (n = 38) F3 (n = 67) F4 (n = 97) Mean SD Median Interquartile range The user digitally paints an ROI along the anterior edge of the liver surface, which the software automatically detects on the selected slice and adjacent contiguous slices (Fig. 1). A series of 10 or more consecutive ROI measurements along the anterior aspect of the liver, totaling a length of 80 cm or more, were made in each case, with the left lateral segment as the default. The tool calculates the distance between the detected liver edge and a smoothed polynomial line (spline) that is measured on a pixel-bypixel basis and measures the mean distance for each slice (expressed as tenths of a millimeter). The final liver surface nodularity score is the arithmetic mean of the measurements from each slice. Statistical Analysis All liver surface nodularity scores were obtained and reported with summary statistics (mean, SD, median, and interquartile range) and calculated separately for each patient cohort according to stage of hepatic fibrosis. A Kruskal- Wallis test was used to assess differences in liver surface nodularity score among the discrete F0 F4 cohorts. Emphasis was placed on the clinically relevant distinctions of significant hepatic fibrosis ( F2), advanced hepatic fibrosis ( F3), and cirrhosis (F4). ROC curves were obtained for each candidate metric, and AUC was calculated with DeLong 95% CI. Exploratory cutoffs for fibrosis categories were derived from ROC analysis. A value of p < 0.05 (two-sided) was the criterion for statistical significance. All statistical analyses were performed with the R program (version 3.2.2, R Core Team 2014). Cases with missing data or tool failure were excluded. Results The final cohort consisted of 367 patients (191 men, 176 women; mean age, 51.1 years) with the following distribution of hepatic fibrosis stages: F0, 118; F1, 47; F2, 38; F3, 67; and F4, 97. The mean Model for End-Stage Liver Disease score in the cohort with cirrhosis (F4) was ± 6.14 (SD). Summary statistics for liver surface nodularity scores are provided in Table 1. There was a progressive increase in mean liver surface nodularity score on MDCT according to fibrosis stage: 2.01 ± 0.28 for F0, 2.34 ± 0.39 for F1, 2.37 ± 0.40 for F2, 2.88 ± 0.68 for F3, and 4.11 ± 0.95 for F4. No patients were excluded owing to technical failure of the liver surface nodularity tool. The corresponding box plots of liver surface nodularity scores according to fibrosis stage are shown in Figure 2. There was little difference in mean liver surface nodularity score between fibrosis stages F1 and F2, but progressive increases were notable between all other ascending fibrosis levels. Examples of the liver surface nodularity tool applied to patients with varying degrees of hepatic fibrosis are shown in Figure 1. The diagnostic performance of the liver surface nodularity score for discriminating between stages of hepatic fibrosis is shown in Table 2. The most relevant ROC curves for significant ( F2) and advanced ( F3) fibrosis levels are shown in Figure 3. For detecting cases of significant fibrosis (F2 F4 versus F0 F1), the ROC AUC was (95% CI, ) with sensitivity of 80.2% and specificity of 80.0% at the liver surface nodularity score threshold of For the differentiation of advanced fibrosis (F3 F4 versus F0 F2), the ROC AUC for the liver surface nodularity score was (95% CI, ) with sensitivity of 89.0% and specificity of 84.2% at a liver surface nodularity score threshold of Finally, for detecting cases of cirrhosis (F4), the ROC AUC was (95% CI, ) with TABLE 2: Diagnostic Performance of Liver Nodularity Score for Predicting Pathologic Stage of Hepatic Fibrosis Fibrosis Score a ROC AUC Cutoff Value Sensitivity (%) Specificity (%) F0 vs F1 F ( ) F0 1 vs F2 F ( ) F0 2 vs F3 F ( ) F0 3 vs F ( ) Note Values in parentheses are 95% CI. a F0 F1 vs F2 4 refers to clinically significant fibrosis ( F2), and F0 F2 vs F3 F4 refers to advanced fibrosis ( F3) AJR:207, December 2016

4 MDCT of Hepatic Fibrosis Nodularity Score Fibrosis Stage sensitivity of 97.9% and specificity of 84.8% at a liver surface nodularity score threshold of Discussion Hepatic fibrosis can have a variety of underlying causes. Although liver damage is likely irreversible when end-stage cirrhosis ensues, earlier stages of fibrosis may be Sensitivity Specificity Fig. 2 Box plot of liver surface nodularity score according to hepatic fibrosis stage (F0 F4) shows progressive overall increase in nodularity with fibrosis stage, most notably for advanced fibrosis ( F3). Boxes represent middle 50% (interquartile range) of data. Line within box represents median of data. Whiskers extend to minimum and maximum values. Circles indicate outliers (defined as being farther from median than multiple of size of box or interquartile range). Outliers for F4 extend above depicted range. reversible in some cases. Accurate noninvasive methods for detecting, staging, and monitoring precirrhotic fibrosis could have a substantial positive impact and possibly reduce the need for liver biopsy [2, 3, 14]. Beyond the invasiveness and potential complications, additional limitations of liver biopsy include high cost and sampling error. A number of potentially useful biomarkers based AUC = ( ) Sensitivity on cross-sectional imaging findings have been proposed and investigated. Elastography performed with either ultrasound or MRI techniques has been extensively studied and validated for assessing and grading hepatic fibrosis according to parenchymal stiffness [14, 27]. Although both ultrasound and MR elastography have proven to be robust clinical techniques, the disadvantages of even MR elastography include the need for dedicated hardware and software, the need for prospective planning, a technical failure rate exceeding 5%, and the fact that inflammation can also increase liver stiffness (reducing specificity) [13, 14, 16, 27]. Intrinsic MDCT imaging biomarkers that correlate with the degree of hepatic fibrosis are attractive because such findings can be retrospectively derived and compared with other scans for interval change. Examples of such markers are hepatosplenic volumetric changes [17 22, 25] and liver surface nodularity [23 25]. Unlike studies of elastography, most investigations into these potential intrinsic CT biomarkers have compared only cohorts with and without cirrhosis, excluding the key element of intermediate fibrosis. In addition, many studies have included only patients with cirrhosis related to viral hepatitis, precluding any conclusions regarding generalizability. In particular, the initial val- AUC = ( ) Specificity A B Fig. 3 ROC curves for performance of liver surface nodularity score. A and B, ROC curves for differentiating significant ( F2) (A) and advanced ( F3) (B) fibrosis show good performance of liver surface nodularity score with AUC values of and Performance for differentiating cirrhosis ( F4) was even better with AUC value of (ROC curve not shown). AJR:207, December

5 Pickhardt et al. idation study of the quantitative liver surface nodularity scoring tool we used this study showed strong reproducible results but only compared patients with hepatitis C induced cirrhosis and healthy control subjects [25]. Our study shows that the simple, objective semiautomated measurement of liver surface nodularity on MDCT images allows accurate and generalizable staging of hepatic fibrosis. Importantly, this discrimination includes noninvasive detection of both significant ( F2) and advanced ( F3) levels of hepatic fibrosis, which are emerging as key clinical indicators that may influence management decisions. The diagnostic performance of liver surface nodularity quantification in our study compares favorably with the reported results for both ultrasound and MR elastography. For example, typical ROC AUC values of ultrasound-based transient and shear wave elastography for differentiating significant fibrosis ( F2) have ranged from to [14], and a meta-analysis of 12 MR elastography studies showed an AUC of [7]. The ROC AUC for liver surface nodularity quantification exceeded in our study for discriminating all stages of fibrosis (including no fibrosis), rising to for cirrhosis (F4). Diagnostic performance in terms of sensitivity and specificity at selected liver surface nodularity score thresholds also compares favorably with reported elastographic results. Combining the robust performance of liver surface nodularity quantification with prospective or retrospective application on MDCT suggests that this tool could play a useful role in patient care. Abdominal CT is a commonly performed examination for a variety of clinical indications, and many patients with known or suspected liver disease are likely to have more than one study on file. This allows not only detection and staging of hepatic fibrosis but also monitoring of disease status over time. Beyond clinical practice, this tool also has potential applications for large-scale research, taking advantage of the ubiquitous nature of CT. Furthermore, other CT biomarkers that are intrinsic to standard scanning, such as volumetric and textural features, could potentially be combined with liver surface nodularity quantification in a multiparametric manner. We intend to investigate this approach in the future. We also plan to compare liver surface nodularity quantification results directly against elastography for staging hepatic fibrosis. Finally, quantification of liver surface nodularity may also be feasible with other cross-sectional imaging techniques, including both MRI and highfrequency ultrasound. It is reasonable to question whether the default use of 5-mm sections and the anterior surface of the left lateral segment in this study might have had an impact on the liver surface nodularity measurements. The issue of CT section thickness was assessed in the initial validation study of use of this tool. In that study, Smith et al. [25] found that there was no gain in accuracy with use of thinner sections for differentiating cirrhotic from noncirrhotic livers. Further studies are underway to evaluate the impact of CT image thickness on the liver surface nodularity score by use of an imaging phantom. The decision to use the left hepatic lobe was initially made to reduce observational bias. By choosing a consistent anatomic region, the user is not biased to choose a portion of liver with the greatest perceived surface nodularity. For the tool to remain objective, it is important to eliminate such bias, which could degrade this quantitative metric. Further unpublished research findings obtained with this liver surface nodularity tool suggest that surface nodularity appears to be almost evenly distributed throughout the liver. In addition, this tool may be used on both unenhanced and contrast-enhanced CT series. We acknowledge limitations to our study. First, liver histologic results were not available in the subset of patients with cirrhosis (F4). However, the presence of end-stage fibrosis was not in doubt in this well-defined clinical cohort. In addition, it is conceivable that some of our healthy control subjects may have had early unsuspected liver disease, although the robust discrimination between the F0 control state and F1 F4 fibrosis argues against this being an important drawback. Although we believe that inclusion of all causes of hepatic fibrosis supports the generalizability of the liver surface nodularity tool, we cannot comment on its performance for individual etiologic factors, such as hepatitis C, and larger subcohorts may be required. Clearly, further investigation with larger cohorts that include patients with specific causes of chronic liver disease is indicated to see whether results vary according to etiologic factor. Conclusion Quantification of liver surface nodularity on MDCT can serve as a useful noninvasive biomarker for staging hepatic fibrosis and, unlike elastography or biopsy, can be performed retrospectively and compared across multiple CT studies. References 1. Afdhal NH, Nunes D. Evaluation of liver fibrosis: a concise review. Am J Gastroenterol 2004; 99: Martínez SM, Crespo G, Navasa M, Forns X. Noninvasive assessment of liver fibrosis. Hepatology 2011; 53: Nguyen D, Talwalkar JA. Noninvasive assessment of liver fibrosis. Hepatology 2011; 53: Holmberg SD, Lu M, Rupp LB, et al. Noninvasive serum fibrosis markers for screening and staging chronic hepatitis C virus patients in a large US cohort. Clin Infect Dis 2013; 57: Friedrich-Rust M, Nierhoff J, Lupsor M, et al. Performance of acoustic radiation force impulse imaging for the staging of liver fibrosis: a pooled meta-analysis. J Viral Hepat 2012; 19:e212 e Friedrich-Rust M, Ong MF, Martens S, et al. Performance of transient elastography for the staging of liver fibrosis: a meta-analysis. Gastroenterology 2008; 134: Singh S, Venkatesh SK, Wang Z, et al. Diagnostic performance of magnetic resonance elastography in staging liver fibrosis: a systematic review and meta-analysis of individual participant data. Clin Gastroenterol Hepatol 2015; 13: Talwalkar JA, Kurtz DM, Schoenleber SJ, West CP, Montori VM. Ultrasound-based transient elastography for the detection of hepatic fibrosis: systematic review and meta-analysis. Clin Gastroenterol Hepatol 2007; 5: Wang QB, Zhu H, Liu HL, Zhang B. Performance of magnetic resonance elastography and diffusionweighted imaging for the staging of hepatic fibrosis: a meta-analysis. Hepatology 2012; 56: Castéra L, Vergniol J, Foucher J, et al. Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 2005; 128: Foucher J, Chanteloup E, Vergniol J, et al. Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study. Gut 2006; 55: Yin M, Talwalkar JA, Glaser KJ, et al. Assessment of hepatic fibrosis with magnetic resonance elastography. Clin Gastroenterol Hepatol 2007; 5: Yin M, Glaser KJ, Talwalkar JA, et al. Elastography: clinical performance in a series of 1377 consecutive examinations. Radiology 2016; 278: Tang A, Cloutier G, Szeverenyi NM, Sirlin CB. Ultrasound elastography and MR elastography for assessing liver fibrosis. Part 2. Diagnostic performance, confounders, and future directions. AJR 2015; 205: AJR:207, December 2016

6 MDCT of Hepatic Fibrosis 15. Sagir A, Erhardt A, Schmitt M, Haussinger D. Transient elastography is unreliable for detection of cirrhosis in patients with acute liver damage. Hepatology 2008; 47: Chen J, Talwalkar JA, Yin M, Glaser KJ, Sanderson SO, Ehman RL. Early detection of nonalcoholic steatohepatitis in patients with nonalcoholic fatty liver disease by using MR elastography. Radiology 2011; 259: Zhou XP, Lu T, Wei YG, Chen XZ. Liver volume variation in patients with virus-induced cirrhosis: findings on MDCT. AJR 2007; 189:[web]W153 W Liu P, Li P, He W, Zhao LQ. Liver and spleen volume variations in patients with hepatic fibrosis. World J Gastroenterol 2009; 15: Li WX, Zhao XT, Chai WM, et al. Hepatitis B virus induced liver fibrosis and cirrhosis: the value of liver and spleen volumetry with multi-detector spiral computed tomography. J Dig Dis 2010; 11: Ozaki K, Matsui O, Kobayashi S, et al. Selective atrophy of the middle hepatic venous drainage area in hepatitis C related cirrhotic liver: morphometric study by using multidetector CT. Radiology 2010; 257: Chen XL, Chen TW, Zhang XM, et al. Quantitative assessment of the presence and severity of cirrhosis in patients with hepatitis B using right liver lobe volume and spleen size measured at magnetic resonance imaging. PLoS One 2014; 9:e Furusato Hunt OM, Lubner MG, Ziemlewicz TJ, Munoz Del Rio A, Pickhardt PJ. The liver segmental volume ratio for noninvasive detection of cirrhosis: comparison with established linear and volumetric measures. J Comput Assist Tomogr 2016; 40: Ferral H, Male R, Cardiel M, Munoz L, Ferrari FQ. Cirrhosis: diagnosis by liver surface analysis with high-frequency ultrasound. Gastrointest Radiol 1992; 17: Kudo M, Zheng RQ, Kim SR, et al. Diagnostic accuracy of imaging for liver cirrhosis compared to histologically proven liver cirrhosis. Intervirology 2008; 51: Smith AD, Branch CR, Zand K, et al. Liver surface nodularity quantification from routine CT images as a biomarker for detection and evaluation of cirrhosis. Radiology 2016; 18: Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C. Hepatology 1996; 24: Tang A, Cloutier G, Szeverenyi NM, Sirlin CB. Ultrasound elastography and MR elastography for assessing liver fibrosis. Part 1. Principles and techniques. AJR 2015; 205:22 32 FOR YOUR INFORMATION The comprehensive book based on the ARRS 2016 Annual Meeting categorical course on Oncologic Imaging: From Diagnosis to Cure is now available! For more information or to purchase a copy, see AJR:207, December