Shear wave elastography in the assessment of liver fibrosis
Department of Gastroenterology & Hepatology
Document TypeMasters Research thesis
Access StatusThis item is embargoed and will be available on 2021-01-10.
© 2018 Dr. David Nadebaum
The accurate quantification of liver fibrosis is essential to the prognostication and clinical management of patients with chronic liver disease (CLD). Whilst liver biopsy remains the gold standard for fibrosis assessment, it has a number of limitations which have seen its use become increasingly substituted by non-invasive techniques. Ultrasound shear wave elastography (SWE) includes some of the most widely used non-invasive technologies in clinical practice. This work evaluates two ultrasound SWE devices which are in differing stages of clinical development and use; the first being a well-validated point SWE technique from Siemens called Acoustic Radiation Force Impulse elastography or ‘ARFI’ and the second a new 2D-SWE platform by Toshiba. The differing study aims for the two technologies were assessed in separate patient cohorts. Hence the thesis is divided in two. ARFI (Siemens): Background: Acoustic Radiation Force Impulse elastography or ‘ARFI’ is a point shear wave elastography (SWE) technique that is in broad clinical use for the quantification of liver fibrosis. Whilst well validated, questions remain for a number of areas of ARFI performance. This includes the magnitude and likely mechanism of obesity’s impact on ARFI performance, the impact of hepatosteatosis on ARFI reliability and whether ARFI performance is dependent on operator experience. There is also conflicting information as to whether ARFI liver stiffness measurements (LSMs) correlate with cirrhosis severity and the presence of cirrhotic complications. Finally, clinicians have limited facility to gauge the validity of obtained ARFI measurements beyond the IQR/Median criteria. An additional study aim was therefore to develop new strategies to aid ARFI reliability assessment; specifically whether inter-operator disagreement predicts the presence of unreliable ARFI measurements. Method: ARFI performance was assessed amongst a cohort of 943 patients with diffuse CLD of mixed aetiology, who had ARFI LSMs taken as part of clinical fibrosis assessment. Patients were scanned independently by either two or three operators, with ARFI results analysed in the context of patient demographic and CLD information obtained from medical records. Anthropometric measures including body mass index (BMI) was recorded at the time of scanning, and the distance from the skin surface to liver capsule (SLD) was measured from ARFI screenshots as a marker of central adiposity. The cumulative number of scans completed by individual operators and the institution overall was recorded. Assessed performance measures included IQR/Median and inter-operator agreement. ARFI accuracy was also assessed amongst a subcohort of 55 patients who had undergone a liver biopsy within 6 months of ARFI. The performance of ARFI in assessing cirrhosis severity was assessed amongst a further subcohort of 186 patients with clinically diagnosed cirrhosis. The presence of cirrhotic complications was determined retrospectively from medical records and endoscopy reports. Prognostic indices including Child Pugh and Model for End stage Liver Disease (MELD) scores were calculated using bloods tests where available. Results: ARFI showed modest accuracy in assessing liver fibrosis, demonstrating an AUROC of 0.67, 0.76 and 0.70 at discriminating the F01/ F2, F2/ F3 and F3/F4 cut-offs, respectively. ARFI showed good sensitivity (80.0 – 88.9%) and NPV (70.6 – 95.3%), but relatively poor specificity (42.9 – 66.3%) and PPV (27.9 – 56.2%) at the three cut-offs. Body habitus, particularly skin-to-liver capsule distance or ’SLD’, was found to be the primary determinant of ARFI performance in multi-regression analyses. SLD had the strongest relationship with ARFI accuracy (R2 = 0.543) followed by necroinflammatory change (R2 = 0.167), whilst all other patient factors, including hepatosteatosis, failed to show an independent association. Patients with a SLD >2.5cm (indicating significant central adiposity) showed particularly poor ARFI performance and was associated with higher IQR/Median ratios (median = 0.363 vs. 0.187, p<0.001), greater deviation between operators (29.8% vs. 15.9%, p<0.001) and poorer correlation with biopsy (rho = -0.242 vs. 0.493) than those with a SLD ≤2.5cm. Individual operator experience showed a weak relationship with ARFI performance, with operators of <25 scans experience having similar median IQR/Median ratios (0.170 vs. 0.165, p=0.13), slightly greater deviation between operators (14.3% vs. 11.06%, p=0.014) and greater deviation from the biopsy reference range (mean deviation = 0.588 vs. 0.279m/s, p=0.004) than more experienced colleagues. There also appeared to be a similarly weak association between overall institutional experience and ARFI performance, with reliability being slightly reduced amongst the first 150 scans performed in the institution. In patients in whom both operators had concordant F score results, ARFI LSM showed greater correlation with biopsy (rho = 0.392) than in cases of inter-operator disagreement (rho = 0.010). When scanned by three operators, patients with three-way operator agreement showed even stronger correlation with histopathology (rho = 0.571). Amongst cirrhotic patients, ARFI showed a moderately strong correlation with prognostic scores of liver function, including both MELD score (rho = 0.342, p<0.001) and Child- Pugh Score (rho = 0.363, p<0.001). ARFI LSMs showed modest accuracy in predicting the presence of ascites (AUROC = 0.58), encephalopathy (AUROC = 0.60) and oesophageal varices (AUROC = 0.69). Conclusion: ARFI showed moderate performance in quantifying liver fibrosis in a clinical Australian setting. The technology’s strength appears to be in the exclusion of liver fibrosis, however the tool is prone to false positive results. Body habitus was found to be the primary determinant of ARFI performance, with necroinflammatory change and operator experience showing a weaker impact on scan reliability. Central adiposity, as indicated by SLD, showed a particularly strong relationship with ARFI performance and the routine measurement and reporting of SLD should be considered to help clinicians gauge the reliability of ARFI results. Scanning patients with multiple independent operators also showed value as a reliability indicator, with inter-operator discordance being a predictor of poor ARFI performance. 2D-SWE (Toshiba): Background: The second technology assessed is a new 2D-SWE platform from Toshiba, which has a number of technical innovations and theoretical advantages over Siemens’ ARFI system. The technology is in the early clinical phases of testing and therefore data on this new technique remains limited. Our study aim was therefore to evaluate specific technical parameters to help assist in the formation of acquisition guidelines. This included assessing the measurement variability of Toshiba 2D-SWE (i.e. IQR/Median), the number of measurements required per patient to yield a precise LSM estimate and whether the uniformity of shear wave velocities within the measurement ROI (i.e. ROI SD/Speed ratio) could be used to assess the reliability of individual 2D-SWE measurements. Method: 2D-SWE was assessed amongst fifty-five patients with mixed aetiology CLD using the Toshiba Aplio 500 ultrasound system. Ten measurements were obtained per patient by an operator blinded to all preceding readings. Measurement variability (i.e. IQR/Median) and the number of measurements required per patient to achieve a LSM estimate within 5% of the existing method using 10 samples was assessed. Results were analysed against scan and clinical information including CLD aetiology, BMI, SLD, presence and severity of hepatosteatosis and measurement depth within the liver. The ratio of the standard deviation of shear wave velocities within the measurement ROI to overall shear wave velocity (i.e. ROI SD/Speed) was calculated for each individual measurement, and its relationship with measurement consistency (i.e. deviation of the measurement from the set’s median) was assessed. Results: The median IQR/Median ratio for 2D-SWE was 0.131 (q1-q3: 0.089–0.174). Five readings provided an approximation within 0.11m/s or 4.2% of the median velocity of ten measurements. Factors associated with increased measurement variability included increasing BMI (rho=0.388, p=0.003), SLD (rho=0.426, p=0.002) and measurements taken within 1.5cm of the liver capsule (p<0.001). Measurements with heterogeneous shear wave profiles (indicated by a ROI SD/Speed >0.15) showed greater deviation from the set’s median velocity than those with a ROI SD/Speed ≤0.15 (0.421 vs. 0.219 m/s, p=0.0001). Conclusion: 2D-SWE showed low overall measurement variability, with a minimum of five readings providing equivalent precision to the existing method using 10 samples. Obesity (i.e. BMI>30kg/m2), increasing abdominal wall thickness (i.e. SLD), sub-capsular measurements and a ROI SD/Speed >0.15 were all associated with increased measurement variability. ROI SD/Speed warrants further evaluation as a quality assessment metric, as it may allow objective operator assessment of individual 2D-SWE measurement reliability in real-time.
Keywordsliver; fibrosis; cirrhosis; elastography; shear wave; obesity
- Click on "Export Reference in RIS Format" and choose "open with... Endnote".
- Click on "Export Reference in RIS Format". Login to Refworks, go to References => Import References
- Medicine (RMH) - Theses