icon_star paper   Hepatitis C Articles (HCV)  
Back grey arrow rt.gif
 
 
Noninvasive Diagnosis of Hepatic Fibrosis in Patients With Chronic Hepatitis C by Splenic Doppler Impedance Index
 
 
  Clinical Gastroenterology and Hepatology
Volume 5, Issue 10, Pages 1199-1206.e1 (October 2007)
 
Chen-Hua Liu_, Shih-Jer Hsu, Jou-Wei Lin, Juey-Jen Hwang, Chun-Jen Liu_, Pei-Ming Yang_, Ming-Yang Lai_, Pei-Jer Chen_, Jun-Herng Chen, Jia-Horng Kao_, Ding-Shinn Chen_
_ Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan
Department of Internal Medicine, National Taiwan University Hospital, Yun-Lin Branch, Yun-Lin County, Taiwan
Department of Pathology, National Taiwan University Hospital, Yun-Lin Branch, Yun-Lin County, Taiwan
Department of Medical Research, National Taiwan University College of Medicine, Taipei, Taiwan
Hepatitis Research Center, National Taiwan University College of Medicine, Taipei, Taiwan
 
Supported by grants from the National Taiwan University Hospital, the National Science Council, and Department of Health, Executive Yuan, Taiwan.
 
ABSTRACT
Background & Aims:
The value of Doppler ultrasonography to evaluate the severity of hepatic fibrosis in patients with chronic hepatitis C (CHC) remains controversial.
 
Methods: Consecutive histologically proven patients with CHC over a 4-year period were divided into training (n = 335) and validation (n = 168) sets. Hepatic Doppler impedance index, splenic Doppler impedance index, and mean portal vein velocity were evaluated for all patients before liver biopsies. Multivariate logistic regression was performed to find the independent factors to predict patients with significant fibrosis (≥F2) and cirrhosis (F4) in the training set. Receiver operating characteristic curves were constructed for these factors to evaluate the diagnostic accuracy of significant hepatic fibrosis and cirrhosis in the training set, and in the validation set to evaluate the reproducibility.
 
Results: Multivariate logistic regression revealed that the splenic arterial pulsatility index (SAPI) and the mean portal vein velocity were predictive of significant fibrosis (≥F2) and cirrhosis (F4). Receiver operating characteristic analysis showed the areas under the curves of regression models and SAPI were comparable in predicting significant fibrosis (0.88 vs 0.87, P = .22) and cirrhosis (0.92 vs 0.90, P = .12) in the training set. Areas under the curves of SAPI were 0.89 and 0.92 in predicting significant hepatic fibrosis and cirrhosis in the validation set. By choosing optimized cut-off levels, 54% and 76% of the patients with significant hepatic fibrosis and cirrhosis could be predicted correctly.
 
Conclusions: SAPI is accurate and reproducible for assessing the severity of hepatic fibrosis in patients with CHC. Applying this simple Doppler index can decrease the need for staging liver biopsy.
 
Discussion
Gray-scale ultrasonography has been used as a tool to detect cirrhosis based on the presence of nodular hepatic surface, coarse echotexture, narrowing of intrahepatic vessels, and increased spleen size index. However, the value of these parameters to discriminate hepatic fibrosis from cirrhosis remains limited.40, 41, 42 With the introduction of DDU, it has improved our knowledge in understanding the hemodynamic changes at different stages of chronic liver diseases by measuring hepatic and splenic Doppler indices. Previous studies have shown that PVV, HARI, HAPI, SARI, and SAPI were correlated with cirrhosis. To overcome the controversial results in previous studies,34, 35, 36, 37 we prospectively evaluated 503 patients with CHC harboring various stages of hepatic fibrosis who received Doppler evaluation in 2 consecutive cohorts, and tried to find an accurate and validated Doppler index to predict the severity of hepatic fibrosis. In this study, we found that SAPI and PVVmean could predict significant fibrosis and cirrhosis in CHC patients. Furthermore, SAPI showed comparable diagnostic accuracy to the regression models and excellent diagnostic concordance in the training and validation sets, implying its clinical usefulness to predict the severity of hepatic fibrosis.
 
The splanchnic impedance indices evaluated the arterial blood flow velocity during the systolic and diastolic cycles. Although systolic arterial blood velocity is determined by the cardiac systole, diastolic blood velocity is affected mainly by the outflow resistance of the corresponding organs.32 With the advance of hepatic fibrosis, the portal resistance increases, causing the increased outflow resistance of the spleen artery. The decreased diastolic velocity of the splenic artery results in increased splenic impedance indices (SARI and SAPI). SAPI, which uses the mean arterial velocity instead of the peak arterial velocity, makes it more sensitive than SARI in detecting the waveform changes on increased portal resistance.32 Although SARI and SAPI were correlated highly in our study, our results consistently showed the superior value of SAPI over SARI in the diagnosis of significant hepatic fibrosis and cirrhosis. Previous studies have shown that hepatic impedance indices (HARI and HAPI) were correlated with the severity of hepatic fibrosis, based on the assumption of distortion of hepatic architecture and reduction of intrahepatic vascular space.29, 30, 31 However, the measurement of the hepatic impedance indices was more difficult than that of splenic impedance indices because of hepatic fatty change and the spatial approximation of the hepatic artery to the portal vein. In addition, hepatic impedance indices also increase with age in healthy subjects, making the indices less reliable in detecting the severity of hepatic fibrosis.43 In our study, the failure rate to detect the hepatic impedance indices was much higher than that to detect the splenic impedance indices (15.3% vs 0.6%), and therefore limited its clinical usefulness in predicting the severity of hepatic fibrosis.
 
It is conceivable that the PVV decreases with the increase of intrahepatic resistance to portal flow. Several studies have shown that patients with cirrhosis or advanced fibrosis had decreased PVV compared with those with mild fibrosis.26, 27, 28 Our study also confirmed the importance of PVV in diagnosing patients with significant hepatic fibrosis and cirrhosis. However, when we compared the overall diagnostic accuracy of the regression models including both SAPI and PVVmean with either SAPI or PVVmean, PVVmean added little to SAPI to predict the severity of hepatic fibrosis.
 
In our study, SAPI was accurate in predicting both the significant hepatic fibrosis and cirrhosis, with AUCs of 0.87 and 0.90, respectively, in the training set. Furthermore, SAPI also showed excellent reproducibility in the validation set, with AUCs of 0.89 and 0.92 in predicting significant hepatic fibrosis and cirrhosis, respectively. Further analysis of different cut-off levels of SAPI in the training and validation sets showed low variability in the overall diagnostic accuracy, implying its potential reliability through different cohorts. SAPI can be measured easily within minutes by ultrasonographic machines equipped with automatic tracing and calculation, and can be measured concomitantly during routine ultrasonographic screening. For predicting cirrhosis, SAPIs set at 1.40 and 1.20 had a high positive predictive value (90%) and negative predictive value (97%), and about 76% of the study population can avoid liver biopsies by choosing the 2 cut-off levels. For predicting significant hepatic fibrosis, SAPI set at 1.10 and 0.85 had a high positive predictive value (98%) but a modest negative predictive value (76%), and 54% of the study population can thus avoid liver biopsies. More than 90% of our patients who were diagnosed erroneously as F less than 2 by SAPI were staged histologically as F = 2. The suboptimal diagnostic power of SAPI to exclude the presence of significant hepatic fibrosis may be explained by the subtle change of portal resistance in mild to moderate hepatic fibrosis, or the sampling as well as interpretation variability of liver biopsies. In patients with SAPI of less than 0.85, another complementary noninvasive test, such as Fibroscan (EchoSens, Paris, France) or aspartate transminase to platelet ratio index,20, 22 should be performed to improve diagnostic accuracy.
 
Although liver biopsy is considered the gold standard for assessing fibrosis, it usually is flawed by sampling error and interobserver and intraobserver variability. A biopsy specimen of 1.5 cm, which contains at least 6-8 portal triads, is considered adequate for interpretation of diffuse liver diseases.44 All biopsies in our study were performed by cutting needles to obtain specimens of 18-20 mm in length without fragmentation to minimize sampling errors. In addition, one experienced pathologist interpreted all the specimens to avoid interobserver variability and to minimize intraobserver variability.
 
In the study, SAPI was shown to be the best parameter, among various Doppler indices in evaluating the severity of hepatic fibrosis. However, some limitations existed. First, interobserver and interequipment variability of arterial Doppler impedance indices exist, which may make it difficult to obtain reproducible results. Two experienced sonographers with excellent intraobserver and interobserver correlation performed DDU examinations with automatic DDU tracing and calculation for all the patients. In addition, nonexperienced sonographers can minimize the interobserver variability to nonsignificant levels by a common training program.45 Second, SAPI may not be measured in some patients because of splenectomy. However, only 0.6% of our patients had undetectable SAPI because of splenectomy. Third, the study was based on the assumption that liver biopsy is the gold standard for the assessment of hepatic fibrosis. Fourth, although logistic regression models have been used widely and accepted in clinical research, other statistical methods, such as neural nets or support vector machines, may generate equal or better performance than logistic regression.46 Further research aimed at comparing the diagnostic accuracy among different heuristic tools is needed.
 
In conclusion, SAPI is accurate and reproducible to predict significant hepatic fibrosis and cirrhosis in patients with CHC. By using optimized cut-off levels, a large proportion of patients with significant fibrosis or cirrhosis can be diagnosed correctly without liver biopsies. Further studies are needed to validate the usefulness of SAPI in other cohorts of patients.
 
BACKGROUND
Chronic hepatitis C virus infection is a global health problem, affecting about 3% of the worldfs population.1 Compared with chronic hepatitis C (CHC) patients with mild hepatic fibrosis, those with significant hepatic fibrosis are at higher risk of progression to cirrhosis over a 10- to 20-year period, suggesting the need of early antiviral therapy to prevent the development of cirrhosis and associated complications.2, 3 For patients with cirrhosis, surveillance endoscopy and ultrasound for gastroesophageal varices and hepatocellular carcinoma are needed to minimize morbidity and mortality.4, 5 Furthermore, reduced treatment response and tolerability to antiviral therapy may be encountered in cirrhotic patients.6 An accurate assessment of the severity of hepatic fibrosis is therefore important for both diagnostic and therapeutic purposes.
 
Liver biopsy has been recognized as the gold standard for assessing the grade of necroinflammation and stage of fibrosis. However, it is costly and harbors risks of complications, with rates as follows: 20% patient discomfort, 0.1%-3% significant morbidity, and 0.02%-0.24% mortality.7, 8, 9, 10, 11 In addition, sampling error caused by the nonuniform distribution of the parenchymal damage, as well as intraobserver and interobserver variability often are encountered.12, 13, 14 A noninvasive tool to evaluate liver disease activity or fibrosis stage would be helpful, particularly in monitoring CHC patients over time.
 
Noninvasive methods to evaluate the hepatic histology in hepatitis C virus-infected patients include symptoms and signs, routine laboratory tests, serum markers of fibrosis and inflammation, quantitative tests of liver function, and radiologic imaging.15 Previous studies have assessed the usefulness of noninvasive tests in predicting hepatic fibrosis.16, 17, 18, 19, 20, 21, 22, 23, 24, 25 However, most of them did not show satisfactory results, either owing to lack of accuracy, accessibility, reproducibility, or because they were expensive.
 
Duplex Doppler ultrasonography (DDU), which is noninvasive and readily available, has been used to assess the splanchnic vascular hemodynamics in patients with chronic liver disease. Portal vein velocity (PVV) has been shown to be correlated with significant hepatic fibrosis or cirrhosis.26, 27, 28 Hepatic artery resistive index (HARI) and hepatic artery pulsatility index (HAPI) were associated with portal vein resistance, hepatic vein-portal vein pressure gradient, and severity of fibrosis.29, 30, 31 Splenic artery resistive index (SARI) and splenic artery pulsatility index (SAPI) were associated with portal vein resistance, significant hepatic fibrosis, cirrhosis, and grade of esophageal varices.32, 33, 34 However, the value of DDU in evaluating liver histology remains controversial, probably because of the small number of patients, diverse grading of liver histology, and large number of missing data.34, 35, 36, 37
 
In this study, we included a large number of patients with histologically proven CHC and aimed to evaluate the value of DDU in predicting significant hepatic fibrosis and cirrhosis.
 
Materials and Methods
Patients

From January 2003 to December 2006, a total of 565 consecutive patients with CHC were evaluated prospectively at the gastroenterological clinics of the National Taiwan University Hospital and the National Taiwan University Hospital Yun-Lin Branch. CHC was defined as the presence of anti-hepatitis C virus and hepatitis C virus RNA for more than 6 months. Patients who were co-infected with human immunodeficiency virus or hepatitis B virus had a history of heavy alcohol use or other causes of liver diseases, declined percutaneous liver biopsy, or who were contraindicated for percutaneous liver biopsy were excluded from the study. DDU was performed for all eligible patients before liver biopsies. After excluding 62 patients, 503 eligible patients were divided into 2 groups: consecutive patients from January 2003 to June 2005 (n = 335) constituted the training set to search for predictors of significant hepatic fibrosis (≥F2) and cirrhosis (F4) by the METAVIR system,38 whereas patients from July 2005 to December 2006 (n = 168) constituted the validation set (Figure 1). The study was performed in accordance with the principles of the Declaration of Helsinki. The study was approved by the Ethical Committee of the National Taiwan University Hospital, and all patients gave informed consent before enrollment.
 
Demographic and Biochemical Data
 
Basic demographic data were recorded for all patients. Hemogram data, basic biochemical data, coagulation profiles, serologic data, and virologic data were collected before liver biopsies.
 
Duplex Doppler Ultrasonography
 
All patients were studied in the morning with an overnight fasting. A Toshiba Aplio (Toshiba Co. Ltd., Tokyo, Japan) with color Doppler and a 3.75-MHz phased-array curved electronic probe was used. The patients were kept in the supine position and suspended normal respiration. The probe initially was placed at the right intercostal space to find the right portal vein and right branch of the hepatic artery. The mean velocity of right portal vein (PVVmean) was measured automatically with time-averaged velocity in 2-3 cardiac cycles by placing the sampling cursor within the right portal vein with angle correction (Figure 2A). Peak systolic, diastolic, and mean velocity of the right branch of the hepatic artery were determined automatically by the machine. The HARI and HAPI of the hepatic artery were calculated according to the following formula (Figure 2B)29:
 
The SARI and SAPI were measured according to the formula as described earlier by placing the sampling cursor in the main branches of the intrasplenic artery near the splenic hilum at the left intercostal space (Figure 2C).32
 
All examinations were performed by 2 sonographers. To evaluate the intraobserver and interobserver variations of the Doppler measurements, 25 patients received independent Doppler examinations by the 2 sonographers on 2 consecutive days.
 
Liver Histology Evaluation
 
After written informed consent was given, each patient without contraindication for liver biopsies received an echo-guided percutaneous liver biopsy from the right hepatic lobe by using a 16-gauge Temno Evolution biopsy needle (Allegiance, McGaw Park, IL). The mean length of the liver specimen was 19 ± 1 mm and the diameter was 1.4 mm. No fragmentation of the liver specimen was found. The sampling tissues were fixed with formalin, embedded with paraffin, and stained with H&E and reticulin silver (Masson trichome method). The staging of hepatic fibrosis was scored by the METAVIR system, which ranged from F0 to F4.38 Significant hepatic fibrosis and cirrhosis are defined as a fibrosis stage of F2 or greater and F4, respectively. The samples were evaluated by one experienced pathologist who was blinded to the clinical status of study subjects.
 
Statistics
 
Statistical analyses were performed using SPSS 11.0 for windows (SPSS Inc., Chicago, IL). Patient characteristics were expressed as mean ± SD and percentage as appropriate. The intraobserver and interobserver variations of Doppler measurements were evaluated by the Pearson correlation coefficient and coefficient of variation. Univariate and multivariate logistic regression with the Wald test for PVVmean, HARI, HAPI, SARI, and SAPI were performed to find the independent factors in predicting patients with significant hepatic fibrosis and cirrhosis. Receiver operator characteristic (ROC) curves were constructed for regression models and reduced models. To evaluate the diagnostic accuracy (DA) in predicting significant fibrosis and cirrhosis, areas under the curves (AUCs) of individual tests with 95% confidence intervals (CIs) were calculated and compared.39 All statistical tests were 2-tailed and the results were considered statistically significant when the P value was less than .05.
 
Results
 
Patients

The clinical characteristics of eligible patients in the training and validation sets are summarized in Table 1. All patients with a fibrosis stage of F4 were in Child-Pugh A class. When performing DDU, 59 patients (11.7%) had undetectable hepatic arterial signals owing to anatomic variation or moderate to severe hepatic steatosis, 18 (3.6%) had both undetectable hepatic arterial and portal venous signals owing to severe hepatic steatosis, and 3 (0.6%) had undetectable splenic arterial signals owing to splenectomy. None of these patients had both undetectable hepatosplenic Doppler signals.
 
Intraobserver and Interobserver Variations of Doppler Measurements
 
Considering intraobserver variation, the Pearson correlation coefficient was 0.86 (P < .01) for PVVmean, 0.88 (P < .01) for HARI, 0.85 (P = .01) for HAPI, 0.93 (P < .01) for SARI, and 0.88 (P < .01) for SAPI. In addition, the coefficient of variation was 7.1% ± 4.4% (range, 4%-13%) for PVVmean, 4.6% ± 2.9% (range, 2%-10%) for HARI, 6.7% ± 5.3% (range, 2%-13%) for HAPI, 2.7% ± 1.8% (range, 2%-6%) for SARI, and 4.1% ± 2.5% (range, 1%-10%) for SAPI. Considering interobserver variation, the Pearson correlation coefficient was 0.82 (P = .02) for PVVmean, 0.86 (P < .01) for HARI, 0.84 (P = .01) for HAPI, 0.90 (P < .01) for SARI, and 0.85 (P = .01) for SAPI. In addition, the coefficient of variation was 6.1% ± 3.6% (range, 3%-12%) for PVVmean, 3.1% ± 1.9% (range, 1%-7%) for HARI, 6.3% ± 4.9% (range, 3%-11%) for HAPI, 2.2% ± 1.6% (range, 1%-5%) for SARI, and 3.4% ± 2.2% (range, 1%-8%) for SAPI.
 
Prediction of Significant Hepatic Fibrosis and Cirrhosis by Duplex Doppler Ultrasonography
 
Appendix 1 (see supplementary material online at www.cgh.journal.org) shows the Doppler indices in patients with different stages of hepatic fibrosis. PVVmean decreased with the advance of hepatic fibrosis. In contrast, the hepatic and splenic impedance indices increased with the advance of hepatic fibrosis. Table 2 shows the univariate and multivariate logistic regression analyses of the Doppler indices to predict patients with significant hepatic fibrosis and cirrhosis in the training set. In univariate analysis, all of the Doppler indices were associated significantly with significant hepatic fibrosis and cirrhosis. In multivariate logistic regression analysis, both SAPI and PVVmean were independent factors in predicting patients with significant hepatic fibrosis (SAPI: odds ratio [OR], 1.10; 95%CI, 1.07-1.13; P < .001; PVVmean: OR, 0.82; 95%CI, 0.75-0.90; P < .001), and cirrhosis (SAPI: OR, 1.08; 95%CI, 1.05-1.10; P < .001; PVVmean: OR, 0.75; 95%CI, 0.66-0.86; P < .001). The regression models in predicting patients with significant hepatic fibrosis (≥F2) and cirrhosis (F4) were shown as follows:
 

logit-1.gif

Table 2. Univariate and Multivariate Logistic Regression Analyses With Wald Tests of Various Doppler Indices in Predicting Patients With Significant Hepatic Fibrosis and Cirrhosis in the Training Set
 

Mult-2.gif

Diagnostic Accuracy of Significant Hepatic Fibrosis and Cirrhosis by Regression Models and Reduced Models
 
As the regression models indicated, the predicted probability of patients having significant hepatic fibrosis and cirrhosis was a function of increased SAPI and decreased PVVmean, which can be used as a new continuous variable. By choosing different cut-off points, ROC curves thus can be constructed. Although SAPI and PVVmean were predictive of significant fibrosis and cirrhosis in patients with CHC, the regression models needed cumbersome calculations and thus were not applicable for clinical routines. For simplicity, we used SAPI or PVVmean as the reduced models and compared them with regression models for diagnostic accuracy by ROC curves analysis.
 
Figure 3, Figure 4 show the ROC curves of regression models, SAPI, and PVVmean for patients with significant hepatic fibrosis and cirrhosis in the training and validation sets. The AUCs of regression models and SAPI showed comparable DA in predicting significant fibrosis and cirrhosis, and showed significantly higher DA than those of PVVmean in predicting significant hepatic fibrosis and cirrhosis.
 

roc-3.gif

Figure 4. ROC curves of regression models, SAPI, and PVVmean in predicting patients with significant hepatic fibrosis (≥F2) and cirrhosis (F4) in the validation set. (A) AUCs of the regression model and SAPI were comparable in predicting patients with significant hepatic fibrosis (0.89 vs 0.89, P = .37). In addition, AUCs of the regression model and SAPI were higher than PVVmean in predicting patients with significant hepatic fibrosis (0.89 vs 0.74, P < .001, and 0.89 vs 0.74, P < .001, respectively). (B) AUCs of the regression model and SAPI were comparable in predicting patients with cirrhosis (0.93 vs 0.92, P = .30). In addition, AUCs of regression model and SAPI were higher than PVVmean in predicting patients with significant hepatic fibrosis (0.93 vs 0.78, P < .001, and 0.92 vs 0.78, P = .004, respectively).
 

AB-4.gif

Sensitivity, Specificity, and Diagnostic Accuracy of the Splenic Arterial Pulsatility Index in Predicting Significant Hepatic Fibrosis and Cirrhosis
 
Appendix 2 (see supplementary material online at www.cgh.journal.org) shows the sensitivity, specificity, and DA of SAPI using the selective cut-off levels in predicting significant hepatic fibrosis and cirrhosis in the training set, validation set, and the overall study population. The overall DA ranged from 73% to 79% in predicting significant hepatic fibrosis when the SAPI was set from 0.85 to 1.10, whereas the overall DA ranged from 83% to 90% in predicting cirrhosis when the SAPI was set from 1.20 to 1.40. In addition, the differences of DA for significant hepatic fibrosis and cirrhosis between the training and validation sets were 0%-3% and 0%-4%, respectively, implying the concordance of diagnostic power by SAPI. When the SAPI was set at 1.10, the specificity and the positive predictive value were 98% and 98%, respectively; when the SAPI was set at 0.85, the sensitivity and the negative predictive value were 94% and 76%, respectively, in predicting significant hepatic fibrosis When the SAPI was set at 1.40, the specificity and the positive predictive value were 99% and 90%, respectively; when the SAPI was set at 1.20, the sensitivity and negative predictive value were 88% and 97%, respectively, in predicting cirrhosis. By these selected cut-off levels, 54% (270 patients) and 76% (381 patients) of the study subjects could be identified correctly as significant hepatic fibrosis or cirrhosis without further invasive liver biopsies.
 
 
 
 
  icon_paper_stack View Older Articles   Back to Top   www.natap.org