| |
The Use of Transient Elastometry for Assessing Liver Fibrosis in Patients with HIV and Hepatitis C Virus Coinfection: only detects cirrhosis
|
| |
| |
Clinical Infectious Diseases Sept 2007
Salvador Vergara,1 Juan Macias,1 Antonio Rivero,5 Alicia Gutierrez-Valencia,3 Mercedes Gonzalez-Serrano,6 Dolores Merino,7 Ma Jose Rios,4 Jose Angel Garcia-Garcia,1 Angela Camacho,5 Luis Lopez-Cortes,3 Josefa Ruiz,6 Julian de la Torre,5 Pompeyo Viciana,3 and Juan A. Pineda2; for the Grupo para el Estudio de las Hepatitis Viricas de la SAEI
1Clinical Unit of Infectious Diseases, Department of Internal Medicine, and 2Clinical Unit of Infectious Diseases, Hospital Universitario de Valme, 3Department of Infectious Diseases, Hospital Universitario Virgen del Rocio, and 4Clinical Unit of Infectious Diseases, Department of Internal Medicine, Hospital Universitario Virgen Macarena, Seville, 5Clinical Unit of Infectious Diseases, Hospital Universita io Reina Sofia, Cordoba, 6Infectious Diseases Unit, Department of Internal Medicine, Hospital Universitario Virgen de la Victoria, Malaga, and 7Department of Internal Medicine, Hospital Juan Ramon Jimenez, Huelva, Spain
"....In the present study, we validated TE to discriminate significant fibrosis and cirrhosis in HIV-HCV-coinfected patients. We selected cutoff values with high diagnostic performance that were previously elaborated in HCV-monoinfected patients. We found that the diagnostic accuracy of TE was very high for detecting cirrhosis and lower for significant fibrosis. Specifically, the performance of TE was low for discrimination of mild fibrosis (stages 0 and 1) from significant fibrosis (stage >2), which may partly limit the applicability of this technique in clinical practice.....The predictive accuracy of TE (as measured by AUROC) for detecting significant fibrosis in the present study was within the range of previous reports involving HCV-monoinfected patients [11-13]. However, the performance in the HIV-HCV-coinfected population was not as good as that among HIV-uninfected patients...."
ABSTRACT
Background. Transient elastometry (TE) is accurate for detecting significant liver fibrosis and cirrhosis in hepatitis C virus (HCV)-monoinfected patients. However, this procedure has been insufficiently validated in patients with human immunodeficiency virus (HIV) and HCV coinfection. The purpose of this study was to validate reported cutoff values of TE that discriminate significant liver fibrosis and cirrhosis in HIV-HCV-coinfected subjects.
Methods. Liver stiffness measurements were obtained for 169 HIV-HCV-coinfected adult patients who had undergone a liver biopsy or who had received a nonhistologic diagnosis of cirrhosis within 12 months before or after a liver stiffness measurement. Patients had received no prior therapy for HCV infection.
Results. TE measurements ranged from 3.6 kPa to 75 kPa. The area under the receiver operating characteristic curve was 0.87 (95% confidence interval, 0.84-0.93) for significant liver fibrosis and 0.95 (95% confidence interval, 0.92-0.99) for cirrhosis. To diagnose significant liver fibrosis, a cutoff value of 7.2 kPa was associated with a positive predictive value of 88% and a negative predictive value of 75%. Thirty-four patients (20%) were misclassified when this cutoff value was used. Thirteen (24%) of 54 patients with liver stiffness values <7.2 kPa had significant liver fibrosis detected by liver biopsy. To diagnose cirrhosis, a cutoff value of 14.6 kPa was associated with a positive predictive value of 86% and a negative predictive value of 94%. Thus, 13 patients (10%) had disease that was misclassified using this cutoff value.
Conclusions. We found that the diagnostic accuracy of TE was high for detecting cirrhosis and good for diagnosis of significant liver fibrosis. However, the performance of TE was low for discriminating mild fibrosis from significant liver fibrosis, which might limit the applicability of this technique in clinical practice.
Since the first decade of widespread use of HAART, the natural history of HIV infection has changed radically. On the one hand, AIDS-related mortality has decreased dramatically [1]. On the other hand, liver disease has emerged as one of the leading causes of morbidity and mortality in patients with HIV and hepatitis C virus (HCV) coinfection [2, 3]. Therefore, the management of HCV-related liver disease has become a major challenge in the context of HIV-HCV coinfection [4].
Currently, liver biopsy is considered to be the gold standard for assessing significant liver fibrosis in the context of HCV-related liver disease [5]. Nevertheless, this is an invasive procedure that could lead to infrequent but serious complications [6, 7]. To undergo a liver biopsy, a patient must be admitted to the hospital, resulting in an increase in health costs [8]. Moreover, this method has significant diagnostic limitations. Liver biopsy is known to be associated with significant interobserver variability [9], and its accuracy for assessing liver fibrosis importantly depends on the size of the specimen [10]. Because of these reasons, new noninvasive methods for evaluating liver fibrosis have emerged over the past few years.
Transient elastometry (TE) is a novel and noninvasive procedure that assesses liver fibrosis by measuring liver stiffness (LS). LS measurement has shown a high accuracy for detecting liver fibrosis in HCV-infected patients [11-13]. Thus, cutoff values of 14.6 kPa and 17.6 kPa have been reported to be reliable in ruling out cirrhosis in HCV-monoinfected patients [11-13]. Nonetheless, there is contradictory data on the ability of TE to discriminate mild from significant liver fibrosis [11-13]. In addition, this procedure has been insufficiently validated in the context of HIV-HCV coinfection [14]. Because of these reasons, we validated the diagnostic performance of previously reported TE cutoff values that discriminate cirrhosis from significant liver fibrosis in HIV-HCV-coinfected patients.
DISCUSSION
In the present study, we validated TE to discriminate significant fibrosis and cirrhosis in HIV-HCV-coinfected patients. We selected cutoff values with high diagnostic performance that were previously elaborated in HCV-monoinfected patients. We found that the diagnostic accuracy of TE was very high for detecting cirrhosis and lower for significant fibrosis. Specifically, the performance of TE was low for discrimination of mild fibrosis (stages 0 and 1) from significant fibrosis (stage >2), which may partly limit the applicability of this technique in clinical practice.
The accuracy of TE in discriminating cirrhosis in patients with HIV infection and chronic hepatitis C is high. This good diagnostic accuracy concurs with studies involving patients with chronic hepatitis C who did not have HIV infection [11-13] and patients with chronic liver diseases of diverse etiologies [13]. Accordingly, a previous study involving a smaller sample of HIV-HCV-coinfected patients also yielded similar results [14]. This was not unexpected. The amount of fibrosis does not increase linearly from one stage to the next [20]. Indeed, fibrosis quantification by digital imaging found similar amounts for specimens exhibiting low stages. The fibrosis area increases clearly in samples showing some bridging or more advanced stages, but there was only a significant difference in the area of liver affected by fibrosis between patients with cirrhosis and those without cirrhosis [20]. This explains the sharp increase in LS measurements in cirrhotic patients, which aids accurate discrimination of cirrhosis by TE.
The predictive accuracy of TE (as measured by AUROC) for detecting significant fibrosis in the present study was within the range of previous reports involving HCV-monoinfected patients [11-13]. However, the performance in the HIV-HCV-coinfected population was not as good as that among HIV-uninfected patients. Thus, the LS cutoff value to discriminate significant fibrosis was highly specific in HIV-uninfected patients, although it yielded a low sensitivity. To the contrary, the LS cutoff value of 7.2 kPa was associated with a specificity of 64%-66% and a sensitivity of 87%-88% in the present study. The most likely reason for this performance of TE is the strategy used to elaborate the cutoff values in studies involving HIV-uninfected patients. In 2 of these studies, the cutoff values were chosen to maximize the sum of sensitivity and specificity [11,12], whereas in a third report, the strategy was to favor specificity [13]. We did not aim to elaborate an additional cutoff value, but we validated for HIV-HCV-coinfected patients a value that had been elaborated by other researchers for patients without HIV infection. To our knowledge, this cutoff value had not been validated for HIV-infected patients before our study. However, this cutoff value is increasingly being assumed as diagnostic of significant fibrosis in studies involving HIV-HCV-coinfected patients [21]. The potential for misdiagnosis associated with using this cutoff value is relatively high, because 17% of the patients classified as having significant fibrosis had a liver biopsy specimen showing a fibrosis stage of 1 or 0. If the cutoff value is used to detect a fibrosis stage of 1 or 0, the rate of misclassification is even higher. One-fourth of the patients with an LS measurement <7.2 kPa had significant fibrosis.
The present study has some limitations. Patients included in the study may not be representative of the HIV-HCV-coinfected population. On the one hand, patients who undergo liver biopsy are selected according to particular characteristics. They usually adhere to clinical visits and therapy and are abstinent from illicit drugs and alcohol. These patients normally have less advanced HIV infection and less alcohol-related liver disease. On the other hand, the inclusion of patients with clinical cirrhosis may have biased the study to patients with advanced liver disease. However, patients with clinically overt cirrhosis often do not require liver biopsy. This has led to these patients not being included in studies comparing TE with histological examination. This is the reason for the higher AUROC for cirrhosis in the present study, compared with previous studies [11-14]. However, the inclusion of patients with mild and advanced fibrosis, as in the present study, allows an evaluation of the performance of TE within a wide range of liver disease. Moreover, these 2 extremes of liver fibrosis are clinically relevant. Thus, the diagnosis of mild liver fibrosis is usually used to guide the decision of treating HCV infection, and the diagnosis of cirrhosis has profound management and prognostic implications.
Disagreement between liver biopsy and TE results could be explained by the limitations of evaluating fibrosis by liver biopsy. In the present study, however, the results were not improved using higher-quality liver biopsy specimens as reference for the diagnosis of fibrosis. There are 2 possible explanations for this. First, the length of liver biopsy specimen that is needed to achieve an optimal sample size for the staging of fibrosis is very long [10] and is seldom found in clinical practice [22]. The patients grouped as having high-quality liver biopsy specimens for the present study had long liver biopsy specimens with numerous portal tracts, but even liver biopsy specimens with a length of 15-30 mm are still associated with variability in the assessment of fibrosis by pathologists [10]. Secondly, the performance of TE is possibly close to the maximum that can be obtained with the technique. The high diagnostic yield of TE for detecting cirrhosis is difficult to improve. On the other hand, the values of LS overlap for fibrosis stages 0, 1, and 2. Thus, even in very selected and large liver biopsy samples, it would be difficult for TE to perform better in discriminating significant fibrosis. This concurs with the fact that liver biopsy is more useful for diagnosing the extreme (stages 0 and 4) than the intermediate grades of fibrosis (stages 1-3) [23].
TE is useful for diagnosing and ruling out cirrhosis in HIV-HCV-coinfected patients. Most patients with discordant TE and liver biopsy results had stage 2 or 3 fibrosis. The limitations of liver biopsy are well-known and could explain these disagreements [9, 10]. On the contrary, TE has some limitations in discriminating mild from significant fibrosis. Panels of experts in the management of HIV-HCV-coinfection recommend postponing treatment with pegylated IFN plus ribavirin in patients with genotype 1 or 4 and a fibrosis stage of 1 or 0, as determined by examination of the liver biopsy specimen [4]. In this regard, if TE were used to decide to defer treatment of HCV infection, one-fourth of the patients classified by LS measurements as having a fibrosis stage of 1 or 0 but classified as having a fibrosis stage of >2 according to examination of a liver biopsy specimen would not receive therapy. On the other hand, TE could be applied to screen hard-to-treat patients. The technique showed good diagnostic accuracy in detecting significant fibrosis. In addition, an error in deciding to treat a small amount of patients with LS values >7.2 kPa and a fibrosis stage of 1 or 0, particularly when most of them have stage 1 fibrosis, is of relatively little clinical importance.
In conclusion, TE is a tool that can be applied to diagnose cirrhosis in HIV-HCV-coinfected patients. In addition, TE can help to select hard-to-treat patients with significant fibrosis who could be candidates for anti-HCV therapy. However, patients with LS values <7.2 kPa still need to undergo additional evaluations (e.g., evaluation of serum markers of fibrosis or liver biopsy) to direct anti-HCV therapy takes liver fibrosis into account.
RESULTS
Characteristics of the study population. One hundred seventy-one patients fulfilled the inclusion criteria. Two patients (1.2%) were excluded because of unsuccessful measurement of LS (because of ascites in both patients). One hundred sixty-nine patients were finally enrolled in the study. Of these 169 patients, 132 (78%) were recruited from 3 of the 6 participating centers. The main characteristics of the study population are summarized in table 1. Eighty-one patients (48%) showed advanced fibrosis. Forty-six patients (27%) received a nonhistologic diagnosis of cirrhosis.
Relation between LS and fibrosis. LS measurements ranged from 3.6 kPa to 75 kPa, with a median value of 9.7 kPa (Q1-Q3, 6.4-33.3 kPa). According to liver fibrosis staging, the distribution of median LS measurements was as follows: 5.2 kPa (Q1-Q3, 4.4-6.9 kPa) for stage 0, 6.3 kPa (Q1-Q3, 4.9-7.9 kPa) for stage 1, 7.6 kPa (Q1-Q3, 6.4-12.5 kPa) for stage 2, 9.9 kPa (Q1-Q3, 7.9-14.9 kPa) for stage 3, and 40.3 kPa (Q1-Q3, 21.2-58.3 kPa) for stage 4 (figure 1). LS significantly and positively correlated with the degree of liver fibrosis (-, 0.64; P < .001).
Figure 1. Relationship between the stage of fibrosis (according to the Knodell histological activity index modified by Scheuer [15]) and liver stiffness (determined by transient elastometry), as determined by the Kruskal-Wallis test. The median values (first and third quartiles) by fibrosis stage are represented by the box plots.
Diagnostic performance of LS cutoff values. The AUROC of LS measurements for significant fibrosis, advanced fibrosis, and cirrhosis in the whole study population, the patients who received reliable liver fibrosis assessment, and in the patients who underwent liver biopsy are shown in table 2. The diagnostic accuracy of cutoff values previously elaborated for detecting significant fibrosis and cirrhosis in HCV-monoinfected patients are shown in table 3.
Using a cutoff value of 7.2 kPa to discriminate significant fibrosis, 34 (20%) of 169 patients had disease that was misclassified. Thirteen (24%) of 54 patients with an LS measurement <7.2 kPa had significant fibrosis (12 patients [92%] had stage 2 fibrosis, and 1 patient [8%] had stage 3). On the other hand, 20 (17%) of 115 patients who had an LS measurement 7.2 kPa had a fibrosis stage of 1 or 0 (18 patients [90%] had stage 1 fibrosis, and 2 patients [10%] had stage 0).
Five (5%) of 98 patients with an LS measurement <14.6 kPa had cirrhosis, and 12 (17%) of 71 patients who had an LS measurement >14.6 kPa had a fibrosis stage of 3 or less (4 patients [33%] had stage 3, 4 patients [33%] had stage 2, and 4 patients [33%] had stage 1 or 0). Using a cutoff value of 17.6 kPa for discriminating cirrhosis, 20 patients (12%) had disease that was misclassified. Ten (9%) of 106 patients with an LS measurement <17.6 kPa had cirrhosis, and 9 (14%) of 63 patients for whom the LS measurement was >17.6 kPa had a fibrosis stage of 3 or less (3 patients [33%] had stage 3, 2 patients [22%] had stage 2, and 4 patients [44%] had stage 1 or 0). When restricted to high-quality samples, the analysis yielded similar results for discriminating significant fibrosis and cirrhosis (table 3).
Table 2. Area under the receiver operating characteristic curve (AUROC) of liver stiffness among the whole study population, patients who received reliable liver fibrosis assessment, and patients who underwent liver biopsy, by stage of fibrosis.
NOTE. Fibrosis was staged on the basis of the Knodell histological activity index (modified by Scheuer [15]), with a stage of >2 indicating significant fibrosis, a stage of >3 indicating advanced fibrosis, and stage 4 indicating cirrhosis.
Table 3. Validity criteria of optimal cutoff values of liver stiffness (LS) for discriminating significant fibrosis and cirrhosis.
PATIENTS AND METHODS
Study design and patients. This was a cross-sectional study that was performed in 6 tertiary care hospitals in Andalusia (southern Spain). From December 2005 through November 2006, we obtained an LS measurement for all HIV-HCV-coinfected patients who (1) were aged >18 years, (2) underwent a liver biopsy to assess the severity of liver disease or received a nonhistologic diagnosis of cirrhosis within 12 months before or after an LS measurement, and (3) did not receive therapy for HCV infection during the time between the liver biopsy and LS measurement. For patients who had received a nonhistologic diagnosis of cirrhosis, a fibrosis stage of 4, according to the Knodell histological activity index (modified by Scheuer [15]), was assumed.
Nonhistologic diagnosis of cirrhosis. Nonhistologic diagnosis of cirrhosis was made on the basis of clinical, biological, and/or imaging criteria. Thus, patients who had had at least 1 episode of ascites, portal hypertensive gastrointestinal bleeding, hepatic encephalopathy, nonobstructive jaundice, spontaneous bacterial peritonitis, or hepatocellular carcinoma, in the absence of any other likely nonhepatic causes, received diagnoses of decompensated cirrhosis. On the other hand, patients who showed 1 clinical and/or biological sign of cirrhosis, in addition to meeting imaging criteria-provided that the sign or signs were not explained by any other nonhepatic causes-were also considered to have cirrhosis, even though they had not developed any decompensation due to cirrhosis. Abdominal wall vascular collaterals, gynecomastia, hepatomegaly, jaundice, palmar erythema, vascular spiders, and splenomegaly were considered to be clinical signs of cirrhosis [16]. Thrombocytopenia, prolonged serum prothrombin time, hyperbilirubinemia, elevated asparate aminotransferase and alanine aminotransferase levels, decreased serum albumin level, and hyperglobulinemia were considered to be biological criteria [17]. Imaging data indicative of cirrhosis included the finding of a small, nodular, and irregular liver with increased echogenicity and/or a doppler flow significantly decreased in the portal circulation on an ultrasound, CT, or MRI [18].
LS measurement. LS measurements were obtained using FibroScan (Echosens) [19]. Measurements were performed on the right lobe of the liver. For this, patients were placed in supine decubitus position with the right arm in abduction. The probe of the system was applied between the ribs. The operator located the measurement area by A-mode images provided by the probe transducer. A minimum of 10 measurements were required. The median value was assumed to be representative of LS. A set of measurements was considered to be reliable if the success rate was >60% and the interquartile range was less than one-third of the median LS value. Unreliable measurements of LS were excluded. All measurements were obtained by 4 trained operators using a single device.
Liver histology. Most patients underwent a liver biopsy to evaluate liver disease and establish the indication of HCV treatment. After written consent was obtained from these patients, percutaneous liver biopsies were performed under either ultrasonography or CT guidance. After processing the samples, liver fibrosis was scored in accordance with the Knodell histological activity index (modified by Scheuer [15]). According to this classification, fibrosis is categorized into 5 stages: absence of fibrosis (stage 0); portal fibrosis without septa (stage 1); portal fibrosis and few septa (stage 2); numerous septa with architectural distortion, without cirrhosis (stage 3); and cirrhosis (stage 4). The length of liver biopsy specimens (expressed in millimeters) and the number of portal tracts contained in each specimen were also recorded. Liver biopsy specimens were evaluated by a single experienced pathologist at each center. Liver biopsy specimens from the 3 centers that included the most patients in the study were reevaluated by a central pathologist blinded to the clinical data. A total of 102 (83%) of the 123 liver biopsy specimens were reevaluated. Good agreement was found between the central pathologist and the pathologist from each center in diagnosing both significant fibrosis ( = 0.78-0.84) and cirrhosis ( = 0.85-0.91).
Statistical analysis. The categorical variables were expressed as number of patients (percentage), and the continuous variables were expressed as median (first and third quartiles, Q1-Q3). The relationship between the stage of fibrosis and the LS measurement was assessed using the Kruskal-Wallis test, and the correlation between both variables was evaluated using Kendall's - correlation coefficient.
The predictive accuracy of LS measurements was tested by calculating the areas under the receiver operating characteristic curves (AUROC). The cutoff points evaluated were previously elaborated for HIV-uninfected patients. The selected threshold values were 14.6 kPa and 17.6 kPa for the diagnosis of cirrhosis and 7.2 kPa for the diagnosis of significant fibrosis. Diagnostic accuracy was calculated by sensitivity, specificity, positive predictive value, and negative predictive value. Patients with significant fibrosis (stage >2), advanced fibrosis (stage >3), or cirrhosis (stage 4) were considered to have the disease. Performance of TE was also assessed using ROC curves for different subpopulations of patients, classified according to the quality of liver biopsy specimens and to liver fibrosis stage shown in the histological analysis. High-quality liver biopsy specimens were defined as samples >15 mm in length or with >7 portal tracts. Patients from whom high-quality liver biopsy specimens were obtained and who had received a nonhistologic diagnosis of cirrhosis were included in the group of patients who received reliable liver fibrosis assessment. Statistical analysis was performed using SPSS, version 14 (SPSS).
Ethical aspects. The study was performed in accordance with the Helsinki Declaration and was approved by the ethics committee at each participating center.
|
|
| |
| |
|
|
|
