Am Fam Physician. 2003;68(9):1831-1834
Clinical Question
How likely is cirrhosis in a patient with hepatitis C?
Applying the Evidence: Richard is a 42-year-old man with chronic hepatitis C virus infection. He has no spider nevi, and laboratory test results are as follows: aspartate transaminase level, 32 U per L; γ-glutamyltransferase level, 92 U per L; total cholesterol level, 250 mg per dL (6.45 mmol per L); and platelet count, 240 × 103 per mm3 (240 × 109 per L). What is the probability that he has cirrhosis?
Answer: Referring to the accompanying table, Richard's risk of cirrhosis is only 1.8 percent. Applying his clinical data to equation 1 yields the following:
3.6 = 7.811 – [3.131 × ln (240)] + [0.781 × ln (92)] + [3.467 × ln (42)] – [0.014 × 250].
Because the score is less than 4.2, Richard has a low likelihood of having cirrhosis.
Evidence Summary
Hepatitis C is a growing health concern. An estimated 3.9 million persons in the United States are infected with hepatitis C virus; 2.7 million of these persons have chronic infection.1,2 Cirrhosis develops in approximately 7 percent of patients with chronic hepatitis C virus infection.1,2 Patients with significant fibrosis are at higher risk for progression to cirrhosis and are candidates for antiviral therapy. Patients with cirrhosis may benefit from screening for hepatocellular cancer, although the actual benefit of screening remains unproved.1
Three clinical decision rules estimate the likelihood of fibrosis or cirrhosis in patients with chronic hepatitis C virus infection. The estimated likelihood of these conditions can be a factor in deciding whether to order a liver biopsy and in prioritizing patients for screening for hepatic carcinoma and other complications of hepatitis C. For example, patients with a very low probability of fibrosis may choose to forego liver biopsy because they are unlikely to benefit from antiviral treatment. The three clinical decision rules were chosen because they were validated in a new set of patients (an important test of a clinical decision rule).
The first clinical decision rule predicts the likelihood of current cirrhosis in patients with hepatitis C.3 This rule is based on data from 264 consecutive patients with hepatitis C virus infection who were evaluated at a specialty center between 1989 and 1998. The diagnosis of cirrhosis was made by pathologists who were blinded to the patients' clinical data. The model was designed to predict the presence of probable or definite cirrhosis (grade 4 on a scale of 0 to 4). The decision model was validated in a group of 102 patients at another hospital; it performed well, as measured by an area under the receiver operating characteristic (ROC) curve of 0.935. This ROC curve value is consistent with an excellent ability to distinguish patients with cirrhosis from those without cirrhosis. The model requires only four simple pieces of data that are easily acquired in the primary care setting: presence or absence of spider nevi, gender, platelet count, and aspartate transaminase (AST) level. Because the exact equation to calculate the probability of cirrhosis is complicated, the researchers created a simplified table for bedside use (see accompanying table). For example, in a male patient with spider nevi, an AST level of 60 U per L, and a platelet count of 180 × 103 per mm3 (180 × 109 per L), the estimated probability of cirrhosis is 89 percent.
| Laboratory test results | Probability of cirrhosis in male patients, % (95% CI) | Probability of cirrhosis in female patients, % (95% CI) | |
|---|---|---|---|
| Platelet count < 140 × 103 per mm3 (140 × 109 per L) | |||
| AST ≤40 U per L | |||
| Spider nevi | 96.6 (83.0 to 99.0) | 35.0 (9.0 to 74.0) | |
| No spider nevi | 56.0 (24.0 to 84.0) | 2.4 (0.5 to 10.6) | |
| AST > 40 U per L | |||
| Spider nevi | 99.8 (98.7 to 100) | 92.0 (71.0 to 98.0) | |
| No spider nevi | 96.4 (88.0 to 99.0) | 34.0 (18.0 to 54.0) | |
| Platelet count 140 × 103 per mm3 | |||
| AST ≤40 U per L | |||
| Spider nevi | 29.0 (9.0 to 63.0) | 0.8 (0.1 to 6.1) | |
| No spider nevi | 1.8 (0.4 to 7.2) | 0.03 (0 to 0.04) | |
| AST >40 U per L | |||
| Spider nevi | 89.0 (70.0 to 97.0) | 14.0 (3.0 to 45.0) | |
| No spider nevi | 27.0 (18.0 to 40.0) | 0.7 (0.2 to 3.1) | |
The second clinical decision rule is designed to identify patients with a low probability of fibrosis. The model was proposed by a group of researchers in France, based on their study of 476 consecutive patients with untreated hepatitis C.4 In this study, 351 patients were used to develop the decision rule, and 125 patients were used to validate it. The researchers excluded patients who had persistently normal alanine transaminase levels, current regular alcohol intake, morbid obesity, or coinfection with human immunodeficiency virus or hepatitis B virus. The reference standard was an ultrasound-guided liver biopsy, which was performed in all 476 patients. A fibrosis grade of 0 or 1 (scale of 0 to 4) was considered “absence of fibrosis.” Using equation 1, the score is calculated as follows (ln = natural logarithm; GGT = γ-glutamyltransferase):
Reading the Numbers: The receiver operating characteristic (ROC) curve is a way to measure the overall accuracy of a test. It plots sensitivity against 1 minus the specificity. An area under the ROC curve (AUROCC) of 0.5 means that the test is no better than a coin toss in discriminating between patients with and without a disease, while an AUROCC of 1.0 means that the test is perfect. As a general rule, an AUROCC from 0.9 to 1.0 describes an excellent test, one from 0.8 to 0.9 a good test, and one from 0.7 to 0.8 a fair test.
Score = 7.811 – [3.131 × ln (platelet count, in 109 per L)] + [0.781 × ln (GGT, in IU per L)] + [3.467 × ln (age, in years)] – [0.014 × cholesterol, in mg per dL]
Significant hepatic fibrosis is not present in 96 percent of patients with a score lower than 4.21. In contrast, 66 percent of patients with a score higher than 6.90 have significant fibrosis (defined as stage 2 through 4). In the validation group, the area under the ROC curve was 0.81, which is consistent with a good ability to discriminate patients with fibrosis from those without fibrosis. This decision rule does not have a bedside model; therefore, it must be programmed into a calculator or computer.
The third clinical decision rule is based on a study of 192 consecutive untreated patients with chronic hepatitis C who underwent liver biopsy. The rule was validated in another 78 patients at a university medical center.5 The reference standard was histopathology from the liver biopsy, with significant fibrosis (presence of bridging fibrosis) defined as 3 or more points on a 6-point scale and cirrhosis defined as a score of 5 or 6 points. Using the first group of 192 patients, the developers of the clinical decision rule found that the ratio of AST level to platelet count was a good predictor of both significant fibrosis and cirrhosis. Equation 2:
Ratio = [AST level, in U per L, as a multiple of the upper limit of normal/platelet count, in 109 per L] × 100
The researchers then used the ROC curve to identify the optimal cutoff for each prediction and applied this cutoff to the validation group of 78 patients. A ratio less than or equal to 0.5 made significant fibrosis unlikely (negative predictive value: 86 percent), while a ratio greater than 1.5 made significant fibrosis likely (positive predictive value: 88 percent). Similarly, 98 percent of patients with a ratio less than or equal to 1.0 did not have cirrhosis.
The major limitation of the three clinical decision guides is that they have not been validated in the primary care or community setting. Thus, they may overestimate the likelihood of the development of cirrhosis over a 20-year period in community-based patients. Most patients with fibrosis do not develop cirrhosis, and patients with persistently normal hepatic transaminase levels were unlikely to be included in the studies. Finally, a low likelihood of having fibrosis or cirrhosis does not mean that the patient should not undergo a liver biopsy. The decision to perform or forego biopsy should be based on a variety of patient and disease factors, as well as the patient's values and preferences.1