Development and validation of a model to diagnose cirrhosis in patients with hepatitis C.

Publication/Presentation Date

10-1-2002

Abstract

OBJECTIVE: Although noninvasive markers predictive of cirrhosis in patients with chronic hepatitis C have been examined, none has proved sufficiently accurate for clinical use. The aim of this study was to develop an accurate model that can be easily used by clinicians to predict the probability of cirrhosis in hepatitis C patients from readily available clinical and laboratory information.

METHODS: We identified 264 consecutive patients with established chronic hepatitis C infection and extracted multiple physical examination and biochemical variables (recorded before liver biopsy). Similar data were extracted from charts at another hospital.

RESULTS: Logistic regression identified the following independent predictors of cirrhosis: platelet count < or = 140,000/ mm3, spider nevi, AST > 40 IU/L, and male gender. Male and female patients with normal values for platelet count and AST and no spider nevi had low probabilities of cirrhosis: 1.8% (95% CI = 0.4-7) and 0.03% (95% CI = 0.003-0.04), respectively. Male patients with abnormal values on all three other predictors had a probability of cirrhosis of 99.8% (95% CI = 98.7-100). Over 48% of study patients had a low (< or = 1.8%) or a very high (> or = 99.8%) predicted probability of cirrhosis. The model had area under the receiver operating characteristic curve of 0.938 (95% CI = 0.91-0.97) and 93.4% in an internal validation. The model accurately distinguished patients with and without cirrhosis (area under the receiver operating characteristic curve = 93.3%) in 102 hepatitis C patients from another hospital.

CONCLUSIONS: In patients with hepatitis C, four readily available variables together predict cirrhosis accurately. Successful validation in hepatitis C patients at another hospital with lower prevalence of cirrhosis suggests this model's potential for broad applicability.

Volume

97

Issue

10

First Page

2623

Last Page

2628

ISSN

0002-9270

Disciplines

Business Administration, Management, and Operations | Health and Medical Administration | Management Sciences and Quantitative Methods

PubMedID

12385450

Department(s)

Administration and Leadership

Document Type

Article

Link to Full Text

Share

COinS