Transient elastography for predicting clinical outcomes in patients with chronic liver disease.

There is increasing interest in developing noninvasive means to evaluate liver fibrosis in patients with chronic liver disease to determine disease severity, prognosis and optimal treatment. Transient elastography (TE) has previously been demonstrated to predict the presence or absence of advanced fibrosis. The current study was conducted to determine whether TE can identify patients with chronic liver disease at risk of clinical decompensation. A total of 667 patients underwent TE and were followed for a median of 861 days and 57 patients achieved the primary outcome, a composite of clinical endpoints including death, ascites, encephalopathy, increased Child Score ≥ 2, variceal bleed, hepatocellular carcinoma or listing for transplant. Overall, TE had an area under the receiver operating characteristic curve of 0.87 for predicting clinical outcome. Using a cut-off of 10.5 kPa, TE has a sensitivity, specificity, positive predictive value and negative predictive value (NPV) of 94.7%, 63.0%, 19.3% and 99.2%, respectively. A predictive model for clinical events was developed using generalized cross-validation for clinical endpoints considering TE, liver biopsy results and multiple other predictors. Individually, TE performed better than biopsy, or any other variable, for predicting clinical outcome [Harrell's C Statistic 0.86 for TE, 0.78 for stage]. Patients with a TE score of >12.5 kPa were found to have a relative hazard for clinical event of 18.99 compared with patients with TE score <10.5. A combined variable model including TE, aspartate aminotransferase/alanine aminotransferase ratio and model for end-stage liver disease (MELD) yielded the highest predictive accuracy with Harrell's C value of 0.93. In the subset of patients with cirrhosis, TE was not found to be independently associated with clinical outcomes in univariate or multivariate analysis although it retained a high sensitivity and NPV of 97.5% and 92.3%, respectively, at a kPa cut-off of 10.5. TE can successfully identify patients with chronic liver disease who are at low risk of clinical decompensation over a time period of 2 years.