The articles to which Dr. Kisloff refers (1, 2) were published after our article had been submitted. Neither study performed systematic liver biopsies on all patients to determine the presence of clinically undetected cirrhosis, so our projections for cirrhosis are not synonymous with mortality or risk for end-stage liver disease. A meta-analysis found a 27.9% progression from chronic hepatitis C to cirrhosis after 8 to 12 years (3). With 20-year likelihoods of 27.5% and 9.5% for compensated cirrhosis, risks for decompensated cirrhosis were 6.2% and 1.1%, respectively, and risks for liver disease-related death were 5.6% and 1.0%, respectively.

Our analysis applies only to patients enrolled in treatment trials who are positive for chronic hepatitis C on polymerase chain reaction and have elevated aminotransferase levels, or to similar patients. In the study by Seeff and colleagues (1), the first of the two studies mentioned by Dr. Kisloff, 11 persons were RNA positive for chronic hepatitis C. Because one quarter to one third of patients with chronic infection have normal aminotransferase levels and a low risk for progression, Seeff and colleagues probably observed one or two liver-related deaths in seven or eight patients with elevated aminotransferase levels. When modified to match this study, our model predicted a 23% mortality rate for patients with liver disease. This translates into an 80% likelihood of observing two or fewer liver-related deaths in the study by Seeff and colleagues. After the study by Thomas and coworkers (2) (the second study mentioned by Dr. Kisloff) was adjusted for the proportion of patients likely to have elevated aminotransferase levels and persistent viremia, 4.0% died of liver disease and 4.6% developed decompensated cirrhosis. In comparison, our model predicted a 3.7% (95% CI, 2.6% to 4.7%) mortality rate for patients with liver disease and a 4.2% (CI, 3.1% to 5.3%) incidence of decompensated cirrhosis. Thus, our model predictions were consistent with these two studies, providing further model validation (4).

In a sensitivity analysis, when we considered the possibility of doing at most three liver biopsies in patients with histologically persistently mild hepatitis, the results did not change. Last, the predictive model for response was based on data from clinical trials, and the effects of dosage reduction and discontinuation on therapeutic efficacy were therefore included. To account for potential side effects in patients receiving therapy, the model reduced their quality of life. This reduction, however, was more than offset by the loss in length and quality of life due to end-stage liver disease.

Sir William Osler once said, "Medicine is a science of uncertainty and an art of probability." Although not all patients respond to treatment and not all develop liver disease, the long-term benefit of preventing the devastating effects of end-stage liver disease with immediate therapy should, on average (5), be preferred over watchful waiting.