We appreciate the concerns expressed in these letters. Although a randomized trial examining the efficacy of interferon for survival or cirrhosis prevention would provide definitive evidence, it would require many patients to be followed for a long period—a costly and risky enterprise. Given the known complications of hepatitis B and the increasing number of antiviral agents, a control group untreated for such a lengthy trial would be difficult to maintain. Treatment decisions must be based on the best available data. We believe that the data available in 1995 support the belief that interferon prolongs life and lowers costs.

The evidence supporting the loss of HBeAg after treatment with interferon is of the highest level (1) and is based on our meta-analysis of randomized trials that establishes a pooled efficacy with low false-positive () and false-negative (ß) rates, high statistical significance, and no marked heterogeneity among trials. On the basis of these data, we assumed that HBeAg-positive patients are more likely to develop cirrhosis than are those who are HBeAg-negative, but we did not assume that HBeAg-positive patients are more likely to develop decompensated cirrhosis (after developing cirrhosis) or hepatocellular carcinoma (independent of their liver histologic findings).

Although clinical factors predict response, studies examining differential responses to interferon within subgroups are unavailable. Our analysis suggests that even if interferon increased the likelihood of losing HBeAg by only 10% in subgroups less likely to respond, life expectancy would still increase by 10 months. In our appendix, we examined the possibility that nonresponders might have a worse prognosis than untreated patients because of either perversity or selection. Interferon would actually have to decrease the likelihood of nonresponders losing HBeAg by approximately 500-fold for it to provide no benefit.

Our meta-analysis differed from the analysis of Drs. Evans and Loudon in that we excluded patients lost to follow-up or death and considered only the treatment period, not 1-year outcomes (we considered reactivation of HBeAg separately). If we examine 1-year outcomes, our model predicts a risk difference of 29%, a value lower than the difference in Drs. Evans and Loudon's analysis. Moreover, if we use their estimate, our results are essentially unchanged: Interferon should increase life expectancy by about 3 years and still save resources.

Although Schering-Plough Laboratories provided funding, they had no input into the structure of our model, any data used, or manuscript content. We were free to publish results regardless of the outcome of our analysis.