We thank Lau et al. for their comments regarding our article,1 as it gives us the opportunity to provide a more in-depth report of some of our results. When the trial was planned, we decided to allow patients to cross over to the opposite treatment or to a new one (i.e., peritoneo-venous shunt) when the assigned treatment had failed, according to the doctors' advice and patients' will, because of ethical reasons. As a result, 11 patients crossed over from paracentesis to TIPS after a variable interval from randomization (2-16 months; mean 4.5 ± 1.5 SE).
Effectively, a 33% crossover could have affected the final outcome of the entire group. However, the probabilities of survival were slightly better in patients who crossed over to TIPS than in the other 22 patients who were randomized to paracentesis (62% vs. 46% at 1 year and 28% vs. 31% at 2 years, respectively). Even though the difference was not significant (log-rank 0.25; P = .61), in an intent-to-treat analysis, the patients who crossed over to TIPS could have slightly improved the probability of early survival in the group assigned to paracentesis, reducing the gap between the two arms of the trial. In fact, we repeated our Kaplan-Meier analysis excluding the 11 patients who crossed over to TIPS, and we found that the survival was still significantly better in the group of patients randomized to TIPS (log-rank 4.28; P = .038).
Treatment was considered as having failed when the patient required at least four large-volume taps over the last month, independently of his/her sodium excretion rate. Nevertheless, at the time of failure, 9 of the 11 patients who crossed over to TIPS showed a daily sodium excretion lower than their daily sodium intake.
With regard to hepatic encephalopathy, the findings of our trial were very similar to those reported by Ginès et al.2: the total number of patients who developed new episodes of encephalopathy was greater (though not significantly) in patients treated with TIPS; however, severe encephalopathy was more frequent in patients with TIPS. In our case, the crossover to TIPS could have reduced the difference between the two groups; therefore, we agree that encephalopathy is a serious clinical problem of patients treated with portal-systemic shunting.
Finally, regarding the observation that numbers of censored patients were different, we would like to note that the average survival of the 20 dead patients randomly assigned to paracentesis was 6.4 ± 1.3 months, whereas that of the 13 dead patients assigned to TIPS was 19.7 ± 6 months. According to these figures, just after the first 6 months of follow-up the number of patients still at risk differed in the two groups.
In all, we agree that although TIPS has been shown to benefit the control of ascites and the survival of some patients, it could be detrimental for others. Accordingly, such an invasive procedure must be preceded by accurate selection.