Viral breakthrough was defined as an HCV RNA increase of ≥1 log10 IU/mL from the lowest level reached during treatment, or HCV RNA >100 IU/mL in patients who previously had <25 IU/mL during treatment. Relapse was defined as detectable HCV RNA during the follow-up period after having undetectable HCV RNA at the end of treatment. On the basis of previous studies,13, 15 commonly observed substitutions in NS3 after treatment failure considered to confer lower-level in vitro resistance to telaprevir (3- to 25-fold increase in replicon 50% inhibitory concentration

[IC50]) were: V36A/M, T54A/S, R155I/K/M/T, and A156S. Substitutions considered to confer higher-level in vitro resistance to telaprevir (>25-fold increase in replicon IC50) were ABT-199 cell line A156T/V and the combination of V36M+R155K.16 Other changes within the NS3·4A region were also investigated. Following sequencing, amino acid positions were assigned with hidden Markov models using HMMer2 software (Howard Hughes Medical Institute, Chevy Chase, MD), which was trained on multiple sequence alignments of HCV reference sequences from the Los Alamos National Laboratory database.17 Pretreatment sequence and sequence at time of failure were compared for all patients with on-treatment virologic failure or relapse. MDV3100 in vitro Potential new resistance-associated mutations were identified as amino acid states whose

frequencies were significantly different between pretreatment and failure sequences. Aspartate Statistical significance was defined as a one-tailed P < 0.05 using Fisher's exact test for unpaired pretreatment and failure sequences, and Liddell's exact test18 for paired sequences. A Bonferroni correction was applied for multiple comparisons. For each patient not achieving an SVR, any nonwildtype variants at positions known to be associated with telaprevir treatment failure (36, 54, 155, and 156)

were indexed from the failure visit. The proportion of patients losing these variants was recorded until the end of study visit (i.e., last available sequence during the study). To ascertain the median time to loss of variants as compared to time of failure at each position, nonparametric (Kaplan-Meier) survival analyses were performed. P-values for other analyses mentioned in this article were generated using the chi-squared test and were not calculated where sample sizes were low. The disposition of patients in the REALIZE trial, and the baseline characteristics of the two telaprevir treatment arms included in this virologic analysis, have been published elsewhere.4 Briefly, 662 patients were randomized: 266 to the T12/PR48 arm, 264 to the lead-in T12/PR48 arm, and 132 to the PR48 control arm. Regarding previous peginterferon/ribavirin response, 53% were prior relapsers, 19% were prior partial responders, and 28% were prior null responders.