Background: With 24 mo f/u, ENESTcmr demonstrated higher rates of stable, deep MRs with nilotinib (NIL) vs IM in pts on long-term (≥ 2 y) IM with residual disease. Here, we present 36-mo results, including crossover from IM to NIL after 2 y on study. Methods: Pts with Philadelphia chromosome–positive CML-CP (N = 207) with complete cytogenetic response but detectable BCR-ABL (by RQ-PCR with a sensitivity of ≥ 4.5 logs) after ≥ 2 y on IM were randomized to NIL 400 mg twice daily (BID; n = 104) or IM 400 or 600 mg once daily (QD; n = 103). Crossover from IM to NIL was allowed for pts with detectable BCR-ABL after 24 mo, treatment failure, or confirmed (≥ 2 consecutive assessments) loss of response at any time. Results: Significantly more pts achieved MR^4.5 by 36 mo with NIL (Table). Median time to MR^4.5 was 24 mo in the NIL arm and not reached in the IM arm with 36 mo f/u. 46 of 103 (45%) pts randomized to IM crossed over to NIL. When accounting only for responses up to crossover, 47% and 24% of pts on NIL and IM, respectively, achieved MR^4.5 (P = .0003). At 24 mo, 52 pts on NIL and 78 pts on IM had detectable disease; 4/52 who continued NIL, 0/35 who continued IM, and 11/43 who crossed over from IM to NIL achieved undetectable BCR-ABL by 36 mo. The rate of MR^4.5 appeared higher in pts randomized to NIL (33% by 1 y in pts without MR^4.5 at baseline [BL]) than in pts who crossed over to NIL (21%) with similar follow-up. Adverse event profile was similar to the 12 mo report. Conclusions: By 36 mo, significantly more pts achieved MR^4.5 by switching to NIL vs remaining on IM and median time to MR^4.5 was accelerated by more than 1 y in the NIL arm. Pts with detectable disease who crossed over from IM to NIL after 24 mo were able to achieve deep MRs by 36 mo on study, whereas no pts who remained on IM achieved undetectable BCR-ABL. Delaying switching from IM to the more potent BCR-ABL inhibitor NIL does not increase the proportion of pts achieving deep MR. Clinical trial information: CAMN107A2405.