3509 Background: Overall response rate (ORR) to temozolomide (TMZ) is ∼10% in refractory mCRC pts with MGMT methylation detected by qualitative assays, e.g. methylation-specific PCR (MSP). ORR to irinotecan/FOLFIRI in second-line trials was 4-16%. The efficacy of TMZ may be improved by its combinatorial use in earlier lines and molecular selection beyond MSP. Lack of MGMT expression by immunohistochemistry (IHC) and high MGMT % methylation by MethylBEAMing (MB) are prognostic for higher ORR/PFS in TMZ-treated pts. Methods: This multicenter, randomized phase 2 trial investigated PFS superiority of second-line CAPTEM (Arm A: capecitabine 750 mg/sqm bid days1-14/TMZ 75 mg/sqm bid days10-14q28 days) over FOLFIRI (arm B) in RAS mutated mCRC pts with MGMT methylation centrally confirmed by MSP. Eligible pts: ECOG PS 0-1, measurable disease, failure of 1^st-line oxaliplatin-based tx (or relapse within 6 mos from oxaliplatin-based adjuvant tx). Randomization was stratified by time from the start of oxaliplatin-based therapy to PD ( < /≥9 months); prior bevacizumab (yes/no). A one-sided log-rank test with a sample size of 82 pts (41 per arm) achieved 90% power at a 5% significance level to detect mPFS increase from 2 to 4 mos. Secondary endpoints: safety, QoL, OS, ORR. Exploratory endpoints: predictive value of MGMT IHC/MB. Results: From Nov 2014 to Feb 2019, 82 pts (arm A/B: 41/41) were enrolled in 18 Italian sites. Baseline characteristics (arm A/B): males 44/56%, median age 70/67, ECOG PS 0 54/51%, right-sidedness 37/39%, 1 metastatic site 44/34%, prior bevacizumab 68/66%, 1^st-line PFS 9,4/10,2 months. At a median follow up of 26.6 mos, 70 PFS/46 OS events were collected. The mPFS was 3.6 vs 4.1 mos in arm A vs B (HR = 1.26;95%CI 0.78-2.02;p = 0.34) and mOS was 9.1 vs 14.2 mos (HR =1.08;95%CI 0.60-1.94;p = 0.79). ORR and DCR (arm A/B): 12/10% and 51/51%. Grade 3-4 adverse events: 15/44% (diarrhea 0/12%, stomatitis 0/7%, anemia 2/10%, neutropenia 2/22%, thrombocytopenia 7/0%). Neither MGMT IHC nor MB status were prognostic. MGMT IHC positive subgroup, arm A (n = 12) vs arm B (n = 22): mPFS, 2.0 vs 4.1 mos (HR = 2.06;95%CI 0.96-4.45;p = 0.06), mOS, 6.4 vs 10.6 mos (p = 0.78), ORR (0% vs 14%) and DCR (25% vs 55%;OR = 0.28;95%CI 0.06-1.31;p = 0.11). In MGMT IHC negative subgroup, no PFS/OS/ORR/DCR differences were noted between the two arms. P interaction IHCxArm: 0.171 for PFS, 0.917 for OS, 0.06 for DCR. Similar accuracy was achieved by MB. Conclusions: The use of TMZ should be explored by phase 3 trials enrolling MGMT IHC-negative +/- high MGMT % methylated mCRC. Clinical trial information: NCT02414009.