In the maintenance phase of AIEOP‐BFM acute lymphoblastic leukemia (ALL) 2009 protocol, mercaptopurine (MP) is given at the planned dose of 50 mg/m2/day; however, dose adjustments are routinely performed to target patients’ white blood cells to the optimal range of 2000‐3000 cells/μl. ALL pediatric patients (n=290, age: median (1st‐3rd quartile): 4.8 (3.0‐8.1) years; male: 56.9%) were enrolled mainly in four medium‐large Italian pediatric hospitals; 14.1% of patients relapsed after a median (1st‐3rd quartile) follow‐up time of 4.43 (3.82‐5.46) years from maintenance beginning. MP metabolites (thionucleotide (TGN) and methyl‐derivatives (MMPN)) were measured in the erythrocytes of 387 blood samples of 200 patients by HPLC‐UV. SNPs (rs1800462, rs1800460 and rs1142345 in TPMT gene, rs116855232 in NUDT15, rs1127354, rs7270101, rs6051702 in ITPA and rs2413739 in PACSIN2) were characterized by Taqman SNP genotyping assays. Cox proportional hazard models did not show an impact TGN levels and variability on relapse. In contrast, after multivariate analysis, relapse hazard ratio (HR) increased in ALL children of the intermediate risk arm compared to those in standard risk arm (3.44, 95% confidence interval (CI), 1.31‐9.05, p= 0.012), and in carriers of the PACSIN2 rs2413739 T allele compared to those with the CC genotype (heterozygotes CT: HR, 2.32; 95% CI, 0.90‐5.97; p=0.081; homozygous TT: HR, 4.14; 95% CI, 1.54‐11.11; p=0.005). Future studies are needed to confirm the lack of impact of TGN levels and variability on relapse in the AIEOP‐BFM ALL trials, and to clarify the mechanism of PACSIN2 rs2413739 on outcome.