To explore contemporarily genetic and non-genetic determinants of long-term immunological memory to hepatitis B (HB) vaccination, we conducted a case-control study nested in an adolescent cohort of booster recipients who had received primary infantile HB vaccination but with residual anti-HBs titers <10 mIU/mL at 15-18 years of age. High-resolution phenotypes of human leukocyte antigen (HLA)-A, -B, and -DRB1 loci were determined by sequence-specific oligonucleotide probe hybridization. After controlling for pre-booster anti-HBs levels, the absences of HLA-A*02 and -DRB1*08, simply expressed as A*02(-) and -DRB1*08(-), and the presence of B*15 were significantly associated with elevated risks of non-response (post-booster anti-HBs titers<10 mIU/mL) to booster vaccination. The adjusted odds ratios (ORs) were 3.85 (CI, 1.82-8.33), 4.55 (CI, 1.23-16.67), 3.59 (CI, 1.40-9.17), respectively. There was multiplicative synergism between A*02 and B*15 on the risk of non-response to booster vaccination. The multivariate-adjusted ORs for A*02(-)/B*15, A*02(-)/B*15(-), A*02/B*15, and A*02/B*15(-) haplotypes were 20.39 (p=0.0003), 3.29 (p=0.007), 1.32 (p>0.05), and 1.0, respectively. Recent cigarette smoking and/or betel-quid chewing was associated with a 12-fold risk of non-response to booster vaccination. Further comparisons between responders and adolescents who had undetectable post-booster anti-HBs titers (<0.1 mIU/mL) demonstrated similar results. Our results indicated that response to booster HB vaccination as well as long-term immunological responses to HB vaccination are closely related with host genetic factors, and probably modified by recent substance use.