AIM: The aim of the study was to detect the genetic predictors of reseponse to haloperidol. BACKGROUND: Haloperidol is a benchmark drug for the pharmacological treatment of schizophrenia, but the genetics of its efficacy is yet to be elucidated. METHODS: A genome-wide association analysis was carried out in a small sample of patients treated with haloperidol (n=96) and the results were replicated in a larger sample of patients treated with second-generation antipsychotics or perphenazine (final n=169, available from the Clinical Antipsychotic Trials for Intervention Effectiveness study). The Positive and Negative Symptom Scale % score decrease was the outcome in both samples. The period of observation was restricted to 1 month in the replication sample and the most severe cases were included to best balance the replication. The quality control (QC) for the investigation and replication sample included a minor allele frequency at least 0.01, call rate at least 0.95, and Hardy-Weinberg equilibrium P at least 0.0001. The source for imputation was the 1000 Genomes Pilot+HapMap 3 dataset. In total 1 080 870 single nucleotide polymorphisms (SNPs) were available after imputation and QC in the investigation sample. After QC of real genotypes, locus-targeted imputations were restricted to windows of 10 kb on either side of the sentinel SNP in the replication sample. Sentinel SNPs were the most significant findings in the investigation sample. Analysis of variance was the test of choice, PLINK, SNPTEST, and GTOOL were used in the analysis. RESULTS: Two SNPs (rs7912580 and rs2412459) were associated with response in both samples, respectively, located in an intergenic region between the AT-rich interactive domain 5B (ARID5B, MRF1-like) gene and rhotekin 2 (RTKN2) gene, an intronic region located in the eukaryotic translation initiation factor 2α kinase 4 (EIF2AK4) gene (P=1.358e-06 and 0.015 for the Positive and Negative Symptom Scale % total score decrease in the investigation and replication samples, respectively). The direction of association was opposite in the two samples, a finding that is sometimes reported as a flip-flop association. CONCLUSION: Heterozygosis for the ancestral allele was associated with the best improvement in the investigation sample and with poorer outcome in the replication sample. This discrepancy can be because of differences in the replication and investigation sample including the drugs used and the severity at baseline. Nevertheless, this finding is in line with two relevant hypothesis of schizophrenia, related to alterations in the immunological system (RTKN2) and in the neurodevelopment of the central nervous system (EIF2AK4). More studies are warranted to further investigate these associations.