Abstract Clozapine (CLZ) is the superior antipsychotic drug for treatment of schizophrenia, but exhibits an extensive interpatient pharmacokinetic variability. Here, we conducted a genome-wide association study (GWAS) of CLZ serum concentration adjusting for known smoking habits, which is a major nongenetic factor reducing CLZ levels. The study included 484 patients with 10,283 steady-state serum concentrations of CLZ and N-desmethylclozapine, prescribed dosing, co-medications and known smoking habits (n = 422; 9284 serum samples) from a therapeutic drug monitoring (TDM) service. The GWAS analyses were performed with and without smoking habits as covariate, where possible hits were assessed in relation to the target CLZ concentration range applied in the TDM service (300–2500 nmol/L). The smoking-independent analysis of N-desmethylclozapine serum concentration and the CLZ-to-N-desmethylclozapine ratio replicated the previously identified locus on chromosome 4. After adjusting for smoking habits in patients confirmed as ‘smokers’ (61%) or ‘nonsmokers’ (39%), a novel variant (rs28379954; minor T>C allele frequency 4.1%; 7.6% CT carriers in the population) within the gene encoding the nuclear factor 1 B-type (NFIB) was significantly associated with reduced CLZ serum concentration (p = 1.68 × 10−8, beta = −0.376; explained variance 7.63%). There was no significant association between rs28379954 and N-desmethylclozapine concentration in the GWAS analysis (p = 5.63 × 10−5). The fraction of CLZ TDM samples below 300 nmol/L was significantly higher in carriers vs. noncarriers of the rs28379954 minor C allele [12.0% (95% CI: 9.4–14.7) vs. 6.2% (95% CI: 5.7–6.8), p < 0.001]. We identified a novel variant in the NFIB gene associated with reduced CLZ levels and increased risk of subtherapeutic serum concentrations. This warrants testing of clinical relevance of screening for this gene variant, and also experimental studies to investigate the biological mechanisms of NFIB involvement in CLZ pharmacokinetics.