From 3116 patients treated with clozapine, we identified patients who had a TDM history of non-clozapine, atypical antipsychotics (aripiprazole, risperidone, olanzapine, or quetiapine) preceding clozapine treatment from a TDM/pharmacogenetics service at the Center for Psychopharmacology, Diakonhjemmet Hospital, Oslo, Norway, during the period of January 2005 and February 2023. This laboratory service performs concentration analyses requested by clinicians for dose adjustments to obtain serum levels in the target concentration range, check for potential nonadherence, and determine a patient’s metabolizer status for specific drugs in psychiatric clinical settings. Longitudinal TDM profiles enable assessment of treatment stability, medications used (analyzed), change in medications (switch), and metabolic phenotypes at the patient level over time.

In TDM research, concentrations are often dose-adjusted to harmonize the data and obtain surrogate measures of drug clearance. For the study, interquartile ranges of population-based CDs were prepared for the relevant antipsychotic drugs from the measurements in the TDM registry at the study site (n = 147,964, Table S1 in the electronic supplementary material [ESM]). Patients were subgrouped based on these ranges. Those with at least one measurement with CD of non-clozapine antipsychotics below the 25th percentile of the respective drug’s population-based CDs were assigned to the LCD (cases) group, whereas patients who exhibited CDs between the 25th and 75th percentiles were assigned to the normal CD group (NCD, controls).

In the current study, adult patients with a history of detectable serum concentration of clozapine and/or its major metabolite, N-desmethylclozapine, were eligible for inclusion. Those patients with measurements of orally administered non-clozapine antipsychotics (and their major metabolites) available in the TDM database within 1 year preceding the first clozapine TDM measurement were included. The included non-clozapine antipsychotics comprised aripiprazole (metabolite: dehydroaripiprazole), olanzapine (N-desmethylolanzapine), risperidone (9-hydroxyrisperidone), and/or quetiapine (desalkylquetiapine). Exclusion criteria were (i) lack of dosing information, (ii) measurements with CDs above the 75th percentile population-based threshold (Table S1, see ESM), and (iii) for risperidone specifically, co-medication with paliperidone. Furthermore, we searched for comedication information that may influence the pharmacokinetics of the included antipsychotics [18]. These were (i) CYP inducers (carbamazepine, phenobarbital, and phenytoin), (ii) CYP2D6 inhibitors (bupropion, fluoxetine, and paroxetine; for patients treated with aripiprazole and risperidone), and (iii) valproate and CYP1A2/3A4 inhibitor fluvoxamine (for patients treated with olanzapine). There were 10 olanzapine measurements with potential drug–drug interactions (n = 1 carbamazepine, n = 9 valproate comedications) and were excluded in a secondary analysis. In addition to the typical interacting drugs (CYP inhibitors/inducers), psychiatric comedications were systematically reviewed for patients assigned to LCD or NCD groups. This was performed to assess potential differences in psychiatric drug comedication patterns between the two groups.

Information on age, sex, prescribed dosage, time between last drug intake and blood sampling for TDM (withdrawal time), comedications, and smoking habits (yes/no) were obtained from the requisition forms filled out by the physicians. The study population comprised only Norwegian inhabitants and were assumed to mainly comprise patients of Caucasian ancestry (ethnicity not confirmed due to privacy restrictions).

Serum concentration of antipsychotics and their metabolites was determined using the liquid chromatography tandem mass spectrometry (LC MS/MS) method, which was validated and certified for TDM routine analyses. Due to the renewal of instrumentation during the study period, the assays had been modified and cross-validated in line with FDA guidelines on precision, accuracy, and detection limits (Table S2, see ESM).

Subtherapeutic serum level status was only applied as descriptive information and a potential measure of under-treatment when comparing LCD and NCD patient groups. To define the subtherapeutic measurements, the lower boundary of the therapeutic reference ranges recommended in the consensus guidelines were used [19, 20]. These were aripiprazole active moiety 335 nmol/L (150 µg/L, aripiprazole plus dehydroaripiprazole), olanzapine 64 nmol/L (20 µg/L), quetiapine 261 nmol/L (100 µg/L), risperidone active moiety 47 nmol/L (20 µg/L, risperidone plus 9-hydroxyrisperidone). Frequency of antipsychotic polypharmacy (at least two) preceding clozapine treatment required detectable levels of the following antipsychotics: amisulpride, aripiprazole, brexpiprazole, cariprazine, chlorprothixene, flupentixol, haloperidol, levomepromazine, lurasidone, olanzapine, paliperidone, perphenazine, quetiapine, risperidone, sertindole, ziprasidone, and zuclopenthixol.

For patients who were using risperidone and aripiprazole before clozapine treatment, we searched the laboratory database for potential access to CYP2D6 genotypes (available for 50–53% of the patients treated with risperidone and aripiprazole, respectively). CYP2D6 genotyping of patients with schizophrenia is quite common in Norway, and offered along with TDM at the laboratory where we sourced data included in the study. The assays for CYP2D6 genotyping at the laboratory are TaqMan-based real-time polymerase chain reaction assays with the CYP2D6 pharmacogenetic panel comprising the lack-of-function variants (Nonf) CYP2D6*3 (rs35742686), CYP2D6*4 (rs3892097), CYP2D6*5 (whole gene deletion), CYP2D6*6 (rs5030655), the decreased function (Decr) variants CYP2D6*9 (rs5030656), CYP2D6*10 (rs1065852), CYP2D6*41 (rs28371725), and duplicated functional CYP2D6 alleles. Normal CYP2D6 allele (Norm, i.e., CYP2D6*1) is interpreted when none of the variant alleles included in the panel are detected.

Risperidone- and aripiprazole-treated patients with available CYP2D6 genotype information were categorized into genotype-predicted metabolic capacity/phenotype according to a previous publication [21] as poor, intermediate, intermediate plus, normal, and ultrarapid metabolizers.

The main outcome measure of the study was the metabolic ratio differences between LCD and NCD groups who were treated with antipsychotics of interest (i.e., aripiprazole, risperidone, olanzapine, and/or quetiapine). As a secondary outcome, we investigated the frequency of subtherapeutic serum levels of these drugs among LCD and NCD patients.

Metabolic ratio was estimated by the ratio of the metabolite (nmol/L) and parent drug concentration (nmol/L). CD was calculated as a ratio of parent drug concentration (nmol/L) and total daily dose (mg/day). For statistical comparisons between LCD and NCD patients, Pearson’s chi-squared tests (sex, polypharmacy status, smoking habits, frequency of subtherapeutic levels) and Fisher’s exact tests (CYP2D6 genotype-predicted phenotypes) were conducted for categorical variables, whereas Mann–Whitney U Tests (CD, metabolic ratio, total daily dose, age, withdrawal time, time between last non-clozapine measurement and first clozapine measurement) were used for continuous variables.

All statistical analyses were conducted using R Statistical Software version 4.3.1 [22].