Clozapine levels and outcomes in Serbian patients with treatment-resistant psychotic disorders previously treated without measuring clozapine levels (CLOSER)

of clozapine is not routinely performed. This study evaluated the distribution of clozapine levels and their relationship with clinical outcomes in Serbian patients who had not undergone prior TDM. 140 Patients with treatment-resistant schizophrenia and schizo-affective disorder were enrolled. Clozapine levels were measured by dried blood spot (DBS) analysis. Side effects were evaluated by GASS-c, severity of symptoms and functional impairment with WHODAS, CGI-S and GAF. Of the patients, 51.2% had subtherapeutic levels, 24.8% were in the therapeutic window, and 24% had supratherapeutic levels. Clozapine levels showed no association with side effects and a weak positive association with symptom severity and functional impairment. No serious side effects were observed in patients with clozapine levels surpassing 1000 ng/ml ( n = 8). Based on these findings, we propose that the upper limit of the therapeutic range should not be regarded as an absolute barrier, and guidelines should allow for a personalized approach when prescribing clozapine.


Introduction
Treatment-resistant schizophrenia (TRS) is characterized by persisting symptoms despite the sequential administration of at least two appropriate antipsychotic treatments and affects approximately 30% of individuals with the classification schizophrenia (Elkis and Buckley, 2016).With response rates of 40-60%, clozapine is the only pharmacological option with proven efficacy for TRS (Correll and Howes, 2021).Furthermore, clozapine has demonstrated effectiveness in managing treatment-resistant schizoaffective disorder (Rey Souto et al., 2021).However, it has side effects ranging from common and cumbersome (including sedation) to relatively rare but potentially life threatening (including ileus, pneumonia, myocarditis and neutropenia) (Miller, 2000).
Clozapine levels over 350 ng/ml are associated with twofold (Siskind et al., 2021) to threefold (Tralongo et al., 2023) higher odds of response.As side effects are also concentration-related (Tan et al., 2021) and the evidence for benefit of levels over 600 ng/ml is limited (Siskind et al., 2021), the therapeutic range has been defined as 350-600 ng/ml (Correll et al., 2022).The laboratory alert level (i.e.toxic threshold) is 1000 ng/ml (Hiemke et al., 2018).These values are not absolute: a subset of patients responds at levels well below 350 ng/ml, and some require and tolerate levels >600 ng/ml and even >1000 ng/ml (2021; Mayerova et al., 2018;Meyer, 2019;Yada et al., 2021).
Although several international clozapine guidelines (Correll et al., 2022;Group, 2013) recommend therapeutic drug monitoring (TDM), both for optimization of treatment efficacy and patient's safety, the implementation of TDM is inconsistent across countries.In the UK, TDM is optional, employed only under specific indications such as dose optimization or compliance assessment (2022).In Japan, TDM is not covered by health insurance and its use is limited to a few research institutions (Yada et al., 2021).In low-and middle-income countries (including Serbia), clozapine concentration measurement is often unavailable due to financial and infrastructural constraints.A recent guideline for clozapine titration provides specific recommendations for safe clozapine titration in absence of TDM (de Leon et al., 2022).Data from countries with and without routine TDM for clozapine show that a considerable proportion of clozapine treated patients are outside the therapeutic range (Flanagan et al., 2023;Mayerova et al., 2018).
The aims of our study are 1) to determine the proportions of Serbian patients -titrated without TDMwith clozapine levels in the subtherapeutic, therapeutic, supratherapeutic and laboratory-alert range and 2) to determine the relationship of clozapine levels with side effects and symptomatic / functional outcomes.

Setting, study population and design
The study was approved by the Ethics Committee, University Clinical Center Kragujevac, Serbia (project number: 01/4897).Patients were recruited from three Serbian psychiatric hospitals: the University Clinical Center of Kragujevac (Kragujevac), the Special Hospital for Psychiatric Diseases Gornja Toponica (Toponica) and the Special Hospital for Psychiatric Disorders Kovin (Kovin).Inclusion criteria were a) an ICD-10 classification of schizophrenia (F20 and its subclassifications) or schizo-affective disorder (F25 and its subclassifications); b) treatment resistance defined as documented treatment failures with at least 2 different antipsychotic medications; c) ongoing clozapine treatment with stable dosage for ≥ 4 weeks; d) age ≥ 18; e) sufficient mastery of the Serbian language; f) capacity and willingness to provide informed consent.Capacity to provide informed consent was assessed using the UCSD Brief Assessment of Capacity to Consent (UBACC).In all cases, clozapine had been titrated without measuring its concentration, but with regular monitoring of ANC and metabolic parameters.
Patients underwent measurement of the clozapine trough levels and clinical assessments of the severity of illness, impairment of functioning and side effects.For each patient, blood sampling and all clinical assessments were performed on one day, by the treating physician (psychiatry resident or psychiatrist).All measurements were done at baseline (t1) and repeated at ~3 months (t2) and ~6 months (t3).Here we report on the baseline assessments.

DBS blood sampling and analysis
Blood samples were taken at 12 h after the previous dosage.Analysis of the DBS-samples was performed at the hospital pharmacy laboratory of the University Medical Center Groningen, the Netherlands.For DBS blood sampling and analyses, previously validated and published procedures were followed (Geers et al., 2017(Geers et al., , 2022)).Briefly, after a lancet finger prick, a few drops of blood were captured with a special filter paper.The sample was allowed to dry at room temperature and stored at − 20 • C until batch transport to The Netherlands by express mail.For analysis, an 8-mm disc was punched out of each blood spot.Each disc was sonified in extraction buffer to extract the sample, which was analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS).Based on the previous validation study, a predefined conversion factor was used to calculate the corresponding plasma level from the measured capillary concentration (Geers et al., 2017(Geers et al., , 2022)).
Illness.Illness severity was assessed with the Clinical Global Impression -Severity of Illness (CGI-S) subscale (Leucht et al., 2006).The CGI-S provides a clinician's assessment of a patient's illness severity, relative to the clinician's experience with other patients with the same diagnosis with scores ranging from 1, "normal, not at all ill" to 7, "among the most extremely ill patients".
Symptom severity and functioning.The DSM IV Global Assessment of Functioning (GAF) scale (Startup et al., 2002) provides a clinician's assessment of a patient's social, occupational, and psychological functioning.Scores range from 1 (severely impaired) to 100 (extremely high functioning).
Disability.The level of disability was assessed by the proxyadministered 12 item World Health Organization Disability Assessment Schedule (WHODAS) (Holmberg et al., 2021).Item scores range from 0 (no difficulty) to 4 (extreme or cannot do), resulting in total scores ranging from 0 (no disability) to 48 (full disability).

Statistical analysis
For continuous data median and interquartile range (IQR) are presented.For group-wise comparisons of continuous data, the Mann-Whitney U test (2 groups) or Kruskal Wallis test (>2 groups) was performed.For group-wise comparisons of categorical data, the Chi 2 test was performed.
To evaluate the relationship between clozapine levels on the one hand and side effects (as measured by GASS-C) and disease severity / functional impairment (as measured by CGI-S, GAF and WHODAS) on the other hand, regression models were fitted.First, four univariate ("crude") models were fitted with clozapine level as the predictor variable and GASS-C, WHODAS, CGI-S and GAF scores as outcome variables.
Exploratory analysis had revealed that the centers differed regarding clozapine levels, side effects and severity of the other clinical outcomes.To account for this, center was added as categorical covariate (partially adjusted models).Center was not added as a level in a linear mixed model given the small number of centers (n = 3).Descriptive data are presented for each center individually as well as for the entire sample.
Finally fully adjusted modes were fitted for all four outcome variables by further adding the following covariates: center, clozapine : norclozapine ratio, use of antidepressants, use of other antipsychotics, use of anticholinergics, use of somatic medication, sex, age, BMI and divided clozapine dosing (i.e.whether the total daily clozapine dose was divided over multiple smaller doses).For the categorical covariates 'center' and 'sex', specific reference categories were used (Kragujevac for center and female for sex).For the other categorical covariates, the absence of a condition (i.e., a value of zero) was used as the reference category.
Graphical diagnostic methods were applied to test assumptions of non-collinear predictor variables, normality of residuals, linearity and homoscedasticity.For the fully adjusted models with CGI, WHODAS, and GAF as outcome variables, the assumptions were sufficiently satisfied.The fully adjusted model with GASS-C as outcome variable showed moderate violations of the linearity and homoscedasticity assumptions, which could not be salvaged by data transformation.Given the between-center differences, the fully adjusted models were refitted with data split by center as sensitivity analyses.This revealed that the combination of the centers in the dataset was partially responsible for the violations of linearity and homoscedasticity assumptions.Results of the sensitivity analyses were consistent with those of the fully adjusted models (data not shown).All regression models were fitted after removal of the cases with one or more missing values in any of the outcome or predictor variables, under the assumption of data missing at random.Representativity analyses were performed; the available and missing cases were compared on various relevant variables.No correction for multiple testing was carried out.All analyses were conducted using R version 4.3.0.and RStudio version 2021.09.1.P values <0.05 were considered statistically significant.

Patient characteristics
Table 1 shows the characteristics of the patients.As patient populations of the three centers showed differences, characteristics are presented for the entire group and for each individual center.A total number of 140 patients were recruited; 50 from Kragujevac, 49 from Toponica and 41 from Kovin.A majority of patients had finished high school, vocational secondary school or lower and most were unemployed or receiving (retirement or disability) benefits.The recruitment of patients across the three centers varied significantly by treatment setting.Kragujevac predominantly recruited outpatients, Kovin mostly inpatients, and Toponica exclusively inpatients.Most patients had classification schizophrenia.Smaller numbers of patients with classification schizo-affective disorder were recruited in Kragujevac and Toponica.Antipsychotics were the most commonly used co-medications, with slightly less than half of the patients using one or more agents of this class as co-medication.Similar co-prescription rates were observed for benzodiazepines.Close to a quarter used at least one mood stabilizer, with varying prescription patterns across clinics.Fewer patients were co-prescribed with anticholinergics, antidepressants, and somatic medications.

Patterns of clozapine prescription and clozapine levels
Table 2 provides an overview of clozapine dosages and levels.Clozapine dosages.22.3% of the patients were prescribed dosages below 100 mg clozapine.58.3% used 100-299 mg, 15.1% used 300-499 mg and the remaining 4.3% used 500-699 mg.No patient used over 700 mg/day.Dosing patterns showed differences across the centers, with a higher proportion of patients on low dosages in Kragujevac.

Relationship between clozapine levels and outcomes
Relationships between clozapine levels and GASS-C, WHODAS, CGI, GAF were assessed using regression models (table 4, supplementary Tables 1-4).Data for the fully adjusted models with GASS-C, WHODAS, CGI-S and GAF as outcome variables had 14, 13, 12 and 12 missing datapoints, respectively (the models regarding CGI-S and GAF had identical missings).For each model, patients with missing datapoints were compared to those without missing data.Other than higher scores on the WHODAS, the excluded patients did not show meaningful differences when compared with the included patients (supplementary Tables 5-7).Thus, the missing datapoints were excluded from analysis under the assumption of data missing at random.GASS-C.The bivariate (crude) regression model did not show a statistically significant relationship between clozapine level and GASS-C (B = − 0.002, CI = − 0.006 -0.002, p = 0.42).This did not change after correcting for center (partially adjusted model, B = 0.002, CI − 0.001 -0.005, p = 0.24) or all covariates (fully adjusted model, B = − 0.0003, CI = − 0.004 -0.003), p = 0.85, Fig. 1).Additional exploratory bivariate regression models did not identify clozapine level as a significant predictor of any of the 16 individual GASS-C items (data not shown).

Discussion
The median clozapine dosage in our study population of 140 Serbian in-and outpatients with TRS, titrated without TDM, was 100-300 mg, with 22% dosed below 100 mg and none dosed above 700 mg.TDM showed clozapine levels within the therapeutic range (350-600 ng/ml) in 36%.Subtherapeutic clozapine levels were found in 56%, and 12% (n = 17) had supratherapeutic levels (≥600 ng/ml).Clozapine levels >1000 ng/ml were found in 5.7% (n = 8).While no significant relationship between clozapine levels and side effects was found, a modest positive association of clozapine levels with disease severity and functional impairment was observed.

Prescription patterns and clozapine levels
A minority (24.8%) of the patients had clozapine levels within the therapeutic range.The differences between the three Serbian centers we observed might reflect the academic status of Kragujevac, with less severely ill patients with higher level functioning.
The finding that approximately half of the patients had levels below 350 ng/ml was unexpected.It is well established that the odds-ratio for Continuous data are presented as median (IQR), categorical data are presented as N (%).P-value superscripts: I Kruskal-Wallis.II Fisher's exact.III Chi-square. 1-6.
Numbers in superscript represent number of missing cases.remission with levels over this threshold is considerably higher than below it.Regrettably, we did not succeed in employing a symptom scale, which would have provided valuable information on remission rates in our group.However, a considerable proportion of patients with clozapine levels <350 ng/ml had a CGI score of 5, 6 or 7 (indicating a severity level of "markedly ill" or worse), and a GASS-C of ≤16 (mild side effects)(data not shown).We suspect that a proportion of this subset of patients with relatively low clozapine levels and low side effects would benefit from (and tolerate) clozapine up titration and that implementation of TDM in Serbia could facilitate this improvement.Three recent studies of clozapine levels in patients titrated without TDM, reported 32%, 33% and 31% of the patients with therapeutic clozapine levels (Mayerova et al., 2018;Nomura et al., 2021;Yada et al., 2021).The lower total percentage of patients with therapeutic clozapine levels in our study (24.8%) is largely attributable to the percentage in the center Kragujevac (15.9%).The percentages in the other two centers enrolled in our study were similar to those found in the other studies (Toponica 29.8% and Kovin 28.9%).
Interestingly, the proportions of patients with therapeutic clozapine levels in countries with and without clozapine TDM were in the same range: the British Isles national TDM database (~370,000 samples from ~50,000 patients) also found 36% of the patients with clozapine levels within the therapeutic range (Flanagan et al., 2023;Mayerova et al., 2018).And a study of Belgian TDM database (330 samples from 171 patients) reported that 22.3% had therapeutic clozapine levels (Patteet et al., 2014).It's noteworthy that the TDM databases include samples taken for specific clinical indications, such as suspected low or high concentration or suspected non-adherence, potentially influencing the results.Table 5 provides an overview of the six recent studies (this study included) of the distribution of clozapine levels, titrated with or without TDM.Abbreviations: B, regression coefficient; CI, confidence interval of regression coefficient; GASS-C, Glasgow Antipsychotic Side-effects Scale for Clozapine; WHODAS, WHO Disability Assessment Schedule 2.0 (12 item, proxy-administered; CGI-S, Clinical Global Impression -Severity Scale; GAF, Global Assessment of Functioning.a Adjusted for center.b Adjusted for center, clozapine : norclozapine ratio, use of antidepressants, use of other antipsychotics, use of anticholinergics, use of somatic medication, male sex, age, BMI and divided clozapine dosing.

Clozapine levels and side effects
The observation of higher GASS-C scores in patients from Kragujevac is intriguing, especially given the relatively lower clozapine levels in this center.Since Kragujevac was the only clinic enrolling outpatients, we initially speculated that this could be the explanatory factor.Supporting this hypothesis, a previous study validating the Serbian version of the GASS-C established that inpatients required more instruction than outpatients (Ignjatovic Ristic et al., 2018).However, when comparing GASS-C levels between inpatients and outpatients from Kragujevac, no difference was observed.The difference in GASS-C scores, like the difference in clozapine dosages, may be due to the academic nature of the Kragujevac center.The less severely ill patients from this center may have been more sensitive to clozapine-related side effects and/or more inclined to report their complaints.This observation is important and warrants attention in follow-up studies.To adjust for this difference, "center" was included as a categorical variable in the multiple regression model evaluating the relationship between GASS-C and clozapine levels.The analysis revealed no significant relationship between these variables; despite significant differences between centers, side effect levels within each center were uniform across the clozapine level range.We observed pronounced interindividual variations: some patients with low clozapine levels experienced considerable side effects, while others with higher levels did not.Our findings are in line with two recent studies with a similar methodology.In the cross sectional study by Yada et al. of 131 TRS patients on clozapine without TDM, no significant difference in mean clozapine levels was found between patients with or without ADRs.However, patients with levels above 1000 ng/ml experienced more severe ADRs (OR 31.72)(Yada et al., 2021).Moreover, Nomura et al. cross-sectionally evaluated 108 patients who had been titrated onto clozapine without TDM, finding no relationship between GASS-c and clozapine peak levels, trough levels or total clozapine exposure assessed by AUC calculation (Nomura et al., 2021).Comparison of GASS-C in patients with levels over versus below 600 ng/ml did not reveal a significant difference.
Contrarily, earlier studies suggest a relationship between clozapine levels and side effects.A recent systematic review and meta-analysis showed a positive relationship between clozapine levels and heart rate, triglycerides and combined ADRs (Tan et al., 2021).Another systematic review reported a positive relationship of clozapine level with EEG slowing, obsessive-compulsive symptoms, heart rate variability, hyperinsulinemia, metabolic syndrome, and constipation but not with (other) cardiac complications (QTc, myocarditis, sudden death), hematological side effects (leucopenia, neutropenia) or sialorrhea (Skokou et al., 2022).Also, in a recent prospective study in 28 inpatients with TRS, the majority could not tolerate clozapine target levels of over 400 ng/ml (Bogers et al., 2023).Taken together, these results point to increasing side effects with rising concentrations.However, there appears to be considerable between-patient variation with some patients developing debilitating side effects within the therapeutic range, and others tolerating levels well over it.This may explain the lack of a relationship between clozapine level and side effects in our study and that of Nomura.As patients were titrated without TDM but based on side effects and clinical response, those with higher tolerance may have reached higher levels, resulting on average in similar side effect levels across the concentration range.

Clozapine levels and efficacy
We observed a weak positive association of clozapine levels with disease severity and functional impairment.This is contrary to two recent systematic reviews reporting higher response rates at clozapine levels over 350 ng/ml (Siskind et al., 2021) and 407 ng/ml (Tralongo et al., 2023).The association between clozapine levels and worse functional outcomes in our sample may be attributed to a tendency of prescribers to use relatively low doses of clozapine and only up-titrate the dose when considerable remaining symptoms are present.

Strengths and limitations
The strength of the current study is that it provides real-world information on levels, effects and side effects of clozapine when titrated without TDM.Relatively high-resolution data were collected on clinical characteristics of the patients, and clinical outcomes were assessed by well-established, validated instruments.The study has several limitations.First, given the cross-sectional nature of the current study, it does not provide information on the within-patient relationship of clozapine levels with side effects and symptoms / functional impairment.For this purpose, a follow-up project is underway.Second, the inclusion of multiple centers in this study resulted in multiple clinicians performing the clinical assessments in multiple, differing patient groups.Although validated and commonly used instruments were used which the clinicians were trained for or had experience with, this has introduced between-center differences to our data.On the other hand, it increases the generalizability of our data.
Third, we used the GASS-C to evaluate the total clozapine-related side effects burden, which has several inherent drawbacks.To begin with, as it is a self-report questionnaire, its results may not always reflect objectifiable side effect burden.Additionally, it was developed to identify and measure individual side effects.It does not employ itemweighing and may therefore lack sum-score validity.However, the fact that additional analyses did not reveal a relationship between clozapine levels and any of the individual GASS-C items indicates that our results are reliable.Moreover, the fact that GASS-C is considered the gold-standard for evaluating clozapine-related side effects and has been used to evaluate the total side effect burden in previous studies, further supports our approach (de Filippis et al., 2021;Gee et al., 2017).

Conclusion
Of 129 Serbian patients titrated without TDM, only a quarter had therapeutic levels; half had subtherapeutic and the remaining quarter had supratherapeutic levels.Clozapine levels were unrelated to side effects (measured by GASS-C), and only weakly related to illness severity, disability and level of functioning (measured by WHODAS, CGI-S and GAF).These results in general support a tailormade approach of clozapine titration, to which (introduction of) TDM can contribute.First, it enables clinicians to detect patients with insufficient response with subtherapeutic levels, in whom clozapine dosages can confidently be increased to improve clinical outcome.Secondly, it can identify patients with supratherapeutic levels, in whom an attempt at dose reduction could be worthwhile.
On the other hand, the observation that side effects rates in patients with levels exceeding 1000 ng/ml were comparable to those found in the entire cohort, supports the consideration that maintaining clozapine levels above 600 ng/ml or even 1000 ng/ml may be relatively safe in selected cases.This is particularly pertinent for patients with persistent debilitating symptoms and mild side effects.Longitudinal studies focusing on within-subject clozapine level optimization (crossing these limits) are needed.

Declaration of generative AI and AI-assisted technologies in the writing process
During the preparation of this work the authors used ChatGPT 4.0 in order to improve grammar and phrasing, and to generate/correct R code.After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Fig. 1 .
Fig. 1.Relationship between side effects and clozapine levels.Horizontal trendlines signify the lack of a significant relationship between clozapine levels and GASS-C and were added for each center.Intercepts of the trendlines are derived from the fully adjusted model, with all continuous covariates set to the mean value and all categorial covariates set to the reference value.Abbreviation: GASS-C, Glasgow Antipsychotic Side-effects Scale for Clozapine.

Table 1
Patient characteristics.
H.J.de Haas et al.

Table 3
Side effects and clinical outcomes.
I Kruskal-Wallis.H.J.de Haas et al.

Table 4
Relationships between clozapine levels and clinical outcomes.

Table 5
Overview of recent studies evaluating clozapine level distributions.Our study used Dried Blood Spot analysis of capillary blood, whereas the other used venous blood samples.Abbreviations: HPLC, High-Performance Liquid Chromatography; UHPLC, Ultra-High-Performance Liquid Chromatography; MS/MS, tandem mass spectrometry.
*H.J.de Haas et al.