Introduction

Clozapine, a second-generation antipsychotic (SGA), is the most effective antipsychotic for treatment-refractory schizophrenia (TRS) [1, 2], especially in reducing positive psychotic symptoms [1]. Approximately 30% of patients with schizophrenia respond poorly to antipsychotics and are thus clinically categorized as TRS [3]. Clozapine is well-known for its neutropenic side effects. Among the clozapine-induced adverse drug reactions, neutropenia is particularly dangerous because it may eventually develop into agranulocytosis, a potentially life-threatening condition due to high risk of infection [4]. The prevalence of neutropenia (absolute neutrophil count [ANC] < 1500/mL or white blood count [WBC] < 3000) and agranulocytosis (ANC < 500/mL) caused by clozapine was determined to be 2–3% and 0.4%-0.7%, respectively [5,

Results

Demographic and clinical data

A total of 1084 patients were included in our analysis. Among them, 796 (73.4%) patients were diagnosed with schizophrenia, 149 (13.7%) with schizoaffective disorder, 78 (7.2%) with bipolar disorder, and the remaining 61 (5.7%) with other psychiatric disorders (Table 1).

Table 1 Major diagnoses of patients receiving clozapine in the present study
Full size table

Table 2 reveals the patients’ demographic characteristics and medications at discharge. Among the included patients, 539 (49.7%) were men. The average age was 41.06 ± 12.81 (mean ± standard deviation) years and average daily clozapine dosage was 242.08 ± 148.86 mg. Concomitant antipsychotics were prescribed to 325 (30.0%) patients, with 245 (22.6%) receiving FGAs and 86 (7.9%) of them receiving SGAs. Antidepressants were prescribed to 164 (15.1%) patients. Ten (0.9%) patients were prescribed with mood stabilizers other than sodium valproate and lithium. Regarding other concurrent medications, 376 (34.7%) patients in the cohort were taking anticholinergics, 559 (51.6%) were taking benzodiazepine, 378 (34.9%) were taking antihypertensives, 29 (2.7%) were taking antidiabetics, 23 (2.1%) were taking antihyperlipidemics, and 22 (2.0%) were taking antiepileptics (Table 2).

Table 2 Demographic characteristics and medications: bivariate analysis
Full size table

Prevalence of neutropenia, leukocytosis, and related factors

A total of 501 events of neutropenia and leukocytosis occurred among all participants during the follow-up period. Specifically, 2.9% of patients in the CLO group and 8.8% of patients in the CLO + VAL group developed neutropenia, whereas 39.6% of patients in the CLO group, 42% of patients in the CLO + VAL group, 71.1% of patients in the CLO + Li group, and 50% of patients the CLO + VAL + Li group developed leukocytosis. Concerning neutropenia, Fisher’s exact test demonstrated that the CLO + VAL group had a significantly higher ratio of neutropenia (p < .001; Table 3). There was no significant difference in the incidence of neutropenia between the group receiving CLO + VAL and the group receiving CLO + Li. In this study, we did not observe any neutropenia event in the CLO + Li group or in the CLO + VAL + Li group. As for leukocytosis, concomitant lithium administration was associated with a higher ratio of leukocytosis event (p < .001; Table 3). We also found seven cases (five in the CLO group and two in the CLO + VAL group) who had developed both neutropenia and leukocytosis during the 3-year follow-up. However, the ratios were not significantly different across the four groups (Table 2).

Table 3 Neutropenia and leukocytosis: p value obtained by Fisher’s exact test (2 × 2 table)
Full size table

Result of logistic regression

After adjustment for covariates (i.e., age, sex, hematological events, daily dosage of clozapine, concomitant psychotropics, and other medications) through logistic regression, the CLO + VAL group (odds ratio [OR] = 3.49, p < .001, compared with the CLO group) and age (OR = 1.03, p = .007) were significantly associated with an increased risk of neutropenia. The CLO + Li group (OR = 3.39, p < .001, compared with the CLO group), younger age (OR = 0.97, p < .001), clozapine daily dosage (OR = 1.01, p = .025), and benzodiazepine use (OR = 1.64, p < .001) were significantly associated with leukocytosis development (Table 4).

Table 4 Multivariate logistic regression analysis of neutropenia and leukocytosis (final model after variable selection of backward elimination)
Full size table

In addition, in the two formulas presented in Table 3, individual multicollinearity diagnostics revealed that all variables involved in the first step of logistic regression were nonsignificant due to multicollinearity, meaning that every variable had a variance inflation factor (VIF) < 1.5.

Trend for neutropenia and leukocytosis according to age, dosage of valproate, and dosage of lithium

To explore whether there was an increasing or decreasing trend of develo** neutropenia and leukocytosis according to age, dosage of valproate, and dosage of lithium, we conducted the Cochran–Armitage test, wherein, individuals were stratified by age (i.e., aged < 30, aged 30–39, aged 40–49, aged 50–59, aged ≥ 60 years). With the group of < 30 years as the reference, neutropenia development did not differ significantly among the subgroups. However, with an increase in age, a trend of increased neutropenic events was observed (p = .039; Fig. 1A). In addition, to examine the relationship between valproate daily dosage and neutropenia, we classified the valproate dosage into three subgroups (i.e., dosage ≤ 1000 mg, 1001–2000 mg, and > 2000 mg), with dosage ≤ 1000 mg serving as the reference group. We found no significant difference in each subgroup nor a trend of decreasing or increasing incidence of neutropenia with an increase in valproate dosage (p = .808; Fig. 1C).

Regarding leukocytosis, compared with the reference group (< 30 years), those aged 40–49, 50–59, and ≥ 60 had significantly lower risks of develo** leukocytosis. Moreover, a decreasing trend of leukocytosis development was observed with increasing age (p < .001; Fig. 1B). Regarding lithium dosage, with a dosage of < 600 mg as the reference group, each of the other two subgroups (i.e., 601–1200 mg and > 1200 mg) demonstrated no significant difference in leukocytosis development; however, an increasing trend of leukocytosis incidence with an increase in lithium daily dosage was observed (p < .001; Fig. 1D).

Fig. 1
figure 1

Linear trend for neutropenia and leukocytosis, stratified by age, valproate use, and lithium use

Full size image

Adjusted variables include clozapine dosage, sex, antipsychotics use, antidepressants use, anticholinergics use, benzodiazepine use, antihypertensives use, antidiabetics use, antihyperlipidemics use, and antiepileptics use.

Relationship between the covariates

The Spearman correlation matrix in Supplementary Fig. 2. reveals that the correlations between many covariates were nonsignificant, and a few covariates were significantly but weakly correlated, as evidenced by the correlation coefficients, which were all less than 0.3. Therefore, the correlation between the covariates in our analysis may not affect our main outcome.

Discussion

Our study found a significantly increased risk of valproate-related neutropenia and lithium-related leukocytosis in patients receiving clozapine. These findings were independent of other variables, such as age, sex, daily dose of clozapine, and other concomitant medications. Younger age, male sex, and concurrent benzodiazepine use might be linked to leukocytosis, and older age is a risk factor for neutropenia. Lithium dosage increase is related to an upward trend of leukocytosis development, but such a trend was not noted between valproate dosage increase and neutropenia development.

On the basis of our literature review, we enrolled a relatively large number (N = 1084) of patients compared with other studies on valproate-related neutropenia and lithium-related leukocytosis. Our research data were from one large psychiatric institution, and the data were collected between 2006 and 2020.

We observed that valproate coadministration is a significant risk factor (OR = 3.49) for the development of clozapine-associated neutropenia. This finding was consistent with those of other studies [16, 26, 27]. Clozapine by itself would induce neutropenia; however, the precise underlying mechanisms remain unclear. One popular hypothesis is that the bioactivation of clozapine produces a chemically reactive nitrenium ion, which causes apoptosis to neutrophils as well as toxicity to stromal cells, the precursors of neutrophils in bone marrow [28]. Clozapine is metabolized by CYP1A2 and CYP3A4, with CYP1A2 playing a major role [26,27,28]. Clozapine metabolism through CYP1A2 is competitively inhibited by valproate [29]. Additionally, valproate by itself may induce blood dyscrasias by affecting the differentiation of normal multipotent hematopoietic progenitors. The risk of blood dyscrasia is higher when valproate is used in combination with clozapine [30]. In contrast to the mentioned UK study [16], we did not find a dose-dependent effect of sodium valproate on neutropenia development.

In the present study, another mood stabilizer, lithium, increased the risk of leukocytosis in patients treated with clozapine (OR = 3.39). The precise mechanism by which lithium increases WBC count remains unclear. Putative theories include an increase in granulocyte production, direct stem-cell stimulation, cytokine stimulation, redistribution of demarginated leukocytes, and increased cortisol production [31]. Lithium has been used to treat clozapine-induced neutropenia [21, 32] and to prevent neutropenia during clozapine rechallenge [33], with a minimum serum lithium level of 0.4 mmol/L required for such effects [34]. Nevertheless, prescribing clozapine to a patient already under lithium may result in leukocytosis [46, 47]. Future studies are warranted to confirm our findings.

A slightly increased risk of leukocytosis was observed with higher clozapine doses (OR = 1.01, 95% confidence interval: 1.00–1.02). Leukocytosis, a generally asymptomatic and benign condition [33], appears to be a potential side effect of clozapine [36]. The mechanisms of clozapine-induced leukocytosis may be related to changes in plasma concentrations of the granulocyte colony-stimulating factor, tumor necrosis factor-α, interleukin (IL)-2 cytokines, and IL-6 cytokines [33]. Moreover, a downward dose titration results in normalization of WBC counts, suggesting a dose-dependent effect of clozapine [48] .

Our study did not replicate a recent finding in the UK demonstrating a reduced risk of neutropenia in patients cotreated with SSRIs [16]. Although the study reported reduced risks of neutropenia in patients treated with SSRIs, a large drug surveillance study with 122,000 psychiatric inpatients [49], which had a similar finding as ours, did not support this finding. In addition, the use of concomitant antidepressants such as mirtazapine and imipramine was associated with neutropenia development. Evidence does not support SSRIs or serotonin–norepinephrine reuptake inhibitors causing neutropenia in patients treated with clozapine [50].

A recent study in Japan involving 3746 patients treated with clozapine reported that 4.9% of the patients developed neutropenia. In our study, the incidence of neutropenia was 2.9%, which was consistent with the 2–3% prevalence described in the literature [5]. Additionally, a recent systematic review reported significantly lower clearance of clozapine in East Asian patients compared with Caucasians [51]. Poor metabolism of clozapine was noted in approximately 10% of Asians; therefore, they needed lower doses to reach therapeutic concentrations [52]. Because of this, Asian patients may be more sensitive to blood dyscrasia (neutropenia or leukocytosis). Data on the concomitant use of clozapine with valproate and lithium in Asian patients are limited. Therefore, a strength of our study is the identification of incidences of neutropenia and leukocytosis between patients treated with clozapine monotherapy alone and subgroups of patients on clozapine cotreated with mood stabilizers.

Our study has some limitations. Because this was a retrospective cohort study, some important predictive factors may have been underestimated. Moreover, because the groups were not randomly assigned, some threats to internal validity may have occurred. Although we analyzed participants’ basic demographic data and common concurrent medications, other factors such as smoking [11], oral contraceptive use [29], obesity [53], and inflammation [54] may have affected the WBC counts. Currently, no clinical registry comprises data on physical conditions and treatments from both psychiatric and general hospitals. Furthermore, because the included patients may have irregular outpatient clinic visits, routine hematologic examination may not be regularly performed. We also did not include the serum levels of both valproate and lithium for our patients at discharge. Additionally, clozapine adherence was not objectively confirmed because monitoring clozapine levels is not the standard practice in Taiwan.

Conclusion

Our study demonstrated that concomitant administration of valproate with clozapine was significantly associated with the neutropenia three-fold risk increase. Concomitant use of clozapine with lithium was associated with three-fold higher risks of leukocytosis. Therefore, regular follow-up hematology profiles are warranted in patients who received combined treatment with clozapine and valproate/lithium.