Impact of smoking behavior on clozapine blood levels – a systematic review and meta‐analysis

Tobacco smoking significantly impacts clozapine blood levels and has substantial implications on individual efficacy and safety outcomes. By investigating differences in clozapine blood levels in smoking and non‐smoking patients on clozapine, we aim to provide guidance for clinicians how to adjust clozapine levels for patients on clozapine who change their smoking habits.


Introduction
Clozapine is the most effective antipsychotic agent for treatment-resistant schizophrenia (TRS) (1,2). Clozapine is one of the few antipsychotics where therapeutic level monitoring is used, with levels above the maximum threshold (1000 ng/ml) being associated with an increased risk of seizures (3). In clinical practice, pseudo-resistance to clozapine may occur as a result of blood levels below the minimum threshold definition (350 ng/ml) and there is an association between clozapine blood levels and response (4). Cytochrome P450 (CYP) 1A2 (CYP1A2) is the major clozapine metabolic enzyme and is responsible for approximately 70% of clozapine's metabolism (5). Clozapine is metabolized to its primary metabolite norclozapine (5). In this regard, polycyclic aromatic hydrocarbons generated by tobacco smoking induce the activity of CYP1A2 (6) which leads to increased clozapine metabolism.
Rates of smoking are up to five times higher among people with schizophrenia compared to the general population, with smoking rates among people with schizophrenia as high as 60% (7). A daily consumption of 7-12 cigarettes may be sufficient for maximum induction of clozapine metabolism (8). Beginning smoking is therefore a clinically relevant risk for relapse and inadequate response to clozapine treatment (9) and smoking cessation among clozapine users can induce severe clozapine intoxication (10). Ethnicity (Asian heritage) (11,12), gender (13), age (13)(14)(15), CYP-polymorphisms (16), caffeine (17), and comedication with clinically relevant CYP1A2-interaction (9) can also influence clozapine blood levels through the CYP-450 system. In this context, for example, people of Asian heritage are presumed to need a lesser clozapine dose compared with Caucasian or American populations (12,18) which might be due to a relatively reduced CYP1A2 activity (11). On a similar note, the clinical relevance of CYP1A2 can be observed in smoking patients after transition to electronic cigarettes, where the termination of CYP1A2 induction also induces a clinically relevant increase in clozapine blood levels (19).
Concentration to dose (C/D) ratio is a measure of clozapine clearance and higher ratios-indicating lower clearance-are associated with females, non-smokers, Asians, genetic poor metabolizers, CYP-inhibitors, obesity, inflammation and possibly with renal impairment and pregnancy (18), whereas lower ratios indicate lack of adherence or are associated with males, smokers, non-Asians, and CYP-inducers (18).
In summary, there is a lack of clarity in the available literature as to how great an influence changes in smoking habit can have on clozapine blood levels. To guide clinical care of people treated with clozapine, and to assist in averting sub-or supra-therapeutic clozapine levels and their associated deleterious effects, we conducted a systematic review and meta-analysis of the impact on changes in smoking habit on clozapine blood levels among people on clozapine. Our primary outcome of interest was impact on clozapine blood levels, with secondary outcomes of impact on clozapine to C/ D ratios, norclozapine/clozapine ratios, and norclozapine levels. We planned sensitivity analyses and meta-regression analyses on other factors which influence clozapine metabolism including ethnicity, age, and gender, to assist in clarifying the role of potential confounders, if any.

Subtitle
To investigate the impact of smoking behavior on clozapine blood levels and clozapine concentration to dose ratios in order to provide guidance for clinicians how to manage patients on clozapine who smoke or stop smoking.

Methods
The methods are based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (20). This study was registered with PROS-PERO (registration number CRD42020185579), an international database of prospectively registered systematic reviews.

Searches
Systematic searches were conducted of publications indexed in PubMed, EMBASE, and Psy-cINFO using the search terms (level OR levels OR concentration OR concentrations OR ratio OR ratios) AND (blood OR serum OR plasma) AND (clozapin* OR clozaril OR zaponex OR denzapin* OR clopine OR norclozapine OR desmethylclozapine). The abstracts and titles of articles identified through electronic searches were independently screened by two reviewers (EW, LM). Publications in all languages were considered for inclusion.

Inclusion criteria
Studies were included if they reported information on mean (and SD) of clozapine blood levels OR C/ D ratios for people with psychiatric disorders who were smoking and not smoking in cohort studies, case-control studies or randomized or non-randomized controlled trials.

Exclusion criteria
Case reports and case series were excluded.

Assessment of reporting strength
As all studies suitable for inclusion were observational studies, we used a modified Newcastle-Ottawa Quality Assessment Scale (21) for assessment of quality. The maximum score was 5, and studies with score of at least 3 were rated as highquality studies. We considered the following domains: representativeness of the sample, sample size, comparability between smokers and nonsmokers, ascertainment of clozapine blood levels (outcome) and quality of descriptive statistics (quality scores are provided in Table S2, modified Newcastle-Ottawa Quality Assessment Scale is provided in Appendix S1).

Data extraction
Two reviewers (EW, LM) independently extracted the data into an electronic spreadsheet and disagreements were resolved by joint examination of the papers. The following data were extracted: 1. Sample size of subjects on clozapine (smokers and non-smokers) 2. Mean (and SD) clozapine blood levels in both groups (ng/ml) 3. Mean (and SD) clozapine dose in both groups (in mg/day) 4. Type of comparison (within-subject or between-subject comparisons between the two groups) The following characteristics of each study were recorded where possible and if available, both for male and female participants in both groups separately: 1. Mean age of subjects in both groups 2. Ratio male:female in both groups 3. Clozapine blood levels and clozapine doses for male and female participants 4. Whether data collection was prospective or retrospective 5. Inpatient or outpatient status of subjects 6. Amount of smoking in the smoking group 7. Concentration to dose (C/D) ratio in both groups 8. Norclozapine levels in both groups 9. Norclozapine to clozapine ratios in both groups

Data synthesis and analysis
The primary outcome was the clozapine blood level (in ng/ml) in the smoker and non-smoker group. Where adequate quantitative data were not reported, corresponding authors were contacted to provide means and SDs. Where confidence intervals were reported, these were converted to SD using the Cochrane Handbook formula (17). Included studies were divided into between-subject studies and within-subject studies. Between-subject studies compared data from subjects on clozapine divided into smokers and non-smokers, whereas within-subject studies investigated effects within the same individuals as smokers and then nonsmokers. Meta-analyses were conducted using RevMan (version 5.3), and meta-regression analyses were conducted using Comprehensive Meta-Analysis (version 3.3). We assessed heterogeneity using the I 2 statistic, a measure that does not depend on the number of studies in the meta-analysis and hence has greater power to detect heterogeneity when the number of studies is small. I 2 provides an estimate of the percentage of variability due to heterogeneity rather than chance alone. An estimate of 50% or greater indicates possible heterogeneity, and scores of 75-100% indicate considerable heterogeneity. Given the observational nature of primary studies and expected high rates of heterogeneity, a random-effects model was used for all the analyses. A significance level of a < 0.05 was applied for all analyses.

Subgroup and sensitivity analyses
Subgroup analyses were undertaken for betweensubject studies for high-quality (as defined by ≥3 out of 5 points on the modified Newcastle-Ottawa Scale) vs. low-quality studies, male populations, female populations, and studies with people from Asian origin vs. studies with people not from Asian origin. Meta-regression analyses were performed for between-subject studies for clozapine blood levels and the moderators (i) difference in clozapine dose between smokers and non-smokers (ii) proportion of male smokers and (iii) proportion of male non-smokers (iv) age (smokers) and (v) age (non-smokers).

Publication bias
Where meta-analyses included at least 10 studies, publication bias was tested using funnel plot asymmetry with Kendall's Tau, where low P-values suggest publication bias. A threshold of 10 for funnel plots was chosen in accordance with the recommendations from the Cochrane Handbook (22).
Nevertheless, a higher proportion of male participants among smokers and non-smokers had increased the effect size (Q = 18, df = 2, P = 0.0001). Scatterplots for moderator variables are displayed in the Figures S19-S25. There was no evidence of publication bias (s = À0.05, P = 0.77, see Table S3).
There was insufficient relevant data to conduct sensitivity analysis or meta-regression for gender.

Discussion
This meta-analysis from a total of n = 23 studies comprising more than 7000 subjects with psychiatric disorders represents the most comprehensive analysis on the relationship of clozapine levels/ratios and smoking to date. We found that clozapine blood levels are reduced by around a third in smokers compared with non-smokers. Our analysis is the first to comprehensively combine and quantify published data on the impact of smoking on clozapine levels. For patients on clozapine who smoke and subsequently quit smoking, dosages should be decreased by 30% and clozapine blood level analyses should be performed. It is important to note that the CYP1A2 activity decrease may be a gradual process over the first three to four days after smoking cessation (44). Conversely, if patients start to smoke clozapine blood levels may fall by 30% resulting in the need to increase the dose and to monitor blood levels. In this regard, clinical signs of clozapine underdosing, such as anxiety, restlessness, and sleep disturbances must be monitored in patients who commence smoking.
In our analyses, clozapine blood levels were significantly lower in smokers. However, there may be unaccounted for confounding factors. Co-occurring caffeine intake might have increased clozapine blood levels due to CYP1A2 inhibition as previously observed in the literature (17). Smoking and caffeine use may co-occur among people with schizophrenia (45). Thus, differences in blood levels between smokers and non-smokers might have been underestimated since none of the included studies controlled for caffeine consumption. We were not able to conduct sensitivity analyses for caffeine as these data were not reported. We were only able to include information on other reported relevant CYP-450 interacting comedication.
C/D ratios were significantly lower in smokers in our analyses and our results remained significant when analyses were restricted to high-quality studies or studies from Asian origin. This allows for adjustment for certain factors associated with CYP activity such as Asian genetic heritage. However, impact of other factors associated with influence on C/D ratios (e.g., gender, poor metabolizer, obesity, and inflammation) was not able to be controlled for in our analyses. Furthermore, we were able to undertake meta-regressions examining the impact of dose, age, and gender on clozapine blood levels between smokers and non-smokers. Our findings regarding the influence of gender on clozapine blood levels with smoking should be treated with caution, since only three studies reported blood levels (33,35,39), one reported C/D ratios (36) between smokers and non-smokers for male and female populations separately and one study only included only one gender (41). In one of the studies which disaggregated data by gender and smoking status, only 14% of men were non-smokers (35), which may have skewed the results. Between-subject analyses may be biased since the included studies, with the exception of Lu et al. (30), were not fixed-dose studies and doses in smokers might have been increased by the treating clinicians. Thus, the true difference induced by smoking might be bigger than estimated, as suggested by within-subject analyses based on limited subject numbers. Furthermore, one study from Rostami et al. reported samples of smokers and non-smokers, and thus various samples might be derived from the same patients increasing the risk of bias.
Our analyses had a high degree of heterogeneity, and as such our results should be viewed with caution. We were able to conduct sensitivity analyses and meta-regression on dose, gender, ethnicity, and age. There were insufficient data to permit meta-regression for comedications, smoking quantity, genetics for fast metabolizer, caffeine use or body weight, since none of the included studies reported these confounders for smokers and nonsmokers homogenously allowing for meta-analyses. Nevertheless, comedication was considered as an item in our assessment of reporting strength and thus high-quality studies excluded clinically relevant CYP1A2 interacting comedication. Furthermore, seven out of our included studies were retrospective analyses which may be more prone to bias than prospective approaches, especially since data may not be available on medication adherence and ascertainment of smoking behavior is less certain. We attempted to address this through our sensitivity analysis of study quality.
Ours is the largest meta-analysis on clozapine blood levels and C/D ratios from a total of four studies (46). In contrast to the work from Tsuda et al., we included studies of patients of Asian heritage, assessed risk of bias and conducted sensitivity analyses and meta-regression analyses, if possible. Based on their meta-analysis on C/D ratios, Tsuda et al. estimated that if 200 and 400 mg per day of clozapine would be administered to smokers, about 100 and 200 mg per day, respectively, should be administered to non-smokers, based on a SMD C/D ratio of 1.1 (46). In our analyses, SMD of C/D ratio between smokers and non-smokers (0.7 ng/ml per mg/day) was lower than in the one from Tsuda et al. suggesting an estimated 30% reduction of dose after a patient stops smoking.

Conclusion
Our meta-analysis confirms that smoking behavior and any change in smoking behavior is associated with substantial clinical implications for patients on clozapine and extends the current knowledge by providing an evidence-based quantification of these effects. According to our analyses, reductions of clozapine dose by 30% are recommended when a patient on clozapine stops smoking. Nevertheless, reductions have to be performed with TDM of clozapine steady-state trough levels and a clinical risk-benefit evaluation since high variability between individuals (95% CI in the range of À0.55 to À0.22 in our analyses) has to be expected. Dose reductions should be combined with instruction of patient and nurses (for signs of intoxication and relapse) and a close monitoring of clozapine blood levels.