Abstract
Objective:
The objective of this study was to examine whether delays in clozapine treatment affect outcomes once clozapine is started and identify factors that affect these outcomes.
Method:
Patients starting clozapine in a four year period at South London and the Maudsley NHS Foundation Trust were included. Clinical details were gathered from clinical notes. Primary outcome was net change in inpatient admissions comparing the periods before and after clozapine was started.
Results:
There was no significant association between the length of clozapine delay (mean clozapine delay = 3.93 years) and number or length of inpatient admissions once clozapine had been started (mean net change in days of admission = 16.74 days), F value = 0.901, p = 0.345. Clozapine reduced the total number of bed days per year, but only if treatment was continued – stopping resulted in inpatient admissions returning to pre-clozapine levels. Younger patients had a greater reduction in bed days when taking clozapine (p = 0.027).
Conclusion:
Clozapine reduces the number of inpatient days, regardless of the chronicity of the illness at the time clozapine was started. Continued compliance with clozapine is necessary to maintain this benefit. Reduction in bed days is greater in younger patients, suggesting early initiation of clozapine may be beneficial.
Introduction
Schizophrenia exerts an enormous cost across multiple levels; including personal, societal and economic. It is associated with significantly increased morbidity and mortality (Barbato, 1998); life expectancy is reduced by almost 15 years (Chang et al., 2011), not only due to increased risk of suicide or violent death, but also because of the association between serious mental illness and cardiovascular disease, cancer and diabetes. There is a detrimental effect on quality of life, primarily through the associated social dysfunction, related to both cognitive symptoms and presence of psychotic symptoms. Self-care, occupational functioning and functioning in personal and community settings are affected (Janca et al., 1996), compounded by the significant social stigma attached to mental illness (NHS Health and Social Care Information Centre, 2011).
The cost to individuals of poorly or incompletely treated schizophrenia is translated into further cost to society. Direct costs include increased bed stay and more frequent and lengthy contact with outpatient and community services, including medical, social, housing and criminal justice. Large indirect costs arise from lost ability to contribute to the workforce, both for the sufferer and for their carers. The sum of these excess costs has been estimated at over $62 billion per annum (Knapp et al., 2004; McEvoy, 2007). Clearly, there is an acute need to treat schizophrenia in an effective and timely manner.
Clozapine is recommended as first-line pharmacological therapy in treatment-resistant schizophrenia (defined as a failure to completely respond to, or tolerate, two different antipsychotics prescribed for an adequate period of time at an adequate dose) by multiple local, national and international guidelines.
Despite clear evidence for the exclusive superiority of clozapine in treatment-resistant schizophrenia in both the short (Kane et al., 1988) and long term (Juul Povlsen et al., 1985; Kuha and Miettinen, 1986; Lindström, 1988), data suggest that clozapine is significantly underutilised (Kerwin and Lofts, 1993). We have shown that for inpatients at South London and the Maudsley Trust (SLaM) in London, clozapine use is delayed for an average of four years (Howes et al., 2012). This delay in offering an evidence-based treatment has clear ramifications for patient’s health, long term outcome and resource utilisation.
There is now a large body of evidence demonstrating that the delay in treating psychosis is associated with increased symptom severity, and longer delays are associated with worse symptoms (Haas et al., 1998). Clarke et al. (2006) showed that the long term consequences of a prolonged untreated psychosis include a reduced likelihood of remission and reduced functional outcome, as well as increased psychopathology. Cognitive function (Alptekin et al., 2005; Tirupati et al., 2004; Üçok et al., 2006; Vyas et al., 2007), occupational functioning (Norman et al., 2007; Penttila et al., 2011), forensic risk (Large and Nielssen, 2008), subsequent psychiatric admissions (Helgason, 1990) and even smoking habits (Kaymak et al., 2012) have all been shown to be adversely affected by a long untreated psychosis. The initiation of antipsychotic medication in the first episode of psychosis results in better outcomes (May and Tuma, 1965; May et al., 1976; May et al., 1981; Wyatt et al., 1997), and the latency of that treatment is a significant predictor of time to treatment response (Loebel et al., 1992). Importantly, a long untreated psychosis is also associated with poorer response to antipsychotic treatment once it is initiated (Perkins et al., 2005).
It is clear that the duration of untreated psychosis in first episode schizophrenia affects eventual response (both in latency and magnitude) to antipsychotics. It is not clear if delays in starting clozapine treatment with consequent periods of poorly or incompletely treated psychosis also result in an adverse effect on response to clozapine and on clinical and functional outcomes. This study aims to examine the relationship between the delay in receiving clozapine and treatment outcomes after clozapine is started, and identify exploratory variables contributing to outcomes in patients prescribed clozapine.
Method
All patients who commenced clozapine for the first time between 1 January 2006 and 15 April 2010 at SLaM were included. Ethical approval for the study was granted by the Trust Drug and Therapeutics Committee. Demographic data and pre-clozapine prescribing histories were gathered as described previously (Howes et al., 2012). Inpatient admission data, including the number and length of admissions, were retrieved from clinical notes by an experienced pharmacist, for the time period from the date of first presentation to services, to the end of the study (1 November 2014). Inpatient admissions to any psychiatric unit were included; admissions to general medical hospitals (even where the patient was assessed by a liaison psychiatry team) were excluded as the primary reason for admission was assumed to be medical rather than a psychiatric relapse.
Incidences of clozapine discontinuation were also gathered from clinical notes. Clozapine was classified as ‘discontinued’ when a deliberate switch to a different medication was made. Periods of non-compliance followed by retitration directly onto clozapine were not considered as clozapine ‘discontinuation’, even if another antipsychotic was started to temporarily ‘cover’ the retitration period. Short term use of other antipsychotics pending retitration to clozapine were also not considered as ‘discontinuation’, where it was clear from the clinical notes that the intention was to retitrate and the delay was caused only by practical or logistical reasons. Patients were classified as ‘clozapine continuers’ where they continued clozapine treatment for the entire study period with no breaks in treatment.
This study may be considered a ‘mirror image’ study, in which the ‘mirror’ was placed at the date of clozapine initiation. The length of time included in the study on either side of the mirror point (‘pre-clozapine’ and ‘post-clozapine’ data) was individualised for each patient, in order to maximise the amount of data available for analysis. The time length chosen to study for each patient was either the time from diagnosis to the clozapine start date or the time from the clozapine start date to the end of the study, whichever was longer. Where this time period extended beyond the date of first presentation to services, or beyond the end of the study, the time period was curtailed accordingly.
Where clozapine was commenced during an inpatient admission (the ‘index’ admission), this admission was excluded from analysis. The maximum length of the ‘pre-clozapine’ analysis period therefore included inpatient admissions from the date of first presentation to services to the date of the start of the index admission or the date of clozapine commencement (if clozapine was started as an outpatient). The ‘post-clozapine’ analysis period included admissions from the date of the end of the index admission, or the day after commencing clozapine (if clozapine was started as an outpatient), to the end of the study. Missing data on the length of the inpatient stay was replaced with the average length of stay for the whole cohort.
Statistical analysis was completed using SPSS version 22. Normality of data distributions was tested using Z-scores. Paired samples t-tests were used to analyse the total group data and linear regression and multivariate regression to investigate factors that affected the primary outcome. Analysis of clozapine continuers and discontinuers was completed using independent samples t-tests and binary logistic regression.
The primary outcome of net change in days of admission per year, or number of admissions per year, was calculated by subtracting the number of days (or admissions) in the time period after clozapine from the number in the time period before clozapine was started. Therefore a negative net change denoted a higher number of days or admissions post-clozapine.
Results
There were 102 patients included in the analysis; of the initial 149 patients, exclusions were 31 patients that were in forensic services, five had died, nine were lost to follow up (discharged out of the Trust) and two were excluded for other reasons (one duplicate patient, one had received clozapine before the start date of the analysis).
The average study period was 3300 days (range 90–7102 days). Data on the length of stay were unavailable for 12 of a total of 405 inpatient admissions included in the analysis (2.96%). In total, two patients were inpatients at the time of the end of the study, and for these patients the end date of the study was used as the end date of the admission (0.49%). Overall, missing data accounted for 3.46% of inpatient admissions included in the analysis.
Patient demographics.
A statistically significant difference between clozapine continuers and discontinuers – see text for details.
Delay and outcome
There was no significant association between the length of clozapine delay (mean clozapine delay = 3.93 years) and number or length of inpatient admissions once clozapine had been started (mean net change in days of admission = 16.74 days) for the entire cohort, F value = 0.901, p = 0.345, clozapine continuers (clozapine delay = 4.13 years, net change in days of admission = 24.73 days, F = 0.140, p = 0.709), or clozapine discontinuers (clozapine delay = 3.53 years, net change in days of admission = 1.42 days, F = 1.279, p = 0.266).
Relationship between delay to clozapine initiation and net change in days of admission per year. Note that a negative net change denotes a higher number of days of admission post-clozapine.
Continuers versus discontinuers
There were proportionally more men than women in the clozapine discontinuation group (77.1%) compared with the continuers group (55.2%), and this difference was statistically significant (p = 0.023). There were no statistically significant differences between clozapine continuers and discontinuers in terms of age (continuers = 39.43 years, discontinuers = 37.11 years, p = 0.235), delay in clozapine use (continuers = 4.13 years, discontinuers = 3.53 years, p = 0.533) or the total number of antipsychotic prescriptions used prior to clozapine (continuers = 5.67, discontinuers = 5.03, p = 0.438).
When the entire cohort is analysed as a single group (i.e. ignoring whether patients continued or discontinued clozapine), there was a significant reduction in the number of admissions per year (p < 0.0005) and in the length of inpatient admissions per year (p = 0.004) after clozapine was started. On average, patients spent 17 fewer days as an inpatient per year, and had 0.34 fewer admissions per year, after clozapine was started.
The reduction in the number of inpatient admissions per year after clozapine was started remained statistically significant in the clozapine continuers cohort (p < 0.0005), as did the reduction in the number of days spent as an inpatient per year (p < 0.0005). Patients who continued to take clozapine spent on average 24.7 fewer days in hospital per year after clozapine was started and had 1.4 fewer admissions per year.
However, patients who discontinued clozapine had no statistically significant reduction in bed days (1.42 fewer inpatient days per year, p = 0.906) or admissions per year (0.25 fewer admissions per year, p = 0.078). Although there was a numerically longer mean clozapine delay for clozapine continuers (4.13 years) compared to discontinuers (3.53 years), this difference was not statistically significant (p = 0.533).
For patients that discontinued clozapine, the mean time to clozapine discontinuation was 1074 days, with a wide range of 1–2841 days. Regression analysis showed no relationship between the time to discontinuation and bed days per year after starting clozapine. There was a significant relationship between the time to discontinuation and the number of admissions per year after clozapine was started (additional inpatient admissions per year = 0.123 + 0.001(time to clozapine discontinuation), R2 = 0.137).
Patient admission and prescription details.
A statistically significant difference in the measure after clozapine was initiated – see text for details.
Admission days per year pre- and post-clozapine.
Factors affecting inpatient bed days
Ethnicity, diagnosis, gender and the number of antipsychotics used prior to clozapine initiation were not statistically significantly associated with the net change in the number of days of admission or the number of admissions per year after clozapine was started. Multivariate analysis of variance indicated that age did have a significant effect on the length of time spent in hospital after clozapine was started (Θ = 0.81, F(5, 20) = 3.22, p = 0.027). Discriminant analysis showed that as age increases, the net change in the number of days of admission per year decreases. This effect was seen for clozapine continuers, but not clozapine discontinuers. Further discriminant analysis of clozapine discontinuers was considered unhelpful due to low patient numbers. Two outlying patients at the extremes of age (>59 years) were removed from this analysis.
Effect of age on net change pre- and post-clozapine initiation in inpatient bed days and admissions per year for clozapine continuers.
Discussion
This study investigated the effect of delay to clozapine treatment on clinical outcomes once clozapine is started, using inpatient admissions as the primary outcome. We hypothesised that a longer delay to commencing clozapine may be associated with a less favourable clinical outcome, based on observations in first episode psychosis where longer durations of untreated psychosis are associated with poorer eventual response to treatment. We also investigated the effect of continuing or stopping clozapine treatment on long term outcomes.
Our study found no association between the length of clozapine delay and the number of inpatient admissions, or days spent as an inpatient per year, once clozapine had been commenced. We found, in common with previous authors, that clozapine reduced the number of bed days per year. We have also shown that this positive effect is maintained only if the patient continues to take the clozapine – if the treatment is stopped, then the number of days spent as an inpatient returns to the level it was before the clozapine was started.
Two previous studies using the same methodology for severity of illness also found no statistically significant association between the delay to clozapine use and the number of inpatient admissions after clozapine was started (Harrison et al., 2010; Nielsen et al., 2012). A recent analysis by Ucok et al. (2015) used clinician, patient and carer’s opinions to assign patients to groups of ‘good’ response and ‘minimal or no’ response. Using these criteria, they found that a ‘good’ response to clozapine was associated with a shorter delay to clozapine initiation, when compared to with those in the ‘minimal or no’ response group. This result may reflect some degree of selection bias, as they also found that ‘good’ responders were more likely to have had fewer admissions prior to clozapine commencement, suggesting perhaps that these patients suffered a less severe illness course than their counterparts in the ‘minimal or no’ response group.
Our study, whilst of a smaller population than those published previously, benefits from wider inclusion criteria and longer follow up times. We did not exclude patients who were in the first few months of treatment, who were acutely unwell, were current inpatients, had received clozapine for less than a defined period of time or were taking less than a defined number of milligrams of clozapine per day, as other studies have done (Harrison et al., 2010; Nielsen et al., 2012; Ucok et al., 2015). Additionally, we have demonstrated the importance of separate analysis of patients who discontinue clozapine, again not considered by other authors.
Beyond continuing to take the medication, the only other factor we found to affect bed days was age, where we found that younger patients tended to have proportionally fewer bed days after clozapine was started. This benefit to clozapine reduced as age increased. Ucok et al. (2015) also found that younger patients were more likely to have a ‘good’ response to clozapine. It is possible that older patients are more likely to have experienced more relapses during their illness course, and more relapses have previously been shown to have an adverse effect on long term outcome (Altamura et al., 2007). Conversely, younger age has also been associated with an increased likelihood of relapse (Altamura et al., 2007).
This study has important implications for prescribers, who can be reassured that clozapine is expected to bring a clinical benefit to their patients irrespective of the length of time they may have spent during their illness taking other antipsychotics. Clozapine should not be withheld from patients even if they have been in contact with mental health services for extended periods of time. We have also shown that the benefits to clozapine only remain if patients continue to take the drug. Future research should address reasons for clozapine discontinuation and strategies for maintaining therapy, as the consequences of ceasing clozapine have clear implications not only for the patient but also for service resource management.
There are several factors which need to be considered with respect to the study design. The use of inpatient admission data as a proxy marker for the mental state of a patient may be inaccurate. Patients may be admitted to psychiatric units for a range of reasons that may not reflect a relapse in mental state, including social reasons. In clinical practice however we feel this is rare, and in our population in South East London the illness severity required for inpatient admission, given the continued pressure on inpatient bed resources, is generally considered to be high. For these reasons we think it unlikely that patients in our cohort were admitted for reasons other than a relapse in mental illness. We also believe that inpatient bed days are an important outcome in their own right, given the impact of inpatient admissions on scarce health system resources. Conversely, it is possible that patients were not admitted despite a psychiatric relapse. We did not include input from community-based home treatment or crisis teams, and it is possible that as a result, inpatient admission data was underestimated. However, it is expected that whilst to some degree access to inpatient services may have changed over the course of the study, in general the severity of illness that would prompt inpatient admission is likely to have remained broadly the same, and so this possible source of underestimation of illness should have remained the same for every patient during the course of the study.
Further, it is possible that the length of inpatient admissions may have been affected by other factors outside mental state – social factors may play a part. The ‘mirror image’ nature of this study, using each patient as their own comparator, is expected to minimise the impact of this on the overall results since, for each individual, social and other factors are likely to remain broadly the same over the study period. There are also difficulties in assessments based on retrospective clinical note review – we have attempted to overcome these by interrogating multiple data sources where available, including clinic letters and pharmacy dispensing systems. Finally, and as noted above in relation to the study conducted by Ucok et al. (2015), it is possible that patients who were started on clozapine earlier were selected to do so by clinicians because they were thought more likely to respond or more likely to comply with treatment. It is therefore possible that patients that started clozapine earlier are systematically different to those who started later.
In conclusion, we have shown that taking clozapine results in a reduction in the number of days spent as an inpatient per year. Patients experience this reduction in bed days regardless of the chronicity of their illness. However, this benefit lasts only as long as the drug is being taken, with people who stop taking clozapine returning to their pre-clozapine levels of inpatient bed days. Clinicians can be confident that prescribing clozapine will confer a positive outcome to their patients regardless of when in the illness course it is started, and they should continue to make every effort to help their patients comply with treatment to maintain this benefit. Further, although we have shown no effect of delay to clozapine initiation on long term outcomes, we have also shown that the benefits to clozapine administration may be attenuated with advancing age, suggesting that early initiation of clozapine in the illness is to be recommended.
Footnotes
Declaration of conflicting interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: SG has received lecturing honoraria from Sunovion, Janssen and Otsuka, and consultancy fees from Sunovion. SS declares no conflicts of interest. DT has served on advisory boards for Lundbeck, Servier and Sunovion, received lecturing fees from Janssen, Otsuka, Servier and Lundbeck, and research funding from BMS, Janssen, Lundbeck and Sunovion.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: SSS was supported by a European Research Council Consolidator Award and by the National Institute for Health Research (NIHR) Mental Health Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London.