Psychiatric hospitalization following antipsychotic medication cessation in first episode psychosis

Abstract

Background

There are questions about the risk-benefit balance of longer-term antipsychotic medication treatment following first episode psychosis, especially in relation to relapse because of dopamine supersensitivity following treatment cessation.

Aim

The purpose of this study was to determine whether hospitalization rates in first episode psychosis patients are associated with length of initial oral antipsychotic medication exposure.

Methods

We examined psychiatric hospitalization rates in patients experiencing first episode of psychosis from the total population of Sweden between 1 January 2007–31 December 2016 (n=7043). We categorised patients by the length of first antipsychotic treatment (<6 months, 6 months to <1 year, 1 year to <2 years, 2 years to <5 years and ⩾5 years).

Results

Compared to those treated for <6 months, individuals receiving oral antipsychotic medications for ⩾5 years had less than half the cumulative incidence of hospitalization at all times between 1–4 years after treatment cessation.

Conclusion

We found no evidence that hospitalization rates increased with increasing baseline antipsychotic exposure.

Introduction

Guidelines recommend continuing antipsychotic treatment following first episode of psychosis (FEP) for 1–2 years (Correll et al., 2018). In recent years, questions have been raised about the necessity of longer-term treatment in this population (Leucht, 2018). It has been argued that prolonged antipsychotic treatment may increase rebound psychosis and relapse risk, potentially via dopamine supersensitivity (Murray et al., 2016). We are aware of only one population-based study which supports this hypothesis (Tiihonen et al., 2018) and there is limited, discrepant evidence from studies in animals and humans (Goff et al., 2017). We aimed to determine hospitalization rates in individuals with FEP following different lengths of first antipsychotic treatment (<6 months, 6 months to <1 year, 1 year to <2 years, 2 years to <5 years and ⩾5 years).

Methods

For this study we used Swedish linked national patient registers, which have complete population coverage from 1973 and 2006 for inpatient and outpatient services respectively, and the Prescribed Drug Register, which is complete from 1 July 2005 onwards. We limited our definition of FEP to individuals receiving a diagnoses of schizophrenia International Classification of Diseases (ICD-9): 295A-E, G, W, X; ICD-10: F20) and non-affective psychoses (ICD-9: 297, 298B-E, W, X; ICD-10: F21-24, F28-F29) in the National Patient Register after 1 January 2007. As we wanted to study an incident FEP cohort we only included individuals aged 16–35 years and excluded individuals who received a diagnosis of schizophrenia, non-affective psychosis or bipolar disorder, or were treated with antipsychotic medication, before 1 January 2007. We categorised first oral antipsychotic medication exposure as: <6 months, 6 months to <1 year, 1 year to <2 years, 2 years to <5 years and ⩾5 years. We used prescription dates and a three-month repeat prescription window to define periods of continuous prescribing in patients with ⩾2 prescriptions. Individuals were followed up from the date they stopped taking antipsychotic medication to the earliest of: psychiatric hospitalization, restarting antipsychotic medication, death, migration out of Sweden, or 31 December 2016. We performed Fine-Gray competing risks regression (Austin and Fine, 2017; Fine and Gray, 1999) with psychiatric hospitalization as the outcome and restarting antipsychotic medication or death as competing events. This produces a subdistribution hazard ratio, which is the difference in rate between exposure groups in patients who are event free or have experienced a competing event (reintroduction of antipsychotic medication, or death), and estimates the cumulative incidence function for hospitalization.

We adjusted for age, sex, calendar year, diagnosis (schizophrenia vs non-affective psychosis), percentage of days spent as inpatient during first antipsychotic exposure (as a proxy for severity), history of substance misuse diagnosis and antipsychotic type at baseline (using the seven antipsychotic medication neuroscience based nomenclature groupings Zohar et al., 2015). Ethical approval for the study was obtained via the Regional Ethical Review Board in Stockholm.

Results

Psychiatric hospitalization occurred in 1323 of the 7043 patients during follow-up (Figure 1, Table 1). There was an inverse relationship between duration of the patient’s first oral antipsychotic exposure and cumulative incidence of hospitalization (Figure 2). Individuals exposed to antipsychotic medication for ⩾5 years had an adjusted subdistribution hazard ratio of 0.44 (95% confidence interval 0.26–0.75) compared to those exposed for <6 months (Table 2).

Figure 1. Flow diagram of potentially included patients.

Table 1. Patient characteristics by duration of antipsychotic treatment.

Baseline characteristics	Duration of antipsychotic treatment in first episode psychosis
	<6 months	6 months to <1 year	1 year to <2 years	2 years to <5 years	⩾5 yearsa
n	3045	1838	1196	821	143
Age, median (IQR)	24.73 (21.19–29.14)	24.91 (21.02–29.69)	25.37 (21.19–29.82)	24.58 (20.70–29.12)	25.67 (22.25–29.68)
Female, n (%)	1011 (33.20)	620 (33.73)	443 (37.04)	308 (37.52)	47 (32.87)
Schizophrenia diagnosis, n (%)	638 (20.95)	478 (26.01)	356 (29.77)	318 (38.73)	76 (53.15)
Percentage of days spent as inpatient during first treatment, range	0–86%	0–82%	0–40%	0–48%	0–46%
Substance misuse, n (%)	1241 (40.76)	685 (37.27)	433 (36.20)	266 (32.40)	46 (32.17)
Second generation antipsychotic, n (%)	2689 (88.31)	1609 (87.54)	1065 (89.05)	746 (90.86)	130 (90.91)

IQR: interquartile range.

^aMean 6.34, standard deviation 1.04 years.

Figure 2. Hospitalization incidence in the first 4 years following antipsychotic cessation by length of baseline exposure, from fully adjusted competing-risks regression model.

Table 2. Psychiatric hospitalization after antipsychotic cessation.

	Duration of antipsychotic treatment in first episode psychosis
	<6 months	6 months to <1 year	1 to <2 years	2 to <5 years	⩾5 yearsa
n (%)	653 (21.44)	341 (18.55)	196 (16.39)	117 (14.25)	16 (11.19)
Follow-up time, person-years (1000 s)	5.05	2.54	1.39	0.74	0.08
Adjusted SHR (95% CI)a	1 (baseline)	0.84 (0.73–0.98)	0.73 (0.60–0.88)	0.55 (0.43–0.69)	0.44 (0.26–0.75)

CI: confidence interval; SHR: subdistribution hazard ratio.

^aAdjusted SHR and 95% CI. Adjusted for age, sex, calendar year, diagnosis (schizophrenia vs other non-affective psychosis), percentage of days spent as inpatient during first antipsychotic exposure, history of substance misuse diagnosis, and antipsychotic type (using neuroscience based nomenclature groupings). The SHR should be interpreted as the difference in rate between exposure groups in patients who are event-free (e.g. not been hospitalised) or have experienced a competing event (restarting antipsychotic medication or death).

Discussion

Duration of baseline antipsychotic treatment was inversely associated with rate of hospitalization after stopping treatment in FEP. Compared to those treated for <6 months, individuals receiving antipsychotic medications for ⩾5 years had less than half the cumulative incidence of hospitalization at all times between one and four years after treatment cessation. In all groups, hospitalization rate was most rapid in the first six months after treatment cessation.

Our results run contrary to the findings of Tiihonen and colleagues (2018), the only previous study of this type in a population cohort of which we are aware. Tiihonen et al. (2018) found lowest hospitalization rates in patients never stopping antipsychotics, rather than those stopping immediately, with increasing rates over increasing duration of treatment before cessation. An analytical difference is that they used a Cox proportional hazards model, whereas we considered the competing risk of restarting antipsychotic medication, which substantially modifies the chance of hospitalization (Fine and Gray, 1999). However, if we treat our data in the same way (censoring at the point of restarting antipsychotic medication), we find no association between baseline antipsychotic exposure length and hospitalization, so this does not explain the discrepancy. Our cohort contained no individual permanently exposed to antipsychotics and no individual stopping immediately; such individuals may be atypical FEP cases. Our competing risks approach gives a more accurate estimate of cumulative incidence of hospitalization.

Despite multivariable adjustment, our results remain vulnerable to unmeasured confounding, especially relating to confounding by indication, and time-modified confounding. Our approach may misclassify antipsychotic exposure periods, and we cannot guarantee that repeat collection of prescriptions means adherence to medication, but this would bias our results toward no effect.

Our study does not support the hypothesis that longer antipsychotic medication exposure increases the rate of relapse after stopping treatment. Although this hypothesis has been widely discussed in recent years (Leucht, 2018; Murray et al., 2016; Tiihonen et al., 2018), there is not yet consistent compelling clinical or preclinical evidence to endorse it (Goff et al., 2017).