Sleep Problems, Suicidal Ideation, and Psychopathology in First-Episode Psychosis

Abstract

Background and Hypothesis

Insomnia occurs frequently in the clinical course of schizophrenia. A growing literature has found associations between insomnia, suicidal ideation and behavior, and psychopathology in schizophrenia. We explored associations between sleep problems, suicidal ideation, and psychopathology in a cohort of patients with first-episode psychosis.

Study Design

We performed a secondary analysis of data for n = 403 subjects with data from the Recovery After an Initial Schizophrenia Episode study using regression models.

Study Results

The prevalence of sleep problems and suicidal ideation at baseline was 57% and 15%, respectively. After controlling for potential confounders, in the study baseline sleep problems were associated with increased odds of suicidal ideation with evidence of a dose-dependent relationship (OR = 2.25, 95% CI 1.15–4.41, P = .018). Over 24 months, sleep problems at any time point were associated with an over 3-fold increased odds of concurrent suicidal ideation (OR = 3.21, 95% CI 1.45–7.14, P = .004). Subjects with persistent sleep problems were almost 14 times more likely to endorse suicidal ideation at least once over the study than those without sleep problems (OR = 13.8, 95% CI 6.5–53.4, P < .001). Sleep problems were also a predictor of higher Positive and Negative Syndrome Scale total (β = 0.13–0.22), positive (β = 0.14–0.25), and general (β = 0.16–0.27) subscale scores at baseline and multiple follow-up visits (P < .01 for each).

Conclusions

Sleep problems are highly prevalent and associated with suicidal ideation and greater psychopathology in first-episode psychosis. Formal assessment and treatment of insomnia appear relevant to the clinical care of patients with psychosis as a predictor of suicidal ideation and symptom severity.

Introduction

Insomnia is common in the clinical course of schizophrenia, affecting approximately one-quarter to one-half of this patient population.^1–3 Although the etiology of insomnia is likely multifactorial, dysfunctional and/or misaligned circadian rhythms may contribute.^4,5

Insomnia is robustly associated with suicidal ideation and behavior (SIB) across multiple psychiatric conditions. There is a 2-fold increased odds of suicidal behavior in psychiatric patients with vs without comorbid sleep disturbance.^6,7 This association is particularly relevant for patients with schizophrenia, given a 27% lifetime prevalence of suicide attempt⁸ and a 5% lifetime prevalence of suicide death⁹ in the disorder. There is evidence from several studies that insomnia is associated with SIB, including suicide death, in patients with chronic schizophrenia, although only one of these studies employed a longitudinal design.^10–17 There is also some evidence for relationships between sleep disturbance and SIB in patients with early psychosis.^18,19 In a cohort of 334 patients with first-episode psychosis, Salagre et al.¹⁸ found that insomnia predicted a trajectory of worsening suicidal ideation over a 2-year follow-up period. Salavatore and colleagues found that prior sleep disturbances were a significant predictor of suicide attempts at first-psychotic episodes.¹⁹ However, a retrospective analysis of 106 subjects with prodromal psychosis found an association between sleep disturbance and suicidal ideation, but not suicide attempts.²⁰ There is also evidence for abnormal sleep architecture and reduced spindles in this patient population.²¹ Several studies have also found associations between insomnia and symptom severity in patients with chronic schizophrenia, including total, positive, and general psychopathology.^10–12 Whether this association between insomnia and symptom severity extends to patients with early psychosis is also unknown. However, there is evidence from both naturalistic and experimental studies that insomnia is associated with increased subclinical psychotic experiences.^22–25 A recent study of 740 youth at clinical high risk for psychosis found that sleep problems were significantly, positively associated with attenuated psychotic symptom severity, including suspiciousness, perceptual abnormalities, and disorganization.²⁶

Given the relative paucity of studies on insomnia, SIB, and psychopathology in early psychosis, we attempted to replicate associations found in chronic schizophrenia, as positive findings would underscore the importance of recognition and treatment of insomnia early in the course of illness. In the present study, we examined associations between sleep problems, suicidal ideation, and psychopathology in a longitudinal study of a well-characterized sample of patients with early psychosis. We hypothesized that at each time point over 2 years, current sleep problems would be associated with an increased odds of current suicidal ideation and greater levels of current psychopathology, and that insomnia at study baseline would be associated with an increased odds of suicidal ideation at follow-up visits.

Methods

Subjects and Procedures

Publicly available data from the National Institute of Mental Health Data Archive were obtained from the Recovery After an Initial schizophrenia Episode—Early Treatment Program (RAISE) early psychosis trial (n = 404), which is described in detail elsewhere.²⁷ Briefly, inclusion criteria were age 15–40; the ability to participate in research assessments in English; ability to provide fully informed consent (or assent, if age <18); and the presence of definite psychotic symptoms and evidence of DSM-IV diagnoses of either schizophrenia, schizoaffective disorder, schizophreniform disorder, brief psychotic disorder, or psychotic disorder not otherwise specified. Subjects were diagnosed using the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV by blinded, remote, centralized personnel using live, and 2-way video. Diagnoses of affective psychosis, substance-induced psychotic disorder, psychosis due to general medical conditions, clinically significant head trauma, or other serious medical conditions were excluded. All participants had experienced only one episode of psychosis (ie, individuals with a psychotic episode followed by full symptom remission and relapse to another psychotic episode were excluded) and had taken ≤6 months of lifetime antipsychotics. Subjects were followed for 24 months. The primary aim of the RAISE early treatment program was to combine state-of-the-art pharmacologic and psychosocial treatments delivered by a well-trained, multidisciplinary team in order to significantly improve the functional outcome and quality of life for patients with first-episode psychosis. This intervention, NAVIGATE, included personalized medication management, family psychoeducation, resilience-focused individual therapy, and supported employment and education. Subjects randomized to NAVIGATE (vs community care) remained in treatment longer, had greater improvement in quality of life, greater improvement in symptoms (PANSS total score mean difference 4.3 points), and greater involvement in school and work. The Augusta University IRB deemed the current study, a secondary analysis of de-identified data from the RAISE early psychosis trial, exempt.

In the present study, we investigated relationships between sleep problems over the past 2-to-4 weeks, suicidal ideation over the past 2 weeks, and psychopathology over the past week for subjects with available data from the RAISE trial at baseline and 6, 12, 18, and 24 months. Data on past suicide attempts were not available. Sleep problems at each time point were defined as the presence of terminal insomnia and/or endorsing sleep problems on the side effects scale. Hypersomnia and daytime sleepiness, which were addressed in separate questions, were excluded. Terminal insomnia over the past 2 weeks at each time point was assessed using the Calgary Depression Scale for Schizophrenia (CDSS, item #7, “EARLY WAKENING: Do you wake earlier in the morning than is normal for you? How many times per week does this happen?”). Sleep problems in the past 4 weeks were also assessed at each time point by patient self-report (“sleeping too little or problems sleeping at night”), which would also likely capture initial and/or middle insomnia. We dichotomized sleep problems as present or absent (yes/no) based on a positive response to the terminal insomnia and/or the self-report sleep problem items at each time point. Suicidal ideation at each time point was assessed using the CDSS (item #8, “SUICIDE: Have you felt that life wasn’t worth living? Did you ever feel like ending it all? What did you think you might do? Did you actually try?”). We dichotomized suicidal ideation at each time point as present/absent (yes/no). Psychopathology was measured with the Positive and Negative Syndrome Scale (PANSS) total and subscale scores at each time point. It is important to note that the PANSS does not have an insomnia item or a suicide-specific item. We included all subjects in the RAISE trial with CDSS scores at any time point. Data on age, sex, race, smoking, education, body mass index (BMI), alcohol, and cannabis use (lifetime abuse, dependence, or none). Data on baseline medications were also available. At baseline, 83.4% of subjects were taking antipsychotics (most commonly risperidone, 36.3%), 31.9% antidepressants, 5.0% sedative-hypnotics, and 12.6% no psychotropic medications.²⁸

Statistical Analysis

Data were analyzed with Statistical Package for the Social Sciences (SPSS) version 27 (SPSS, Inc.; Chicago, Illinois). Descriptive statistics (means, standard deviations, and proportions) were calculated for demographic and clinical variables. A one-sample Kolmogorov–Smirnov test was used to examine each variable for normality. We first compared sleep problems and suicidal ideation at each time point (baseline and 6, 12, 18, and 24 months) using a Chi-square test, 2-sided. We then investigated the association between sleep problems and suicidal ideation at each time point using logistic regression models. We also investigated baseline sleep problems as a predictor of suicidal ideation at 6, 12, 18, and 24 months. Next, we investigated sleep problems at any of the study visits as a predictor of concurrent suicidal ideation. Finally, we investigated whether persistent sleep problems (“yes” vs “no” at all study visits) as a predictor of suicidal ideation at any point during the study. The assumptions of logistic regression were met. Age, sex, race, smoking, education, BMI, alcohol use, PANSS total score, and depressive symptoms (CDSS total score minus insomnia and suicide items) were considered as potential confounding and/or moderating factors in these analyses.

We then investigated relationships between sleep problems and psychopathology. We first calculated bivariate correlations (Spearman’s rho) between baseline PANSS total and positive, negative, and general subscale scores and sleep problems at each time point. For any bivariate correlations with P < .10, we investigated these associations in linear regression models, controlling for potential confounding effects of age, sex, race, smoking, education, BMI, and alcohol use.

Results

The total number of subjects in the RAISE trial was 404, of whom 403 had data on baseline sleep problems and suicidal ideation. Data on sleep problems and suicidal ideation were also available for 291 subjects at 6 months, 256 at 12 months, 221 at 18 months, and 205 at 24 months. The baseline demographic and clinical characteristics of the study sample are presented in table 1. Additional information of subjects who dropped out of the study are included as supplementary material (supplementary table 1). There were no significant differences in age, sex, race, education, BMI, quality of life, past psychiatric hospitalization, or cannabis use based on subject group. However, there were significant differences in diagnosis, smoking, alcohol use, age of onset of psychosis, duration of untreated psychosis, lifetime antipsychotic exposure, CDSS total scores, PANSS total, positive, and general scores based on subject group. Subjects with (vs without) sleep problems had higher CDSS total scores with a large effect size (Cohen’s d = 0.80), and PANSS total, positive, and general scores with a small-to-medium effect sizes (Cohen’s d = 0.29–0.32).

Table 1.

Demographic and Clinical Characteristics of the Study Sample

Variable	Baseline Sleep Problem		P-value
	Yes	No
	Mean (SD)	Mean (SD)
	N = 229	N = 175
Age	23.7 (4.9)	23.3 (5.2)	.31
BMI	27.1 (7.3)	25.7 (5.5)	.16
Age of psychosis onset	18.5 (6.0)	20.0 (6.2)	.02
Duration of untreated psychosis (days)	230 (287)	146 (218)	<.01
Lifetime antipsychotic exposure (days)	38 (42)	52 (49)	<.01
CDSS total scoreb	15.0 (4.5)	11.9 (3.2)	<.01
PANSS totalb	78 (15)	74 (15)	<.01
PANSS positiveb	19 (5)	18 (5)	<.01
PANSS negativeb	20 (5)	20 (5)	.54
PANSS generalb	39 (8)	36 (8)	<.01
Clinical Global Impression (CGI)-severityb	4.1 (0.8)	4.0 (0.8)	.04
Quality of Life Scale total scoreb	52 (19)	53 (19)	.67
	n (%)	n (%)
Diagnosis			.03
Schizophrenia	113 (49.3)	101 (57.7)
Schizoaffective disorder	58 (25.3)	23 (13.1)
Schizophreniform disorder	35 (15.3)	32 (18.3)
Brief psychotic disorder	1 (0.4)	1 (0.6)
Psychotic disorder Not Otherwise Specified (NOS)	22 (9.6)	18 (10.3)
Sex			.11
Male	158 (69.3)	134 (76.6)
Female	70 (30.7)	40 (23.4)
Race			.09
Caucasian	121 (52.8)	97 (55.4)
African descent	94 (41.0)	58 (33.1)
American Indian/Alaska native	10 (4.4)	11 (6.3)
Asian	3 (1.3)	9 (5.1)
Hawaiian/Pacific Islander	1 (0.4)	0 (0.0)
Smoker			.01
Yes	130 (56.8)	77 (44.3)
No	99 (43.2)	97 (55.7)
Highest education			.71
Four-year college degree or higher	10 (4.4)	10 (5.7)
Some college	54 (23.6)	51 (23.6)
Completed high school	79 (34.5)	54 (31.0)
Attended high school	72 (31.4)	53 (30.5)
Eighth grade or less	14 (6.1)	6 (3.4)
Past psychiatric hospitalization (yes)	174 (76.0)	142 (81.1)	.21
Alcohol use			.01
Dependence	67 (29.3)	28 (16.0)
Abuse	28 (12.2)	24 (13.7)
None	134 (58.5)	123 (70.3)
Cannabis use			.48
Dependence	51 (22.3)	32 (18.3)
Abuse	34 (14.8)	23 (13.1)
None	144 (62.9)	120 (68.6)

Notes: PANSS, Positive and Negative Syndrome Scale.

^a P-values for continuous variables are from either Student’s t-test (2-sided) or Mann–Whitney U test. P-values for categorical variables are from Chi-square test.

^bPossible ranges for scores for these rating instruments are as follows:

CDS total score 0–27.

PANSS Total score 30–210.

PANSS Positive score 7–49.

PANSS Negative score 7–49.

PANSS General score 16–112.

CGI-Severity 1–7.

Quality of Life Scale (QOLS) total score16–112.

Sleep Problems and Suicidal Ideation

table 2 compares sleep problems and suicidal ideation at each time point. The prevalence of sleep problems was 57% at baseline and ranged from 40% to 48% at the follow-up visits (figure 1). The prevalence of suicidal ideation was 15% at baseline and ranged from 5% to 11% at the follow-up visits (figure 1). There were 5 subjects who reported a suicide attempt on the CDSS during the study, all of whom reported concurrent sleep problems. There were significant differences in the prevalence of suicidal ideation between subject with vs without sleep problems at baseline, 6 months, and 18 months (table 2). ln logistic regression models, after controlling for potential confounding factors, sleep problems were associated with a significant increased odds of suicidal ideation at baseline (OR = 2.25, 95% CI 1.15–4.41, P = .018) and 18 months (OR = 4.64, 95% CI 1.1–19.6, P = .037; see table 3). These associations were no longer significant after controlling for depressive symptoms. Over the 24-month study, sleep problems at any time point were associated with an over 3-fold increased odds of concurrent suicidal ideation (OR = 3.21, 95% CI 1.45–7.14, P = .014). This association was also attenuated after controlling for depressive symptoms (OR = 2.23, 95% CI 0.95–5.23, P = .066). Baseline sleep problems did not predict suicidal ideation at any individual follow-up visit. However, subjects with persistent sleep problems (“yes” at all study visits, n = 50) were almost 14 times more likely to endorse suicidal ideation at least once over the 24-month study than those without sleep problems (“no” at all study visits, n = 61) (prevalence of suicidal ideation 50% vs 8%, OR = 13.8, 95% CI 6.5–53.4, P < .001). This association remained significant after controlling for baseline depressive symptoms (OR = 7.7, 95% CI 1.8–33.5, P = .007).

Table 2.

Sleep Problems and Suicidal Ideation

Suicidal Ideation	Total	Concurrent Sleep Problem		P-value
by Visit	N	Yes n (%)	No n (%)
Baseline	403			.02
Yes		45 (40.0)	15 (25.1)
No		183 (60.0)	170 (74.9)
6 months	291			<.01
Yes		22 (17.9)	10 (6.0)
No		101 (82.1)	158 (94.0)
12 months	256			.06
Yes		11 (10.7)	7 (4.6)
No		92 (89.3)	146 (95.4)
18 months	221			.02
Yes		12 (11.4)	4 (3.4)
No		93 (88.6)	112 (96.6)
24 months	205			.17
Yes		7 (8.1)	4 (3.6)
No		79 (91.9)	145 (98.4)
Any	404			<.01
Yes		79 (25.6)	8 (8.3)
No		229 (74.4)	88 (91.7)

^a P-values are from Chi-square test.

Fig. 1.

Longitudinal changes in sleep problems and suicidal ideation.

Table 3.

Logistic Regression Models of Sleep Problems as an Indicator of Suicidal Ideation

Sleep Problem	Suicidal Ideation (CDSS Item #8)	Model 1a	P-value	Model 2b	P-value
Visit	Visit	OR (95% CI)
Baseline	Baseline	2.25 (1.15–4.41)	.018	1.38 (0.63–3.02)	.423
6 months	6 months	1.96 (0.82–4.71)	.133	0.84 (0.28–2.50)	.749
12 months	12 months	2.60 (0.84–8.11)	.099	2.33 (0.65–8.39)	.196
18 months	18 months	4.64 (1.10–19.6)	.037	2.21 (0.38–12.9)	.381
24 months	24 months	1.31 (0.30–5.68)	.718	0.14 (0.02–1.41)	.095
Baseline	6 months	1.26 (0.51–3.09)	.616	0.82 (0.27–2.48)	.730
Baseline	12 months	2.46 (0.77–7.76)	.126	1.36 (0.37–5.03)	.649
Baseline	18 months	1.47 (0.40–5.37)	.557	1.00 (0.22–4.63)	.998
Baseline	24 months	3.46 (0.68–17.7)	.136	2.56 (0.38–17.5)	.337
Any	Any (concurrent)	3.21 (1.45–7.14)	.004	2.23 (0.95–5.23)	.066
All (vs none)	Any	13.8 (6.5–53.4)	<.001	7.7 (1.8–33.5)	.007

Notes: PANSS, Positive and Negative Syndrome Scale; BMI, body mass index.

^aAfter adjustment for confounders (age, sex, race, smoking, education, BMI, alcohol use, and PANSS total score).

^bAfter adjustment for confounders from Model 1 plus depression (CDS total score minus suicide and insomnia items).

In post hoc analyses, we investigated whether the intervention in RAISE (NAVIGATE vs community care) moderated associations between sleep problems and suicidal ideation. Subjects randomized to NAVIGATE vs community care had a significantly lower prevalence of sleep problems at baseline (52% vs 62%, P = .043), and 6 (35% vs 49%, P = .015), 12 (33% vs 48%, P = .017), and 24 (34% vs 57%, P < .001) months, but not 18 months (44% vs 50%, P = .430). By contrast, there was no difference in the prevalence of suicidal ideation at any study visit based on NAVIGATE vs community care. Including treatment group in the logistic regression models did not change the pattern of findings.

Current Sleep Problems and Psychopathology

As shown in table 4, in linear regression models, after controlling for potential confounding factors, baseline sleep problems were a significant predictor of higher baseline PANSS total (β = 0.13, P = .010), positive subscale (β = 0.18, P < .001), and general subscale (β = 0.20, P < .001) scores, but not negative subscale scores. Concurrent sleep problems were a significant predictor of PANSS total and general scores at 6, 12, 18, and 24 months, and PANSS positive scores at 6, 12, and 18 months. Additional information on PANSS total and subscale scores and CDSS scores at each visit based on the presence or absence of sleep problems is included as supplementary material (supplementary table 2).

Table 4.

Linear Regression Models of Sleep Problems as a Predictor of Psychopathology

Sleep Problems by Visit	Dependent Variable PANSS score Visit		Final Linear Regression Model
			β	P-value
Baseline	Total	Baseline	0.134	.010
	Positive	Baseline	0.184	<.001
	Negative	Baseline	0.058	.242
	General	Baseline	0.199	<.001
6 months	Total	6 months	0.215	<.001
	Positive	6 months	0.235	<.001
	Negative	6 months	0.026	.673
	General	6 months	0.241	<.001
12 months	Total	12 months	0.127	.048
	Positive	12 months	0.148	.023
	Negative	12 months	−0.014	.821
	General	12 months	0.164	.011
18 months	Total	18 months	0.221	.001
	Positive	18 months	0.252	<.001
	Negative	18 months	−0.012	.870
	General	18 months	0.274	<.001
24 months	Total	24 months	0.186	.011
	Positive	24 months	0.144	.059
	Negative	24 months	−0.013	.865
	General	24 months	0.270	<.001

^aAfter adjustment for confounders (age, sex, race, smoking, education, BMI, and alcohol use). BMI, body mass index

Given these associations, in post hoc analyses, we also investigated associations between sleep problems and individual PANSS items. In linear regression analyses, after controlling for potential confounding effects of age, sex, race, smoking, education, BMI, and alcohol use, baseline sleep problems were associated with significantly greater levels of hallucinations, hostility, anxiety, guilt, tension, depression, impulse control, and active social avoidance at baseline (P ≤ .04 for each). The strength of these associations with most individual PANSS items were small (β = 0.11–0.16), although effects were slightly greater for hallucinations (β = 0.20), anxiety (β = 0.20), and depression (β = 0.28), all 3 of which remained significant after Bonferroni correction for multiple (23) comparisons (P < .001 for each).

Discussion

We found that sleep problems were a common symptom in a large study of patients with first-episode psychosis, affecting over half of subjects at baseline. Over a 2-year period, the prevalence of sleep problems ranged from 40% to 57%, and the prevalence of suicidal ideation ranged from 5% to 15%. After controlling for potential confounders: (1) sleep problems were associated with significantly increased odds of concurrent suicidal ideation at baseline and 18 months, (2) sleep problems at any time point were associated with an over 3-fold increased odds of concurrent suicidal ideation, (3) baseline sleep problems were not a predictor of incident suicidal ideation at follow-up visits, (4) subjects with persistent sleep problems were almost 14 times more likely to endorse suicidal ideation at least once over the study than those without sleep problems, (5) sleep problems were associated with higher concurrent PANSS total, positive, and general psychopathology scores at baseline and all follow-up visits, also with a dose-dependent relationship, and (6) at an individual symptom level, sleep problems were most strongly associated with levels of hallucinations, anxiety, and depression. Taken together, these findings replicate and extend previous associations between sleep problems, SIB, and psychopathology in early psychosis and chronic schizophrenia.^{10–12,18,19}

A strength of studies in patients with early psychosis is the minimization of potential residual confounding by other illness and treatment factors (eg, subjects had ≤6 months of lifetime antipsychotic exposure). The availability of longitudinal data over a 2-year period is another notable strength. Other strengths of the present study include the large sample size and the consideration of multiple potential confounding factors. We found a 27% prevalence of terminal insomnia and a 57% prevalence of sleep problems (more broadly defined), which is consistent with estimates from several recent studies in patients with chronic schizophrenia.^1–3 There was a higher prevalence of sleep problems in females vs males with first-episode psychosis, which is consistent with a diagnostic meta-analytic data.²⁹

Study limitations include the self-reported measures of insomnia and sleep problems and suicidal thinking and behavior in the RAISE trial. We considered the more broad category of sleep problems, as the CDSS only asks specifically about terminal insomnia. We included difficulty sleeping, which might capture other phases of insomnia (ie, initial and/or middle) but potentially also other sleep problems. However, we were able to exclude hypersomnia, which was assessed in a separate item (and not associated with suicidal ideation at any time point).

Although sleep problems were a significant predictor of concurrent suicidal ideation at baseline and 18 months, their association was not significant at the other visits (although the OR was in the hypothesized direction). It is possible that due to dropouts, there was inadequate statistical power to detect associations at follow-up visits. Study findings did not support our hypothesis that baseline sleep problems would be associated with suicidal ideation at follow-up visits. This implies that sleep problems at a single time point are not a viable predictor of subsequent suicidal ideation. However, we found that persistent sleep problems were a strong predictor of suicidal ideation over the course of the study. Across all analyses, we found that associations between insomnia and suicidal ideation were attenuated by controlling for depression. This is consistent with a bidirectional association between insomnia and depression, as insomnia may precede a depressive episode, but also persist after treatment for depression.³⁰

The available data did not permit consideration of medication status throughout the trial, although >80% of the study sample was taking antipsychotics at baseline. The prevalence of sleep problems was notably higher The small number of suicide attempts precluded a formal analysis of its association with sleep problems, although all 5 patients who reported suicide attempts also endorsed concurrent sleep problems. Although the CDSS items for insomnia and suicidal ideation are scored on a 4-point Likert scale and the “difficulty sleeping” item was scored on a 3-point Likert scale, we dichotomized “sleep problems” and “suicidal ideation” variables because small numbers of subjects with more severe ratings delimited comparisons across a larger number of categories. We also controlled for PANSS total score in the primary analyses, suggesting the associations between insomnia and suicidal thinking and behavior were not driven by illness acuity.

The concordance between current sleep problems and current suicidal ideation across longitudinal visits raises the possibility of treatment of insomnia as a potentially modifiable risk factor to reduce suicide risk. Clozapine, which is commonly associated with sedation, is the only antipsychotic with an indication for suicide prevention in schizophrenia from the US Food and Drug Administration.³¹ Using data from the US Food And Drug Administration Adverse Event Reporting System, we found that compared to clozapine, other second-generation antipsychotics were associated with a significantly increased reported odds ratio of insomnia and SIB as a suspected psychiatric adverse drug reaction, and these reported odds ratios were positively correlated.³² In a meta-analysis of 8 randomized controlled trials, we also found a 2.2-fold significant increased odds of insomnia in patients treated with other antipsychotics vs clozapine.³³ These findings suggest beneficial effects on sleep as one potential pathway underlying the anti-suicidal properties of clozapine (and potentially other antipsychotics).

A review of risk factors for suicide in schizophrenia found strong associations with increased positive symptoms, particularly auditory hallucinations and delusions, and low negative symptoms.⁹ This is consistent with evidence for lower suicide risk in patients with primary, enduring negative symptoms.³⁴ Concordantly, in the present study we found that sleep problems were associated with greater severity of positive but not negative symptoms. Furthermore, at the individual symptom level, we found that sleep problems were most strongly associated with the severity of hallucinations, anxiety, and depression. Thus, sleep problems may contribute to suicide risk through both positive and depressive symptoms. However, a recent survey study found that <20% of clinicians formally assessed sleep disturbances in patients with schizophrenia.³⁵ Future studies should use standard instruments in the assessment of insomnia (eg, the Insomnia Severity Index.)³⁶

Several proposed psychological and physiological mechanisms relevant to psychosis—including cognitive distortions and dysfunction, serotonergic dysfunction, hyperarousal, and hypothalamic-pituitary-adrenal axis dysfunction—may mediate the association between insomnia and suicide.^37–39 For example, sleep disturbance attributable to positive symptoms may have associated cognitive distortions.⁴⁰ Psychosis is also associated with circadian rhythm misalignment in sleep–wake and melatonin cycles, which may contribute to insomnia, and altered circadian clock gene expression.^4,5,41,42 Mechanistic studies in this patient population are warranted. These studies should assess circadian rhythm dysfunction (eg, Morningness–Eveningness Questionnaire⁴³), which is brief and easily administered, increasing its relevance to clinical practice.

The association between sleep disturbance and greater severity of psychopathology suggests that insomnia may represent an important treatment target in psychosis. Despite state-of-the-art pharmacologic and psychosocial treatments, sleep problems remained highly prevalent over the course of the trial. The present study replicates an association between sleep disturbance and the severity of anxiety and depression found in chronic schizophrenia.¹² Findings underscore the need for recognition and treatment of these symptoms in patients with schizophrenia.⁴⁴ There is evidence for efficacy of pharmacologic and non-pharmacologic treatments for insomnia in patients with psychosis,^45,46 including sedating antipsychotics, melatonin, eszopiclone,⁴⁷ and cognitive behavioral therapy for insomnia.^48–51 Both first- and second-generation antipsychotics are associated with improvements in sleep time, efficiency, and quality.^52,53 Long-acting injectable antipsychotics may also modify several key risk factors for suicide in psychosis, including adherence, psychopathology (eg, command hallucinations or agitation), and illness relapse.⁵⁴

Our findings provide additional evidence that formal assessment and treatment of insomnia and sleep disturbance is relevant to the clinical care of patients with early psychosis as a predictor of suicidal ideation and symptom severity. They also underscore the need for comprehensive suicide risk assessment in patients with early psychosis. Future, large, longitudinal studies in early psychosis are needed to replicate associations. These studies should also assess for different phases of insomnia, comorbid sleep disorders (eg, obstructive sleep apnea), and different types of suicidal thoughts.

Funding

None.

Author Contribution

Dr. Miller designed the study. Mr. Ayers and Dr. Miller managed the literature searches, analyses, and wrote the first draft of the manuscript. All authors contributed to and have approved the final manuscript.

Previous Presentation

This manuscript was presented at the Schizophrenia International Research Society virtual meeting, April 17-21, 2021.