Rest-activity rhythm disruption and metabolic health in schizophrenia: a cross-sectional actigraphy study of community-dwelling people living with schizophrenia and nonpsychiatric comparison participants

Abstract

Study Objectives:

People living with schizophrenia (PLWS) have increased physical comorbidities and premature mortality which may be linked to dysregulated rest-activity rhythms (RARs). This study aimed to compare RARs between PLWS and nonpsychiatric comparison participants (NCs) and to examine the relationships of RARs with age, sleep, metabolic, and physical health outcomes and, among PLWS, relationships of RARs with illness-related factors.

Methods:

The study sample included 26 PLWS and 36 NCs, assessed with wrist-worn actigraphy to compute RAR variables and general sleep variables. Participants completed assessments for clinical symptoms, physical health, sleep quality, medication use, and assays for fasting glycosylated hemoglobin (hemoglobin A1c) levels. We examined group differences in RAR and sleep variables, relationships of RAR variables with metabolic and physical health measures, and, among PLWS, relationships between RAR variables and illness-related measures.

Results:

PLWS had significantly shorter active periods, lower relative amplitude, and lower mean activity during their most active 10 hours compared to the NCs (Cohen’s d = 0.79, 0.58, and 0.62, respectively). PLWS had poorer sleep quality, greater mean percent sleep, less wake after sleep onset, and higher total sleep time variability compared to NCs. PLWS had higher rates of antidepressant, anxiolytic, and antipsychotic medication use compared to NCs, which may have impacted sleep quality and objective sleep measures. Across both groups, more fragmented and variable RARs were associated with higher HbA1c levels (η_p² = .10) and worse physical health (η_p² = .21). Among PLWS, RARs were correlated with total sleep time (r_s = .789, P < .01) and percent sleep (r_s = .509, P < .05), but not with age, sleep quality, or other illness-related factors.

Conclusions:

RARs provide unique information about sleep and activity for PLWS and have the potential for targeted interventions to improve metabolic health and mortality.

Citation:

Mahmood Z, Ramsey A, Kidambi N, et al. Rest-activity rhythm disruption and metabolic health in schizophrenia: a cross-sectional actigraphy study of community-dwelling people living with schizophrenia and nonpsychiatric comparison participants. J Clin Sleep Med. 2024;20(9):1505–1516.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Rest-activity rhythm dysregulation is a particularly important and understudied health-related factor for people living with schizophrenia and may be linked to the disproportionately high rates of physical comorbidities and premature mortality seen in schizophrenia. The current study examined differences in actigraphy-derived rest-activity rhythms among people living with schizophrenia and nonpsychiatric comparison participants and characterized the relationships of rest-activity rhythm variables with physical health, sociodemographic variables, and clinical variables.

Study Impact: This study extends our understanding of the full clinical syndrome of rest-activity rhythm disruption among people living with schizophrenia to better direct health care resources and inform targeted interventions to improve metabolic health. Consideration of sleep disturbances and medication use are important cofactors of rest-activity rhythms and physical health.

INTRODUCTION

Despite modern advances in health care driving the longevity revolution in the general population, premature mortality has persisted in people living with schizophrenia (PLWS).^1–4 Furthermore, the mortality gap has widened since the 1970s⁵ and has been attributed to high rates of metabolic and cardiovascular disease among PLWS.^6–8 Lifestyle interventions to improve metabolic and cardiovascular health have been limited due to the high prevalence of unhealthy lifestyles (eg, smoking, poor diet, sedentary habits), suboptimal health care perpetuated by social stigma against mental illnesses, adverse effects of antipsychotics, and biological factors (eg, accelerated aging).^8–13 Thus, to increase longevity among this high-risk population, identification of novel and modifiable cardiometabolic risk factors is needed.

One potentially modifiable and understudied contributor to cardiometabolic health is dysregulation of the 24-hour rest-activity cycle.¹⁴ Rest-activity rhythms (RARs) are the 24-hour pattern of resting and being active, with many quantifiable characteristics such as amplitude, regularity, timing, and durations. Earlier-onset of the active period has been associated with elevated blood pressure among people living with bipolar affective disorder.¹⁵ Among the older adult population, individuals with more irregular RARs had higher odds of developing metabolic syndrome, independent of obesity.¹⁶ While the link between metabolic health and RARs has been established in the general population, it has not been well studied among PLWS, despite their high-risk group for metabolic health diseases.

In addition to RAR dysregulation, disrupted and variable sleep are also associated with poor metabolic outcomes. One large and diverse study of community-dwelling United States adults found that greater variability of sleep onset times and sleep duration were associated with worse central obesity, lower high-density lipoprotein cholesterol, and higher fasting glucose.¹⁷ Longer sleep duration (> 8 hours) has been associated with higher body mass index, waist circumference, and weight among individuals with serious mental illness who were treated with antipsychotic medications.¹⁸ Among PLWS, eveningness chronotype has been associated with poor self-reported sleep quality, and poor self-reported sleep quality has been linked to higher total and low-density lipoprotein cholesterol.¹⁹ Few studies have examined the influences of both RAR variables and sleep measures on health among PLWS.

RAR dysregulation is a particularly important and understudied health-related factor for PLWS. RAR dysregulation is estimated to occur in nearly 80% of PLWS with considerable heterogeneity in patterns,²⁰ has genetic links to schizophrenia phenotypes,²¹ and is a common prodrome of schizophrenia.²² Extant research shows that schizophrenia symptom severity is associated with the extent of circadian rhythm disruption.^22–24 The impact of antipsychotic medications on RAR regulation remains inconclusive given mixed findings within a limited literature and evidence of RAR disruptions in schizophrenia regardless of medication status.^20,22,25,26 Sleep patterns and RARs become weaker and more fragmented with aging,²⁷ thus PLWS may be at a further disadvantage as they age due to the cumulative effects of aging and chronic RAR dysregulation. Another potential contributor to abnormal RAR is obstructive sleep apnea (OSA). While one study noted lower intradaily stability among patients with OSA,²⁸ other studies did not find differences in circadian rhythms among people with OSA based on dim-light melatonin onset,²⁹ repeated measures,^30,31 or cortisol trends.^31,32 RAR dysregulation is one proposed mechanism linking OSA and metabolic disorders,^33–35 although OSA has been linked to poor metabolic outcomes independent of RAR dysregulation—through microvascular changes, inflammation, oxidative stress, adipocyte dysfunction, and gut dysbiosis.^36,37 In this paper, we first examine the role of RARs on metabolic health, with the aim of examining OSA and other contributing factors for dysregulated RARs in future studies. RAR disruption has an established strong, bidirectional association with adverse neurobehavioral and physiological symptoms within general and other clinical populations,³⁸ although links to sociodemographic and clinical factors among PLWS warrant further exploration.

Beneficial effects of chronotherapies in both general and clinical populations suggest RARs are modifiable; however, current evidence from clinical trials remains mixed.^39–48 Tools that can sustainably and reliably manipulate RARs with the specificity demanded by the complex alterations in psychiatric illness remain elusive. As such, understanding modifiable factors related to circadian disruptions in schizophrenia may help identify those at greatest risk for poor health outcomes and uncover additional treatment targets.

Understanding the full clinical syndrome of RAR disruption among PLWS could better direct health care resources and inform targeted interventions to improve metabolic health. As such, the purpose of the current study is to examine differences in actigraphy-derived RARs (particularly activity and timing of most/least active periods, relative amplitude (RA), and stability and fragmentation of rhythms) among PLWS and nonpsychiatric comparison participants (NCs) and characterize the relationships of RAR variables with physical health, sociodemographic variables, and clinical variables. We hypothesize that PLWS would have more dysregulated RARs compared to NCs, and that dysregulated RARs would be associated with negative metabolic and physical health outcomes among both PLWS and NCs. We will explore the relationships of RAR variables with age, sleep, and, among PLWS, illness-related factors (eg, antipsychotic medication dosing, severity of positive and negative symptoms).

METHODS

Participants

The current study is a secondary analysis of baseline data from a larger, prospective study of accelerated aging in schizophrenia that was approved by the University of California San Diego Human Research Protections Program (Protocol #101631). All study participants provided written informed consent to participate. The aims of the present study are novel from prior published works utilizing this dataset, and current analyses were restricted to participants who had available actigraph data (detailed parameters described below).⁴⁹

We recruited English-speaking participants from the greater San Diego area who had either a diagnosis of schizophrenia or schizoaffective disorder or no history of major psychiatric illness. The participants with schizophrenia or schizoaffective disorder were screened using the Diagnostic and Statistical Manual of Mental Disorders, fourth edition - text revision Structured Clinical Interview⁵⁰ (as the Diagnostic Statistical Manual, fifth edition criteria had not yet been published). The NC participants were screened with the Mini International Neuropsychiatric Interview.⁵¹ Study exclusion criteria included: other present Diagnostic and Statistical Manual of Mental Disorders, fourth edition - text revision Axis I diagnoses; nontobacco-related substance abuse or dependence within the last 3 months; diagnosis of intellectual disability, major neurological disorder, or dementia; and/or a medical disability that would impede the participant’s ability to complete research assessments.

Sociodemographic and clinical characteristics

Sociodemographic (sex, age, race/ethnicity, education, and smoking status) and schizophrenia-related clinical factors (age of onset, duration of illness, and antipsychotic medication dosages) were gathered during individual interviews in combination with review of medical records (with Health Insurance Portability and Accountability Act authorization). Daily dose equivalents of antipsychotic medications were calculated according to World Health Organization guidelines.⁵²

Sleep

We assessed objective sleep measures with a wrist-worn actigraph (Actisleep-BT; Actigraph, Pensacola, Florida) worn on the nondominant hand in combination with nightly sleep diaries of bedtimes and wake times. The raw data were processed using validated algorithms for total sleep time (TST), percent sleep (percent of total time in bed that is spent sleeping), wake after sleep onset (total duration of overnight awakenings after initially falling asleep), sleep-onset time (time at which sleep begins), and sleep-offset time (time at which sleep ends).⁵³ We calculated mean and variability (root-mean square successive difference) measures of the sleep measures.

RAR assessments

We assessed 24-hour sleep-wake activity patterns (minimum of 3 consecutive 24-hour periods) using a panel of variables from both nonparametric⁵⁴ and extended-cosine modeling approaches.⁵⁵ Through the nonparametric methods using R Statistical Software (v4.1.3; R Core Team 2021; Vienna, Austria) and the R package “nparACT,”⁵⁶ we assessed interdaily stability (IS), fragmentation of rhythm, or intradaily variability (IV), activity level during the most active 10 hours (M10), and activity level during the least active 5 hours (L5), and RA, which is the difference between M10 and L5 in the average 24-hour pattern and represents overall robustness of the RAR. We conducted extended-cosine modeling using the R package “RAR”⁵⁷ to generate estimates about activity-rhythm timing and shape, specifically active-period length, up-mesor (t-left or start-time of the active period), and down-mesor (t-right or start-time of the resting period). To reduce the number of RAR variables examined, we report on the active period duration (difference between down-mesor and up-mesor), IS, IV, RA, M10, and L5.

Psychiatric symptom severity and general mental health

Self-reported mental health was assessed through the mental health composite score from the RAND Medical Outcomes Study 36-Item Short Form Health Survey,⁵⁸ with higher scores indicative of better self-reported mental health. Depressive symptom severity was measured with the 9-item Patient Health Questionnaire,⁵⁹ with higher scores indicating greater severity of depression. We assessed anxiety using the Brief Symptom Inventory – Anxiety,^60,61 with higher scores indicating worse anxiety symptoms. The severity of positive and negative symptoms among the PLWS group was evaluated with the Scale for the Assessment of Positive Symptoms and the Scale for the Assessment of Negative Symptoms.⁶²

Physical health measures

Self-reported physical health was assessed through the physical health composite score from the 36-Item Short Form Health Survey,⁵⁸ with higher scores indicative of better self-reported physical health. The presence and severity of common medical comorbidities were assessed with the Cumulative Illness Rating Scale total score (range: 0–56, higher scores indicating a greater number of conditions and higher severity of illness).^63,64 Participant’s height and weight were collected to determine their body mass index. Fasting hemoglobin A1c (HbA1c) levels, a standard clinical measure of hyperglycemia, were assayed at the University of California San Diego Hospital laboratory using standard laboratory assays. Higher HbA1c levels suggest worse blood sugar control and higher risk of diabetes and related complications. Finally, the number of metabolic syndrome criteria was based on the American Heart Association/National Heart, Lung, and Blood Institute criteria: elevated waist circumference (≥ 102 cm for men, ≥ 88 cm for women), elevated triglycerides (≥ 150 mg/dL or on medications for elevated triglycerides), elevated high-density lipoprotein cholesterol levels (< 40 mg/dL for men, < 50 mg/dL for women), or elevated blood pressure readings (systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 85 mmHg or on medications for hypertension).⁶⁵

Statistical analysis

All variables were assessed for normality prior to analyses. HgA1c levels were log-transformed for all analyses to reduce variance and improve efficiency. Sociodemographic characteristics and clinical variables were summarized, and differences were compared between the 2 diagnostic groups using independent sample t tests and χ² tests.

General linear models were conducted to assess the association of RAR variables with metabolic and physical health outcomes, while controlling for age, diagnostic group, and sex due to their links with metabolic and physical health outcomes. We used semiparametric linear models, and normality is not required for these models. However, a variable is transformed appropriately using log-transformation if its distribution is highly skewed to make it less skewed to improve efficiency (power). We log-transformed the following variables: HbA1c levels. We used false discovery rate (FDR) methods to correct for multiple comparisons testing.⁶⁶ We performed power analysis for regression analyses involving the RAR predictor for the 3 primary health outcomes (HbA1c, number of metabolic syndrome criteria and self-reported physical health). The study sample n = 62 has 80% power for a significant IV effect, if the true effect size η_p² = .137 based on a 2-sided type I error α = .05. The study sample does provide sufficient power for a medium effect size.

We examined how taking sleep medications more than once a week was related to the Pittsburgh Sleep Quality Index (PSQI) and actigraphy measures using independent samples t tests to compare these subgroups within the PLWS sample. While the number of participants in each subgroup was small, we examined the relationships for medium-large effect sizes. Finally, Spearman’s correlations were conducted to examine the differential association of RAR variables with sociodemographic variables, sleep, and (specifically among PLWS) illness-related factors. All analyses were performed in R and SPSS. Significance was defined as α < .05 (2-tailed) for all analyses.

RESULTS

The study sample consisted of 62 individuals (PLWS: n = 26, NC: n = 36; Table 1). On average, participants had 7 consecutive 24-hour periods of actigraphy data, with a range of 4–12 24-hour periods and the vast majority (80%) had 6–8 24-hour periods of data. The two groups were comparable by proportion of female and White participants. Compared to NCs, the PLWS had fewer years of education, as well as worse scores on anxiety and depression scales (Ps < .05). Among the PLWS group, 0% had full-time employment, 8% had part-time employment, 16% were permanently disabled, 16% were unemployed, with the rest unknown. Among the NC group, 28% had full-time employment, 8% had part-time employment, 3% were retired, 3% were unemployed, with the rest unknown.

Table 1.

Demographic and clinical characteristics of PLWS and NCs.

	NCs			PLWS			t or χ2	df	P	Cohen’s d
	n	Mean	SD	n	Mean	SD
Sociodemographic
Age at visit	36	51.4	12.6	25	52.8	11.1	−0.43	59	.67	−0.11
Sex (% female)	18	50.0		14	56.0		0.213	1	.64
Race (% non-Caucasian)	7	19.4		9	36.0		2.09	1	.15
Education (years)	36	15.5	2.1	25	12.8	2.3	4.77	59	< .001	1.24
Marital status (% partnered)	15	42		1	4		10.4	1	.001
Living situation (% with others)	28	78		21	88		9.09	1	.34
Mental health factors
Antipsychotic dose (WHO DDD)				25	2.4	2.3
Positive symptoms (SAPS)				25	6.1	3.8
Negative symptoms (SANS)				25	6.4	3.6
Anxiety (BSIA)	34	0.9	1.7	24	6.6	5.2	−5.71	56	< .001	−1.59
Depression (PHQ-9)	34	2.3	3.7	24	9.5	6.3	−4.99	56	< .001	−0.89
Physical health
Physical well-being (SF-36)	34	52.2	8.7	24	44.6	12.0	2.65	39.4	.011	0.75
Medical comorbidities (CIRS)	34	3.5	3.2	22	6.3	2.9	−3.27	54	.002	−0.89
Body mass index (kg/m2)	26	25.6	4.9	22	31.5	7.5	−3.25	33.9	.003	−0.97
Hemoglobin A1c (%)*	31	5.37	0.34	21	6.26	1.64	−2.46	21.1	.023	−0.84
Metabolic syndrome criteria (#)	28	1.54	1.29	22	3.32	1.70	−4.01	38.2	< .001	−1.20
Self-reported sleep
Sleep quality score (PSQI)	31	4.9	3.2	20	7.9	4.0	−2.93	49	.005	−0.84
Self-reported sleep quality (% bad)	3	6.5		5	25		3.35	1	.06
Sleep latency (% 31+ min)	4	12.9		9	45		6.59	1	.01
Sleep duration (% 6 or fewer hours)	1	3.2		1	5		0.10	1	.75
Sleep efficiency (% < 85%)	10	32.3		8	40		0.32	1	.57
Sleep disturbance (% ≥ 1/wk)	4	12.9		10	50		8.40	1	.004
Sleep medication use (% ≥ 1/wk)	3	9.7		8	40		6.61	1	.01
Daytime dysfunction (% difficult)	1	3.2		2	10		1.01	1	.32
Objective sleep measures
Mean TST (min)	36	367.3	65.0	25	379.6	131.0	−0.43	32.3	.67	−0.13
Mean percent sleep (%)	36	79.1	9.8	25	84.0	8.1	−2.05	59	.045	−0.53
Mean WASO (min)	36	92.0	54.8	25	60.5	31.1	2.86	57.1	.006	0.68
TST RMSSD (min)	36	77.9	42.4	25	122.5	74.5	−2.72	34.8	.01	−0.78
Percent sleep RMSSD (%)	36	8.4	4.7	25	9.4	8.1	−0.63	59	.53	−0.17
WASO RMSSD (min)	36	46.7	31.0	25	40.5	23.3	0.85	59	.40	0.22
Mean latency (min)	36	9.33	7.91	25	6.01	4.32	1.91	59	.04	0.50
Bedtime (24 hours clock)	36	23:20	1:23	25	23:05	2:46	0.46	59	.64	0.12
Waketime (24 hours clock)	36	05:28	1:25	25	05:11	01:39	0.72	59	.47	0.19

*Analyses performed on log-transformed values. BSIA = Brief Symptom Inventory – Anxiety, CIRS = Cumulative Illness Rating Scale, NCs = nonpsychiatric comparison participants, PHQ-9 = Patient Health Questionnaire, PLWS = people living with schizophrenia, PSQI = Pittsburgh Sleep Quality Index, RMSSD = root-mean square successive difference, SANS = Scale for the Assessment of Negative Symptoms, SAPS = Scale for the Assessment of Positive Symptoms, SD = standard deviation, SF-36 = 36-Item Short Form Health Survey, TST = total sleep time, WASO = wake after sleep onset, WHO DDD = World Health Organization defined daily dose.

The PLWS were more likely to receive antidepressant and anxiolytic medications compared to the NCs. Among the PLWS, 19 (76%) received 1 or more antidepressant medications, compared to 3 (8%) of NCs. Among the PLWS, 8 (32%) received an anxiolytic medication (benzodiazepine, buspirone, hydroxyzine), compared to none of the NCs. Among the PLWS, 4 (16%) received trazodone for sleep, compared to none of the NCs. Among PLWS, 91.3% of the participants were currently taking atypical antipsychotic medications and no participants were currently on monotherapy with typical antipsychotic medications. Sleep complaints were common in both groups, with over 40% of PLWS reporting issues with sleep latency, sleep efficiency, sleep disturbances, and sleep medication use. Eight PLWS reported taking any sleep medications more than once per week, while 12 PLWS did not. We found that the individuals on medications had worse sleep quality scores (total PSQI score) (P = .004, d = 1.52), longer self-reported sleep latency (P = .36, d = 0.43), and more self-reported sleep disturbances (P = .35, d = 0.44). For the actigraphy data, we also found that the individuals on sleep medications had greater percent sleep (P = .20, d = 0.61), compared to those not taking sleep medications. We did not observe relationships with sleep latency (P = .90, d = 0.10) or wake after sleep onset (P = .83, d = 0.10).

The PLWS had worse self-rated physical well-being, greater number/severity of comorbid medical conditions, higher body mass index, higher HbA1c levels, and greater number of metabolic syndrome criteria, compared to the NCs. Within this sample, 59% percent of PLWS and 21% of NCs met criteria for metabolic syndrome diagnosis (χ² = 7.42, P = .006).

Differences in sleep and RARs between PLWS and NCs

The PLWS reported worse sleep quality compared to the NCs (t₄₉ = −2.93, P = .0005, Cohen’s d = −0.84; Table 1). PSQI-assessed sleep complaints of sleep latency > 30 minutes were not associated with actigraph-derived measures of sleep latency. While 4 NCs and 9 PLWS reported impaired sleep latency (PSQI), none of these individuals had actigraphy-assessed sleep latency > 30 minutes. Of the subset of participants (n = 31) with sleep apnea risk scores, 54% of the PLWS group and 28% of the NC group had high risk for sleep apnea. Using objective actigraphic sleep measures, the PLWS had higher percent sleep, less wake after sleep onset, and increased variability of TST (Cohen’s d = −0.53, 0.68, and −0.78, respectively).

In terms of RAR variables, the PLWS had a shorter active period, lower RA, and lower activity during the M10 period compared to the NCs (Cohen’s d = 0.79, 0.58, and 0.62, respectively). IS, IV, and L5 were similar between the 2 groups (Table 2).

Table 2.

Comparison of rest-activity rhythm variables between PLWS and NCs.

	NCs			PLWS			t	df	P	Cohen’s d
	n	Mean	SD	n	Mean	SD
Active period length (hours)	36	15.5	1.37	25	14.1	2.24	2.78	36.3	.009	0.79
Interdaily stability	36	0.57	0.12	25	0.54	0.14	0.78	59	.44	0.20
Intradaily variability	36	0.80	0.22	25	0.84	0.25	−0.56	59	.58	−0.15
Relative amplitude	36	0.88	0.09	25	0.80	0.16	2.03	34.2	.05	0.58
M10	36	2650.1	672.9	25	2200.5	806.8	2.36	59	.02	0.62
L5	36	166.2	111.9	25	256.2	268.4	−1.58	29.8	.12	−0.47

L5 = mean activity during the least active 5-hour period, M10 = mean activity during the most active 10-hour period, NCs = nonpsychiatric comparison participants, PLWS = people living with schizophrenia, SD = standard deviation.

Associations of RAR variables with physical health measures

Controlling for diagnostic group, age, and sex, higher IV was associated with higher HbA1c levels (B = 1.39, standard error = 0.60, P = .02, FDR-P = .15, η_p² = .10), although the relationship was not significant after FDR correction (Table S1, Table S2, and Table S3 in the supplemental material). Lower IS and lower RA were associated with worse self-rated physical well-being (B = 35.6, standard error = 9.51, P < .001, FDR-P = .006, η_p² = .21 and B = 22.1, standard error = 10.8, P = .04, FDR-P = .15, η_p² = .07, respectively). Linear regression models that included an interaction term of RAR variables by diagnostic group were not found to be significant, so the models were trimmed as presented above.

Correlations of RAR variables with sociodemographic, mental health, and other sleep variables stratified by diagnostic group

RAR variables were associated with anxiety, depression, and sleep quality measures among NCs, but not among PLWS (see Table 3). Among NCs, higher IS (ie, better crossdaily stability of rhythms) was associated with lower anxiety and depression scores (r_s = −.433, P < .05 and r_s = −.402, P < .05, respectively), while lower IV (less fragmentation of rhythms) was associated with lower anxiety and better sleep quality (r_s = .405, P < .05 and r_s = .539, P < .01, respectively).

Table 3.

Correlations of rest-activity rhythm variables in PLWS and NCs.

		NCs (n = 36)						PLWS (n = 25)
		Active Period Length	Interdaily Stability	Intradaily Variability	Relative Amplitude	M10	L5	Active Period Length	Interdaily Stability	Intradaily Variability	Relative Amplitude	M10	L5
Sociodemographic
Age	rs	−.204	.237	−.169	−.033	−.123	−.018	.056	.077	−.093	−.068	.014	.015
Mental health
Antipsychotics dose (WHO DDD)	rs							−.128	−.217	.156	.052	−.327	−.060
Positive symptoms (SAPS)	rs							−.105	.109	−.021	.016	.032	.060
Negative symptoms (SANS)	rs							−.014	.151	.188	−.201	−.326	.153
Anxiety (BSIA)	rs	.177	−.433*	.405*	−.187	.008	.303	−.284	.328	.172	−.135	−.132	.019
Depression (PHQ-9)	rs	.116	−.402*	.246	−.273	−.118	.277	.086	−.133	−.099	.027	.007	.014
Self-reported sleep
Sleep quality (PSQI)	rs	.164	−.274	.539**	−.107	−.046	.131	−.013	.372	−.179	.033	−.027	−.108
Objective sleep
Total sleep time (mean)	rs	−.268	.255	−.136	.636**	−.074	−.807**	−.204	.307	−.319	.789**	.080	−.752**
Percent sleep (mean)	rs	.175	.113	.027	.457**	−.153	−.545**	−.070	.084	−.023	.509*	−.237	−.588**
WASO (mean)	rs	−.262	−.010	−.106	−.313	.136	.347*	−.155	.079	−.205	.015	.278	.122
Total sleep time (RMSSD)	rs	.038	−.324	−.118	−.272	−.070	.234	.056	−.164	−.165	−.055	.155	.142
Percent sleep (RMSSD)	rs	−.118	−.094	−.159	−.181	.135	.244	.450*	−.251	.068	−.344	.067	.318
WASO (RMSSD)	rs	−.228	−.035	−.169	−.288	.116	.366*	.218	−.149	−.192	−.108	.136	.158
Mean latency (min)	rs	−.220	.103	.028	−.076	.083	.161	.086	−.015	−.324	.117	.419*	−.002
Bedtime (24 hours clock)	rs	.184	−.187	−.072	.074	−.173	−.151	.017	.246	.061	.204	−.152	−.263
Waketime (24 hours clock)	rs	.387*	−.259	.113	−.241	−.044	.291	.280	−.192	.197	−.595**	−.108	.528**

*Correlation is significant at the .05 level (2-tailed). **Correlation is significant at the .01 level (2-tailed). BSIA = Brief Symptom Inventory – Anxiety, L5 = mean activity during the least active 5-hour period, M10 = mean activity during the most active 10-hour period, PHQ-9 = Patient Health Questionnaire, NCs = nonpsychiatric comparison participants, PLWS = people living with schizophrenia, PSQI = Pittsburgh Sleep Quality Index, RMSSD = root-mean square successive difference, SANS = Scale for the Assessment of Negative Symptoms, SAPS = Scale for the Assessment of Positive Symptoms, SD = standard deviation, SF-36 = 36-Item Short Form Health Survey, WASO = wake after sleep onset, WHO DDD = World Health Organization defined daily dose.

In both groups, greater RA was associated with longer TST (NCs: r_s = .636, P < .01; PLWS: r_s = .789, P < .01) and increased percent sleep (NCs: r_s = .457, P < .01; PLWS: r_s = .509, P < .05). Among both groups, lower L5 was associated with longer TST (NCs: r_s = −.807, P < .01; PLWS: r_s = −.752, P < .01) and increased percent sleep (NCs: r_s = −.545, P < .01; PLWS: r_s = −.588, P < .02). Among PLWS only, longer active period length was associated with greater variability of percent sleep (r_s = .450, P < .05), higher M5 was associated with greater sleep latency (r_s = .419, P < .05), and higher RA and lower L5 was associated with earlier wake time (r_s = −.595, P < .01 and r_s = .528, P < .01, respectively). RAR variables were not correlated with age, antipsychotic dose, or severity of positive and negative symptoms.

DISCUSSION

The purpose of the current study was to examine differences in actigraphy-derived RARs in PLWS and NCs and to characterize the physical health and clinical correlates of RARs. The current study findings partially supported our hypotheses. PLWS had shorter active periods, less robust RARs (ie, lower RA), and lower mean activity during their M10 compared to the NCs. Also, greater fragmentation of rhythms and lower IS of rhythms were associated with worse metabolic and physical health outcomes in both groups. Among PLWS, RAR variables were associated with sleep measures, but not with age or illness-related factors.

The findings of the current study are consistent with several prior studies that have reported lower M10 activity, greater variability in TST, and less robust RARs among PLWS.^20,67–69 In contrast, one study by Berle and colleagues reported that institutionalized PLWS had higher IS and lower IV (ie, more structured RARs) compared to NCs.⁷⁰ The disparate findings of the Berle et al study could reflect the different routines for inpatient PLWS compared to community-dwelling PLWS, as reflected in the current sample. More structured RARs among inpatient PLWS may be the product of the fixed institutional schedules for meals, medications, and nighttime activities.

To our knowledge, this is the first study to explore the physical and mental health correlates of RAR in PLWS. Consistent with polysomnography findings that circadian rhythm disruption-related sleep changes (eg, rapid eye movement sleep modulation⁷¹ and altered sleep timing^20,72) are frequently associated with poorer psychological and physiological health, the current study found that greater fragmentation of rhythms and lower IS of rhythms were associated with worse metabolic and physical health outcomes among both PLWS and NCs. These preliminary findings are consistent with several studies among other populations (ie, older adults, patients with bipolar disorder) that found that RARs with less rhythmicity, lower amplitudes, or more variability of sleep duration or sleep timing were associated with worse cholesterol levels, obesity, hyperglycemia, and higher blood pressure.^16,17,73 This finding highlights the potential importance of RAR variables to metabolic health among PLWS. Potential mechanisms linking RAR disruption and metabolic dysfunction include effects mediated by sleep-disordered breathing⁷⁴ and molecular changes due to dysregulated transcription of metabolic genes and abnormal rhythmnicity of glucose homeostasis and insulin secretion.⁷⁵ Irregular RARs may be associated with changes in behaviors that promote risk for metabolic syndrome. For example, higher sleep variability may lead to dysregulated patterns of mealtimes and eating frequency, which has been associated with weight gain and diabetes risk.^76,77 In contrast to other studies that reported older age was associated with RAR dysregulation,^27,78 we did not find an association between age and RAR variables. This may be due to the limited number of participants older than age 70 (current sample age range 29–72 years) relative to these other studies (age 21–91) as well as due to potential survivor effects, where the older PLWS within our sample have survived to that older age due to being overall healthier than the younger PLWS group.

Of note, the PLWS within this sample had some indicators of better objectively assessed sleep, although they also had worse subjectively assessed sleep compared to NCs. PLWS had better mean percent sleep and less wake after sleep onset which would indicate fewer overnight awakenings. However, PLWS had worse sleep quality measures and increased variability of TST compared to NCs. Similarly, reports of sleep latency > 30 minutes on the PSQI were not associated with actigraphy-derived measures among both PLWS and NCS, possibly reflecting limitations of actigraphy in accurately assessing sleep latency without EEG sensors, cognitive deficits, or medication effects among PLWS. Our subgroup analysis of PLWS who reported taking sleep medications more than once a week showed that these individuals had worse sleep quality scores, longer self-reported sleep latency, more self-reported sleep disturbances, and greater actigraphy-measured percent sleep compared to not taking sleep medications. Differences between self-reported and objective sleep reports have been reported among PLWS and other psychiatric populations and may provide additional information about the mood state and functional status of the participants. One study found discrepancies between self-reported and objective sleep parameters among PLWS, with wider discrepancies linked to worse psychosocial functioning.⁷⁹ Similarly, studies of people living with bipolar disorder have shown greater self-reported and objective sleep measure discrepancies to be linked to lower functional status⁸⁰ and more severe depression.^81,82 Also variability of sleep measures, rather than mean sleep measures, may be an important indicator of sleep quality and functioning status among PLWS and other individuals living with serious mental illnesses.⁸³ Furthermore, the influence of medications on sleep quality and objective sleep parameters require further evaluation among PLWS.

Our exploratory analysis examining sociodemographic and other clinical correlates of RAR variables found overlapping and distinct relationships among these variables between the 2 groups. Specifically, in both groups, more robust RAR, as reflected by higher RA, was associated with longer sleep duration and better sleep efficiency. In NCs only, better IS of RARs was associated with lower severity of anxiety and depression, whereas less fragmentation of intradaily rhythms was linked with lower anxiety and better sleep quality. In contrast, in PLWS, longer active period length was associated with more variable percent sleep, but RAR variables were not associated with antipsychotic dosage, or positive and negative symptom severity.

Consistent with prior literature in nonclinical populations,^84,85 more stable and less fragmented rhythms were associated with better emotional functioning and sleep quality in the NC group, although no such positive relationship was found within the schizophrenia group. Moreover, RAR variables were not associated with positive and negative symptom severity in this sample, which is inconsistent with the extant polysomnography literature that shows sleep parameters, such as deficits in slow wave sleep, are associated with greater symptom severity, particularly negative symptoms of schizophrenia.⁸⁶ The current null finding could reflect the relative stability of these outpatient PLWS, heterogeneity in clinical rating scales among studies, and/or the need for more focused analyses to investigate the different subdomains of illness symptomatology.

The lack of association between antipsychotic medications and RARs in the current study is consistent with some prior studies,^22,87–89 although there are contrasting findings within a limited research base of primarily observational cross-sectional studies that lack randomization to medications.^25,26,90 Historically, the influence of antipsychotic medications on the rest-activity cycle has been difficult to determine given heterogeneity in the class of antipsychotic drugs and additional potential confounds (eg, use of other circadian-system alternating drugs, variability in sleep behaviors, and environmental influences). Variability in the pharmacological profile of various antipsychotic drugs can differentially impact sleep-wake states,⁹¹ with studies showing a higher efficacy of atypical antipsychotics in improving sleep and RAR abnormalities.^25,92 Nevertheless, the current study found more RAR disruption in a sample in which nearly all PLWS were taking atypical antipsychotics. Furthermore, the PLWS group’s higher rates of antidepressant and anxiolytic medication use may have further impacted rapid eye movement latency, sleep continuity, as well as duration of slow wave sleep and rapid eye movement sleep—with downstream consequences on sleep quality and actigraphy measures.^93–95 Due to heterogeneity of antidepressant and anxiolytic effects on sleep, we were unable to perform subgroup analyses of the impact of these medications. This finding, when considered in the context of the extant literature, suggests that PLWS experience RAR abnormalities at disproportionate rates compared to the general population even with pharmacological intervention, underscoring the need for alternative avenues of propagating benefits that come with stabilization of sleep and RAR dysfunction.

RARs are a modifiable target.⁹⁶ Several interventions have been developed to strengthen RARs and improve outcomes for different patient populations. Timed light exposure has shown improvements in RARs and sleep consolidation for patients with Alzheimer’s disease and other dementias.^44,45,47 Bright-light treatment coupled with melatonin improved daytime wake time and activity levels and strengthened RAR in patients with dementia.⁴⁶ Other treatments include filtering blue light exposure, transcutaneous electrical nerve stimulation,⁴¹ adherence to sleep and wake time schedule, vitamin B12 supplementation,³⁹ and stimulant medications which can be helpful for circadian dysregulation disorders, jet lag, and shift work.⁹⁷ Implementation of RAR therapies in PLWS has historically not been a clinical or research priority. Available literature suggests the need for integrated therapies given the complexity introduced by the phenotypic overlay between the circadian system and the pathophysiology of schizophrenia. For example, there is some evidence supporting the efficacy of melatonin on sleep, circadian rhythm and metabolic outcomes in PLWS, but the literature is scarce and difficult to draw firm conclusions.⁹⁸ Disparate findings regarding the efficacy of melatonin as an adjunct therapy may be due to altered MT1 melatonin receptor expression given that a polymorphism in the MT1 gene is associated with schizophrenia.⁹⁹ While the current study cannot examine the influence of OSA risk or diagnosis on the RAR variables, we found high rates of OSA among the PLWS, that were 2-fold higher than those in NCs. These findings highlight the importance of assessing for and treating OSA among PLWS, given the high prevalence of OSA and linkages to both RARs and metabolic dysregulation.

The current study was limited by the small sample size and we may have been underpowered to detect true relationships. Its cross-sectional design limited exploration of causal relationships between RAR and metabolic, physical, or mental health. Nearly all PLWS were on atypical antipsychotics so any potential differential impact of class of antipsychotics could not be examined. Similarly, we were unable to perform subgroup analyses of different antidepressant and anxiolytic medications on sleep outcomes. Due to the limited data available on sleep apnea risk, the current study was not able to disentangle the independent contributions of OSA to metabolic health. The current findings may have also been influenced by the technical limitations of the actigraphs and the inherent bias present in self-report measures. Nevertheless, the use of both objective and self-reported measures of RAR and related factors (sleep) is consistent with current guidelines to include both types of behavioral assessment tools as each may capture related but unique constructs.

A particular strength of the study is its novel investigation of mental and physical health correlates, including metabolic health, of RAR in PLWS. Our findings underscore the importance of regular sleep health screenings to identify patients in need of targeted, integrated behavioral and medical treatments for RAR stabilization. This may lead to improved sleep and reduce risk for physical comorbidities, which may, in turn, further improve and/or sustain stability in the circadian rest-activity cycle.

DISCLOSURE STATEMENT

All authors have reviewed and approved the final manuscript. This work was supported, in part, by the National Institute of Mental Health (R01 grant R01MH094151-01 and K23 grant K23 MH119375-01), the National Institute on Aging T35 grant AG26757 (PI: Ellen Lee, MD, Benjamin Han, MD, MPH), the National Institutes of Health (NIH UL1TR001442 of CTSA [PI: Gary Firestein, MD]), a Havens Established Investigator Grant from the Brain & Behavior Research Foundation, the American Psychiatric Association, the Desert-Pacific Mental Illness Research Education and Clinical Center at the VA San Diego Healthcare System, and the Stein Institute for Research on Aging at the University of California San Diego. The funding sources had no other role in this publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Dr. Malhotra is funded by the National Institutes of Health. He reports income related to medical education from Jazz, Zoll, Livanova, and Eli Lilly. ResMed provided a philanthropic donation to University of California San Diego. The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors have no other conflicts of interest to report. At the time of the study, Dr. Mahmood was affiliated with the San Diego State University/University of California San Diego Joint Doctoral Program, San Diego, California and the VA San Diego Healthcare System, San Diego, California. She is currently affiliated with the VA Greater Los Angeles Healthcare System, Los Angeles, California.

ABBREVIATIONS

FDR

false discovery rate

HbA1c

hemoglobin A1c levels

IS

interdaily stability

IV

intradaily variability

L5

least active 5 hours

M10

most active 10 hours

NC

nonpsychiatric comparison participant

OSA

obstructive sleep apnea

PLWS

people living with schizophrenia

PSQI

Pittsburgh Sleep Quality Index

RA

relative amplitude

RAR

rest-activity rhythm

TST

total sleep time