https://orcid.org/0000-0002-2874-658X
Abstract
Abnormalities in the stress system have been implicated in insomnia. However, studies examining physiological stress regulation in insomnia have not consistently detected differences in the hypothalamic–pituitary–adrenal (HPA)-axis response to stress. One explanation may be that deficits in the stress system are associated specifically with a biological vulnerability to insomnia rather than the phenotypic expression of insomnia. To examine stress response as a function of vulnerability to insomnia, this study tested response to the Trier Social Stress Test in a sample of healthy sleepers with varying familial risks for insomnia.
Thirty-five healthy individuals with and without familial risk for insomnia were recruited to complete a laboratory stressor. Participants with one or both biological parents with insomnia were categorized as positive for familial risk, whereas those without biological parents with insomnia were categorized as negative for familial risk. Participants completed the Trier Social Stress Test in the laboratory, and psychological and physiological (autonomic and HPA-axis) responses were compared.
Despite self-reported increases in anxiety, those positive for familial risk exhibited a blunted cortisol response relative to those without familial risk for insomnia. Individuals with blunted cortisol also reported heightened reactivity to personal life stressors, including increased sleep disturbances, elevated cognitive intrusions, and more behavioral avoidance.
Findings from this study provide initial evidence that abnormal stress regulation may be a biological predisposing factor conferred via familial risk for insomnia. This deficit may also predict negative consequences over time, including insomnia and the associated psychiatric comorbidities.
Although a growing body of evidence supports the diathesis–stress model of insomnia that explores how biological or genetic traits interact with environmental influences to produce insomnia, the biological underpinnings of predisposition to stress and insomnia have not been adequately characterized. Abnormalities in hypothalamic–pituitary–adrenal (HPA)-axis functioning have been posited as a mechanism of vulnerability for insomnia, though replication of this has been inconsistent. Results from this study suggest that deficits in HPA-axis functioning may be specifically related to familial risk for insomnia, specifically in the form of blunted cortisol response to stress. Furthermore, the deficit in stress regulation may be specific to the HPA-axis and not the autonomic nervous system. The pattern of deficit found in this study suggests adverse implications (e.g., over-reliance on other allostatic response systems) that elevate the risk for medical and psychiatric morbidities.
INTRODUCTION
Research in the etiology of insomnia has pointed to a diathesis–stress model that proposes that the disorder arises from a combination of latent predispositional vulnerabilities that can be triggered by—and interact with—environmental stress.1,2 It is clear that individuals have varying degrees of vulnerability to insomnia, and individuals who are prone to sleep disturbance following stress are more likely to develop insomnia.3,4 However, the biological underpinnings of vulnerability to insomnia have yet to be determined. This is important to delineate because pathobiology can further inform intervention and prevention of insomnia along with the myriad of psychiatric and medical comorbidities.
Emergent evidence suggests that abnormalities in the stress response systems may be an underlying mechanism of vulnerability to insomnia. Theoretically, individuals with deficits in stress regulation may experience impairments that can, in turn, interfere with sleep initiation or continuity. Indeed, people with insomnia report more intense subjective stress responses compared with good sleepers,5 and numerous longitudinal studies show that stress exposure predicts the development of new onset insomnia.3,6,7 To corroborate these findings, studies have also searched for physiological markers of abnormal stress regulation (e.g., hypothalamic–pituitary–adrenal (HPA)-axis functioning), though with varied results. A few earlier studies have examined basal cortisol functioning in individuals with insomnia, and some have found higher evening and nocturnal circulating levels of adrenocorticotropic hormone and cortisol compared with healthy sleepers,8,9 with higher cortisol levels predicting sleep disturbance.9,10 It has been posited that this suggests a deficit in the recovery process of the HPA-axis following activation (e.g., from morning circadian stimulation),11 leading to high levels of nocturnal cortisol. However, other studies have failed to detect differences in evening and nocturnal cortisol between healthy controls and people with insomnia,12–15 and experimental studies have also indicated that elevated cortisol may actually be a consequence of sleep deprivation.16,17 This makes it difficult to determine whether elevated cortisol is a precursor to the development of insomnia, or whether it is a consequence of the sleep disturbance that is unrelated to the etiology of insomnia.
There may be a number of factors that contribute to the inconsistent detection of differences in stress response systems associated with insomnia disorder. First, deficits in stress response systems may not be consistently detected in the absence of an acute stressor. In fact, such deficits may be specific to the response to a stressor18; however, very few studies have specifically examined physiological responses to acute stressors in the laboratory. Moreover, the existing studies have utilized a wide range of stressors, and this may also explain the sporadic findings because not all stressors uniformly elicit a response across the various stress systems.11,19 For example, a recent study of adults with insomnia did not detect differences in cortisol response to a physical stressor (electric shock) compared with healthy controls; however, it was unclear whether the physical stressor elicited an effective stress response.20 Furthermore, the stressor also lacked a strong social-evaluative element, despite these being most strongly associated with a cortisol response.21 Prior research has also demonstrated that use of an electric shock effectively induced a response in α-amylase but not in cortisol.22 These limitations result in gaps in knowledge and indicate a need for the use of standardized stress challenge paradigms, which reveal stress-response abnormalities obscured by individual differences in resting baseline measures.23,24
To determine whether deficits in the stress response systems are implicated in vulnerability to insomnia, research should also focus on premorbid individuals with specific risk factors. In accordance with the diathesis–stress model, insomnia arises from an interaction of individual predisposition and environmental triggers (i.e., stress); however, genetic predisposition to insomnia (i.e., diathesis) may predominate in some individuals, whereas environmental triggers may play a more prominent role in others (i.e., stressful life event triggers). Therefore, studies using samples that are organized based on the phenotypic expression of insomnia disorder may have significant heterogeneity in the mechanisms of vulnerability, which can result in sporadic detection of differences in physiological stress response. Alternatively, reducing sample heterogeneity by examining the stress response systems in individuals who are at risk and premorbid for insomnia can reveal the biological underpinnings of the diathesis for insomnia independent of the disease process. Among the various risk factors, familial risk is among the strongest predictors of incident insomnia and thus is a clear marker of the diathesis for insomnia.7,25 Those with familial risk of insomnia show up to a six-fold increase in insomnia relative to the general population, reflecting the strong heritability of insomnia disorder4,26 and sleep reactivity to stress.4,27,28
Current Study
To examine whether abnormalities in the stress response systems may be implicated in the development of insomnia, this study compared psychological and physiological responses with a standardized laboratory stressor in good sleepers with and without familial risk for insomnia. Specifically, this study employed the Trier Social Stress Test (TSST), which is among the most well-studied standardized stress challenge paradigms and which activates both the autonomic nervous system and the HPA axis.29 The autonomic nervous system is immediately responsive to stressors and activates the classic “fight or flight” response. In contrast, the HPA-axis response involves a cascade of processes that eventually lead to the production and release of glucocorticoids that protects against the long-term threat of stress exposure. Among various functions, glucocorticoids complete the negative feedback loop that is imperative in restoring physiological homeostasis. This adaptive response prevents damage resulting from prolonged exposure to stress.30,31 As such, abnormalities in the HPA-axis response to stress may be particularly relevant in the diathesis–stress model.
Although various profiles of abnormalities in HPA-axis response to stress have been documented,11 prospective studies have indicated that a profile of blunted cortisol response to stress is predictive of negative outcomes. An inadequate cortisol response to stress can result in overcompensation of other allostatic systems (e.g., increased inflammatory cytokines) that are normally counter-regulated by glucocorticoids, leading to a wide range of pathologies and other adverse outcomes. For example, a longitudinal study in police officers found that blunted cortisol responses to stress during training predicted higher distress and worse coping following exposure to work stress and trauma. In comparison, those who were initially able to mount a robust cortisol response demonstrated a trajectory of resilience.32 Similarly, cortisol response following a traumatic event has been shown to affect the subsequent development of post-traumatic stress symptoms.33 For example, blunted salivary cortisol measured in the hospital following a motor vehicle accident predicted more severe depression and post-traumatic stress symptoms following discharge.34,35 In contrast, administration of hydrocortisone in the intensive care unit led to reduced incidence of post-traumatic stress disorder.36–38 Importantly, when correlated with specific symptoms, low serum cortisol measured in the intensive care unit was most strongly correlated with sleep disturbance.39 Additionally, blunted cortisol has also been prospectively associated with poor outcomes across a range of other disorders, including in children with attention-deficit hyperactive disorder40 and adult males with alcohol dependence.41 There is also evidence that a blunted cortisol response may be a heritable risk factor, as has been documented in non-alcoholic children of alcoholics.42,43 Finally, studies in sleep disruption—albeit different from insomnia disorder—have also found a blunted cortisol response to the TSST.44–46 Together, the evidence indicates the plausibility that familial risk for insomnia may also be associated with a blunted cortisol response to stress.
In this study, both the autonomic nervous system and the HPA-axis responses to a standardized laboratory stressor were compared to examine specificity in deficits to the two stress systems. It was hypothesized that individuals across familial risk groups would show comparable responses in the autonomic nervous system, but those with familial risk for insomnia would exhibit more blunted HPA-axis stress responses compared with those negative for familial risk for insomnia.
METHODS
Participants
Participants were recruited from the Evolution of Pathways to Insomnia Cohort study, which was a longitudinal investigation examining insomnia in a large representative sample of adults living in southeastern Michigan.3 Participants from this study were prescreened for eligibility only if they consented to be contacted for future studies, and exclusionary criteria included history of shift work, indication of insomnia based on an Insomnia Severity Index (ISI) score of greater than eight, and indication of sleep disorders (e.g., obstructive sleep apnea, restless legs syndrome). The prescreening yielded a total of 111 participants who were recruited for participation in this study (see Figure 1) and a total of 42 individuals volunteered for participation.
Recruitment flow chart. EPIC = Evolution of Pathways to Insomnia Cohort study. ISI = Insomnia Severity Index.
Eligibility was further determined via telephone screening followed by in-person interviews conducted by a clinical psychologist or a specialist in sleep medicine. Exclusion criteria included any acute or chronic disease; current or history of insomnia, sleep disorders (e.g., obstructive sleep apnea, restless legs syndrome), or psychiatric illnesses (e.g., major depressive disorder, substance use disorders); use of any steroidal contraceptive, hypnotic, or medication acting on the central nervous system; and current evening, night, or rotating work shift. Screening for sleep apnea was completed using the Berlin questionnaire.47 Individuals identified as high risk for sleep apnea on the Berlin questionnaire (two or more categories were positive) were excluded from study participation. Restless legs syndrome was also screened out using the restless legs syndrome questionnaire48 if participants met the four essential diagnostic criteria as recommended by Allen and colleagues.49 Participants also completed the Hamilton Depression Rating Scale,50 and those with a score of 10 or higher were excluded. All study procedures were in accordance with the Declaration of Helsinki and approved by the Institutional Review Board. Informed consent was obtained from all participants before the experimental procedures. Participants were compensated for their participation.
A total of 42 participants completed an in-person interview, and seven were excluded due to insomnia assessed via the Insomnia Severity Index and clinical interview using the International Classification of Sleep Disorder (2nd edition) criteria. A final sample of 35 individuals were enrolled in the study and were subsequently categorized into negative and positive familial risk groups based on parental history for insomnia, which was assessed via self-report. Participants were considered positive for familial risk if they reported that one or both biological parents experienced difficulty falling or staying asleep or had non-refreshing sleep at least three times a week for a duration of one month or longer. Participants without any parental history of insomnia were classified as negative for familial risk. There were 17 individuals who were negative for familial risk of insomnia, and 18 who were positive (five reported both parents experienced insomnia). Familial risk groups did not differ by age, body mass index (BMI), or sex (see Table 1).
Mean, Standard Deviation, and Effect Size for Group Differences (Cohen’s d) for Demographic and Sleep Variables.
| Variables | Negative FR | Positive FR | Cohen’s d | p | EPIC sample |
|---|---|---|---|---|---|
| Age | 47.76 (11.05) | 45.56 (10.26) | 0.21 | ns | 44.71 (13.78) |
| Sex (% F) | 41.2% | 61.1% | n/a | ns | 58.6% |
| Body mass index | 26.76 (3.94) | 25.33 (3.64) | 0.38 | ns | 28.2 (6.57) |
| Bedtime (hour) | 2301 (1.60) | 2317 (1.56) | 0.17 | ns | 2227 (3.23) |
| Waketime (hour) | 0621 (1.80) | 0618 (2.09) | 0.02 | ns | 0759 (2.82) |
| Time in bed (minute) | 465.88 (35.76) | 454.17 (36.31) | 0.32 | ns | 479.90 (169.35) |
| Total sleep time (minute) | 427.06 (30.52) | 425.17 (43.75) | 0.05 | ns | 407.42 (75.90) |
| Sleep onset latency (minute) | 14.82 (7.46) | 18.89 (16.64) | 0.32 | ns | 45.28 (80.30) |
| Number of night awakenings | 0.68 (0.64) | 1.08 (1.96) | 0.28 | ns | 2.82 (5.46) |
| Wake after sleep onset (minute) | 10.18 (17.11) | 3.47 (4.21) | 0.54 | ns | 27.17 (50.66) |
| Insomnia Severity Index | 1.94 (2.30) | 2.67 (3.20) | 0.26 | ns | 3.81 (3.58) |
| FIRST | 16.76 (3.82) | 19.94 (5.73) | 0.65 | .06 | 16.24 (4.35) |
| Arousal predisposition scale | 22.35 (5.37) | 26.28 (6.18) | 0.68 | .05 | n/a |
| Variables | Negative FR | Positive FR | Cohen’s d | p | EPIC sample |
|---|---|---|---|---|---|
| Age | 47.76 (11.05) | 45.56 (10.26) | 0.21 | ns | 44.71 (13.78) |
| Sex (% F) | 41.2% | 61.1% | n/a | ns | 58.6% |
| Body mass index | 26.76 (3.94) | 25.33 (3.64) | 0.38 | ns | 28.2 (6.57) |
| Bedtime (hour) | 2301 (1.60) | 2317 (1.56) | 0.17 | ns | 2227 (3.23) |
| Waketime (hour) | 0621 (1.80) | 0618 (2.09) | 0.02 | ns | 0759 (2.82) |
| Time in bed (minute) | 465.88 (35.76) | 454.17 (36.31) | 0.32 | ns | 479.90 (169.35) |
| Total sleep time (minute) | 427.06 (30.52) | 425.17 (43.75) | 0.05 | ns | 407.42 (75.90) |
| Sleep onset latency (minute) | 14.82 (7.46) | 18.89 (16.64) | 0.32 | ns | 45.28 (80.30) |
| Number of night awakenings | 0.68 (0.64) | 1.08 (1.96) | 0.28 | ns | 2.82 (5.46) |
| Wake after sleep onset (minute) | 10.18 (17.11) | 3.47 (4.21) | 0.54 | ns | 27.17 (50.66) |
| Insomnia Severity Index | 1.94 (2.30) | 2.67 (3.20) | 0.26 | ns | 3.81 (3.58) |
| FIRST | 16.76 (3.82) | 19.94 (5.73) | 0.65 | .06 | 16.24 (4.35) |
| Arousal predisposition scale | 22.35 (5.37) | 26.28 (6.18) | 0.68 | .05 | n/a |
FR = Familial risk, EPIC sample = Evolution of Pathways to Insomnia Cohort study sample, FIRST = Ford Insomnia Response to Stress Test, n/a = not applicable, ns = not significant. Significance set at p = .05.
Mean, Standard Deviation, and Effect Size for Group Differences (Cohen’s d) for Demographic and Sleep Variables.
| Variables | Negative FR | Positive FR | Cohen’s d | p | EPIC sample |
|---|---|---|---|---|---|
| Age | 47.76 (11.05) | 45.56 (10.26) | 0.21 | ns | 44.71 (13.78) |
| Sex (% F) | 41.2% | 61.1% | n/a | ns | 58.6% |
| Body mass index | 26.76 (3.94) | 25.33 (3.64) | 0.38 | ns | 28.2 (6.57) |
| Bedtime (hour) | 2301 (1.60) | 2317 (1.56) | 0.17 | ns | 2227 (3.23) |
| Waketime (hour) | 0621 (1.80) | 0618 (2.09) | 0.02 | ns | 0759 (2.82) |
| Time in bed (minute) | 465.88 (35.76) | 454.17 (36.31) | 0.32 | ns | 479.90 (169.35) |
| Total sleep time (minute) | 427.06 (30.52) | 425.17 (43.75) | 0.05 | ns | 407.42 (75.90) |
| Sleep onset latency (minute) | 14.82 (7.46) | 18.89 (16.64) | 0.32 | ns | 45.28 (80.30) |
| Number of night awakenings | 0.68 (0.64) | 1.08 (1.96) | 0.28 | ns | 2.82 (5.46) |
| Wake after sleep onset (minute) | 10.18 (17.11) | 3.47 (4.21) | 0.54 | ns | 27.17 (50.66) |
| Insomnia Severity Index | 1.94 (2.30) | 2.67 (3.20) | 0.26 | ns | 3.81 (3.58) |
| FIRST | 16.76 (3.82) | 19.94 (5.73) | 0.65 | .06 | 16.24 (4.35) |
| Arousal predisposition scale | 22.35 (5.37) | 26.28 (6.18) | 0.68 | .05 | n/a |
| Variables | Negative FR | Positive FR | Cohen’s d | p | EPIC sample |
|---|---|---|---|---|---|
| Age | 47.76 (11.05) | 45.56 (10.26) | 0.21 | ns | 44.71 (13.78) |
| Sex (% F) | 41.2% | 61.1% | n/a | ns | 58.6% |
| Body mass index | 26.76 (3.94) | 25.33 (3.64) | 0.38 | ns | 28.2 (6.57) |
| Bedtime (hour) | 2301 (1.60) | 2317 (1.56) | 0.17 | ns | 2227 (3.23) |
| Waketime (hour) | 0621 (1.80) | 0618 (2.09) | 0.02 | ns | 0759 (2.82) |
| Time in bed (minute) | 465.88 (35.76) | 454.17 (36.31) | 0.32 | ns | 479.90 (169.35) |
| Total sleep time (minute) | 427.06 (30.52) | 425.17 (43.75) | 0.05 | ns | 407.42 (75.90) |
| Sleep onset latency (minute) | 14.82 (7.46) | 18.89 (16.64) | 0.32 | ns | 45.28 (80.30) |
| Number of night awakenings | 0.68 (0.64) | 1.08 (1.96) | 0.28 | ns | 2.82 (5.46) |
| Wake after sleep onset (minute) | 10.18 (17.11) | 3.47 (4.21) | 0.54 | ns | 27.17 (50.66) |
| Insomnia Severity Index | 1.94 (2.30) | 2.67 (3.20) | 0.26 | ns | 3.81 (3.58) |
| FIRST | 16.76 (3.82) | 19.94 (5.73) | 0.65 | .06 | 16.24 (4.35) |
| Arousal predisposition scale | 22.35 (5.37) | 26.28 (6.18) | 0.68 | .05 | n/a |
FR = Familial risk, EPIC sample = Evolution of Pathways to Insomnia Cohort study sample, FIRST = Ford Insomnia Response to Stress Test, n/a = not applicable, ns = not significant. Significance set at p = .05.
Procedures
Prior to study participation, volunteers were asked to refrain from naps and to maintain a consistent sleep schedule based on habitual sleep times for one week. Adherence to the sleep schedule was verified using a sleep diary (see Table 1 for descriptive statistics on sleep variables by familial risk groups). Participants were also asked to refrain from use of alcohol, caffeine, tobacco, and any illicit substance for 24 hours prior to testing. The experimental procedure included three components, beginning with collection of baseline measures, followed by the speech preparation and stress tasks, and ending with a 40-minute recovery period. Psychological and physiological measures of stress were collected throughout the experimental protocol, including the baseline period, prespeech preparation, pre- and post-TSST, and every 5 minutes during the recovery period.
Psychological Stress Response
Subjective experiences of stress in response to the experimental procedure were measured using a visual analog scale. Participants were instructed to “Please make a mark on the line indicating how anxious you feel right now,” with the line anchored by “Not at all” (score of 0) on the left and “Extremely” (score of 100) on the right.
Physiological Stress Response
The autonomic stress response was measured using heart rate, mean arterial pressure, and salivary α-amylase. Heart rate was monitored using a Masimo SET pulse oximeter, and blood pressure was monitored using a portable HealthSmart blood pressure machine. Mean arterial pressure (MAP) was calculated by adding one-third of the pulse pressure (subtracting the diastolic pressure from the systolic pressure) to the diastolic pressure.51 α-Amylase was also assayed in salivary samples to index autonomic nervous system activation.52 HPA-axis was measured using cortisol, which was also assayed in salivary samples. Saliva samples were obtained using an oral swab for quick and hygienic sampling (Salimetrics, State College, PA) and frozen immediately following collection. Saliva samples for the last five samples were lost for one participant due to technical difficulties.
TSST Protocol
Baseline
The TSST29 experimental sessions were run between the hours of 1100–1600. Upon arrival, subjects remained seated for ~30–45 minutes while baseline measurements were recorded. A baseline measure of psychological stress was collected using the visual analog scale measuring current anxiety. Physiological measures of stress were collected using heart rate, blood pressure, α-amylase, and cortisol. Baseline heart rate and blood pressure were indexed by an average of three samples obtained during this period, and α-amylase and cortisol were assayed using one saliva sample.
Speech Preparation
Upon completion of the baseline measures, participants were taken to the speech room, where three hospital staff members (both males and females) were sitting at a rectangular table, and a video camera was installed and aimed at the head of the table. The participant was instructed to stand in front of the three staff members while the research assistant instructed the participant to assume the role of a job applicant invited for a personal interview with company’s staff managers. Participants were informed that following a preparation period, they were to present a 5-minute speech persuading the hospital staff that he or she was the perfect applicant for the vacant position. The staff was also introduced as being specially trained to monitor nonverbal behavior, and that the audio and visual recording would be subjected to both voice and video frequency analyses.
Following these instructions, the participants returned to the previous room and were given 10 minutes for speech preparation. Each participant was provided with paper and pencils for outlining their speech; however, written material was not allowed during the speech. Communication with the research assistant was limited during the speech preparation phase. Psychological and physiological measurements were taken following speech preparation.
Stress Tests
Following speech preparation, participants were returned to the speech room in a wheelchair to control for postural variance in blood pressure and seated at the head of the table in front of the three hospital staff. The lead manager welcomed the participant as a job applicant and asked him or her to deliver the speech for the next 5 minutes. If the speech concluded in <5 minutes, the manager would prompt continuation by stating, “You still have some time left. Please continue.” Once the designated 5 minutes was complete, the previously enumerated measurements were promptly taken. An oral mental arithmetic challenge immediately followed the speech task. Participants were instructed to serially subtract 13 from 1022 with high speed and accuracy for 5 minutes. Following each error, participants were instructed by the research assistant to restart at 1022. Psychological and physiological measurements were taken again and continued throughout the recovery period.
Recovery Period
Participants were allowed 40 minutes for recovery following the stress tests, during which they remained in the speech room along with only a research assistant present and preselected National Geographic magazines. Psychological and physiological measurements were repeated every five minutes until the end of the recovery period. At the end of the study, participants were debriefed about the goal of the study and informed that neither a voice frequency nor a video analysis would be performed.
Response and Management of Personal Stressors
In addition to the experimental stress task, participants also completed questionnaires that assessed for management of personal stressors experienced outside of the laboratory. These were completed prior to the TSST protocol during baseline measurements and included the Impact of Events Scale (IES), Sense of Control Scale, and the Ford Insomnia Response to Stress Test (FIRST).
Impact of Events Scale
The IES is a fifteen-item, self-report measure of intrusive and avoidant cognitions occurring in response to stressful or traumatic events.53 The cognitive intrusion subscale is exemplified by items such as “I thought about it when I didn’t mean to,” and “Any reminder brought back feelings about it.” The avoidance subscale is exemplified by items such as “I tried not to think about it,” and “I was aware that I still had a lot of feelings about it, but I didn’t deal with them.” Items are rated on a five-point scale ranging from “not at all” (0) to “extremely” (4).
Control Beliefs
A twelve-item self-report measure of personal control was also delivered. This scale was modified using Pearlin and Schooler’s54 mastery scale and was used originally for the Midlife in the United States Survey.55 This measure includes two subscales that measure a sense of mastery and constraint. The mastery subscale is exemplified by items such as “I can do just about anything I really set my mind to,” and “When I really want to do something, I usually find a way to succeed at it.” The constraint subscale is exemplified by items such as “I often feel helpless in dealing with the problems in life,” and “There is little I can do to change the important things in my life.” Items are rated on a seven-point scale by how strongly they agreed with each statement (1 = strongly agree and 7 = strongly disagree).
Ford Insomnia Response to Stress Test
The FIRST is a validated measure of trait sleep reactivity to stress and is operationalized as a propensity for sleep disruption in response to stressful events.56 The FIRST asks how likely a respondent is to have difficulty sleeping in nine distinct situations on a four-point scale (1 = not likely and 4 = very likely): before an important meeting the next day, after a stressful experience during the day, after a stressful experience in the evening, after getting bad news during the day, after watching a frightening movie or TV show, after having a bad day at work, after an argument, before having to speak in public, and before going on vacation the next day. Participants were asked to rate the likelihood of sleep disturbance even if they had not experienced the situation recently.
Statistical Approach
Changes in psychological and physiological stress responses across the experiment were modeled using a mixed-effects approach because it is more robust in handling unequal sample sizes and missing data than the traditional repeated-measures analysis of variance57; however, one participant was dropped from analyses with α-amylase and cortisol because half of the salivary samples were lost due to laboratory error. Stress responses across the experiment were modeled primarily using Time as a quadratic function (Time2) to appropriately capture non-linear changes in stress across the study (e.g., an inverted-U function to model an initial increase in anxiety during the TSST that subsequently decreases throughout the recovery period); Time as a linear function was also included in the model as the lower order polynomial. Group differences were modeled using Familial risk groups as a categorical factor (positive or negative), entered as both a main effect and interactions with Time (both linear and quadratic). Participant was included as a random effect to account for individual differences in stress responses.
Statistical testing was initiated using a full model that first tested familial risk group as a moderator variable. The interaction of interest was Time2 × Familial risk group, and all lower order terms (i.e., Time, Time2, Familial risk group, and Time × Familial risk group) were included in the model. A significant interaction effect of Time2 × Familial risk group indicated that stress responses to the TSST differed by familial risk groups, and the lower-order terms (i.e., marginal effects) were only examined to aid the interpretation of the Time2 × Familial risk group interaction.
In cases where familial risk was not a significant moderator, the final models were reduced to a main effects model that included Time, Time2, and Familial risk group as fixed effects. In the main effects model, a statistically significant effect of Time2 indicated elicitation of a stress response across groups, and a statistically significant effect of Familial risk group indicates differences in average stress values between familial risk groups. In presence of Time2, the linear effect of Time represented the specific rate of change when Time2 was zero, and was examined only to aid the interpretation of the effect of Time2.
Finally, the area under the curve (AUC) was also calculated for both cortisol and α-amylase values to examine global HPA-axis and autonomic nervous system response to the TSST. While the initial analytical model is sensitive to change across time, there are specific circumstances in which different patterns of change are associated with comparable cumulative exposure (e.g., a steep increase and decrease in cortisol compared with a slow but sustained increase); as such, a secondary analytical approach using the AUC was also employed. AUC was calculated using the trapezoid formula correcting for baseline values.58 AUC was also tested using a mixed-effects model with Familial risk group as the fixed effect and participant as the random effect.
RESULTS
Group Characteristics
Comparisons between familial risk groups using a one-way analysis of variance (ANOVA) revealed no differences on demographic variables (age, sex, and BMI), self-reported sleep variables (sleep timing, sleep duration, sleep quality, etc.), or by scores on the Insomnia Severity Index (ISI; see Table 1). As expected, marginal differences were detected on sleep reactivity to stress as assessed by scores on the FIRST, with FIRST scores increasing with familial risk, F(1, 33) = 3.68, p = .06. Familial risk groups also differed by scores on the arousal predisposition scale (APS), with APS scores increasing by familial risk, F(2, 33) = 4.00, p = .05.
Laboratory Subjective Stress Response
To confirm the successful induction of anxiety in response to the stress test, analyses examined scores from the visual analog scale during the TSST. Results from the mixed-effects model revealed a significant main effect of Time2, F(1, 348) = 95.17, p < .0001, confirming that the TSST elicited a significant increase in subjective anxiety that then subsided during the recovery period. Results did not reveal a main effect of Familial risk, or an interaction of Time2 × Familial risk, indicating that familial risk groups did not differ in subjective stress response trajectories following the TSST.
Laboratory Autonomic Stress Response
Mean arterial pressure, heart rate, and salivary α-amylase were examined to confirm induction of an autonomic stress response to the TSST. Results from the mixed-effects models confirmed a significant response for all three outcome variables following the stressor (see Table 2); however, none of the three markers of autonomic stress response differed by familial risk for insomnia. Additionally, results also showed no differences between familial risk groups in AUC of α-amylase across the experimental period.
Differences in Autonomic Stress Response (Mean Arterial Pressure (MAP), Heart Rate, α-Amylase, and Cortisol) by Familial Risk Groups.
| Moderator model | Main-effects model | ||
|---|---|---|---|
| Dependent Variable | Time2 × Familial risk | Time2 | Familial risk |
| MAP | β = 0.19 ± 0.17, ns | β = −0.17 ± 0.08, p = .05 | β = −0.23 ± 0.26, ns |
| Heart rate | β = 0.09 ± 0.16, ns | β = −0.31 ± 0.11, p < .01 | β = −0.31 ± 0.25, ns |
| α-Amylase | β = −0.03 ± 0.22, ns | β = −0.70 ± 0.11, p < .0001 | β = 0.17 ± 0.27, ns |
| Cortisol | β = 0.37 ± 0.19. p = .05 | – | – |
| Moderator model | Main-effects model | ||
|---|---|---|---|
| Dependent Variable | Time2 × Familial risk | Time2 | Familial risk |
| MAP | β = 0.19 ± 0.17, ns | β = −0.17 ± 0.08, p = .05 | β = −0.23 ± 0.26, ns |
| Heart rate | β = 0.09 ± 0.16, ns | β = −0.31 ± 0.11, p < .01 | β = −0.31 ± 0.25, ns |
| α-Amylase | β = −0.03 ± 0.22, ns | β = −0.70 ± 0.11, p < .0001 | β = 0.17 ± 0.27, ns |
| Cortisol | β = 0.37 ± 0.19. p = .05 | – | – |
Standardized coefficients reported for comparison of effect size across models. The moderator model was tested first to examine whether the stress response differed by familial risk groups (Time2 × Familial risk). If not significant, a main-effects model was completed. Only variables of interest are included in this table (see supplementary tables S1 through S4 for complete models). ns = not significant. Significance set at p = .05.
Differences in Autonomic Stress Response (Mean Arterial Pressure (MAP), Heart Rate, α-Amylase, and Cortisol) by Familial Risk Groups.
| Moderator model | Main-effects model | ||
|---|---|---|---|
| Dependent Variable | Time2 × Familial risk | Time2 | Familial risk |
| MAP | β = 0.19 ± 0.17, ns | β = −0.17 ± 0.08, p = .05 | β = −0.23 ± 0.26, ns |
| Heart rate | β = 0.09 ± 0.16, ns | β = −0.31 ± 0.11, p < .01 | β = −0.31 ± 0.25, ns |
| α-Amylase | β = −0.03 ± 0.22, ns | β = −0.70 ± 0.11, p < .0001 | β = 0.17 ± 0.27, ns |
| Cortisol | β = 0.37 ± 0.19. p = .05 | – | – |
| Moderator model | Main-effects model | ||
|---|---|---|---|
| Dependent Variable | Time2 × Familial risk | Time2 | Familial risk |
| MAP | β = 0.19 ± 0.17, ns | β = −0.17 ± 0.08, p = .05 | β = −0.23 ± 0.26, ns |
| Heart rate | β = 0.09 ± 0.16, ns | β = −0.31 ± 0.11, p < .01 | β = −0.31 ± 0.25, ns |
| α-Amylase | β = −0.03 ± 0.22, ns | β = −0.70 ± 0.11, p < .0001 | β = 0.17 ± 0.27, ns |
| Cortisol | β = 0.37 ± 0.19. p = .05 | – | – |
Standardized coefficients reported for comparison of effect size across models. The moderator model was tested first to examine whether the stress response differed by familial risk groups (Time2 × Familial risk). If not significant, a main-effects model was completed. Only variables of interest are included in this table (see supplementary tables S1 through S4 for complete models). ns = not significant. Significance set at p = .05.
Laboratory HPA-Axis Stress Response
Results from the mixed-effects model with cortisol response to the TSST as the dependent variable (see Table 2) indicated that a significant cortisol response was detected across the sample (main effect of Time2), with an increase following the stress tasks and a decrease during the recovery period. Average cortisol values across the TSST in the positive familial risk group were also lower relative to the negative familial risk group. Additionally, change in cortisol level across the experimental period differed by familial risk groups. Comparisons at individual time points further indicated that the slopes at time points +10, +15, +20, +25, and +30 were significantly lower for the positive versus negative familial risk group, indicating a blunted cortisol response in those at risk for insomnia (see Figure 2). Examination of the AUC for cortisol across the experimental period also revealed that those with familial risk exhibited a 60% decrease in cortisol response compared with those without familial risk, F(1, 32) = 5.45, p < .05.
Cortisol response by negative and positive familial risk groups across the TSST timeline. The baseline period represents cortisol levels collected during 30–45 minutes of rest before any experimental procedure. Cortisol values were taken at the beginning and end of the 10-minute speech anticipation period, during which participants outlined their speech after experimental instructions were provided. Cortisol was also collected at the beginning and end of the TSST, prior to the speech task and then immediately following the mental arithmetic task. A recovery period of 40 minutes occurred following the conclusion of the TSST, where participants remained in the speech room. Significant differences in the rate of change in cortisol between familial risk groups are indicated by asterisks (***p < .001, *p < .05).
A post hoc exploratory analysis was also completed to examine whether cortisol responses (both change across time and AUC) differed in a linear dose-response fashion with respect to varying degrees of familial risk for insomnia. To test this, the positive familial risk group was further divided into medium (N = 13) and high (N = 5) risk groups if one or both biological parents experienced insomnia, respectively. Familial risk was entered as an ordinal variable using orthogonal polynomial coding to preserve the ordered relationship between the three groups. Results revealed a significant linear trend for both changes across time, Time2 × Familial Risk, β = 0.52, p < .01, and for AUC, β = −1.03, p < .01, indicating that cortisol decreased linearly by equal proportions based on familial risk for insomnia (see Figure 3 for cortisol values across the experiment by negative, medium, and high familial risk groups).
Cortisol response by negative, medium, and high familial risk groups across the TSST timeline. The baseline period represents cortisol levels collected during 30–45 minutes of rest before any experimental procedure. Cortisol values were taken at the beginning and end of the 10-minute speech anticipation period, during which participants outlined their speech after experimental instructions were provided. Cortisol was also collected at the beginning and end of the TSST, prior to the speech task and immediately following the mental arithmetic task. A recovery period of 40 minutes occurred following the conclusion of the TSST, where participants remained in the speech room. Results indicate that cortisol response decreased linearly with increasing familial risk.
Cortisol, Familial Risk, and Management of Personal Stressors
HPA-axis functioning in the lab was also compared with participants’ responses to personal life stressors that have occurred in the preceding year. Responses on the Impact of Events Scale and Control Beliefs Scale were tested as the dependent variables via multiple linear regressions in relationship with AUC for cortisol, and its interaction with familial risk as independent variables. Results indicated that a lower AUC for cortisol was independently associated with more avoidance, β = −0.48, p < .01, and intrusive cognitions, β = −0.37, p < .05, following personal stressors that had occurred during the past year. The lower AUC for cortisol was also associated with higher feelings of constraint, β = −0.34, p < .05. Results also indicated that higher familial risk for insomnia exacerbated intrusive cognitions in relationship to lower AUC for cortisol, AUCCORT × Familial Risk, β = −0.97, p < .05, and also exacerbated feelings of constraint, AUCCORT × Familial Risk, β = −0.64, p = .05. The relationship between cortisol and avoidance was only marginally associated with familial risk for insomnia, β = −0.51, p = .10.
DISCUSSION
This study examined whether differences in stress regulation are associated with the diathesis for insomnia. We hypothesized that individuals with familial risk for insomnia might exhibit abnormalities in HPA-axis response to the laboratory stressor, but not in the autonomic nervous system response. Results confirmed our hypothesis, revealing that those with familial risk for insomnia exhibit a blunted cortisol response to stress. Additional findings also suggested that blunting of cortisol in response to stress may occur in a linear dose response manner with increasing familial risk (i.e., one or both biological parents with insomnia).
The pattern of results indicates that although the laboratory stressor successfully induced a psychological and autonomic nervous system response, those with familial risk for insomnia exhibited a deficit in mounting an adaptive cortisol response. The release of glucocorticoids plays a major role in energy mobilization and also restores homeostasis by initiating a negative feedback response that reduces production of catecholamines and glucocorticoids.30,59 If this regulatory response is blunted, consequences may include inadequate allocation of energy toward a psychological or behavioral response to the stressor and may also lead to incomplete physiological restoration that over time can increase allostatic load and vulnerability to illnesses.60,61 For example, blunted cortisol response to stress is associated with various forms of psychopathology, such as depression,62 alcohol and substance abuse,63 and schizophrenia.64 This is also consistent with evidence from the present study showing that individuals with blunted cortisol response to stress engaged in less effective coping strategies and reported increased cognitive intrusion, which can in turn increase the risk for depression and anxiety (i.e., rumination). Furthermore, prior studies have also shown that a blunted cortisol response prospectively predicts a trajectory of emotion dysregulation and poor coping over time.32 Future research may utilize a prospective design in larger samples to replicate this effect and to further examine the potential clinical significance and downstream consequences of the blunted cortisol response associated with familial risk for insomnia.
Importantly, findings from this study provide initial evidence that diathesis for insomnia may be a heritable characteristic that includes abnormalities in the physiological stress response system. This is consistent with prior studies indicating that vulnerability to sleep disruption in response to stress has a significant genetic component.4,27,28 Further studies have suggested that a length polymorphism in the 5HTTLPR serotonin transporter may be associated with the abnormalities in stress reactivity65 and has also been implicated as a candidate gene for insomnia. In fact, a prior study found that stress in caregivers produced worse sleep in individuals with the s-allele polymorphism in the 5HTTLPR serotonin transporter compared with those who did not.66Finally, extant insomnia studies do not typically account for familial risk, and thus may result in significant heterogeneity in the distribution of familial risk between study samples. If HPA axis reactivity is different in insomniacs with high versus low heritability, then variability in heritability across study samples may lead to inconsistent results with respect to HPA functioning. Indeed, this may explain the inconsistent replication of differences in HPA axis response to stress in insomnia. Furthermore, the effect of familial risk may also be diluted when group means are calculated using samples with and without familial risk for insomnia.
Results from this study also suggest that this heritable vulnerability may be associated predominantly with HPA-axis functioning rather than the autonomic nervous system, at least early in the evolution to insomnia. This would suggest that individuals at risk for insomnia do not show differences in immediate responses to stress (i.e., autonomic nervous system), but rather exhibit inadequate recovery from stressors (i.e., HPA-axis). This biological predispositional factor is reflected psychologically in the self-reported increased predisposition for arousal to stress and residual cognitive intrusion after stressors. In fact, these psychological factors have also been identified as moderators in the relationship between stress and incident insomnia 67. Over time, those with a blunted cortisol response who have increased exposure to accumulating or acute environmental stressors may develop incident insomnia related to inadequate stress recovery. Indeed, this is consistent with the phenomenology of insomnia, often described as an “inability to shut off the mind” to achieve or maintain sleep. Additionally, blunted cortisol response and the associated deficits in coping have also been associated with psychiatric illnesses such as depression,62 and thus may represent the biological underpinnings of sleep disturbance as a transdiagnostic risk factor.
Given that this is a preliminary study, we relied on a cross-sectional design in a small sample of premorbid healthy sleepers with self-reported familial risk. The use of a mixed-effects model conserved degrees of freedom (and thus statistical power) by modeling time as a continuous instead of a discrete variable; however, the relatively small total sample size may impact the reliability of the estimates and thus should be interpreted with caution. An additional consideration is that the sample size of individuals with high familial risk was smaller than the sample size of the medium and low familial risk groups, though this may reflect the small population of premorbid adults without insomnia despite a strong family history. Future studies may dedicate more resources toward replicating these findings in larger studies with more stringent controls. This will likely entail a longitudinal design in larger samples of individuals with genetically verified familial risk, with objective measures of sleep (e.g., polysomnography or actigraphy) and circadian rhythms (e.g., dim light melatonin onset). To make specific and more generalizable inferences regarding the role of cortisol in the pathophysiology of insomnia, individuals with genetic differences in candidate genes should also be studied. Although the present results are associated with familial risk for insomnia, it is also possible that the blunted cortisol response to stress is a consequence of physiological wear-and-tear over time, rather than an inherited trait deficit. A longitudinal research design would further clarify this and would have potential implications for intervention and prevention.
CONCLUSION
This study examined stress regulation in individuals with and without familial risk for insomnia to determine whether diathesis for insomnia is characterized by deficits in the physiological stress response system. Results indicated that despite an increase in anxiety and autonomic nervous system activity following the laboratory stressor, individuals with familial risk for insomnia exhibited a blunted cortisol response relative to those without familial risk for insomnia. Furthermore, individuals with blunted cortisol response to a laboratory stressor were also more likely to report heightened response to personal life stressors, including increased sleep disturbances, elevated cognitive intrusions, and more behavioral avoidance. Given the adaptive function of cortisol in restoring physiological homeostasis, those with blunted cortisol in response to stress may need to rely on other allostatic response systems, such as the inflammatory cytokines, elevations of which can exacerbate vulnerability to illnesses.11
SUPPLEMENTARY MATERIAL
Supplementary material is available at SLEEP online.
DISCLOSURE STATEMENT
None declared.
ACKNOWLEDGMENTS
This study was supported by a National Institute of Mental Health grant (MH-082785) to C.L.D. The authors would like to thank the technical staff of Henry Ford Hospital Sleep Center for their invaluable assistance in the completion of the present study.
REFERENCES
Comments