The ICU is a complex and stressful environment with significant, unfavorable physical, cognitive, psychologic, and functional consequences for both patients and families (1–3). Patients who survive critical illness experience high rates of anxiety, depression, and posttraumatic stress disorder (PTSD) that persist months to years after hospital discharge (1). This constellation of psychologic morbidities is an important component of postintensive care syndrome (PICS) (4). Studies also demonstrate that family members of ICU patients commonly suffer from psychologic disorders including persistent anxiety, depression, and PTSD (2 , 5–7) and may experience a decrease in quality of life that may persist for 2 years or more (7). Such psychologic disorders among ICU family members are characteristics of PICS-family (PICS-F) (8) and are often associated with significant financial and emotional burden (9).
Multiple recent studies have begun to elucidate risk factors for PICS-F, including female sex, age of family member or patient, history of anxiety, and amount of social support (6 , 7 , 10). Family members also appear to be at higher risk for persistent psychologic distress if they were involved in medical decision-making or perceived that communication with clinicians was inadequate (2). Previous studies have not evaluated whether physiologic markers of acute stress in the family members of adult patients are associated with subsequent PICS-F.
Evidence from other types of traumatic events (e.g., automobile collisions, combat exposure, psychosocial or workplace stress, myocardial infarction) suggests that acute physiologic stress may manifest in aberrations in the hypothalamus-pituitary axis (HPA) and may predispose an individual to persistent psychologic distress (11 , 12). The HPA regulates the stress response by modulating cortisol secretion (12), and physical, emotional, and intellectual stresses are individually associated with alterations in the normal pattern of serum cortisol secretion (13). Previous studies in non-ICU populations have shown that cortisol secretion is elevated in patients with acute and chronic anxiety (14 , 15).
We sought to determine whether salivary cortisol measured early in an ICU admission is associated with subsequent anxiety among ICU family members. We hypothesized that elevation in salivary cortisol would be associated with anxiety 3 months later among ICU family members.
In a prospective cohort study, we enrolled adult family members of patients newly admitted to a multidisciplinary medical/surgical ICU with an admission Acute Physiology and Chronic Health Evaluation (APACHE) II (16) score greater than 15 (a score associated with at least a 10–20% hospital mortality in contemporary environments). Study inclusion criteria were ability to speak and read English. Participants were enrolled within the first 24 hours of the patient’s ICU admission. Study eligibility criteria excluded pregnant or breastfeeding females, prisoners, and children (age < 18 yr) or a known history of PTSD, dementia, or schizophrenia (17). Study participants who used steroid-containing medications, which interfere with cortisol secretion, were also excluded (18). To ensure enrollment within the acute phase of ICU exposure, family members of patients who had been transferred from another ICU after an inpatient stay longer than 24 hours or had a previous ICU or long-term acute care hospital admission within the 90 days prior to the index admission were also excluded. One study participant per patient was eligible for study enrollment. If more than one family member was available, participants were prioritized based on their relationship with the patient according to the algorithm depicted in Figure E1 (Supplemental Digital Content 1, http://links.lww.com/CCM/D18), with preference for a spouse or family member living with the patient.
This study was approved by the Intermountain Medical Center Institutional Review Board (number 1040305). Written consent was obtained from all participants. Each participant received a $50 gift card upon return of the cortisol samples, and another $50 gift card on completion of the 3-month telephone interview.
Demographic characteristics and medical history of study participants were obtained, including self-reported history of anxiety or depression and use of medications or substances known to interfere with salivary cortisol secretion or collection (18). A modified Perceived Stress Scale (PSS) was also collected at study enrollment to record participant’s levels of perceived stress (19).
Characteristics of the ICU patient were also collected, including admission APACHE II score, Elixhauser comorbidity score (20), age, sex, ICU and hospital length of stay, in-hospital mortality, and 3-month mortality, as well as the patient’s number of prior ICU admissions in the last 5 years, based on participant report.
We obtained five consecutive saliva samples from participants beginning the morning after study enrollment. Samples were obtained upon awakening and at prespecified time intervals throughout the day (immediately upon awakening, 30 min post awakening and before breakfast, 30 min before lunch, 30 minutes before dinner, and just before bedtime) (21). Saliva samples were collected using a Salivette (Sarstedt AG & Co, Nümbrecht, Germany), a sampling device for measuring salivary cortisol. Participants were instructed to follow standard protocol for salivary cortisol measurements using the salivette (21). Samples were collected the next day from participants by a research coordinator and stored at –20°C until analysis (22). Cortisol levels were evaluated using quantitative commercial enzyme immunoassay with chemiluminescence detection (chemiluminescence immunoassay; IBL-Hamburg, Germany). The assay has a lower detection limit of 0.1 nmol/L with intra- and interassay coefficients of variation less than 8% (21). Cortisol via saliva sampling has been shown to have high correlation (r = 0.90) with cortisol from serum and plasma samples (23). The area under the curve with respect to ground (AUCg) was the prespecified primary measure of salivary cortisol (24). The AUCg is an established method of reporting cortisol levels and is obtained by calculating the area under the curve and above the baseline (ground) when results are graphed over time (for a sample calculation, see Fig. E2, Supplemental Digital Content 1, http://links.lww.com/CCM/D18). We normalized the AUCg values for total time elapsed between the first and last samples. Cortisol awakening response (CAR) was a prespecified secondary measure of salivary cortisol. CAR represents the increase in cortisol level from the first (awakening) sample to the following sample 30 minutes later (25).
Cortisol values were excluded from analysis if any value was greater than 3 SDs from the mean of all samples for the entire population or if the elapsed time between the first two samples (used to calculate the CAR) was greater than 90 minutes (i.e., > 1 hr late). Participants’ data were excluded entirely from analysis if they did not provide the cortisol samples necessary to calculate either AUCg or CAR.
The primary outcome was anxiety, assessed during a telephone follow-up 3 months (± 1 mo) after enrollment, as measured by the Hospital Anxiety and Depression Scale-Anxiety (HADS-A) (26). The Hospital Anxiety and Depression Scale (HADS) screens for anxiety and depression using a 14-item scale with scores ranging from 0 to 21 for anxiety and depression; a HADS score greater than or equal to 8 on either section is indicative of possible or probable anxiety or depression, respectively (26). In addition to the primary outcome, we also screened for depression (HADS-depression [HADS-D] ) and symptoms of posttraumatic stress (Impact of Event Scale-Revised [IES-R] ) during the 3-month follow-up. The IES-R is a 22-item scale measuring intrusion, avoidance, and hyperarousal; a mean score of greater than or equal to 1.6 indicates PTSD symptoms (27).
Our prespecified primary analysis was a logistic regression of AUCg on anxiety at 3 months, controlling for a history of self-reported anxiety. In a prespecified secondary analysis, we repeated the primary analysis, using the CAR instead of AUCg. We also report secondary outcomes including depression (HADS-D score ≥ 8) and symptoms of PTSD (IES-R score ≥ 1.6).
Sensitivity analyses were performed to determine whether patient severity of illness or participant PSS altered the association between CAR and anxiety after correcting for a history of self-reported anxiety. Sensitivity analyses were also conducted to control for potentially relevant covariates using backward stepwise logistic regression. In addition to our predictors of interest (AUCg, CAR), candidate covariates included age, sex, tobacco history, alcohol use, obesity, history of depression, history of anxiety, as well as whether the ICU patient had a previous ICU admission within the last 5 years. Covariates were retained in the model using a backward stepwise feature selection strategy that optimized Akaike information criterion.
Sample Size and Power Calculation
The power calculation was performed assuming that the mean estimated HADS-A score among individuals with minimal anxiety was 5 (± 2), while individuals with significant anxiety had mean 8 (± 3). We therefore estimated, with 90% power (alpha = 0.05), to detect a slightly more conservative but still clinically important difference of 2 (26 , 28) on the HADS-A with 100 enrolled participants with a 15% attrition and a ratio of 1.5:1 in abnormal to normal AUCg (based on studies documenting the distribution of cortisol levels in different populations) (29–31). Sample size calculations were performed in PASS v12 (32).
Of 1,031 admissions to the study ICU, 137 (13%) had a family member eligible for the study and 100 participants were enrolled; one participant withdrew after enrollment and before completing study procedures. Ninety-two participants completed follow-up. See Figure 1 for consort diagram.
Study participants’ mean age was 54 (SD 14) years, and 64% were female. The majority (71%) lived with the patient prior to ICU admission, and 53% were the patient’s spouse. A history of anxiety was self-reported by 26 participants (26%). Mean (SD) PSS score at enrollment was 5 (4) and was significantly higher for those with a history of anxiety (4.5 vs 7.0; p = 0.002). Demographic information for study participants are found in Table 1 with additional information available in Table E1 (Supplemental Digital Content 1, http://links.lww.com/CCM/D18).
The ICU patients had a mean age of 60 (17) years, and 51% were female. Mean APACHE II at admission was 30 (7). Forty-seven of these patients (47%) had been admitted to an ICU within the last 5 years. In-hospital mortality was 21%. Further detailed demographic and medical information for patients are displayed in Table E2 (Supplemental Digital Content 1, http://links.lww.com/CCM/D18).
Outcomes and Data Analysis
For all participants, median cortisol levels by normalized AUCg were 3.9 nmol/L (interquartile range [IQR], 2.9–4.7) and by CAR were 1.9 nmol/L (IQR, –2.3 to 8.1). Of the 92 participants who completed follow-up, we had complete samples for AUCg in 80 participants and for CAR in 77 participants, with seven being excluded because more than 90 minutes elapsed between the samples. No cortisol measurements were greater than 3 SDs from the mean.
At 3-month follow-up, among the 92 participants with recorded outcomes, 29 participants (32%) had HADS-A score greater than or equal to 8 indicating possible or probable anxiety. Fifteen (16%) participants had HADS-D score greater than or equal to 8, indicating possible or probable depression, and 14 participants (15%) had IES-R scores consistent with PTSD (≥ 1.6). For our primary outcome, we used a dichotomous assignment instead of continuous measures in all analyses due to an influential outlier in HADS-A scoring.
In our primary analysis, normalized AUCg was not significantly associated with anxiety (odds ratio [OR], 0.94; CI, 0.69–1.26; p = 0.70), after controlling for a history of anxiety. In the prespecified secondary analysis, however, CAR was significantly associated with anxiety (OR, 1.08; CI, 1.01–1.15; p = 0.02) after controlling for a history of anxiety (Fig. 2). In sensitivity analysis, CAR was also significantly associated with 3-month anxiety after controlling for the patient’s admission APACHE II score and history of anxiety (OR, 1.08; CI, 1.01–1.16; p = 0.02). CAR also remained significantly associated with 3-month anxiety after controlling for PSS and history of anxiety (1.09; CI, 1.02–1.17; p = 0.02). An effect plot which depicts the relationship between predictor (CAR) and outcome (3-mo anxiety) at different levels of the predictor while controlling for a history of anxiety is represented in Figure 3. In a sensitivity analysis using backward stepwise logistic regression, the association between CAR and anxiety remained significant (OR, 1.10; CI, 1.03–1.18; p = 0.009) when controlling for other relevant covariates (sex, history of anxiety, and previous ICU admission) (Table 2).
In a prospective, observational study of family members of ICU patients, the CAR, a measure of hypothalamic-pituitary activation commonly associated with physiologic stress (33), was associated with anxiety at 3 months. Although this finding was a prespecified secondary analysis, the association between CAR and anxiety persisted in sensitivity analyses that controlled for patient baseline severity of illness, a participant self-reported history of anxiety, and PSS scores at study enrollment.
The impact of a critical illness on family members of ICU patients has received increasing recognition (6 , 34). Many questions remain about the best interventions, both in and after the ICU, to ameliorate or prevent PICS-F and how to identify those family members most at risk for PICS-F. To our knowledge, ours is the first study in family members of adult ICU patients investigating the link between an acute physiologic disturbance and subsequent anxiety.
Our results confirm the concept that a family member’s vulnerability to anxiety after a loved one’s ICU admission is comprised of both physiologic and psychologic characteristics. Our findings additionally suggest that acute physiologic stress (measured here by salivary cortisol) might identify individuals at highest risk for post-ICU anxiety and raise the possibility that reducing physiologic stress through specific support strategies may lessen post-ICU anxiety. CAR may serve as a physiologic indicator of identifying individuals who may respond to support interventions during the ICU experience. Previous studies of family members of chronically ill patients found that providing communication and emotional support to family members reduces anxiety (35 , 36).
The ideal procedures for measurement of physiologic stress in ICU family members remain unclear and are likely dependent on the complex relationship between acute and chronic stress and cortisol secretion. High stress can impair HPA functioning, resulting in either chronically elevated or blunted cortisol, both of which may be linked to important health outcomes (37 , 38). This blunting effect may have been seen in the participants with chronic stress in our study, possibly due to a chronically ill loved one or other stressors that were not measured in this study. Within each individual, the response to stress is multifaceted and variable—our results reflect these complexities. Given that this population is presumed to be under acute stress, we predicted an activated HPA and elevated levels of salivary cortisol (reflected in both an increased AUCg and CAR) would predict later anxiety. The CAR was associated with post-ICU anxiety in this population, while our primary analysis using AUCg was nonsignificant. The CAR and AUCg reflect discrete parts of the cortisol secretory cycle (39); because the CAR is distinct from other earlier and later circadian cortisol secretion, it may be more sensitive to sleep quality (40)—a potentially salient influence in an ICU setting.
We are unable with this study design to distinguish between acute and chronic stress as causes of patterns of cortisol secretion. As indicated, 47% of participants had a family member with a prior admission to an ICU and as such may have acute stress superimposed on chronic stress. Research is needed to understand the interplay between acute and acute-on-chronic stress for ICU patients and their family members.
Our study has several potential limitations. First, as part of a substudy, participants in this study were asked to leave the patient’s room for additional testing. Some family members declined the study, as they did not want to leave the patient’s room, even for a short time, to meet with the research coordinator. It is possible that those family members have higher stress, and they were not included in this study. This selection bias may limit the generalizability of our findings. However, our rates of declined consent were low, suggesting relatively minor risk of bias. Owing to resource limitations, we were only able to enroll one participant per day, as such many potentially eligible participants were not approached. We do not believe that the process of approaching participants introduced systematic bias, as coordinators approached family members without any preferences for patient or participant types.
We did not collect detailed data on participant’s other potential stressors or baseline medical history, so we may have missed some factors that may influenced the cortisol values and study outcomes. Additionally, it is possible that some participants did not accurately record the timing of their cortisol collection, and our results may have been skewed by inaccurate salivary cortisol data. However, no values in our participants were outside the expected range, suggesting that such an effect was unlikely. To reduce the burden on family members, we only collected salivary cortisol samples on 1 day and thus did not capture intraindividual fluctuations in CAR (41). Finally, we did not measure sleep or sleep quality (42) or regulate all substances and habits known to interact with cortisol secretion.
Family members of ICU patients are a select population vulnerable to long-term effects from the stress precipitated by a loved one’s critical illness. Not all family members develop PICS-F, and identification of at-risk individuals is a key aspect of interventions to prevent or ameliorate the morbidities associated with this syndrome. We identified elevated CAR as a predictor of subsequent anxiety in family members of adult ICU patients. This study is a novel, important step in characterizing physiologic processes that could allow for early targeted therapies to mitigate post-ICU anxiety.
1. Hopkins RO, Weaver LK, Collingridge D, et alTwo-year cognitive, emotional, and quality-of-life outcomes in acute respiratory distress syndrome. Am J Respir Crit Care Med 2005; 171:340–347
2. Azoulay E, Pochard F, Kentish-Barnes N, et alFAMIREA Study Group: Risk of post-traumatic stress symptoms in family members
of intensive care
unit patients. Am J Respir Crit Care Med 2005; 171:987–994
3. Iwashyna TJ, Ely EW, Smith DM, et alLong-term cognitive impairment and functional disability among survivors of severe sepsis. JAMA 2010; 304:1787–1794
4. Elliott D, Davidson JE, Harvey MA, et alExploring the scope of post-intensive care
syndrome therapy and care: Engagement of non-critical care providers and survivors in a second stakeholders meeting. Crit Care Med 2014; 42:2518–2526
5. Anderson WG, Arnold RM, Angus DC, et alPosttraumatic stress and complicated grief in family members
of patients in the intensive care
unit. J Gen Intern Med 2008; 23:1871–1876
6. Cameron JI, Chu LM, Matte A, et alRECOVER Program Investigators (Phase 1: towards RECOVER); Canadian Critical Care Trials Group: One-year outcomes in caregivers of critically ill patients. N Engl J Med 2016; 374:1831–1841
7. Wintermann GB, Weidner K, Strauß B, et alPredictors of posttraumatic stress and quality of life in family members
of chronically critically ill patients after intensive care
. Ann Intensive Care
8. Davidson JE, Jones C, Bienvenu OJFamily response to critical illness: Postintensive care syndrome
-family. Crit Care Med 2012; 40:618–624
9. Van Pelt DC, Milbrandt EB, Qin L, et alInformal caregiver burden among survivors of prolonged mechanical ventilation. Am J Respir Crit Care Med 2007; 175:167–173
10. Hickman RL Jr, Douglas SLImpact of chronic critical illness on the psychological outcomes of family members
. AACN Adv Crit Care 2010; 21:80–91
11. Kudielka BM, Schommer NC, Hellhammer DH, et alAcute HPA axis responses, heart rate, and mood changes to psychosocial stress (TSST) in humans at different times of day. Psychoneuroendocrinology 2004; 29:983–992
12. Smith SM, Vale WWThe role of the hypothalamic-pituitary-adrenal axis in neuroendocrine responses to stress. Dialogues Clin Neurosci 2006; 8:383–395
13. Mouthaan J, Sijbrandij M, Luitse JS, et alThe role of acute cortisol and DHEAS in predicting acute and chronic PTSD symptoms. Psychoneuroendocrinology 2014; 45:179–186
14. Vasiliadis HM, Forget H, Préville MThe association between self-reported daily hassles and cortisol levels in depression and anxiety in community living older adults. Int J Geriatr Psychiatry 2013; 28:991–997
15. Heaney JL, Phillips AC, Carroll DAgeing, depression, anxiety, social support and the diurnal rhythm and awakening response of salivary cortisol
. Int J Psychophysiol 2010; 78:201–208
16. Knaus WA, Draper EA, Wagner DP, et alAPACHE II: A severity of disease classification system. Crit Care Med 1985; 13:818–829
17. Nater UM, Hoppmann CA, Scott SBDiurnal profiles of salivary cortisol
and alpha-amylase change across the adult lifespan: Evidence from repeated daily life assessments. Psychoneuroendocrinology 2013; 38:3167–3171
18. Adam EK, Kumari MAssessing salivary cortisol
in large-scale, epidemiological research. Psychoneuroendocrinology 2009; 34:1423–1436
19. Cohen S, Williamson GSpacapan S, Oskamp SPerceived stress in a probability sample of the United States. The Social Psychology of Health. 1988, pp Newbury Park, CA, Sage, 31–67In
20. Elixhauser A, Steiner C, Harris DR, et alComorbidity measures for use with administrative data. Med Care 1998; 36:8–27
21. Holt-Lunstad J, Steffen PRDiurnal cortisol variation is associated with nocturnal blood pressure dipping. Psychosom Med 2007; 69:339–343
22. Kirschbaum C, Hellhammer DHSalivary cortisol in psychoneuroendocrine research: Recent developments and applications. Psychoneuroendocrinology 1994; 19:313–333
23. Aardal E, Holm ACCortisol in saliva–reference ranges and relation to cortisol in serum. Eur J Clin Chem Clin Biochem 1995; 33:927–932
24. Pruessner JC, Kirschbaum C, Meinlschmid G, et alTwo formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 2003; 28:916–931
25. Clow A, Hucklebridge F, Stalder T, et alThe cortisol awakening response: more than a measure of HPA axis function. Neurosci Biobehav Rev 2010; 35:97–103
26. Zigmond AS, Snaith RPThe hospital anxiety and depression scale. Acta Psychiatr Scand 1983; 67:361–370
27. Weiss DS, Marmar CRThe impact of event scale-revised. Assessing Psychological Trauma and PTSD 2004; 2:168–189
28. Puhan MA, Frey M, Büchi S, et alThe minimal important difference of the hospital anxiety and depression scale in patients with chronic obstructive pulmonary disease. Health Qual Life Outcomes 2008; 6:46
29. Savla J, Granger DA, Roberto KA, et alCortisol, alpha amylase, and daily stressors in spouses of persons with mild cognitive impairment. Psychol Aging 2013; 28:666–679
30. Raff H, Homar PJ, Burns EAComparison of two methods for measuring salivary cortisol
. Clin Chem 2002; 48:207–208
31. Dismukes AR, Shirtcliff EA, Hanson JL, et alContext influences the interplay of endocrine axes across the day. Dev Psychobiol 2015; 57:731–741
32. Hintze JPASS 12. 2013Kaysville, UT, NCSS, LLC,
33. Golden SH, Sánchez BN, Wu M, et alRelationship between the cortisol awakening response and other features of the diurnal cortisol rhythm: The Multi-Ethnic Study of Atherosclerosis. Psychoneuroendocrinology 2013; 38:2720–2728
34. Haines KJ, Denehy L, Skinner EH, et alPsychosocial outcomes in informal caregivers of the critically ill: A systematic review. Crit Care Med 2015; 43:1112–1120
35. Legg LA, Quinn TJ, Mahmood F, et alNon-pharmacological interventions for caregivers of stroke survivors. Cochrane Database Syst Rev 2011; (10):CD008179
36. Parker D, Mills S, Abbey JEffectiveness of interventions that assist caregivers to support people with dementia living in the community: A systematic review. Int J Evid Based Healthc 2008; 6:137–172
37. Saxbe DEA field (researcher’s) guide to cortisol: Tracking HPA axis functioning in everday life. Health Psychology Review 2008; 2:163–190
38. Duan H, Yuan Y, Zhang L, et alChronic stress exposure decreases the cortisol awakening response in healthy young men. Stress 2013; 16:630–637
39. Clow A, Thorn L, Evans P, et alThe awakening cortisol response: Methodological issues and significance. Stress 2004; 7:29–37
40. Stalder T, Kirschbaum C, Kudielka BM, et alAssessment of the cortisol awakening response: Expert consensus guidelines. Psychoneuroendocrinology 2016; 63:414–432
41. Wüst S, Wolf J, Hellhammer DH, et alThe cortisol awakening response - normal values and confounds. Noise Health 2000; 2:79–88
42. Wright CE, Valdimarsdottir HB, Erblich J, et alPoor sleep the night before an experimental stress task is associated with reduced cortisol reactivity in healthy women. Biol Psychol 2007; 74:319–327