Salivary Cortisol Responses to Household Tasks among Couples with Unexplained Chronic Fatigue

Abstract

This study examined salivary cortisol levels in couples in which one member had unexplained chronic fatigue (CF). The couples completed questionnaires and seven household activities in a laboratory setting, and provided salivary cortisol samples prior to and immediately after the activities, and again after completing additional questionnaires and debriefing. The couples rated their interactions as similar to those at home, suggesting ecological validity, and patients with CF experienced the activities as involving more exertion than did their partners. The multilevel model results indicated that patients with CF had overall lower cortisol levels and flatter slopes across repeated measurements than did their significant others. Patients’ and significant others’ (SOs) cortisol concentrations were significantly associated with each other over time. Furthermore, SOs’ cortisol was associated with greater relationship satisfaction and greater observed rates of patients’ illness/pain behaviors per minute, but patients’ levels of cortisol were not associated with relationship variables. This study is the first to examine cortisol in couples with CF; the results are discussed in terms of implications for future research.

Syndromes of unexplained chronic fatigue (CF) include chronic fatigue syndrome (CFS) and idiopathic chronic fatigue (ICF). CF is characterized by at least six months of profound fatigue accompanied by four or more (with CFS) or less than four (with ICF) other symptoms (muscle and joint pain, headache, memory and concentration difficulties, unrefreshing sleep, sore throat, lymphadenopathy, and postexertional malaise) and significantly impaired functioning (Fukuda et al., 1994). The cause of CF and related dysfunction is unknown but may be multifactorial and encompass psychosocial as well as biological factors. Prior research has suggested that CF-related dysfunction and illness behavior may be associated with psychosocial factors, including SOs’ responses (Romano, Jensen, Schmaling, Hops, & Buchwald, 2009; Schmaling, Smith, & Buchwald, 2000). Hypothalamic-pituitary-adrenal (HPA) axis function is implicated in the experience of fatigue. Its end product, cortisol, has been an area of investigation in CF. Cortisol normally peaks soon after awakening and then gradually decreases throughout the day, reaching its lowest point in the evening (Adam & Kumari, 2009). Meta-analytic reviews of case-control studies have concluded that persons with CF have more flattened profiles and less diurnal variability of cortisol activity compared to healthy controls (Papadopoulos & Cleare, 2012; Powell, Liossi, Moss-Morris, & Schlotz, 2013). These previous studies compared independent samples of matched cases and control participants. CF occurs within a social context: a more powerful comparison of cases and controls would be to study couples and examine variability within and between couples (Atkins, 2005).

The primary purpose of this study was to examine cortisol levels over time as patients with CF and their intimate partners/significant others (SOs) did seven household activities. Based on the meta-analyses summarized above, it was hypothesized that patients with CF would evidence flatter slopes over time compared to their SOs. Secondarily, it was hypothesized that couple variables would be associated with cortisol levels over time. Partners in healthy couples synchronize their diurnal cortisol slope (Liu, Rovine, Cousino Klien, & Almeida, 2013), and flatter cortisol slopes have been associated with less relationship satisfaction for women only (Saxbe, Repetti, & Nishina, 2008) and for all participants with no gender difference (Barnett, Steptoe, & Gareis, 2005). Positive couple interactions have resulted in increased cortisol levels whereas negative couple interactions have resulted in relatively little change in cortisol (Fehm-Wolfsdorf, Groth, Kaiser, & Hahlweg, 1999). The associations between cortisol levels and patients’ overt expressions of illness (e.g., verbal complaints of fatigue, grimacing during exertion) have not been studied. In couples with asymmetrical health statuses such as those in the present study, healthy SOs may be largely responsible for compensating for their ill partners’ limitations, which may be reflected in more cortisol variability among SOs. Given that SOs’ cortisol is expected to vary more than patients’, SOs’ observations of the patients’ illness behaviors as they work on household tasks could be relatively stressful; therefore, patient illness behavior was anticipated to be associated with SOs’, but not patients’, cortisol level over time, in addition to global relationship satisfaction.

Method

Participants

Potential participants were patients at a regional tertiary care clinic for chronic fatigue. The clinic records listed 714 patients that met CF diagnostic criteria (Fukuda et al., 1994) based on medical, laboratory, and psychiatric evaluations and had previously indicated a willingness to be contacted about research projects. These patients were sent a letter inviting them to express their interest in the study if they had unexplained chronic fatigue, were between 18 and 70 years of age, lived in the region (within two hours’ drive), and were living with a significant other for at least three months. Interested participants returned a postage-paid form, and were then contacted by research staff that interviewed them to confirm eligibility, including symptom questions to confirm the presence of CFS/ICF symptoms, height and weight information, and exclusionary diagnoses, such as steroid dependent asthma and medication dependent diabetes.

Of these 714 patients, 238 were unable to be contacted or did not respond to study recruiters’ attempts to contact them. Of the remaining 476 patients, 245 were contacted but were deemed ineligible for the study when information obtained by interview disqualified them from participation, such as no longer meeting CF criteria. From the remaining 231 eligible patients, 125 were recruited for the study. Of these, 85 were unwilling to travel to the laboratory to participate in the procedures described below. Of the remaining 40 subjects, three discontinued participation prior to the laboratory procedures, and six did not have intimate SOs, leaving 31 participants. Participants’ honoraria varied based on the research components they completed.

To participate in the study, SOs also had to be between the ages of 18-70 years at the time of entry into the study and not have a diagnosis of CF or other physical condition that significantly limited their functional abilities. Both patient and SO needed to be able to speak, read, and write English.

Thirty-one couples participated: the patients were predominantly female (n = 29, 93.5%), consistent with the gender predominance in the population (van't Leven, Zielhuis, van der Meer, Verbeek, & Bleijenberg, 2010), and the SOs were predominantly male (n = 26, 83.9%); there were 26 heterosexual and five same-sex couples. The patients averaged 43.3 years of age (SD = 10.2), and the SOs averaged 44.0 years of age (SD = 11.9); the patients were non-Hispanic white (96.8%), had completed two years of college, and 45% were employed in full- or part-time positions. The patients characterized their relationship with their SOs as follows: married, 74.2%, and living together, 25.8%. They had lived together an average of 14.26 years (SD = 10.99). The patients reported having had CFS for an average of 12.4 years (SD = 7.1), and the majority met case criteria for CFS (n = 29; 93.5%) with the rest meeting case criteria for ICF.

Procedure and Materials

Eligible and interested participants were mailed a consent form prior to the scheduled research appointment. Upon arrival at the laboratory, the patient and SO completed consent forms, and then began completing a set of questionnaires (see below). Thirty minutes after arrival at the laboratory, they were asked to provide a saliva sample. They then were videotaped as they performed seven household activities together: sweep the floor, pick up toys from the floor, make a bed, fold laundry, carry and stack fireplace logs, shelve books, and play a game requiring concentration, Triominos© (Pressman Toy Corporation). After completion of each task they rated their perceived exertion on a 15-point scale. After completion of all seven activities, they provided a second saliva sample. Then, the couple completed a final questionnaire, including questions regarding their perceptions of how similar their interactions with their SO were compared to home, participated in a debriefing procedure about the study, and finally, they provided a third saliva sample.

Self-report measures

Participants provided demographic information and a set of questionnaires. Global relationship satisfaction was measured by the 7-item version Abbreviated Dyadic Adjustment Scale (ADAS) (Sharpley & Rogers, 1984); total ADAS scores range from 0-36. In this sample, patient and SO DAS scores showed good internal consistency (α = 0.86 and 0.81, for n = 27 and 28, respectively, among participants who answered all seven items; one additional patient and one SO had one item missing; their extrapolated total scores based on their responses to non-missing items were included in other analyses). The total score was retained for analysis: higher scores indicated more satisfaction. Patients’ scores averaged 19.20 (SD = 6.09) and SOs’ scores averaged 24.77 (SD = 4.71). Sharpley & Rogers (1984) suggested using the SD (SD = 5.4) and means of their samples (the average score of their married and living together samples was 23.45 and the average of the separated and divorced samples was 14.30) to classify relationship adjustment. Twelve patients (41.38%) and five SOs (16.67%) had relatively low ADAS scores of 18 or less, or one SD below the married and living together samples’ average in the Sharpley & Rogers study (1984).

As described above, participants rated their perceived exertion, and that of their partner, after completing each of the seven activities. The well-established rate of perceived exertion (RPE) scale (Borg, 1970) ranges from a value of 6 (“very, very, very light”) to 20 (“more than very, very hard”), and has been used in previous studies of chronic fatigue, including during submaximal exercise testing (Schmaling et al., 2005). The average RPE across tasks was retained for analysis. There was one missing data point for the SO ratings of patient exertion after one task.

Finally, patients and SOs rated the degree of similarity of the couples’ interactions during the seven tasks to their interactions at home on a 5-point scale ranging from 0 (not at all similar) to 4 (completely similar) as a measure of the ecological validity of the couples’ interactions.

Observational assessment

The videotapes were coded with the Living in Family Environments (LIFE; Hops, Davis, & Longoria, 1995) coding system. Details regarding this procedure and the results from the couple interactions have been published previously (Romano et al., 2009). For the purposes of this study, patients’ rate per minute of illness/pain behaviors was retained for analysis (M = 0.88, SD = .67). Illness/pain behaviors included verbal content (“That hurts.” “I’m so tired.”) and nonverbal signs, such as facial expressions of discomfort, sighs, or stiff movements.

Salivary cortisol

Participants were instructed to provide three saliva samples by using Salivettes® (Sarstedt, Inc., Newton, NC). A Salivette consists of a cotton swab that participants placed in their mouth and chewed for 30-60 seconds until saturated with saliva, and then expressed the swab into its plastic container without touching it. Samples were refrigerated and sent to the laboratory in five batches for processing. On arrival at the laboratory, the Salivettes were centrifuged at 4°C at approximately 1800 × g for 30 minutes. Salivary cortisol concentration was measured using ¹²⁵I cortisol kit from Pantex (Santa Monica, CA), modified for use with saliva samples as described previously (Seeman et al., 2001), for two batches of samples. For the remaining three batches the Pantex kits were no longer available; the samples were assayed using the Cortisol ImmuChem Coated Tube ¹²⁵I RIA Kit from MP Biomedicals (Orangeburg, NY), modified for use with saliva samples. The commercial protocol was modified by increasing the sample volume to 100 μl and adding 25 μl of human steroid-free serum to each sample to bring protein concentrations to levels equivalent to those in the standards. The detection limit of the assay is 0.4 ng/ml. The intra- and inter-assay coefficients of variation were 7.7% and 9.1%, respectively. In order to determine the agreement between the two assay systems, seven samples covering a broad concentration range were assayed with both systems. The inter-assay coefficient of variation between the two assays was 11.5%, comparable to the inter-assay coefficients of variation found for each method above. There were two missing data points: the third sample from two patients did not contain enough saliva to be assayed. To normalize the distribution of the cortisol data, natural logarithm transformations were applied prior to analysis (Keene, 1995); the exponential function, which is the inverse of the natural logarithm transformation, was applied to return predicted values to the original ng/ml measurement scale to facilitate interpretation. The elapsed time between the first and second saliva sample collections averaged 1 hour and 47 minutes (SD = 19 minutes), and the elapsed time between the second and third collections averaged 24 minutes (SD = 9 minutes). On average, the first saliva sample was collected about noon (1237h, range = 0905h – 1725h), within couple samples were collected at the same time. Patients’ and SOs’ average cortisol concentrations compare favorably to previously published samples’ values from the same laboratory during midday (Li et al., 2006). Average cortisol values across time and slopes over time (by fitting a regression line) were calculated for each person. A series of t-tests compared cortisol slope and average level by the use of hormones (yes/no) or antidepressants (yes/no) and gender (female/male). None of these tests achieved statistical significance so these variables were not included as covariates in subsequent analyses.

Data analysis

Descriptive statistics were used to characterize the perceived representativeness of the experimental activities vis-à-vis the participants’ usual home interactions, and the level of perceived exertion to engage in the household activities. The approach to examining cortisol concentration was based on a three-level model: cortisol concentration nested within time (i.e., cortisol was sampled three times within each couple), repeated assessments nested within person, and persons nested within couple (patient with CFS versus SO). IBM SPSS version 19 MIXED was used with fixed effects for the time of day of the salivary collection (based on a 24 hour clock), type of person within couple (patient with CFS versus SO), and the time of day by type of person within couple interaction; a repeated measure of the cortisol sample (first, second, or third); and random effects for individuals at baseline and change in cortisol over time. The second analysis was a two-level model: cortisol concentration of the SO nested within time (sampled three times), repeated assessments nested within person, with time-varying covariate of the patients’ cortisol concentration, and fixed covariates of the level of relationship satisfaction based on patients’ reports (low versus high) and the patient illness/pain behavior per minute.

Results

Ecological validity

Patients and SOs rated their interactions in the laboratory as similar to their interactions at home (M = 2.76, SD = 0.83 and M = 2.48, SD = 1.00, respectively; 4 = “completely similar”). Across the seven activities, patients self-rated RPEs averaged 11.27 (SD = 1.96); these average values correspond to describing their exertion as “fairly light to somewhat hard.” SOs’ self-rated RPEs averaged 8.75 (SD = 1.64); these values correspond to describing their exertion as between “very, very light” and “very light.” Given the patients’ fatigue, it is not surprising that they perceived these household activities as demanding modest effort whereas their SOs felt they had exerted themselves very lightly.

Differences in salivary cortisol

Table 1 shows the average cortisol concentration for patients with CF and their SOs at each of the three collections. The fit of the multilevel model with the addition of relationship satisfaction was better than the fit of the model without: BICs = 289.46 and 301.16, respectively, for a difference of 11.7. Raftery (1996) characterized that a BIC difference of > 10 as “very strong” evidence for a preferred model, therefore, the results of the model including relationship satisfaction are presented here. The effects for type of person within couple and the interaction of the type of person within couple and time (repeated cortisol samples) were statistically significant (F(1, 292.30) = 5.78, p < .05; F(1, 319.57) = 5.44, p < .05, respectively). The effect for the time of day of the samples and relationship satisfaction were not statistically significant (F(1, 315.58) = 0.02, ns, and F(1, 259.48) = 0.89, ns, respectively). Therefore, the differences in patients’ and SOs’ slopes over time (cortisol sampled three times over the course of the experimental tasks) are robust to the variability in the time of day that different couples participated in the study. Consistent with the previously observed flattening of patients’ cortisol concentrations, SOs’ cortisol concentrations decreased more steeply over time than patients’; this interaction is graphically depicted in Figure 1. Cortisol concentrations based on predicted values from the multilevel model are shown in Table 1. SOs’ predicted cortisol concentrations decreased over twice as much as did patients’ predicted cortisol concentrations.

Table 1.

Mean Cortisol Concentrations ng/ml by Person and Time as Measured and Predicted based on the Multi-level Model

	Sample 1		Sample 2		Sample 3
	Actual	Predicted	Actual	Predicted	Actual	Predicted
Patients with CF	2.09 (0.97)	1.85 (0.46)	1.68 (0.77)	1.60 (0.44)	1.48 (0.65)	1.55 (0.44)
Significant Others	2.70 (1.70)	2.34 (1.09)	1.90 (1.60)	1.75 (0.89)	1.75 (1.98)	1.63 (0.81)

Note: standard deviations are in parentheses.

Figure 1. Predicted Cortisol Concentrations for Patients with CF and their Significant Others over Time.

Predicted Cortisol Concentrations for Patients with CF and their Significant Others over Time. The inverse of the natural logarithm transformation, the exponential function, was applied to return predicted natural logarithm values to the original ng/ml measurement scale to facilitate interpretation. Each line represents an individual's cortisol concentration over time.

Prediction of cortisol by couple variables

Over time, SO cortisol concentrations were significantly associated with the time-varying values of the patients’ cortisol concentrations (F(1, 71.30) = 4.11, p < .05). In addition, the level of relationship satisfaction (F(1, 73.93) = 8.71, p < .05) and the rate per minute of patient illness/pain behavior (F(1, 73.96) = 5.29, p < .05) were also related to SO cortisol concentrations; the time of the day of the cortisol collection was not significantly associated with SO cortisol concentration (F(1, 73.85) = 3.90, ns). Greater relationship satisfaction (t (73.93) = 2.95) and higher rates of patient illness/pain behavior were related to greater levels of SO cortisol (t(73.96) = 2.30). Figure 2 shows the SOs’ cortisol concentrations based on predicted values from the multilevel model plotted against the patients’ rate of illness/pain behavior per minute, by level of relationship satisfaction: less satisfied couples are represented by diamonds and more satisfied couples are represented by circles. A visual inspection of the graph reveals that SOs’ in more satisfied relationships have higher levels of cortisol than SOs in less satisfied relationships at comparable rates of illness/pain behavior per minute by their partners. Stated otherwise, the average cortisol concentrations of SOs in more satisfied relationships were nearly 1.5 times greater than SOs in less satisfied relationships [M = 2.18 ng/ml (SD = 0.62) and M = 1.55 ng/ml (SD = 0.57), respectively] whereas patients in more satisfied relationships averaged about half the rate of illness/pain behavior per minute compared to patients in less satisfied relationships [M = 0.61 ng/ml (SD = 0.40) and M = 1.05 ng/ml (SD = 0.78), respectively].

Figure 2. Averaged Significant Others’ Predicted Cortisol Concentrations By Patients’ Rate of Pain/Illness Behavior Per Minute, Grouped by Level of Relationship Satisfaction.

Averaged Significant Others’ Predicted Cortisol Concentrations By Patients’ Rate of Pain/Illness Behavior Per Minute, Grouped by Level of Relationship Satisfaction. The exponential function was applied to return predicted natural logarithm values to the original ng/ml measurement scale to facilitate interpretation. Diamonds represent couples with lower levels of relationship satisfaction (◆ = ADAS ≤ 12); the solid regression line fitted to their values was R² = 0.75. Circles represent couples with higher levels of relationship satisfaction (○ = ADAS > 12); the dashed line fitted to their values was R² = 0.49.

When patients’ cortisol concentrations were used as the dependent variable, SO cortisol concentrations were associated with patient cortisol concentrations as in the previous analysis (F(1, 65.73) = 5.70, p < .05), but neither of the relationship variables nor the time of the day of the cortisol sample were significantly associated with patients’ cortisol concentrations (relationship satisfaction: F(1, 67.79) = 0.10, ns; rate of patient illness/pain behavior: F(1, 66.11) = 0.40, ns; the time of the day of the cortisol sample: F(1, 67.56) = 0.43, ns).

Discussion

This study examined 31 couples in which one person had chronic fatigue and the changes in cortisol concentration as they interacted in a series of household activities. The interactions appeared to have ecological validity on the basis of participants rating their activities during the study as similar to their activities at home. Furthermore, patients with CF perceived the household activities as requiring more effort than did their healthy partners. Consistent with our hypothesis, patients with CF had flatter slopes over time compared to their SOs. These results are consistent with the meta-analytic reviews of previous case-control studies that found flatter diurnal cortisol slopes among patients with CF compared to controls (Papadopoulos & Cleare, 2012; Powell et al., 2013).

Furthermore, this study uniquely examined changes in cortisol concentration across multiple time points within couples as they engaged in the same activities. Previous studies of cortisol in patients with CF have not used such closely related and yoked controls and methodology as in the present study. Whereas meta-analyses of previous studies have not found convincing evidence of lower average cortisol concentrations in persons with CF (Papadopoulos & Cleare, 2012; Powell et al., 2013), in the present study patients with CF had, overall, significantly lower cortisol concentrations than their SOs. These results provide some evidence of HPA axis dysfunction in CF.

Despite the flattened slopes of the patients’ cortisol over time, patients’ and SOs’ levels of cortisol were related to each other, akin to a previous report of couples’ synchrony in cortisol levels across time among healthy couples (Liu et al., 2013). In addition, greater SO cortisol levels were associated with satisfied relationships and with more frequent illness/pain behaviors by the patients during the household tasks. Figure 2 suggests that SOs in satisfied relationships may have been more physiologically sensitive to their partners’ illness behaviors whereas among less satisfied couples, it was only at much higher rates of patient illness behaviors that SOs’ cortisol concentrations approached those of the more satisfied SOs. Patients’ cortisol levels were not associated with relationship variables, perhaps due to their lack of variability over time and the dampening effects of CF on HPA activity.

There were several potential limitations to this study. These results from a sample of partnered tertiary care patients who were willing to interact in the laboratory may have limited generalizability to all chronically fatigued persons. There are other parameters that can influence cortisol secretion and it would have been ideal to experimentally control for more of these variables, although it was not possible to do so. The first such variable affecting cortisol is the time of day: cortisol gradually decreases throughout the day (Adam & Kumari, 2009). Multilevel models are able to deal robustly with time variation, but the logistics of scheduling couples precluded being able to test everyone at the same time of day. Within-couple saliva samples were taken at the same time, although different couples were tested across a wide range of times, which provided evidence for lower cortisol among patients over a considerable period of the day. The second variable is menstrual status and phase. Only two of the female patients still menstruated regularly, precluding the investigation of differences in menstrual status and phase within our sample. Reassuringly, studies have found no differences in salivary cortisol concentrations sampled during follicular versus luteal phases (Groschl, Rauh, Schmid, & Dorr, 2001; Kudielka & Kirschbaum, 2003; McCormick & Teillon, 2001). Thirdly, we did not find differences in cortisol concentration by gender or the use of certain medications, and ideally they would be experimentally controlled. Finally, participants were asked to refrain from eating or drinking at least 30 minutes before the start of their research appointment, but it is possible that some participants had ingested substances previously that might have affected cortisol concentrations.

Finally, this study represented only a small snapshot of the interactions of patients with CF and their SOs; patients had been ill an average of a dozen years, and lived with their SOs for 14 years on average. There has been interest in identifying subtypes of CF, in part because its heterogeneity can decrease the likelihood of identifying effective treatment and management strategies. A number of potential subtypes and subtyping dimensions have been suggested, but subtyping on the basis of characteristics of patients’ intimate relationships has not been among the suggestions (Jason et al., 2005) and could be considered. The interesting results obtained during approximately three hours in our laboratory suggest that studies that focus on the interaction of patients with their SOs are worthy of longer term research to add to our understanding of the complexity and chronicity of CF.