Salivary cortisol profiles in patients remitted from recurrent depression: One-year follow-up of a mindfulness-based cognitive therapy trial

Journal of Psychiatric Research 46 (2012) 80e86

Contents lists available at SciVerse ScienceDirect

Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires

Salivary cortisol profiles in patients remitted from recurrent depression: One-year follow-up of a mindfulness-based cognitive therapy trial Marianne Gex-Fabrya, *, Françoise Jermanna, Markus Kosela, Michel F. Rossierb, c, Martial Van der Lindend, Gilles Bertschye, Guido Bondolfia, Jean-Michel Aubrya a

Department of Mental Health and Psychiatry, Geneva University Hospitals, 2 chemin du Petit-Bel-Air, CH-1225 Chêne-Bourg, Geneva, Switzerland Department of Genetics and Laboratory Medicine, Geneva University Hospitals, Geneva, Switzerland Department of Internal Medicine, Geneva University Hospitals, Geneva, Switzerland d Cognitive Psychopathology and Neuropsychology Unit, Faculty of Psychology and Educational Sciences, University of Geneva, Geneva, Switzerland e INSERM Unit 666 & University of Strasbourg, Department of Psychiatry & Mental Health, University Hospital of Strasbourg, Strasbourg, France b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 May 2011 Received in revised form 2 September 2011 Accepted 21 September 2011

Few studies have examined changes of diurnal cortisol profiles prospectively, in relation to nonpharmacological interventions such as mindfulness-based cognitive therapy (MBCT). Fifty-six patients remitted from recurrent depression (
3 episodes) were included in an 8-week randomized controlled trial comparing MBCT plus treatment as usual (TAU) with TAU for depression relapse prophylaxis. Saliva samples (0, 15, 30, 45, 60 min post-awakening, 3 PM, 8 PM) were collected on six occasions (pre- and post-intervention, 3-, 6-, 9-, 12-month follow-up). Cortisol awakening response (CAR), average day exposure (AUCday) and diurnal slope were analyzed with mixed effects models (248 profiles, 1e6 per patient). MBCT (n ¼ 28) and TAU groups (n ¼ 28) did not significantly differ with respect to baseline variables. Intra-individual variability exceeded inter-individual variability for the CAR (62.2% vs. 32.5%), AUCday (30.9% vs. 23.6%) and diurnal slope (51.0% vs. 34.2%). No time, group and time by group effect was observed for the CAR and diurnal slope. A significant time effect (p ¼ 0.003) was detected for AUCday, which was explained by seasonal variations (p ¼ 0.012). Later wake-up was associated with lower CAR (11.7% per 1-hour later awakening, p < 0.001) and lower AUCday (4.5%, p ¼ 0.014). Longer depression history was associated with dampened CAR (15.2% per 10-year longer illness, p ¼ 0.003) and lower AUCday (8.8%, p ¼ 0.011). Unchanged cortisol secretion patterns following participation in MBCT should be interpreted with regard to large unexplained variability, similar relapse rates in both groups and study limitations. Further research is needed to address the scar hypothesis of diminished HPA activity with a longer, chronic course of depression. Ó 2011 Elsevier Ltd. All rights reserved.

Keywords: Cortisol awakening response Diurnal cortisol profile Depression Mindfulness-based cognitive therapy

1. Introduction Hyperactivity of the hypothalamic-pituitary-adrenal axis (HPA) in depression has been challenged in recent reviews that pointed at modest effects and large heterogeneity between studies (Knorr et al., 2010; Stetler and Miller, 2011). Debate has been ongoing about whether HPA dysfunction should be considered as a vulnerability factor for depression, as a correlate of depressive state or as a scar marker of the chronic course of the disorder. In recent years, focus has shifted from challenge

* Corresponding author. Tel.: þ41 22 305 57 93; fax: þ41 22 305 57 99. E-mail address: [email protected] (M. Gex-Fabry). 0022-3956/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jpsychires.2011.09.011

conditions such as the dexamethasone/corticotropin releasing hormone test (DEX/CRH test) to less invasive basal saliva cortisol measurements, with special emphasis on the cortisol awakening response (CAR), i.e. a concentration peak observed within the first 15e45 min after awakening. On the one hand, increased CAR in unaffected subjects at familial risk for depression lent support to the hypothesis of cortisol hypersecretion as a possible endophenotype for depression (Mannie et al., 2007; Vreeburg et al., 2010). Furthermore, elevated CAR in adolescents was observed to prospectively predict onset of depression during follow-up (Adam et al., 2010). On the other hand, support for the state marker hypothesis arose from normalization of the HPA axis during antidepressant treatment, with larger effect associated with larger improvement of depression and anxiety severity (Lenze et al., 2011; McKay and Zakzanis, 2010). There is strong

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

evidence that antidepressants modulate the glucocorticoid receptor (GR) through different mechanisms and that restoring GR function might play a role in their therapeutic action (Anacker et al., 2011). Studies on the relationship between cortisol profiles and characteristics of depression have been controversial. A large cohort study first showed no association between the CAR and depression severity, chronicity and symptom profile, but significantly higher CAR among patients with comorbid anxiety disorders (Vreeburg et al., 2009a). The same group later focused on a dimensional model of depression and anxiety and suggested that a non-linear relationship between CAR and symptom dimensions might explain earlier negative findings (Wardenaar et al., 2011). Finally, a history of stressful life experiences might lead to persistent changes in cortisol profiles, as observed for early life events (Gerritsen et al., 2010), low socioeconomic trajectories (Gustafsson et al., 2010), and job and general life stress (Chida and Steptoe, 2009). Such long-term effects might be relevant to persistent CAR elevation among patients in remission from depression (Aubry et al., 2010; Bhagwagar et al., 2003). A recent meta-analysis pointed at the important overlap of morning salivary cortisol levels between depressed patients and controls, while a small but statistically significant difference was confirmed (Knorr et al., 2010). Several large epidemiological surveys focused on inter-individual variability and its determinants (Kumari et al., 2010; Lederbogen et al., 2010; Vreeburg et al., 2009b). Numerous variables have been documented to influence cortisol secretion, including sociodemographic variables (e.g. sex and age), lifestyle parameters (smoking, physical activity, sleep duration), health factors (cardiovascular disease) and sampling conditions (working day vs. weekend, season). Information about intra-individual variability is scarcer. According to a large survey with saliva samples collected on four consecutive days, up to 78% of the total CAR variation might be attributed to day-to-day variability (Almeida et al., 2009). A study in older adults reported that the CAR might be sensitive to short-term effects, with prior-day feelings of loneliness, sadness, threat and lack of control associated with higher CAR the next day (Adam et al., 2006). Thus, lack of adjustment for relevant between- and within-subject variability factors might have accounted for low signal-to-noise ratio and inconsistent results in some earlier studies. Not surprisingly, cortisol has been evaluated as a marker for improvement with various types of non-pharmacological treatment, such as interventions aimed at reducing stress. A recent article reviewed accumulating evidence that cortisol levels tend to decrease after a Mindfulness-Based Stress Reduction program (MBSR) (Matousek et al., 2010). Salivary cortisol was also included among possible predictors of brief cognitive therapy effectiveness in preventing relapse in recurrent depression (Bockting et al., 2006). The present investigation was part of a study designed to confirm the efficacy of Mindfulness-Based Cognitive Therapy (MBCT) (Segal et al., 2002) compared with Treatment As Usual (TAU) in reducing relapse risk in patients remitted from at least 3 episodes of depression. We previously reported that participation in the MBCT program was associated with delayed relapse but unchanged relapse rate over the 14-month observation period (Bondolfi et al., 2010). A second aim of the study was to examine possible changes of diurnal salivary cortisol profiles, which were measured on 6 occasions over the follow-up period. Objectives of the present report were 3-fold: quantify within- and betweenpatient variability of different indices related to cortisol profiles; test for a possible change over time that might be specific to participants in the MBCT program; and examine the role of variables previously documented or as yet undocumented as relevant variability factors.

81

2. Materials and methods 2.1. Patients Participants in the MBCT trial were recruited through media announcements and mailings to psychiatrists and general practitioners in the French speaking part of Switzerland. Inclusion criteria have been described earlier (Bondolfi et al., 2010) and can be summarized as follows: history of recurrent major depressive disorder according to DSM-IV (American Psychiatric Association,1994); at least three past depressive episodes; remission since at least 3 months, with a score 13 on the Montgomery-Asberg Depression Rating Scale (MADRS) (Montgomery and Asberg, 1979); history of treatment by a recognized antidepressant drug, but being off medication for at least 3 months. The following non-inclusion criteria were considered: history of schizophrenia or schizoaffective disorder; current substance abuse, eating disorder or obsessive compulsive disorder; organic mental disorder, pervasive developmental disorder, borderline personality disorder; dysthymia with onset before age 20; >4 sessions of cognitive behavioral therapy ever; current psychotherapy or counseling; current practice of meditation or yoga. The study was designed as a replication of two previous MBCT trials, with identical inclusion criteria (Ma and Teasdale, 2004; Teasdale et al., 2000). The rationale for including patients with 3 or more episodes was based on previous findings that MBCT allowed reducing relapse risk specifically in the ones with
3 episodes. The study flow chart has been provided in our earlier publication (Bondolfi et al., 2010). Briefly, of 142 patients who completed a selection interview with an experienced clinical psychologist, 71 eligible participants entered a 3-month run-in period. Phone contact was maintained on a monthly basis to ascertain that remission was stable, in the absence of antidepressant medication. After the run-in phase, an enrolment interview took place to check that inclusion criteria were still met. Eleven patients were excluded (7 had relapsed; 2 refused to participate; 2 could not be contacted) and 60 were randomized to MBCT plus TAU or TAU. The study protocol received approval from the ethics committee of the Geneva University Hospitals and each participant provided written informed consent before being enrolled. 2.2. Study design and intervention Detailed information about the MBCT program and study procedures has been provided in our earlier publication (Bondolfi et al., 2010). Briefly, MBCT consists of 8 weekly sessions of a group intervention that integrates components of the MBSR program with elements of Cognitive Behavioral Therapy (CBT) to prevent depressive relapse (Segal et al., 2002). Participants were randomly assigned to TAU (unrestricted access to any type of treatment or help) or MBCT plus TAU. All therapists had participated to a training program, had experience running MBCT groups and had ongoing personal mindfulness practice. Four MBCT booster-sessions were provided at 3-month intervals during the 1-year follow-up. Based on the main efficacy hypothesis of MBCT allowing to reduce relapse risk, sample size was estimated at 28 subjects per group (Bondolfi et al., 2010). Post-hoc power analysis indicated that for 28 MBCT participants, power would be 77% to detect a CAR normalization toward values observed in controls (effect size 0.53 in log-scale, based on data in Aubry et al., 2010; paired Student t-test, two-tailed, significance level at 5%). 2.3. Instruments Patients were assessed at baseline (T1), at the end of the MBCT program (T2, month 2) and at 3-month intervals during a 1-year

82

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

follow-up (T3eT6; months 5, 8, 11 and 14). Depression severity was assessed with the MADRS and the self-rated Beck Depression Inventory II (BDI-II) (Beck et al., 1996). Relapse or recurrence were assessed with the depression section of the Structured Clinical Interview for DSM-IV (First et al., 2002) by two raters blind to group assignment and further confirmed by a senior psychiatrist. 2.4. Cortisol analysis Participants were instructed to collect 7 saliva samples per day at home (awakening, 15, 30, 45 and 60 min post-awakening, 3 PM and 8 PM) on 6 occasions (T1eT6). In order to avoid sample contamination, they were told not to eat and brush their teeth during the first hour after waking. Participants reported actual sampling times on a specific form, together with bed and waking time. As a delay exceeding 15 min between wake-up and first saliva sample significantly affects the CAR (Okun et al., 2010), we excluded 6 cortisol profiles corresponding to delays >15 min. Similarly, we excluded 9 profiles with delays >15 min between actual and recommended sampling times at 15, 30, 45 and 60 min post-awakening. Saliva samples were collected using Salivettes (Sarstedt, Nümbrecht, Germany) and kept in the refrigerator until transport to the laboratory by priority mail (delivery on the next day). Prior to the study, we found that cortisol sampling was stable for at least one week at 4  C (data not shown). Upon arrival at the laboratory, samples were immediately centrifuged (10 min at 3000 rpm), the swab was discarded and saliva was frozen at 20  C until cortisol assay. Upon thawing, samples were centrifuged again and 100 ml saliva aliquots were collected. Cortisol concentration was measured directly (without extraction) using a modified commercial solidphase radioimmunoassay (Coat a CountÒ, Diagnostic Product Corporation, Los Angeles, CA). Calibrators were adapted to the lower salivary cortisol concentrations, as compared to serum, by successive dilutions in water. Under these conditions, the analytical detection limit was 0.7 nmol/l and inter-assay imprecision at 7.5 nmol/l was below 10% (coefficient of variation). The whole analytical procedure was accredited according to ISO15189 laboratory norms. For data analysis, concentrations below detection limit were arbitrarily set to half this value (0.35 nmol/l). 2.5. Data analysis In keeping with recommendations (Adam and Kumari, 2009), we considered three different indices of diurnal cortisol profiles. The CAR provided an estimate of exposure to free cortisol in response to awakening. It was calculated as the area under the curve (AUC) above the minimum concentration in the first hour, according to the trapezoidal rule and taking into account actual sampling times. Based on previous data (Aubry et al., 2010), we showed that this CAR measure is largely equivalent to the one obtained from the difference between maximum and minimum cortisol values in the same 1-hour interval (Spearman’s r ¼ 0.97, p < 0.001; unpublished). Average exposure across the day was estimated through the AUC calculated using the trapezoidal rule and all available samples (AUCday). To correct for differences in the sampling interval, AUCday was normalized to the median 13 h sample collection period. The diurnal slope assessed cortisol decline over the day. It was estimated as the difference between wake-up and last evening values divided by the time interval between these samples. All CAR estimates (n ¼ 248) included 5 saliva samples, while AUCday and diurnal slope (n ¼ 238) were considered as missing if the sample at 8 PM was missing. Comparison between MBCT and TAU groups proceeded with the Fisher’s exact test for categorical variables and the ManneWhitney U-test for continuous variables. Pre- (T1) and post- (T2)

intervention cortisol indices were compared with Wilcoxon signed ranks tests. After log-transformation because of positively skewed distributions (i.e. log10 of CAR and AUCday; log10 of diurnal slope þ 1 in order to avoid negative values), the whole data set (T1eT6) was then analyzed with 3 series of mixed effects models (also known as multi-level or hierarchical models). Firstly, components of variance models including participant as a random factor were used to estimate the intraclass correlation coefficient (ICC) and inter- and intra-individual variability (variance estimates in the log-scale were back-transformed to coefficients of variation in the original scale). Secondly, group (MBCT vs. TAU), time (T1eT6) and time by group interaction were entered as fixed factors. Similar models were estimated for group, season and season by group interaction. Goodness of fit of models including time and season were compared using the Akaike information criterion (smaller-is-better form). Thirdly, screening for additional predictors proceeded by entering the following variables into the model: gender, comorbid anxiety disorder (constant person-level factors), weekend vs. weekday, relapse, antidepressant medication during the study (time-dependent, day-level factors), age, years of education, illness duration, number of previous episodes (constant covariates), and wake-up time, sleep duration, MADRS and BDI scores (timedependent covariates). The final models included all variables identified as significant in the screening process. Statistical significance was set at 0.05 (two-tailed tests). Bonferroni adjustment for multiple comparisons was used in post-hoc tests. Data analysis was performed using SPSS 17 (SPSS Inc., Chicago, IL). 3. Results 3.1. Sample characteristics Of 60 patients randomized to MBCT plus TAU or TAU, 28 participants in each group provided at least one valid cortisol profile (248 CAR data; n ¼ 48, 47, 47, 39, 38 and 29 at T1, T2, T3, T4, T5 and T6, respectively). No difference between groups was observed for sociodemographic variables and baseline assessments (Table 1). In the whole sample, median duration of depressive illness was 17 years (range 4e50) and number of previous depressive episodes ranged from 3 to 14. Comorbid anxiety disorder was present in 35.7% of patients. Relapse rates over the 14-month observation period were 32.1% and 35.7% among MBCT plus TAU and TAU participants, respectively. The proportion of patients who started antidepressant medication during the study was 68.4% among patients who relapsed and 16.2% among those who did not. Cortisol indices at study inclusion (T1) did not significantly differ between groups (ManneWhitney U-tests, p ¼ 0.27 for CAR, p ¼ 0.21 for AUCday, p ¼ 0.42 for diurnal slope). Pre- and postintervention (T2) indices are described in Table 2 for patients with valid profiles at both times. No significant change was observed, whether among MBCT plus TAU or TAU participants. 3.2. Intra- vs. inter-individual variability of cortisol indices Components of variance models for the whole data set (T1eT6; Table 3, model 1) indicated that intra-individual, inter-occasion variability exceeded inter-individual variability for the CAR (62.2% vs. 32.5%, ICC ¼ 0.21) and diurnal slope (51.0% vs. 34.2%, ICC ¼ 0.31), whereas both intra- and inter-individual variability were lower for AUCday (30.9% vs. 23.6%, ICC ¼ 0.37). 3.3. Time and time by group effects The three cortisol indices were first investigated for a possible change over time in unadjusted mixed models that considered

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

83

Table 1 Patient characteristics in MBCT þ TAU and TAU groups.

Female Age (years) Education (years) Duration of depressive illness (years) Number of previous episodes Baseline MADRS Baseline BDI Comorbid anxiety disorder Relapsed during follow-up Antidepressant medication during follow-up

(n, %) (median, (median, (median, (median, (median, (median, (n, %) (n, %) (n, %)

range) range) range) range) range) range)

MBCT þ TAU group (n ¼ 28)

TAU group (n ¼ 28)

20 46 16 16 4 4 8 12 9 10

20 49 15 17 4 3 8 8 10 9

71.4% 27e63 10e21 6e43 3e14 0e13 0e32 42.9% 32.1% 35.7%

p-valuea 71.4% 24e66 10e28 4e50 3e8 0e13 0e38 28.6% 35.7% 32.1%

1 0.61 0.80 0.93 0.67 0.81 0.85 0.40 1 1

Abbreviations. MADRS, Montgomery & Åsberg Depression Rating Scale; BDI, Beck Depression Inventory; MBCT, mindfulness-based cognitive therapy; TAU, treatment as usual. a Fisher’s exact test for categorical variables; ManneWhitney U-test for other variables.

time, group and time by group as fixed factors and subject as a random factor (Table 3, model 2). No significant effect was observed for the CAR and diurnal slope. A significant time effect was detected for AUCday, without any group effect or time by group interaction. Post-hoc tests revealed that AUCday at 12-month postMBCT or TAU (T6) was significantly higher than measurements at T1, T2, T3 and T4 (p < 0.05 after Bonferroni adjustment for multiple comparisons). The maximum difference was observed between T4 and T6, i.e. assessments at 6-month and 12-month postintervention (p ¼ 0.001). Because T4 assessments were mostly performed in the fall (83.8%) and T6 evaluation mostly occurred in the spring (88.5%), we examined whether the time effect was attributable to a seasonal effect by considering an alternative model including season, group and season by group interaction (not shown). Season significantly contributed to the AUCday model (p ¼ 0.012), with no significant group effect or season by group interaction. Post-hoc tests revealed that AUCday was significantly higher in the spring than in the fall (þ21.6%, p ¼ 0.012). The model including season as a predictor provided a better fit to the data than the one including assessment time according to the Akaike criterion (177 vs. 169, n ¼ 236), suggesting that seasonal variation might be a confounding factor in the observed time effect. 3.4. Variability determinants

4. Discussion

Screening for additional variability factors indicated that sociodemographic variables (sex, age, years of education) did not significantly contribute to the variability of cortisol indices. Among variables associated with sampling conditions, later wake-up (p ¼ 0.001) and longer sleep duration (p ¼ 0.009) were significantly associated with lower CAR, whereas weekday vs. weekend Table 2 Pre- and post-intervention cortisol indices in MBCT þ TAU and TAU groups.

n

differences did not reach statistical significance. Later wake-up time also predicted lower AUCday (p ¼ 0.019), while no effect was observed for diurnal slope. Depression severity (MADRS, BDI), comorbid anxiety disorder, occurrence of a depressive relapse and starting antidepressant treatment during the study were not significantly associated with cortisol indices. However, a significant association was observed between a longer depression history and both lower CAR (p ¼ 0.006) and lower AUCday (p ¼ 0.014). No association was found with respect to number of previous depressive episodes. Final models describing CAR, AUCday and diurnal slope after adjustment for wake-up time and illness duration are provided in Table 3 (model 3). As in unadjusted models, no time, group or time by group interaction was observed for the CAR and diurnal slope, whereas the time effect remained significant for AUCday. One-hour later awakening was associated with an 11.7% lower CAR (p < 0.001) and 4.5% lower AUCday (p ¼ 0.014), whereas a 10-year longer history of depression was accompanied with a 15.2% lower CAR (p ¼ 0.003) and 8.8% lower AUCday (p ¼ 0.011). These two variables only explained a small fraction of variability, as indicated by minor changes of variability estimates between unadjusted and adjusted models (Table 3).

Pre-intervention (T1)

Post-intervention (T2)

median range

median range

p-valuea

MBCT þ TAU group CAR (h,nmol/l) 22 AUCday (h,nmol/l) 21 21 diurnal slope (h1,nmol/l)

4.6 99.9 0.71

0.7e15.9 39.9e245.2 0.05e2.13

4.3 97.4 0.67

1.2e20.1 40.6e261.2 0.00e1.80

TAU group CAR (h,nmol/l) 22 AUCday (h,nmol/l) 21 21 diurnal slope (h1,nmol/l)

5.5 96.5 0.66

0.9e13.1 32.6e200.6 0.10e1.64

5.9 92.1 0.54

1.0e18.8 0.63 55.5e172.0 0.81 0.11e1.37 0.19

0.73 0.97 0.36

Abbreviations : MBCT, mindfulness-based cognitive therapy; TAU, treatment as usual. a Wilcoxon signed ranks test.

Taking advantage of the prospective design of a randomized controlled trial aimed at assessing the effectiveness of MBCT to prevent depressive relapse, the present study showed that cortisol indices displayed larger variability within persons, upon repeated measures over 14 months, than between persons. This observation contrasts with early salivary cortisol studies, which generally concluded to moderate to high intra-individual stability (Pruessner et al., 1997). However, it is in agreement with more recent longitudinal studies showing that the highest part of variability might be attributed to within person sources (Almeida et al., 2009; Mommersteeg et al., 2006). Intra-individual variability might also be larger in patients with depressive disorders than in healthy controls. Particularly erratic patterns in patients with more severe or recurrent episodes led to hypothesize that high variability per se might represent a core feature of HPA dysregulation in depression (Peeters et al., 2004). Search for differences between groups might thus be hampered by a low signal-to-noise ratio, while additional focus on within-person time-varying variables might be needed. This is particularly true for the CAR, which is considered to reflect processes specific to the sleepewake transition (Wilhelm et al., 2007). A recent study highlighted the value of proximal measures in explaining variability, by showing increased CAR and total cortisol output associated with state prior-day rumination and worry, but not trait characteristics (Zoccola et al., 2011). Research

84

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

Table 3 Inter- and intra-individual variability of CAR, AUCday and diurnal slope in 56 patients remitted from depression (1e6 cortisol profiles per patient).

Parameter estimatea 1 2

3

Inter-individual variability, CV Intra-individual variability, CV Group (MBCT vs. TAU) Time T1, pre-MBCT/TAU T2, post-MBCT/TAU T3, 3-month follow-up T4, 6-month follow-up T5, 9-month follow-up T6, 12-month follow-up Time x Group Inter-individual variability, CV Intra-individual variability, CV Group (MBCT vs. TAU) Time T1, pre-MBCT/TAU T2, post-MBCT/TAU T3, 3-month follow-up T4, 6-month follow-up T5, 9-month follow-up T6, 12-month follow-up Time x Group Wake-up time (1-hour later) Duration of depressive illness (10-years longer) Inter-individual variability, CV Intra-individual variability, CV

(%) (%)

32.5 62.2 0.99

p-value

0.65 0.54

1.00 1.03 1.05 0.92 1.13 1.23

Parameter estimatea 23.6 30.9 1.07

32.0 63.1 0.91

0.41 0.69

1.00 1.07 1.05 0.94 1.14 1.17 0.88 0.85 (%) (%)

p-value

0.58 0.003

1.00 0.99 0.99 0.91 1.01 1.27 0.98

(%) (%)

Diurnal slope (h1,nmol/l) n ¼ 238

AUCday (h,nmol/l) n ¼ 238

CAR (h,nmol/l) n ¼ 248

Model

29.8 60.9

34.2 51.0 0.98

0.82 24.2 30.0 1.04

0.95 0.91 23.8 29.3

p-value

0.28 0.092

1.00 0.99 0.75 0.82 0.81 0.88

0.45 0.004

1.00 1.01 0.98 0.92 1.01 1.26 0.99 <0.001 0.003

Parameter estimatea

0.52 34.3 50.7 1.01

0.23 0.12

1.00 0.98 0.76 0.82 0.81 0.89 0.81 0.014 0.011

1.05 1.01

0.46 0.10 0.86

33.2 50.8

Abbreviations : MBCT, mindfulness-based cognitive therapy; TAU, treatment as usual. a Fixed effects provided as multiplying factors after back-transformation from mean differences in the log-scale; random effects provided as coefficients of variation (CV) obtained from variance estimates in the log-scale.

investigating cortisol indices in relation with specific mechanisms of change, such as enhancement of mindfulness and selfcompassion (Kuyken et al., 2010), would be of particular relevance to the understanding of MBCT action. Because baseline CAR was about 50% higher in patients remitted from depression than in controls (Aubry et al., 2010), our hypothesis was that cortisol patterns would normalize with MBCT. In contrast, the present study did not indicate any significant change in cortisol profiles over the 8-week MBCT program or 12-month follow-up, except for a seasonal effect for average exposure across the day. Such a negative finding deserves several comments. First, a type II error cannot be excluded, provided that sample size was relatively small and changes during MBCT might be subtle. Second, the study was not specifically designed to assess the impact of MBCT on cortisol profiles, so that no definite conclusion can be reached. Third, a possible relationship between unchanged cortisol profiles and similar relapse rates with MBCT plus TAU and TAU might be invoked. Indeed, although time to relapse was significantly delayed in the MBCT group, relapse rates did not significantly differ over the 14-month observation period (Bondolfi et al., 2010). Alternatively, cortisol indices might be related with some specific mechanism of MBCT action, as yet insufficiently investigated and understood. Whereas some studies failed to report changes, a recent literature survey found accumulating evidence for reduced cortisol levels following MBSR in various samples of healthy subjects and patients with somatic disorders (Matousek et al., 2010). It was stressed that many studies had methodological limitations, such as small sample size, uncontrolled design and lack of adjustment for confounding variables. The same group later reported a CAR increase from pre- to post-MBSR in breast cancer patients, which was accompanied by significant improvements in self-reported stress, depressive symptomatology, and medical symptoms (Matousek et al., 2011). This increase was postulated to reflect normalization from a blunted CAR at baseline. Effect of non-pharmacological treatment

on cortisol secretion patterns has been little investigated in depressed patients, unlike the impact of antidepressant treatment (McKay and Zakzanis, 2010). Among 21 depressed participants to a CBT intervention, no significant change of cortisol measures was observed between pre- and post-treatment (Taylor et al., 2009). As far as we know, no previous study specifically focused on patients in remission from several depressive episodes. Determinants of intra- and inter-individual variability of cortisol indices have been reviewed on several occasions (Fries et al., 2009; Hansen et al., 2008; Kudielka and Wust, 2010). The observed association between lower CAR and later awakening or longer sleep duration in remitted patients is in agreement with results from several large community studies (Kumari et al., 2009; Lederbogen et al., 2010; Vreeburg et al., 2009b). While data in depressed patients are scarce, one study reported no association and suggested that loss of regulatory control over morning cortisol response might be one facet of HPA axis dysregulation in depression (Stetler and Miller, 2005). Other relevant sleep-related parameters might include sleep quality (Lasikiewicz et al., 2008), waking in the dark or in the light (Thorn et al., 2004), and sampling on workdays or weekends (Thorn et al., 2006). They are obviously not independent from each other and from stress-related factors such as anticipation of the upcoming day (Fries et al., 2009). Literature has been inconsistent with respect to a possible seasonal effect on HPA activity. On the one hand, results in the present study are consistent with a study of month-to-month changes in healthy adults in Sweden. Highest cortisol concentrations were observed in the spring (February to April) and lowest levels in summer (July and August), without any seasonal difference for the CAR (Persson et al., 2008). On the other hand, they contrast with a large study in the Netherlands which reported higher morning and evening cortisol levels during months with less daylight, yet without any effect on the CAR (Vreeburg et al., 2009b). While effects of season and daylight remain insufficiently

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

understood, they deserve to be taken into account and adjusted for in prospective intervention studies. With focus placed on patients with at least three past depressive episodes, the present study revealed an as yet undocumented association between a longer, chronic course of depression and a decreased CAR and cortisol release over the day. As far as we know, the relationship between the course of depressive illness and cortisol patterns has been little investigated. Long-term longitudinal studies are lacking with respect to a possible adaptative change of the HPA activity with the chronic course of depression. A meta-analysis nevertheless suggested that time might be a critical element, with an initial activation of the HPA axis when a chronic stressor arises and a subsequent decrease as time passes (Miller et al., 2007). Hypocorticolism may occur after a prolonged period of hyperactivity of the HPA axis due to chronic stress, as observed in chronic fatigue syndrome, fibromyalgia and PTSD (Fries et al., 2005). Interestingly, hypoactive HPA axis has also been documented in patients with atypical depression (Antonijevic, 2006). Limitations of the present study need to be acknowledged. First, even though saliva samples were collected on several occasions over a 14-month period, it might have been of interest to collect samples on consecutive days as well, in order to distinguish day-today variability from longer-term effects of time. Indeed, a longitudinal study in burnout patients showed that variability across two consecutive weekdays was of the same order of magnitude as variability across three occasions, more than 6 months apart (Mommersteeg et al., 2006). Secondly, some degree of noncompliance with the sampling protocol cannot be excluded and might have inflated intra-individual variability, even though patients were reminded about the importance of complying with instructions and carefully reporting actual sampling times. Thirdly, several variables documented to influence cortisol profiles in earlier studies were not controlled for in the present study. These include variables at the patient-level, such as smoking status (Badrick et al., 2007), and variables at the day-level, such as priorday or same-day emotional experience (Adam et al., 2006). Fourthly, the present study, designed to compare MBCT and TAU with respect to relapse risk, provided limited statistical power to study the interplay between numerous variability factors, when compared with large studies in community samples. Finally, generalization of the present results deserves caution. Patients were in remission from recurrent depression and off antidepressant medication at study inclusion. We had no precise information about severity of past depressive episodes, most often treated in private practice, and reasons for discontinued antidepressant treatment (e.g. moderate severity, adverse events or patient preference). Thus, the study sample was most likely heterogeneous in this respect. In addition, participants had free access to any type of medical or alternative support during the study. In our earlier publication (Bondolfi et al., 2010), we postulated that high availability and easy access to mental health care in Switzerland may have played a role in the similar relapse rates among MBCT plus TAU and TAU participants. Possible help received from various sources during the study should similarly be kept in mind when considering unchanged cortisol profiles with MBCT. In summary, cortisol indices measured prospectively over 14 months displayed large within-person and between-person variability, that was only partly explained by factors such as awakening time, duration of depression history and seasonal effects. While the scar hypothesis of diminished HPA activity with a longer, chronic course of depression deserves further investigation, larger carefully designed studies, with better control of relevant variability factors, are awaited to determine the possible role of cortisol as a marker for improvement with MBCT and other non-pharmacological interventions.

85

Contributors Marianne Gex-Fabry performed the literature searches, statistical analysis and data interpretation; she wrote the manuscript. Françoise Jermann was involved in protocol writing, data collection and manuscript revision. Markus Kosel critically revised the article. Michel Rossier contributed to the protocol, supervised the laboratory procedures and participated to manuscript writing. Martial Van der Linden participated to the study design and revision of the article. Gilles Bertschy contributed to the protocol, patient inclusion, and manuscript revision. Guido Bondolfi contributed to the study protocol, supervised the MBCT clinical trial and revised the manuscript. Jean-Michel Aubry designed and supervised the cortisol study and substantially contributed to the article. All authors contributed significantly to and have approved the final manuscript. Role of the funding source The study was supported by a grant of the Swiss National Science Foundation (Grant N 3200BO-108432 to Guido Bondolfi, Gilles Bertschy, Jean-Michel Aubry and Martial Van der Linden). The funding source had no further role in the study design; in the collection, analysis and interpretation of data; in the writing of the report; nor in the decision to submit the article for publication. Conflict of interest All authors declare that they have no conflict of interest relevant to the present study. Acknowledgment We thank Christiane Gonzalez for technical help with saliva collection and Liliane Bockhorn for performing cortisol assays. References Adam EK, Doane LD, Zinbarg RE, Mineka S, Craske MG, Griffith JW. Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 2010;35:921e31. Adam EK, Hawkley LC, Kudielka BM, Cacioppo JT. Day-to-day dynamics of experienceecortisol associations in a population-based sample of older adults. Proceedings of the National Academy of Sciences USA 2006;103:17058e63. Adam EK, Kumari M. Assessing salivary cortisol in large-scale, epidemiological research. Psychoneuroendocrinology 2009;34:1423e36. Almeida DM, Piazza JR, Stawski RS. Interindividual differences and intraindividual variability in the cortisol awakening response: an examination of age and gender. Psychology and Aging 2009;24:819e27. American Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-IV). Washington, DC: American Psychiatric Association; 1994. Anacker C, Zunszain PA, Carvalho LA, Pariante CM. The glucocorticoid receptor: pivot of depression and of antidepressant treatment? Psychoneuroendocrinology 2011;36:415e25. Antonijevic IA. Depressive disorders e is it time to endorse different pathophysiologies? Psychoneuroendocrinology 2006;31:1e15. Aubry JM, Jermann F, Gex-Fabry M, Bockhorn L, Van der Linden M, Gervasoni N, et al. The cortisol awakening response in patients remitted from depression. Journal of Psychiatric Research 2010;44:1199e204. Badrick E, Kirschbaum C, Kumari M. The relationship between smoking status and cortisol secretion. Journal of Clinical Endocrinology and Metabolism 2007;92: 819e24. Beck AT, Steer RA, Brown GK. Manual for the Beck depression inventory-II. San Antonio, TX: Psychological Corporation; 1996. Bhagwagar Z, Hafizi S, Cowen PJ. Increase in concentration of waking salivary cortisol in recovered patients with depression. American Journal of Psychiatry 2003;160:1890e1. Bockting CL, Spinhoven P, Koeter MW, Wouters LF, Visser I, Schene AH. Differential predictors of response to preventive cognitive therapy in recurrent depression: a 2-year prospective study. Psychotherapy and Psychosomatics 2006;75: 229e36. Bondolfi G, Jermann F, Van der Linden M, Gex-Fabry M, Bizzini L, Weber Rouget B, et al. Depression relapse prophylaxis with mindfulness-based cognitive

86

M. Gex-Fabry et al. / Journal of Psychiatric Research 46 (2012) 80e86

therapy: replication and extension in the Swiss health care system. Journal of Affective Disorders 2010;122:224e31. Chida Y, Steptoe A. Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biological Psychology 2009;80:265e78. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical interview for DSM-IV-TR axis I disorders. New York: Biometrics Research, New York State Psychiatric Institute; 2002. Patient Edition (SCID-I/P). Fries E, Dettenborn L, Kirschbaum C. The cortisol awakening response (CAR): facts and future directions. International Journal of Psychophysiology 2009;72:67e73. Fries E, Hesse J, Hellhammer J, Hellhammer DH. A new view on hypocortisolism. Psychoneuroendocrinology 2005;30:1010e6. Gerritsen L, Geerlings MI, Beekman AT, Deeg DJ, Penninx BW, Comijs HC. Early and late life events and salivary cortisol in older persons. Psychological Medicine 2010;40:1569e78. Gustafsson PE, Janlert U, Theorell T, Hammarstrom A. Life-course socioeconomic trajectories and diurnal cortisol regulation in adulthood. Psychoneuroendocrinology 2010;35:613e23. Hansen AM, Garde AH, Persson R. Sources of biological and methodological variation in salivary cortisol and their impact on measurement among healthy adults: a review. Scandinavian Journal of Clinical and Laboratory Investigation 2008;68:448e58. Knorr U, Vinberg M, Kessing LV, Wetterslev J. Salivary cortisol in depressed patients versus control persons: a systematic review and meta-analysis. Psychoneuroendocrinology 2010;35:1275e86. Kudielka BM, Wust S. Human models in acute and chronic stress: assessing determinants of individual hypothalamus-pituitary-adrenal axis activity and reactivity. Stress 2010;13:1e14. Kumari M, Badrick E, Ferrie J, Perski A, Marmot M, Chandola T. Self-reported sleep duration and sleep disturbance are independently associated with cortisol secretion in the Whitehall II study. Journal of Clinical Endocrinology and Metabolism 2009;94:4801e9. Kumari M, Badrick E, Sacker A, Kirschbaum C, Marmot M, Chandola T. Identifying patterns in cortisol secretion in an older population. Findings from the Whitehall II study. Psychoneuroendocrinology 2010;35:1091e9. Kuyken W, Watkins E, Holden E, White K, Taylor RS, Byford S, et al. How does mindfulness-based cognitive therapy work? Behaviour Research and Therapy 2010;48:1105e12. Lasikiewicz N, Hendrickx H, Talbot D, Dye L. Exploration of basal diurnal salivary cortisol profiles in middle-aged adults: associations with sleep quality and metabolic parameters. Psychoneuroendocrinology 2008;33:143e51. Lederbogen F, Kuhner C, Kirschbaum C, Meisinger C, Lammich J, Holle R, et al. Salivary cortisol in a middle-aged community sample: results from 990 men and women of the KORA-F3 Augsburg study. European Journal of Endocrinology 2010;163:443e51. Lenze EJ, Mantella RC, Shi P, Goate AM, Nowotny P, Butters MA, et al. Elevated cortisol in older adults with generalized anxiety disorder is reduced by treatment: a placebo-controlled evaluation of escitalopram. American Journal of Geriatric Psychiatry 2011;19:482e90. Ma SH, Teasdale JD. Mindfulness-based cognitive therapy for depression: replication and exploration of differential relapse prevention effects. Journal of Consulting and Clinical Psychology 2004;72:31e40. Mannie ZN, Harmer CJ, Cowen PJ. Increased waking salivary cortisol levels in young people at familial risk of depression. American Journal of Psychiatry 2007;164: 617e21. Matousek RH, Dobkin PL, Pruessner J. Cortisol as a marker for improvement in mindfulness-based stress reduction. Complementary Therapies in Clinical Practice 2010;16:13e9. Matousek RH, Pruessner JC, Dobkin PL. Changes in the cortisol awakening response (CAR) following participation in mindfulness-based stress reduction in women who completed treatment for breast cancer. Complementary Therapies in Clinical Practice 2011;17:65e70.

McKay MS, Zakzanis KK. The impact of treatment on HPA axis activity in unipolar major depression. Journal of Psychiatric Research 2010;44:183e92. Miller GE, Chen E, Zhou ES. If it goes up, must it come down? Chronic stress and the hypothalamic-pituitary-adrenocortical axis in humans. Psychological Bulletin 2007;133:25e45. Mommersteeg PM, Heijnen CJ, Verbraak MJ, van Doornen LJ. A longitudinal study on cortisol and complaint reduction in burnout. Psychoneuroendocrinology 2006;31:793e804. Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. British Journal of Psychiatry 1979;134:382e9. Okun ML, Krafty RT, Buysse DJ, Monk TH, Reynolds CF, Begley A, et al. What constitutes too long of a delay? Determining the cortisol awakening response (CAR) using self-report and PSG-assessed wake time. Psychoneuroendocrinology 2010;35:460e8. Peeters F, Nicolson NA, Berkhof J. Levels and variability of daily life cortisol secretion in major depression. Psychiatry Research 2004;126:1e13. Persson R, Garde AH, Hansen AM, Osterberg K, Larsson B, Orbaek P, et al. Seasonal variation in human salivary cortisol concentration. Chronobiology International 2008;25:923e37. Pruessner JC, Wolf OT, Hellhammer DH, Buske-Kirschbaum A, von Auer K, Jobst S, et al. Free cortisol levels after awakening: a reliable biological marker for the assessment of adrenocortical activity. Life Sciences 1997;61:2539e49. Segal ZV, Williams JMG, Teasdale JD. Mindfulness-based cognitive therapy for depression: a new approach to preventing relapse. New York: Guilford Press; 2002. Stetler C, Miller GE. Blunted cortisol response to awakening in mild to moderate depression: regulatory influences of sleep patterns and social contacts. Journal of Abnormal Psychology 2005;114:697e705. Stetler C, Miller GE. Depression and hypothalamic-pituitary-adrenal activation: a quantitative summary of four decades of research. Psychosomatic Medicine 2011;73:114e26. Taylor CB, Conrad A, Wilhelm FH, Strachowski D, Khaylis A, Neri E, et al. Does improving mood in depressed patients alter factors that may affect cardiovascular disease risk? Journal of Psychiatric Research 2009;43:1246e52. Teasdale JD, Segal ZV, Williams JM, Ridgeway VA, Soulsby JM, Lau MA. Prevention of relapse/recurrence in major depression by mindfulness-based cognitive therapy. Journal of Consulting and Clinical Psychology 2000;68:615e23. Thorn L, Hucklebridge F, Esgate A, Evans P, Clow A. The effect of dawn simulation on the cortisol response to awakening in healthy participants. Psychoneuroendocrinology 2004;29:925e30. Thorn L, Hucklebridge F, Evans P, Clow A. Suspected non-adherence and weekend versus week day differences in the awakening cortisol response. Psychoneuroendocrinology 2006;31:1009e18. Vreeburg SA, Hartman CA, Hoogendijk WJ, van Dyck R, Zitman FG, Ormel J, et al. Parental history of depression or anxiety and the cortisol awakening response. British Journal of Psychiatry 2010;197:180e5. Vreeburg SA, Hoogendijk WJ, van Pelt J, Derijk RH, Verhagen JC, van Dyck R, et al. Major depressive disorder and hypothalamic-pituitary-adrenal axis activity: results from a large cohort study. Archives of General Psychiatry 2009a;66: 617e26. Vreeburg SA, Kruijtzer BP, van Pelt J, van Dyck R, DeRijk RH, Hoogendijk WJ, et al. Associations between sociodemographic, sampling and health factors and various salivary cortisol indicators in a large sample without psychopathology. Psychoneuroendocrinology 2009b;34:1109e20. Wardenaar KJ, Vreeburg SA, van Veen T, Giltay EJ, Veen G, Penninx BW, et al. Dimensions of depression and anxiety and the hypothalamo-pituitary-adrenal axis. Biological Psychiatry 2011;69:366e73. Wilhelm I, Born J, Kudielka BM, Schlotz W, Wust S. Is the cortisol awakening rise a response to awakening? Psychoneuroendocrinology 2007;32:358e66. Zoccola PM, Dickerson SS, Yim IS. Trait and state perseverative cognition and the cortisol awakening response. Psychoneuroendocrinology 2011;36:592e5.