Longitudinal effects of the SSRI paroxetine on salivary cortisol in Major Depressive Disorder

Psychoneuroendocrinology (2015) 52, 261—271

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/psyneuen

Longitudinal effects of the SSRI paroxetine on salivary cortisol in Major Depressive Disorder Henricus G. Ruhé a,b,∗,1, Sharina J. Khoenkhoen a,1, Koen W. Ottenhof a, Maarten W. Koeter a, Roel J.T. Mocking a, Aart H. Schene a,c,d a

Program for Mood Disorders, Department of Psychiatry, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands b University of Groningen, University Medical Center Groningen, University Center for Psychiatry, Mood and Anxiety Disorders, Groningen, The Netherlands c Department of Psychiatry, Radboud University Medical Center, Nijmegen, The Netherlands d Donders Institute for Brain, Cognition and Behavior, Radboud University, Nijmegen, The Netherlands Received 8 July 2014; received in revised form 10 October 2014; accepted 30 October 2014

KEYWORDS SSRI; Depressive disorder; HPA-axis; Cortisol; Treatment response; Remission

Summary Hypothalamic-pituitary-adrenal (HPA)-axis dysregulation is a prominent finding in more severe Major Depressive Disorder (MDD), and is characterized by increased baseline cortisol levels at awakening (BCL), blunted cortisol awakening response (CAR) and increased area under the cortisol curve (AUC). Selective serotonin reuptake inhibitors (SSRIs) appear to normalize HPA-axis dysfunction, but this is hardly investigated longitudinally. We studied salivary BCL, CAR and AUC at awakening and 30 min thereafter. We compared measurements in initially drug-free MDD-patients with healthy controls (HCs) at study-entry. In patients, we repeated measures after 6 and 12 weeks’ treatment with the SSRI paroxetine. Non-responding patients received a randomized dose-escalation after six weeks’ treatment. We found no significant study-entry differences in BLC, CAR or AUC between MDD-patients (n = 70) and controls (n = 51). In MDD-patients, we found general decreases of BCL and AUC during paroxetine treatment (p ≤ 0.007), especially in late and non-responders. Importantly, while overall CAR did not change significantly over time, it robustly increased over 12 weeks especially when patients achieved remission (p ≤ 0.041). The dose-escalation intervention did not significantly influence CAR or other cortisol parameters.

∗ Corresponding author at: Room 5.16, Mood and Anxiety Disorders, University Center for Psychiatry, University Medical Center Groningen Hanzeplein 1, 9700 RD Groningen, The Netherlands. Phone: +31 50 361 2367; fax: +31 50 361 1699. E-mail address: [email protected] (H.G. Ruhé). 1 Both authors contributed equally to the manuscript.

http://dx.doi.org/10.1016/j.psyneuen.2014.10.024 0306-4530/© 2014 Elsevier Ltd. All rights reserved.

262

H.G. Ruhé et al. In conclusion, paroxetine seems to interfere with HPA-axis dysregulation, reflected in significant overall decreases in BCL and AUC during treatment. Paroxetine appears to decrease HPA-axis set-point in MDD, which might result in increased HPA-axis activity over time, which is further improved when patients achieve remission (ISRCTN register nr. ISRCTN44111488). © 2014 Elsevier Ltd. All rights reserved.

1. Introduction The hypothalamic-pituitary-adrenal-axis (HPA-axis) is the main endocrinological regulator of allostasis. A robust shortterm HPA-axis response to acute stress may confer optimal physiological function and reflect adaptability or reactivity to environmental demands (Kudielka et al., 2009; Clow et al., 2004). However, more enduring HPA-axis hyperactivity is seen in approximately 73% of patients with Major Depressive Disorder (MDD) (Vreeburg et al., 2009; Holsboer, 2000). In general in MDD, HPA-axis alterations are found as (I) hyperactivity: increased cortisol in blood and cerebrospinal fluid, (II) non-suppression: higher rates of non-suppression to the dexamethasone suppression test and dexamethasoneCRH (DEX-CRH) test (Heuser et al., 1994; Stetler and Miller, 2011), and/or (III) decreased HPA-axis feedback inhibition. These factors can be integrated by the hypothesis that impaired glucocorticoid receptor (GR)-mediated feedback inhibition (non-suppression) leads to a higher baseline cortisol secretion (hyperactivity), which diminishes reactive capacity (Pariante, 2009). These HPA-disturbations are especially observed in severe/melancholic MDD and/or inpatients with MDD (Kunugi et al., 2010; Stetler and Miller, 2011). However, decreased HPA-axis activity was also found (Stetler and Miller, 2005; Huber et al., 2006), in particular during chronic depressive episodes (duration >2 years) (Kunugi et al., 2010). This differential stress-effect might explain contradictory findings of HPA-axis activity in MDD studies. Selective Serotonin Reuptake Inhibitors (SSRIs) are antidepressant drugs extensively used in MDD. Interestingly, besides increasing serotonergic neurotransmission, SSRIs might also change HPA-axis disturbances (Vermetten et al., 2006; Aihara et al., 2007). Only one randomized trial has been performed — in generalized anxiety disorder — and reported reductions in HPA-axis hyperactivity during treatment with the SSRI escitalopram versus placebo (Lenze et al., 2010). Furthermore, in healthy controls (HCs), HPA-axis awakening response significantly increased after six days of citalopram versus placebo (Harmer et al., 2003). In MDD, the SSRI fluoxetine decreased corticotropin-releasing hormone (CRH) (indicative of HPA-hyperactivity) in a small open study (De Bellis et al., 1993). A 16 weeks open study with citalopram in 20 MDD-patients showed significant decreases in cortisol (after DEX-CRH) (Nikisch et al., 2005). However, after 5 weeks of treatment with the SSRI paroxetine or the non-SSRI amitriptyline, HPA-axis activity only decreased in amitriptyline responders (Deuschle et al., 2003). Importantly, effects of antidepressants, including SSRIs, on the HPA-axis seem to occur mainly in MDD-patients responsive to treatment (Deuschle et al., 2003; Nikisch et al., 2005). Therefore, it remains largely unclear whether

HPA-axis abnormalities resolve as a result of SSRI treatment, or due to clinical improvement (state effect). Nevertheless, HPA-axis dysregulation has also been observed in remitted MDD patients, indicating a persistent trait (Bhagwagar et al., 2003; Lok et al., 2011). Therefore, it has been suggested that resolving HPA-axis abnormalities during MDD treatment indicates SSRI response. This may have important clinical implications, also explaining why persistent HPA-axis hyperactivity during SSRI treatment predicts MDD relapse (Appelhof et al., 2006; Hardeveld et al., 2014), although relapse has also been associated with HPA-hypoactivity (Bockting et al., 2012). These HPA-axis abnormalities may therefore form novel targets for (add-on) interventions (Pariante, 2009). However, to date, studies longitudinally investigating the HPA-axis in MDD-patients during SSRI treatment to disentangle antidepressant and state effects are largely lacking, especially with a randomized controlled trial design. To investigate (I) HPA-axis activity and awakening response in MDD, (II) its modulation by SSRIs, and (III) its association with treatment response, we repeatedly measured salivary cortisol responses to awakening in patients treated for 12 weeks with paroxetine. First, we compared patients to matched HCs at study entry. Subsequently, in the patients, we investigated changes in HPA-axis activity over time during treatment. Moreover, we included a randomized, double-blind, placebo-controlled SSRI dose-escalation in non-responders after six weeks of treatment (Ruhe et al., 2009), enabling investigation of causal effects of different paroxetine doses on HPA-axis activity. We hypothesized that (I) MDD-patients would have increased HPA-axis activity and decreased awakening response compared to HCs; (II) in patients, HPA-axis dysregulation would improve during 12 weeks paroxetine-treatment; (III) this change would be especially present in treatment responders (and remitters), and (IV) effects on the HPA-axis would be more outspoken after dose-escalation for non-responders.

2. Methods 2.1. Participants This study reports a secondary research question in a larger study (ISRCTN register nr. ISRCTN44111488; http://www. trialregister.nl/trialreg/admin/rctview.asp?TC=193) (Ruhe et al., 2009). Following approval by the institutional ethical committee and written informed-consent, we recruited drug-free outpatients. Inclusion criteria were: MDD determined by the structured clinical interview for DSM-IV (SCID), age 18—70 years and a Hamilton Depression Rating Scale score (HDRS17 -score) >18. All participants were drug-naive or drug-free for >4 weeks and, if treated before, had received

Longitudinal effects of the SSRI paroxetine on salivary cortisol in MDD ≤1 antidepressant (other than paroxetine; at an effective dose for >4 weeks) for the present MDD-episode. Exclusion criteria were: bipolar disorder, psychotic features, neurological cognitive impairments, primary anxiety and/or substance abuse disorders, severe suicidal ideation and pregnancy. We allowed (low dose) benzodiazepine use, and excluded patients who used drugs which directly affect HPA-axis activity (e.g. systemic corticosteroids). We recruited HCs by advertisements. HCs were in good physical health and never used psychotropic medication. Exclusion criteria for HCs were SCID-positive current or lifetime psychiatric disorder(s), including addiction disorders, Beck Depression Inventory score >9, >4 alcoholic beverages/day (last month) or a 1st-degree relative with psychiatric disorder(s). HCs could have incidentally used illicit drugs; drug-use one month prior to participation was not allowed.

2.2. Intervention Patients were treated at the outpatient department of the Program for Mood disorders of the Academic Medical Center. After study-entry, patients were treated open label with paroxetine 20 mg/day for 6 weeks. After 6 weeks (T0), responders (≥50% decrease of symptoms) continued treatment with paroxetine 20 mg/day for another 6 weeks (T1). Non-responders were randomized at T0 by a computer-program and received either a true paroxetine, or a placebo dose escalation, added to the initial dose of paroxetine 20 mg/day (Supplemental Fig. S1). Dose-escalation consisted of incremental steps of one capsule every 5 days (to a maximum of 50 mg/day). Paroxetine serum concentrations (PSC) were measured at T0 and T1 with a validated HPLC-MS/MS method (therapeutic range 10—75 ␮g/L). For more details see Ruhe et al. (2009). Clinical measurements for patients were scheduled at study-entry, T0 and T1. HCs were measured at study-entry only.

2.3. HPA-axis Several studies have indicated that salivary cortisol levels are valid measures of HPA-axis activity (Westermann et al., 2004; Kirschbaum and Hellhammer, 1994; Gallagher et al., 2006; Wust et al., 2000), particularly the physiologic rise in cortisol within 30 min after waking up (Clow et al., 2004). This rise has been hypothesized to be able to quantify both HPA-activity (cortisol at awakening; baseline cortisol level; BCL), and HPA-axis feedback inhibition (difference between cortisol 30 min after awakening and awakening; Cortisol Awakening Response; CAR), which are both reflected in the area under the curve (AUC). These measures show good intra-individual stability across time and can quantify subtle changes in HPA-axis regulation, thereby providing a valid method to study changes in HPA-axis activity over a treatment period (for review see Fries et al. (Fries et al., 2009)). After instruction, all subjects collected salivary samples at home, immediately after awakening and 30 min thereafter. Waking time and the time of the second saliva sample were recorded. Samples were refrigerated and brought to the hospital at the next visit. To avoid blood

263

contamination, participants refrained from brushing their teeth or eat before collecting both samples. Cortisol measurements were obtained at study-entry, T0 and T1, and stored at −20 ◦ C until analysis by radioimmunoassay (IBL Hamburg; designed for saliva samples). Intra- and inter-assay variations were 5.1% and 6.5%, respectively.

2.4. Statistics 2.4.1. Data cleaning and missing values All cortisol-measures were normally distributed. As done previously (Vreeburg et al., 2009), we considered 160% of the highest BCL (41.4 nmol/L) as maximum possible CAR value (70 nmol/L) (Clow et al., 2004). Because salivary cortisol levels above this limit are likely contaminated by blood and may distort further analyses, we considered 3/235 measurements as missing. We calculated CAR as the saliva cortisol level measured at 30 ± 10 min after awakening minus the BCL. Nineteen measurements outside this ±10 min range were also considered missing. AUC was calculated as BCL + ½*CAR (Fig. S2). In order to quantify how missing data influenced results we also performed our analyses in imputed datasets, which is explained in the supplemental. 2.4.2. Treatment response We operationalized treatment response by dividing patients into three response groups based on HDRS17 -score relative to the pre-treatment score: (I) early responders (≥50% decrease on HDRS17 -score at T0 ), (II) late responders (≥50% decrease on HDRS17-score at T1 ) and (III) non-responders (<50% decrease on HDRS17 -score at T0 and T1 ). Remission was defined as a HDRS17 score ≤7. 2.4.3. Analyses We used IBM-SPSS v.19.0 for all analyses. Clinical data were analyzed on intention to treat (ITT) basis and for clinical measures of drop-outs we used a last observation carried forward approach. For comparisons of patients vs. HCs, we used a propensity score (PS1 ) (Bartak et al., 2009) representing the predicted probability for a case to belong to a certain group, calculated in a binary logistic model with the desired confounders (including age, gender, alcohol-use, smoking (Kudielka et al., 2009) and race (Hajat et al., 2010)) as predictors (Rosenbaum and Rubin, 1983). For comparison of the three response groups, we considered age, gender, alcoholuse, smoking, race, and body mass index (BMI) at study-entry as covariates, for which we finally included only variables that differed trendwise (p < 0.1) univariately. In addition, we included awakening time as a fixed covariate in all analyses (Stalder et al., 2009). Because non-responding patients at T0 were randomized to receive a dose-escalation or not, confounders were assumed to be equally distributed (Table S1) resolving the need for propensity scores for analyses of dose-escalation. For comparisons between groups, we used Linear Regression Analysis with BCL, CAR and AUC as dependent variables and group (patients and HCs, final remitters and non-remitters, dose-escalation) as independent variable (including the propensity score and/or covariates as indicated).

H.G. Ruhé et al. Enrollment

264 Referred for assessment of eligibility (n= 278)

Study-entry

Excluded before open phase - Invited but withdrew (n= 27) - Ineligible by criteria (n= 88) - Ineligible by current use of antidepressants (n= 37) - Overall refusal to participate (n= 47) - Excluded other reasons (n=9)

Start open paroxetine 20mg/day (n= 70)

Open phase

Dropout open phase paroxetine 20mg/day (n= 11) - Inefficacy (n= 1) - Adverse effects (n= 3) - Other reasons (n= 6) - Withdrawn after interim (n= 1)

Randomized phase or continuation

T0

T1

Randomization of Non-responders open phase (n= 42)

Paroxetine dose-escalation (n= 21) [paroxetine30-50mg/day]

Placebo dose-escalation (n= 21) [paroxetine 20mg/day + placebo 10-30mg/day]

Dropout Rand. Phase (n= 1) - Inefficacy (n= 0) - Adverse effects (n= 0) - Other reasons (n= 1)

Dropout Rand. Phase (n= 5) - Inefficacy (n= 3) - Adverse effects (n= 3) - Other reasons (n= 1)

Analyzable clinical data (n= 20) - Wk 12 completer (n= 20)

Analyzable clinical data (n= 18) - Wk 12 completer (n= 16)

Figure 1

Responders (n= 17)

Paroxetine continuation (n= 17) [paroxetine 20mg/day] Dropout Rand. Phase (n= 0) - Inefficacy (n= 0) - Adverse effects (n= 0) - Other reasons (n= 0)

Analyzable clinical data (n= 17) - Wk 12 completer (n= 17)

Patient disposition.

To investigate changes in cortisol parameters over time (during treatment) in all patients and patient subgroups, we used Linear Mixed Model Analyses, which enable to measure changes of parameters over time and the interaction effects of subgroups*time (Jaeger, 2008), while correcting for covariates. In the mixed models investigating treatment response, cortisol parameters were the dependent variable and response group (early, late and non-response), follow-up time, measurement moment and their interactions were independent variables. We corrected results for age, gender and for variables that differed between the three response groups (p < 0.1; i.e. alcohol use and smoking) and awakening time as a time-dependent covariate. Improvement of modelfit was judged by decreasing Akaike Information criterion (AIC). With respect to quantification of remission effects, two models were applied. One model investigated differences between ‘final remitters and non-remitters’ as a groupvariable (i.e. differences between effects in final remitters and non-remitters as a fixed variable over time), while the second model investigated the effect of ‘remission as a timedependent state-factor’ (i.e. difference between subjects in remission vs. non-remitters at a certain time point during treatment which is changing over time).

3. Results 3.1. Patient disposition Seventy patients (mean age 42.5 ± 7.8) started open label paroxetine at study-entry (see patient-flow in Fig. 1). At T0 , 11 patients had dropped out before week 6. The ITT response rate in the open phase was 17/70 (24.3%); so 42 non-responding patients were randomized at T0 and entered the double-blind phase. A total of 36 patients completed the 6 week randomization phase; in addition the 17 T0

responders also completed another 6 weeks of treatment. We obtained at least 1 HDRS17 score after randomization for 38 patients, therefore yielding 55 patients for the longitudinal analyses. Of these 55 patients 17 (30.9%) were early responders, 18 (32.7%) were late responders and 20 (36.4%) were non-responders. After data-cleaning we obtained 63, 52 and 48 valid cortisol measurements at study-entry, T0 and T1 , respectively. We recruited 51 controls of whom we obtained valid cortisol measurements at study-entry for 47 subjects. Subjects without valid cortisol measures were significantly younger, and of non-Caucasian heritage, but did not differ on other demographic characteristics, or (for patients) depression-related characteristics (data available on request).

3.2. Population characteristics The 70 patients had a mean HDRS17 of 25.2 ± 4.4 (SD) at study entry, on average indicating severe MDD. Fourteen (20%) had a comorbid (secondary) anxiety disorder (eleven panic disorder/agoraphobia, three specific phobia, but none had a posttraumatic stress disorder (PTSD)), 52 (74.3%) had melancholic features, 31 (41.3%) had a recurrent MDE and 9 (12.9%) had a MDE for >2 years. Compared to controls, patients had significantly lower education-levels and were more often married and/or divorced and of non-Caucasian ethnicity (Table 1). After randomization at T0, patient groups did not differ significantly regarding study-entry characteristics (Table S1). With regard to further analyses, early and late responders smoked and drank trendwise more than non-responders (p ≥ 0.06; Table S2). Response groups did not significantly differ for other variables (all p > 0.12)., Final remitters (n = 15) had significant lower study-entry HDRS17 scores than final non-remitters (n = 55; 22.4 ± 2.4 vs. 26.0 ± 4.5; p = 0.004) but did not differ otherwise (data available on request).

Longitudinal effects of the SSRI paroxetine on salivary cortisol in MDD Table 1

265

Characteristics of patients and healthy controls.

Age (yrs ± SD) at study entry Sex Female, n (%) Marital state, n (%) Never married Married Divorced Widowed Education, n (%) Low Intermediate High Alcohol use (units/wk ± SD) Current smoking (cigarettes/day ± SD) Race, n (%) Caucasian Creole Asian Other MDE descriptives HDRS17 (± SD) Recurrent MDD (n(%)) Episode >2 years (n(%)) Melancholic subtype (n(%)) Waking up time (h ± SD) Cortisolparameters BCL (nmol/L ± SE) CAR (nmol/L ± SE) AUC (nmol/L ± SE)

MDD patients (n = 70)

Healthy controls (n = 51)

Significancea

42.5 ± 7.8

42.4 ± 8.4

0.991

46 (65.7)

33 (64.7)

1.000

22 (31.4) 25 (35.7) 21 (30.0) 1 (1.4)

28 (54.9) 15 (29.4) 4 (7.8) 1 (2.0)

0.010

13 (18.8) 42 (60.9) 14 (20.3) 5.8 ± 10.7 10.7 ± 12.4

0 (0.0) 16 (31.4) 35 (68.6) 6.9 ± 6.9 6.9 ± 9.6

<0.001

38 (54.3) 14 (20.0) 11 (15.7) 7 (10.0)

45 (88.2) 5 (9.8) 1 (2.0) 0 (0.0)

0.001

25.2 ± 4.4 30 (42.9) 9 (13.0) 52 (74.3) 07:22 ± 1:45

N/A

07:20 ± 1:27

0.523 .079

0.912 b

12.9 ± 1.1 5.9 ± 1.2 16.2 ± 1.0

14.4 ± 1.3 6.3 ± 1.4 17.6 ± 1.2

0.388 0.818 0.399

Abbreviations: AUC = area under the curve; BCL = baseline cortisol level; CAR = cortisol awakening response; HDRS17 = 17-item Hamilton Depression Rating Scale; MDD = Major Depressive Disorder. a Determined by chi-square/independent samples t-test/ANOVA/mixed models as appropriate. b Corrected for propensity score and awakening time. Means based on dataset without multiple imputation. See also Fig. S2.

3.3. Cortisol parameters at study-entry Although mean BCL, CAR and AUC were numerically lower in patients compared to HCs, these differences were not significant at study-entry (all p > 0.24 uncorrected; p ≥ 0.39) corrected for PS1 and awakening time (Table 1/Fig. S2). In the patient-group, cortisol parameters at study-entry were not associated with HDRS-scores (all p ≥ 0.11; Linear Regression). At study entry, response groups did not differ significantly with respect to BCL, CAR or AUC at study entry (p > 0.103). We neither observed significant differences in study-entry BCL, CAR and AUC between final remitters and final non-remitters (all p > 0.19; corrected for HDRS17 at study entry and awakening time).

3.4.1. Association with treatment response As a next step, we investigated differences in changes in BCL, CAR and AUC over time between the three response

3.4. Changes in cortisol parameters over time during paroxetine treatment On average, BCL and AUC values both decreased significantly over time (F1,50.703 = 8.090; p = 0.006 and F1,51.921 = 7.756; p = 0.007, respectively, Fig. 2), while CAR did not significantly change during the study (F1,49.135 = 0.047; p = 0.830).

Figure 2 patients).

Changes in cortisol parameters over time (all

266

H.G. Ruhé et al.

Figure 4

CAR in final remitters vs non-remitters.

decrease in AUC while in early responders the AUC slightly increased over 12 weeks of treatment. 3.4.2. Association with remission of the depressive episode For final remitters and non-remitters, changes in CAR over time differed significantly, despite comparable CAR-values at study-entry. Final remitters showed a significant increase in CAR, while CAR in final non-remitters did not change (final-remission*time interaction F1,52.571 = 4.400; p = 0.041; Fig. 4). Although by itself PSCs were not significantly associated with CAR (F1,62.130 = 0.031; p = 0.860), adding this variable contributed significantly to this model,(decreasing the AIC). AUC and BCL did not differentially change over time for final remitters versus non-remitters (all p > 0.37). Additionally, to investigate whether remission status (either at T0 or T1) was associated with higher CARs, we included remission as a time-dependent variable in the Linear Mixed Model Analyses. Indeed, this showed that remission, at any time-point, was associated with higher CARs (F1,80.519 = 5.682; p = 0.019). This association did not change over time as the remission status*time interaction was not significant (p = 0.673).

3.5. True paroxetine versus placebo dose-escalation after randomization

Figure 3 Changes in cortisol parameters over time by response groups.

groups in a mixed model, by testing response-group*time interactions (correcting for alcohol use, smoking and awakening time). Because by design late and non-responding patients received different dosages of paroxetine after T0 (randomization of only non-responders), we added PSCs as additional time-dependent covariate, which improved the models (based on lower AIC). We found significant response*time interactions for AUC (F2,42.433 = 3.976; p = 0.026) and a trendwise interaction for and BCL (F2,42.832 = 2.335; p = 0.11), but not for CAR (F2,44.753 = 0.201; p = 0.819, Fig. 3). In late and non-responders we observed a

After randomization groups did not differ significantly for HDRS17 or cortisol measures at T0 (all p > 0.18; data available on request). Remission rates were 3 (14.3%) in the true and 2 (9.5%) in the placebo-dose-escalation groups (p = 1.000 Fisher’s Exact). Dose-escalation after 6 weeks of treatment did not significantly affect BCL, AUC or CAR-courses (condition*time interaction; all p > 0.08; Table 2).

4. Discussion This study investigated differences in salivary cortisol between HCs and MDD-patients and the effects on cortisol measurements of treatment with paroxetine in the MDDpatients. We found no significant differences in BCL, CAR or AUC between controls and MDD-patients. Before treatment, BCL, CAR and AUC at study-entry were also similar

Longitudinal effects of the SSRI paroxetine on salivary cortisol in MDD Table 2

267

Changes in cortisol parameters for placebo- and true dose-escalation groups. Placebo dose-escalation

AUC nmol/L BCL nmol/L CAR nmol/L

True dose-escalation

T0

T1

T0

T1

12.4 ± 1.5 8.9 ± 1.3 7.1 ± 1.8

11.7 ± 1.6 9.6 ± 1.5 2.6 ± 1.8

13.2 ± 1.3 10.3 ± 1.2 5.7 ± 1.6

11.3 ± 1.6 7.5 ± 1.5 6.7 ± 1.7

Between T0 (randomization to dose escalation) and T1 we found no significant differences in changes in cortisolparameters (mean ± SE) between patients who received placebo dose-escalation versus patients who received real dose-escalation (p > 0.08; Linear Mixed Model Analysis). Analyses based on 17 patients receiving a placebo dose-escalation and 19 patients receiving a true dose-escalation. Means based on dataset without multiple imputation. AUC = area under the curve, BCL = baseline cortisol level, CAR = cortisol awakening response.

in (early/late) final responders versus non-responders and in final remitters and non-remitters. More specifically, after paroxetine-treatment, we observed a decrease of BCL and AUC in MDD-patients over time. Especially in early responders the AUC increased over time, and interestingly, robust increases in CAR were found in patients who achieved remission from MDE while no increases in AUC or CAR, were observed in late/non-responders and non-remitters, respectively. Furthermore, remission status during treatment was significantly associated with higher CARs. The randomized dose-escalation in 42 non-responders 6 weeks after studyentry did not significantly influence BCL, AUC or CAR.

4.1. Cortisol parameters at study-entry 4.1.1. Hyperactivity Our hypothesis of HPA-axis hyperactivity in MDD-patients compared to controls had to be rejected. Moreover, our numerical results even indicate lower HPA-axis activity. However, given the small differences, our study was clearly underpowered (power = 0.12 for a double-sided test at ˛ = 0.05) to significantly detect these differences relative to HCs at study-entry (Fig. S2). Previous studies comparing MDD-patients and HCs concerning cortisol parameters showed inconsistent results, although a recent meta-analysis of 361 studies in 18,454 individuals reported HPA-axis hyperactivity with small effect sizes in outpatient populations with MDD (Stetler and Miller, 2011). 4.1.2. Awakening response Although the CAR was indeed numerically lower in our patient group, we could not confirm our hypothesis of decreased HPA-axis awakening response in patients versus controls. Evidence exists for a blunted CAR in MDD (Stetler and Miller, 2005; Huber et al., 2006) and Seasonal Affective Disorder (SAD) during winter (Thorn et al., 2011) which might be associated with a change in diurnal sleepwake rhythm or lack of social contacts (Stetler and Miller, 2005). However, also higher CARs (Pruessner et al., 2003; Bhagwagar et al., 2005; Vreeburg et al., 2009) were reported in MDD-patients compared to HCs. 4.1.3. Explaining inconsistencies Some explanations for these inconsistencies should be considered. First, previous studies show larger effects in more severe MDD-populations (Stetler and Miller, 2011). However,

since we also selected an average severe group (mean HDRS17 = 25.2), this does not seem plausible. Second, different subpopulations and (omission of) correction for different confounding variables (Vreeburg et al., 2009) may provide clarification. Indeed, different disease duration, comorbidity, depression severity and subtypes of MDD affect findings in HPA-axis abnormalities (Kunugi et al., 2010; Shelton, 2007; Lamers et al., 2013). Atypical depression and SAD result in HPA-axis hypoactivation (Thorn et al., 2011), in contrast to melancholic depression being characterized by HPA-axis hyperactivation (Kunugi et al., 2010; Lamers et al., 2013), but this is equivocal (Stetler and Miller, 2011). Higher CARs were especially found in association with, more severe anhedonia (Wardenaar et al., 2011), which was replicated by Veen et al. (2011) reporting significant correlations between CAR and general distress symptoms. Therefore, correction for differences in depression duration or subtype, anhedonia and general distress should ideally be applied, but are statistically not meaningful (and even inappropriate) in relatively small samples like ours. Nevertheless, in our moderate to severely ill, predominantly melancholic (>74% of the patients were melancholic) depressed patients without comorbid diseases like PTSD, these factors unlikely will explain our results. Although — indeed — mean CAR at study-entry was numerically lower in chronic MDDpatients (0.60 ± 3.56 nmol/l [SE]; n = 9) versus non-chronic (6.14 ± 1.21 nmol/l) MDD-patients, these differences in BCL, CAR or AUC were not significant (p = 0.61; p = 0.18; p = 0.73, respectively; post hoc). Therefore, in our MDD-sample it remains speculative why we found non-significantly lower BCL- and CAR-levels at study entry. Instead of HPA-axis hyperactivity, these observations might indicate blunted HPA-axis activity and diminished awakening response (Stetler and Miller, 2005; Huber et al., 2006; Chida and Steptoe, 2009), which improved after remission (see below).

4.2. Changes in cortisol parameters during treatment and association with treatment response One robust finding is the significant decrease in BCL (and AUC) during treatment, suggesting that SSRI treatment decreases the cortisol ‘setpoint’ (Holsboer, 2000; Schule, 2007). Previous studies also reported a decreasing effect of SSRIs on HPA-axis hyperactivity (Nikisch et al., 2005; Buhl et al., 2010; Lenze et al., 2010), although contradictory findings exist (Deuschle et al., 2003; Bschor et al., 2012).

268 In theory, lowering the cortisol setpoint could facilitate improvement of HPA-axis awakening response which was suggested to be blunted in MDD (Pariante, 2009). While other studies predominantly reported effects on BCL only in treatment responders (Nikisch et al., 2005; Deuschle et al., 2003), we observed decreases in AUC especially in late and non-responders, which was significant for the AUC and showed a trend for BCL. Furthermore, at all measurements over the treatment period, CAR was numerically higher in the early response group (n.s.; Fig. 3). Because the AUC represents a combined measure for BCL and CAR, these data together are suggestive of an increase in CAR without changes in BCL in early responders and a decrease in BCL and no change in CAR in late and non-responders. Since groups did not differ significantly on study-entry HDRS17 -scores, other MDD-characteristics (duration of current episode, melancholic features) and HPAmeasures, it remains elusive what differentiates late and non-responders from early responders and/or whether the late/non-responders are more severely disturbed (Fig. 3 and Table S2). Importantly, we found that especially achieving remission was associated with increases in CAR during paroxetinetreatment. This corroborates findings that in a large sample, significantly higher CARs were found in remitted patients relative to controls (Vreeburg et al., 2009) which was also found in drug-free remitted patients (Aubry et al., 2010). Post hoc, in our sample fifteen remitters had higher CARs at endpoint than controls at study-entry (11.1 ± 2.5 nmol/l [SE] versus 6.2 ± 1.3 nmol/l [SE], respectively (p = 0.08)). CAR increases could be interpreted as a return of normal HPA-axis activity (Kudielka et al., 2009; Clow et al., 2004; Pariante, 2009). A decrease in BCL might in fact increase the range of the awakening response, as reported before (Holsboer, 2000; Buhl et al., 2010; Lenze et al., 2010; Schule, 2007). Indeed, when we post hoc associated (repeated) measures of CAR with BCL, we found a significant inverse association between BCL and CAR, which, in addition, became stronger over time (BCL*time interaction F1,91.707 = 4.388; p = 0.039), while remission-status at T0 or T1 remained a significant independent predictor for CAR (F1,82.130 = 4.639; p = 0.034; all corrected for awakening time). These results suggest that paroxetine treatment improves HPA-awakening response (CAR) by a stronger inverse association with BCL over time of treatment (when BCL decreases), where in addition occurrence of remission further improves HPA-axis responsiveness. Increases in CAR as a representation of restored HPA-axis activity seems at odds with associations of higher CAR with risks for MDD. Furthermore, CAR was also increased in individuals with a parental history of MDD (Vreeburg et al., 2010) and in people with SAD (Thorn et al., 2011). In a prospective study, young adults with higher CARs had significant more risk to develop a depressive episode during one (Adam et al., 2010) and 2.5-year follow-up (Vrshek-Schallhorn et al., 2013). We previously showed that increased HPA-axis activity (at 8:00 and 22:00 h) exists in highly recurrent but remitted MDD-patients (Lok et al., 2011). In a large population based study, higher CARs predicted recurrence (Hardeveld et al., 2014). These findings could be interpreted as that high CARs also represent a trait marker for MDD-vulnerability or recurrence. Alternatively, although speculative, high CAR is

H.G. Ruhé et al. associated with improved cognition (Law et al., 2013; Evans et al., 2011). Higher CARs might thus represent an adaptive phenomenon: as a ‘preparation of the day’ to increase cognitive compensation (e.g. for daily hassles) and improve executive functioning in patients with (recurrent) MDD even when in remission. If this adaptive phenomenon collapses (e.g. after exhaustion and decreased/blunted CARs), a (new) depressive episode might occur, which is corroborated by the finding that especially HPA-hypoactivity was associated with relapse (Bockting et al., 2012). In the present study we additionally investigated the dose—response relationship of paroxetine with respect to changes in HPA-axis disfunction. However, the second phase of six weeks true paroxetine versus placebo dose-escalation revealed no significant differences in BCL, AUC or CAR between dose-groups. This null-finding corroborates with previous observations that dose-escalation of paroxetine (and other SSRIs) does not improve symptomatology (Ruhe et al., 2006; Adli et al., 2005) and neither increases SERToccupancy (Ruhe et al., 2009). An explanation for the changes in cortisol measures during treatment with paroxetine and the absence of associations with PSC or paroxetine dose, might lie in the non-serotonergic SSRI-effects on HPA-axis activity. Acute administration of antidepressants causes increases in GR expression and function, but after chronic treatment of rodents, GR-expression returns to control levels, while basal HPA-axis activity is still reduced. Currently, changes in HPA-axis activity after chronic treatment with SSRIs are thought to inhibit the multidrug resistance P-glycoprotein. This P-glycoprotein also expels glucocorticoids from cells, so SSRIs might indirectly increase intracellular concentrations of glucocorticoids, which can improve GR-function with increased awakening response as a result. It is of interest that this P-glycoprotein inhibition might not occur in treatment resistant patients (Pariante, 2009). Especially the non-responders and/or non-remitters in this study might have been early refractory patients, who might not have achieved final remission after several successive treatmentsteps (Ruhe et al., 2012). Unfortunately these patients were not followed up routinely to confirm this.

4.3. Limitations and strengths Most importantly, paroxetine treatment was not fully placebo-controlled, and only patients were followed-up. Measuring controls during the same follow-up would have quantified effects of repeated sampling. A full placebo group could have differentiated true drug effects from nonspecific effects of remission and/or over time. Full placebo treatment was not pursued because of ethical reasons and the original aim of the study (efficacy of dose-escalation). Nevertheless, the placebo-controlled dose-escalation phase of this study did not identify dose-related drug-effects on HPA-axis parameters. Second, although we corrected for awakening-time and age, gender, alcohol-use, smoking, race, BMI, HDRS17 -scores at study-entry in the propensityscores, we did not have data to control for potential other confounders like menstrual cyclus, late evening/night working hours or childhood adversity. Neither did we instruct subjects specifically to sample outside weekend-days (which

Longitudinal effects of the SSRI paroxetine on salivary cortisol in MDD was however mostly done the day before baseline visits during the week), nor did we quantify or specifically restrict alcohol use the night before sampling. Third, a more advanced HPA-axis sampling protocol would have enabled us to additionally detect more subtle changes. We only sampled at one day during the week, while nowadays it is considered good practice to base assessment upon measures repeated on two consecutive weekdays. This could have provided some measure of trait stability. We neither obtained saliva samples at 15, 45 and 60 min after awakening, nor a diurnal profile beyond these timepoints, which would have enabled us to determine more complete parameters of the cortisol awakening curve. Especially 45 min post waking samples are important for females who peak later than males. Our restriction to obtain two cortisol measurements only, might also have obscured a difference between patients and controls. We did not use objective electronic monitoring or an actimeter to reduce uncertainty of the actual wake and saliva sampling times. However, despite these limitations, our methods resulted in interesting changes in repeated cortisol measures that were longitudinally associated with clinical course of depressive symptoms. Fourth, we did not differentiate other sub-types than melancholic/atypical depression. Differentiation of other symptom-axes in MDD might better corresponds with particular HPA-axis dysregulation (Veen et al., 2011; Wardenaar et al., 2011). In future studies sub-types should differentiated with the Mood and Anxiety Symptoms Questionnaire (Wardenaar et al., 2011). Fifth, this type of research is susceptible to confounding by missing data, which also occurred in our sample. However, the main results of this study remained the same when we performed multiple imputation (see supplemental). Sixth, the slight different definitions of BCL- and CAR-measurements between studies complicate the discussion of comparisons between studies. This urges for uniform nomenclature and standardization of HPA-axis measures. Major strengths of our study are that (I) we investigated the longitudinal effects of SSRI treatment on HPA-axis activity, (II) in patients drug-free at study-entry and (III) treated uniformly according to a study protocol, and (IV) our results appeared robust when applying MI to resolve potential bias by missing values (drop-outs and outliers). Altogether, these strengths allowed us to longitudinally disentangle antidepressant and state effects on HPA-axis activity in MDD.

5. Conclusion This study investigated changes in salivary cortisol over time during treatment with paroxetine and a randomized doseescalation. Without evidence for baseline differences compared to controls, we found significant decreases in BCL and AUC in all patients over 12 weeks of paroxetine treatment, suggesting that SSRIs decrease HPA-axis set-point in MDD, which was most prominent in late and non-responders, while early responders showed no change. The robust CAR increase especially when patients became remitters versus no change in non-remitters during paroxetine treatment, suggests that improved HPA-axis feedback inhibition may result from a combination of overall decreasing cortisol values due to SSRI treatment and additional increased activity of the HPA-axis when patients remit from a depressive episode.

269

Role of the funding source This study was supported by a grant from the Netherlands Organization for Health Research and Development (ZonMw), program Mental Health, Education of Investigators in Mental Health (OOG; #100-002-002) to Dr. H.G. Ruhé. Dr. H.G. Ruhé is currently supported by a NWO/ZonMW VENI-Grant #016.126.059. The funding source by no means influenced the design or results of the study described in this manuscript.

Conflicts of interest statement None declared.

Acknowledgements The authors wish to thank the patients who participated in this study. Mrs. M. Haages managed randomization and maintained blinding. This study was supported by a grant from the Netherlands Organization for Health Research and Development (ZonMw), program Mental Health, education of investigators in mental health (OOG; #100-002-002) to Dr. H.G. Ruhé. Dr. H.G. Ruhé is currently supported by a NWO/ZonMW VENI-Grant #016.126.059.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10. 1016/j.psyneuen.2014.10.024.

References Adam, E.K., Doane, L.D., Zinbarg, R.E., Mineka, S., Craske, M.G., Griffith, J.W., 2010. Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 35, 921—931. Adli, M., Baethge, C., Heinz, A., Langlitz, N., Bauer, M., 2005. Is dose escalation of antidepressants a rational strategy after a medium-dose treatment has failed? A systematic review. Eur. Arch. Psychiatry Clin. Neurosci. 255, 387—400. Aihara, M., Ida, I., Yuuki, N., Oshima, A., Kumano, H., Takahashi, K., et al., 2007. HPA axis dysfunction in unmedicated major depressive disorder and its normalization by pharmacotherapy correlates with alteration of neural activity in prefrontal cortex and limbic/paralimbic regions. Psychiatry Res. 155, 245—256. Appelhof, B.C., Huyser, J., Verweij, M., Brouwer, J.P., Van Dyck, R., Fliers, E., et al., 2006. Glucocorticoids and relapse of major depression (dexamethasone/corticotropin-releasing hormone test in relation to relapse of major depression). Biol. Psychiatry 59, 696—701. Aubry, J.M., Jermann, F., Gex-Fabry, M., Bockhorn, L., Van der Linden, M., Gervasoni, N., et al., 2010. The cortisol awakening response in patients remitted from depression. J. Psychiatr. Res. 44, 1199—1204. Bartak, A., Spreeuwenberg, M.D., Andrea, H., Busschbach, J.J.V., Croon, M.A., Verheul, R., et al., 2009. The use of propensity score methods in psychotherapy research. Psychother. Psychosom. 78, 26—34.

270 Bhagwagar, Z., Hafizi, S., Cowen, P.J., 2003. Increase in concentration of waking salivary cortisol in recovered patients with depression. Am. J. Psychiatry 160, 1890—1891. Bhagwagar, Z., Hafizi, S., Cowen, P.J., 2005. Increased salivary cortisol after waking in depression. Psychopharmacology (Berl.) 182, 54—57. Bockting, C.L., Lok, A., Visser, I., Assies, J., Koeter, M.W., Schene, A.H., 2012. Lower cortisol levels predict recurrence in remitted patients with recurrent depression: a 5.5 year prospective study. Psychiatry Res. 200, 281—287. Bschor, T., Ising, M., Erbe, S., Winkelmann, P., Ritter, D., Uhr, M., et al., 2012. Impact of citalopram on the HPA system. A study of the combined DEX/CRH test in 30 unipolar depressed patients. J. Psychiatr. Res. 46, 111—117. Buhl, E.S., Jensen, T.K., Jessen, N., Elfving, B., Buhl, C.S., Kristiansen, S.B., et al., 2010. Treatment with an SSRI antidepressant restores hippocampo-hypothalamic corticosteroid feedback and reverses insulin resistance in low-birth-weight rats. Am. J. Physiol. Endocrinol. Metab. 298, E920—E929. Chida, Y., Steptoe, A., 2009. Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biol. Psychol. 80, 265—278. Clow, A., Thorn, L., Evans, P., Hucklebridge, F., 2004. The awakening cortisol response: methodological issues and significance. Stress 7, 29—37. De Bellis, M.D., Gold, P.W., Geracioti Jr., T.D., Listwak, S.J., Kling, M.A., 1993. Association of fluoxetine treatment with reductions in CSF concentrations of corticotropin-releasing hormone and arginine vasopressin in patients with major depression. Am. J. Psychiatry 150, 656—657. Deuschle, M., Hamann, B., Meichel, C., Krumm, B., Lederbogen, F., Kniest, A., et al., 2003. Antidepressive treatment with amitriptyline and paroxetine: effects on saliva cortisol concentrations. J. Clin. Psychopharmacol. 23, 201—205. Evans, P., Fredhoi, C., Loveday, C., Hucklebridge, F., Aitchison, E., Forte, D., Clow, A., 2011. The diurnal cortisol cycle and cognitive performance in the healthy old. Int. J. Psychophysiol. 79, 371—377. Fries, E., Dettenborn, L., Kirschbaum, C., 2009. The cortisol awakening response (CAR): facts and future directions. Int. J. Psychophysiol. 72, 67—73. Gallagher, P., Leitch, M.M., Massey, A.E., lister-Williams, R.H., Young, A.H., 2006. Assessing cortisol and dehydroepiandrosterone (DHEA) in saliva: effects of collection method. J. Psychopharmacol. 20, 643—649. Hajat, A., Diez-Roux, A., Franklin, T.G., Seeman, T., Shrager, S., Ranjit, N., et al., 2010. Socioeconomic and race/ethnic differences in daily salivary cortisol profiles: the multiethnic study of atherosclerosis. Psychoneuroendocrinology 35, 932—943. Hardeveld, F., Spijker, J., Vreeburg, S.A., Graaf, R., Hendriks, S.M., Licht, C.M., Nolen, W.A., Penninx, B.W., Beekman, A.T., 2014. Increased cortisol awakening response was associated with time to recurrence of major depressive disorder. Psychoneuroendocrinology 50C, 62—71, http://dx.doi.org/10.1016/j. psyneuen.2014.07.027. Harmer, C.J., Bhagwagar, Z., Shelley, N., Cowen, P.J., 2003. Contrasting effects of citalopram and reboxetine on waking salivary cortisol. Psychopharmacology (Berl.) 167, 112—114. Heuser, I., Yassouridis, A., Holsboer, F., 1994. The combined dexamethasone/CRH test: a refined laboratory test for psychiatric disorders. J. Psychiatr. Res. 28, 341—356. Holsboer, F., 2000. The corticosteroid receptor hypothesis of depression. Neuropsychopharmacology 23, 477—501. Huber, T.J., Issa, K., Schik, G., Wolf, O.T., 2006. The cortisol awakening response is blunted in psychotherapy inpatients suffering from depression. Psychoneuroendocrinology 31, 900—904.

H.G. Ruhé et al. Jaeger, T.F., 2008. Categorical data analysis: away from ANOVAs (transformation or not) and towards logit mixed models. J. Mem. Lang. 59, 434—446. Kirschbaum, C., Hellhammer, D.H., 1994. Salivary cortisol in psychoneuroendocrine research: recent developments and applications. Psychoneuroendocrinology 19, 313—333. Kudielka, B.M., Hellhammer, D.H., Wust, S., 2009. Why do we respond so differently? Reviewing determinants of human salivary cortisol responses to challenge. Psychoneuroendocrinology 34, 2—18. Kunugi, H., Hori, H., Adachi, N., Numakawa, T., 2010. Interface between hypothalamic-pituitary-adrenal axis and brain-derived neurotrophic factor in depression. Psychiatry Clin. Neurosci. 64, 447—459. Lamers, F., Vogelzangs, N., Merikangas, K.R., de, J.P., Beekman, A.T., Penninx, B.W., 2013. Evidence for a differential role of HPA-axis function, inflammation and metabolic syndrome in melancholic versus atypical depression. Mol. Psychiatry 18, 692—699. Law, R., Hucklebridge, F., Thorn, L., Evans, P., Clow, A., 2013. State variation in the cortisol awakening response. Stress 16, 483—492. Lenze, E.J., Mantella, R.C., Shi, P., Goate, A.M., Nowotny, P., Butters, M.A., et al., 2010. Elevated cortisol in older adults with generalized anxiety disorder is reduced by treatment: a placebocontrolled evaluation of escitalopram. Am. J. Geriatr. Psychiatry 19, 482—490. Lok, A., Mocking, R.J., Ruhe, H.G., Visser, I., Koeter, M.W., Assies, J., et al., 2011. Longitudinal hypothalamic-pituitary-adrenal axis trait and state effects in recurrent depression. Psychoneuroendocrinology 37, 892—902. Nikisch, G., Mathe, A.A., Czernik, A., Thiele, J., Bohner, J., Eap, C.B., et al., 2005. Long-term citalopram administration reduces responsiveness of HPA axis in patients with major depression: relationship with S-citalopram concentrations in plasma and cerebrospinal fluid (CSF) and clinical response. Psychopharmacology (Berl.) 181, 751—760. Pariante, C.M., 2009. Risk factors for development of depression and psychosis. Glucocorticoid receptors and pituitary implications for treatment with antidepressant and glucocorticoids. Ann. N. Y. Acad. Sci. 1179, 144—152. Pruessner, M., Hellhammer, D.H., Pruessner, J.C., Lupien, S.J., 2003. Self-reported depressive symptoms and stress levels in healthy young men: associations with the cortisol response to awakening. Psychosom. Med. 65, 92—99. Rosenbaum, P.R., Rubin, D.B., 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70, 41—55. Ruhe, H.G., Booij, J., Weert, H.C., Reitsma, J.B., Franssen, E.J., Michel, M.C., et al., 2009. Evidence why paroxetine dose escalation is not effective in major depressive disorder: a randomized controlled trial with assessment of serotonin transporter occupancy. Neuropsychopharmacology 34, 999—1010. Ruhe, H.G., Huyser, J., Swinkels, J.A., Schene, A.H., 2006. Dose escalation for insufficient response to standard-dose selective serotonin reuptake inhibitors in major depressive disorder: systematic review. Br. J. Psychiatry 189, 309—316. Ruhe, H.G., van, R.G., Spijker, J., Peeters, F.P., Schene, A.H., 2012. Staging methods for treatment resistant depression. A systematic review. J. Affect. Disord. 137, 35—45. Schule, C., 2007. Neuroendocrinological mechanisms of actions of antidepressant drugs. J. Neuroendocrinol. 19, 213—226. Shelton, R.C., 2007. The molecular neurobiology of depression. Psychiatr. Clin. North Am. 30, 1—11. Stalder, T., Hucklebridge, F., Evans, P., Clow, A., 2009. Use of a single case study design to examine state variation in the cortisol awakening response: relationship with time of awakening. Psychoneuroendocrinology 34, 607—614.

Longitudinal effects of the SSRI paroxetine on salivary cortisol in MDD Stetler, C., Miller, G.E., 2011. Depression and hypothalamicpituitary-adrenal activation: a quantitative summary of four decades of research. Psychosom. Med. 73, 114—126. Stetler, C., Miller, G.E., 2005. Blunted cortisol response to awakening in mild to moderate depression: regulatory influences of sleep patterns and social contacts. J. Abnorm. Psychol. 114, 697—705. Thorn, L., Evans, P., Cannon, A., Hucklebridge, F., Clow, A., 2011. Seasonal differences in the diurnal pattern of cortisol secretion in healthy participants and those with self-assessed seasonal affective disorder. Psychoneuroendocrinology 36, 816—823. Veen, G., van Vliet, I.M., Derijk, R.H., Giltay, E.J., van, P.J., Zitman, F.G., 2011. Basal cortisol levels in relation to dimensions and DSM-IV categories of depression and anxiety. Psychiatry Res. 185, 121—128. Vermetten, E., Vythilingam, M., Schmahl, C.D.E.K.C., Southwick, S.M., Charney, D.S., et al., 2006. Alterations in stress reactivity after long-term treatment with paroxetine in women with posttraumatic stress disorder. Ann. N. Y. Acad. Sci. 1071, 184—202. Vreeburg, S.A., Hartman, C.A., Hoogendijk, W.J., van, D.R., Zitman, F.G., Ormel, J., et al., 2010. Parental history of depression

271

or anxiety and the cortisol awakening response. Br. J. Psychiatry 197, 180—185. Vreeburg, S.A., Hoogendijk, W.J., van, P.J., Derijk, R.H., Verhagen, J.C., van, D.R., et al., 2009. Major depressive disorder and hypothalamic-pituitary-adrenal axis activity: results from a large cohort study. Arch. Gen. Psychiatry 66, 617—626. Vrshek-Schallhorn, S., Doane, L.D., Mineka, S., Zinbarg, R.E., Craske, M.G., Adam, E.K., 2013. The cortisol awakening response predicts major depression: predictive stability over a 4-year follow-up and effect of depression history. Psychol. Med. 43, 483—493. Wardenaar, K.J., Vreeburg, S.A., van, V.T., Giltay, E.J., Veen, G., Penninx, B.W., et al., 2011. Dimensions of depression and anxiety and the hypothalamo-pituitary-adrenal axis. Biol. Psychiatry 69, 366—373. Westermann, J., Demir, A., Herbst, V., 2004. Determination of cortisol in saliva and serum by a luminescence-enhanced enzyme immunoassay. Clin. Lab. 50, 11—24. Wust, S., Wolf, J., Hellhammer, D.H., Federenko, I., Schommer, N., Kirschbaum, C., 2000. The cortisol awakening response — normal values and confounds. Noise Health 2, 79—88.