Joke H. M. Tulen, PhD, Department of Psychiatry, Erasmus MC, P.O. Box 2040, 3000 CA Rotterdam, the Netherlands. Email: email@example.com
Aims: It has been suggested that schizophrenic patients are more vulnerable to stress than healthy persons, and that stressors can trigger a psychotic episode or worsen symptoms. The biological system often studied in relation to stress is the hypothalamic–pituitary–adrenal (HPA) axis, which controls the release of cortisol. We investigated whether the diurnal basal activity of the HPA axis differed between young male patients with schizophrenia and healthy controls.
Methods: Twenty-seven male patients (mean age 22 ± 5 years) and 38 healthy male control subjects (mean age 22 ± 3 years) were included in the present study. Saliva was sampled at five time points during the day: directly after awakening, 30 min thereafter, and at 12.00 hours, 16.00 hours and 22.00 hours.
Results: The cortisol concentration decreased significantly more during the day in the patient group thanin the control group. Patients also showed a significantly decreased area under the curve with respect to the increase, again indicating that the cortisol concentrations decreased more during the day in patients than in controls. Both the morning increase and the area under the curve with respect to the increase were significantly negatively correlated with negative symptom severity.
Conclusions: Patients with schizophrenia showed a different daytime sensitivity of the HPA axis. Our findings further suggest that an increase in negative symptom severity is related to a decreased HPA axis sensitivity.
SEVERAL AUTHORS HAVE suggested that patients with schizophrenia are more vulnerable to stress than healthy persons, and that stressors can trigger a psychotic episode or worsen symptoms.1–3 According to the Neural Diathesis Stress Model proposed by Walker and Diforio,2 schizophrenia is the result of an interaction between the genetic vulnerability of a person and the (stressful) life events a person encounters. It has been found that persons with schizophrenia report greater subjective stress to both major as well as minor life events compared with healthy control subjects,4–6 possibly including the social difficulties these patients experience.7 The biological system often studied in relation to stress is the hypothalamic–pituitary–adrenal (HPA) axis, which controls the release of several hormones in response to stress, including cortisol.4
HPA axis functioning can be investigated by inducing physical, cognitive or psychosocial stress, or simply by measuring basal cortisol levels at several time points during the day. Most studies investigating baseline cortisol have found increased levels of cortisol in patients with schizophrenia (for a review, see Walker and Diforio2). However, the majority of these studies did not take the diurnal cortisol pattern into account, which may lead to measurement biases, as cortisol concentrations fluctuate during the day. Unfortunately, studies investigating the diurnal pattern of cortisol in schizophrenia are scarce, and some of these studies measured cortisol concentrations at predefined time points, such as 08.00 hours, 09.00 hours, etc.,8–10 whereas cortisol increase in the morning is related to time of awakening rather than to a predefined time point, introducing measurement bias due to the variations in time of awakening of the participants.11
Diurnal cortisol levels have been found to be related to a variety of trait characteristics, such as anxiety, depression and social impairments.4,12 Cortisol has also been found to be related to a person's current mood: positive affect was significantly associated with lower salivary cortisol levels13 and negative affect was significantly associated with higher salivary cortisol levels.12,13 Wüst et al.14 have found a relationship between morning cortisol increase and several aspects of perceived chronic stress: worries, social stress and lack of social recognition.
The aim of the present study was to investigate whether young adult male inpatients with schizophrenia demonstrated an increased basal activity of the HPA axis, assessed by means of free cortisol concentrations in saliva, based on multiple samples throughout the day, with measurements starting immediately after awakening. In addition, we investigated whether HPA axis functioning was related to symptom severity and social integration scores.
Twenty-seven male patients and 38 healthy male control subjects were included in the present study. The participant characteristics are presented in Table 1. The patients were recruited from the Unit for Psychotic Disorders at the Department of Psychiatry of Erasmus MC, University Medical Center Rotterdam. All patients were inpatients and most of the patients were experiencing their first psychotic episode or were receiving treatment for the first time. The patients were screened by a senior psychiatrist and included in the study if they were diagnosed with schizophrenia according to the criteria of the DSM-IV.15 Patients with a duration of illness shorter than 6 months were included in the study and the diagnosis was confirmed after six months to comply with the corresponding DSM-IV criteria. Patients were excluded from the study when they suffered from comorbid psychiatric disorders. In order to assess symptom severity, the Dutch translation of the Positive and Negative Syndrome Scale (PANSS16) was used. In order to assess subjective well-being of the patients, the Subjective Well-being on Neuroleptics (SWN17) questionnaire was used. All patients experienced a psychotic episode at the time of the study. The details on medication use are presented in Table 2.
Table 1. Participant characteristics
Patients with schizophrenia
Healthy control subjects
NA, not applicable; PANSS, Positive and Negative Symptoms Score; SWN, Subjective Well-being on Neuroleptics.
Age (years ± SD)
22 ± 5
22 ± 3
Duration of illness (months [range])
Total PANSS score (mean ± SD)
72 ± 17
18 ± 6
17 ± 6
37 ± 11
Total SWN score (mean ± SD)
87 ± 14
16 ± 4
Table 2. Mean dosages and duration of current antipsychotic treatment
Mean dosage mg ± SD
Mean duration week ± SD
NA, not applicable.
Lorazepam (n = 2)
11.3 ± 4.4
3.3 ± 4.0
Lorazepam (n = 5)
2.8 ± 0.8
3.9 ± 2.9
Oxazepam (n = 1) Lorazepam (n = 1)
3.0 ± 1.7
1.7 ± 1.5
Lorazepam (n = 1)
300 ± 10
4.7 ± 3.1
Lorazepam (n = 2)
366.7 ± 208.2
1 ± 0
Controls were recruited by means of advertisements in and around the hospital. All controls were healthy and unmedicated at the time of testing as assessed by means of an interview and a questionnaire. None of the controls or their first-degree relatives had experienced past or current psychiatric illnesses.
Exclusion criteria for both patients and controls were neuroendocrine disorders, neurological disorders, and any use of medication that may influence HPA-functioning.
After the subjects were given a complete description of the study, written informed consent was obtained. The study was approved by the Medical Ethics Committee of Erasmus MC, University Medical Center Rotterdam, and was carried out in accordance with the Declaration of Helsinki.
Saliva was sampled at five different time points during the day: directly after awakening when the subject was still in bed, 30 min thereafter, and at 12.00 hours, 16.00 hours and 22.00 hours. Subjects were asked to chew on cotton swabs until the swab was saturated with their saliva, and then to save the cotton swab in the accompanying plastic vessel.
Subjects were free to use an alarm clock or wake up spontaneously. They were instructed not to brush their teeth immediately before sampling, in order to avoid contamination of the saliva with blood from the oral cavity. Subjects were also instructed to collect their saliva before any meal, in order to avoid contamination of the saliva samples with food and beverages. Control subjects received an envelope containing five plastic vessels, a written instruction and a registration form on which they were to write down the exact time they sampled their saliva. The controls were instructed to save their saliva samples at home in the refrigerator or, if possible, in the freezer, and to send or bring the five samples to the laboratory as soon as possible. Patients were aided by the nursing staff, who offered them the plastic vessels at the correct times during the day and who wrote down the exact time of saliva collection. The collected saliva was stored at –20°C at the clinic. Subsequently, saliva was transported to the laboratory, where all samples were stored at –20°C.
Salivary cortisol was assayed in duplicate by an enzyme-linked immunosorbent assay (ELISA) using a commercial kit (DRG Diagnostics, Marburg, Germany). The lower limit of detection of this method is 1.14 nmol/L at the 95% confidence limit. All samples from an individual participant were analyzed in a single assay run. A comparison between cortisol levels in 75 pairs of serum and saliva samples collected at the same moment yielded a correlation coefficient of 0.71 and a slope of 0.05, indicating that saliva concentrations were 5% of the levels in serum.
The area under the curve (AUC) was calculated according to the formulas provided by Pruessner et al.,18 and represents the AUC with respect to the ground (AUCg) and the AUC with respect to the increase (AUCi). The AUCg represents the total hormonal output, whereas the AUCi represents the changes over time. To calculate the AUC, we used the cortisol concentration sampled directly after awakening as a starting point, and we used the reported sampling times for the calculation of the time periods between samples. The Morning Increase in cortisol was calculated by subtracting the cortisol concentration directly after awakening from the concentration measured 30 min later.19
To evaluate whether the shape of the cortisol day curves differed between patients with schizophrenia and controls, a random intercept model for repeated measures was fitted to the data using MLwiN20 software for multilevel modeling (MLwin, Centre for Multilevel Modelling, Bristol, UK). The model included a random effect for Subjects (i.e. allowing the intercept to vary between subjects), and fixed effects for Group (patients vs controls), Time (awakening, 30 min thereafter, 12.00 hours, 16.00 hours, and 22.00 hours), and the interaction effect of Group × Time. Because of the non-linear shape of the cortisol curve, the relationship with Time was modeled as a third-order function by including the model time, time squared and cubic time. To control for possible confounders, we added the following covariates: Smoking (yes/no), Time of Awakening, Duration of Illness (the duration was set to 0 for the controls), Antipsychotic Medication use (yes/no), and Benzodiazepine use (yes/no). We added smoking in our analyses because nicotine stimulates the activity of the HPA axis and alters the reactivity of the HPA axis to psychosocial stress.21 We included Time of Awakening as a covariate because this has also been found to influence cortisol levels.22 Duration of Illness, Antipsychotic Medication use and Benzodiazepine use were added to control for possible effects of these factors within the patient sample. We also used ancova (using the same covariates as described above) with Group as the between-subjects factor to investigate whether the cortisol levels differed between groups at specific time points and whether the Morning Increase, AUCg and AUCi differed between patients and control subjects.
We performed Pearson's correlations between cortisol measures, PANSS Total, Positive, Negative and General Psychopathology subscales and the SWN scale for Social Integration. We used the Pearson correlation coefficient rather than the non-parametric Spearman correlation coefficient as the scores on the PANSS and SWN were at interval level and were normally distributed (Kolmogorov–Smirnov's Z ranged from 0.388 to 0.998, all P > 0.25 for each of the subscales of the PANSS and the SWN Social Integration scale).
The statistical analyses were performed using spss version 13.0 (spss, Chicago, IL, USA) and MLwiN 2.10.20 All analyses were two-tailed and the alpha was set at 0.05. If the assumption of sphericity was violated, we used the Greenhouse–Geiser correction.
Time of awakening
Time of Awakening did not differ significantly between patients and controls, with patients awakening on average at 07.57 hours, and controls awakening at 08.25 hours (t[53.5] = 1.69, P > 0.05). However, because of its potential importance, we included Time of Awakening in the subsequent analyses as a covariate.
Cortisol day curves
As can be seen in Table 3 and Figure 1 (presenting the estimated marginal means), all polynomials of Time were significant, indicating that the cortisol levels changed over time, showing linear, quadratic and cubic trends. We also found a significant interaction effect of Group by Time (linear), indicating that the cortisol levels showed steeper decreases over time in the patient group than in the control group, and, as can be seen in Figure 1, this decrease was most prominent in the morning. No significant main effect for Group was found; overall, cortisol levels did not differ significantly between groups. In addition, none of the covariates had significant effects on the cortisol day curves: the estimates for Smoking, Time of Awakening, Duration of Illness, Antipsychotic Medication use and Benzodiazepine use were all non-significant.
Table 3. Beta coefficients of the model fitted for the cortisol day curve
CI, confidence interval.
Group * Time^1
Group * Time^2
Group * Time^3
Duration of disease
When we investigated whether the groups differed in their cortisol levels at any of the specific time points, only the cortisol level at 12.00 hours was significantly lower in patients than in controls (F[1,57] = 4.92, P < 0.05, partial η2 = 0.08).
Morning increase and AUC
The Morning Increase and AUC were investigated using an ancova in which Smoking, Time of Awakening, Antipsychotic Medication use, Duration of Illness and Benzodiazepine use were entered as covariates.
No significant difference between groups was found when we investigated the Morning Increase (F[1,57] = 1.88, P = 0.18). The mean Morning Increase ± SD of the patients was 2.21 ± 9.65 nmol/L and that of the healthy controls was 3.59 ± 8.00 nmol/L. Of the covariates, only Time of Awakening (F[1,57] = 4.21, P < 0.05, partial η2 = 0.07), and Medication use (F[1,57] = 4.34, P < 0.05, partial η2 = 0.07) were significant. Also, patients with schizophrenia and healthy controls did not differ in their AUCg (F[1,55] = 0.76, P = 0.39), and none of the covariates were significant. The mean AUCg ± SD of the patients was 183.70 ± 49.85 nmol/L and that of the healthy controls was 177.51 ± 56.69 nmol/L. However, patients did show a significantly smaller (i.e. more negative) AUCi (F[1,55] = 4.80, P < 0.05; partial η2 = 0.08), indicating that the cortisol concentrations decreased more during the day in patients with schizophrenia relative to controls. Only the covariate Medication use showed a non-significant trend (F[1,55] = 3.49, P = 0.07, partial η2 = 0.06). After having established that there were no outliers included in the data, the mean AUCi ± SD of the patients was −72.46 ± 108.57 nmol/L and that of the healthy controls was −13.11 ± 118.44 nmol/L.
Clinical parameters and cortisol
In patients, both the Morning Increase and the AUCi showed significant negative correlations with the scores on the PANSS subscale for negative symptoms (r = −0.42 and r = −0.45, respectively, both P < 0.05). No other significant correlations between PANSS scores or the SWN Social Integration score and cortisol measures were found.
The aim of the present study was to evaluate whether the functioning of the HPA axis in young male patients with schizophrenia differed from the functioning of the HPA axis in healthy control subjects. Overall, the patients and controls did not differ in their absolute levels of cortisol at any specific time point. However, as shown by the differences in AUCi, the patients with schizophrenia showed steeper decreases in cortisol concentration during the day relative to controls. The AUCi is an indicator of changes over time, i.e. the sensitivity of the HPA axis, whereas absolute cortisol levels are more indicative of the basal HPA axis functioning, i.e. total hormonal output.18 The steeper decrease in cortisol levels in the patient group suggests that the sensitivity of the HPA axis in patients is increased, resulting in a system that may be more reactive to stressors. This finding is in line with the neural diathesis stress model of Walker and Diforio,2 who assume that exposure to psychosocial stressors exacerbates the symptoms of schizophrenia and that vulnerability to schizophrenia is associated with heightened behavioral responsivity to stressors.
Additionally, we found that the PANSS subscale for negative symptoms correlated negatively with the Morning Increase and the AUCi in the patient group. As stated above, the AUCi and Morning Increase are indicators of the sensitivity of the HPA axis.18 As four of the seven items on the PANSS subscale for negative symptoms concern affective flattening and social and emotional withdrawal, it appears that the HPA axis of patients with more negative symptoms was less affected by the emotional and social symptoms (social stress) associated with the disease than patients with fewer negative symptoms. Most previous studies have linked increased morning cortisol responses with perceived stress and negative affect, but contradicting results have also been found (see for a review Clow et al.19). Our findings are consistent with the findings of Wüst et al.,14 who suggested that morning cortisol is related to aspects of social integration and behavior: they found a positive relationship between the Morning Increase and social stress. Studies investigating baseline cortisol failed to find significant relationships between HPA axis functioning and schizophrenia symptoms,23,24 whereas studies using the dexamethasone suppression test have found relationships between hypercortisolemia and the severity of negative and cognitive symptoms.25,26
We did not find overall higher cortisol concentrations during the day in patients with schizophrenia compared with controls, suggesting that the basal HPA axis functioning in patients with schizophrenia was not different. Some previous studies investigating cortisol concentrations at predefined time points in unmedicated, relatively young patients with schizophrenia have found overall increased cortisol concentrations during the day relative to controls,8,10,27,28 whereas others found normal diurnal cortisol concentrations in patients with schizophrenia.29,30 Most of our patients used antipsychotic medication. Studies investigating cortisol concentrations at predefined time points in medicated patients have found mixed results. Patients using first-generation antipsychotics showed increased cortisol levels relative to controls in one study,9 but not in another,31 and patients using second generation antipsychotics did not show significantly different cortisol day curves relative to healthy control subjects.9,32 Most of the medicated patients in our study used second-generation antipsychotics, and, consistent with the studies mentioned above,9,32 we did not find overall higher cortisol concentrations during the day in patients with schizophrenia compared with healthy subjects. Furthermore, antipsychotic medication use did not affect the diurnal cortisol levels significantly. We also controlled for smoking, time of awakening, use of benzodiazepines, and duration of illness, but none of these factors explained the variation in cortisol levels significantly.
Our results may be limited due to a few methodological issues. First, we did not ask the participants to refrain from smoking during the sampling day because patients would not cooperate if we did. It should be noted, however, that we did control for smoking behavior in our statistical analyses. Another limitation was the fact that the patients with schizophrenia were hospitalized, whereas the healthy controls were not. It is possible that the differences we found between groups were the result of hospitalization rather than of schizophrenia itself. We only included male participants in our study, which limits the potential to generalize the findings of the study. Also, some of the patients did not use antipsychotic medication, whereas others used either typical or atypical medication. It was not possible to investigate all patients under medication-free conditions due to their symptom severity. Another issue was participant adherence: it was unclear to what extent the participants were compliant to our saliva sampling instructions, which is an important factor when investigating saliva samples.33 However, we did stress the importance of correct sampling during the oral and written instructions, and we asked participants to write down the exact time at which each saliva sample was taken, thereby controlling for possible deviations from our instructions. The patients were assisted by the nursing staff during their saliva sampling. For future research, a solution to this issue would be the use of digital sampling devices that record the exact time of sampling.
Concluding, we found that patients with schizophrenia only differed from healthy control subjects regarding the sensitivity of their HPA axis, and not in their basal HPA axis functioning. The Morning Increase and AUCi were negatively associated with the severity of the negative symptoms in schizophrenia, possibly due to the lack of (social) stress these patients experience.