Cortisol response to psychological stress in patients with schizophrenia in the Chinese population

Accumulating evidence suggests that hypothalamic-pituitary-adrenal (HPA) axis dysfunction might play an important role in the pathophysiology of schizophrenia. This study aimed to explore the cortisol response to psychological stress in patients with schizophrenia. In this study, patients with schizophrenia (n = 104) and healthy volunteers (n = 59) were asked to complete psychological stress challenge tasks, which included the Paced Auditory Serial Addition Task and Mirror-Tracing Persistence Task, and then, saliva was collected to measure cortisol levels. Emotions and psychopathology were assessed by the Positive and Negative Affect Schedule (PANAS) and Positive and Negative Syndrome Scale (PANSS). The results showed (1) that the cortisol stress response and negative emotions in patients with schizophrenia differed signicantly from those in healthy controls; (2) there were signicant interactions between sampling time and diagnosis for saliva cortisol levels; (3) there were signicant interactions between scoring time and diagnosis for the negative affect score of the PANAS; and (4) the changes in salivary cortisol levels and negative affect scores before and after the psychological stress challenge tasks were not correlated with clinical symptoms in patients with schizophrenia. These ndings indicated that an abnormal cortisol stress response might be a stable biological characteristic of schizophrenia.


Introduction
Schizophrenia is a severe neuropsychiatric disorder for which the etiology remains unclear. However, the interaction of multiple biological and environmental factors plays an important role in the pathophysiology of this illness. Several models of stress-vulnerability have been proposed for schizophrenia [1][2][3][4] to better understand the relationships between vulnerability, stress, and disease. The hypothalamic-pituitary-adrenal (HPA) axis is one of the primary systems that moderate the physiological response to psychological and physiological stressors [5] , and cortisol secretion levels can re ect the function of the HPA axis.
Cortisol secretion is indexed using morning cortisol levels, diurnal cortisol levels, cortisol awakening response (CAR), and cortisol stress response. Morning cortisol levels re ect basal HPA axis function, diurnal cortisol levels re ect the uctuation of the HPA axis throughout the day [6] , and CAR and cortisol stress response re ect the mobilization capability and response to stress of the HPA axis, as awakening can be considered a stressor [7,8] . The effective cortisol response to stress is marked by a rapid rise in cortisol levels right after the introduction of the stressor, followed by a rapid decline. This physiological process plays a key role in the body's ability to adapt to the environment and deal with threats [9] . Cortisol stress response is an important indicator used to evaluate HPA axis function.
The association between cortisol and schizophrenia has been widely recognized [9] . In general, patients with schizophrenia have abnormally elevated cortisol levels, but the results of current studies are inconsistent [9][10][11][12] . A recent meta-analysis pointed out that aberrant morning cortisol levels measured peripherally (blood or saliva) were observed more frequently in patients with schizophrenia than in controls; however, they might have been in uenced by factors such as illness stage, mood-state, and psychotropic medication [10] . Studies have also noted a blunted CAR in patients with schizophrenia [13][14][15] . Although antipsychotic treatment can alleviate the effects of abnormal diurnal cortisol levels [4,16] , the blunted CAR persists [17] . These results suggest that changes in cortisol levels might be a more stable indicator.
In addition to CAR, another indicator of changes in cortisol levels is cortisol stress response. The cortisol response to psychosocial stress is reportedly observed but blunted in patients with schizophrenia when compared to that in controls [9] . Girshkin et al. reported patients with schizophrenia did not show any cortisol increase in response to magnetic resonance imaging (MRI) as a stressor [12] , unlike healthy individuals [18] . This differed from the cortisol stress response shown among patients with schizophrenia in most studies using laboratory-based psychological stress challenge tasks [9,11] . One possible explanation for this was that the stress during an MRI mainly involved passive perception of noise and did not require the patient to participate in the execution; it was merely an uncomfortable environment, and a lack of concern about the environment is a manifestation of apathy in patients with schizophrenia.
Other studies also found heightened cortisol secretion or attened CAR among individuals who could not be diagnosed with schizophrenia but who met the clinical high-risk (CHR) criteria for psychosis when compared to that in healthy controls [7,[19][20][21][22] . Similar results have been reported among relatives of patients with schizophrenia [21][22][23] . These ndings suggest abnormal cortisol secretion predates the onset of psychiatric symptoms and may be genetic. Cortisol-related abnormalities and the HPA axis dysfunction they represent may serve as biomarkers for predicting the occurrence and development of schizophrenia.
In this study, we aimed to investigate the difference in cortisol response to psychological stress between individuals with schizophrenia and healthy controls in the Chinese population and to test the hypothesis that patients with schizophrenia had an abnormal cortisol response to stress, which represents dysregulated HPA function, and the HPA axis dysfunction has the potential to serve as a biomarker for schizophrenia. Salivary cortisol secretion is a reliable indicator of free cortisol in plasma and has been used as an index for HPA axis activity in many studies [24] , thus we noninvasively examined cortisol levels by testing saliva collected before and after a laboratory-based psychological stress challenge task.

Participants
Patients were recruited from inpatients or outpatients at the Beijing HuiLongGuan Hospital from 2017 to 2019. The inclusion criteria for patients were as follows: (1) diagnoses con rmed according to the Patients were excluded if they had: (1) evidence of a signi cant physical illness or neurological condition; (2) substance dependence or current substance abuse (except nicotine) at the time of screening; (3) a psychiatric disorder other than schizophrenia according to DSM-IV criteria; (4) any endocrine disorder; or (5) received any immunotherapies.
Age-and sex-matched healthy volunteers were recruited from the local community. Healthy volunteers were excluded if they had (1) evidence of a signi cant physical illness or neurological condition; (2) current and ongoing substance abuse (except nicotine) at the time of screening; (3) any psychiatric disorder according to the DSM-IV criteria; (4) any endocrine disorder; or (5) received any immunotherapies.

Psychological stress challenge task
We used a psychological stress challenge task to assess the capability of participants to persist in goaldirected behavior while under stress. The task consisted of two tests each lasts for 7 minutes: the Paced Auditory Serial Addition Task (PASAT) and the Mirror-Tracing Persistence Task (MTPT), both of which have been applied in psychiatric research [25] . The task is monotonous, and there is a danger of being punished. This makes the task somewhat di cult, so some participants may choose to drop out. During the PASAT, multiple numbers are displayed in sequence on a computer screen; the participant is asked to sum the numbers that appear and then use the mouse to select the correct answer. When the participant answers incorrectly, the computer plays annoying sounds. The computer program adjusts the number display speed based on the participant's performance to ensure that the test is challenging. During the MTPT, participants are asked to use the mouse to trace the outline of a star on the computer screen.
However, the moving direction of the cursor on the screen is opposite to that of the mouse in the participant's hand, creating a challenge. If the cursor leaves the track or stands still, it is judged as an error, causing the computer to play annoying noises and reset the drawing process. The computer adjusts the width of the star according to the participant's performance to ensure the task is challenging for everyone.
Since these two tasks involve two different skills (computing and hand-eye coordination), their combination helps avoid biases that may be caused by individual expertise. The two tasks were performed sequentially and in random order. The whole process was completed in the form of humancomputer dialogue, thus avoiding the subjective in uence of researchers.
Participants could opt-out at any time during the task. They were told in advance that they would be compensated for participating. Participants received 40 Yuan for completing both tasks. If they did not complete the tasks, the compensation was halved.
We determined distress tolerance when a participant persisted in completing either of the tasks; we determined distress intolerance when a participant failed to complete both tasks.

Saliva collection and processing
To reduce the effects of changes in cortisol over time, all tasks and saliva sample collection were held between 12:00 and 4:00pm. We collected three saliva samples from each participant: before starting the tasks and after resting for 20 and 40 minutes following the tasks. To reduce interference factors, all participants were asked not to do strenuous exercise, eat, drink, smoke, or brush their teeth for 1 hour before the test, and then to gargle 10 minutes before the test.
Saliva samples were collected using a Salivette device (Sarstedt AG & Co Kommanditgesellschaft SarstedtstraBe 1 51588 Nümbrecht Germany). Participants were asked to chew the swabs thoroughly for 2 minutes at each sample collection time to make sure more than 1 mL of saliva was collected. The saliva sample was then frozen at -20°C for 24 hours. After thawing the frozen saliva, it was centrifuged at 4°C for 10 minutes (4,000 rpm/min) to obtain more than 0.5 mL saliva from each sample. Then, the processed saliva sample was placed into a centrifuge tube; the internal standard solution was added, followed by the precipitant; and the tube was vortexed and centrifuged for 10 minutes. The supernatant was then taken out and blown dry with nitrogen. Derivatization reagent was added to the dried sample, and this was placed in the oven for derivatization. After derivatization, the supernatant was used for liquid chromatography and tandem mass spectrometry using standard protocols. The intra-trial coe cient of variations for high-, medium-, and low-quality of salivary cortisol were 8.8%, 5.6%, and 10.1%, respectively.

Clinical assessment
All participants completed the Positive and Negative Affect Schedule (PANAS) [26] before the start of the rst task, after completing (or aborting) the rst task, and after completing (or aborting) the second task. This scale contains ten positive affect and ten negative affect (NA) items. Considering the stress response mainly involves negative emotions, we only used the NA scores.
A total of 176 participants, including 112 patients with schizophrenia and 64 healthy volunteers, completed the psychological stress challenge task, PANAS, and saliva collection. Psychopathology was assessed by an attending psychiatrist using the Positive and Negative Syndrome Scale (PANSS) [27] . The process is shown in Figure 1.

Statistical Analyses
The saliva cortisol data collected in the patient and control groups were analyzed. Data that deviated from the 25%-75% range for each set in each group of data by more than three times the interquartile range (IQR) of the set were de ned as extreme values and excluded from the nal analysis. Finally, the data of 104 patients and 59 controls were included.
Salivary cortisol data and PANAS score data are presented as the mean and standard deviation. A oneway analysis of variance (ANOVA) was used to examine the impacts of each factor on variables, and a repeated-measures ANOVA was used to examine the interaction between multiple factors, such as diagnosis (patients or controls), sampling time (before, and 20 minutes and 40 minutes after testing), and distress intolerance (yes or no) on variables. The chi-squared test was used to assess differences in categorical variables between groups. Spearman's rank correlation coe cient was also used to examine the correlation between variables. Nonparametric tests were used to examine group differences for dependent variables that did not exhibit normal distributions. The signi cance level was set at 5% (p < 0.05

Participant characteristics
In terms of age, sex, smoking status, education level, and PASAT error rate, there were no signi cant differences between the patients with schizophrenia and the healthy controls. The error rate of MTPT in the patients with schizophrenia was signi cantly higher than that of the controls. In terms of stress tolerance, there was a signi cant difference between the patients with schizophrenia and the controls ( Table 1). For ease of description, we assigned subscripts to the variables of salivary cortisol data collected at different times: 1 represented data before the task, 2 represented data 20 minutes after the task, and 3 represented data 40 minutes after the task.
In the patient group, saliva cortisol levels rst increased and then decreased. In the control group, saliva cortisol levels were highest before the task began, and then gradually decreased (Figure 2A).
There was a signi cant interaction between sampling time and diagnosis (F = 6.961, p = 0.002). Using the NA score before the task as covariates, the above results remained valid. To exclude the in uence from the interaction, we used one-way ANOVA to examine the impacts of sampling time and diagnosis on the salivary cortisol levels of the two groups separately. Before the task, the saliva cortisol level of the control group tended to be higher than that of the patient group ( This indicated the stress response of cortisol during the psychological stress challenge task test was blunt among patients with schizophrenia and more active among the healthy controls.

NA scores
For ease of description, we assigned subscripts to the variables of NA scores evaluated at different times: 1 represented data before the task, 2 represented data after one task, and 3 represented data after both tasks.
There was a signi cant interaction between scoring time and diagnosis (F = 14.744, p < 0.001). To exclude the in uence from the interaction, we used one-way ANOVA to examine the impacts of scoring time and diagnosis on the NA scores of the two groups separately. The NA scores of the patients with schizophrenia were signi cantly higher before the task (F 1 = 18.442, P 1 < 0.001) than those of the controls. There was no signi cant difference between patients and controls in the NA scores after completing one task (F 2 = 0.006, p 2 = 0.937) and the NA scores after completing two tasks (F 3 = 0.066, p 3 = 0.798). The main effect of scoring time in the patient group was not signi cant (F = 0.307, p = 0.712), while the main effect of scoring time in the control group was signi cant (F = 16.168, p < 0.001). These results indicated that patients with schizophrenia had more negative emotions before the test, while the psychological stress challenge task test had no signi cant effect on this experience. The controls had fewer negative emotions before the test, and their negative emotions increased during the test ( Figure   2B).

Distress tolerance and distress intolerance
In the control group, most subjects persisted in completing the task, and only one subject chose to drop out (showed distress intolerance). The rate of distress intolerance in the schizophrenia group was signi cantly higher ( Figure 3A).
As there was only one participant who showed distress intolerance in the control group, further grouping and comparison were not necessary. Therefore, we only divided the patients with schizophrenia into subgroups based on stress tolerance and made comparisons among the subgroups.
Although the trends of cortisol-level change differed between the subgroups, there was no signi cant interaction between sampling time and tolerance (F = 1.985, p = 0.145) and the main effects of sampling time or tolerance were not signi cant (F = 2.991, p = 0.058 and F = 0.017, p = 0.897, respectively) ( Figure  3B).
In the NA score, there was a signi cant interaction between scoring time and tolerance (F = 4.338, p = 0.019), but the main effects of scoring time (F = 2.604, p = 0.083) and tolerance (F = 0.180, p = 0.672) were not signi cant ( Figure 3C).

Clinical symptoms
The PANSS positive subscale scores, general psychopathological subscale scales, and total scores differed signi cantly between patients who showed distress tolerance and those who did not. A higher score for the distress-intolerance group indicated that these patients' symptoms may be more severe ( Figure 3D).
We performed a correlation analysis of the total PANSS scores, positive subscale scores, negative subscale scores, general psychopathological subscale scales, and salivary cortisol level of the patients with schizophrenia. Most of the data showed no signi cant correlations. The saliva cortisol level before the task was correlated with negative subscale scores (r = 0.233, p = 0.017). The cortisol level 20 minutes after the task was correlated with PANSS total scores (r = 0.198, p = 0.044) and negative subscale scores (r = 0.277, p = 0.004). However, mathematically speaking, the correlation is generally considered meaningful only if the correlation coe cient is greater than 0.3; the above correlation coe cients were all too low, indicating no practical signi cance. There was no correlation between patients' salivary cortisol level 40 minutes after the task and the PANSS total scores, negative subscale scores, positive subscale scores, or general psychopathological subscale scales. None of the PANSS scales were correlated with the patients' NA scores.

Rigor
We reran the analyses including the data of the 13 participants with extreme values that were previously excluded, and this did not change the outcome.

Cortisol and schizophrenia
Changes in saliva cortisol level before and after stress represent the function of the cortisol stress response and HPA axis. The previous stress-vulnerability model theoretically proposed the relationship between HPA axis function and schizophrenia [1][2][3][4] , which has been veri ed in several studies [9] . In this study, the existence of time and diagnosis interactions showed that the trends in the cortisol level change of the two groups differed greatly as the psychological stress challenge task progressed. The patients' cortisol stress response was not obvious: although it tended to rst rise and then fall, the change was not signi cant (p > 0.05). The overall response was attened, which was similar to the blunted CAR previously reported in patients with schizophrenia [13][14][15]17] . The change trend of salivary cortisol levels in the control group differed greatly from that of the patient group; the salivary cortisol level was clearly high before the task began, and then gradually decreased. The difference in cortisol stress response between patients with schizophrenia and healthy controls was obvious, con rming the hypothesis that patients with schizophrenia had HPA axis dysfunction.

Cortisol, stress, and distress intolerance
We chose to measure cortisol stress response, as it better re ects the pathological process than a baseline cortisol level [9] .
Prior to the start of the psychological stress challenge task, the baseline cortisol level increased due to the effects of anticipation anxiety. Although our study did not take samples at the same time another day for comparison, a previous study performed this comparison, con rming the existence of the expected anxiety [25] . We found pre-task cortisol levels were signi cantly higher than cortisol levels after the task in the controls, which was inconsistent with the ndings of previous studies in an American population [11] . Since we controlled the same initial experimental conditions for all participant, such as experimenting at the same time in the afternoon, avoid strenuous exercise, eating, drinking, smoking, brushing teeth, and other actions that may affect cortisol test values before the experiment, we believe that cortisol levels represent stress levels, so a high level of cortisol suggests that healthy controls already had signi cant sensations of stress before the task began. Considering that previous studies were conducted in the Western parts of the world, some differences may have been due to our participants being from a Han Chinese population. Many studies revealed racial differences in cortisol levels [28][29][30][31] , one study also found that people of different ethnicity have different responses to psychological support under pressure [32] . All these suggest that cortisol and stress-related research may have ethnic differences. But there are no speci c conclusions for racial differences in the effects of expected anxiety on cortisol levels yet. This point warrants further study.
The in uence of culturally related psychological factors is another possible reason for the differences. Studies have shown that cultural differences between the East and the West affect people's attitudes toward examinations. In some cultural atmospheres that emphasize examinations, test anxiety is more obvious and common [33,34] ; China has such a cultural atmosphere, and compared with participants in previous Western studies [11,25] , our subjects had a lower task withdrawal rate, which may have also re ected this cultural atmosphere. After learning they would take a test, the healthy controls might have had higher expectations for their performance, which could also have caused severe anticipation anxiety. However, the patients felt less stressed before the task, which is consistent with the apathy and decreased self-expectation caused by schizophrenia.
Cortisol levels in both groups showed a downward trend after the 40-minute rest following the task, which is a normal part of stress response. However, the patients' cortisol level decreased at a much slower rate.
The cortisol level of the patient group before the task was signi cantly lower than that of the controls, while the cortisol level 40 minutes after the end of the task tended to be higher than that of the controls, indicating the response of the patients' HPA axis was insensitive or delayed, or that the negative feedback mechanism was damaged and unable to escape from stress. This nding mirrors earlier studies, where some patients showed a signi cant delay in cortisol stress response [11] .
Among the patients, the trends of changes in cortisol levels in distress tolerance and distress-intolerance patients were different. Among patients with distress intolerance, there was an upward trend in cortisol levels during the task, and then they dropped more sharply after the task. This may indicate that these patients felt that the pressure during the task greatly increased and exceeded their tolerance threshold, leading them to opt-out. When they decided to give up, most of the stress disappeared, and cortisol levels dropped rapidly. However, patients with distress tolerance may have been insensitive to the stress caused by the task, leading to the indifference of the process for completing the task and a smoother curve in the change of their cortisol levels. As this study was limited by a small sample size of patients with distress tolerance, our conclusions warrant further veri cation.

Clinical symptoms and other indicators
We found that cortisol stress response was not signi cantly correlated to patient symptoms, nor was it correlated with negative emotion scores before the task. Education level and smoking status had no signi cant effect on cortisol stress response, which is consistent with previous research [11,12] .
Although there are few studies on cortisol stress response in patients with schizophrenia, there are some studies on other cortisol indicators, and the results are inconsistent. Most of them found no signi cant correlation between cortisol and symptoms [35][36][37] , with similar ndings for CHR patients [7,19] . However, a study of treatment-resistant patients with schizophrenia found that patients' serum cortisol levels were signi cantly reduced in the morning after treatment, and, although the correlation was weak, it was related to improvement in PANSS negative symptoms [38] . One study also found that, in chronic schizophrenia, cortisol levels were weakly correlated with the PANSS negative-symptom subscale [39] . Another study found that, after treatment, the increase in cortisol during the day in patients with schizophrenia was relieved, but the blunted CAR did not change [17] . These differences may have been due to varying research methods. Previous studies showed that symptoms may be related to cortisol levels within a certain period, re ecting the daily function of the HPA axis, while cortisol stress response represents the mobilization and coping ability of the HPA axis under stress (which has something in common with the CAR). This may indicate that HPA axis mobilization and abnormal coping ability may be a stable biological indicator that can indicate the disease, will not uctuate with the severity of symptoms, and will be less affected by other factors. Therefore, it has clinical predictive signi cance and may even become an endophenotype.
We found that the patients with distress intolerance had more severe clinical symptoms, which also indicates that distress intolerance may represent a state that is more deviated from the healthy state, as distress intolerance in the healthy control group was very rare. However, whether distress intolerance is the result of symptoms or one of the factors that cause obvious symptoms remains to be studied.

Negative emotions
The difference in NA scores between patients and controls is understandable. The psychological stress challenge task is challenging, requires concentration and patience, and plays annoying stimuli in response to a mistake. This process is obviously unpleasant. The NA score of the controls increased as the task progressed, as expected. However, patients with schizophrenia were already experiencing negative emotions before starting the task. The feeling remained almost unchanged during the task, which indicated the patients with schizophrenia were less satis ed with their daily life or less satis ed with their participation in the test, relatively speaking, the content of the task was not their focus.

Strengths and limitations
To our knowledge, no large-sample study has previously examined the correlation between cortisol stress response and schizophrenia among the Chinese population. Possible racial differences make our research more meaningful. We collected data of more than 160 cases, providing a sample larger than those of previous studies. However, an even larger sample is needed to establish cortisol stress response as a biomarker for schizophrenia.
Cortisol levels may also be affected by other factors, such as alcohol intake or body weight. These factors were not considered in our study. Because the main focus of our study was the change of cortisol levels under stress, there was a self-control effect, and these factors would have had a limited impact on the interpretation of the results. If complete information could be obtained from the subjects, it would help improve the rigor of our conclusions. Cross-sectional studies cannot rule out differences in individual sensitivity to cortisol. Although there were self-control effects and a relatively large sample size, which can minimize the effect of sensitivity differences on the results, more information on sensitivity is needed.
Data to date have been obtained from cross-sectional studies, but disease and treatment conditions may progress or change. Therefore, long-term follow-up data are required to con rm the stability of cortisol stress response in individual treatment.

Conclusions
Schizophrenia is a major focus of psychiatry research, and the identi cation of biological indicators is important for clinical diagnosis. As a potential disease diagnostic biomarker, cortisol is convenient for sampling and detection, non-invasive, and inexpensive to measure. The ability to use it alone or in combination to predict the occurrence and development of the disease is an exciting prospect. Figure 1 Page 17/17

Figures
Outline of the experimental test Analysis related to distress tolerance (*p < 0.05)