Associations between hypothalamic–pituitary–adrenal axis function and facial emotion processing in depressed and control participants



Aim:  The current study examined the relation between facial emotion processing accuracy and an aspect of hypothalamic–pituitary–adrenal axis function in 64 inpatients with major depression and 49 healthy controls over a 2-week period.

Methods:  The Dexamethasone Suppression Test and a Facial Expression Recognition Task were completed at baseline and 10–14 days after baseline. Treatment response was determined 6 weeks after baseline by change in the Montgomery–Asberg Depression Rating Scale.

Results:  Increased cortisol response to dexamethasone was significantly correlated with reduced ability to recognize facial expressions of anger, sadness and disgust within the total sample, but these correlations did not remain significant at 10–14 days. Surprisingly, cortisol response to dexamethasone was comparable in acutely depressed inpatients and healthy controls, and did not change over time in relation to treatment response.

Conclusion:  The study findings provide preliminary evidence that hypothalamic–pituitary–adrenal axis functioning and processing threat-related facial expressions are related, perhaps through involvement of the amygdala.

THE HYPOTHALAMIC–PITUITARY–ADRENAL (HPA) axis is a complex system which is activated in response to physical and/or psychological stress. It is sometimes referred to as the limbic HPA axis because of the crucial role the limbic system plays in the perception of various forms of stress.1 A key structure in the limbic system is the amygdala. After activation in response to threatening sensory information, the amygdala projects threat-signaling impulses to the hypothalamus; this, in turn, activates the HPA axis.

Neuroimaging studies have provided evidence of the amygdala's role in conscious and non-conscious processing of fearful expressions.2 Debate continues as to whether the amygdala is involved in the perception of fear specifically, or whether it contributes to perception of a variety of emotions. Studies suggest that the amygdala may respond to emotionally salient information, regardless of emotional valence or type of task, but that it may have a special role in the accurate identification of facial expressions that convey imminent threat to survival.3

Although strong evidence exists for the important role of the amygdala in the HPA axis stress response, and in processing emotional material, there is much less research examining the relation between HPA axis activation and emotion processing. Studies in healthy controls have found release of glucocorticoids following stress induction to influence selective attention, threat-related information processing, and the consolidation of memory of emotionally arousing events.4 These findings indicate a close relation between processing emotional information and the HPA axis stress response. van Marle, Hermans, Qin and Fernandez5 examined facial emotion processing in healthy female subjects (n = 27) after successful induction of psychological stress through movie clips and found increased amygdala sensitivity to emotional facial expressions, but equal specificity to fearful and happy faces. That is, amygdala function shifted towards heightened, but indiscriminate, sensitivity to emotional facial expressions under stress.

Abnormalities in HPA axis function and facial emotion processing are features of major depression. HPA axis dysfunction manifests as excess cortisol secretion and/or abnormalities of HPA axis feedback, the latter of which is typically measured using the dexamethasone suppression test (DST) or the combined dexamethasone/corticotrophin releasing hormone (dex/CRH) test.1 Most consistent evidence of HPA axis dysfunction in major depression comes from studies examining more severe subtypes, such as inpatient depression.6 Individuals with depression also show dysfunctional facial emotion processing. Relatively consistent evidence has been found for negative interpretation biases when processing neutral faces and increased attention towards negative facial expressions and away from positive expressions.7 Le Masurier, Cowen and Harmer8 reported that unaffected first-degree relatives of individuals with major depression show increased waking salivary cortisol levels on working days and significantly faster responses to facial expressions of fear, compared with participants with no family history of major depression. However, correlations between these two variables were not conducted in their study.

The current study examined facial emotion processing ability and an aspect of HPA axis function, response to dexamethasone (DST), in healthy volunteers and individuals with major depression. Interrelations between HPA axis function and facial emotion processing performance were also conducted, thereby extending previous findings. As well as directly comparing the two groups, healthy volunteers and individuals with major depression were included to maximize the chances of obtaining a wide range of DST and facial emotion processing scores. Studies of facial emotion processing following manipulation of the HPA axis have tended to use paradigms that examine biases in facial emotion processing, or vigilance/selective attention towards certain emotions. The current study assessed accuracy in recognizing facial expressions of emotion using the Facial Expression Recognition Task,9 and thus, adds new data to the area. Our findings comparing performance of depressed and healthy participants on the Facial Expression Recognition Task have been reported previously,10,11 while our DST findings and their relation to facial emotion processing have not.



Sixty-four consecutively admitted inpatients gave informed consent. They were aged between 18 and 60 years, with a DSM-IV12 diagnosis of major depressive episode (unipolar or bipolar) and were recruited from the Acute Inpatient Unit at Hillmorton Hospital, Christchurch, New Zealand. This unit is a public inpatient service within the Canterbury District Health Board which provides intensive psychiatric care of adults aged between 18 and 65 years with acute psychiatric needs. Reasons for exclusion were current serious alcohol abuse or dependence, comorbid endocrine, neurological or chronic medical conditions, pregnancy, previous serious head injury, electroconvulsive therapy in the past 12 months, or taking medications likely to interfere with neuroendocrine function, such as β-blockers or steroids. Patients were treated as deemed appropriate by the treating psychiatrists at Hillmorton Hospital. Type of antidepressant treatment and dosage was adjusted according to clinical improvement, and in some cases, plasma levels.

The control group consisted of 49 psychologically healthy individuals without a personal history, or a history in a first-degree relative, of major mental illness, including major depression. Personal current or past psychiatric conditions were screened using the Mini International Neuropsychiatric Interview.13 Controls were excluded from the study for the same reasons as depressed patients. Depressed patients and healthy controls were matched for age, sex, premorbid verbal IQ (National Adult Reading Test14) and years of secondary and tertiary education. Women were matched for phase of menstrual cycle. All participants were fluent in English. The study was approved by the National Health and Disability Ethics Committee.

Study design

Depressed patients and healthy controls completed neuroendocrine assessment (DST) at two time-points: baseline (for patients, within 5 days of admission to hospital) and 10–14 days after baseline. The Facial Expression Recognition Task was administered to both groups at these same time-points. The Montgomery–Asberg Depression Rating Scale (MADRS)15 measured depression severity in the patient group and was completed at three time-points; baseline, 10–14 days after baseline, and 6 weeks after baseline. The Structured Clinical Interview for DSM-IV Axis I Disorders16 assessed the presence of all Axis I psychiatric conditions at 6 weeks in patients. All participants undertook a 90-min neuropsychological assessment at baseline and 10–14 days after baseline. These data are published elsewhere.10,11

Neuroendocrine assessment

The procedure for the DST has been described elsewhere.17 Briefly, participants received an oral dose of 1 mg dexamethasone (Douglas Pharmaceuticals, Auckland, New Zealand) at 23.00 hours. Depressed patients received their dose of dexamethasone in hospital, and controls took dexamethasone at home. The following day, two blood samples were collected in ethylenediaminetetraacetic acid at 08.00 hours. Following blood collection, samples were centrifuged at 3000 revolutions per min for 10 min to prepare plasma, which was immediately frozen at −20°C until assayed. Cortisol levels were measured using enzyme linked immunosorbent assay (ELISA) and dexamethasone levels were analyzed using liquid chromatography-mass spectrometry (LC-MS/MS) method.

Facial emotion processing assessment

All participants performed a modified version of the Facial Expression Recognition Task, developed by Harmer and colleagues.9 Assessment took place between 11.00 hours and 15.00 hours on the day of testing. Participants were presented with individual faces displaying varying intensities (50–100% full emotion) of five basic emotions (angry, happy, sad, fearful and disgusted), and neutral expressions. Recognition accuracy, reaction time and neutral misinterpretation bias (the percentage of neutral expressions misclassified as an emotion) were recorded (for a more detailed description, see10).

Missing data

Nine patients did not complete the second DST or Facial Expression Recognition Task because of: commencement of electroconvulsive therapy (n = 3); being unable to be contacted (n = 3); withdrawal of consent (n = 2); or being hospitalized with a serious physical injury (n = 1).

Statistical analysis

Statistical analyses were performed using pasw (Predictive Analytic Software) Statistics 18 (spss, Chicago, IL, USA). Demographic data were assessed using chi-squared tests or anova, with group (depressed or control) as the between-subjects factor.

The relation between categorical variables (such as group) and cortisol levels (post-dexamethasone) at baseline was examined using univariate anova. Analysis of changes in post-dexamethasone cortisol levels over time used repeated measures anova. In both analyses, dexamethasone levels and age were included as covariates, and sex and smoking status were between-subjects factors, as these factors can influence HPA axis function.18

Repeated measures anova was the primary analysis for examining associations between facial expression recognition performance (within-subjects factor), group (between-subjects factor) and post-dexamethasone cortisol levels (covariate). If interactions between emotion and cortisol levels were significant, and if there was no interaction with group, Pearson's correlations were conducted for each of the six facial expressions and cortisol levels within the total group (depressed and control).

The proportion of DST suppressors and non-suppressors in the depressed and control groups was compared with Fisher's exact test (two-tailed), using a cut-off value of 110 nmol/L.19


Demographic and clinical characteristics

Depressed and control groups did not differ in any demographic characteristics (see Table 1). None of the depressed or control participants had consumed alcohol or marijuana in the 24 h prior to assessment. The proportion of smokers differed significantly between groups, thus, smoking status (‘smoker’ or ‘non-smoker’) was added as a between-subjects factor in analyses. Of the smokers in each group, the number of cigarettes smoked prior to baseline assessment was not significantly different (depressed group, 5.1 cigarettes [SD = 5.0]; control group, 4.2 cigarettes [SD = 3.4]; F1,30 = 0.2, P = 0.7). The most prevalent comorbid psychiatric disorders in the depressed sample were: post-traumatic stress disorder (16.1%), panic disorder with agoraphobia (13.2%) and alcohol abuse (8.8%).

Table 1. Demographic and clinical characteristics of depressed and healthy control groups, treatment responder and non-responder groups, and study completer and non-completer groups
 Depressed (n = 64)Control (n = 49) F / χ2 P Responder (n = 25)Non-responder (n = 30) F / χ2 P Completer (n = 55)Non-completer (n = 9) F / χ2 P
  • One-way anova.

  • Pearson's χ2-test.

  • §

    Fisher's exact test (no statistical value available).

  • MADRS, Montgomery–Asberg Depression Rating Scale; NZ, New Zealand.

Age39.6 (10.9)37.9 (11.4)0.70.4239.3 (11.2)38.4 (10.3)0.10.838.9 (10.8)40.3 (12.3)0.10.7
Sex (M : F)26:3817:320.30.5812:139:212.10.121:343:61.0§
Predicted verbal IQ106.8 (8.6)107.3 (6.7)0.10.72106.7 (10.1)106.5 (7.6)0.01.0106.7 (8.7)105.6 (9.4)0.70.7
Secondary education (years)4.2 (0.9)4.5 (0.8) (0.9)4.2 (0.9) (0.9)4.1 (1.1)0.50.5
Tertiary education (years)1.7 (1.9)1.7 (1.2)0.20.632.2 (2.0)1.5 (2.0) (2.0)1.1 (1.7)1.20.4
Ethnicity (NZ/European : other)57:741:70.30.5623:227:30.7§50:57:20.4§
Smoker (Y : N)22:427:427.0 0.008 8:1710:200.10.818:373:60.2§
Menstrual phase (follicular : luteal : menopause)16:13:5 (4 unknown)13:15:35.00.46:4:1 (2 unknown)9:8:2 (2 unknown)0.5§14:13:4 (3 unknown)3:2:10.8§
MADRS (baseline)35.7 (8.9)35.6 (9.4)35.8 (8.5)0.00.935.7 (8.8)35.4 (8.4)0.00.9
MADRS (10–14 days)25.6 (9.1)21.3 (7.2)28.5 (8.8)10.50.00225.2 (8.8)
MADRS (6 weeks)21.3 (12.3)9.8 (4.6)29.8 (8.2)118.4<0.000120.8 (12.0)
Age at depression onset30.8 (11.5)32.0 (10.8)27.9 (11.2) (10.9)32.0 (15.6)0.50.6
Years since depression onset9.1 (9.1)7.7 (9.3)11.5 (9.3) (9.3)8.3 (7.6)0.10.8
No. previous hospitalizations0.7 (1.4)0.6 (1.1)1.0 (1.7) (1.8)0.3 (0.4)1.40.3
Previous episodes (single : multiple)21:4312:138:223.70.0619:362:70.6§
Unipolar : bipolar depression57:721:428:20.7§48:79:00.6§
Psychotic features (Y : N)8:563:224:261.0§5:501:81.0§
Melancholic features (Y : N)31:3311:1417:130.80.429:262:70.1§
Atypical features (Y : N)3:611:242:281.0§3:550:91.0§

Twenty-one patients were unmedicated at baseline assessment. The remaining patients (n = 43) were taking selective serotonin reuptake inhibitors (SSRI; n = 20), serotonin-noradrenaline reuptake inhibitors (SNRI; n = 17), tricyclic antidepressants (TCA; n = 4) or monoamine oxidase reuptake inhibitors (MAOI; n = 2). Unmedicated depressed patients did not differ from medicated depressed patients on any demographic or clinical variables (all P > 0.2).

Treatment response (>50% reduction on the MADRS between baseline and 6 weeks) was used to categorize treatment outcome, resulting in 25 responders and 30 non-responders (nine out of 64 patients were lost to follow up). Treatment responders and non-responders were compared to determine whether they differed on demographic, clinical or psychoactive substance consumption variables. No significant differences were found (all P > 0.1). Importantly, responders and non-responders did not differ in MADRS scores at baseline (F1,55 = 0.0, P = 0.9; see Table 1).

Neuroendocrine findings

Dexamethasone was well tolerated by almost all participants. Some participants experienced slight sleeplessness on the night of dexamethasone administration and facial flushing the day after.

Depressed and healthy control groups did not differ in their post-dexamethasone cortisol levels (depressed group = 70.0 nmol/L [SD = 37.9]; control group = 70.6 nmol/L [SD = 28.3], F1,112 = 0.3, P = 0.6) or in the proportion of suppressors to non-suppressors in each group (depressed group = 7.8% non-suppressors, healthy control group = 2.0% non-suppressors; Fisher's exact test, P = 0.2). No significant effects of dexamethasone levels (F1,112 = 0.1, P = 0.8), sex (F1,112 = 0.2, P = 0.7), or smoking status (F1,112 = 1.6, P = 0.2) were found. The effect of age reached trend level (F1,112 = 3.9, P = 0.06), as did the correlation of age with cortisol response to dexamethasone (r = 0.20, P = 0.06).

No group by time interaction (F2,103 = 1.1, P = 0.4) was found in a repeated measures anova of cortisol response to dexamethasone over time in responders, non-responders and healthy controls.

Association between HPA axis function and facial emotion processing

HPA axis and facial expression recognition accuracy

The main findings from repeated measures anova, with group as the between-subjects factor, emotion as the within-subjects factor and cortisol level (post-dexamethasone) as a covariate, were significant effects of emotion (F5,535 = 34.3, P < 0.0001), and a significant interaction between emotion and cortisol levels (F5,535 = 2.4, P = 0.04). No effect of group (F1,109 = 0.2, P = 0.6), or emotion × cortisol levels × group (F5,535 = 1.1, P = 0.4) were found. Because the latter interaction was not significant, correlations between post-dexamethasone cortisol levels and facial expression recognition accuracy were initially conducted within the total group (depressed and healthy control participants).

Six analyses were conducted which correlated baseline post-dexamethasone cortisol levels with recognition accuracy of the six facial expressions of emotion (angry, happy, sad, fearful, disgusted and neutral expressions; see Table 2). Significant negative correlations were found between cortisol levels and the recognition of angry (r = −0.20, P = 0.03), sad (r = −0.24, P = 0.01), and disgusted (r = −0.22, P = 0.02) facial expressions. That is, as cortisol response to dexamethasone increased, recognition of angry, sad, and disgusted faces decreased.

Table 2. Correlation of facial expression recognition performance with post-dexamethasone cortisol levels in the total group at baseline and 10–14 days later
 Baseline (n = 113)10–14 days (n = 104)
R P r P
  1. r = Pearson's correlation coefficient.

Anger0.20 0.03 −0.170.1
Sadness0.24 0.01 −0.010.9
Disgust0.22 0.02 −0.180.08

To further understand the nature of the association between facial expression recognition accuracy and post-dexamethasone cortisol levels, correlations were conducted separately for depressed and healthy control groups. The negative correlation between recognition of angry faces and cortisol levels remained significant in the depressed group (r = −0.29, P = 0.02). In the healthy control group, the correlation was in the same direction, but was not significant (r = −0.08, P = 0.6). Depressed and healthy control groups showed similar negative correlations between the recognition of disgusted faces and cortisol levels (r = −0.24, P = 0.06 and r = −0.23, P = 0.09, respectively), but neither correlation reached significance due to reduced sample sizes. The negative correlation between recognition of sad faces and cortisol levels appeared to be driven by the depressed group, with the separate analyses showing the correlation to gain in strength in the depressed group (r = −0.40, P = 0.001), and being in the opposite direction for the healthy control group (r = 0.16, P = 0.3).

Further repeated measures anova was conducted for the 10–14-day assessment, with the same variables included as above. Again, the effects of emotion (F5,480 = 33.1, P < 0.0001), and emotion × cortisol levels (F5,480 = 2.9, P = 0.02) were significant, while group (F1,96 = 0.2, P = 0.6), and emotion × cortisol levels × group (F5,480 = 1.1, P < 0.3) did not reach significance. As can be seen in Table 2, no correlations between recognition accuracy of any of the five emotions (and neutral) and post-dexamethasone cortisol levels were significant. Inspection of the data showed that the correlations were occurring in the same direction at this time-point, but were weaker in strength.

HPA axis and neutral misinterpretation bias

Repeated measures anova, with group as the between-subjects factor, misinterpreted emotion (that is, the emotion that neutral faces were being misinterpreted as) as the within-subjects factor, and post-dexamethasone cortisol levels as a covariate, was conducted. A significant effect of emotion was found (F4,428 = 17.9, P < 0.0001), while all other main effects and interactions were not significant.

HPA axis and facial expression recognition reaction time

Other than a significant effect of emotion (F5,535 = 19.8, P < 0.0001), repeated measures anova of facial expression recognition reaction time (within-subjects factor) and cortisol levels (covariate) did not produce significant effects of group (F1,109 = 0.2, P = 0.7), cortisol levels (F1,109 = 0.3, P = 0.6), emotion × cortisol (F5,535 = 1.0, P = 0.4), or group × emotion × cortisol levels (F5,535 = 0.6, P = 0.7).

Effects of medication

Post-dexamethasone cortisol levels in patients taking antidepressant medications (n = 21), compared with patients not taking antidepressant medication (n = 43), were not significantly different at baseline (medicated = 72.0 nmol/L [SD = 44.3], unmedicated = 64.1 nmol/L [SD = 30.1], F1,63 = 0.9, P = 0.4). No significant differences were found between these groups on any baseline facial emotion processing measures (facial expression recognition, facial expression reaction time, and misinterpretation bias; P > 0.2).

Depressed patients were treated with various types of antidepressant medication (SSRI, SNRI, TCA or MAOI). It may have been of interest to investigate any effects that specific types of antidepressant medication had on cortisol response to dexamethasone, facial emotion processing, or treatment response. However, such an analysis would have required splitting the depressed group into small subgroups based on the type of antidepressant being taken, and then comparing these subgroups. This would have lowered the statistical power substantially, and thus, these analyses were not conducted.


In the current study, severely depressed inpatients and healthy controls completed an assessment of HPA axis function (DST) and facial emotion processing (Facial Expression Recognition Task) at two time-points (baseline and 10–14 days later). Here our primary focus is on the association between HPA axis function and facial expression recognition, which to our knowledge, has not been examined previously.

These results should be viewed in the context of facial emotion processing abnormalities in depression, which have been shown in previous studies, and which have also been demonstrated in the current sample. Previous studies have tended to show negative interpretation biases when processing neutral faces and increased attention towards negative facial expressions in depression.7 In this sample, depressed patients displayed a specific deficit in the recognition of disgusted facial expressions, compared with healthy controls.10 A negative interpretation bias in our sample has also been found, with depressed patients being more likely to interpret neutral facial expressions as sad, and being more likely to interpret posed sad expressions as real sadness, compared with healthy controls.10,20

Cortisol response to dexamethasone was found to be significantly related to aspects of facial emotion processing. When controlling for post-dexamethasone cortisol levels, there was a significant relation between cortisol and recognition of certain emotions which was not modulated by group (depressed or control). Given this lack of an interaction between cortisol, emotion and group, correlations were conducted within the total sample (depressed and healthy control participants). Reduced accuracy in processing angry, sad and disgusted facial expressions was significantly associated with increased cortisol levels following dexamethasone administration. This finding suggests that reduced HPA axis feedback may be associated with deficits in recognizing threat-related facial expressions, regardless of the presence of major depression. Although these correlations did not remain significant at the 10–14-day assessment, inspection of the data indicated that correlations were in the same direction but slightly weaker in strength (see Table 2). Reduced sample size at this time-point probably explains this finding.

The amygdala plays a role in HPA axis activation and processing of threatening emotional information and thus it is likely that the current findings can be explained, at least in part, by amygdala involvement. However, it is difficult to explain why increased cortisol response to dexamethasone would lead to reduced accuracy in recognizing threatening facial expressions from an evolutionary viewpoint. van Marle et al.5 reported heightened sensitivity, but not specificity, to facial expressions of emotion in their healthy sample. It may be that those individuals with reduced HPA axis feedback in the current study were showing indiscriminate sensitivity to threatening expressions, with angry, sad and disgusted expressions being incorrectly interpreted as another threatening emotion, such as fear. Given the severity of depression in this sample, we restricted the Facial Expression Recognition Task to a relatively low number of stimuli (16 faces per emotion, morphed between 50% and 100% full emotion). This is not a sufficient number of stimuli to adequately analyze the issues of what other emotion each emotion was misclassified to when it was incorrectly labeled. A further study with a larger number of stimuli would be necessary to conduct this analysis but would be practically difficult in patients with this severity of depression.

An alternative explanation for the current findings is that significant correlations between post-dexamethasone cortisol levels and facial emotion processing accuracy may not have been specific to sadness, disgust and anger per se. These facial expressions were the most difficult expressions to recognize for all participants (see 10) thus, increased cortisol response to dexamethasone may reduce effortful processing and result in these participants being more likely to guess or give up on difficult cognitive tasks. Again, it would have been of interest to determine what emotions angry, disgusted and sad faces were being misinterpreted as, in order to determine whether certain trends existed, or whether patients were simply guessing.

We were surprised to find no evidence of HPA axis abnormality in our depressed sample given that HPA axis dysfunction is reported consistently in depression research.1 Several studies have shown reduced HPA axis feedback in depressed compared with non-depressed individuals by examining plasma cortisol levels following the DST or the dex/CRH test, or by determining the proportion of dexamethasone suppressors and non-suppressors in each group.19,21–23 Although there were numerically more dexamethasone non-suppressors in the current depressed group compared with the healthy control group, the difference was not statistically significant. Furthermore, the percentage of non-suppressors in the depressed group (8%) was substantially lower than in previous studies, which have found rates of dexamethasone non-suppression to be at least 40% in inpatient major depression.19,24

There are several possible explanations for this finding. First, antidepressant medications may have interfered with cortisol response to dexamethasone. However, previous studies have found no evidence that antidepressant medications influence cortisol response to the dex/CRH test18,23 and in the current sample, medicated and unmedicated depressed patients had comparable cortisol responses to dexamethasone. Second, it could be that the DST was not sensitive enough to detect differences in this sample. Yet while it has been suggested that the dex/CRH test may be more sensitive than the DST in identifying HPA axis abnormalities,21 several studies have found abnormal cortisol responses to dexamethasone using the DST.6,17,19 Furthermore, studies using the dex/CRH test have found an elevated cortisol response to dexamethasone before CRH injection in depressed samples.23 Importantly, there was no evidence to suggest that differences in dexamethasone metabolism were biasing the results. Sensitivity may have been increased by measuring baseline plasma cortisol levels and cortisol levels at later time-points the following day, as has been done in previous studies differentiating depressed and control participants.17,25 However, multiple sampling was not possible in this severely depressed group on a busy in-patient unit.

The explanation which we favor for the lack of HPA axis activation found in our depressed sample is that the sample was fundamentally different from inpatient samples in other studies that have shown HPA axis abnormalities. Bed occupancy at Hillmorton Hospital is usually greater than 100%. Individuals with severe but uncomplicated major depression are unlikely to be hospitalized. Those hospitalized with major depression have typically experienced major psychosocial problems prior to admission. Thus, it may be that individuals in the depressed group had a less biological illness than those in previous studies, reflected by normal HPA axis functioning. Watson et al.26 found that a sample of patients with chronic depression also showed a normal cortisol response to the sensitive dex/CRH test, an unexpected finding given the persistence of depressive symptoms in this subtype of depression. The authors suggested that the absence of HPA axis abnormality in their sample was due to depressive symptoms in chronic depression being maintained through social and cognitive dysfunction, rather than biological dysfunction.26

The DST was conducted twice in depressed and healthy control participants in the current study to determine whether the test could be used as an early neuroendocrine marker of treatment response in major depression. Studies have produced preliminary evidence to suggest that early attenuation (within 2 weeks from baseline) of cortisol response to the dex/CRH test may predict longer-term improvement in clinical state.22 The current depressed group was similar to the sample in the study by Ising et al.,22 also consisting of depressed inpatients who were heterogeneous in antidepressant medications and subtypes of depression. However, as noted, in contrast to our study, cortisol levels post-dexamethasone and pre-CRH injection was significantly higher in the depressed group than in control participants. In the study by Ising et al.,22 no attenuation of cortisol response to dexamethasone (before CRH injection) was evident with treatment which is then in accord with our study.

This study was originally designed to examine the relation between HPA axis feedback and facial expression recognition in depression, with the expectation that both would be abnormal and that HPA axis abnormality would mediate abnormalities in facial expression recognition. However, while there were differences between groups in facial expression recognition, we found no difference in DST results. In addition, analysis of covariance indicated that group was not a factor modulating the relation between facial expression recognition and DST results. Analyses of correlation between cortisol and facial expression recognition therefore were in a large group of participants, some of whom suffered from depression, and across a relatively normal range of post-dexamethasone cortisol levels. Thus, the study does not examine facial expression recognition in the face of abnormal activation of the HPA axis.

The DST examines an aspect of HPA axis function: autoregulatory feedback. This does not necessarily correspond to ambient cortisol levels. It would be useful to examine the relation between measures of cortisol (stimulated and unstimulated) in order to examine the relation between HPA axis function and facial expression recognition further.

This study presents preliminary evidence of an aspect of HPA axis function being related to recognition accuracy of threat-related facial expressions, perhaps through involvement of the amygdala. Replication of this finding is required, using more in-depth examination of both facial emotion processing and HPA axis function.


We thank the patients and staff at Hillmorton Hospital for their cooperation in this study. We are also grateful to the Tertiary Education Commission for their financial support by providing K.D. with a Bright Future Doctoral Scholarship. There is no conflict of interest.