Hypothalamic–pituitary–adrenal (HPA) axis response in borderline personality disorder is controversial. Several studies published two decades ago reported high rates of non-suppression with the 1 mg dexamethasone test in patients with borderline personality features, suggesting an association of the condition with affective disorders (Reference Sternbach, Fleming and ExteinSternbach et al, 1983). However, recent studies suggest some similarities with HPA axis disturbances found in post-traumatic stress disorder (PTSD). Studies of people with PTSD have reported lower cortisol levels than in a normal comparison group, increased lymphocyte glucocorticoid receptor density and enhanced cortisol suppression with a 0.5 mg dexamethasone test (Reference Rinne, de Kloet and WoutersRinne et al, 2002; Reference Yehuda, Halligan and GolierYehuda et al, 2004). On the other hand, several studies with the 0.5 mg dexamethasone test have reported enhanced cortisol suppression in patients with borderline personality disorder and comorbid post-traumatic symptoms (Reference Grossman, Yehuda and NewGrossman et al, 2003, Reference Lange, Wulff and BereaLange et al, 2005), but not in patients with this personality disorder but without PTSD. However, cortisol suppression was very high in the normal control group in these studies, ranging from 70% to 85%, which might have reduced the discriminatory power of the 0.5 mg test. A preliminary study with a lower dose (0.25 mg) of dexamethasone (Reference Carrasco, Diaz-Marsa and PastranaCarrasco et al, 2003) reported enhanced cortisol suppression in borderline personality disorder with no comorbid PTSD compared with other personality disorders and demonstrated high specificity for detection of cortisol suppression. However, the conclusions were limited by the absence of a comparison group of normal individuals.

METHOD

Patients were selected at the emergency room for repetitive self-aggressive behaviour (at least two episodes in the preceding 6 months) and evaluated by a senior psychiatrist with the Structured Clinical Interview for DSM–IV Axis I Disorders (SCID–I; Reference First, Spitzer and GibbonFirst et al, 1995) and Axis II Personality Disorders (SCID–II; Reference First, Gibbon and SpitzerFirst et al, 1997), as well as with the Childhood Trauma Questionnaire (Reference Bernstein, Stein and NewcombBernstein et al, 2003) and the Zanarini Rating Scale for Borderline Personality Disorder (ZAN–BPD; Reference Zanarini, Vujanovic and ParachiniZanarini et al, 2003).

Fifty-one patients with a diagnosis of borderline personality disorder were selected. Those with comorbid PTSD (n=4) were excluded, as were those with current major depression (n=9) or substance dependence (n=4) and a lifetime history of bipolar disorder (n=1) or schizophreniform disorder (n=1). Finally, 32 patients with no major metabolic or hormonal disease entered the study and were admitted to the hospitalisation unit for a wash-out period to eliminate medication and other drugs. The control group included 18 healthy individuals recruited from a healthcare prevention programme and matched with the patients for age and gender.

Biological tests followed a wash-out period of at least 1 week for anxiolytic medication and 3 weeks for other medications and illicit drugs (5 weeks for fluoxetine). Participants were admitted to the psychoendocrinology research unit at 07.30 h on day 1. An intravenous catheter was inserted at 08.00 h, allowing subsequent blood sampling exempt from the stress-inducing effects of needle-sticks. After 30 min, a blood sample was taken for measurement of plasma cortisol. At 23.00 h on day 1 the participants were administered an oral capsule containing 0.25 mg dexamethasone, and on day 2 at 08.00 h blood samples were taken again for cortisol measurement. Participants were tested under strictly controlled conditions, including minimal activity, 8 h fasting and sleep from 23.00 h the previous night.

Between-group comparisons employed analysis of covariance (ANCOVA) for differences in percentage cortisol suppression and chi-squared tests for differences in the rate of non-suppressor participants. Within the patient group relationships between variables were explored using Pearson's correlation coefficients. All statistical analyses were two-tailed with a 0.05 level of significance.

RESULTS

Distribution of age (patients, mean 30.6 years, s.d=6.4; controls, mean 29.7 years, s.d.=5.5) and gender (patients, 59% female; controls, 61%) showed no significant difference between groups. Across all groups, cortisol suppression was 55% (s.d.=31.4) in women and 50% (s.d.=35.0) in men, with no significant difference (t=0.52, d.f.=41, P=0.6).

Cortisol non-suppression was defined as a post-test cortisol level >5 μg/dl. In the borderline personality disorder group, 10 out of 29 patients (34%) were non-suppressors, v. 14 of 16 control participants (88%) (χ²=11.6, d.f.=1, P<0.01). Analysis of covariance revealed that the patient group had significantly lower cortisol levels pre-test (F=4.0, d.f.=1, P<0.05) and post-test (F=19.8, d.f.=1, P<0.01) compared with controls, as well as a significantly greater percentage cortisol suppression suppression (F=11.09, d.f.=1, P>0.01) (Table 1). The difference in cortisol levels was not significantly altered by reanalysis with age, anxiety or depression scores as covariates.

Table 1

Cortisol suppression in the 0.25 mg dexamethasone test

	Borderline personality disorder group	Control group
Cortisol level, μg/dl: mean (s.d.)
Baseline 1	17.2 (5.4) *	20.8 (7.0)
Post-test 2	5.7 (5.5) **	13.9 (6.4)
Cortisol suppression, %: mean (s.d.) 2	63.2 (30.1) **	31.5 (30.5)
Post-test cortisol levels, n (%)
0–≤5 μg/dl	19 (66)	2 (12)
> 5–10 μg/dl	5 (17)	3 (19)
> 10 μg/dl	5 (17)	11 (69)

1. Borderline personality disorder group n=32, control group n=18

2. Borderline personality disorder group n=29, control group n=16

^* P < 0.05

^** P < 0.01

Percentage cortisol suppression significantly correlated with severity of disorder on the Zanarini scale (r=0.42, P<0.05) but not with scores on the Childhood Trauma Questionnaire (r=0.068, P=0.73), which were higher in the patient group (mean 30.1, s.d.=8.1, range 16–54) than the controls (mean 23.2, s.d.=9.7, range 15–48; t=3.0, d.f.=66, P<0.01).

DISCUSSION

Significantly low baseline cortisol levels and enhanced cortisol suppression with 0.25 mg dexamethasone in borderline personality disorder without comorbid PTSD seem to be contrary to findings in previous studies (Reference Rinne, de Kloet and WoutersRinne et al, 2002; Reference Grossman, Yehuda and NewGrossman et al, 2003; Reference Lange, Wulff and BereaLange et al, 2005) reporting enhanced cortisol suppression in borderline personality disorder to be associated specifically with PTSD symptoms. The relationship of enhanced cortisol suppression and borderline personality disorder in our study is further supported by significant correlation with severity of disorder (ZAN–BPD) and the lack of significant correlation with childhood trauma scores.

Interestingly, the rate of comorbid PTSD in our patient sample (10%) was lower than that reported in previous studies (Reference Grossman, Yehuda and NewGrossman et al, 2003; Reference Lange, Wulff and BereaLange et al, 2005), which found rates of PTSD comorbidity of 25–50%. This suggests that the patients in our sample might have been clinically different from those in other studies, who were recruited by advertisements or from veterans' hospitals or facilities for abused women. This might have increased homogeneity, thereby favouring detection of biological abnormalities in borderline personality disorder, which is composed of different clinical and biological domains.

Also, methodological differences might explain the discrepancy of our results with other studies reporting high salivary baseline cortisol levels and increased rates of non-suppression in borderline personality disorder (Reference Lieb, Rexhausen and KahlLieb et al, 2004). In addition, methodological differences in cortisol measurement might also lead to different results. We used a laboratory method with strictly controlled conditions of environmental stress, including 30–60 min of relaxed delay before sample extraction. Ambulatory conditions for cortisol salivary sampling in the previously mentioned study might not be free of external stress influences, which could account for increased cortisol levels.

The very low dose (0.25 mg) of dexamethasone might be responsible for the strong group effect which was not found in previous studies using 0.5 mg doses. As predicted, 0.25 mg dexamethasone produced milder cortisol suppression in the control group than the 0.5 mg dose as reported in previous studies (31% v. 60–80%) (Reference Rinne, de Kloet and WoutersRinne et al, 2002; Reference Grossman, Yehuda and NewGrossman et al, 2003; Reference Lange, Wulff and BereaLange et al, 2005). Furthermore, 65% of the patients in our study were categorised as complete suppressors (cortisol level <5 μg/dl) compared with only 12% of controls, suggesting that the 0.25 mg dexamethasone test could be useful for discriminating abnormal HPA axis functioning in personality disorders.

In conclusion, our results show low plasma cortisol levels and enhanced cortisol suppression in a 0.25 mg dexamethasone suppression test in individuals with borderline personality disorder with no PTSD compared with healthy controls. Increased HPA feedback inhibition probably reflects increased lymphocyte glucocorticoid receptor density, which might be secondary to previous intense and persistent stress in both disorders (Reference Yehuda, Halligan and GolierYehuda et al, 2004). However, these results apply to a specific subgroup of patients selected by the presence of repetitive self-aggressive behaviours in the previous 6 months, and should not be generalised to all people with borderline personality disorder.