Metyrapone as Additive Treatment in Major Depression: Title and subTitle BreakA Double-blind and Placebo-Controlled Trial FREE

Our understanding of the neuroendocrine pathophysiology of depressionhas made significant progress in recent years, which should help to developnew remedies. Alterations of the hypothalamic-pituitary-adrenal (HPA) axisare the most consistent pathological endocrine findings in depression.^{1 - 3} The key regulator ofHPA axis activity, corticotropin-releasing hormone (CRH), is increased indepression.^{4 - 6} Theeffects of CRH are modulated by neuropeptides^{7 - 9} andmonoaminergic transmitters.^{10 - 11} Theexaggerated HPA function in depression appears to be also the result of animpaired glucocorticoid receptor (GR)^{12 - 13} andmineralocorticoid receptor (MR)^{14 - 15} feedbackcontrol. Whether HPA-axis hyperactivity is initially caused by a CRH overdriveresulting in MR/GR dysfunction or vice versa by dysregulation at another levelis yet undecided.^{16 - 17}

Hence, attempts have been made to treat depression by directly targetingHPA-axis activity. Currently, 3 major pathways are investigated: (1) administrationof CRH antagonists like R121919¹⁸ ; (2) administrationof GR antagonists like RU 486 or Org 34517^{19 - 21} ;and (3) treatment with steroid-synthesis inhibitors like ketoconazole, aminoglutethimide,or metyrapone.²²

Preclinical studies nurture the hope for new therapeutic strategiesbased on steroid-synthesis inhibition,²³ butclinical data about the antidepressive efficacy of these compounds are mainlyconfined to small open-label trials and case reports.^{24 - 30} Firstreports date back to the 1970s when Jeffcoate et al³¹ treateddepressed Cushing patients with metyrapone. In 1991, Murphy³² reportedthe first case of a successfully treated depressed patient. One single-blindcrossover trial of metyrapone and 2 small double-blind studies on ketoconazolein the treatment of depression have been published.^{33 - 35}

Our aim was to conduct the first prospective, randomized, placebo-controlled,and double-blind clinical trial of metyrapone as additive treatment in depression.Metyrapone was preferred because this compound inhibits selectively the 11β-hydroxylaseand the 11β-hydroxysteroid-dehydrogenase type 1 (11β-HSD-1),^{36 - 37} thereby exerting direct effects withinthe central nervous system.³⁸ The additiveapproach was applied because the intended inclusion of severely depressedpatients made a pure placebo group ethically challenging.³⁹ Furthermore,the continuous use of an antidepressant allowed a standardized follow-up afterthe double-blind period.

The questions to be answered were whether metyrapone exerts potentiatingeffects during a standard antidepressant therapy and whether an earlier onsetof action and an improved overall and sustained treatment response can beachieved. Because GR/MR distribution⁴⁰ as wellas 11β-HSD-1 activities^{41 - 42} aresubject to sexual dimorphism in humans, the sample was prospectively stratifiedfor sex and balanced for treatment with 2 selected serotonergic antidepressants.

A total of 352 inpatients referred within a 3-year period from May 1998to May 2001 to 2 specialized wards for affective disorders at the UniversityHospital Hamburg met DSM-IV⁴³ diagnosticcriteria for major depressive disorder according to assessments by 3 experiencedpsychiatrists (H.J., M.S., and K.W.). These prospective participants werescreened for inclusion and exclusion criteria. A total of 63 inpatients gaveinformed consent and were included in the trial (Figure 1).

Inclusion criteria were (1) a diagnosis of major depressive disorder,single or recurrent according to DSM-IV criteria(diagnoses 296.2 or 296.3); (2) a minimum baseline score of 18 points on theHamilton Rating Scale for Depression, 21-item version (HAMD-21)⁴⁴ ;(3) age from 18 to 75 years; (4) a period of at least 5 days free from antidepressants,antipsychotics, mood stabilizers, and all other medications except mild antihypertensiveagents; and (5) a negative result on urinary drug screening. Criteria forexclusion were (1) a current DSM-IV diagnosis forother axis I psychiatric disorders; (2) serious medical conditions, especiallythose associated with adrenal insufficiency; and (3) pregnancy, nursing, orrefusal by women to use a reliable method of birth control.

Participants were randomly assigned to a study group if they satisfiedthese criteria. The study sample was stratified for sex and balanced for the2 serotonergic antidepressants used (nefazodone and fluvoxamine). The studywas approved by the local ethics committee. Written informed consent was obtainedfrom all subjects.

The 63 participants who met the inclusion criteria were randomly assignedto the treatment groups. The pharmacological reason for the use of serotonergiccompounds is based on the presumed hippocampal effects of metyrapone on 5HT-1Areceptors. The clinical rationale for the use of these compounds is basedon activating properties of fluvoxamine and sedative effects of nefazodone,allowing inclusion of inhibited as well as agitated patients. The respectiveantidepressant was selected according to clinical symptomatology.

Following baseline assessments, subjects entered a 3-week, double-blindtreatment period with either metyrapone (250 mg given orally 4 times a day)or placebo (4 times a day) at 8 AM, 12 PM, 6 PM, and 10 PM, in addition to a standard antidepressanttreatment with nefazodone or fluvoxamine. The dose range for fluvoxamine was150 to 200 mg/d and for nefazodone 300 to 400 mg/d after 1 week of treatment.

The medication with metyrapone or placebo and each of the serotonergicantidepressants was started simultaneously on day 1 of the study. After day21, patients continued to take the antidepressant, but metyrapone or placebomedication was stopped. The study ended after 5 weeks.

Both serotonergic antidepressants were given as commercially availableoriginal medication. Metyrapone was capsuled, and identical placebo capsuleswere produced. Allocation codes were provided in sealed envelopes for eachpatient at the pharmacy of University Hospital Hamburg, where formulationand blinding were conducted. The randomization was organized by a computer-generatedlist using the PLAN procedure from the SAS/STAT software (SAS Institute Inc,Cary, NC). The concomitant use of lorazepam was allowed for a maximum of 8days from day 0 to day 7 of the 35-day treatment period.

Following the assessment of inclusion and exclusion criteria on day0, data were collected, including information about the history of illnessand sociodemographic data concerning family and social, psychological, andmedical problems. The psychopathometric assessments were performed at days0, 3, 7, 14, 21, 28, and 35 between 10 AM and 1 PM.External ratings were the HAMD-21,⁴⁴ the Montgomery-AsbergDepression Rating Scale (MADRS),⁴⁵ and theClinical Global Impressions scale.⁴⁶ Self-ratinginstruments were the Beck Depression Inventory (revised 21-item version)⁴⁷ and the Zung Self-Rating Depression Scale (20-itemversion).⁴⁸ Adverse effects were assessed bya German adaptation of the Udvalg for Kliniske Undersogelser (UKU) side effectscale.⁴⁹ External rating instruments were appliedby 2 authors (H.J. and M.S.), who had training sessions to assure high qualityand interrater reliability.

Blood samples were drawn between 8:30 and 9:30 AM ondays 0, 1, 3, 7, 10, 14, 21, 28, and 35 for clinical chemistry, endocrineparameters (cortisol, 11-deoxycortisol, corticotropin [ACTH], and dehydroepiandrosterone[DHEA]), and drug monitoring of fluvoxamine and nefazodone during the steady-statephase (data given for day 14).

Endocrine parameters were determined by commercial radioimmunoassayswith coated tube techniques (cortisol, 11-deoxycortisol, DHEA; DRG-Instruments,Marburg, Germany) or immunoradiometrically (ACTH; Nichols Institute, San JuanCapistrano, Calif). The cross-reactivity of cortisol determinations with 11-deoxycortisolwas less than 0.1%.

Fluvoxamine and nefazodone were determined after automated extractionvia column switching by reverse-phase high-performance liquid chromatographyusing UV detection.⁵⁰

During the trial, only 1 author (K.W.) had access to laboratory datato control for clinically relevant changes.

The intention-to-treat sample of 63 patients was estimated to be sufficientto detect large effect sizes with a power of 85% at an α level of.05.For the statistical evaluation of outcome criteria, the intention-to-treatsample of 63 patients with the dropouts classified as nonresponders has beentaken into consideration, whereas for hormones, the statistical evaluationwas based on those 60 patients who actually received medication during thestudy. Missing values were substituted using a last observation carried forwardapproach.

A priori primary outcome criteria were (1) 2 psychometric criteria definedby the number of responders and the time to onset of action and (2) the courseof concentrations of ACTH, cortisol, 11-deoxycortisol, and DHEA. The numberof responders was considered twice after 3 and 5 weeks by defining the treatmentresponse as a 30% and 50% reduction, respectively, from baseline HAMD scores.Onset of action was determined by the survival analytical approach of Stassen,⁵¹ which defines the onset of action as the time pointat which at least a 20% reduction of baseline HAMD scores occurred. Otherpsychometric scores, demographic parameters, and adverse effects were consideredas secondary variables.

Differences in the responder rates between the treatment groups weretested by Fisher exact test. The other main criterion, the time to onset ofaction, was analyzed with a Kaplan-Meier survival analysis and embedded log-ranktest as described,⁵¹ considering dropouts ascensored cases.

Multivariate analyses of covariance (MANCOVA) with sex as a covariatewere further applied for testing the effects of treatment, antidepressantmedication, and time on secondary variables like HAMD and MADRS. Accordingto the underlying data structure, differences between metyrapone and placeboin demographic baseline variables or adverse effects were tested by MANCOVAor nonparametric tests (Fisher exact test or median test). Associations betweensome variables in the various experimental conditions were tested with thePearson correlation coefficients. To characterize the study sample on a descriptivelevel, rates of remission defined as a HAMD score of less than 8 at day 35were calculated, and a Quitkin pattern analysis based on the Clinical GlobalImpressions scale to define the time point, when at least a marked improvementwas detectable, was performed.⁵¹ Also, therelapse rates after week 3, defined as a persistent 20% increase in HAMD scores,and the treatment effect sizes were derived.

For the hormonal variables, sex was considered a covariate. Multivariateanalyses of covariance with a repeated-measures design were applied to testthe effects of treatment and antidepressant medication (between-subject factorswith 2 levels: metyrapone or placebo; nefazodone or fluvoxamine) and time(a within-subject factor with 9 levels) on ACTH, cortisol, deoxycortisol,and DHEA as dependent variables. A complementary statistical evaluation (posthoc analysis) by MANCOVA focused on possible differences in the endocrineparameters between responders and nonresponders at the end of the study (day35). Response, treatment, and time were considered as influential variablesand sex and baseline concentrations as covariates.

In cases of significant factor effects in the MANCOVA, univariate Ftests and tests with contrasts were conducted to identify those variablescontributing significantly to these effects and to locate the time pointsof significant differences. As a nominal level of significance, α = .05was accepted. All post hoc tests (univariate F tests and tests with contrasts)were performed at a reduced level of significance (Bonferroni correction)to keep the type I error less than or equal to 0.05. Measures are given inmean ± SEM unless otherwise stated.

Eleven patients of the metyrapone group and 9 of the placebo group receiveda DSM-IV 296.2x diagnosis, and 22 of the metyrapone-treatedand 21 of the placebo-treated patients were classified as recurrently depressedpatients (DSM-IV 296.3x).Regarding diagnoses, historyof depression (duration of illness, age at time of onset, number of episodes,severity of current episode, and duration of current episode), and baselinepsychopathometric scores (HAMD, MADRS, Clinical Global Impressions-SeverityGrade scale, Beck Depression Inventory,and Zung Self-Rating Depression Scale),no significant differences between metyrapone and placebo were detected (Table 1 and Table2). Women had suffered more previous episodes than men.The medianfor previous episodes in women was 3 (minimum/maximum, 1/21), compared with2 in men (minimum/maximum, 1/8). Furthermore, the duration of the currentindex episode prior to study inclusion was longer in men (median, 4 months;minimum/maximum, 1/24 months) than in women (median, 2 months; minimum/maximum,0.5/19 months).

Of the 63 patients randomized, 56 patients completed the trial and 7patients dropped out. Three dropouts revoked their consent after inclusionand before the treatment was started, and 4 patients left during the studybecause of adverse effects. Four dropouts were assigned to the placebo group,3 to the metyrapone group, 4 to the fluvoxamine group, and 3 to the nefazodonegroup. Five of 7 dropouts were women.

With regard to premedication, 2 of 33 patients in the metyrapone groupreceived fluoxetine 4 weeks prior to the inclusion date. Both patients tookit for fewer than 10 days. In the placebo group, 1 of 30 patients had takenfluoxetine for a few days 2 weeks before inclusion. In the metyrapone group,1 patient received nefazodone prior to inclusion and was switched to fluvoxaminebecause of previous adverse effects; 1 patient received fluvoxamine for 3days, 11 days before inclusion in the trial. In the placebo group, 1 patientwas exposed to fluvoxamine for 1 day, a week before entering the hospital.Regarding data of all antidepressants taken within a 4-week period prior toinclusion as well as lifetime exposure to antidepressants, no differencesbetween the treatment groups were detectable. One patient in each group receiveda β-blocker in low dose.

Considering the main outcome, the responder rates in the metyraponegroup at day 21 and day 35 were significantly greater than those in the placebogroup. Twenty-three of 33 patients in the metyrapone group vs 13 of 30 patientsin the placebo group showed a positive treatment response using the 30% reductioncriterion of the HAMD at day 21 (Fisher exact P = .031).At day 35, 19 responders in the metyrapone group vs 10 responders in the placebogroup yielded a significantly better response rate and better treatment outcomein the metyrapone group (50% reduction criterion; Fisher exact P = .047).The course of the HAMD scores is shown for bothtreatments in Figure 2, and the treatmentefficacy is similarly detectable with HAMD, MADRS, Beck Depression Inventory,and Zung Self-Rating Depression Scale with effect sizes ranging from 0.37to 0.73 (Table 2).

Concerning the time to onset of action, the Kaplan-Meier survival analysis(Figure 3) indicated a significantlyshorter time to improvement for the metyrapone group than for the placebogroup (log-rank test, P<.006). Fifty percent ofthe metyrapone-treated patients showed early improvement by a HAMD score reductionof at least 20% at day 7. In the placebo group, a 20% improvement was reachedby 50% of the patients only at day 14. At day 35, all patients from the metyraponegroup showed at least a 20% improvement, whereas in the placebo group, 5 of30 did not show any improvement. The lower panel of Figure 3 depicts the Quitkin pattern⁵¹ ofthe Clinical Global Impressions scale scores also indicating a faster onsetof action induced by metyrapone. About the same proportion of patients inboth groups showed no marked improvement on the Clinical Global Impressionsscale until the study’s end (metyrapone, 6/33; placebo, 10/30). Theremission rates at week 5, defined as a total HAMD score of less than 8 points,were 10/33 in the metyrapone group and 7/30 in the placebo group. The relapserate after day 21, defined as a persistent 20% increase in HAMD scores today 35, was very low and similar for the metyrapone group (3/33) and the placebogroup (3/30). Analyzing the influence of treatment, antidepressant medication,and time on the sum scores of the HAMD and MADRS, with sex and baseline scoresas covariates, we observed significant effects of treatment (F_2,52 = 4.52;significance of F = 0.02) and time (F_10,47 = 6.16;significance of F<0.001) on each of the considered scales (univariate Ftests, P<.05). Fluvoxamine and nefazodone didnot significantly differ in their antidepressant efficacy, neither alone norin interaction with the factors treatment and time. Tests with polynomialcontrasts revealed significantly stronger reductions of HAMD (F_1,56 = 42.63;significance of F<0.001) and MADRS (F_1,56 = 52.75;significance of F<0.001) scores by metyrapone compared with placebo ateach time point until day 35 (Figure 2).

The MADRS scores showed an earlier and more pronounced reduction thanthe HAMD scores in the metyrapone group, verifying the statement that theMADRS is sensitive to early antidepressant actions.⁴¹ Item-3(inner tension) and item-10 (suicidal ideation) showed especially fast reductionsfor patients treated with metyrapone.

The MANCOVA of the endocrine data showed significant main effects oftreatment (F_4,52 = 8.05; significance of F = 0),time (F_32,1642 = 9.18; significance of F = 0),and antidepressant medication (F_4,52 = 3.61; significanceof F = 0.01), as well as a significant interaction effect of all3 factors (F_32,1642 = 1.83; significance of F = 0.003)attributed to all hormones except DHEA (univariate F tests, P<.05). Subsequent univariate F tests showed that cortisol showedslightly higher concentrations in the fluvoxamine group after commencing treatment.Both antidepressants did not show different effects on ACTH, 11-deoxycortisol,and DHEA. By analyses of simple effects of antidepressant medication withinthe factors treatment and time, significant differences did not emerge atthe various time points of the study for any of the endocrine variables. Therefore,it is not necessary to differentiate between the antidepressants regardingthe endocrine variables. Also, sex as a covariate did not significantly influencethe investigated hormones.

By analysis of the simple effects of treatment and time, we found thatduring metyrapone treatment, ACTH, 11-deoxycortisol, and DHEA showed significantelevations of plasma concentrations compared with baseline (test with polynomialcontrasts; P values for a second-degree polynomial<.05) (Table 3). These elevationswere already significant at day 1 for 11-deoxycortisol, ACTH, and DHEA. Thesehormones were also significantly increased during the entire 3-week treatmentperiod compared with placebo (tests with contrasts, P<.05).For cortisol, there was a trend toward slightly higher cortisol concentrations(P values for a first-degree polynomial = .052).

After discontinuation of metyrapone, the concentrations of ACTH, 11-deoxycortisol,and DHEA decreased reasonably, and, apart from still-increased ACTH concentrationsat day 35 in the metyrapone group, no significant differences from placeboremained (test with contrasts, P<.05). Duringplacebo, analyses of time effects also showed slightly increased ACTH and11-deoxycortisol concentrations compared with baseline, which were similarfor fluvoxamine and nefazodone. Cortisol for the whole placebogroup and DHEA concentrations remained largely unchanged.

With metyrapone, a significant correlation between ACTH and 11-deoxycortisolplasma concentrations emerged (r_beta = 0.953; P<.001 at day 21), and a similar but less robust correlation wasfound between ACTH and DHEA (r_beta = 0.518; P<.01 at day 21). No significant correlations between these parametersappeared in the placebo group. Also, no significant correlation was detectedbetween ACTH and cortisol levels during metyrapone administration as evidenceof a sufficient enzyme block.

Comparing responders and nonresponders in the metyrapone group, a positivetreatment response after 5 weeks appeared to be associated with greater elevationsof ACTH and 11-deoxycortisol plasma concentrations (Table 4) (Figure 4), althoughthis difference failed to reach statistical significance in a post hoc MANCOVA.

There was no significant association between basal cortisol levels atday 0 and treatment outcome at day 21 or day 35, although baseline cortisolconcentrations were higher for the responders taking metyrapone. No differenceemerged in the response rate by splitting the patient sample at a cutoff ofgreater than 30 μg/dL of cortisol at baseline. Higher basal cortisol concentrationswere not associated with a more favorable outcome during metyrapone treatment.

The metyrapone treatment was well tolerated, and no serious adverseeffects occurred. Minor adverse effects had a low incidence (Table 5) and were predominantly reported by women (mean ± SEMadverse events reported: women, 3.32 ± 0.47; men, 2.55 ± 0.40).Only nausea and headaches were reported significantly more often during metyraponetreatment compared with placebo (Fisher exact P = .037and P = .048, respectively). Adverse effectswere mainly due to the serotonergic antidepressant treatment, especially inthe first 2 weeks. Patients receiving fluvoxamine reported more nausea andrestlessness, while patients taking nefazodone complained more frequentlyof a dry mouth. We did not observe any alterations in general clinical chemistryparameters.

All patients taking metyrapone showed significant elevations of 11-deoxycortisol;therefore, full compliance can be assumed. At the end of the treatment course,both raters and patients guessed whether placebo or metyrapone had been taken.An association analysis (κ coefficient) showed no agreement betweenraters’ and patients’ guesses and the identity of the given medication.

Plasma concentrations of both antidepressants showed that all patientswere compliant with the respective treatment. Mean ± SEMplasma concentration at steady state for nefazodone was 691 ± 23ng/mL, for the metabolite meta-chlorophenylpiperazine (mCPP) 41 ± 1.6ng/mL, and for fluvoxamine 77 ± 3.4 ng/mL. No effect of metyraponeon these concentrations was detected, and no associations between plasma concentrationsof antidepressants and outcome parameters emerged.

Lorazepam was restricted to the first 8 days (day 0 to day 7) of thestudy, according to our protocol. Thirteen of 30 patients in the placebo group(mean ± SEM dose, 1.18 ± 0.24 mg/d of lorazepam)and 11 of 33 patients in the metyrapone group (mean ± SEMdose, 1.47 ± 0.28 mg/d of lorazepam) received lorazepam.No association between administration of lorazepam and treatment responseemerged.

Major results of this study were that the addition of metyrapone toa standard serotonergic antidepressant treatment led to a significantly betteroverall treatment outcome and a significantly more rapid onset of therapeuticaction. These beneficial effects outlasted the 21-day treatment period independentlyof the serotonergic antidepressant used, ie, nefazodone or fluvoxamine.

Our study is the first placebo-controlled, double-blind trial of metyraponeused as an augmentation agent in the treatment of patients with major depression.Our findings validate the results of an earlier open-label study²⁷ andthe only other placebo-controlled, single-blind crossover study³³ onmetyrapone. This latter study is not directly comparable to ours because 30mg/d of hydrocortisone, which itself has psychotropic effects, was co-administered.Antidepressant effects were also observed applying the cortisol-synthesisinhibitor ketoconazole in a controlled study with 20 depressed patients,³⁴ where merely 3 hypercortisolemic patients improved.Malison et al³⁵ found only limited efficacyin treatment-refractory major depression. In our substantially larger sample,metyrapone enhanced antidepressant efficacy independently of eucortisolemicor hypercortisolemic basal morning plasma cortisol concentrations.

As expected, the major findings were that metyrapone intervened withcortisol synthesis, inducing marked increases of plasma ACTH, the cortisolprecursor 11-deoxycortisol, and the neurosteroid DHEA. During metyrapone treatment,we did not observe significant decreases in basal morning plasma cortisolconcentrations, which is in line with the findings of Raven et al.³⁷ These observations can be explained by the dosageregimen, administering metyrapone during daytime until the nadir of cortisolsecretion, when HPA-axis feedback mechanisms are especially responsive. Aconsiderable rebound of plasma cortisol concentrations occurs in the morninghours, driven by the elevated ACTH concentrations. During placebo, both antidepressants,which are known to interact with the Cytochrom P450 system, increased ACTHand 11-deoxycortisol concentrations compared with baseline, while cortisolfor the whole group and DHEA remained largely unchanged. However, no differentialinteraction of both antidepressants with metyrapone was detected.

No apparent correlation emerged between basal plasma cortisol concentrationsbefore starting metyrapone treatment and the overall outcome. The respondersin the metyrapone group had insignificantly higher cortisol concentrations,but morning cortisol concentrations show a large variability and are a weakmeasure of overall HPA-axis activity. Patients showing improvement with metyraponedeveloped larger increases of ACTH and 11-deoxycortisol compared with nonresponders,although this effect did not reach significance. One explanation would bethat patients taking metyrapone improved according to the extent of steroid-synthesisinhibition. However, leaner patients who received higher average doses ofmetyrapone per kg showed the same differences in treatment response, and nocorrelation with the body weight index emerged (data not shown). Hence, ourfindings in the responder faction may reflect a more sensitive hypothalamicor hippocampal feedback, leading to a reset of the circadian rhythm of theHPA axis or unmasking of central CRH overdrive.³²

Metyrapone also significantly increased the plasma concentrations ofthe cortisol precursor 11-deoxycortisol, which itself has psychotropic effects,⁵² and of the neuroactive steroid DHEA, which exertsantiglucocorticoid, anxiolytic, and antidepressive actions.^{53 - 56} Neuroactivesteroids act on γ-aminobutyric acid A and progesterone receptors inhumans.⁵⁷ Metyrapone induces profound long-termchanges of synthesis and concentrations of such steroids, and a correlationbetween a reduction in MADRS scores and an increase of neuroactive urinaryand plasma steroids has been found.³⁷ WhereasRaven et al³⁷ did not find changes in the urinarysecretion of DHEA, we observed significant elevations of plasma DHEA.

Although the antidepressive and neuroendocrine effects of metyraponeare clearly demonstrated by our study, the mechanisms of actions are stillfar from elucidated. Besides the reset of HPA-axis activity and the enhancedrelease of neuroactive steroids, we surmise that alternative mechanisms cooperateto yield the therapeutic effects. Steroid actions and feedback regulationare determined by their tissue concentrations, which are largely regulatedby prereceptor metabolism. The enzyme 11β-HSD regulates the steroid accessto receptors, catalyzing the conversion of active 11-hydroxy-glucocorticoidslike cortisol into their inactive keto-forms. One form, the 11β-HSD-1,is widely expressed in the hypothalamus, hippocampus, cerebellum, and neocortex,³⁶ maybe acting as a tissue-specific modulator of glucocorticoidaction.⁵⁸ This enzyme’s reaction is bidirectional,and in intact neurones the reduced nicotinamide-adenine dinucleotide phosphate(NADPH)–dependent reduction of 11-deoxysteroids is the predominant reaction,regenerating active glucocorticoids in neuronal target cells.⁵⁹

Whereas ketoconazole mainly blocks steroid synthesis at the adrenallevel, metyrapone crosses the blood-brain barrier⁶⁰ andinhibits the conversion of the endogenous precursor 11-deoxycortisol to cortisolboth at the adrenals and in the brain, acting on the 11-beta-hydroxylase.In addition, metyrapone blocks the 11-oxoreductase activity of the 11β-HSD-1³⁸ either directly or indirectly by increased formationof endogenous inhibitors, eg, progesterone-derivates.⁶¹ Thus,metyrapone is able to decrease cortisol concentrations in the brain independentlyof circulating steroid levels. This potential to decouple central nervoustissues from peripheral steroid concentrations⁶² hasbeen demonstrated in humans,⁶³ and behavioralconsequences of altered hippocampal 11β-HSD-1 activity were recentlyshown in an animal model.⁶⁴

A specific site of action of metyrapone could be the hippocampus, where11β-hydroxylase and 11β-HSD-1 are colocalized with MR.^{36 ,65} Inhibitionof cortisol synthesis in this region would deplete MR from their ligands,and MR should subsequently be up-regulated. This regulatory sequel can bedemonstrated for many antidepressant drugs, which increase the binding capacityand gene expression of MR in the hippocampus in the rat.^{66 - 68} Increasesof hippocampal MR levels precede the decrease in CRH messenger RNA in thehypothalamic paraventricular nucleus⁶⁹ andthe readjustment of the HPA-axis activity.^{68 ,70} Complementarily,the MR antagonist spironolactone hampered antidepressive effects of amitriptylinein humans.¹⁵ Therefore, metyrapone might assista faster restoration of MR function, which accelerates the attainment of newallostatic equilibrium.¹⁶

The reduced occupation of GR and MR by cortisol in the hippocampus couldconsecutively accelerate the up-regulation of 5HT-1A receptors, which areessential for the action of serotonergic antidepressants.¹⁴ Areduced availability of cortisol in the central nervous system, especiallyduring the evening, could directly promote the reset in positive feedbackloops. Such cortisol-dependent circuits are demonstrated in several brainareas, like the amygdala and hypothalamus^{71 - 73} andnotably the hippocampus.⁷⁴

Furthermore, the neogenesis of neurones within the hippocampus may beinvolved in the etiology of depression.⁷⁵ Preliminaryexperiments show that metyrapone increases the number of new cells in thegyrus dentatus of the hippocampal formation in mice,⁷⁶ maybepartly mediated by neurosteroids like DHEA.⁷⁷ Moreover,metyrapone induces c-fos expression in limbic regions.⁷⁸ Again,a restoration of hippocampal feedback onto HPA-axis activity could be assisted.

Several mechanisms are potentially responsible for the antidepressantefficacy of metyrapone. Shortcomings of our study were that HPA-axis activitywas characterized merely by morning cortisol concentrations. Seemingly eucortisolemicpatients frequently show other signs of HPA dysregulation in depression.¹⁵ More sensitive tests, like the combined dexamethasonsuppression test (DST)/CRH challenge, would allow a more refined analysisof hidden perturbations. The wash-out phase was short because our severelyill patients needed acute treatment.

Although steroid-synthesis inhibitors are not quite ready for routineclinical application, the findings of this study clearly warrant further studiesaimed at identifying subgroups of depressed patients who will benefit mostfrom this approach and surrogate markers to find the optimal dose regimen.

Correspondence: Holger Jahn, MD, Departmentof Psychiatry and Psychotherapy, University Hospital Hamburg-Eppendorf, Martinistr.52, 20246 Hamburg, Germany (jahn@uke.uni-hamburg.de).

Submitted for Publication: March 28, 2003;final revision received January 6, 2004; accepted June 10, 2004.

Funding/Support: Dr Schick was supported bya fellowship of the Graduiertenkolleg 255 (Deutsche Forschungsgemeinschaft,Bonn, Germany).

Acknowledgment: We thank Sebastian Schultz,PhD, and Michael Baehr, PhD (Pharmacy of the University Hospital of Hamburg,Hamburg, Germany) for preparation, randomization, and coding of the studymedication and Christoph Hiemke, PhD (University of Mainz, Mainz, Germany)for measurements of the plasma concentrations of antidepressants. We are especiallygrateful to Bernhard Menke, PhD (Novartis, Nuremberg, Germany) for the generoussupply of metyrapone.