ORIGINAL INVESTIGATION

Rapid tryptophan depletion following cognitive behaviouraltherapy for panic disorder

Caroline Bell & Sean Hood & John Potokar & Jon Nash &

Mariona Adrover & Chris Frampton & Dana Hince &

Ann Rich & Spilios Argyropoulos & David Nutt

Received: 13 May 2009 /Accepted: 23 September 2009 /Published online: 13 October 2009# Springer-Verlag 2009

AbstractObjective The aim of this study was to examine the effectof rapid tryptophan depletion (RTD) combined with apanicogenic challenge in patients with panic disorder whohad responded to treatment with cognitive behaviouraltherapy (CBT). We hypothesised that RTD (compared withthe control drink) would result in an increase in anxietysymptoms when provoked by a panicogenic challenge withthe benzodiazepine antagonist, flumazenil.Methods Nine patients with panic disorder who hadresponded to CBT received a tryptophan-free amino aciddrink on one occasion and a control drink on the other in a

double-blind crossover design. In addition, they receivedflumazenil and placebo infusions on each day.Results Our hypothesis regarding the effects of RTD wassupported by findings of a significant interaction betweenRTD and flumazenil on measures from visual analoguesscales (total) and the Spielberger State Anxiety inventory. Asomewhat unexpected finding was that in this group ofCBT responders, the panicogenic effect of flumazenil wasnot completely blocked by treatment. This meant thatalthough four of the nine subjects (44%) reported apanicogenic effect of flumazenil on the RTD day, this wasnot significantly different from the rate of panic attacks inresponse to flumazenil on the control day.Conclusion We suggest that the partial return of symptomsin response to flumazenil reflects a vulnerability to RTD inthis group of panic disorder patients who had responded totreatment with CBT.

Keywords 5-HT. Panic disorder . Tryptophan depletion .

Cognitive behavioural therapy

Introduction

It is proposed that panic disorder involves the activation ofa complex fear circuitry (Gorman et al. 2000). Effectivetreatments for panic disorder include both antidepressantsand cognitive behavioural therapy (CBT; American Psychi-atric Association for the Work Group on Panic Disorder2009), which would appear to have very different modes ofaction. Although theories involving the mechanism ofaction of these treatments have tended to emphasis eithera biological or cognitive model, it is likely that the two arenot mutually exclusive. Current theories propose that thereis a biological abnormality (involving a hypersensitive fear

C. Bell : C. FramptonUniversity of Otago,Christchurch, New Zealand

J. Potokar : J. Nash :A. Rich : S. Argyropoulos :D. NuttPsychopharmacology Unit, University of Bristol,Bristol, UK

S. Hood :D. HinceSchool of Psychiatry & Clinical Neurosciences,University of Western Australia,Perth, Australia

M. AdroverLong Stay Unit, Hospital Psiquiatrico (GESMA),Palma de Mallorca, Spain

D. NuttImperial College London,London, UK

C. Bell (*)Department of Psychological Medicine,University of Otago, Christchurch,P.O. Box 4345, Christchurch, New Zealande-mail: caroline.bell@otago.ac.nz

Psychopharmacology (2011) 213:593–602DOI 10.1007/s00213-009-1696-z

network) that predisposes the individual to have distortedcognitive parameters and to the generation of panic, andthat effective treatment strategies using medication or CBTintervene in different ways or at different parts of the fearcircuitry (Gorman et al. 2000; Sakai et al. 2006). Thistheory has been developed to propose that CBT worksthrough cortical “top down” effects, whereas antidepres-sants are proposed to work from a subcortical or “bottomup” effects (Goldapple et al. 2004).

Although the neurobiological basis of panic disorder andits treatments are not fully understood, a role for serotonin(5HT) has long been proposed (Deakin and Graeff 1991).We have previously investigated the involvement of 5HT inpanic disorder using the technique of rapid tryptophandepletion (RTD) in a group of patients who had respondedto treatment with selective serotonin reuptake inhibitors(SSRIs) and reported that RTD results in an increase inpanic symptoms in response to a panicogenic challenge(Bell et al. 2002). These results of a relapse in symptomswith RTD are similar to those reported by studies ofpatients with depression, social anxiety disorder and PTSDwho had responded to treatment with SSRIs (Argyropouloset al. 2004; Bell et al. 2005; Booij et al. 2002; Corchs et al.2009). These findings have been proposed to reflect either arole for 5HT in the mechanism of therapeutic effect of thesedrugs or to a particular 5HT vulnerability in these patients(Booij et al. 2002). Interestingly, studies of patients withobsessive-compulsive disorder (OCD) who had respondedto treatment have reported that RTD does not result in arelapse of symptoms (Barr et al. 1994; Berney et al. 2006).

The effects of RTD in subjects who have responded toCBT have been less investigated than responders tomedication. Two studies of RTD in patients with depressionwho had responded to treatment with CBT reported thatRTD did not reverse the effects of CBT (Haynes et al.2004; O’Reardon et al. 2004). A study of RTD in patientswith OCD who had responded to either CBT or SSRIsreported that RTD did not result in a relapse of obsessionalsymptoms in patients treated by CBT (although, asdescribed above, the SSRI responders also failed to relapsesuggesting that RTD may have different effects in OCDcompared with other disorders; Berney et al. 2006). Therehave been no other studies to date investigating the effectsof RTD in CBT-treated anxiety disorders.

The aim of this study was to examine the effect of RTDin patients with panic disorder who had responded totreatment with CBT. Previous studies in panic disorder havesuggested that panic symptoms are often only reported inthe experimental setting when provoked by a panicogenicchallenge, a finding that has been confirmed by RTDstudies in untreated (Miller et al. 2000; Schruers et al.2000) and treated panic disorder (Bell et al. 2002). In viewof this, we therefore replicated the design of our previous

study (in patients with panic disorder who had responded totreatment with SSRIs) and combined RTD with a pharmaco-logical panic challenge using the benzodiazepine antagonist,flumazenil, given at the time of peak tryptophan depletion.Wehypothesised that RTD (compared with the control drink)would result in an increase in anxiety symptoms whenprovoked by flumazenil.

Methods

Subjects

We studied ten patients (four females; mean age 33.2 years,range 20–43 years) who had responded to treatment withCBT and had remained well for at least 12 weeks. Allpatients were recruited from an outpatient anxiety clinicspecialising in the treatment of anxiety disorders. Referralsto this clinic are typically received from community mentalhealth teams and general practitioners, as well as self-referrals in response to local media awareness campaigns.The patients for this study had been diagnosed with aprimary diagnosis of panic disorder according to Diagnosticand Statistical Manual of Mental Disorders (DSM) IVcriteria assessed by an experienced clinician (CB, JN, SH,MA, and JP) and by means of a semi-structured interview(Mini International Neuropsychiatric Interview; Sheehan1998). They had chosen a psychological (rather thanmedication) treatment for their panic disorder and weretreated in the outpatient clinic. Treatment involved 12weekly sessions of CBT in the first 3 months followed bythree booster sessions over the next 3 months. CBTfollowed a standard protocol (Clark et al. 1994) and wasdelivered by a psychologist or a nurse therapist who hadgained a postgraduate diploma in cognitive therapy andwho had regular clinical supervision. Only subjects whohad responded to treatment were included in the study.Response was defined by subjects showing a very good orgood improvement (score 1 or 2) on the Clinical GlobalImpressions Improvement Scale (Guy 1976) which hadbeen maintained for a minimum of 12 weeks. These criteriawere the same as used to define clinical response in ourprevious study of RTD in SSRI responders (Bell et al.2002). Subjects were taking no psychotropic medicationand had not received any benzodiazepines in the 12 weeksprior to the test days. Participants also had no activesubstance abuse disorder (alcohol or illicit drugs).

The clinical characteristics of the patients included in theRTD study are summarised in Table 1. The frequency ofpanic attacks when unwell i.e. prior to treatment rangedfrom 1–28 per month (mean 12). One patient reportedsevere agoraphobia prior to treatment. Four patients had aprevious history of major depression. None of the patients

594 Psychopharmacology (2011) 213:593–602

had a history of suicidal thoughts or deliberate self-harm.Two patients had a family history of panic disorder, one ofOCD, and one of depression. The duration of illness priorto treatment was wide ranging from 4 to 156 months (mean52.3 months). The mean duration of treatment was6.6 months (range 5–8 months), and the mean number ofCBT sessions completed was 12.3 (range 10–18).

In the week before the first test day, all subjects werecomprehensively screened to ensure physical fitness. Thenature of the project was described to them as an attempt tounderstand the role of tryptophan and 5HT in anxiety. Theywere made aware that the process of RTD or thepanicogenic challenge used i.e. flumazenil might cause atransient increase in anxiety and/or depression. They werealso informed of the possible side effects of flumazenil.Subjects were aware that they could withdraw from thestudy at any time without this decision influencing theirtreatment. All subjects gave written informed consent, andthe study was approved by the local ethics committee.

The recruitment for this study was extremely prolonged.Forty-two participants were enrolled in the initial CBTtreatment phase of the study and were informed at the startof this process about the RTD study at the end of treatment.There was however marked attrition over the course ofCBT treatment with only ten participants beginning theRTD phase. The reasons for not entering the RTD phaseincluded eight patients declining to participate despite beingeligible responders, 13 not completing CBT treatment (rangetwo to nine CBT sessions completed), one patient relapsingand being prescribed medication, and ten being lost to follow-up despite completing CBT. There was however no significant

difference between those patients that did and did not enter thechallenge phase on baseline measures of severity of illness(Clinical Global Impressions scale), anxiety (State-TraitAnxiety Inventory (STAI)) or depression (Beck DepressionInventory (BDI); p > 0.1), or with reference to the number ofCBT sessions completed.

Procedure

RTD was performed according to a double-blind, placebo-controlled, crossover design with each patient taking part intwo similar test days 1 week apart. The order of the RTDand control tests was randomly allocated by unweightedblocking, with allocation concealed until completion of thestudy.

Each test day was conducted according to the followingprotocol. Subjects followed a low tryptophan (TRP) diet onthe day prior to the study (as in Delgado et al. 1990) andfasted from midnight before attending the research clinic at9:00 a.m. An intravenous cannula was then inserted into anantecubital vein, and this kept patent using heparinisedsaline. Subjects rested in a semi-supine position on a bedand were only allowed to consume water for the duration ofthe study. Prior to consuming the amino acid drink, a bloodsample was taken to measure baseline plasma TRPconcentration, and questionnaires were completed to mea-sure baseline anxiety and mood symptoms. Subjects thenconsumed an amino acid drink that either contained 2.3 gof TRP (control day) or not (RTD day) (as described in Bellet al. 2002). Women were only required to drink 80% ofthe above amounts because of their lower body weight inaccordance with the protocol of Smith et al. (1997).

Psychological ratings (detailed below) were recordedevery 60 min for the first 240 min of the study i.e. prior toany infusions. At 270 and 330 min after consumption of theamino acid drinks (i.e. around time of expected peakplasma TRP depletion), patients received an intravenousinfusion as a bolus over 1 min of flumazenil (Annexate-Roche, UK; 2 mg/20 ml) or normal saline (20 ml).Psychological ratings in response to these were then againrecorded (Fig. 1). The order of the infusions was alsorandomly allocated by unweighted blocking, with the orderbeing kept constant for each patient for the two test days.

At the end of each test day, patients were given a mealand after having been assessed as recovered were allowedhome. Subjects were given access to telephone contact ifrequired over the following 24 h.

Two physicians supervised each test day. Both wereblind to the composition of the amino acid drink, whichwas made up by the research assistant. The physicianadministering the questionnaires was also blind to whichinfusion was given; the other physician who prepared theinfusions was not.

Table 1 Clinical characteristics of participants

CBT responders(n=10)

Age (years) 33.2a (20–43)b

Gender 4F/6M

Panic disorder

Panic attacks per month pre-treatment 12a (1–28)b

Duration of illness (months) 52.3a (4–156)b

Duration of treatment (months) 6.6a (5–8)b

Number of CBT sessions completed 12.3a (10–18)b

Comorbidity

BDI pre-treatment 10.6a (4.9)c

BDI post-treatment (i.e. at time of RTD) 5.8a (3.8)c

Number subjects with lifetime historyof depression

4

Number subjects with family history ofdepression

1

aMeanb Rangec Standard deviation

Psychopharmacology (2011) 213:593–602 595

Measures

Psychological measures

Rating scales for subjective anxiety symptoms consisted ofthe visual analogue scales (VAS) of anxiety, tachycardia,tremor, flushing, dizziness, nausea, sleepiness, alertness andfeeling tearful, hopeless, sad or depressed on a 0–100 mmline; the Panic Symptom Inventory (PSI), which ratestypical somatic and psychological symptoms of anxiety ona 0–4 severity scale; and the Spielberger State-Trait AnxietyInventories (SSAI and STAI; Spielberger et al. 1970), as anindicator of transitory changes to anxiety. Two morecognitive scales were also used as a test of the cognitivemodel of anxiety which would predict that the patient’stendency to catastrophically misinterpret interoceptive cues,as measured by these questionnaires, would be associatedwith a fearful response to the challenge (in this case, toflumazenil). These included the Agoraphobic Cognitionsquestionnaire (ACQ) which is a 14-item scale comprisingtwo seven-item factors concerning thoughts (a) of physicalcatastrophe due to anxiety symptoms (e.g. “I am going tohave a stroke”) or (b) of mental, social or behaviouraldisaster from loss of control (e.g. “I am going to babble ortalk funny” or “I am going to go crazy”). Items are rated onfive-point scales ranging from “thought never occurs whenI’m anxious” to “thought always occurs when I’m anxious”(Chambless et al. 1984).The BSQ is a 17-item scaleconcerning the degree to which patients fear somaticsymptoms commonly associated with panic e.g. dizziness.Items are rated on five-point scales ranging from “notfrightened or worried by this sensation” to “extremelyworried by this sensation” (Chambless et al. 1984).Depressive symptoms were measured by the BDI (Beck etal. 1961) and by the unhappy and tearful/sad items of thePSI and by VAS scores of feeling tearful, hopeless, sad ordepressed.

The rating scales were modified for the test days to beused as state measures in that subjects were asked to rate

how they felt at that time rather than over e.g. the last2 weeks. At baseline (i.e. 0 min), ratings of anxiety anddepression were recorded using the VAS, PSI, STAI, SSAI,ACQ, BSQ and BDI. VAS scores were measured hourlyuntil the infusions. The SSAI, PSI, ACQ and BSQ ratingswere recorded again at 240 min which was 30 min beforethe first infusion. At 5, 15, 30 and 45 min after eachinfusion, patients were asked to rate the peak intensity oftheir symptoms using the VAS. Fifteen minutes after eachinfusion, patients were asked to rate the peak symptomresponses since the infusion on the PSI. Thirty minutesafter each infusion, patients completed the SSAI againrating the peak response since the infusion. Forty-fiveminutes after each infusion, patients completed the ACQ,BSQ and BDI again rating the peak response since theinfusion.

The primary psychological outcome variables were theVAS total and VAS anxiety scores (as these havepreviously been reported to be the best scales to measurereactivity to a panicogenic challenge; Battaglia and Perna1995). Because of the difficulties in assessing laboratory-induced panic attacks, we applied two definitions of whatconstituted a panic attack. These were rated from the PSIand were as used in previous studies with flumazenil (Bellet al. 2002; Coupland et al. 1997, 2000). The first definedthe occurrence of a panic attack when there was anincrease in the anxiety or fear item of the PSI of ≥2,together with an increase in score of ≥2 in a least fourother DSM panic symptoms (definition 1). A lowerthreshold anxiety criterion was also used (in view of thefact that these patients were responders to treatment)which required an increase of ≥1 in both anxiety or fearitems and at least 4 other DSM panic symptoms(definition 2).

Plasma tryptophan levels

Blood samples for measurement of total and free plasmaTRP and other large neutral amino acids (LNAA) were

Amino acid drink Infusion 1 Infusion 2

Home

0 60 120 180 240 270 300 330 360 390

Time (minutes)

Baseline TRP levels TRP levels TRP levels

Fig. 1 Study design

596 Psychopharmacology (2011) 213:593–602

taken just before patients consumed the amino acid drinksand immediately prior to each infusion (i.e. at 0, 260 and320 min). The samples were collected in lithium heparintubes and centrifuged for 10 min at 2,400 rpm and 4°C. Theseparated plasma was frozen and stored at −20°C until itwas analysed. Plasma tryptophan was measured in plasmaby high-performance liquid chromatography (HPLC) withfluorescence detection. For total tryptophan, the plasma wasfirst deproteinised with trichloroacetic acid before HPLCanalysis, whereas for the free measure, the plasma protein-bound tryptophan was separated by use of 12 K cut-offfilters by centrifugation.

Data analysis

The effect of RTD on plasma tryptophan levels wasexamined using repeated-measures analysis of variance(ANOVA) with condition (depleted/non-depleted) and time(0, 260 and 320 min) as the within-subject factors. Baselinevariables between the RTD and control days in both studieswere compared using t tests. Two time periods were thenanalysed—the pre-challenge period (0–270 min) and thepost-challenge (infusion) period (270–375 min).

In the pre-challenge period, the effects of RTD on VAS,PSI, SSAI, ACQ and BSQ scores before the first infusionof flumazenil/saline were examined using repeated-measures ANOVA with condition (RTD/control) and timeas the within-subject factors.

The post-challenge period addressed the central questionof whether or not RTD altered patients’ responses toflumazenil. The primary outcome variables were the post-infusion VAS total and anxiety ratings. The secondaryoutcomes were the post-infusion PSI, SSAI, ACQ and BSQscores. The analysis was by repeated-measures ANOVAwith three within-subject factors of condition (RTD orcontrol), infusion (flumazenil or saline) and time. Thenumber of panic attacks (as defined above) followingflumazenil/saline was compared on the RTD and controldays using the McNemar test. Depressive symptoms asmeasured by the BDI were analysed by repeated-measuresANOVA.

Correlations between anxiety measures at baseline andpre-infusion with an anxiety response post-infusion wereinvestigated by examining the effect of the challenge(flumazenil, placebo) for each subject for each of theanxiety measures. In addition, in order to investigate theassociation between the level of panic-related cognitionspost-treatment and fearful responding to the challenge, weinvestigated the correlation between the mean BSQ andACQ scores at baseline and pre-infusion with the effect ofthe challenge (flumazenil, placebo) for each subject foreach of the anxiety measures. These correlations weretested by Spearman’s rho correlation coefficient.

The between-subject effects of the order of the depletionand the challenge were assessed.

SPSS 11.0 for Windows was used for the analysis.

Results

One person failed to complete the challenge phase becauseof an adverse response (an episode of bradycardia andsyncope) to the infusion of flumazenil on the control day.Subsequent to the study, the patient disclosed that he hadsuffered previous episodes of syncope during panic attacks,and we concluded that this patient fell into the smallcategory of patients who suffer vasovagal episodes duringpanic attacks. He had not reported this upon specificquestioning in the screening interview. This subject hadonly completed the pre-challenge phase of the control day,and his data from the challenge days were thereforeexcluded from further analysis because there was too muchmissing data.

This meant that the number of subjects in the RTD phaseof the study was nine.

Plasma tryptophan levels

There were no significant differences between the RTD andcontrol days in baseline measures of plasma-free or totalTRP or the free TRP/LNAA ratio. Significant depletion×time interactions occurred for total TRP (F = 74.6,df = 2,16, p < 0.001), and free TRP levels (F = 8.3,df = 2,16, p = 0.003) with similar but not statisticallysignificant trends for the free TRP/LNAA ratio. Afteringestion of the TRP-depleted mixture, there was a meanfall in total (and free) plasma TRP of 83% (75%) and 83%(72%) at 260 and 320 min, respectively. There was asimilar fall in free plasma TRP/LNAA ratio of 90% and88% at 260 and 320 min, respectively. This was signifi-cantly different from the increases seen after the controldrink of 129% (126%) and 63% (85%) in total (and free)plasma TRP at the same time points. The control drink alsocaused a reduction (but much less than the RTD condition)in the free plasma TRP/LNAA ratio of 37% and 29% at thesame time points.

Baseline ratings (before amino acid drink)

Baseline psychological ratings reflected the fact that thesepatients had responded to treatment. They had a mean SSAIscore of 35.1, which compares with scores of 52.0–54.7 forpatients with untreated panic disorder, 31.2–41.0 for patientswith treated panic disorder and 29.2–30.8 for normal controls(Gorman et al. 1997; Miller et al. 2000; Nutt et al. 1990). Themean scores from the ACQ (1.7), BSQ (1.43) and BDI (6.5)

Psychopharmacology (2011) 213:593–602 597

were in the range of healthy controls. There were nodifferences in baseline ratings of any of the measures usedbetween the depletion and control days.

Pre-challenge period (0–270 min)

The VAS total, VAS anxiety, SSAI, PSI, ACQ and BSQscores showed no significant effects of tryptophan depletionover this time period.

Flumazenil/saline challenge period (270–375 min)

The VAS total score post-challenge showed significant effectsof flumazenil (F = 6.7, df =1,8, p = 0.03) and a significantflumazenil×depletion interaction (F = 6.4, df=1,8, p = 0.03;Fig. 2). VAS scores for anxiety showed a significant effect offlumazenil (F = 6.4, df = 1,8, p = 0.03) and a trend for aflumazenil×depletion interaction (F = 3.9, df = 1,24,p = 0.08). Analysis of the other VAS scores post-challengeshowed significant effects of flumazenil on ratings offlushing (F = 18.6, df = 1,8, p < 0.005), dizziness (F = 9.0,df = 1,8, p < 0.05), alertness (F = 9.3, df = 1,8, p < 0.001),sleepiness (F = 6.8, df = 1,8, p < 0.05) and tremor (F = 5.2,df = 1,8, p = 0.05), but no significant interactions.

The SSAI scores showed a significant effect of flumazenil(F = 4.6, df = 3,24, p = 0.01) and a significant TRPdepletion×flumazenil interaction (F = 3.1, df = 3,24,p = 0.04; Fig. 2). The total PSI scores showed a significanteffect of flumazenil (F = 7.2, df = 3,24, p = 0.001), but nosignificant interaction between TRP depletion and the

flumazenil challenge (F = 1.5, df = 3,24, p = 0.24; Fig. 2).The ACQ scores showed no significant main or interactioneffects. The scores from the BSQ showed a significant effectof flumazenil (F = 3.35, df = 3,24, p = 0.04), but nosignificant interactions.

Flumazenil produced a panic attack (defined by the strictcriteria above) in four of the nine patients (three femalesand one male) when tryptophan depleted and two subjects(one female and one male) on the control day (Fig. 3).Thesedifferences were not significant (p = 0.625). When the lessrestrictive definition (described above) was applied, fourpatients reported a panic attack with flumazenil on the RTDday with four on the control day. No patients reported panicsymptoms with both infusions on the RTD day. One subjectpanicked with the placebo infusion on the control day. Inthree patients, the infusions had no effect on either day.

There were no evident effects of the order of either thedepletion or infusion.

Correlations

There were no significant correlations between anxietyresponses after the challenges and baseline and pre-infusionmeasures. Similarly, no significant correlations were foundwhen subjects were limited to the four panickers.

Mood effects

There were no significant effects of RTD on mood asmeasured by changes to the BDI, the unhappy and tearful/

0

50

100

150

200

250

300

350

0mins

60mins

120mins

180mins

240mins

5mins

15mins

30mins

45mins

5mins

15mins

30mins

45mins

Mea

n V

as T

ota

l Sco

res

Control RTD

Saline infusion

25

35

45

55

0 mins 240 mins After Saline AfterFlumazenilM

ean

Sp

ielb

erg

er S

tate

Sco

res

Control RTD Control RTD

Flumazenil infusion

05

1015202530354045

0 mins 240 mins After Saline AfterFlumazenilM

ean

Pan

ic In

ven

tory

Sco

res

Fig. 2 Psychological ratingsbefore and after flumazenil/saline challenge

598 Psychopharmacology (2011) 213:593–602

sad items of the PSI and by VAS scores of feeling tearful,hopeless, sad or depressed.

Discussion

The aim of this study was to investigate the serotonergicmechanisms involved in panic disorder patients who hadbeen successfully treated by CBT using the same RTDparadigm as used in a similar study of panic disorderpatients who had responded to treatment with SSRIs (Bellet al. 2002). We had hypothesised that RTD (compared withthe control drink) would result in an increase in anxietysymptoms when provoked by flumazenil. This hypothesiswas partially supported by findings of a significantinteraction between RTD and flumazenil on the primaryoutcome variable, the VAS total score post-infusion but notby the VAS anxiety scores (although there was a trend inthis direction). This effect was also shown by the significantinteraction between RTD and flumazenil on the post-infusion SSAI scores but not by the PSI. As shown inFig. 2, the increases on the VAS total and SSAI wereclinically significant and considerable by psychometricstandards. Four of the nine subjects (44%) did report apanicogenic effect of flumazenil on the RTD day, althoughthis was not significantly different from the rate of panicattacks in response to flumazenil on the control day. Thelack of significance of this finding is likely to be explainedby the higher than expected frequency of panic attacks inresponse to flumazenil on the control day. The finding ofpartial relapse of symptoms in a proportion of subjects withRTD in this study is similar to the effects of RTD inpatients with depression who had responded to treatment inwhich about 50% of subjects are reported to experience a

relapse of symptoms in response to RTD with only 20% ofsubjects reporting a full relapse (Booij et al. 2002).

The results from this CBT responder study are slightlydifferent from our previous study in SSRI responders withpanic disorder where we found RTD resulted in asignificant increase in the categorical definition of panicattacks in response to flumazenil compared with the controlday (when no patients panicked in response to flumazenil).In this study, there were, however, no significant inter-actions on other anxiety ratings i.e. the VAS, SSAI and PSI.The results we suggested reflected a vulnerability to RTD inthis group of patients, which may reflect a role for 5HT inmaintaining the response to SSRIs. The results from thisCBT-treated group of patients, although different, alsosuggest that RTD increases some measures of anxiety inresponse to a panicogenic challenge i.e. the VAS total andSSAI. This could be interpreted in two ways. It is possiblethat the RTD response reflects a role for 5HT in thetherapeutic response to CBT. CBT has been shown to resultin neurobiological changes although the mechanismsinvolved and whether the effects involve 5HT have notbeen fully determined. Recent neuroimaging studies ofpatients with panic disorder before and after treatment withCBT have reported that CBT results in adaptive metabolicchanges in medial prefrontal (Sakai et al. 2006) andtemporal regions (Prasko et al. 2004), which fits withtheories of CBT treatment acting through cortical “topdown” effects on a hypersensitive fear network. It ispossible that serotonergic processes may be involved inthis, although this has not been established. The other, andwe feel more likely, explanation of the anxiogenic effects ofRTD in some treated panic disorder patients is that thesesubjects have a vulnerability to changes to 5HT functionand that RTD uncovers or activates a neural circuit, as haspreviously been suggested to explain the effects of RTD intreated depression (Booij et al. 2002; Neumeister et al.2006). RTD would therefore, in reducing 5HT function,result in a more activated anxiety network, and this wouldexplain the results in both the SSRI and CBT responderstudies.

The anxiety response to RTD in this group of CBTresponders is interesting because two previous studies ofRTD in CBT-treated depressed patients have reportedminimal effects of RTD on mood symptoms (Haynes etal. 2004; O’Reardon et al. 2004). The authors of thesestudies have suggested that these findings suggestedresistance to RTD effects in patients who had respondedto CBT and implied that maintenance of therapeuticresponse in depressed patients treated with CBT is notdependent on an intact serotonin system. Interestingly, aneuroimaging study of CBT in depression (Goldapple et al.2004) reported changes to cortical functioning in theopposite direction to those in a study of panic disorder

0

1

2

3

4

5

6

7

8

9

10

RTD + Saline RTD +Flumazenil

Control Day +Saline

Control Day +Flumazenil

Nu

mb

er o

f S

ub

ject

s

Panic No Panic

Fig. 3 Number of subjects reporting panic attacks after flumazenil onrapid tryptophan depletion and control days

Psychopharmacology (2011) 213:593–602 599

patients who had responded to CBT (Sakai et al. 2006).This may reflect that CBT for different disorders with theemphasis on different cognitive changes may result indifferent changes to neural functioning, and this may alsoexplain the different response to RTD.

As discussed above, it was an unexpected but interestingfinding that the anxiogenic effect of flumazenil was seen onboth the RTD and control days. We had hypothesised thatthe panicogenic effect of flumazenil would be blocked byeffective CBT treatment as in previous studies whereeffective treatment of panic disorder by either drugs (TCAs,SSRIs or benzodiazapines) or CBT (two out of threestudies) results in a reduction in the sensitivity topanicogenic challenges with responses approaching thoseseen in normal controls (Gorman et al. 1997; Schmidt et al.1997; Shear et al. 1991). These studies also suggested thatthese effects were greater in patients who have responded totreatment. The results from our study, however, suggest thatCBT was less effective than we had expected in blockingthe panicogenic effects of flumazenil. The ratings of anxietyfrom the VAS, SSAI and PSI showed similar patterns withclinically significant anxiety responses to flumazenil beingseen on both RTD and controls days, although the meanvalues on the control days were consistently lower thanthose from the RTD days. Unfortunately, we do not havedata on the flumazenil response in this group of CBT-treated patients before therapy but if the current findings arecompared with those of Nutt et al. (1990) in untreatedpatients in which eight of ten patients panicked, this wouldsuggest that effective treatment of panic disorder with CBThas some although incomplete effect in blocking thepanicogenic effects of flumazenil.

The reports of CBT resulting in changes to reactivity e.g.reducing the anxiogenic response to panicogenic challengesare interesting. The mechanism of this change is not clearbut has been suggested to reflect a strengthening of corticalpathways restraining the fear network. Cognitive theories ofpanic suggest that panic symptoms develop because ofpatients’ catastrophic misinterpretation of somatic sensa-tions (Clark 1986). A key component of CBT for panicdisorder is to use interoceptive exposure exercises tochallenge these misinterpretations. Measures that assessbeliefs about the fear of anxiety and the harmfulness of bodilysensations (such as the ACQ and BSQ as used in this study)have generally supported the hypotheses reporting a correla-tion between both high scores on these measures and subjectshaving panic disorder and experiencing a panic response tochallenges (McNally et al. 1990; Rassovsky et al. 2000).Studies have also reported that treatment with eithermedication or CBT normalises scores on these measuresand reduces the likelihood of patients panicking in responseto a challenge (Mavissakalian et al. 1998). The patients inour study at the time of the RTD reflected the fact that they

had responded to treatment and had ACQ and BSQ scoressimilar to those seen in normal controls. Despite thisapparently full recovery, on the RTD and control days,flumazenil resulted in both anxiety and physical symptomse.g. flushing, dizziness and tremor, suggesting that CBT didnot block the provocation of these effects by flumazenil. Thiswas contrary to a previous study (Schmidt et al. 1997) whichreported that CBT markedly reduced CO2 induced panicsymptoms.

Some but not all (Perna et al. 2002) previous studies havesuggested that a panicogenic response to a challenge isassociated with the level of anxiety or panic-relatedcognitions at the time of the study (i.e. the patients tendencyto catastrophically misinterpret interoceptive cues; Miller etal. 2000; Schmidt et al. 1997). We did not find this, however.As in our previous study of SSRI responders (Bell et al.2002), we did not find a correlation between baseline or pre-challenge anxiety scores and a response to flumazenil. Wealso specifically investigated the association between thelevel of panic-related cognitions post-treatment using scoresfrom the ACQ and BSQ and fearful responding to thechallenge and found no significant correlations.

No significant mood effects of RTD were found in thisgroup of subjects. These findings are different from a studyof RTD in remitted OCD patients, which reported signif-icant mood effects with RTD (without significant worsen-ing of obsessive-compulsive symptoms; Berney et al.2006).

As previously reported by other RTD studies, somepatients were completely unaffected by RTD. In this study,four patients fell into this category (three males and onefemale). They could not distinguish the two test days, and alsoreported no significant difference between the flumazenil orplacebo infusions on either day. This lack of effect of RTD insome subjects has been reported in other disorders e.g.depression, and raises the intriguing question of what factorsdetermine a response to RTD. As discussed by Booij et al.(2002), it may be that these subjects are particularly resilientto 5HT manipulations. A lack of response to panicogenicchallenges has also been reported previously suggesting thatthis may reflect a subgroup of panic disorder patients whoare insensitive to the effects of biological challenges (Klein1993; Schmidt et al. 1997).

Ethical issues

The provocation of anxiety and depressive symptoms inrecovered panic patients potentially raises some importantissues. Patients who participated in the study all gaveinformed consent and, as part of this, were made aware thatthe process of TRP depletion or the chemical challengeused might cause a transient increase in anxiety. Weacknowledge that this may have introduced a cognitive

600 Psychopharmacology (2011) 213:593–602

bias to the study but made the decision that ethicallypatients needed to have this information to be able to giveconsent to take part in the study. The other concern was thatalthough the changes (in TRP and brain 5HT) produced bythe technique of TRP depletion were transient, the fact thatsome patients experienced panic and/or depression raisedthe concern that an anxiety or depression circuit maysomehow get reactivated. This did not happen, however. Inthe four patients who became anxious, the anxiety wasshort lived (i.e., maximum duration 10 min). In the onesubject who became depressed, although the duration ofthis state was longer (i.e. up to 90 min), she was assessed tobe fully recovered by the time she left the session. Weensured that the subsequent evening was spent with a carerand advised patients that they were able to contact theinvestigator by telephone at any time. Patients werereassessed the next day and at several weeks follow-upand in no case had the anxiety or depressive symptomsreappeared.

Limitations

The major limitation to this study is that our sample size issmall which limits the power of the study. A calculationfrom the non-significant VAS anxiety score (p = 0.08)would suggest that 18 subjects would have provided 80%power to detect a significant difference. We found recruit-ment for this study however to be very difficult whichresulted in the small numbers. The study duration was5 years with experienced researchers actively recruitingfrom an anxiety disorders clinic and primary care. Themajor problem involved people agreeing to take part in theRTD phase of the study with the rate of recruitment beingconsiderable slower than for the similar study in SSRIresponders. This may reflect a variety of factors includingthe perception from researchers that subjects who chose apsychological treatment modality were particularlyconcerned about the biological/invasive nature of the RTDprocedure. Other studies of RTD in CBT-treated subjectshave interestingly also noted marked attrition in recruitment(Haynes et al. 2004). Clearly, it would have been ideal tohave more patients, but this was not attainable despite ourbest efforts. The marked attrition from patients startingtreatment with CBT and agreeing and being suitable for theRTD part of the study also means that we are unclear howrepresentative our subjects were in comparison with CBT-treated patient in general. While the subjects agreeing toRTD did not differ from non-completers on either thebaseline demographic or psychological measures we tested,it cannot be ruled out that other factors may have beeninvolved. Subjects were also not randomly allocated toCBT (rather than another treatment option) but self-selectedto have this treatment modality, and we cannot therefore

exclude that subject preference for a psychological treat-ment might have had some effect on the subsequentresponse to RTD. The subjects tended to have a longduration of illness prior to starting CBT with one having aduration of 4 months, but all the others having a duration ofgreater than 2 years. This was not however descriptivelydifferent from our previous study in SSRI responders andemphasises the clinical message that even in subjects with along history of illness, a good clinical outcome can beachieved. The subject with a short duration of illness priorto CBT did not have and an anxiogenic response toflumazenil on the RTD, but the numbers are too small tocomment on any possible effects of duration of illness orthe brain mechanisms involved in this and RTD.

In conclusion, we have found that in subjects with panicdisorder who have responded to CBT, the panicogeniceffect of flumazenil is not blocked. Rapid reductions in5HT function by RTD resulted in a partial return of anxietysymptoms in response to flumazenil, which we suggestreflects a vulnerability to RTD in this group of patients ashas previously been suggested as the explanation for itseffects in treated or remitted depression.

Acknowledgements Wewould like to acknowledge study participants,CBT therapists Ruth Williams and Jo Hurst, Mike Franklin formeasurement of plasma tryptophan, David Wilcox and Hannah Slatterywho helped with data entry, the Wellcome Trust and Avon and WiltshirePartnership Trust.

References

American Psychiatric Association for the Work Group on Panic Disorder(2009) Practice guideline for the treatment of patients with panicdisorder (second edition). http://www.psychiatryonline.com/pracGuide/pracGuideTopic_9.aspx

Argyropoulos S, Hood SD, Adrover M, Bell CJ, Rich A, Nash JR,Rich NC, Witchel HJ, Nutt DJ (2004) Tryptophan depletionreverses the therapeutic effect of selective serotonin reuptakeinhibitors in social anxiety disorder. Biol Psychiatry 56:503–509

Barr LC, Goodman WK, McDougle CJ, Delgado PL, Heninger GR,Charney DS, Price LH (1994) Tryptophan depletion in patientswith obsessive-compulsive disorder who respond to serotoninreuptake inhibitors. Arch Gen Psychiatry 51:309–317

Battaglia M, Perna G (1995) The 35% CO2 challenge in panicdisorder: optimization by receiver operating characteristic (ROC)analysis. J Psychiatr Res 29:111–119

Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961) Aninventory for measuring depression. Arch Gen Psychiatry 4:561–571

Bell C, Forshall S, Adrover M, Nash J, Hood S, Argyropoulos S, Rich A,Nutt DJ (2002) Does 5-HT restrain panic? A tryptophan depletionstudy in panic disorder patients recovered on paroxetine. JPsychopharmacol (Oxf) 16:5–14

Bell CJ, Hood SD, Nutt DJ (2005) Acute tryptophan depletion. Part II:clinical effects and implications. Aust N Z J Psychiatry 39:565–574

Berney A, Sookman D, Leyton M, Young SN, Benkelfat C (2006)Lack of effects on core obsessive-compulsive symptoms of

Psychopharmacology (2011) 213:593–602 601

tryptophan depletion during symptom provocation in remittedobsessive-compulsive disorder patients. Biol Psychiatry 59:853–857

Booij L, Van der Does W, Benkelfat C, Bremner JD, Cowen PJ, FavaM, Gillin C, Leyton M, Moore P, Smith KA, Van der Kloot WA(2002) Predictors of mood response to acute tryptophandepletion. A reanalysis. Neuropsychopharmacology 27:852–861

Chambless DL, Caputo GC, Bright P, Gallagher R (1984) Assessmentof fear of fear in agoraphobics: the body sensations questionnaireand the agoraphobic cognitions questionnaire. J Consult ClinPsychol 52:1090–1097

Clark DM (1986) A cognitive approach to panic. Behav Res Ther24:461–470

Clark DM, Salkovskis PM, Hackmann A, Middleton H, AnastasiadesP, Gelder M (1994) A comparison of cognitive therapy, appliedrelaxation and imipramine in the treatment of panic disorder. Br JPsychiatry 164:759–769

Corchs F, Nutt DJ, Hood S, Bernik M (2009) Serotonin and sensitivityto trauma-related exposure in selective serotonin reuptakeinhibitors-recovered posttraumatic stress disorder. Biol Psychiatry66:17–24

Coupland NJ, Lillywhite A, Bell CE, Potokar JP, Nutt DJ (1997) Apilot controlled study of the effects of flumazenil in posttraumaticstress disorder. Biol Psychiatry 41:988–990

Coupland NJ, Bell C, Potokar JP, Dorkins E, Nutt DJ (2000)Flumazenil challenge in social phobia. Depress Anxiety 11:27–30

Deakin JF, Graeff FG (1991) 5HT and mechanisms of defence. JPsychopharmacol (Oxf) 5:305–315

Delgado PL, Charney DS, Price LH, Aghajanian GK, Landis H,Heninger GR (1990) Serotonin function and the mechanism ofantidepressant action. Reversal of antidepressant-induced remis-sion by rapid depletion of plasma tryptophan. Archives ofGeneral Psychiatry 47:411–418

Goldapple K, Segal Z, Garson C, Lau M, Bieling P, Kennedy S,Mayberg H (2004) Modulation of cortical-limbic pathways inmajor depression: treatment-specific effects of cognitive behaviortherapy. Arch Gen Psychiatry 61:34–41

Gorman JM, Browne ST, Papp LA, Martinez J, Welkowitz L, CoplanJD, Goetz RR, Kent J, Klein DF (1997) Effect of antipanictreatment on response to carbon dioxide. Biol Psychiatry 42:982–991

Gorman JM, Kent JM, Sullivan GM, Coplan JD (2000) Neuroana-tomical hypothesis of panic disorder, revised. Am J Psychiatry157:493–505

Guy W (1976) ECDEU Assessment manual for psychopharmacology.US Department of Health, Education, andWelfare, Washington, DC

Haynes PL, McQuaid JR, Kelsoe J, Rapaport M, Gillin JC (2004)Affective state and EEG sleep profile in response to rapid tryptophandepletion in recently recovered nonmedicated depressed individuals.J Affect Disord 83:253–262

Klein DF (1993) False suffocation alarms, spontaneous panics, andrelated conditions. An integrative hypothesis. Arch Gen Psychiatry50:306–317

Mavissakalian MR, Perel JM, Talbott-Green M, Sloan C (1998)Gauging the effectiveness of extended imipramine treatment forpanic disorder with agoraphobia. Biol Psychiatry 43:848–854

McNally RJ, Riemann BC, Kim E (1990) Selective processing ofthreat cues in panic disorder. Behav Res Ther 28:407–412

Miller HE, Deakin JF, Anderson IM (2000) Effect of acute tryptophandepletion on CO2-induced anxiety in patients with panic disorderand normal volunteers. Br J Psychiatry 176:182–188

Neumeister A, Hu XZ, Luckenbaugh DA, Schwarz M, Nugent AC,Bonne O, Herscovitch P, Goldman D, Drevets WC, Charney DS(2006) Differential effects of 5-HTTLPR genotypes on thebehavioral and neural responses to tryptophan depletion inpatients with major depression and controls. Arch Gen Psychiatry63:978–986

Nutt DJ, Glue P, Lawson C, Wilson S (1990) Flumazenil provocationof panic attacks: evidence for altered benzodiazepine receptorsensitivity in panic disorder. Arch Gen Psychiatry 47:917–925

O’Reardon JP, Chopra MP, Bergan A, Gallop R, DR J, Crits-Christoph P (2004) Response to tryptophan depletion in majordepression treated with either cognitive therapy or selectiveserotonin reuptake inhibitor antidepressants. Biol Psychiatry55:957–959

Perna G, Bertani A, Caldirola D, Gabriele A, Cocchi S, Bellodi L(2002) Antipanic drug modulation of 35% CO2 hyperreactivityand short-term treatment outcome. J Clin Psychopharmacol22:300–308

Prasko J, Horacek J, Zalesky R, Kopecek M, Novak T, Paskova B,Skrdlantova L, Belohlavek O, Hoschl C (2004) The change ofregional brain metabolism (18FDG PET) in panic disorder duringthe treatment with cognitive behavioral therapy or antidepressants.Neuroendocrinol Lett 25:340–348

Rassovsky Y, Kushner MG, Schwarze NJ, Wangensteen OD (2000)Psychological and physiological predictors of response to carbondioxide challenge in individuals with panic disorder. J AbnormPsychology 109:616–623

Sakai Y, Kumano H, Nishikawa M, Sakano Y, Kaiya H, Imabayashi E,Ohnishi T, Matsuda H, Yasuda A, Sato A, Diksic M, Kuboki T(2006) Changes in cerebral glucose utilization in patients with panicdisorder treated with cognitive-behavioral therapy. Neuroimage33:218–226

Schmidt NB, Trakowski JH, Staab JP (1997) Extinction of panicogeniceffects of a 35% CO2 challenge in patients with panic disorder. JAbnorm Psychology 106:630–638

Schruers K, Klaassen T, Pols H, Overbeek T, Deutz NEP, Griez E(2000) Effects of tryptophan depletion on carbon dioxideprovoked panic in panic disorder patients. Psychiatry Res93:179–187

ShearMK, Fyer AJ, Ball G, Josephson S, FitzpatrickM, Gitlin B, FrancesA, Gorman J, Liebowitz M, Klein DF (1991) Vulnerability tosodium lactate in panic disorder patients given cognitive-behavioraltherapy. Am J Psychiatry 148:795–797

Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E,Hergueta T, Baker R, Dunbar GC (1998) The Mini-InternationalNeuropsychiatric Interview (M.I.N.I.): the development andvalidation of a structured diagnostic psychiatric interview forDSM-IV and ICD-10. Journal of Clinical Psychiatry 59:22–33

Smith KA, Fairburn CG, Cowen PJ (1997) Relapse of depression afterrapid depletion of tryptophan. Lancet 349:915–919

Spielberger C, Gorsuch R, Lushene R (1970) Manual for the state traitanxiety inventory. Consulting Psychologists Press, Palo Alto

602 Psychopharmacology (2011) 213:593–602