Lawrence H. Price, Andrew W. Goddard, Linda C. Barr, and Wayne K. Goodman

INTRODUCTION

The pharmacological challenge strategy involves administering a test agent under controlled conditions to elucidate some aspect of biological or behavioral function in the organism being studied. It is based on the assumption that true functional abnormalities may not be evident in the basal state because of the action of compensatory mechanisms. Under such circumstances, pharmacological perturbation of a specific target system may reveal information about the functional integrity of both that system and systems that modulate it. Uses of this approach include (a) generation and testing of hypotheses regarding the pathophysiology of a disorder, (b) delineation of the effects and mechanisms of action of treatments, (c) identification of pathophysiologically distinct diagnostic subtypes, and (d) clinical applications (as a diagnostic test, as a predictor of treatment response, as a means of assessing treatment adequacy, and as a predictor of relapse).

In studies of neuropsychiatric disorders, the ideal challenge probe should have a mechanism of action that is well-characterized at the preclinical level, be pharmacologically selective for the system under investigation, have no active metabolites, and induce responses that are sensitive, reliable, accessible to clinical measurement, and reflective of brain function. Safety and convenience are desirable qualities. Drug dosage, route and rate of administration, environmental conditions of the testing situation, and rater–observer characteristics are additional factors that should be standardized, because they may contribute to unwanted response variance.

In practice, few challenge probes meet all of the ideal criteria. Despite this, the pharmacological challenge paradigm continues to enjoy popularity among investigators. Clinically relevant neurobiological processes are otherwise difficult to study in vivo in patients, and the promise of the paradigm for clarifying pathophysiology still seems great. In some instances, that promise has been at least partially realized [e.g., probes of serotonin (5-HT) function in depression], whereas other lines of investigation have foundered [e.g., prolactin (PRL) responses to dopamine (DA) agonists and antagonists in schizophrenia].

In light of these factors, this chapter reviews the use of the pharmacological challenge strategy in patients with anxiety disorders. The chapter is divided into two major sections: panic and related anxiety disorders and obsessive–compulsive disorders (OCD). Within each major section, the use of specific probes is considered according to the major neurotransmitter systems they engage. Doses and routes of administration are specified to underscore the dependence of response sensitivity and specificity on these parameters. This chapter updates the detailed review of Gorman et al. (34) on the pharmacological provocation of panic attacks and additionally considers other response measures to pharmacological challenges.

The use of pharmacological challenges in panic disorder is unique in that the clinical phenomenon of central interest (i.e., the panic attack) can be readily provoked and assessed in the clinical laboratory setting (34). This has proved far more difficult in other disorders (e.g., depression and schizophrenia). The primary focus of this section on panic disorder reflects the extensive study of this condition by means of the pharmacological challenge strategy. Consideration of related anxiety disorders, such as social phobia, generalized anxiety disorder (GAD), and posttraumatic stress disorder (PTSD), is included where appropriate.

Historically, the noradrenergic system has dominated preclinical theoretical and empirical approaches to anxiogenesis (79). The extensive clinical literature on the involvement of norepinephrine (NE) in panic has been made possible through the long-standing availability of several relatively selective probes of this system.

Yohimbine is an indole alkaloid with a2-adrenoceptor antagonist properties. It increases NE function by blocking inhibitory a2-adrenoceptors located presynaptically on NE neurons in the locus coeruleus (LC), resulting in activation of those neurons (79). Although generally considered selective for a2-adrenoceptor blockade, yohimbine also has effects on the DA and 5-HT systems.

In a 1984 study, Charney et al. (14) gave yohimbine (20 mg po) to 39 drug-free patients with panic disorder or agoraphobia and 20 healthy controls. Ratings of anxiety and nervousness, physical symptoms, and blood pressure responses were greater in the patients. Yohimbineinduced increases in plasma levels of the NE metabolite 3-methoxy-4 hydroxyphenylglycol (MHPG) correlated with increases in anxiety in patients but not in controls. Patients with frequent panic attacks had greater MHPG responses than controls. In a subsequent study in 1987 using yohimbine (20 mg po), panic attacks were provoked in 37 out of 68 (54%) panic patients and 1 out of 20 (5%) controls (19). Patients reporting yohimbine-induced panic attacks had greater increases in plasma MHPG, cortisol, blood pressure, and heart rate than controls. More recently, Charney et al. (20) reported that yohimbine (0.4 mg/kg iv) provoked panic attacks in 24 out of 38 (63%) panic patients and 1 out of 15 (7%) controls, again finding that MHPG responses were greater in patients who panicked. Gurguis and Uhde in 1990 replicated the relationship between anxiety and MHPG responses to yohimbine (20 mg po) in 11 panic patients and 7 controls. Albus et al. (3) administered yohimbine (20 mg po) to panic patients treated with placebo (n = 8) or alprazolam (n = 7) and 12 healthy controls. Patients had greater anxiety responses than controls, but none experienced panic attacks, apparently because instructions during the test were designed to minimize the likelihood of an attack. The central effects of yohimbine (0.4 mg/kg iv) in six drug-free panic patients and six controls were studied by Woods et al. (95) using single-photon emission computed tomography (SPECT) imaging with 99mTc-HMPAO. Decreased frontocortical cerebral blood flow occurred in patients compared to controls, with six out of six patients and one out of six controls experiencing increased anxiety following yohimbine.

These studies strongly suggest involvement of the NE system in some panic patients. However, there are limitations to yohimbine as a panicogen. Over one-third of panic patients are insensitive to its panicogenic effects. Moreover, such effects are not specific to panic disorder: in a study of 20 drug-free PTSD patients and 18 controls utilizing yohimbine (0.4 mg/kg iv), Southwick et al. (83) observed panic attacks in 14 out of 20 (70%) and flashbacks in 8 out of 20 (40%) patients, with none in controls. Contrary to expectation, Charney and Heninger found in 1985 (17) that long-term treatment with the antipanic agent imipramine, which significantly affects NE function, did not block yohimbine-induced panic in 11 patients. However, the behavioral response to yohimbine is attenuated by long-term treatment with other antipanic drugs, such as the benzodiazepine alprazolam (n = 14) (17) and the selective serotonin reuptake inhibitor (SSRI) fluvoxamine (n = 16) (31). Effects of yohimbine on 5-HT and DA function could account for the lack of reversal of yohimbine's behavioral effects by imipramine.

Phenomenologically, yohimbine does not produce a panic attack that is identical to naturally occurring episodes. Patients and controls describe euphoria, lacrimation, and rhinnorhea, which are not typical of naturalistic panic. Finally, Albus et al. (4) have noted the influence of expectancy and cognitive set on behavioral responses to yohimbine (20 mg po). While this may be viewed as a limitation of the challenge paradigm in general, it also reflects the real modulation of anxiety by such factors in the natural environment.

Clonidine, an imidazoline derivative, is an a2adrenoceptor agonist with some anxiolytic properties. It decreases neuronal activity in the LC by stimulating presynaptic autoreceptors, consequently reducing sympathetic outflow (78). Sedation and hypotension limit its clinical use as an anxiolytic.

Charney and Heninger found in 1986 (18) that clonidine (0.15 mg iv) caused greater hypotension, greater decreases in plasma MHPG, and less sedation in 26 drug-free panic patients than in 21 controls. In 1989 Nutt (see ref. 66) replicated these observations in 16 panic patients and 16 controls using clonidine (1.5 mg/kg iv). Recently, these finding were also replicated with clonidine (2 mg/kg iv) in a subgroup of panic patients who had manifested yohimbine-induced panic (20). These data suggest that presynaptic a2-adrenoceptor sensitivity is increased in panic disorder, although it is unclear whether this reflects changes in receptor affinity (Kd), receptor binding (Bmax), or second-messenger function. Uhde et al. (90), in a study of 11 drug-free panic patients, 11 depressed patients, and 11 controls, first reported a blunted growth hormone (GH) response to clonidine (2 mg/kg iv) in panic disorder, suggesting subsensitivity of central postsynaptic a2adrenoceptors. This finding, already well established in depression, was confirmed by other investigators (18, 66). A recent study has found blunted GH responses to both clonidine (2 mg/kg iv) and GH-releasing factor (GHRF; 1 mg/kg iv) in 13 panic patients compared to 20 controls (86), suggesting that intrinsic hypothalamic– pituitary dysfunction may be present in panic disorder. Other anxiety disorders may also manifest a blunted GH response to clonidine. Abelson et al. (1) observed this phenomenon in a study of 11 drug-free GAD patients and 14 controls given clonidine (2 mg/kg iv). However, normal GH responses to clonidine (5 mg/kg po) were found in a study of 21 social phobia patients and 22 controls (85).

In summary, panic patients consistently manifest altered responses to a2-adrenoceptor agonists and antagonists. Whether this reflects some primary defect in the a2-adrenoceptor itself is unclear. The large number of non-NE neurotransmitter systems with significant input to the LC leaves open the possibility that dysregulation of one or more of these systems could lead to abnormalities of LC function manifested as panic attacks (79).

Isoproterenol, a synthetic sympathomimetic amine, is a peripherally acting agent with selectivity for the b-adrenoceptor. Early studies suggested that patients predisposed to anxiety-related symptoms experienced exacerbation of such symptoms when given isoproterenol, and that this could be reversed with the b-adrenoceptor antagonist propranolol (34).

Extending their own earlier findings, Pohl et al. (70) reported that 57/86 (66%) panic patients and 4/45 (9%) controls developed panic after isoproterenol 1 mg/min iv. These findings support the hypothesis of increased b-adrenoceptor sensitivity in panic disorder. In contrast, Nesse et al. in 1984 (64) observed no differences between 14 drug-free panic patients and 6 controls in behavioral responses to isoproterenol 0.06–4.0 mg iv. The discrepancy could be explained by differing methods of isoproterenol administration (i.e., continuous infusion (70) vs multiple boluses (64). Differences in panic attack criteria may also have contributed to conflicting results. Nesse et al. (64) additionally found smaller heart rate responses in patients than controls, which suggests decreased b-adrenoceptor sensitivity and a generalized increase in NE function in the nonpanic state. In another study casting doubt on the necessity of b-adrenoceptor supersensitivity for the occurrence of panic, Gorman et al. observed in 1983 (34) that pretreatment with propranolol (0.2 mg/kg iv) failed to block lactate-induced panic in all six panic patients studied.

The reliability and mechanism of isoproterenol-induced panic remain to be clarified. A singular limitation of isoproterenol is its inability to cross the blood–brain barrier. In addition, it is not differentially selective between b1- and b2-adrenoceptor subtypes. Finally, the fact that it is metabolized by monoamine oxidase, which could be altered in panic disorder, constitutes a disadvantage relative to some other b-adrenoceptor agonists.

Epinephrine and NE are endogenous amines that are secreted in response to stress; they do not cross the blood–brain barrier. Epinephrine is a more potent agonist of b- than a-adrenoceptors, whereas NE is primarily an a-adrenoceptor agonist with some b activity. This lack of specificity compromises their use as pharmacological challenge agents.

The effects of these substances in anxiety disorders are not established, as they have rarely been studied in well-diagnosed patients. Early studies suggested that epinephrine tended to induce social withdrawal rather than panic, whereas NE was seen as weakly anxiogenic (34). In 1986, Pyke and Greenberg (73) evaluated the response of six panic patients to a NE infusion starting at 0.25 mg/min and increasing up to 16 mg/min. All six patients experienced panic attacks, but the study included neither a placebo infusion nor healthy subjects as controls.

Preclinical interest in the role of 5-HT in the pathogenesis of anxiety has been building over the past decade, but clinical studies have begun to appear in number only in the past several years.

Often described as a 5-HT partial agonist, m-CPP has complex effects on brain 5-HT systems. It binds equipotently and with greatest affinity to 5-HT1C and 5-HT3 receptors, and less potently to 5-HT2 and 5-HT1A receptors (52). m-Chlorophenylpiperazine appears to act as an agonist at the 5-HT1C receptor, and perhaps at the 5-HT1A receptor as well. Effects at the 5-HT3 receptor seem primarily antagonistic, whereas mixed agonist and antagonist activity has been found at the 5-HT2 site. Serotonin releasing properties have been reported. Although m-CPP has little affinity for most non-5-HT receptors, it does bind to a2-adrenergic sites, the functional significance of which is unknown.

Kahn et al. found in 1988 (52) that m-CPP (0.25 mg/kg po) caused panic attacks in six out of ten drug-free panic patients but not in 10 depressed patients or 11 healthy controls. Cortisol responses to m-CPP were also greater in panic patients (52). These investigators posited 5-HT receptor supersensitivity in panic disorder. However, no differences were found in the behavioral and neuroendocrine responses of 23 drug-free panic patients and 19 controls to m-CPP (0.1 mg/kg iv) (19, 52), or of 27 panic patients and 22 controls to m-CPP (0.05 mg/kg iv) (27). A possible explanation for these discrepant findings is that intravenous administration at these doses leads to receptor overstimulation in both patients and controls, with loss of resolution of the behavioral and neuroendocrine signal demonstrated with the oral route.

Germine et al. (27) administered m-CPP (0.1 mg/kg iv) to ten drug-free GAD patients and 19 controls. Patients showed greater increases in anxiety symptoms, subjective anger, and GH, consistent with supersensitivity of postsynaptic 5-HT receptors. Similarly, Krystal et al. (54) found that 14 drug-free PTSD patients experienced more panic and dissociative symptoms (including flashbacks) following m-CPP (0.1 mg/kg iv) than placebo.

These findings are consistent with other evidence that m-CPP may exacerbate psychopathology across several disorders (52). However, m-CPP's complex neuropharmacology makes it difficult to draw conclusions about specific abnormalities within the 5-HT system. Moreover, the drugged feeling reported by many panic patients following m-CPP administration suggests an important dissimilarity between m-CPP-induced and naturalistic panic. Nonetheless, m-CPP may prove useful in evaluating interactions between 5-HT and other neurotransmitter systems. For example, Asnis et al. (6) recently tested 22 drug-free panic patients, 17 depressed patients, and 10 healthy controls with m-CPP (0.25 mg/kg po) and desipramine (75 mg im). Cortisol responses to each challenge were negatively correlated in the total sample, but particularly in the patient groups, suggesting some dysregulation of 5-HT/NE interactions in these disorders.

Fenfluramine is a phenylethylamine derivative with marked effects on brain 5-HT function. Upon acute administration, it potently releases presynaptic 5-HT and inhibits 5-HT reuptake, with weaker action as a postsynaptic 5-HT agonist. Long-lasting reductions in brain levels of 5-HT and its principal metabolite, 5-hydroxyindoleacetic acid (5-HIAA), have been observed following a single dose. Although structurally similar to amphetamine and classified as a sympathomimetic, fenfluramine causes little activation or euphoria.

Targum and Marshall in 1989 (87) administered DL-fenfluramine (60 mg po) to nine drug-free panic patients, nine depressed patients, and nine healthy controls. Panic patients exhibited greater anxiety, prolactin, and cortisol responses than the other groups, with six (67%) experiencing a panic attack. No depressed patients had an anxiety reaction, and two healthy controls experienced a mild increase in anxiety. In a subsequent study by Targum in 1991 (88), 26 drug-free panic patients and 12 controls were given 0.5-M sodium DL-lactate (10 ml/kg iv) followed one day later by fenfluramine (60 mg po). Nine out of twelve patients (75%) with frequent panic attacks had prominent anxiety responses to both challenges, compared with none of the 14 patients with infrequent panic attacks. The investigator concluded that these responses reflected an elevated level of anticipatory anxiety due to frequent recent attacks, rather than some trait of panic patients. More recently, Targum (88) reported cortisol responses to lactate and fenfluramine in 12 drug-free panic patients with known anxiety responses to both agents and eight nonresponsive healthy controls. Patients had greater cortisol responses to fenfluramine than controls, but were similar to controls in showing no cortisol response to lactate. Some authors have remarked that the quality of the fenfluramine-provoked anxiety in panic patients is more suggestive of anticipatory or generalized anxiety than true panic (44). Targum's 1991 finding is consistent with this view. Fenfluramine might be conceptualized as a probe of one facet of panic disorder, that is, anticipatory fear, whereas lactate might be a better probe of true panic. The differential cortisol response of panic patients to these challenges is consistent with this view.

Tancer and Golden (85) gave fenfluramine (60 mg po) to 21 social phobia patients and 22 controls, observing an elevated cortisol response in the patients. This preliminary evidence suggests that social phobia patients exhibit dysregulated 5-HT neurotransmission, consistent with the clinical efficacy of the MAO inhibitor phenelzine in this population.

L-tryptophan is the initial dietary precursor of 5-HT. After competing with other large neutral amino acids for uptake into the brain, it is converted by the rate-limiting enzyme tryptophan hydroxylase into 5-hydroxytryptophan (5-HTP), which is then decarboxylated to 5-HT. Neuroendocrine effects of intravenous L-tryptophan are believed to result from the central synthesis and release of 5-HT, although there is debate as to whether other mechanisms (e.g., decreased availability of tyrosine for DA synthesis, activation of kynurenine metabolism) might be involved.

Serotonin-precursor loading is not anxiogenic. Charney and Heninger (18) gave L-tryptophan (7 g iv) to 23 drug-free panic patients and 21 controls, observing no difference in behavioral or prolactin responses between the groups. Westenberg and Den Boer in 1989 (91) administered 5-HTP (60 mg iv) to seven drug-free panic patients and seven controls, also finding little behavioral or neuroendocrine difference between groups. The discrepancy of these data with the m-CPP findings could be due to differential 5-HT receptor subtype activation by these agents or to the presynaptic effects of the precursors.

Administration of a tryptophan-free amino acid mixture lowers plasma tryptophan, brain tryptophan, and 5-HT in laboratory animals, with behavioral effects consistent with a state of 5-HT depletion (24). Similar effects on plasma tryptophan have been demonstrated in humans by giving a tryptophan-free amino acid mixture preceded by a 24-hour low tryptophan diet (24).

Double-blind sham-controlled tryptophan depletion in six drug-free panic patients demonstrated no effects on anxiety symptoms (30). In a related study evaluating the role of 5-HT/NE interactions in anxiogenesis, a combination challenge of tryptophan depletion followed by yohimbine (0.4 mg/kg iv) caused greater nervousness in 11 healthy subjects than did either challenge condition alone (29). The NE metabolite MHPG and cortisol responses to the combination test were not augmented above those in the yohimbine alone condition. Investigation of panic patients with this paradigm should further clarify recent evidence of dysregulated 5-HT/NE interactions in panic disorder (6).

Ipsapirone is an azapirone derivative that acts selectively as a full agonist at presynaptic 5-HT1A autoreceptors and as a partial agonist at postsynaptic 5-HT1A sites. Lesch et al. (57) gave ipsapirone (0.3 mg/kg po) to 14 drug-free panic patients and 14 controls. Corticotropin (ACTH), cortisol, and hypothermic responses were blunted in the patients, but anxiety responses did not differ from controls. These findings support some role, perhaps modulatory in nature, of 5-HT1A receptors in the pathogenesis of panic disorder.

This area, constituting the largest single body of pharmacological challenge research into the anxiety disorders, has recently been reviewed (34, 53, 66). Although the findings are impressively robust and reproducible, their theoretical impact has been limited by a paucity of relevant preclinical data regarding mechanism(s) of action.

Lactic acid plays a key role in carbohydrate and energy metabolism. Pitts and McClure first showed in 1967 (69) that infusing sodium lactate caused panic attacks in anxious patients but not controls. This finding has been replicated often, with few contradictions. The challenge probe generally consists of 0.5-M sodium DL-lactate (10 ml/kg) given iv over 20 min. Lactate-induced anxiety appears specific to panic disorder (23) and is antagonized by treatment with antipanic agents (34).

Early investigators suggested that lactate might provoke panic by causing hypocalcemia, but this has not been substantiated. Carr and Sheehan hypothesized in 1984 (13) that panic patients have enhanced sensitivity of ventral medullary chemoreceptors to fluctuations in pH. They argued that lactate infusion could lower the pH in these cells by increasing the local lactate:pyruvate ratio, a consequence of passive diffusion of lactate into these areas and of hypoxemia secondary to cerebrovascular vasoconstriction caused by lactate-induced metabolic alkalosis. Panic attacks would result from the chemoreceptors' "misperception" of life-threatening central hypoxia and acidosis. To date, the assumptions underlying this model are unproven.

In an alternative model, Gorman et al. speculated in 1989 (37) that as lactate is metabolized to bicarbonate, resulting in a metabolic alkalosis, bicarbonate is in turn metabolized to CO2, which stimulates both medullary chemoreceptors and the LC, causing panic in vulnerable individuals. Consistent with this is evidence that clonidine partially attenuates lactate-induced panic (20). Against this theory is the finding by Gorman et al. in 1990 (38) that D-lactate is also panicogenic, although it is not metabolized to CO2. The lactate–CO2 hypothesis has nonetheless been of great value in considering the relationship between respiration and panic (53).

Voluntary hyperventilation, causing hypocapnia, can precipitate panic in panic patients, although some authorities suggest that this is only weakly panicogenic. Panic may also be induced by hypercapnia. Gorman et al. found in 1988 (36) that CO2 inhalation (5% in air) precipitated panic in 12 out of 31 (39%) drug-free panic patients, one out of 13 (8%) healthy controls, and none of 12 patients with other anxiety disorders. Woods et al. in 1986 (93) provoked panic attacks in 8 out of 14 (57%) drug-free panic patients rebreathing 5% CO2 in air; anxiety responses were similar in 8 controls given 7.5% CO2, but less in 11 controls given 5% CO2, suggesting greater sensitivity in the patients. These investigators also observed marked attenuation of the anxiety response to 5% CO2 in seven panic patients treated with alprazolam. Griez et al. in 1987 (39) reported that one or two deep breaths of a 35% CO2/65% O2 mixture could induce panic in panic patients (n = 12) but not healthy controls (n = 11).

The ability of both hypo- and hypercapnia to trigger panic remains an enigma. Carr and Sheehan (13) suggested that both states may lead to a fall in brainstem pH, either directly (hypercapnia leading to respiratory acidosis) or indirectly (hypocapnia leading to respiratory alkalosis leading to cerebrovascular vasoconstriction leading to hypoxia). Related to this is the theory of enhanced chemoreceptor or LC sensitivity to CO2 in panic patients (37), although this does not address the issue of hypocapnia. A less specific, but more encompassing, psychological explanation is that both states produce internal somatic cues that are misinterpreted as "dangerous" and therefore evoke panic (77). Nutt and Lawson (66) have enumerated the flaws in this explanation, which might be advanced to account for all pharmacological panicogens. Among these flaws is the failure to explain why panic patients are prone to such misinterpretations in the first place. Moreover, Goetz et al. (32) have shown that cardiorespiratory activation occurs even in panic attacks during placebo pharmacological challenges, suggesting that these changes are a manifestation, rather than a cause, of panic. Most recently, Klein (53) has advanced a comprehensive theory suggesting that both CO2 and lactate induce panic by triggering a suffocation false-alarm in individuals with a hypersensitive suffocation detector. An attractive feature of this theory is its attempt to differentiate the neurobiological and phenomenological underpinnings of fear, acute and chronic anxiety, and panic.

The major behavioral effects of benzodiazepine (BZ) agonists and antagonists are mediated through saturable high-affinity BZ receptor sites located on a subunit of the g-aminobutyric acidA (GABAA) receptor (45). In this location, BZ receptors allosterically modulate GABAA receptor function to increase conductance through the associated chloride channel. Benzodiazepine receptor agonists (e.g., diazepam) have marked anxiolytic effects. Inverse agonists (e.g., the b-carboline FG 7142) are proconvulsant and anxiogenic. Antagonists (e.g., flumazenil) have little intrinsic activity, but block the effects of agonists and inverse agonists.

Roy-Byrne et al. (80) administered iv diazepam in dosages of 25, 25, 50, and 100 mg/kg 15 minutes apart to nine drug-free panic patients and ten controls, measuring saccadic eye movement velocity as an index of brainstem BZ receptor function. Patients had less diazepam-induced slowing of saccadic eye movement than controls, suggesting subsensitivity of BZ receptors.

Nutt et al. (65) gave flumazenil (2.0 mg iv) to ten drug-free panic patients and ten controls. Panic attacks occurred in 80% of patients but not in the controls. Studying oral flumazenil in 11 drug-free panic patients, however, Woods et al. (94) obtained panic attack rates of four out of ten at 200 mg, none of 11 at 600 mg, and none of eight with placebo. Nutt and Lawson (66) speculated that BZ receptor functioning is shifted in panic patients so that antagonists are recognized as partial inverse agonists. This is consistent with evidence of BZ receptor subsensitivity in panic disorder (80). An alternative interpretation is that flumazenil exacerbates a functional deficiency of some endogenous anxiolytic in the patients. The data do not support the hypothesis that panic patients have increased levels of some endogenous inverse agonist, since this would predict anxiolytic effects of the antagonist. In a recent study of flumazenil given in doses of 2.0 mg iv to seven drug-free panic patients and seven controls, Wilson et al. (92) found that flumazenil decreased saccadic eye movements in both groups, suggesting a slight partial agonist effect.

Two novel compounds acting at the GABAA/BZ receptor deserve mention. The partial inverse agonist Ro-16-0154 is being evaluated for behavioral effects in humans, and the partial agonist abecarnil (ZK 112–119) may have clinical efficacy as an anxiolytic. These new agents, together with the developments that have recently occurred in this area, will significantly advance understanding of BZ receptor pathophysiology in anxiety disorders.

Cholecystokinin (CCK) is an octapeptide found regionally in the gastrointestinal tract and brain, where it acts as a neurotransmitter and neuromodulator. In some neurons, it is colocalized with other neurotransmitters, particularly DA and GABA, and GABA seems to be involved in the regulation of CCK release. Interest in CCK's role in anxiogenesis arose from evidence that it stimulates rat cortical and hippocampal neurons, effects that are blocked by benzodiazepine agonists (25).

DeMontigny (25) reported that seven out of ten healthy subjects experienced panic anxiety after receiving 20 to 100 mg iv of CCK-4, a selective CCK-B receptor agonist that crosses the blood–brain barrier. Bradwejn et al. found in 1991 (12) that CCK-4 (50 mg iv) induced panic in all eleven drug-free panic patients; patients rated these attacks as very similar to their naturalistic panics. Bradwejn et al. (12) subsequently observed that 50 mg iv of CCK-4 provoked panic in all 12 drug-free panic patients and 7 out of 15 (47%) controls, whereas 25 mg iv of CCK-4 induced panic in 10 out of 11 (91%) patients and 2 out of 12 (17%) controls. These investigators concluded that panic patients were more sensitive than controls to the anxiogenic effects of CCK-4. Abelson and Nesse (2) have also found that the CCK-B agonist pentagastrin (0.6 mg/kg iv) provokes panic symptoms in four out of five drug-free panic patients and one out of four controls.

Cholecystokinin-4 is an attractive new probe of anxiety. Its anxiogenic effects are reliable and dose-dependent, with 25 mg iv of CCK-4 producing panic responses similar to 35% CO2 inhalation in patients and controls (11). Its ease of administration and similarity of effect to naturalistic panic are also strengths. Studies of CCK-4's interactions with non-CCK neurotransmitter systems have already begun. DeMontigny (25) reported that lorazepam prevented CCK-4–induced anxiety in four healthy subjects. However, Couetoux-Dutertre et al. (22) recently found that flumazenil (2.0 mg iv) did not block the anxiogenic effects of CCK-4 (50 mg i.v) in 30 healthy subjects, indicating that CCK-4 is not acting as an inverse agonist at the GABAA/BZ receptor. Studies with selective antagonists (e.g., the CCK-B antagonists CI-988, L-365,260, and LY-262691) will help clarify the mechanism of CCK-4's anxiogenic effects. The clinical efficacy of CCK-B antagonists is currently under investigation.

Preclinical evidence strongly implicates corticotropin-releasing hormone (CRH) in the mediation of stress responses in animals (26). Roy-Byrne et al. in 1986 (80) first reported blunted ACTH and cortisol responses to CRH (1 mg/kg) in eight drug-free panic patients compared with 30 controls, a finding that was replicated by Holsboer et al. in 1987 (45). These data are consistent with a chronic hypercortisolemic state in panic disorder. However, Rapaport et al. (75) observed no difference between eight drug-free panic patients and 11 controls in response to ovine CRH (0.03 mg/kg iv). Reconciliation and extension of these findings awaits further clinical studies.

Rapaport et al. reported in 1989 (76) that the GH response to 1 mg/kg iv of growth hormone-releasing factor (GHRF) was markedly blunted in 11 drug-free panic patients compared with 11 controls. This finding, recently replicated by Tancer et al. (86) (cf. above), suggests that hypothalamic–pituitary dysfunction merits further scrutiny as a possible factor in the pathophysiology of panic disorder.

In addition to the voluminous literature documenting thyroid axis abnormalities in affective disorders, reports with small samples have suggested that the thyrotropin response to thyrotropin-releasing hormone (TRH) is blunted in panic disorder. Stein and Uhde (84) recently conducted a careful investigation of the effects of TRH (500 mg iv) in 26 drug-free panic patients and 22 controls. Prolactin, thyrotropin, blood pressure, and heart rate responses were similar between groups. Despite robust increases in heart rate and blood pressure following TRH, only one patient experienced a panic attack. This finding contradicts the theory that cognitive elaboration of interoceptive cues is sufficient to trigger panic attacks in these patients.

Caffeine is a methylxanthine with central effects as an adenosine receptor antagonist. It is a mild psychostimulant, well-known for its dose-dependent anxiogenic properties in healthy subjects and psychiatric patients.

In a 1985 study utilizing caffeine (10 mg/kg po), Charney et al. (15) provoked panic attacks in 15 out of 21 (71%) drug-free panic patients and none of the 17 controls. Uhde et al. (89) observed panic in 9 out of 24 (54) panic patients and none of the 14 controls after 480 mg po of caffeine.

Caffeine-induced anxiety could be from blockade of central adenosine receptors followed by activation of the LC, resulting in increased release of brain NE. Other possible mechanisms include inhibition of phosphodiesterase and BZ receptor antagonism. Unfortunately, the pharmacological and behavioral effects of caffeine are neither selective nor potent, which limits its utility as a probe for anxiety disorders. Major advances in this area await the availability of adenosine receptor ligands with more selectivity than the xanthines or purines.

Neuroglycopenia has many symptoms in common with panic, but Schweizer et al. (81) found that insulin (0.1 units/kg) did not precipitate panic attacks in ten drug-free panic patients, even though prominent symptoms of hypoglycemia were evident. This study argues against the hypothesis that nonspecific sympathetic activation is sufficient to cause panic in predisposed individuals via cognitive elaboration of interoceptive cues.

Early reports suggested that cholinergic agonists such as the central cholinesterase inhibitor physostigmine had anxiogenic properties in healthy subjects. Following pretreatment with propantheline (45 mg po) to block peripheral effects, Rapaport et al. (74) administered physostigmine (0.022 mg/kg iv) to nine drug-free panic patients and nine controls. Behavioral, cortisol, and cardiovascular responses did not differ between groups, and no subjects experienced panic attacks. These data do not support a role for the central cholinergic system in the pathogenesis of anxiety disorders.

Pitchot et al. (68) observed a greater GH response to the DA agonist apomorphine (0.5 mg sc) in nine drug-free panic patients compared with nine major depressive and nine minor depressive patients. These investigators concluded that DA function was increased in panic disorder compared with depression, but the lack of healthy controls limits the interpretation of this study. In their study of 21 social phobia patients and 22 controls, Tancer and Golden (85) found no difference between groups in the prolactin response to oral L-dopa (500 mg). Abnormal neuroendocrine responses to fenfluramine, but not clonidine, were observed, suggesting that social phobia patients manifest dysregulated 5-HT neurotransmission in the face of normal DA and NE function.

Use of the pharmacological challenge strategy in OCD is a relatively recent development, and the high context dependency of OC symptoms has made studies of behavioral responses more problematic than in panic disorder. Despite these limitations, a significant literature has begun to emerge.

Potent SRIs are superior to other agents in the treatment of OCD. This, rather than theoretical considerations, accounts for the fact that studies of 5-HT function have dominated research into the clinical neurobiology of OCD over the past decade (7).

m-Chlorophenylpiperazine has been the most frequently used probe in studies of OCD, appearing in six reports involving drug-free patients. Zohar et al. in 1987 (96) gave m-CPP (0.5 mg/kg po) to 12 OCD patients and 20 controls. Patients manifested a greater anxiety response than controls and specific exacerbation of their obsessive–compulsive (OC) symptoms. Cortisol responses were blunted in patients, whereas increased prolactin and hyperthermia occurred equally in both groups. Charney et al. in 1988 (16) administered m-CPP 0.1 mg/kg iv to 21 OCD patients and 21 controls. Basal prolactin levels were lower and prolactin responses were blunted in female, but not male, patients. There were no differences between patients and controls in cortisol, GH, or anxiety responses, and OC symptoms were unchanged. Hollander et al. (43) studied m-CPP (0.5 mg/kg po) in 20 OCD patients and 10 controls. Obsessive–compulsive symptoms were exacerbated and prolactin responses attenuated in patients, with no difference between groups in anxiety or cortisol responses. To clarify the mechanism of oral m-CPP's behavioral effects, Pigott et al. in 1991 (67) administered m-CPP (0.5 mg/kg po) to 12 OCD patients following acute pretreatment with the 5-HT antagonist metergoline (4 mg po). Metergoline blocked the worsening of OC symptoms and the prolactin increase caused by m-CPP. However, in a subsequent study designed to reconcile conflicting behavioral findings with oral and intravenous m-CPP, Pigott et al. (67) administered m-CPP (0.5 mg/kg po) to 17 OCD patients and m-CPP (0.1 mg/kg iv) to a separate group of 10 OCD patients. Oral m-CPP did not increase OC or anxiety symptoms, but intravenous m-CPP increased both. Exacerbation of OC and anxiety symptoms caused by intravenous m-CPP was blocked by metergoline (4 mg po). Recently, Goodman et al. in 1991 (33) gave 0.1-mg/kg iv and 0.5-mg/kg po doses of m-CPP to the same 12 OCD patients on two different occasions. Neither route of administration resulted in worsening OC symptoms.

Two studies have described the effects of long-term treatment with SRIs on responses to m-CPP in OCD patients. Zohar et al. (96), using a 0.5-mg/kg po dose of m-CPP found that clomipramine treatment abolished the m-CPP–induced exacerbation of OC and anxiety symptoms and the hyperthymic response observed pretreatment in nine patients; prolactin and cortisol responses were unaffected. Using the same m-CPP dosage and route of administration, Hollander et al. in 1991 (43) reported that fluoxetine treatment also abolished pretreatment m-CPP–induced exacerbation of OC symptoms in six patients, although no anxiety response was detected pre- or posttreatment. In this study, prolactin and cortisol responses were increased during fluoxetine treatment, but this may have reflected increased m-CPP plasma levels.

To summarize, the most controversial finding with m-CPP is its putative ability to exacerbate OC symptoms, as reported by two groups (43, 96) but contested by a third (16). Differences in dosage and route of administration have been adduced to explain the discrepancy, but this argument has recently been weakened by evidence that oral and intravenous routes have comparable effects (7). Additionally, one group (67) has failed to replicate its own finding of symptom exacerbation with oral m-CPP, but now observes this effect with the intravenous route, in contrast to the other group which has studied intravenous m-CPP (16). Behavioral and neurobiological variability across patients, interacting with the multiplicity of m-CPP's pharmacological effects, could account for these disparate findings, which are reminiscent of the inconsistencies in the m-CPP–panic literature. In contrast, attenuated neuroendocrine responses to m-CPP is a consistent finding, although not all studies report blunting of the same hormone.

Three studies have utilized DL-fenfluramine as a probe in drug-free OCD patients. Hollander et al. (43) gave fenfluramine (60 mg po) to 20 OCD patients and 10 controls. Prolactin and cortisol responses did not differ between groups. In a study of 21 OCD patients and 27 controls, McBride et al. (63) also found no difference in prolactin responses to fenfluramine (60 mg po). In neither study did fenfluramine significantly affect OC symptoms. However, Hewlett et al. (41) reported that fenfluramine (60 mg po) caused a blunted prolactin response in 26 OCD patients compared with 20 controls, a finding significant in females only.

Lucey et al. (62) studied D-fenfluramine (30 mg po) in 10 OCD patients, 10 depressed patients, and 10 controls, all drug-free. Preference for the D-isomer over the racemate has been advocated based on its greater potency and specificity for 5-HT (as opposed to DA) systems. Both OCD and depressed patients had blunted prolactin and cortisol responses compared with controls, with no differences between the patient groups.

These four studies provide some support for the hypothesis that net 5-HT neurotransmission is impaired in OCD. Fenfluramine studies have yielded similarly equivocal evidence of blunted 5-HT function in depression. Fenfluramine's ability to provoke panic attacks in panic patients (88) contrasts with its lack of behavioral effects in OCD.

In the only study using this agent in OCD, Charney et al. (16) gave L-tryptophan (7 g iv) to 21 drug-free OCD patients and 21 controls. The prolactin response to L-tryptophan was slightly but significantly greater in patients than controls, with no difference in GH response and no exacerbation of OC symptoms. The enhanced prolactin response in OCD contrasts with findings in panic disorder patients, who do not differ from controls (18), and depressed patients, who manifest blunted responses in some subtypes (72). However, Price et al. (72) found that melancholic depressed patients show an enhanced prolactin response. This disappeared when baseline plasma tryptophan levels, which were nonsignificantly lower in melancholic patients than controls, were used as covariates. Baseline plasma tryptophan levels were not measured by Charney et al. (16), but this factor might also account for the finding in OCD patients.

Lesch et al. (56) administered ipsapirone (0.3 mg/kg po) to 12 drug-free OCD patients and 22 controls. Cortisol, ACTH, hypothermic, and behavioral responses did not differ between groups. Lesch et al. (57) found that long-term treatment with fluoxetine in 10 OCD patients attenuated the cortisol, ACTH, and hypothermic responses to ipsapirone. This suggests that SRI treatment down-regulates 5-HT1A receptor sensitivity, although the lack of 5-HT1A–mediated differences between drug-free patients and controls raises doubts regarding the mediation of antiobsessional effects through this mechanism. Cortisol, ACTH, and hypothermic responses to ipsapirone are blunted in panic disorder (58) and depression relative to controls.

Like ipsapirone, buspirone is also an azapirone derivative with high affinity for the 5-HT1A receptor. However, mediation of its neuroendocrine effects through the 5-HT1A site is questionable, since it also antagonizes DA receptors.

Lucey et al. (60) studied buspirone (30 mg po) in 10 drug-free OCD patients and 10 controls. There was no difference in prolactin responses between groups, and OCD patients showed no worsening of symptoms. Acknowledging the lack of pharmacological selectivity of buspirone, these findings are nonetheless consistent with the ipsapirone findings (56) in arguing against 5-HT1A involvement in the pathogenesis of OCD.

MK-212 [6-chloro-2-(1-piperazinyl)-pyrazine] is a 5-HT agonist that binds to 5-HT1A, 5-HT1B, 5-HT1C, and 5-HT2 receptors (9). Bastani et al. (9) gave MK-212 (20 mg po) to 17 drug-free OCD patients and 9 controls. Compared with controls, patients had blunted cortisol responses to the probe, but prolactin and behavioral responses did not differ. Since selective 5-HT2 antagonists block the cortisol response to MK-212 in laboratory animals, this study suggests that 5-HT2 receptors may be subsensitive in OCD.

Metergoline is a nonselective 5-HT1/5-HT2 antagonist that blocks neuroendocrine and hyperthermic responses to m-CPP in humans. Three studies have examined responses to metergoline (4 mg po) in drug-free OCD patients. Zohar et al. in 1987 (96) observed no behavioral changes in 12 patients. In subsequent efforts to assess the effects of metergoline on m-CPP-induced responses in OCD (cf. above), Pigott et al. (67) found no behavioral effects of metergoline alone in 12 patients in a 1991 study or in 10 patients in a later study.

Benkelfat et al. in 1989 (10) administered metergoline (4 mg/day) or placebo for 4 days to 10 OCD patients whose symptoms had responded to long-term clomipramine treatment. Patients experienced worse OC symptoms during metergoline administration than during placebo, but this reflected a slight improvement from baseline in symptoms during placebo treatment; worsening of symptoms from baseline during metergoline treatment was even smaller and nonsignificant.

These studies indicate that nonspecific antagonism of 5-HT receptors has no unique effects on the neurobiology or phenomenology of OCD. Although such antagonism may interfere with the antiobsessional action of SRIs, relevant data are equivocal.

Barr et al. (8) administered double-blind shamcontrolled tryptophan-depletion tests to 15 OCD patients who had responded to long-term SRI treatment. Tryptophan depletion had no effect on OC symptoms, but depressive symptoms increased compared with the sham test. A similar worsening of depressive symptoms has been reported in remitted depressed patients being treated with antidepressants (24). These findings suggest that the vulnerability unmasked by tryptophan depletion is more central to the regulation of mood than of core OC symptoms.

Theoretical and pharmacotherapeutic considerations have traditionally generated interest in NE mechanisms of depression and panic disorder. Phenomenological similarities of these conditions to OCD have stimulated some studies of NE function in OCD.

Clonidine has been used as a pharmacological probe in three studies of drug-free OCD patients. Siever et al. in 1983 (82) gave clonidine (2 mg/kg iv) to nine OCD patients and nine controls and found the GH response blunted in patients. Lee et al. in 1990 (55) administered clonidine (2 mg/kg iv) to 10 OCD patients and 13 controls. They observed no differences between groups in GH or behavioral responses. Hollander et al. in 1991 (42) gave clonidine (2 mg/kg iv) to 18 OCD patients and 10 controls. Cortisol, GH, MHPG, and blood pressure responses did not differ between groups, but patients manifested transient improvement in OC symptoms.

These studies do not support major involvement of a2-adrenoceptors in the pathogenesis of OCD. Independent replication of the improvement in OC symptoms noted by Hollander et al. (42) would be important, but it will be difficult to establish that this does not reflect nonspecific sedation. These findings contrast with the blunted GH response to clonidine seen in depression and panic disorder (90).

Rasmussen et al. in 1987 (78) administered yohimbine (20 mg po) to 12 drug-free OCD patients and 12 controls. Groups did not differ in behavioral or MHPG responses to yohimbine, but patients had an increased cortisol response. This recalls the increased cortisol responses to yohimbine in depression and panic disorder (20). In panic disorder, however, yohimbine also induces panic attacks associated with an enhanced MHPG response (20).

Desipramine is a selective inhibitor of presynaptic NE reuptake. Given acutely, it increases GH secretion, apparently reflecting a2-adrenoceptor stimulation. Lucey et al. (59) gave desipramine (1 mg/kg po) to 10 drug-free OCD patients and 10 controls. Growth hormone and behavioral responses did not differ between groups. This contrasts with studies showing blunted GH responses to desipramine in depression.

Psychostimulants were among the first agents used in challenge studies of mood disorders. Although specific mechanisms vary, the principal effect of these drugs is to increase synaptic availability of monoamines via enhanced presynaptic release and/or reuptake inhibition. Their action as euphoriants is often attributed to their effects on DA function, although they have comparable effects on NE and 5-HT.

Three studies have examined the effects of psychostimulants in drug-free OCD patients. Insel et al. in 1983 (48) administered D-amphetamine (30 mg po) or placebo to 12 OCD patients; a significant improvement in OC symptoms was observed after active drug administration only. Joffe and Swinson in 1987 (50) found no change in OC symptoms in 13 OCD patients given open methylphenidate (40 mg po). In a subsequent placebo-controlled study of 11 OCD patients, Joffe et al. (51) again found that D-amphetamine (30 mg po) improved OC symptoms, whereas methylphenidate (40 mg po) had no effect.

Insel and Pickar reported in 1983 (49) that administration of the nonselective opiate antagonist naloxone (0.3 mg/kg iv) to two drug-free OCD patients in a placebo-controlled study caused worsening of OC symptoms. This contrasts with reports that high-dose naloxone infusions transiently decrease psychotic symptoms in schizophrenia.

Gorman et al. (35) gave 0.5-M sodium DL-lactate (10 ml/kg iv) to seven drug-free OCD patients and 48 drug-free panic disorder patients. Only one of the 7 (14%) OCD patients developed panic attacks, whereas 26 of the 48 (56%) panic disorder patients did so. These findings are consistent with other studies showing that known panicogens have little behavioral effect in OCD (7, 41, 43, 63).

In laboratory animals, arginine vasopressin (AVP) is associated with abnormal persistence of behaviors acquired under aversive conditioning and expression of stereotypic grooming behaviors. To examine the role of AVP in OCD, Altemus et al. (5) administered 3% hypertonic saline (0.1 mg/kg) over 2 hr to 12 drug-free OCD patients and 24 controls. Patients showed greater secretion of plasma AVP than controls, and, in many patients, the normal linear relationship between AVP and serum osmolality was disrupted. Additional findings of elevated cerebrospinal fluid levels of AVP and CRH in OCD patients in this study suggest that the effect of these synergistic arousal-producing hormones may be increased in OCD.

Lucey et al. (61) recently reported that the cholinesterase inhibitor pyridostigmine (120 mg po) caused greater GH responses in nine drug-free OCD patients than in nine controls. This evidence suggesting cholinergic supersensitivity in OCD is tempered by knowledge of pyridostigmine's poor entry into the brain.

SUMMARY

The pharmacological challenge strategy has proven remarkably productive in the study of anxiety disorders. Unfortunately, some of the best-replicated findings (e.g., the panicogenic effects of lactate and CO2) remain most resistant to simple explanations of mechanism. However, this may accurately reflect the complexity of the etiopathologic factors underlying these disorders.

Particularly encouraging are emerging data suggesting real differences in responses to the same challenge agents in different anxiety disorders. Panic disorder differentiates clearly from OCD, with recent evidence suggesting that social phobia and PTSD may also manifest distinctive response profiles. These observations underscore the lack of specificity inhering in the concept of anxiety that serves as the purported common factor linking these conditions to each other. Although it is disappointing that these findings so far have had relatively little clinical application, they could ultimately serve as the basis for a diagnostic nosology based on neurobiology as well as phenomenology. This process will be greatly facilitated if progress in understanding the neuroanatomy of anxiety as a phenomenon (71), as well as that of the discrete disorders (37, 47), can be married to pharmacological challenge findings.

Equally exciting are recent developments in which clinical treatments have led to preclinical advances (e.g., BZ receptor studies) and preclinical findings have suggested new treatments (e.g., CCK antagonists). The reciprocal nature of these interactions epitomizes an ideal in the relationship between preclinical and clinical neuroscience that is rarely realized. In the anxiety disorders, the pharmacological challenge strategy continues to serve this ideal well.

This work was supported in part by grants MH-25642, MH-30929, MH-45802, and MH-50641 from the U.S. Public Health Service and by grant DA-04060 from the State of Connecticut Department of Mental Health. Elizabeth Kyle prepared the manuscript.

published 2000