Human models as tools in the development of psychotropic drugs.

Despite the growing means devoted to research and development (R α D) and refinements in the preclinical stages, the efficiency of central nervous system (CMS) drug development is disappointing. Many drugs reach patient studies with an erroneous therapeutic indication andlor in incorrect doses. Apart from the first clinical studies, which are conducted in healthy volunteers and focus only on safety, iolerability, and pharmacokinetics, drug development mostly relies on patient studies. Psychiatric disorders are characterized by heterogeneity and a high rate of comorbidity. It is becoming increasingly difficult to recruit patients for clinical trials and there are many confounding factors in this population, for example, those related to treatments. In order to keep patient exposure and financial expenditure to a minimum, it is important to avoid ill-designed and inconclusive studies. This risk could be minimized by gathering pharmacodynamic data earlier in development and considering that the goal of a phase 1 plan is to reach patient studies with clear ideas about the compound's pharmacodynamic profile, its efficacy in the putative indication (proof of concept), and pharmacokinetic/pharmacodynamic relationships, in addition to safety, tolerability, and pharmacokinetics. Human models in healthy volunteers may be useful tools for this purpose, but their use necessitates a global adaptation of the phase scheme, favoring pharmacodynamic assessments without neglecting safety. We are engaged in an R α D program aimed to adapt existing models and develop new paradigms suitable for early proof of concept substantiation.

Phase 1 studies constitute a pivotal step in drug development. TTicir goal is to gather enough information to warrant the scientific value of phase 2 studies. The information to be collected includes the pharmacological actions of the drug, its side effects with increasing doses, its pharmacokinetics (PK) and metabolism, its mechanisms of action, and, if possible, early evidence of effectiveness.[1] 'The classic method of conducting phase 1 studies is much more limited (Table I). First-time-in-man, single -dose, and repeated-dose studies are carried out in healthy volunteers (HV), according to a parallel, double -blind (DB), placebocontrolled design. They are focused on PK, safety, and tolerability, seeking the maximal tolerated dose (MTD), which will be the basis for the choice of doses in subsequent patient studies. Using this scheme, many drugs have been developed in the wrong indication[2] or using inappropriate doses,[3] which led to failures or irrelevant studies, which then had to be replicated leading to delays, increased costs, and overexposure of patients to drugs.

It seems clear that, gathering data on pharmacodynamics (PD) and PK/PD relationships earlier would minimize these risks, bearing in mind that, in any case, further steps will face other major issues such as patient heterogeneity and placebo response.

Our usual way of conducting phase 1 studies takes these needs into account (Table II). As early as in the first-inman study, in addition to PK and safety/tolerability evaluation, we collect, basic, central nervous system (CNS) PD data, as well as peripheral PD data (eg, evidence of blood-brain barrier crossing, QTc or cardiac rhythm changes, minimal active dose, and dose effect), and attempt to sketch PK/PD relationships. This information is expanded in repeated-dose studies, which can be followed by PD studies in HV, conducted according to a crossover, DB, placebo-controlled design and using the most, appropriate tools, such as wake or sleep electroencephalography (EEG), cognition or functional imaging according to the molecule and its putative indication (see, for example, references 4 to 10). This allows patient studies to be undertaken with a better knowledge of the drug profile and the most appropriate doses.

In the last years, the necessity for a proof of concept (POC) approach has emerged. Simply stated, POC means that at some stage of the development, before launching large patient studies, it should be demonstrated that the drug does what it is supposed to do. Classically, POC studies are patient studies. Psychiatric disorders are highly heterogeneous syndromes, with a high rate of comorbidity (eg, dementia or schizophrenia coexist with affective or anxiety disorders, affective disorders with anxiety disorders, different anxiety disorders together, etc). Therefore, essentially two strategies are possible to improve patient recruitment. The first is to recruit, small, homogeneous groups of highly characterized patients (no comorbidity, homogeneous symptomatic profile, imaging or biological characteristics, such as genotype, if applicable, etc), which is an ambitious and lengthy operation. The second is to increase the sample size, with the hope that, number will compensate for heterogeneity. This also makes the trial longer and more expensive, and there is no guarantee that. the compensation will be obtained. In addition, concern about, withdrawing an ongoing treatment on the part of ethics committees, physicians, nurses, families, and patients further hampers recruitment.

Studies in HVs have advantages that, compensate for these difficulties. HVs arc easier to recruit. They make more homogeneous - or less heterogeneous - populations than patients and, since more subjects are available, homogeneity can be improved through specific selection criteria. By definition, these subjects are healthy, have a lower risk of complications, and tolerate procedures better. In addition, they have no expectations about, the treatment, which can minimize placebo/nocebo effects. By definition, they arc also volunteers and get paid, and are thus more compliant.

Conducting POC studies in HVs implies the recourse to models or symptom provocation. These challenges can be understood as the provocation not of a complete clinical picture, but of some core signs and symptoms. The goal can be to produce and study functional markers because they are often more sensitive - and hopefully more reliable - than clinical signs and symptoms. Therefore, it is possible to use less stressful provocative procedures, which further increases subjects' comfort and compliance. Symptom provocation in HVs must comply with some basic rules (Table II), as stated by D'Souza et al.[11] In the context of drug development, a model must, also induce target signs and symptoms in a reasonable proportion of HVs. 'These signs and symptoms must be accompanied with reliable functional changes, which can be used as biomarkcrs and display low intcrindividual and mostly intraindividual variability, to warrant good test-retest reliability and permit, the assessment, of drugs effects in a crossover, placebo-controlled design.

Although the principle of a POC approach in HVs is appealing, we must be cautious in putting it into practice. Even when the provocation procedure is simple (eg, a single agent with well-known neurochemical properties, in the case of a pharmacological model), the totality of the neurochemical consequences of its administration are seldom known. The same holds for the new compound studied and for its interaction with the challenge. As a consequence, a positive result (ie, reversal or prevention of the challenge's effects by the new drug) is undoubtedly a clue to efficacy, but. a negative result can hardly be taken as the basis for a “no-go” decision. This often makes pharmaceutical companies reluctant to add a POC study in HVs to their development plan, arguing that, in case of negative results, it could merely delay it. and increase costs. In fact, introducing POC studies in HVs implies further enhancing the global phase 1 scheme (Table I). In this “enhanced development plan,” the single-dose study has the same design and goals as the regular one. The repeated-dose study merges the former repeated-dose and PD HV studies, ie, it. is conducted according to a crossover (per dose), placebo-controlled design. Provided that a single administration study has shown good tolerability in an HV group close to the target population (eg, elderly HVs for cognitive enhancers), this study can be conducted in such a group. A model, if available, can also be used in this study, by adding an administration the day after the classic PK and PD assessments. This avoids wasting HVs and resources, and maximizes the chances of positive results; however, it requires paying close attention to tolerability and safety, which must, be verified before a challenge is added to a repeated-administration study.

Models at FORENAP

Few models are available for routine use in drug development. We have launched a program to adapt existing models for this purpose (Table III), following the principles described above. Below we discuss the rationale for each of these models, as well as preliminary results when available, at. least those which are not covered by confidentiality agreements.

Alzheimer's disease and age-related cognitive impairment

The scopolamine model

The scopolamine model is based on the cholinergic hypothesis of aging and Alzheimer's disease (AD). Its theoretical drawback is that scopolamine is a nonselective muscarinic blocker, whereas selective muscarinic Mr blockade could be considered to better modelize the status of the cholinergic system in AD.[12] Nevertheless, it is a well-established model, producing cognitive defects close to those observed in mild AD and FRG changes consisting of an increase in 5 and - to a lesser extent - 6 bands, and a decrease in a and p power.[13] The scopolamine model has been widely used in clinical experimental pharmacology,[14-23] as well as in the assessment, of cognitive enhancers.[24-40] In our hands, a 0.5-mg subcutaneous injection of scopolamine in young HVs induced impairment, in immediate and delayed word recall, multiple choice reaction time and accuracy, and the digit symbol substitution test. In quantified EEG, it. reduced total power and induced an increase in S and a decrease in 6, a, and p absolute power; in relative power analysis, the 8 and p band activity was increased and that of the 6 and a bands decreased () An interesting feature of this model is that is can be reversed or prevented not. only by cholinomimetic drugs, but. also, as we have found, by compounds without direct, cholinergic effects.[30,31,38,41]

The lorazepam model

Benzodiazepines (BZDs) are known to induce sedation, psychomotor impairment, and anterograde amnesia, leaving retention and retrieval spared.[42] Although cognitive impairment is a class effect, differences between different BZDs have been reported, independently of their elimination half-lives.[43-45] A dissociation between the cognitive and sedative effects of drugs has also been described.[46] Lorazepam has dose-related memory- and attention-impairing effects.[43,44,47] It has been suggested[48] that the profile of lorazepam-induced cognitive impairment is close to that observed in Korsakoff's syndrome, whereas scopolamine rather mimics AD. Some studies[47,49] were unable to distinguish the effects of lorazepam from those of scopolamine. Both drugs were shown to have similar effects on verbal priming[50] and in a face-name associative encoding task,[51] and as well as on associated functional magnetic resonance imaging (fMRI) activation patterns. On the other hand, differential effects were found on logical reasoning, immediate and delayed recall,[52] and priming for human faces.[53]

BZDs have well-known effects on EEG. Changes in p amplitudes seem to reflect their interaction and intrinsic efficacy at the GABAA-BZD (GABA, y-aminobutyric acid) receptor complex; their effects on p and a activity their anxiolytic, anticonvulsant, and sedative properties; and 8-induced changes their hypnotic action.[54]

In our hands, an oral dose of 2 mg lorazepam impaired immediate and delayed word recall, multiple choice reaction time and accuracy, and digit, symbol substitution test, but. had no effect, on flicker fusion frequency. In quantified EEG, lorazepam's effects were dose -dependent in length and intensity, increasing 8 and P power, and decreasing the power of the 6 and a frequency bands ()

The low-dose ketamine model

Ketamine infusion produces positive, negative, and cognitive symptoms reminiscent, of those observed in schizophrenia.[55-65] A hypoglutamatcrgic state has also been proposed as the substratum of late-stage AD.[66] Studies focused on ketamine-induced cognitive impairment, should separate the latter from the psychotomimetic effects of ketamine, which is possible using lower doses.[64]

Nonpharmacological approaches

Functional (positron emission tomography [PET] and fMRI) studies on the neural correlates of cognitive aging basically describe two cases.[67] In one, performance and brain activation during the task are lower than in young controls; this is also the case for episodic memory and conflict, resolution tasks. The second consists of preserved performances associated with enlarged activation, engaging more brain regions, such as during working memory tasks. Our fMRI activation maps, obtained during a spatial “n-back” working memory challenge are in agreement, with these data () Our hypothesis is that activation patterns in elderly volunteers should be closer to those of young volunteers after administration of a cognitive enhancer. Indeed, PPT scan and fMRI studies in young volunteers have shown that physostigmine infusion improved working memory performances and reduced task -related activation.[68-70]

Anxiety

Panic attack model: CCK-4

The idea of using cholecystokinin tetrapeptide (CCK-4) as a panic probe came from experiments showing that BZDs antagonized CCK-8S in the rat,[71] as well as from the serendipitous finding that a 70-ug CCK-4 injection produced panic-like feeling in healthy humans.[72] In subsequent studies,[73-91] CCK-4 induced panic attacks in 0% to 70% of HVs and these attacks were quantitatively and qualitatively similar to those reported by patients. Attack incidence and severity of symptoms were dose-dependent, although discordant results have been described with the same dose and a considerable overlap exists in the rate of response to different doses. The dose of 50 ug seems to give the most homogeneous response rate, ranging from 47% to 65%. Test-retest reliability has been poorly assessed. Two studies - although not specifically designed for this purpose - reported a decrease in the number and intensity of panic symptoms,[79,88] as well as in the incidence of panic attacks.[79]

In HVs, lorazepam prevented CCK-4-induced panic,[73] as did the CCK-4 receptor antagonist. CI988,[80] propranolol,[87] ondansetron after acute but not repeated administration,[88] atrial natriuretic peptide,[89] and vigabatrin.[90] PET scan studies found an increased cerebral blood flow concomitant of anxiety response in the claustrum-insularamygdala region bilaterally and in cerebellar vermis and anterior cingulate gyrus.[92,93] Our fMRI results confirm these data.

Anticipatory anxiety: behavioral model

Recent research suggests that the neurophysiological mechanisms underlying anxiety disorders are closely related - if not identical - to those underlying the emotion of fear.[94] This provides the rationale for using behavioral models based on fear induction or anticipation of an avcrsive stimulus. In behavioral models, an anxious state is induced by presentation of stimuli having an aversive emotional content, via any sensory modality. A major drawback of using intrinsically fearful stimuli is that, the fear or aversion elicited can vary according to volunteers' traits and experiences. Aversive conditioning (in which an emotionally neutral or conditioned stimulus [CS] is paired with an aversive - or unconditioned - one [UCS], usually in different sensory modalities) allows for a more homogeneous response within the subject population by adjusting the aversive nature of the UCS on an individual basis. Although amygdala activation is considered to be central to anticipatory anxiety,[94-96] other regions are also activated during classical conditioning, eg, right, orbitofrontal, dorsolateral prefrontal, inferior and superior frontal, inferior and middle temporal cortices, and left superior frontal cortices,[9798] anterior cingulatc, and insula,[99] according to the paradigm used. The study of many regions together can lessen the consequences of “missing” the amygdaloid complex activation, which is transient even when the UCS continues to be presented in association with CS.[95,98,99] Our results confirm the merits of this approach.

Depression: the tryptophan depletion challenge

The rationale and results of a recent study with this model are given elsewhere in this volume.[100]

Schizophrenia

The apomorphine model

For decades, dopamine transmission abnormalities have been thought to be involved in the pathophysiology of schizophrenia,[101] justifying the stimulation of dopaminergic pathways as a model of schizophrenia in HVs. Apomorphine, a nonselective dopaminergic agonist, has a rapid phase of absorption and distribution in the periphery (20 min) as well as the brain compartment. (30 min) in humans[102] and is an ideal pharmacological tool because it has minor psychotropic effects in both HVs and psychiatric patients. We have characterized apomorphineinduced topographic changes in neurophysiological markers using a 28-lead multielectrode montage in HVs. To ensure that, observed modifications are of central and not of peripheral origin, subjects were pretrcated with domperidone, a dopamine antagonist that does not cross the blood-brain barrier. We assessed drug-induced modifications in EEG/event-related potential measurements at different time points after subcutaneous injection of apomorphine.[103] As expected, the effects of apomorphine on EEG were partially opposite to those reported for classic[104] and atypical[105,106] neuroleptics, with an increase in fast P power and decrease in 6 relative power. This model was validated using haloperidol, which antagonized the acute effects of apomorphine.[107]

The ketamine model

N-Methyl-D-aspartate (NMDA) receptor blockade by ketamine infusion in HVs is acknowledged to be a good model of schizophrenia, reproducing positive, negative, and cognitive symptoms.[55-65] Despite evidence that ketamine modulates dopamine striatal concentration,[108-111] its clinical effects were not reversed by haloperidol in patients[112] or in HVs,[61] or olanzapine,[113] but were blunted by clozapine in patients with schizophrenia.[114] This inconsistent effect of antipsychotics could be dose-related. The above studies used ketamine doses of 0.1 to 0.9 mg/kg in bolus or 1-h infusion, whereas we use 0.16 to 0.54 mg/kg in a 2-h infusion.

Conclusion

There is an agreement on the need to increase the efficiency of drug development. Whatever the improvements in the chemical and preclinical steps, clinical development strategy remains critical. Human models in HVs are obviously not a panacea. They are not applicable to any situation and the validity of the different provocation procedures is uneven. Their optimal use is within what we call an “enhanced development plan,” which requires improvements in safety data processing. Nevertheless, when properly used, human models can secure phase 1 study results, be of help in a “go” (more than in a “no-go”) decision, and therefore improve the safety and efficiency of patient studies, leading to a reduction in both time and resources.

Table I

Three ways of conducting phase 1 studies. MTD, maximal tolerated dose; PK, pharmacokinetics; PD, pharmacodynamics; BBB, blood-brain barrier. *Basic PD includes BBB crossing, minimal active dose, dose effect, and non-central nervous system (CNS) PD. **Basic PD without BBB crossing. ***Refined PD (model): proof of concept, minimal active dose, dose effect, and non-CNS PD.

	Classic	curent (usual)	Echanced
Single-dose study
Population	Young, healthy volunteers	Young, healthy volunteers	Young, healthy volunteers
Design	Parallel, double-blind, vs placebo	Parallel, double-blind, vs placebo	Parallel, double-blind, vs placebo
Objectives	Safety/tolerability (MTD) PK	Safety/tolerability (MTD) PK Basic PD* PK/PD relationships	Safety/tolerability (MTD) PK Basic PD* PK/PD relationships
Repeated-dose study
Population	Healthy volunteers	Healthy volunteers	Healthy volunteers
Design	Parallel, double-blind, vs placebo	Parallel, double-blind, vs placebo	Grossover, double-blind, vs placebo
Objectives	Safety/tolerability (MTD) PK	Safety/tolerability (MTD) PK Basic PD* PK/PD relationships	Safety/tolerability (MTD) PK Basic PD* PK/PD relationships
Next step
Population	Patients	Health volunteers	Patients
Design	Parallel, double-blind, vs placebo	Grossover, double-blind, vs placebo	Parallel, double-blind, vs placebo
Objectives	Safety Efficacy	Refined PD PK/PD relationships	Safety Efficacy

Table II

Criteria to justify symptom provocation in humans.[11]

The safety and protection of subjects is ensured

The study is expected to provide information of critical value

The proposed methodology is undoubtedly adequate to answer the question (testable hypotheses, reliable and valid measures, adequate power, appropriate data analysis, and relevant controls)

The study is the best of only feasible method of answering the question

Table III

Models available or in development at FORENAP. AD, Alzheimer's disease; CCK-4, cholecystokinin tetrapeptide; EEG, electroencephalography; ERP, event-related potential; fMRI, functional magnetic resonance imaging; MEG, magnetoencephalography.

Indication	Method	Marker	Status
AD/cognition
	Scopolamine	EEG/ERPs	Routine
	Lorazepam	EEG/ERPs	Routine
	Low-dose ketamine	EEG/ERPs	Validation underway
	Nonpharmacological method	fMRI	Validation underway
Anxiety
Panic	CCK-4	fMRI	Validation underway
Anticipatory	Behavioral	fMRI	validation underway
Depression
	Tryptophan depletion	Sleep EEG	validation completed
Schizophrenia
	Apomorphine	EEG	Routine
	Ketamine	EEG/MEG	validation underway