Phase II clinical trials have progressively become a cornerstone of the drug development process since beyond dose-effect studies, phase II trials also largely participate in the evaluation of drug efficacy. Due to the development of randomized, double-blind, placebo-controlled studies, phase II trials are actually very close to phase III studies in terms of methodology. Indeed, groups of patients should be uniform and representative with regard to the disease, baseline characteristics and expected response to treatment. Irrespective of single or repeated doses and the evolution modalities of the pathological process, phase II trials are usually conducted according to a parallel protocol whereas a crossover approach remains strictly limited. For psychopharmacology, as in many other fields, great attention is also paid to the evaluation of the clinical effect as well as to the sources of variability. In this way, best knowledge on both pharmacokinetic and pharmacodynamics genetic polymorphisms will perhaps play an important role in the near future.
 

Keywords: Phase II; clinical trial; methodology; psychopharmacology.   

 

Phase II clinical trials, which are mainly designed to detect, in selected patients, the pharmacodynamics activity of a drug and the range of active doses, have progressively become the cornerstone of the drug development process. Beyond dose-effect studies, phase II trials could also contribute to demonstrating the potential efficacy of a drug where phase III trials will subsequently indicate whether such pharmacodynamics activity could be therapeutic. Unlike phase I trials which assess tolerated doses and pharmacokinetics over the broadest range possible in a small number of healthy volunteers, phase II trials are conducted in larger groups, using a wide range of really active doses, leading to some similarities with the methodology of phase III studies. In this brief review, we will discuss general aspects as well as some specific concerns regarding the methodology of phase II clinical trials, giving some examples from psychopharmacology.

2. General Principles of Phase II Trials
2.1 Main objective: dose-effect studies
 
The purpose of studying the relation between drug dose and its effect in humans is first to define the limits of effective doses. The lower limit or threshold of clinical efficacy is determined as the minimal dose producing a significant effect when the upper limit or toxicity threshold is defined as the maximal dose that can be administered without producing significant toxicity. Defining such thresholds could be very easy in drugs in which a clear dose-effect relationship is shown as illustrated by the example of the benzodiazepine hypnotic drug temazepam {C}[1]{C}. Thus, for most patients, 10 mg of temazepam in the evening gives the same response as the placebo. If the temazepam dose is increased to 20 mg, there are about 70% of responders with few side effects (sedation the following morning). With 30 mg of temazepam, more than 70% of responders report an increase in sleep time or a decrease in night awakenings but more than 30% of these patients present difficult waking or cognitive impairment the following morning {C}[1] . In such a context, it is very easy to use the highest efficacy and the lowest toxicity to determine the average effective dose. When applicable, dose-effect studies could also include data on changes in effects as a function of the dose administered and of the drug concentrations in an easily sampled biological medium (most often plasma but also urine or saliva). To complement phase I data, such studies could then suggest which administration route would provide the best expected effect, detect the possible presence of active metabolites of the drug, reveal any phenomena of tolerance or sensitization of receptor sites and allow the analysis of possible interactions between various drugs.

 

2.2 Medication doses and placebo-arm

 

During phase II clinical trials, the drug doses to be studied range from a minimal dose, most often corresponding to a dose that has produced a measurable pharmacological effect in healthy volunteers during phase I trials, to a maximal dose generally lower than the toxicity threshold as also defined in phase I trials. The number of doses to be tested could be variable, most often from 3 to 6, but should remain within the limits of feasibility of the protocol. In spite of some controversy, particularly when a proven therapy is already available, the use of a placebo-arm should be discussed {C}[2, 3]{C}{C}. Such an approach remains one of the strongest ways of providing irrefutable evidence that the drug is really effective. In pathologies such as depression or bipolar disorder, there is growing evidence that placebo response as well as response to the medication have increased significantly in recent years suggesting that the inclusion of a placebo-arm has major scientific importance in trials {C}[4, 5]{C}{C}. Moreover, the use of a placebo-arm could increase the statistical power of the study allowing inclusion of fewer patients in each dose regimen or shorter study duration. Hence, if fewer patients are included, the overall risk related to a potentially toxic treatment or placebo when treatment is very effective, is powerfully reduced thus providing an important benefit for subsequent studies as well as for patients with the same pathology. In this way, except for dose-escalation protocols, phase II clinical trials in psychopharmacology are usually conducted with 3 doses of the drug compared to a placebo and a control or reference product. Many clinical trials for new psychiatric drugs fail to detect a statistical difference in efficacy between treatment and placebo. In antidepressant studies, failure rates of confirmatory trials are estimated to be over 50% even for treatments that are registered by the Food and Drug Administration {C}[6]. High placebo response, common in psychiatric trials, is considered among the main reasons for trial failures.

 

2.3 Population of patients and randomization

 

Except for certain pathologies such as schizophrenia or melancholic depression, most patients included in phase II clinical trials in psychopharmacology are generally not hospitalized. Such situations lead to clinical follow-up which is relatively close to usual clinical practice as well as to similar problems or difficulties as phase III studies regarding trial management. One of the main conditions in such trials is that groups of patients must be sufficiently uniform and representative with regard to the natural history of the disease, baseline characteristics such as age or gender as well as expected treatment response. To ensure homogeneity, regarding the definition of eligibility and exclusion criteria, patients are randomized in one of the different treatment groups. Sometimes, due to the important heterogeneity of the disease as in dementia where natural course or severity of symptoms could be extremely variable, it would be useful to perform a stratified randomization [7, 8]. Nevertheless, since the requirements of homogeneity and representativeness are often contradictory, a compromise must usually be found in agreement with the goal of the study.

 

2.4 Number of subjects

 

Calculating the sample size is another problem encountered during study conception. This should be defined in advance and should be sufficient to detect a difference between treatment regimens. The size of the sample should not be too big either for the patient because of exposure to potential toxicity or for the promoter because of cost. The number of subjects to be included mainly depends on the variability and the amplitude of the variation of the parameter chosen to measure the expected effect. In the case of phase II trials, with a five-arm design, such a high number of patients could rapidly lead to difficulties with recruitment and implicate the development of very large multicenter studies. Because it is sometimes difficult to accurately predict the number of patients needed to demonstrate a clear difference according to treatments and dosages used, it might be helpful to perform an intermediate analysis requiring specific adaptation of statistical analysis procedures [9].

 
3. Other Methodological Concerns
3.1 Single versus repeated doses
 
To improve our understanding of dose-effect relationship it is important to keep in mind some differences related to the use of single or repeated doses. During the administration of a single dose to a group of patients, changes in the effect and plasma concentrations of the drug are studied over a period of time. If the effect is measured in binary mode, the proportion of patients who attain the efficacy threshold is determined as well as the time required to reach this threshold and maintain it. The plasma concentration corresponding to the efficacy threshold is also determined. In continuous mode, the appearance of the effect is noted as well as the maximum reached and the fall. The dose-effect relation can be studied by considering the maximal effect observed for each drug dose. Assessment of the plasma concentration-effect relationship can be based on analysis of the plasma concentration corresponding to the maximal effect or on correlation between the area under the curve for plasma concentration time and effect time. It could be applicable for example by studying hypnotics in patients on one night, particularly in healthy volunteers. Simultaneous measurement of the drug effect and plasma concentrations sometimes reveals a hysteresis phenomenon when the effect in plasma concentrations persists during the fall (figure 1). This phenomenon can be related to the persistence of the drug at the activity site, the late elimination of an active metabolite or more rarely to the sensitization of receptor sites during the study {C}[10].

 

During repeated administrations of the drug, changes in effect are observed as plasma concentrations gradually increase until a plateau is reached. The effect is then measured in steady state after a period of administration equal to at least 5 times the elimination half-life of the drug. If the effect is measured in binary mode, the number of doses and plasma concentrations required to reach the efficacy threshold are noted. The proportion of responders attaining clinical efficacy is determined for the steady state of plasma concentrations corresponding to each dose. If the effect is measured in continuous mode, changes are studied in terms of the time-points during administration of repeated doses when plasma concentrations increase and then during the steady state of these concentrations. In such conditions, the minimum treatment period should be equal to at least 5 times the elimination half-life of the drug, whereas in nonlinear kinetic drugs the treatment period should be longer according to the existence of active metabolites.

 

 

 

Figure 1: Schematic plots showing hysteresis phenomenon. The arrows indicate the direction of increasing time.

 

3.2 Experimental protocols

 

Whereas in the past, phase II clinical trials in psychopharmacology were often open and non-controlled studies involving few patients, current regular practice is to perform controlled and randomized studies in large parallel groups. A crossover approach can also be used in some studies but the conditions for its application remain strictly limited.

 

3.3 Protocol in parallel groups

 

In this study design, the basal state of patients is determined during an initial study most often performed single-blind with a placebo then after patients are randomly divided into parallel groups and, depending on the administration procedure, receive a single dose per group or repeated doses for a period sufficient to reach the steady state of plasma concentrations. When possible, a 5-arm controlled study is performed: three doses of the compound are then compared with a control drug and a placebo (figure 2). The parallel dose-response study gives group mean dose responses, not the distribution or shape of individual dose-response curves. The disadvantage of such a parallel dose-response study is that the precision of the inference is driven by between-subject variability, which usually requires greater sample sizes to increase the precision of the estimates for drug effect {C}[10]{C} . A good example of a study with a parallel group design that includes a placebo, a control product (imipramine) and a test compound is the study of the effects of phenylpiperazine antidepressant nefazodone (dose range 100-600 mg/day) in 284 patients suffering from major depression {C}[11]{C}. This 8-week study showed that the antidepressant efficacy of nefazodone was comparable with that of imipramine, with both drugs significantly better than placebo in a variety of outcome measures {C}[11]{C}.

 

 

Figure 2: Schematic representation of parallel protocol in phase II dose-effect study.

3.4 Crossover protocol
​
In this type of protocol, the patient is his own control. Basal state is determined for each patient and the different doses of the drug are administered successively and studied for their effect (figure 3). Such a protocol can be used for administration of the drug in a single dose, in which case each dose is followed by a sufficient wash-out period, so that all effects are dissipated before administration of the following dose (figure 3). It can also be used for administration of repeated doses, in which case each dose level is of sufficient duration to allow the steady state of plasma concentrations to be reached. In these crossover protocols, the order of the doses administered to each subject can be determined (e.g. in increasing order) or randomized (figure 3). Placebo periods could also be included, either during the protocol when administration is randomized or most often at the end of the trial when doses are administered in increasing order. These placebo periods allow the stability of the pathologic state of patients to be verified and to demonstrate the possible existence of a placebo effect. In fact, this type of protocol is very difficult to carry out, except with healthy volunteers in phase I trials when some pharmacodynamic parameters are measured. One example of such an approach is the study of the effects of fexonadine (H1-receptor blocker, 80-180 mg) on cognitive and psychomotor functions as compared with controls (loratidine, 10 mg ; promethazine, 30 mg) [1]. This 6-way crossover study of 24 healthy volunteers showed that fexonadine, at doses up to 180 mg, appears free from disruptive effects on psychomotor and cognitive functions unlike the effect of promethazine [1].

 

 

Figure 3: Schematic representation of crossover protocol to evaluate dose-effect relationship.

 

Another more recent example, is the study of the effects of the agent 3-(24-dimethoxybenzylidone) anabaseine (DMXB-A) a partial alpha7-nicotinic agonist. Thirty-one subjects with schizophrenia received DMXB-A at two different doses and placebo for periods of 4 weeks in a three-arm, two-site, double-blind, crossover phase II trial. The MATRICS Consensus Battery assessed cognitive effects and the Scale for assessment of Negative Symptoms (SANS) and Brief Psychiatric Rating Scale (BPRS) assessed clinical effects. Subjects continued their current antipsychotic drug during the trial and were nonsmokers [12].

 

4. Evaluation of the clinical effect
4.1 Criteria of therapeutic efficacy

Unlike phase III studies, the criteria for assessing effect in phase II trials may not be strictly therapeutic. However, if the effect measured is therapeutic in nature, the type of effect observed (binary or continuous variation mode) could strongly influence the analysis of trial results. It is thus advisable to define the criteria of therapeutic efficacy carefully and before the beginning of the study. The same is true for the criteria of clinical and biological tolerance of the drug. Determination of the clinical effect of a drug requires that the variation of a drug-related parameter be distinguished from variations related to a spontaneous change in the state of the patient or to an error in measuring the parameter. On the one hand, this implies that the spontaneous variability of the parameter is known, either by measurement of the parameter both before and after drug administration or in the same experimental conditions when patients receive a placebo. On the other hand, determination of the clinical effect implies that the sensitivity and reliability of the method used to ascertain the effect are known, which ensures that the error in measurement is slight relative to variations of the drug-related parameter. Subsequently, the method of measuring the clinical effect depends on the type of effect observed.

 

4.2 Effect analyzed according to a binary mode

 

In this situation, the effect is either present or absent. This is true, for example, for the therapeutic effects of hypnotic drugs (induction of sleep or not). The effect can also correspond to a significance threshold dependent on spontaneous variability of the pathologic situation. The application is possible in psychopharmacology using the notion of responders or non-responders but such a definition is not always so easy (European College of Neuropsychopharmacology, 1995). When the effect is evaluated according to a binary mode (all or nothing), the probability of the presence or absence of the effect is studied for each dose. The dose-effect curve shows the proportion of subjects responding to treatment for each of the doses studied (figure 4). The dose-effect relationship is most often of sigmoid form and the maximal effect is defined as the value where all of the patients are treated efficaciously (figure 4).

 

 

 

Figure 4: Representation of the dose-effect relationship according to a binary mode analysis (presence or absence of the effect)

 

ED50 defines the dose necessary to produce the effect in 50% of the population studied. Emax corresponds to the maximum effect of the tested drug when observed in the overall studied population.

 

4.3 Effect analyzed according to a continuous mode

When the effect shows a continuous variation from 0 (absence of the effect) to a maximum during the increase of doses, the concentration-effect relationship is defined by a parametric model (figure 5). In practice, it is rarely possible in human to measure a maximal effect and the effect is not always proportional to the blood concentration of the drug. Nevertheless, the dose-effect relationship most often assumes a sigmoid form whose middle part corresponds to a linear relation between effect and dose. More complex models have been proposed for the analysis of concentration-effect relationship, either by using logarithmic transformation of the concentration (corresponding to a linear approximation of the hyperbolic model in the zone of effects representing 20 to 80% of the maximal effect ; figure 5) or by integrating the pharmacokinetic and pharmacodynamic model by calculating the concentrations of the drug at the theoretical site of the effect {C}[10]{C}.

 

 

 

Figure 5: Concentration-effect curves as representations of the dose-effect relationship according to a continuous mode analysis

 

EC50 corresponds to the concentration which produces 50% of the maximum effect in the studied population. Emax represents the theoric maximal effect which is usually not measurable in clinical condition.

 

4.4 Factors of Variation in the Dose-Effect Relationship

Although the main purpose of the study of drug-effect relationship seems simple in a given patient, the situation may be much more complex when the relationship is defined in a population of patients. The purpose of clinical trials of new drugs conducted with relatively small groups of patients is to determine a general law applicable to the majority who will ultimately receive the drug. In fact, there is considerable difficulty in applying this general law to individual treatment since the dose-effect relationship is subject to the influence of many factors.

 
4.5 Impact of age and gender

There is increasing evidence that age and gender could possibly influence the dose-effect relationship. Older adults usually respond less predictably than younger adults to most medications hence requiring lower daily doses to achieve desired therapeutic effects and minimize adverse effects and toxicity {C}[13]{C}. Such differences in response could be the result of pharmacokinetic and pharmacodynamic changes associated with aging as for example declining renal function, alteration of body composition, acquired inhibition of drug metabolism or specific receptor and neurotransmitter changes {C}[14]{C}. The modification of response to a drug is also supported by experimental data since older animals for instance seem less responsive to selective serotonin reuptake inhibitors (SSRI) as evaluated in a despair model {C}[15]{C}. Despite a possible decrease in serotonin receptor in the aging brain of humans {C}[16, 17]{C}{C}, the low number of clinical studies which have included patients older than 75 years, provided no definitive conclusions regarding differences in response between antidepressant drugs {C}[18, 19]{C}{C}. Some similar concerns could also be discussed about the impact of gender in antidepressant response, particularly from the possible implication of antidepressant-metabolizing enzymes{C}[20]{C}. Nevertheless, beyond the problem of metabolism, there are also arguments in favor of pharmacodynamic gender-related differences since in women the response to antidepressants might be better when using SSRI or MAO inhibitors rather than tetracyclic molecules {C}[21, 22]{C}{C}.

 

4.6 The role of underlying or associated pathologies

Beyond patient characteristics, some conditions related to the underlying pathological state such as etiology, severity or time course of the disease could have a significant impact on the dose-effect relationship. For example, some differences might exist in terms of drug response between patients suffering from acute reactive depression, patients presenting with a depressive episode during the course of a bipolar disorder as well as according to the severity of depressive symptoms {C}[23]{C}{C}. The same might be observed in patients suffering from dementia according to the severity of their disease as well as the etiology of their cognitive decline or the presence of associated diseases {C}[8, 24]{C}{C}. In such conditions, as previously discussed, it might be helpful to perform a stratified randomization as well as ensuring that placebo groups do not present unexpected progression due to a possible recruitment bias as recently discussed for the efficacy of cholinesterase inhibitors in vascular dementia {C}[24]{C}.

 

4.7 Pharmacokinetic: an important source of variability

The problem of inter and/or intra-individual variations in drug pharmacokinetics sometimes depending on a genetic polymorphism should be taken into account. This is particularly the case for all drugs that are metabolized through P450 cytochrome system {C}[25, 26]{C}{C}. Due to the possible impact of genetic traits, there is actually a trend to integrate some pharmacogenetic data both in clinical trials and practice {C}[27]{C}. Beyond the problem of P450 metabolism, pharmacokinetic variability could also be based on the chemical properties of the drug, especially on chiral form. In this way, the drug development process of antidepressants actually provides increasing evidence for different pharmacokinetic profiles based on the enantiomeric form of the drug {C}[28, 29]{C}{C}. Such recent data lead to the design of clinical trials based on pure enantiomer which could be associated with higher efficacy and lower toxicity {C}[30]. Finally, an important role could also be played by the presence of one or more active metabolites whose kinetics might differ from that of parent products whereas the effect could also vary depending on environmental factors, schedules, and food or living conditions.

 

4.8 What about the pharmacodynamic target?

​
The presence of non-responders to a treatment during a clinical trial could play an important role as a confounding factor in estimating the dose-effect relationship {C}[31]{C}. In parallel to the definition of responders according to appropriate evaluation criteria, recent literature argues for a possible impact of genetic factors on drug-response through polygenic factors as well as specific mutations of drug targets {C}[32, 33]{C}{C}. Indeed, it seems that among patients with Alzheimer disease, cholinesterase inhibitor responders could present a preferential combination of alleles in apolipoprotein E, pre-seniline 1 and pre-seniline 2 genes {C}[34]{C}. Concerning mutations of drug targets, there are numerous clinical data arguing that specific drug receptor or transporter polymorphisms could be potent predictors of drug-response in schizophrenia, depression as well as in anxiety {C}[35, 36]. Such an approach, which is not yet applicable in clinical practice, could have important economic and ethical consequences in the future.

The classic phase II methodology as described above is ideally applicable to situations where a clear drug-response could be demonstrated. For a long time, psychopharmacology had very poor phase II studies due to insufficiency in evaluation criteria. Nowadays, in many fields such as depression or anxiety, a dose-effect relationship could be shown with various drugs such as venlafaxine or SSRI [37, 38]. Nevertheless, the determination of optimal dose could remain a major concern when there is no readily apparent therapeutic effect or predictive pharmacologic effect as in the treatment of schizophrenia. In this context, there is a need to improve the methodology of clinical trials [39] including the quality of rater scoring in clinical trials could also contribute to dose-effect variability [40]. In the near future, we could also postulate that a pharmacogenetic approach will help to optimize drug development and therapeutics, increasing efficacy and safety and reducing side-effects and unnecessary costs.

We are grateful to Nikki Sabourin-Gibbs, Rouen University Hospital, for her help in editing the manuscript.

1. Hindmarch I (1979) Effects of hypnotic and sleep-inducing drugs on objective assessments of human psychomotor performance and subjective appraisals of sleep and early morning behaviour. Br J Clin Pharmacol 8:43S-46S.

2. Lavori PW (2000) Placebo control groups in randomized treatment trials: a statistician's perspective. Biol Psychiatry 47:717-23.

3. Young SN, L Annable (2002) the ethics of placebo in clinical psychopharmacology: the urgent need for consistent regulation. J Psychiatry Neurosci 27:319-321.

4. Keck PE Jr (2000) Placebo effect in randomized, controlled studies of acute bipolar mania and depression. Biol Psychiatry 47:748-755.

5. Walsh BT (2002) Placebo response in studies of major depression: variable, substantial, and growing. JAMA 287: 1840-1847.

6. Dodd S (2014) Application of the Gradient Boosted method in randomised clinical trials: Participant variables that contribute to depression treatment efficacy of duloxetine, SSRIs or placebo. J Affect Disord 168:284-293.

7. Chaumet-Riffaud P (1993) [Therapeutic trials in Alzheimer disease. Selection--recruitment and stratification]. Therapie,. 48:201-215.

8. Dib M (2001) Methodological issues and therapeutic perspectives in vascular dementia: a review. Arch Gerontol Geriatr 33:71-80.

9. Boutitie F (1992) [Monitoring of clinical trials and interim analysis. 2. Statistic methods]. Therapie,  47:351-355.

10. Emilien GW, van Meurs,  JM Maloteaux (2000) The dose-response relationship in phase I clinical trials and beyond: use, meaning, and assessment. Pharmacol Ther 88:33-58.

11. Rickels K (1994) Nefazodone and imipramine in major depression: a placebo-controlled trial. Br J Psychiatry 164:802-5.

12. Freedman R (2008) Initial phase 2 trial of a nicotinic agonist in schizophrenia. Am J Psychiatry 165: 1040-1047.

13. Zubenko GS, Sunderland T (2000) Geriatric psychopharmacology:  why does age matter? Harv Rev Psychiatry 7: 311-333.

14. Pollock BG (1998) Psychotropic drugs and the aging patient. Geriatrics 53:S20-S24.

15. Bourin M (1998) Evaluation of efficacies of different classes of antidepressants in the forced swimming test in mice at different ages. Prog Neuropsychopharmacol Biol Psychiatry 22: 343-351.

16. Shih JC, Young H (1978) the alteration of serotonin binding sites in aged human brain. Life Sci  23:1441-1448.

17. Middlemiss DN (1986) Binding of the novel serotonin agonist 8-hydroxy-2-(di-n-propylamino) tetralin in normal and Alzheimer brain. J Neurochem 46:993-996.

18. Entsuah AR, Huang H, ME (2001) Thase, Response and remission rates in different subpopulations with major depressive disorder administered venlafaxine, selective serotonin reuptake inhibitors, or placebo. J Clin Psychiatry 62:869-77. 

19. Bourin M (2003) Use of paroxetine for the treatment of depression and anxiety disorders in the elderly: a review. Hum Psychopharmacol 18:185-190.

20. Yonkers KA, Brawman-Mintzer O (2002) The pharmacologic treatment of depression: is gender a critical factor? J Clin Psychiatry 63:610-615.

21. Martenyi F (2001) Gender differences in the efficacy of fluoxetine and maprotiline in depressed patients: a double-blind trial of antidepressants with serotonergic or norepinephrinergic reuptake inhibition profile. Eur Neuropsychopharmacol 11: 227-32.

22. Quitkin FM (2002) Are there differences between women's and men's antidepressant responses? Am J Psychiatry 159: 1848-1854.

23. Khan A (2002) Severity of depression and response to antidepressants and placebo: an analysis of the Food and Drug Administration database. J Clin Psychopharmacol  22: 40-45.

24. Schneider LS (2003) Cholinesterase inhibitors for vascular dementia? Lancet Neurol 2: 658-659.

25. Dailly E (2002) Evidence from a population pharmacokinetics analysis for a major effect of CYP1A2 activity on inter- and intraindividual variations of clozapine clearance. Prog Neuropsychopharmacol Biol Psychiatry 26: 699-703.

26. Tanaka E, Hisawa S (1999) Clinically significant pharmacokinetic drug interactions with psychoactive drugs: antidepressants and antipsychotics and the cytochrome P450 system. J Clin Pharm Ther 24: 7-16.

27. Kirchheiner J (2001) CYP2D6 and CYP2C19 genotype-based dose recommendations for antidepressants: a first step towards subpopulation-specific dosages. Acta Psychiatr Scand 104:173-92.

28. Hindmarch I (2001) the enantiomer debate: current status and future directions. Hum Psychopharmacol 16:S101-S104.

29. Baumannb P, Zullino DF, Eap CB (2002) Enantiomers' potential in  psychopharmacology--a critical analysis with special emphasis on the antidepressant escitalopram. Eur Neuropsychopharmacol 12:433-444.

30. Leonard BE (2001) An introduction to enantiomers in psychopharmacology. Hum Psychopharmacol, 16: S79-S84.

31. Girard P (1995) Influence of confounding factors on designs for dose-effect relationship estimates. Stat Med,. 14:987-1005. 

32. Shaikh S, Kerwin RW (2002) Receptor pharmacogenetics: relevance to CNS syndromes. Br J Clin Pharmacol, 54: 344-348.

33. Johnson JA, Lima JJ (2003) Drug receptor/effector polymorphisms and pharmacogenetics: current status and challenges. Pharmacogenetics 13:525-34.

34. Cacabelos R (2002) Pharmacogenomics for the treatment of dementia. Ann Med 34:357-379.

35. Scharfetter J (2001) Dopamine receptor polymorphisms and drug response in schizophrenia. Pharmacogenomics, 2001. 2:251-261.

36. Lotrich FE, Pollock BG, RE Ferrell (2001) Polymorphism of the serotonin transporter: 

37. Implications for the use of selective serotonin reuptake inhibitors. Am J Pharmacogenomics 1:153-64.

38. Smith D (2002) Efficacy and tolerability of venlafaxine compared with selective serotonin reuptake inhibitors and other antidepressants: a meta-analysis. Br J Psychiatry 180:396-404.

39. Baker CB (2003) Evidence that the SSRI dose response in treating major depression should be reassessed: a meta-analysis. Depress Anxiety 17:1-9.

40. Klein DF (2002) Improving clinical trials: American Society of Clinical Psychopharmacology recommendations. Arch Gen Psychiatry 59:272-278.

41. Petkova E (2000) A method to quantify rater bias in antidepressant trials. Neuropsychopharmacology, 22:559-565.