Quantifying Drug-Drug Interactions in Pharmaco-EEG

INTRODUCTION

A drug interaction refers to an event in which the usual therapeutic effect of a drug is modified by factors such as diet, environment, or additional drugs. When two drugs are administered simultaneously or within a short time of each other, an interaction can occur that may either increase or decrease the intended magnitude and/or duration of the effects of one or both drugs. The result may be either a harmful or a beneficial action on the patient. Psychopharmacology is one of the clinical areas where several drugs are most often prescribed concurrently. Drug interactions in the body can be categorized into several classes, depending on the mechanisms responsible for the interaction. Drugs may interact on a pharmaceutical, pharmacokinetic or pharmacodynamic basis.

Pharmaceutical interactions are physicochemical interactions that occur prior to absorption. Drugs may interact when they are mixed inappropriately in syringes and in infusion fluids before administration and become inactivated. For example, there are chemical incompatibilities in solution between aminophylline + chlorpromazine or between dopamine + sodium bicarbonate. In most cases, a visible change such as formation of a precipitate or change in color manifests these interactions. Occasionally, however, these interactions may occur without any observable signs, possibly resulting in undetected loss of potency. In the gut, drugs may chelate with metal ions or adsorb to medicinal resins. Thus, Ca²⁺ and other metallic cations contained in antacids are chelated by tetracycline, and the complex is not absorbed.

When a pharmacokinetic interaction occurs, the result may be an increase or decrease of the expected blood plasma levels, thus resulting in an increase or decrease of drug concentration at the biophase, the place where the receptors are located and the effect will start. This happens because one drug interferes with the absorption, distribution or elimination, either metabolism or excretion, of the other.

Although there are several mechanisms whereby the absorption of a drug may be altered by another drug (for example: (i) changes in pH of gastric fluid: pH elevation by antacids can decrease the absorption of weakly acidic drugs like aspirin or barbiturates; (ii) changes in intestinal motility and function: decreases in gastrointestinal motility by narcotics or anticholinergics can increase absorption), such effects are usually of little clinical importance. Many drugs are extensively bound to plasma albumin (acidic drugs) or α1-acid glycoprotein (basic drugs). In general, only an unbound drug is free to exert an effect or to be distributed to the tissues. Thus, a drug displaced from its binding site by another drug might be expected to result in a change in drug effects. Interactions involving drug metabolism can increase or decrease the amount of drug available for action by inhibition or induction of metabolism, respectively. Examples of drugs that inhibit the metabolism of others include inhibitors of several cytochrome P₄₅₀ isozymes (cimetidine, amiodarone, phenylbutazone, isoniazid, sodium valproate, and erythromycin), dopa decarboxylase (benserazide and carbidopa), and monoamine oxidase (MAO) inhibitors. Drugs that accelerate the metabolism of other agents include barbiturates, rifampicine, phenytoine, carbamazepine, chronic smoking, and certain chlorinated hydrocarbons. The ability of one drug to inhibit the renal excretion of another is dependent on an interaction at active transport sites. Drugs that alter the ability of the proximal renal tubule to reabsorb Na⁺ can affect the excretion of Li⁺. Thus, clearance of Li⁺ is reduced and concentrations of Li⁺ in plasma are increased by diuretics that cause volume depletion and by nonsteroidal anti-inflammatory drugs that enhance proximal tubular reabsorption of Na⁺.

Pharmacodynamic interactions arise when the effects undergo alteration at the site of action. Such interactions may originate at the receptor level, at the intraneuronal level, or at the interneuronal level within the Central Nervous System (CNS). At the receptor level, these interactions may occur either when the drugs possess different degrees of intrinsic activities or when the receptor sensitivity is altered. Different degrees of intrinsic activities may lead to antagonism, when one of the compounds has no intrinsic activity at all (opioids and naloxone) or to a partial agonism, when one of the compounds has less intrinsic activity than the so-called full agonist (benzodiazepines and bretazenil). Receptor sensitivity can decrease with prolonged stimulation and can increase with chronic blockade. Prolonged use of antipsychotic drugs, which can lead to hypersensitivity of central dopamine receptors, is thus thought to be responsible for the appearance of tardive dyskinesia. At the intraneuronal level, the interactions may occur either in steps after the drug-receptor interaction, that is, at the transduction process, or by uptake mechanisms, as for example when tricyclic antidepressants inhibit the active transport of some antihypertensive drugs (i.e., bethanidine, guanethidine, debrisoquine) into sympathetic nerve-endings resulting in a loss of blood pressure control. At the interneuronal level, drug interactions take place in different neurones which are physiologically connected within a network, as for example, the interaction observed when benzodiazepines and alcohol are taken together. Interactions can result in either agonism (additivity or synergism) or antagonism. Additivity occurs when the same level of effect (E*) is achieved with any combination of drugs (A and B) whose sum of percentages is 100%.

Antagonism occurs when E_A+E_B < additivity and synergism when E_A +E_B > additivity. The isobologram is a procedure to make such discrimination. It consists of the graphic representation of different combinations of concentrations of A and B producing the same effect E, allowing the location of the combination (A, B) producing the same effect E* (Figure 1).

OPEN IN VIEWER

Figure 1.

Isobologram as graphic representation of different combinations of concentrations of A and B producing the same effect E, which allows the classification of an interaction as: additivity, antagonism or synergism.

QUANTITATIVE APPROACH

Concerning the CNS, it is important to determine which effect is to be measured in order to assess an interaction. The effect measurement should have different characteristics, the most important being a) non-invasiveness, to make it ethically acceptable, b) continuity, to apply intensive quantitative analysis methods, c) sensitivity, to detect small changes in doses and concentrations, d) reliability, both intra-subject and inter-researcher, e) robustness, i.e., simple to assess and unsusceptible to extraneous influences, f) selectivity, in relation to the function evaluated, g) specificity, in relation to the mechanism of drug action, h) repeatability, to cover time-courses, and finally, i) validity.

The recording of the brain spontaneous electrical activity through the scalp, the so-called electroencephalogram (EEG), is a measurement that accomplishes most of the previously described characteristics. It has been argued that in order to classify a compound as brain active it should be able to induce statistically significant EEG changes in comparison with placebo. In fact, the EEG can be viewed as a multidimensional measurement. Depending on the drug administered, the pharmacodynamic effect can be observed through changes in different target variables derived from several mathematical methods applied to summarize the information conveyed. This information may disclose a non-unique final result following modification of the functional connectivity within widely distributed networks among brain regions. Even though these considerations are true as general statements, in practice the close fit with the previous measurement characteristics has been proved only in a few instances, such as for benzodiazepines, in relation to increases in fast activities, or for opioids, in relation to increases of slow activities. When this fit is achieved, relationships between drug concentrations and effects, such as that shown in Figure 2, can be observed. In these cases the relationship can be fully described through the following equation:

OPEN IN VIEWER

Figure 2.

Schematic representation relating drug concentrations to drug effects when it follows the so-called E_max model. The characteristic parameters of the model are E_max, EC₅₀ and γ.

Where E is effect, E_max is the maximum effect attributable to the drug, C is the concentration to which the effect is related, EC₅₀ is the concentration corresponding to 50% of maximum response and γ is the slope factor (sigmoidicity). Theoretically, EC₅₀ indicates the potency, E_max the intrinsic efficacy, and y indicates the number of binding sites for the drug at the receptor. This is what it is known as the sigmoid E_max model. Other equations that relate the concentrations with the effects have been described, providing the bases for different models to explain the pharmacodynamics, such as the hyperbolic E_max model or the linear model. This modelization is the basis through which a quantitative approach to interactions between drugs can be developed. ¹ On such occasions, interactions can be carefully and specifically assessed, enlarging the scope to be studied. Not only drug-drug interactions can be investigated, but also drug-metabolites, possible enantiomers with different pharmacodynamics, as well as phenomena like tolerance and sensitization. ²

The in vivo pharmacodynamic interaction between the benzodiazepine antagonist flumazenil and agonist midazolam was quantified in humans ³ and in rats ⁴ by a competitive interaction model, using the total number of waves and the total amplitudes in the β frequency band as derived by a periodic analysis as a measure of drug effect. Both in humans and rats a parallel shift in the concentration-EEG effect relationship of midazolam has been observed with increasing flumazenil concentrations, confirming the competitive nature of the interaction. Subsequently, this interaction model was generalized in experiments in rats including compounds with partial and inverse agonistic activity. ⁵

The possibility that metabolites contribute to the effects of parent compounds should be considered. Midazolam is rapidly eliminated from the body by metabolism to two substances, α-hydroxy-midazolam and 4-hydroxy-midazolam, both pharmacologically active. Both metabolites are rapidly conjugated by glucuronic acid to form an inactive product. After intravenous administration of midazolam, relatively low concentrations of the metabolites are found. However, relatively high concentrations of α-hydroxymidazolam are observed after oral administration, so the interaction between midazolam and α-hydroxy-midazolam can reach significant relevance. The concentration-effect relationships of both compounds intravenously administered could be quantified very well in individual subjects by measuring EEG effects (amplitudes of the 11.5 to 30 Hz frequency band), yielding approximately the same pharmacodynamic parameters. However, the effects observed after oral administration were substantially larger than those predicted by an additive and competitive interaction between midazolam and its α-hydroxy-metabolite. ⁶ In most subjects the observed effects were considerably larger at each time point than the predicted effects, suggesting a synergistic interaction between the two compounds. However, other factors, such as formation of other metabolites or an interaction in plasma protein binding, may also have contributed to the discrepancy between predicted and observed effects after oral administration.

Another interesting field related to interactions deals with racemates, as enantiomers can differ both in their pharmacokinetic and pharmacodynamic properties. Schüttler et al ⁷ evaluated ketamine behavior after intravenous administration in humans using the change in the median frequency of the EEG as the measured effect of the drug on the CNS. A sigmoid E_max model could describe the concentration-EEG effect relationship of the racemate and the two enantiomers. No significant differences were seen in either enantiomer in EC₅₀, but the maximal effect (E_max) of the R-enantiomer was significantly smaller than the value found for the S-enantiomer, suggesting a partial agonist activity of the former. However, the EEG effect of the racemate could not be fully explained by an additive and competitive interaction between the two enantiomers, but suggests a synergistic interaction.

Tolerance and sensitization as time-dependent pharmacodynamic changes, which are consequences of previous drug intake, can also be viewed as special kinds of drug interaction phenomena. In a randomized, double-blind, cross-over (wash-out period: 1 month), placebo-controlled design, 16 healthy young volunteers received three active treatments: single oral doses of diazepam 30mg (two periods) and nordiazepam 30mg (one period). Blood samples and 3-minute eyes-closed EEG were obtained before and 18 times after drug intake (last: +48 hours). ⁸ EEG effects (16-lead means of relative beta power [12–30Hz] from 5 second free-of-artifact epochs) after single high doses of diazepam followed a sigmoid relationship, but E_max clearly increased the 2nd time subjects received the drug with a non-overlapping 95% confidence interval (Figure 3) showing a sensitization phenomenon. In addition, EEG effects after single high doses of nordiazepam also followed a sigmoid relationship. This time, however, we observed acute tolerance in the only period in which volunteers received the drug, meaning that higher drug concentrations were needed to achieve the same level of effect in the decreasing limb as in the increasing limb of the relationship (Figure 4).

OPEN IN VIEWER

Figure 3.

Real and modelled data of the relation between free diazepam plasma concentrations and % beta activity in humans (E_max model), comparing the first with the second period the subjects received the drug (at least 1 month apart). A sensitization phenomenon is observed: significant increase of E_max.

OPEN IN VIEWER

Figure 4.

Real and modelled data of the relation between free nordiazepam plasma concentrations and % beta activity in humans (E_max model). An acute tolerance phenomenon is observed: to achieve the same effect, much higher concentrations are needed at the decreasing than at the increasing limb.

QUALITATIVE APPROACH

Up to this point, we have seen some examples of what can be observed using the EEG within the framework of drug interactions. As previously stated, this approximation can be applied when it has been possible to clearly identify a specific target variable derived from the EEG, which undoubtedly meets the characteristics proposed as ideal for an effect measurement. However, this is not the case for the majority of compounds which are active in the brain. Alternatively, QEEG assessment of psychotropic drug interactions can be viewed according to the broad underlying objectives of the study, which may be safety-oriented, strictly pharmacologically-oriented or broadly neurophysiologically-oriented.

When safety data are the major argument of an interaction study, results are mainly related to vigilance assessments. Different possibilities in combining the pharmacodynamic effects have been evaluated using this approach. One example would be to assess the interaction when two compounds have a clear, defined profile. In one study, ketoprofen, an asthma prophylactic compound with sedative properties, was administered together with theophylline, a bronchodilator with alerting properties, anticipating an antagonist effect. ⁹ In another study, lorazepam, a benzodiazepine with a sedative action, was administered together with diphenhydramine, an old histamine H₁ receptor antagonist also with sedative properties and used in allergic diseases, this time pursuing a synergistic effect. ¹⁰ Yet another example would be when the individual profiles of the compounds are not well established and the result of the combination is more difficult to hypothesize in advance, as was observed between granisetron, a 5-hydroxytryptamine 5-HT₃ receptor antagonist used in the treatment of emesis associated with cancer chemotherapy, and haloperidol, a classical neuroleptic compound of the butyrophenone group. ¹¹

Strictly pharmacological trials are focused on eliciting the molecular drug mechanism. For example, if the effects of a compound (or a portion of such effects) are mediated by a certain kind of receptor type, the administration of a known antagonist of such receptors would decrease the EEG effects of the former. Ethanol aggravates benzodiazepine-induced central nervous depression, and flumazenil, a selective competitive antagonist at the benzodiazepine recognition site, promptly reverses the hypnotic effects of benzodiazepines. Flumazenil could therefore also influence ethanol-induced brain effects. The acute effects of an intravenous bolus of 0.5 mg flumazenil on steady state alcohol blood levels (0.9 to 1.2 g/l) effects were studied in six healthy male subjects. ¹² There was no pharmacokinetic interaction between the two agents since their elimination parameters (half-life time: t_1/2) were in good agreement with control values. Furthermore, the marked sedative effects of ethanol as assessed by visual analogue scales and choice reaction time impairment were not affected by flumazenil. However, the pharmaco-EEG indicated that flumazenil seems to transiently reverse the ethanol-induced increases in theta- and slow-alpha-bands and decreases in delta-fast-alpha- and beta-bands, specifically the induced increase in the slow-alpha band. EEG data suggest that EEG might be a more sensitive indicator than both psychometric tests and subjective self-reports. The short duration of the antagonistic effect could be explained by the low dose and the fast elimination of flumazenil. A higher dose might therefore have more pronounced effects, which in turn could be evident through other tests. Thus it could be concluded that in the well-known and clinically relevant interaction between ethanol and benzodiazepines a direct event at the benzodiazepine-GABA-ionophore receptor complex could not be fully excluded.

Neurophysiological grounds can also be examined. This is based on the assumption that behavioral changes can be directly related to the biochemical changes produced by drugs in the brain and that EEG changes can directly be related to these biochemical changes. In a double-blind randomised cross-over placebo-controlled study twelve healthy young volunteers of both sexes received single oral doses of diazepam 10 mg, buspirone 20 mg and diazepam 10 mg + buspirone 20 mg, and EEG was assessed at pre and +1, +2, +4, +6 hours after drug administration. ¹³ After diazepam, brain maps of drug-induced pharmaco-EEG changes showed a decrease in total power, attenuation of alpha activity and augmentation of beta activity, as well as an increase in the centroid and centroid deviation of the total activity, a decrease in the centroid of the combined delta-theta activity, and an increase of the centroid of alpha activity. In contrast, buspirone produced an increase in total power, augmentation of theta and alpha-1 activity, attenuation of alpha-2 and beta activities as well as an increase in the centroid of combined delta and theta activity, and a decrease in the centroid of alpha activity and centroid deviation of total activity. There was a clear interaction with the combination of diazepam + buspirone in such a way that diazepam effects were diminished (Figure 5: part A). Neurobiological data support these findings. Buspirone and diazepam have differential effects on central monoaminergic activity. In electrophysiological studies in rats, both drugs have been found to produce inhibition of serotonergic neural activity in the dorsal raphe nucleus. Although buspirone and benzodiazepines exert similar effects on serotonin neurones, differences are observed in their actions on dopaminergic and noradrenergic neurones. Whereas benzodiazepines inhibit the spontaneous firing of noradrenergic cells in the locus coeruleus, buspirone produces increases in their activity. The firing rate of midbrain dopaminergic neurones is dramatically increased after administration of buspirone. In contrast, benzodiazepines decrease dopaminergic neural activity. The decrease in serotonergic activity produced by these agents may account for anxiolysis, whereas their inhibition of catecholaminergic influences may produce sedation. These nonselective alterations in central monoaminergic activity could explain benzodiazepine side effects as well as its efficacy. On the other hand, the differential effects of buspirone on serotonergic and catecholaminergic neurones may account for its ability to relieve anxiety without causing sedation, inducing a different pharmaco-EEG profile from that observed after benzodiazepine intake.

OPEN IN VIEWER

Figure 5:

Statistical Probability Maps of topographic pharmaco-EEG changes in: A) absolute power of slow alpha (7.5–10.5 Hz) activity after diazepam 10 mg, buspirone 20 mg and the combination of both compared to placebo after the 1st hour. B) relative power variables after ambulatory doses of morphine (oral administration of 30 mg modified formulation) compared to placebo after the 9th hour. C) centroid of alpha activity after buspirone 20 mg and the combination of buspirone 20 mg + fluoxetine 60 mg compared to placebo after the 1st, 2nd and 4th hours. D) relative power of the beta (13–30 Hz) activity after fluoxetine 60 mg and the combination of buspirone 20 mg + fluoxetine 60 mg compared to placebo after the 6th, 8th and 10th hours. E) absolute power of beta (13–30 Hz) activity after yohimbine 30 mg, caffeine 0.8 mg/Kg and the combination of both (two periods) compared to placebo after the 1.5st hour. The eight-color scale represents drug-induced changes based on t-values expressed in p-values. Increases: yellow, orange (p< 0.10), red (p< 0.05), dark violet (p< 0.01). Decreases: mild green, dark green (p< 0.10), mild blue (p< 0.05), dark blue (p< 0.01). A) A full topographic interaction is observed. B) In contrast to anesthetic doses a decrease in slow activities is shown. In addition, increases in alpha-1 and decreases in alpha-2 and fast activities are observed. C) A characteristic change after buspirone (slow-down of the centroid of alpha activity) is increased in magnitude and duration after the combination. D) A characteristic change after fluoxetine (increase in fast activities) is similar after the drug alone or in combintation with buspirone. E) A partial topographic interaction is observed.

COMBINING THE QUANTITATIVE AND THE QUALITATIVE APPROACH

To our knowledge, no previous attempts have been made to apply both the quantitative approach (modelling the relationship between drug plasma concentrations and the pharmacodynamic effects) and the qualitative approach (describing drug effects by means of topographic brain maps of significant probability changes) to the same set of data. This is especially true if we focus on human data and drug-drug interaction trials. Our group has begun to explore the application of both approaches to the same results. Data came from a randomised, double-blind, double-dummy, placebo-controlled, repeated-dose, 4-period crossover study. Twenty-two healthy young subjects of both sexes received paroxetine (active paroxetine 20mg plus placebo alprazolam), alprazolam (placebo paroxetine plus active alprazolam 1mg), the combination of both (active paroxetine 20mg plus active alprazolam 1mg), and placebo (placebo paroxetine plus placebo alprazolam). Drugs were administered early morning, once daily for 15 consecutive days. At day 1 and day 15 topographic EEG (16 leads) was recorded pre and +1, +2, +4, +6 and +8 hours. Results obtained through psychomotor performance tests and all night polysomnographic sleep recordings have been separately published ( ^14,15 respectively). Only EEG data of the acute effect (day 1) have been analysed so far. ¹⁶ Main conclusions were that all active treatments induced significant EEG effects in comparison to placebo, and the EEG profile after the combination mostly resembles the EEG profile after alprazolam. In order to apply the quantitative approach, two target variables were selected: beta and alpha activities. Both fulfill the effect measurement characteristics described in order to be used in such an approach and, in addition, they have already been previously successfully modelled after benzodiazepine intake in humans. ^3,6,8,17 Analysis has been focused on the comparison between acute alprazolam vs. acute alprazolam plus paroxetine effects. No interaction was evidenced at the pharmacokinetic level, meaning that the alprazolam plasma concentration time course was similar when alprazolam was taken alone or when it was taken together with paroxetine.

One issue as yet unsolved, like several others which are possibly not yet recognized as such, is which lead to use in the quantitative approach. All earlier studies dealing with the quantitative approach have only used data coming from a single lead. The lead differs depending on the author, without any explanation to justify the choice. When applying the quantitative approach to data obtained in a topographic setting we have used the average of all leads as a representative value. ⁸ This procedure was that used to analyze beta activity. A linear model was successfully fitted to the data when the quantitative approach was applied. When comparing the model's parameters after alprazolam and the combination intake, no differences were obtained (Figure 6). However, when comparing the beta changes after alprazolam with the changes after the combination within the qualitative approach, significant differences were observed (Figure 7: part A). The differences were such that a higher increase in beta activity at the anterior part of the scalp was recorded after the paroxetine plus alprazolam intake in comparison to alprazolam alone. This is contrary to what occurred at the posterior part of the scalp where the increases were lower after the combination. These qualitative results are in line with what would be expected taking into account the increase in relative beta activity at the anterior part of the scalp observed after antidepressants belonging to the selective serotonin re-uptake inhibitor group, particularly at the fastest beta subbands. ¹⁸ Nevertheless, if the mean of the leads is calculated no changes were assessed. Thus, the quantitative approach was again applied, but this time using the averages of both anterior and posterior leads as representative values. Now the comparison of the model's parameters after alprazolam and the combination intake using the average of the anterior leads showed a significant steeper slope after alprazolam plus paroxetine in relation to alprazolam alone [Baseline: 16.1; Slope for alprazolam: 0.70, Slope for alprazolam when paroxetine is present: 0.92 (p< 0.0001)] (Figure 6), displaying the effects observed when the qualitative approach was used.

OPEN IN VIEWER

Figure 6.

Individual measured effect vs. individual predicted alprazolam plasma concentrations, obtained after the administration of alprazolam 1mg (empty circles) and alprazolam 1 mg plus paroxetine 20 mg (full circles) (n= 22, times= 6). Lines respresent population typical prediction for this relationship after the administration of alprazolam alone (dashed line) and the combination (solid line). Upper part: measured effect is the mean of all 16 electrodes of relative power of beta (13–30 Hz) activity. Lower part: measured effect is the mean of the 10 anterior electrodes of relative power of beta (13–30 Hz) activity. No differences in model's parameters after alprazolam and the combination intakes are obtained when the mean of all 16 electrodes is used. When only the mean of the 10 anterior electrodes is used, a significant steeper slope after alprazolam plus paroxetine in relation to alprazolam alone is observed.

OPEN IN VIEWER

Figure 7.

Statistical Probability Maps of topographic pharmaco-EEG changes after paroxetine 20mg (PAR), alprazolam 1mg (ALP) and paroxetine 20mg plus alprazolam 1mg (INT) compared either to placebo (- PLA) or alprazolam 1mg (- ALP) after the 1st, 2nd, 4th, 6th and 8th hours in: A) relative power of beta (13–30 HZ) activity and B) relative power of alpha (7.5–13 Hz) activity (upper part) and in the dominant frequency (lower part). The six-color scale represents drug-induced changes based on t-values expressed in p-values larger than 1.72 or smaller than −1.72: p< 0.10, larger than 2.08 or smaller than −2.08: p< 0.05, and larger than 2.83 or smaller than −2.83: p< 0.01. A) When compared to placebo increases of fast activities are observed after paroxetine (anterior part), alprazolam and the combination of paroxetine plus alprazolam. When compared to alprazolam, the combination shows higher increases in beta activity at the anterior part of the scalp while at the posterior part of the scalp its increases are lower. B) When compared to placebo, increases of alpha activity are observed after paroxetine (first time periods) while decreases are obtained after alprazolam and the combination. When compared to alprazolam, the combination does not show any significant effect. However, when looking at the dominant frequency an increase is observed after the combination in relation to alprazolam alone (first time periods).

Regarding alpha activity, a linear model was also successfully fitted to the data when the quantitative approach was applied. When comparing the model's parameters after alprazolam and the combination intake, a tendency to a steeper slope was observed after alprazolam plus paroxetine in relation to alprazolam alone [Baseline: 51.8; Slope for alprazolam: −0.57, Slope for alprazolam when paroxetine is present: −0.73 (p= 0.05)] (Figure 8), indicating a tendency to a higher reduction of vigilance when the combination was administered. This is in agreement with the results from the psychomotor performance tests where the combination showed a trend to develop a slower rate of tolerance to the sedative effects compared to alprazolam alone. ¹⁴ However, it is widely accepted that EEG measures have a greater sensitivity in comparison to psychomotor measures when assessing changes in vigilance, specially when they move towards alertness. ¹⁹ Thus, even if no changes are observed at an objective behavioral level, an increase in alpha activity is seen after paroxetine.

OPEN IN VIEWER

Figure 8.

Individual measured effect vs. individual predicted alprazolam plasma concentrations, obtained after the administration of alprazolam 1mg (empty circles) and alprazolam 1 mg plus paroxetine 20 mg (full circles) (n= 22, times= 6). Lines respresent population typical prediction for this relationship after the administration of alprazolam alone (dashed line) and the combination (solid line). Measured effect is the mean of all 16 electrodes of relative power of alpha (7.5–10.5 Hz) activity. A tendency to a steeper slope is observed after alprazolam plus paroxetine in relation to alprazolam alone.

Consequently, an antagonistic effect at the neurophysiological level could be advanced. No such effect was obtained when applying the quantitative approach. Neither were differences observed when comparing the alpha changes after alprazolam with the changes after the combination within the qualitative approach (Figure 7: part B). However, when reviewing the EEG target variables associated with the vigilance level in a qualitative approach, there was an increase in the dominant alpha frequency after the combination in relation to alprazolam alone (Figure 7: part B). This evidences that the alerting effects of paroxetine contributed in some way when alprazolam and paroxetine were taken together. The dominant alpha frequency does not fulfill the measurement characteristics to be used in a quantitative approach. The dominant alpha frequency change range is too narrow and the continuity of its values is questionable as they depend on the epoch length used when calculating the power spectra.

Having read the anterior quantitative and qualitative approach sections but before coming to the present section, one would probably hypothesize that if the quantitative approach is applied, the qualitative approach is useless, in other words, that the qualitative approach adds no further relevant information. As the state-of-the-art stands at the present time, however, the two examples presented show the importance of combining both the quantitative and the qualitative approaches to avoid misinterpreting data from multi-lead EEG recordings in drug-drug interaction trials in humans. Once again, this illustrates the importance of not restricting the number of evaluable EEG-target variables if an accurate appraisal of the results is to be achieved. ²⁰

METHODOLOGICAL ISSUES

Several methodological topics should be stressed when dealing with drug interaction EEG studies. Dose-effect relationships cannot be the same throughout the continuum range of concentrations. Drug plasma concentrations and the time-course of the effects should be integrated. These considerations should be taken into account in order to have an accurate explanation of what is going on.

The opioid effects can be selected as an example of a non-monotonic drug effect on the EEG. While anesthetic doses produce well quantifiable modelable increases in voltage and slow-down of frequencies either using total power, power in the delta band, median frequency or spectral edge (SEg_95%), ²¹ ambulatory doses are characterized by a decrease in slow activities, an increase in alpha-1 activity and decreases in alpha-2 and in slow-beta activities ²² (Figure 5: part B).

The relevance of integrating drug plasma concentrations and the time course of the effects is evident when interpreting the EEG results obtained after co-administration of buspirone and fluoxetine. In a double-blind randomised cross-over placebo-controlled trial, eighteen healthy young subjects of both sexes received single oral doses of buspirone 20 mg, fluoxetine 60 mg and buspirone 20 mg + fluoxetine 60 mg. EEG was assessed at pre and +1, +2, +4, +6, +8, +10 hours after drug administration. Plasma levels of buspirone, an active metabolite of buspirone: Pmp, and fluoxetine were also evaluated by means of HPLC methods. ²³ In comparison to placebo, results showed a highly significant differentiation with buspirone in the 1st hour, with fluoxetine in the 8th hour (mostly in anterior regions), and with buspirone + fluoxetine from the 1st until the 4th hour, re-appearing in the 8th hour (Figure 9: part A). Maps of drug-induced pharmaco-EEG changes as compared to placebo-induced alterations demonstrated buspirone changes in the first hours together with fluoxetine changes in the late hours after the combination (Figure 5: parts C and D). Plasma levels of buspirone were 3 times higher after the combination than after buspirone alone; Pmp plasma levels were similar in both conditions. In addition, plasma levels of fluoxetine were 2 times higher after the combination than after the drug alone. Taken together pharmacokinetic and pharmacodynamic data suggest that the interaction observed in the sense of an increase of buspirone-like effects could be mediated by a pharmacokinetic mechanism. But in addition, fluoxetine pharmacokinetics clearly pointed to a decrease in pharmacodynamic effects when the drug is ingested concomitantly with buspirone, as there was no difference in the magnitude of effects in relation to the single ingestion but its plasma level concentrations were significantly augmented. There are several explanations: from a possible lack of dose-effect relationship between drug plasma levels and EEG effects, the unforgettable possible role of fluoxetine active metabolite concentrations - which were not measured - or, even a possible true pharmacodynamic interaction. In fact, fluoxetine is a selective serotonine (5-HT) reuptake inhibitor, and consequently it increases bioavailability of the neurotransmitter at the synapses. Buspirone, on the other hand, is a partial 5-HT_1a agonist, meaning that in presence of full agonists (such as 5-HT), the net overall results consist of pharmacodynamic antagonism.

OPEN IN VIEWER

Figure 9.

Statistical Probability Maps showing differences: A) between drug-induced and placebo-induced central effects after buspirone 20 mg, fluoxetine 60 mg and the combination of both at 1, 2, 4, 6, 8, 10 hours based on the absolute power of five (delta, theta, alpha-1, alpha-2, beta) frequency bands. B) Between drug-induced and placebo-induced central effects after alcohol 0.8 g/Kg, hydroxyzine 25 mg + alcohol, cetirizine 10 mg + alcohol at 1, 2, 4, 6, 8, 10 hours based on the relative power of five (delta, theta, alpha-1, alpha-2, beta) frequency bands. Images are based on Hotelling's T ² obtained from multivariate tests in repeated measures ANOVA. Significant T ² larger than 2.347: p< 0.10, larger than 3.025: p< 0.05 and larger than 4.862: p< 0.01. A) Buspirone is significantly different from placebo in the 1st hour, fluoxetine in the 8th, and the combination from the 1st until the 4th, reappearing in the 8th hour. B) When H₁-antihistamines are taken together with alcohol, changes are of lower magnitude and less maintained in time than those observed after alcohol alone, being hydroxizine + alcohol the combination that shows the fewest significant changes.

UNCERTAINTIES YET TO SOLVE

The distribution of topographic pharmaco-EEG changes varies depending on the compound administered. If two compounds with opposite pharmaco-EEG effects are administered simultaneously, the pattern of EEG changes may differ depending on each drug's individual topography. In a previously described study, ¹³ diazepam 10 mg decreased slow alpha activity while buspirone 20 mg increased it, both drugs all over the brain. When both drugs were taken simultaneously changes tended to be counteracted all over the brain. Different patterns of EEG changes were observed in a double-blind randomised five-period cross-over placebo-controlled study, in which twelve healthy young volunteers of both sexes received single oral administrations of yohimbine 30 mg (one period), caffeine 8 mg/kg (one period) and the combination of both compounds (two periods). An EEG was recorded prior to administration and at +1.5, +3, +6 hours after drug administration. ²⁴ Yohimbine 30 mg increased fast activity while caffeine 8 mg/kg induced a decrease, though the former effect was confined to the posterior region and the latter to the anterior region. In this study, when the drugs were taken simultaneously these changes persisted in both regions, resulting in significant opposite changes within the same brain map (Figure 5: part E). Although methodological questions may arise, such as spatial aliasing-related issues, these data emphasize the relevance of topographic analysis for accurate assessment of drug effects on human EEGs. However, several theoretical questions are yet to be answered, such as the relationship between topographic differences in drug-induced EEG changes with the selectivity in drug action, or the specificity of the EEG variable evaluated.

Finally, it is interesting to present the results from a drug interaction study which question a basic assumption in neuropsychopharmacology: if a compound is CNS active, the changes in the EEG differ significantly from those observed after placebo. In a phase I randomised crossover double-blind placebo-controlled study, eighteen healthy young volunteers of both sexes received—at 2-week intervals—alcohol 0.8 g/Kg alone, and alcohol 0.8 g/Kg in combination with hydroxyzine 25 mg or cetirizine 10 mg. Topographic pharmaco-EEG mapping together with psychomotor performance tests were performed before and after +1, +2, +4, +6, +8, +10 hours. ²⁵ Psychomotor impairment observed after alcohol was increased when alcohol was taken together with both H₁-antihistamines, and the hydroxyzine + alcohol combination showed the greatest alteration (Figure 10). However, different results were obtained by EEG. Changes with alcohol were maximally different from placebo at +1 h, decreasing thereafter but still significant at +4 h. When H₁-antihistamines were taken together with alcohol, changes were of lower magnitude and less maintained over time (only significant until +2 h), hydroxyzine + alcohol was the combination with fewest significant changes (Figure 9: part B). Maps of drug-induced pharmaco-EEG changes as compared to placebo-induced changes demonstrated main alcohol effects in the alpha band (increase alpha-1, decrease alpha-2, slowing-down alpha centroid). These effects are opposite to the changes previously described as the profile of “low frequency sedation H₁-antihistamines” (decrease alpha activity, mainly alpha-1, with an acceleration of alpha centroid). No interaction was observed in pharmacokinetics. Thus, in view of the opposite pharmaco-EEG profile of alcohol and H₁-antihistamines, few EEG changes are observed when these two compounds are taken simultaneously. This clearly contrasts with the appearance of marked behavioral consequences, questioning whether or not CNS activity is always associated with significant pharmaco-EEG changes. However, one must keep in mind that the EEG is a multidimensional signal. In fact, the spectral analysis used in this study - and in most studies dealing with drug effects on EEG - is only one of many analytical techniques available to extract the information conveyed by the signal. ²⁶ Furthermore, linear and non-linear approaches have provided new results when applied to other EEG fields, such as sleep-EEG. ^27,28 Therefore, the application of these “non-traditional” quantification procedures to drug-drug interactions in pharmaco-EEG, will probably allow a deeper and wider understanding of drug-brain interactions in humans.

OPEN IN VIEWER

Figure 10.

Time course of mean values in a psychomotor performance test (“d2” cancellation) expressed as differences from basal values after alcohol 0.8 g/Kg (▴, A), hydroxyzine 25 mg + alcohol (▪, H), cetirizine 10 mg + alcohol (♦, C), and placebo (•, P). Statistical results after ANOVA: D= treatment factor, T= time factor, DxT= interaction factor. **p< 0.01, *p< 0.05. Psychomotor impairment observed after alcohol is increased when alcohol is taken together with both H₁-antihistamines, being hydroxyzine + alcohol the combination that shows the greatest alteration.