Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

BZs are safe and widely prescribed for the chronic treatment of epilepsy, movement and panic disorders. They are given for the acute and subchronic treatment of insomnia, agitated psychosis and in surgery as effective preanesthetic and anesthetic agents (Martin and Haefely, 1995). However, interactions can occur with combinations of BZs and central nervous system stimulants (e.g., caffeine, cocaine), where one agent is prescribed and the other is ingested by choice (Boulenger et al., 1984; Charney et al., 1985). BZs exert their effects through the GABA-BZ receptor complex (Haefely et al., 1985). It is generally recognized that the antagonism of adenosine receptors at least partly underlies the pharmacological effects of low doses of MXs, whereas phosphodiesterase inhibition and calcium mobilization become more significant at higher doses (Choi et al., 1988; Daly, 1993; Snyder et al., 1981). Several studies have indicated that caffeine competes for binding at BZ sites, and conversely, that BZ may interact with adenosine receptors, although with low affinity in both cases, raising the question of the physiological relevance for these kinds of interactions (Bruns et al., 1983; Marangos et al., 1979; Weir and Hruska, 1983). Thus, the interactions between caffeine and the GABA-BZ system remain poorly defined.

Additive (Beer et al., 1972; Coffin and Spealman, 1985; Valentine and Spealman, 1983), antagonistic (Kaplan et al., 1990; Polc et al., 1981; Rush et al., 1994; Tang et al., 1989), functional antagonistic (Baldwin and File, 1989; Roache and Griffiths, 1987) and synergistic (Falk and Lau, 1991; Katims et al., 1983; Lau and Falk, 1991) interactions have been reported after concurrent BZ and MX administration using various kinds of behavioral paradigms in animals and humans. One of the major reasons for these differences arose from describing the combined effects qualitatively rather than from using quantitative methods specifically developed for that purpose.

Inasmuch as the pharmacological response often can be predicted from the respective PK, it is rational to investigate the role of PK on drug action and interaction before receptor mechanisms. However, the predictability of PK during drug interaction is not as simple and direct as it is when a drug is given alone. One needs not only to analyze the resultant PK changes of the parent drugs and their active metabolites after concurrent drug administration but also consider the potency relation of the two agents before inferring the role of PK in drug interaction. Failure to explore the potency relation may account, in part, for the conflicting findings relating PK to PD in BZ-MX combined effects (Ghoneim et al., 1986; Henauer et al., 1983; Kaplan et al., 1990; Tuncok et al., 1994).

Differential reinforcement of low rate schedules (e.g., DRL 45-s) produce low rates of responding as only those responses that occur after a minimum time interval (45 sec) after a previous response are reinforced. Responses that occur before this time has elapsed are not reinforced, and they reset the timing of the interval. DRL behavior reaches baseline performance after sufficient training, and the effects of drug treatments can be compared to the performance baseline. The DRL schedule contingency not only involves time discrimination but also requires an appropriate inhibition of responding for reinforcement to occur, and involves other memory, sensory and motor capacities (Kramer and Riling, 1970). It has been suggested that the effect of many kinds of drug is to reduce the inhibition of behavior associated with signals of punishment or nonreward in DRL behavior (Gray, 1981). DRL performance satisfies many of the criteria proposed as ideal for PD measurement (Dingemanse et al., 1988; Laurijssens and Greenblatt, 1996). The fulfilling of a required, objectively defined, behavioral contingency by the subject, rather than using a passive measure of an unconditioned drug effect (e.g., EEG recording), affords the DRL method a distinct advantage. The performance measure is a continuous process rather than one limited to temporally discrete trials. Furthermore, it is sensitive to drug effects, and the effects are reproducible, an important feature for defining and evaluating drug interaction (Lau et al., 1996). Finally, after drug administration, reinforced and nonreinforced responses, which generally exhibit decreases and increases, respectively, can be used to evaluate the combined drug effects.

Recently, we used the DRC method proposed by Pöch and his colleagues (Pöch, 1993, 1992; Pöch and Pancheva, 1995; Pöch et al., 1990) to quantitatively analyze the combined effects of alprazolam and caffeine by using 3-hr sessions of DRL 45-sec performance (Lau and Wang, 1996). This method permits the evaluation of the combined effects not only from a phenomenologic (e.g., larger or smaller effect) but also from a mechanistic (additivity or independence) point of view. The assumptions used in the DRC method also can be applied to behavior-time profiles to extend the results obtained from DRC analyses. Values derived from the usual dose-response analyses (e.g., potency ratio of these agents) can aid in predicting the outcome of the combined effect, whereas the response-time curve describes the ongoing interaction profile.

An independent or additive interaction, which was neither synergistic nor antagonistic, characterized the combined effects of alprazolam and caffeine by the i.p. route using reinforcement rate as the PD measure (Lau and Wang, 1996). In that study, PK interaction was also characterized by using tail-tip blood samples between 15 and 180 min. It was concluded that the PK of alprazolam, caffeine and their combination were predictive of the resultant behavior-time profiles. The differences in potency and PK between the two drugs accounted for the expressions of the combined effects. The PK of alprazolam was not altered by the presence of caffeine, but the PK of caffeine was affected by alprazolam. Inasmuch as the PK drug interaction was not evaluated by the i.v. route in that study, the effects of alprazolam on caffeine PK were difficult to interpret.

Different types of interaction for BZ-MX sometimes were obtained from the same laboratory with the use of different behavioral measures or paradigms (De Angelis et al., 1982; Ghoneim et al., 1986; Loke et al., 1985). Our study is an expansion of the previous work on both behavior and PK, which aims to validate the interaction of alprazolam and caffeine by using: 1) different routes of administration; 2) not one but two different kinds of response measures, reinforced and nonreinforced, to investigate whether they were in conformity with each other; 3) blood samples from jugular vein between 2 and 180 min after drug administration to characterize the respective PK; 4) the i.v. route to calculate the PK parameters (e.g., volume of distribution, clearance, and bioavailability) and to define the pure PK drug interaction without having to consider drug absorption and 5) integration of PK and PD to delineate the nature of BZ-MX interaction and the predictive ability of the model.

Both alprazolam and caffeine are metabolized by the P-450 cytochrome enzyme system (Aldridge et al., 1977; von Moltke et al., 1993). Factors (e.g., food restriction) affecting this enzyme system will affect the PK of these agents and will lead to PD changes (Lau et al., 1995; Lau et al., 1996; Sachan, 1982). Both alprazolam and caffeine are absorbed rapidly in rats with an elimination half-life of 0.5 to 0.9 and 3 hr, respectively (Lau and Wang, 1996; Lau et al., 1995; Owens et al., 1991). In humans, food deprivation or restriction can occur for cosmetic, health or economic reasons. In DRL behavior, a food-deprivation regimen is applied to animals to implement a food-reinforced behavioral DRL performance baseline. Thus, it is important to investigate the PK of alprazolam, caffeine and their combinations in food-limited rats, especially because these drugs are metabolized by the P-450 cytochrome enzyme system. Furthermore, based on their half-lives, a 3-hr session was used, a period necessary to investigate the interaction at the onset, peak and disappearance of serum alprazolam concentration, while that of caffeine remained constant, so that we could achieve a better understanding of the mechanisms of drug action and interaction.

Different routes of administration provided an opportunity to examine the interaction of drug concentration-time profiles that might differ from that of the i.p. route, as PK parameters are generally route dependent (e.g., absorption rate constant, metabolite formation and bioavailability). In our study, alprazolam and caffeine were given s.c. and p.o., respectively. There are considerations for choosing these routes of administration. As in the case of midazolam (Lau et al., 1996), we found the s.c. route to be the route of choice owing to its high absolute bioavailability, as well as its dependability in producing consistent within-subject serum concentration-time profiles for repeated doses, whereas for caffeine, the oral route is used by humans for its consumption with high bioavailability (Axelrod and Reichental, 1953).

Although DRL performance has been used extensively in behavioral pharmacology to study the effects of various drugs from different classes, it has not been used for PK-PD studies, except in our laboratory. We have found not only that the DRL 45-sec reinforcement rate-time profiles correlated well with serum alprazolam, caffeine and midazolam concentration-time profiles but also that bioavailability values derived from those profiles mirrored those estimated from PK for midazolam following i.v., s.c., i.p. and p.o. routes of administration (Lau and Wang, 1996; Lau et al., 1996). Integrating PK and PD permits the investigation and possible prediction of drug concentration-effect relations, which are sensitive to variables such as drug interaction, aging and the disease state. Examination of the alprazolam concentration-effect relation in the presence and absence of caffeine can shed light on the nature of the interaction. The competitive interaction between flumazenil and midazolam was demonstrated in humans (Breimer et al. 1991) and in rats (Mandema et al. 1991) with PK-PD modeling by parallel shifts in the concentration-electroencephalography effect relation of midazolam with increasing flumazenil concentration.

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

DRL Performance

Animals. Seven male, albino, Sprague-Dawley rats from HSD (Indianapolis, IN) were used. They were housed individually in a temperature-regulated room with a daily cycle of illumination from 7:00 A.M. to 7:00 P.M. They were reduced to 80% of their initial, adult free-feeding body weights (mean = 383 g; range: 380-388 g) over a 2-wk period by limiting daily food rations: 5 g for the first day, 10 g for the next 5 days and a food supplement (range 14-16 g) to maintain their 80% body weights. Water was continuously available in the living cages. Experiments were executed in accordance with the Guide for the Care and Use of Laboratory Animals (National Institute of Health Publ. no. 85-23, revised 1985).

Drugs. Alprazolam was obtained from Upjohn Laboratories (Kalamazoo, MI). Alprazolam (5 mg) was dissolved in 50 µl of 1.2 N HCl and further diluted to working concentration with 0.9% NaCl solution. Caffeine was purchased from Sigma Chemical Co. (St. Louis, MO) and was dissolved in sodium benzoate (37.5 mg/ml) solution. Alprazolam and caffeine were administered s.c. and p.o. by gavage, respectively, in an injection volume of 1 ml/kg body weight.

Apparatus. Four operant Plexiglas chambers were used and have been described previously (Lau and Wang, 1996). Each chamber, equipped with a response lever and a stainless steel food-pellet receptacle into which 45-mg dustless pellets (BioServ, Frenchtown, NJ) could be delivered, was enclosed in a sound-attenuating shell and was controlled by an IBM-type 486 X computer. Session contingencies were programmed and data recorded using QuickBasic.

Procedure. Animals were magazine trained on a noncontingent random-time schedule initially for 15 min and responses on the lever were shaped by successive approximation and reinforced when IRTs were greater than 3 sec. The temporal requirement was slowly increased to an IRT of 45 sec over 10 to 20 sessions. A 3-hr operant session was conducted daily. After performance had stabilized, a drug-administration series began. The series consisted of: 1) Alprazolam dose-response determination (vehicle, 0.125, 0.4, 1.25, 4 and 7 mg/kg); 2) caffeine dose-response determination (vehicle, 5, 10, 20, 40, 80 and 120 mg/kg); 3) alprazolam-caffeine combinations: (a) alprazolam + 30 mg/kg caffeine, i.e., vehicle + vehicle; vehicle + 30 mg/kg caffeine; 0.125 to 7 mg/kg alprazolam + 30 mg/kg caffeine (b) alprazolam + 20 mg/kg caffeine, i.e., vehicle + 20 mg/kg caffeine; 0.125 to 7 mg/kg alprazolam + 20 mg/kg caffeine. At the end of each combination series, the caffeine dose for that series (e.g., vehicle + 20 mg/kg caffeine) was redetermined. Injections were given immediately before the start of a session and separated by 3 to 5 days. Injections within each series were given in a quasirandom order. Each drug series was separated by 10 noninjection sessions.

Data analyses. The IRT distributions after the administration of vehicle, alprazolam, caffeine and alprazolam-caffeine combinations were analyzed for 3-hr sessions, omitting the first 2 min, which was treated as the settling time. Baseline IRT distributions for each session that immediately preceded an injection also were analyzed. Behavioral parameters were derived from the IRT distributions: shorter (nonreinforced)-response rate, reinforcement rate, total response rate and efficiency. Total number of responses consisted of responses with IRT  45 and < 45 sec, which are the reinforced and nonreinforced responses, respectively. These responses were calculated as rates (responses per min). Efficiency was calculated as the ratio of reinforcement rate to the total response rate. We have found that both the reinforcement rate in the 45-to-55- and 45-sec bins decreased equivalently as a function of dosage for alprazolam and caffeine by the i.p. route. The 45- to 55-sec bin function required a lower dose to reach E_max for both drugs, and consequently resulted in smaller ED₅₀ values (Lau and Wang, 1996). The 45- to 55-sec bin function has been used successfully to characterize the alprazolam-caffeine interaction, and justification will be given in "Results" referring to figures 4A-C. Specific attention was given to the 45- to 55-sec bin data in this study, facilitating the comparison of our results with those from the previous study.

View larger version (27K): [in this window] [in a new window]   Fig. 4.   Mean reinforcement rate-, efficiency-, total-, and shorter-response-rate-time profiles: (A) baseline; (B) 1.25 mg/kg alprazolam and (C) 20 mg/kg caffeine. Reinforcement rate and efficiency measures on left ordinate scales. Total respond and shorter-response rate measures on right ordinate scales.

Characterization of alprazolam-caffeine interaction by DRC method proposed by Pöch. Alprazolam-caffeine interactions were evaluated by comparing the dose-response curves for alprazolam in the presence and absence of two caffeine doses following the method of combined effects proposed by Pöch and his colleagues (Pöch, 1993; Pöch et al., 1990) and have been described previously (Lau and Wang, 1996). From the results of seven-animal medians, rather than means, values of both observed and expected DRCs of the DRL behavior were used for statistical evaluation of observed versus expected frequencies by the ² goodness-of-fit test.

Pharmacokinetics of Alprazolam, Caffeine and their Combinations

Animals. Eight male, albino rats of the same strain were used under the conditions and food-limitation regimen used above. The mean initial, adult free-feeding body weight was 388 g (range 380-391 g).

Drugs and reagents. Alprazolam, -hydroxyalprazolam and 4-hydroxyalprazolam were obtained from Upjohn Laboratories, Kalamazoo, MI. Caffeine, theobromine, paraxanthine, theophylline and -hydroxyethyltheophylline were purchased from Sigma Chemical Company Co., St. Louis, MO. Reagents were obtained from standard commercial sources.

HPLC determination of alprazolam, caffeine and their metabolites and serum sampling. HPLC. Serum microsample HPLC methods for determination of alprazolam, caffeine and their metabolites have been described previously (Jin and Lau, 1994; Lau and Falk, 1991). Separation for both drugs was performed on Beckman Ultrasphere C₁₈ columns (5-µm particle size, 150 × 2 mm I.D.). Programmable absorbance UV detectors 785A (Applied Biosystems Instruments, Foster City, CA) were operated at 230 and 270 nm for alprazolam and caffeine methods, respectively. The capacity factors for demoxepam used as internal standard, 4-hydroxyalprazolam, -hydroxyalprazolam and alprazolam were 2.08, 2.73, 3.37 and 4.43, respectively, whereas for theobromine, paraxanthine, theophylline, -hydroxyethyltheophylline (internal standard) and caffeine were 1.31, 2.52, 2.97, 3.73 and 6.45, respectively. There was no mutual interference between these two agents or among their metabolites with respect to the HPLC methods.

0-

0-

0-

0-

0-

0-

0-

0-

0-

PK-PD Modeling: PK and DRL Performance

Data Analysis. Integration of PK and PD was based on the relation between mean serum alprazolam concentration-time profiles for the three s.c. doses (1.25-7 mg/kg) in the presence and absence of p.o. 20 mg/kg caffeine of group 1 in PK studies (N = 4) and the respective mean behavior-time profiles in PD studies (N = 7).

View larger version (24K): [in this window] [in a new window]   Fig. 1.   Diagrammatic representation of the integrated PK-PD model used to describe the reinforcement rate in the 45- to 55-sec bin (effect-link model) after s.c. administration of a single dose of alprazolam. The k(1, 2) and k(2, 1) are the inter-cpt rate constants, and k(0, 1) is the elimination rate constant.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

DRL Performance

Figure 2A-B, as an example, show the effects of alprazolam and caffeine on IRT distributions for the first 30 min of the sessions. For baseline days and vehicle administration, the highest response rate occurred in the 40- to 50-sec band. Both alprazolam and caffeine decreased the reinforced, and increased the nonreinforced response rate in a dose-related fashion.

View larger version (35K): [in this window] [in a new window]   Fig. 2.   Mean effects of (A) s.c. alprazolam (0-7 mg/kg); (B) p.o. caffeine (0-120 mg/kg) on IRT distributions for 2 to 30 min after drug administration. All the responses were nonreinforced (<45 sec) before the first arrow and reinforced (45 sec) after the first arrow. Responses between the two arrows were the 45- to 55-sec bin responses.

Figure 3A-F show an overview of DRL performance for the 3-hr session after vehicle and drug administration. Decreases in reinforcement rate in the 45- to 55-sec bin, and in bins larger than 45 sec, were linear with respect to alprazolam dose, whereas these functions for caffeine reached a plateau at higher doses (fig. 3A). At higher doses, both alprazolam and caffeine increased shorter IRTs (<45 sec); however, the increases were more profound for caffeine than for alprazolam (fig. 3B). The opposing relation between the reinforced and nonreinforced response rate after drug administration resulted in a higher total response rate only at the 40-mg/kg caffeine dose (fig. 3C). Consequently, efficiency for both drugs was similar to the reinforcement-rate function across doses (fig. 3D). For both alprazolam and caffeine, dose-response relations for the peak area measure were similar to those in the 45- to 55-sec bin (fig. 3E), whereas the center of the IRT distribution peak shifted to the shorter IRTs as shown by the peak location measure (fig. 3F).

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Mean (S.E.) % baseline dose-response curves of alprazolam and caffeine for 3-hr sessions: (A) reinforcement rate in the 45- and 45- to 55-sec bins; (B) shorter-response rate; (C) total response rate; (D) efficiency; (E) peak area and (F) peak location. % B, percent baseline.

Figure 4A shows that the mean behavioral performance measures (reinforcement rate for both the responses >45 sec and in the 45- to 55-sec bin, shorter response rate, total response rate and efficiency) during baseline days were similar across the duration of the 3-hr sessions. The respective performance measures for 1.25 mg/kg alprazolam and 20 mg/kg caffeine (one dose from each drug as examples) are shown in Figures 4B and C, respectively. During baseline days, the ratio of reinforcement rate in the 45- to 55-sec bin to the total reinforcement rate was approximately 0.9 (e.g., 0.69/0.79 = 0.87 for time point at 60 min, fig. 4A), which implied that 90% of the reinforced responses occurred in the 45- to 55-sec bin. For the 1.25-mg/kg dose, the ratios were 0.25 and 0.9 at 15 and 180 min, respectively (fig. 4B). The smaller the ratio, the more IRTs occurred in the bins larger than 55 sec. Although both alprazolam and caffeine decreased reinforcement rate, it is apparent that alprazolam effects were short-lived, whereas they remained relatively constant for caffeine across the session. For example, at time point 150 min, the reinforcement rate in the 45- to 55-sec bin was approximately at the baseline level for alprazolam (0.64 min¹), whereas it remained low for caffeine (0.34 min¹). Reinforcement rate in the 45- to 55-sec bin was more sensitive to drug effects than the total reinforcement rate was, especially during the phase when performance was returning to baseline. It also required lower doses to reach E_max than the total reinforcement rate measure did. Therefore, the 45- to 55-sec bin was used to characterize the effects of drugs when given alone and in combination to minimize the possibility of behavioral toxicity that might occur if higher doses were necessary to perform the analysis. The highest efficiency occurred at the time when the shorter-response rate was the lowest.

Inasmuch as effects of alprazolam were short-lived, DRCs for alprazolam and caffeine in the 45- to 55-sec bin were constructed by ALLFIT using four time periods (fig. 5A-D). Performance attained a plateau for alprazolam for the first two time periods, but they differed in E_max values, 13.08 and 4.9% for 2 to 30 and 31 to 60 min, respectively. The E_max value in the second time period was used for the two later time periods because the dose-response relation for alprazolam is unlikely to change across time after it had reached E_max. DRCs of alprazolam shifted to the right across the four time periods, whereas those curves remained similar for caffeine (fig. 5A-D). Thus, ED₅₀ values for alprazolam changed across the four time periods from 0.26, 0.5, 1.72 to 5.26 mg/kg (i.e., 0.84, 1.62, 5.57 to 17.04 µmol/kg, respectively), whereas for caffeine those values, 14.13, 18.04, 20.42 and 16.71 mg/kg (i.e., 72.8, 92.9, 105.2 and 86.1 µmol/kg, respectively) remained relatively similar throughout the session. As a result, the potency ratios of these two agents in terms of µmol/kg changed during a session from 86 to 5. The slope values for alprazolam and caffeine were similar, 1.82 and 1.86, respectively. For the first hour, the effect of alprazolam on DRL performance plateaued at 4 mg/kg, but approximately linear dose-response relations occurred for the second and third hours, with a disappearance of effect for the lower alprazolam doses.

View larger version (28K): [in this window] [in a new window]   Fig. 5.   Alprazolam and caffeine DRCs constructed by using median % baseline reinforcement rate in the 45-55 s bin for four time periods: (A) 2 to 30 min; (B) 31 to 60 min; (C) 61 to 120 min; (D) 121 to 180 min.

As described in "Methods," for each drug combination series, s.c. saline + a fixed dose of p.o. caffeine (20 or 30 mg/kg) was given not only in the beginning, but also at the end of a series, and these points are also shown in figure 5A-D. There were only minor variations observed for the 30-mg/kg caffeine dose in the time periods 31 to 60 and 61 to 120 min, which implied that the effects of a fixed dose of caffeine did not vary across a combination series. Thus, mean value of the two treatments in the series of a given caffeine dose was used to characterize the combined effects below.

Using the 31- to 60-min time period as an example, the effects of alprazolam in the presence of two fixed doses of caffeine (20-30 mg/kg) in the reinforcement rate in the 45- to 55-sec bin were analyzed by the Pöch DRC method for the combined effects (Fig. 6B-C). Both the expected independent and additive curves for all the caffeine combinations (10, 20, 30, 40, 80 and 120 mg/kg) can be obtained simply and predicted from the DRCs of alprazolam and caffeine (fig. 5A-D) as described in "Methods," an advantage of the Pöch method, and the expected curves are shown in sequence as pairs in figure 6A. The combined effects of alprazolam in the presence of two doses of caffeine did not differ from either the theoretical expected independent or additive curves as reflected by the ² statistics, although there were two observed median values deviant from the expected (e.g., 1.25 mg/kg alprazolam + 30 mg/kg caffeine). Thus, the combined effects were neither synergistic nor antagonistic.

View larger version (35K): [in this window] [in a new window]   Fig. 6.   Alprazolam DRCs in the presence and absence of a fixed dose of caffeine (20 or 30 mg/kg) constructed by using median % baseline reinforcement rate in the 45- to 55-sec bin for time period 31 to 60 min. (A) Theoretical expected additive and independent curves shown as pairs for each caffeine dose (10-120 mg/kg); (B) alprazolam + 20 mg/kg caffeine and (C) alprazolam + 30 mg/kg caffeine. The observed values and the two expected curves are shown for each combination series with 2 statistics.

The mean performance of behavior-time profiles in the 45- to 55-sec bin for caffeine, alprazolam in the presence and absence of two fixed doses of caffeine (20-30 mg/kg) and the expected independent curves from 15 to 180 min, are shown in figs. 7 and 8. The effects of vehicle administration (saline, sodium benzoate and their combination) were close to baseline (100%) on the DRL behavior-time profiles except at the 15 min for the vehicle combinations. Each of these vehicle treatments is shown in separate quadrants for a clear view and to avoid repetition. Generally, the decrements in the 45- to 55-sec bin for the three highest alprazolam doses (1.25-7 mg/kg) in the presence of 20 mg/kg caffeine were similar to those occurring when alprazolam was given alone. Although, in the presence of 30 mg/kg caffeine the combined effects deviated from alprazolam effects after 60 min in a dose-related fashion and approached independence. However, for the lowest alprazolam dose (0.125 mg/kg), the combined effects were closer to those of caffeine or independent effects rather than to those of alprazolam. The mean expected additive curves are not shown for the combination series as those curves were not separable from the independent curves. The two drug combination series showed independent interaction for all the time points across the 3-hr session (figs. 7 and 8), except the time point at 180 min for 7 mg/kg alprazolam + 20 mg/kg caffeine; the decrement in reinforcement rate was less than the expected independent effect, although it did not differ from the effect of alprazolam given alone. These results demonstrate that comparing observed combined effects to calculated expected curves is crucial for characterizing drug interaction.

View larger version (36K): [in this window] [in a new window]   Fig. 7.   Mean (S.E.) % baseline reinforcement rate- (45- to 55-sec) time profiles after 20 mg/kg caffeine alone, and alprazolam (0.125-7 mg/kg) in the presence and absence of 20 mg/kg caffeine. Expected independent curves (S.E.) are shown for each combination. Saline and sodium benzoic acid are the vehicles for alprazolam and caffeine, respectively. *P < .05 relative to the respective independent curve.

View larger version (35K): [in this window] [in a new window]   Fig. 8.   Mean (S.E.) % baseline reinforcement rate- (45- to 55-sec) time profiles after 30 mg/kg caffeine alone, and alprazolam (0.125-7 mg/kg) in the presence and absence of 30 mg/kg caffeine. Saline, and sodium benzoic acid are the vehicles for alprazolam and caffeine, respectively. Expected independent curves (S.E.) are shown for each combination.

For the three higher doses of alprazolam (1.25-7 mg/kg) in the presence and absence of caffeine (20 mg/kg p.o.), the shorter-response rate decreased to baseline level after the initial stimulation, but again increased in a dose-related fashion in terms of its time to peak and the duration of the peak (fig. 9). For example, the peak times were at 60, 90, and 150 min for 1.25, 4 and 7 mg/kg, respectively, and the peak durations progressively increased. In each case the second peak lasted longer, but was less elevated, compared to the first peak. Caffeine at 20 mg/kg produced a milder stimulation (150% of baseline) across the session except a 350% increase was observed in shorter response rate at 5 min. Thus, the pattern of effects for the shorter-response rate differed for the two drugs. However, the presence of caffeine did not alter the dynamics of the above two-peak phenomenon (P > .05).

View larger version (26K): [in this window] [in a new window]   Fig. 9.   Mean (S.E.) % baseline shorter-response rate- (<45-sec-time profiles after alprazolam (1.25-7 mg/kg) in the presence and absence of 20 mg/kg caffeine.

Pharmacokinetics of Alprazolam, Caffeine and their Combinations

Alprazolam PK by the s.c. route in the presence and absence of 20 mg/kg p.o. caffeine. After i.v. administration, alprazolam was eliminated according to a biphasic process. Alprazolam was rapidly distributed with a mean distribution t_1/2 of 5.14 min, and was eliminated with a mean terminal elimination t_1/2 of 40.58 min (table 1). The V_c, V_ss and clearance were 1.65 liter/kg, 3.85 liter/kg and 6.15 liter/hr/kg, respectively. Alprazolam metabolites, the two oxidative metabolites, 4-hydroxyalprazolam and -hydroxyalprazolam, were not detectable.

(0-)

View larger version (21K): [in this window] [in a new window]   Fig. 10.   Mean (S.E.) serum alprazolam concentration-time profiles and the fitted curves after PK modeling following administration: (A) alprazolam s.c. 1.25 to 7 mg/kg and (B) alprazolam s.c. 1.25 to 7 mg/kg + p.o 20 mg/kg caffeine.

Oral caffeine PK in the presence and absence of three s.c. doses of alprazolam (1.25-7 mg/kg). The serum caffeine and its three DMX metabolites (theobromine, paraxanthine and theophylline) concentration-time profiles after five doses of p.o. caffeine (10-120 mg/kg) are shown in Fig. 11A-D. For the doses of 10, 40 and 120 mg/kg caffeine, not all the serum samples were obtained from jugular vein catheters. In two of the four animals, their jugular vein catheters became occluded and blood could not be withdrawn after nine blood-sampling series as described in "Materials and Methods" Thus, for these two animals, tail-tip blood samples were used for determining the serum concentrations of caffeine and the three DMXs. Blood samples of one animal in this group, who had completed the blood sampling series, was used to determine whether the values estimated from the tail-tip samples were in accordance with those values obtained from the jugular vein samples by simultaneously collecting both samples at 5, 15, 30 and 60 min after 40 mg/kg p.o. caffeine administration. Serum caffeine concentrations at 5, 15, 30 and 60 min for tail-tip and jugular vein samples, respectively, were 3.26 and 8.88 µg/ml; 13.21 and 15.91 µg/ml; 17.8 and 19.62 µg/ml; 20.85 and 21.73 µg/ml, respectively. Serum caffeine concentration was much lower in tail-tip than in the jugular vein sample at 5 min, but progressively indifferent for the two samples with time. Similar results were found for the serum DMX concentrations. Thus, for these three caffeine doses (10, 40 and 120 mg/kg), mean serum caffeine and DMX concentrations were only calculated between 15 to 180 min.

View larger version (28K): [in this window] [in a new window]   Fig. 11.   Mean (S.E.) serum concentration-time profiles after p.o. caffeine (10-120 mg/kg) administration: (A) caffeine; (B) theobromine; (C) paraxanthine and (D) theophylline.

View larger version (32K): [in this window] [in a new window]   Fig. 12.   Mean (S.E.) serum caffeine and total DMX concentration-time profiles after p.o. 20 mg/kg caffeine in the presence and absence of 1.25 to 7 mg/kg alprazolam administration. *P < .05 relative to the respective serum caffeine and DMX concentrations without alprazolam.

PK interaction between alprazolam and caffeine by the i.v. route. Both alprazolam and caffeine i.v. serum concentration-time profiles were not altered by concurrent administration of i.v. caffeine and alprazolam, respectively (fig. 13A-B). The PK parameters for alprazolam were not influenced by caffeine (table 2). After i.v. administration, alprazolam was eliminated according to a biphasic process and the PK parameter values estimated from the concentration-time profiles were similar to those for the group 1 (table 1). Caffeine PK parameters ± alprazolam could not be determined accurately using the data in figure 13B, as the t_1/2 of caffeine was much longer than that of alprazolam. Caffeine PK parameters listed in table 2 were obtained from a different group of animals (N = 4) under feeding conditions similar to those used in this experiment (C.E. Lau, Y. Wang and F. Ma, unpublished data). After i.v. administration, caffeine was eliminated according to a monophasic process. V_c and V_ss for alprazolam were larger than those for caffeine. Alprazolam clearance was markedly greater than for caffeine, 6.9 vs. 0.29 liter/hr/kg, which accounted for its shorter t_1/2 compared to caffeine (24.8 vs. 187 min). The two hydroxy metabolites of alprazolam were not detectable by the i.v. route in the presence or absence of caffeine. Alprazolam also did not alter the AUC_{(0-6 hr)} values of caffeine or the three DMXs (table 2, bottom panel).

View larger version (19K): [in this window] [in a new window]   Fig. 13.   Mean (S.E.) serum concentration-time profiles for: (A) alprazolam (1.25 mg/kg alprazolam ± 10 mg/kg caffeine); (B) caffeine (10 mg/kg caffeine ± 1.25 mg/kg alprazolam) by the i.v. route.

PK-PD Modeling: PK and DRL performance

PK-PD model of serum alprazolam concentration-time profiles in the presence and absence of caffeine (20 mg/kg, p.o.). Alprazolam metabolite concentrations were either low or not detectable, and with their relative low potency compared to the parent compound, these metabolites were not included in the PD analysis. Alprazolam distribution and elimination characteristics were determined initially for the i.v. 1.25-mg/kg dose using the mean alprazolam serum concentration-time profile. The bioavailability values of the three s.c. alprazolam doses were complete using the mean data. All the values of the intercompartmental rate constants derived from the i.v. route describe the three s.c. alprazolam doses ± 20 mg/kg caffeine profiles well, except the elimination rate constant values from the central cpt varied somewhat for the two higher alprazolam doses when given alone, as shown in table 3. Figure 10A-B show the mean observed and fitted serum alprazolam concentration-time profiles of the three alprazolam doses ± 20 mg/kg caffeine using these PK parameters, respectively.

View larger version (32K): [in this window] [in a new window]   Fig. 14.   Mean % baseline reinforcement rate in the 45- to 55-sec bin time profiles and the predicted curves after PK-PD modeling following administration of alprazolam (1.25-7 mg/kg). (A) alprazolam alone; (B) alprazolam + p.o. 20 mg/kg caffeine and (C) simulated effect-time profiles for alprazolam doses ± p.o. 20 mg/kg caffeine by using the estimated PK-PD parameters.

Relation between serum drug concentrations and DRL performance. The data for the first 10 min for both PK and PD (shorter-response and reinforcement rate in the 45- to 55-sec bin) measures were not used in these analyses owing to the equilibration time required for serum alprazolam concentration to the effect cpts (fig. 14). Figure 15 shows the relations between mean serum alprazolam or mean caffeine concentration (N = 4 rats) and mean DRL performance in the 45- to 55-sec bin (N = 7 rats) as constructed using ALLFIT. The 0.125- and 0.4-mg/kg serum alprazolam concentrations were obtained from simulation of the PK parameters (table 3, fig. 10A). The effects of alprazolam on DRL performance were concentration related regardless of alprazolam doses (0.125-7 mg/kg). For example, for the 1.25-mg/kg alprazolam dose, a full concentration-effect relation (from E₀ to E_max) was observed, whereas other doses exhibited only partial functions, i.e., high or low doses associated with larger or smaller effects, respectively.

View larger version (25K): [in this window] [in a new window]   Fig. 15.   Mean % baseline reinforcement rate in the 45- to 55-sec bin (N = 7) vs. mean serum alprazolam or caffeine concentrations (N = 4). Lines are the curves fitted by ALLFIT.

View larger version (26K): [in this window] [in a new window]   Fig. 16.   Mean (S.E.) % baseline shorter-response rate vs. alprazolam serum concentrations for the three doses of alprazolam (1.25-7 mg/kg) ± p.o. 20 mg/kg caffeine.

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

This study investigated PK and PD interactions between alprazolam and caffeine initiated by our previous research (Lau and Wang, 1996), but extended the findings to additional routes of administration, doses, as well as two indices of DRL performance. Two DRL performance measures, reinforced and nonreinforced response rates, not only yielded similar conclusions with respect to drug interaction, but also bore interesting differential relations to serum alprazolam concentrations. Behavior-time profile is the method of choice for studying this kind of drug action and interaction. It would have been simpler to analyze the 3-hr session data in a collapsed form to make inferences, but that would have omitted the dynamics of the on-going behavior and its relation to PK. The bioavailabilities of s.c. alprazolam in the presence and absence of caffeine were high, and were not determined for the i.p. route in the previous study. The lack of PK interaction between alprazolam and caffeine by the i.v. route suggested that the effect of s.c. alprazolam on p.o. caffeine PK was an indirect effect of caffeine absorption. PK models permit the prediction of serum alprazolam concentration for other doses in the linear range, especially for lower doses that yielded drug concentrations below analytical sensitivity. Although the combined effects are not distinguishable in terms of additivity or independence by using the Pöch DRC method, independent interaction is suggested by PK-PD modeling as reflected in the IC₅₀ values (table 3).

To study drug interaction, drug effects need to be reproducible, otherwise tolerance and sensitization might be interpreted as antagonism and synergism, respectively. We have found that within-subject variability in reinforcement rate on DRL 45-sec was not different after two consecutive s.c. doses of midazolam separated by 3 to 5 days (Lau et al., 1996). The effects of alprazolam on DRL performance were similar to those of midazolam, not only in reinforcement rate, but also in shorter-response rate (unpublished data). Furthermore, effects of caffeine + saline at the beginning and end of a combination series on reinforcement rate in the 45- to 55 sec bin (fig. 5) and on shorter-response rate were approximately similar (data not shown), suggesting no tolerance or sensitization occurred as a result of the acute repeated caffeine administration. Thus, the observed combined effects resulted from drug interaction.

The PK of s.c. alprazolam, p.o. caffeine and their combinations mainly mirrored the respective behavior-time profiles of the reinforcement rate in the 45- to 55-sec bin in 3-hr sessions. The onset of alprazolam action was rapid and its duration of action short, although the effect of caffeine remained mainly constant throughout the session, except for the 10-mg/kg dose that showed acute tolerance (fig. 15). As a result, the potency ratio of these two drugs changed markedly during the session, which determined the expression of the combined effects. The combined effects of alprazolam and caffeine were not distinguishable in terms of additivity or independence, neither were they synergistic nor antagonistic, as shown by the DRCs (fig. 6), which were in agreement with the findings reported previously (Lau and Wang, 1996). Similar results were obtained for most of the behavior-time profiles (figs. 7 and 8), except on one occasion diminished effects were observed only in the latter part of the session (at time point 180 min for 7 mg/kg alprazolam + 20 mg/kg caffeine), which could be attributable to the lower serum caffeine and DMX concentrations produced by alprazolam (fig. 12). This was not surprising, for as the potency ratio of alprazolam decreased across the session, caffeine and its active metabolites became more dominate in the expression of the combined effects. These results suggest that the interaction between alprazolam and caffeine occurs by PK, rather than by a receptor mechanism. The conclusion is warranted by the PK results and the comparison of the observed combined effects to the theoretical expected values, crucial operations for the characterization of combined effects.

The reason for using the reinforcement rate in the 45- to 55-sec bin instead of total reinforcement rate as an evaluation of DRL performance is evident in figure 4. Total responses are comprised of both the reinforced and nonreinforced responses, and closely parallel the shorter-response rate (<45 sec) for both alprazolam and caffeine (fig. 4). Thus, the shorter-response, rather than the total response rate was chosen as the second measure to characterize the combined effects. In the previous study, the effects of alprazolam ± caffeine on shorter-response rate were not explored (Lau and Wang, 1996). The shorter-response rate increased for the higher doses after alprazolam and caffeine administration (fig. 3), but differed in pattern for the two drugs. For 20 mg/kg caffeine, the shorter-response rate remained at baseline level after the initial stimulation (fig. 9). Conversely for alprazolam, the shorter-response rate decreased to baseline level after the initial stimulation, but increased again in a dose-related fashion in terms of its time to peak and duration of the peak (fig. 9). It is interesting that for the combination series (1.25-7 mg/kg alprazolam + 20 mg/kg caffeine), caffeine did not alter the dynamics of the shorter-response rate, evidence of independence (figs. 9 and 16). Similar results also were found for the combination series of 1.25 to 7 mg/kg alprazolam + 30 mg/kg caffeine (data not shown). These results further demonstrated that alprazolam, but not caffeine, is more potent in determining the pattern of the shorter-response rate in the combined effects.

Alprazolam is the most widely prescribed BZ, and is used as an anxiolytic, antipanic and antidepressant agent (Dawson et al., 1984; Fawcett and Kravitz 1982), but adverse side effects of BZs have been increasingly recognized in recent years clinically, e.g., early-morning insomnia and daytime anxiety, tension or panic (Vgontzas et al., 1995). The occurrence of these two kinds of effects may depend upon the levels of BZ concentration in the body. For example, hypnotic effects may be associated with higher, whereas early-morning insomnia with the approach to lower BZ concentrations. The second peak of the shorter-response rate for alprazolam was related to the lower range of serum alprazolam concentrations (fig. 16). Inasmuch as alprazolam is shorter lived (t_1/2 = 25-57 min) in rats than in humans (t_1/2 6-16 h) (Greenblatt et al., 1983; Smith et al., 1984), the concentration-dependent increases constituting the second peak in the shorter-response rate, and decreases in the reinforcement rate in the 45- to 55-sec bin, suggests this as the possible mechanism of the two kinds of effects observed in humans, therapeutic and adverse side effects. With the use of PK-PD modeling, one could incorporate both kinds of effects and link them to alprazolam PK to predict therapeutic and adverse side effects. To our knowledge, no explicit PK-PD modeling has been developed that attempts to describe these relations.

Our study along with the previous results (Lau and Wang, 1996) demonstrated that, to characterize the combined effects properly, it is important to compare the experimental values to the calculated expected effects across time. The interaction between BZ and MX reported in the literature mainly has been characterized qualitatively rather than quantitatively; this may partially account for the differences found across studies. Several reports have sampled limited portions of such general dose-combination and temporal functions, yielding results consistent with the independent actions expected from physiological (i.e., functional) antagonism, rather than the competitive antagonism or synergistic actions sometimes inferred (Marrosu et al., 1985; Stirt, 1981; Wangler and Kilpatrick, 1985). Furthermore, the use of different terms to characterize and interpret combined effects complicates this issue. Terms such as "antagonism" and "additivity" sometimes are used to refer to mechanisms of action, and at other times simply to indicate the direction of an observed effect.

Both alprazolam and caffeine are rapidly absorbed from extravascular routes (Arnaud, 1993; Lau and Wang, 1996) and are highly lipophilic (Arnaud, 1993; Greenblatt and Shader, 1987). There is a rapid equilibrium between drug concentration in blood and at central sites of action for both drugs, a factor important in the onset of drug action. The equilibration half-lives between serum alprazolam concentration in the central and effect cpts (t_1/2k_eo) for alprazolam ± 20 mg/kg caffeine are 4.4 and 3.0 min, respectively (table 3). These values were similar to the value reported for midazolam, 2.2 min (Breimer et al., 1991). No acute tolerance in the reinforcement rate in the 45- to 55-sec bin was observed for alprazolam and caffeine, except for 10-mg/kg caffeine dose (fig. 15).

The PK of alprazolam was not altered by caffeine, whereas the PK of caffeine was altered by alprazolam and resulted a significantly decreased formation of its three active DMX metabolites (fig. 12). Orally administered caffeine is absorbed from the small intestine and the stomach (Chvasta and Cook, 1971). The acute PK interaction observed in our study was not a metabolic one, as no interaction was observed by the i.v. route (fig. 13). Rather, alprazolam affected factors influencing the absorption of caffeine, such as gastrointestinal motility (Fargeas et al., 1984). Therefore, this kind of indirect interaction would not be avoided by using a BZ that is metabolized by conjugation with glucuronic acid instead of the cytochrome P-450 enzyme system, e.g., lorazepam (Schillings et al., 1975). However, when alprazolam and caffeine were given by the i.p. route, the decreased caffeine and DMXs AUC_{(0-3 hr)} observed in a previous study might have resulted from a different mechanism (Lau and Wang, 1996). After i.p. administration, both drugs pass from the peritoneal cavity through intercellular gaps in the mesenteric wall and the surrounding capillaries into general circulation. Alprazolam and caffeine might compete with each other during the absorption phase. Nevertheless, the decreases in serum caffeine and DMX concentrations accounted for the diminished action observed for the combined effects at 180 min for 7 mg/kg alprazolam + 20 mg/kg caffeine (fig. 8). It is interesting that they were not different from the effects of alprazolam given alone. However, the effects of alprazolam on caffeine PK were alprazolam dose dependent. This was apparent not only from the PK results (fig. 12), but also from the combined effects of lower doses of alprazolam (0.125 and 0.4 mg/kg) + two caffeine doses (20-30 mg/kg), which were not diminished as shown in figures 6, 7, 8. These dose-dependent effects of alprazolam on caffeine PK might account for the controversial results reported for PK BZ-MX interaction that were based on single-dose combinations with no information regarding the concentration changes of active metabolites (Ghoneim et al., 1986; Henauer et al., 1983).

Although drug interactions on a receptor level can be readily studied in vitro (Möhler and Richards, 1981), the predictive value of such studies is limited. Even beyond PK considerations, not only may in vivo drug-drug interactions occur at the receptor level, but drugs may also interact at the level of post-receptor events (Ariens et al., 1956). This study aimed to quantify the interaction between alprazolam and caffeine using DRL 45-sec performance for the endpoints. The DRL performance satisfies all criteria as a relevant PD measure (Laurijssens and Greenblatt, 1996). The effect-link sigmoid E_max model method has successfully described the relation of PD and PK for many drugs, including midazolam (Breimer et al., 1991). Drug concentration in the biophase (e.g., effect site) is not in rapid equilibrium with the blood and can produce a delay. The delay of drug action also may be attributed to any mechanism of action requiring appreciable time, e.g., protein synthesis, resulting in a lag between the concentration-time and effect-time. The IC₅₀ values were 0.0201 and 0.0199 µg/ml for alprazolam alone and in the presence of 20 mg/kg caffeine, respectively, from the PK-PD models (table 3) which is consonant with the values reported in a previous study by using tail-tip blood samples (Lau and Wang, 1996). All these values were derived from the mean data that did not permit assessing intersubject variability. Nevertheless these results demonstrated that alprazolam IC₅₀ values were not affected by the presence of caffeine, whereas E₀ differed in the presence of caffeine. It suggested that the interaction is not competitive, but independent. The competitive interaction between flumazenil and midazolam was demonstrated by an increase in the midazolam EC₅₀ value in the presence of flumazenil using PK-PD modeling in humans (Breimer et al., 1991).

Inasmuch as E_max for alprazolam was 4.9, which was close to zero (the largest possible E_max), the effects of caffeine on E_max were not detected. If caffeine synergized the effects of alprazolam, they could only be detected in combined effects when serum alprazolam concentrations were at the lower end (e.g., for all the combinations after 2 hr, figs. 7 and 8), but these were not different from independence. However, if caffeine antagonized the disruptive effects of alprazolam, it would be apparent in all the combined effects when serum alprazolam concentrations were at the higher end (e.g., the combined effects of 1.25-7 mg/kg alprazolam at first hour, figs. 7 and 8), but this did not occur. Thus, the determinants of the combined effects depend on the potency ratio and efficacy of the two drugs, and the relation between serum alprazolam and caffeine concentrations. Inasmuch as behavior performance and PK were conducted in different groups of animals, the minor between-subject variability might reflect the diminished effects observed only at 180 min for 7 mg/kg alprazolam + 20 mg/kg caffeine rather than for the other alprazolam dose combinations (1.25-4 mg/kg). However, if caffeine had altered alprazolam PK, then the combined effects would have yielded a totally different perspective than the present one.

In summary, the present approach can be applied to any drug-drug interaction study, and use other behavioral paradigms. In humans, serum monitoring is often done to maximize treatment effectiveness, but these data can function only as an uncertain guide for animal behavioral research. Alprazolam is a drug with a markedly different half-life in rats and humans (Greenblatt and Wright, 1993; Jin and Lau, 1994; Owens et al., 1991). Similar species differences in half-lives have been found for other BZs: diazepam (Hironaka et al., 1984; Igari et al. 1982; Sethy et al., 1987) and flurazepam (Lau et al., 1987). Without assessing the potency ratio of the two agents across a session, comparing the combined effects to the expected calculated effects and acquiring the parallel PK data, our study would have been difficult to interpret.