Behavioral economics has been used in the field of behavioral pharmacology to assess the reinforcing functions of drugs (e.g., Bickel et al. 1993; Hursh and Winger 1995). Specifically, analyses of demand provide a way to mathematically quantify a drug’s reinforcing effectiveness when it is available alone (e.g., Hursh and Winger 1995) or when another drug is concurrently available (e.g., Bickel et al. 1992; Spiga et al. 2005). In a demand-curve analysis, consumption of a drug is measured as a function of unit price. Unit price is expressed as the response requirement divided by the dose (Bickel et al. 1990). Thus, the price of a drug can be increased by increasing the response requirement, decreasing the dose, or both. As the price of a commodity increases, subjects initially increase their response to maintain a relatively constant level of consumption. At these prices, demand is inelastic. With further increases in price, consumption decreases markedly. Demand is termed elastic at these prices, and the rate at which consumption decreases as a function of increases in price is referred to as demand elasticity. Elasticity represents a dose-independent measure of the reinforcing effectiveness of the drug; that is, the effectiveness of the drug as a reinforcer is inversely related to the speed at which consumption decreases with increases in price. Another measure which corresponds to elasticity is the price at which demand shifts from inelastic to elastic, termed Pmax. A higher Pmax is indicative of a more effective reinforcer.

Using demand-curve analysis, it is possible to compare the elasticity of different drugs. Hursh and Winger (1995) compared demand elasticity for alfentanil, nalbuphine, cocaine, and methohexital in rhesus monkeys using a normalized demand-curve analysis. Normalization allows for a dose- and potency-independent measurement of elasticity by equating the consumption of each drug at its lowest price and comparing elasticity independent of the initial level of behavior maintained by each drug. Rank-ordered in terms of their reinforcing effectiveness according to Pmax, alfentanil had the highest Pmax, followed by cocaine, nalbuphine, and methohexital.

Demand-curve analyses also have been used to assess changes in demand for a drug as the result of the concurrent availability of a second drug (Bickel et al. 1992; Spiga et al. 2005). In these procedures, the price for one drug is increased (resulting in a corresponding decrease in its consumption) while the price of a concurrently available drug is held constant. Three types of economic interactions may result: The drugs may be substitutes, independent, or complements (Hursh 2000). That is, as the price of one drug is increased, consumption of the other drug may increase (substitutes), stay the same (independent), or decrease (complements). For example, Bickel et al. (1995) reanalyzed data from previous experiments in which two concurrent commodities (at least one was a drug) were available at varied prices. Sucrose and ethanol were found to be substitutes in rats; as the price of sucrose was increased, consumption of ethanol increased. In this review paper, Bickel et al. also showed that cigarettes and coffee were independent in human subjects; increasing the price of coffee had no effect on cigarette consumption (and vice versa).

Drug interactions also can be determined by assessing changes in demand for the drug available at escalating prices. Spiga et al. (2005) conducted an economic analysis of nicotine and methadone self-administration in humans. Demand for methadone was more inelastic when nicotine was concurrently available, while demand for nicotine was unaffected by the availability of methadone. Thus, methadone and nicotine were asymmetrical complements; while nicotine was a complement to methadone (methadone was a more effective reinforcer with nicotine available), the availability of methadone did not affect nicotine consumption.

Determining (a) how elasticity of drug demand changes with the availability of another drug, and (b) how an increase in the price of one drug alters consumption of an alternative drug can provide important information in determining the abuse liability of drugs and predicting behavior under circumstances in which individuals have access to multiple drugs. In phase 1 of the present experiment, the elasticity of demand for cocaine and the ultra-short acting μ-opioid agonist remifentanil was determined. Remifentanil was chosen because its short duration of action minimizes any direct effects of the drug in suppressing behavior as well as side effects (e.g., weight loss) commonly associated with the self-administration of longer-acting opiates such as heroin. In phase 2, cocaine and remifentanil were available concurrently. Across conditions, the price for either cocaine or remifentanil was increased while the price of the other drug was held constant to determine how demand for each drug changed as the result of the availability of the other drug and to determine the type of economic interaction that existed between them.

Materials and methods

Subjects

Six adult rhesus monkeys (Macaca mulatta), four males and two females, served in phase 1 of the experiment. Five of these monkeys (the exception being one female, Monkey 20996) served in phase 2. All monkeys had a history of drug self-administration which primarily included responding for various drugs, including cocaine and remifentanil, according to fixed-ratio (FR) schedules. Monkeys lived in the experimental chambers and were fed high-protein monkey biscuits twice daily, supplemented by fresh fruit once daily. Water was continuously available. Environmental enrichment was provided on a regular rotating basis. The University of Michigan is accredited by the American Association for the Accreditation of Laboratory Animal Care (AAALAC), and all procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council 1996) and approved by the University Committee on Care and Use of Animals.

Apparatus

Monkeys were housed in stainless steel cages that were 83.3 × 76.2 × 91.4 cm deep. A response panel containing three levers and three stimulus lights (one above each lever) was located on the wall to the left of the barred front door, approximately 10 cm from the front and 19 cm from the bottom of the cage. The stimulus lights, formed by three horizontally aligned circular openings, were 2.5 cm in diameter and spaced 1.5 cm apart. Each opening was covered with translucent plastic and was capable of being illuminated from behind by 5-W colored bulbs. Below each light was a lever capable of being pressed by approximately 0.10–0.15 N of force.

Each monkey wore a Teflon mesh jacket (Lomir, Quebec, Canada). A flexible stainless-steel tether was attached to the back of the jacket at one end and to the back wall of the cage at the other end. Monkeys were implanted with an indwelling intravenous catheter that passed subcutaneously from the implanted vein to an exit site on the monkey’s back. The catheter was fed through the tether to the outside of the rear of the cage where it was connected via a series of tubing to three stock solutions; one with cocaine, one with remifentanil, and one with saline. The tubing leading to each stock solution passed through its own infusion pump which delivered 0.2 ml of drug per second. As described in the methods below, each drug delivery lasted 5 s (totaling a 1-ml injection volume) and was followed by a 5-s (1 ml) saline infusion. Control of experimental sessions was provided by Med-PC (Med-Associates, Georgia, VT) interfacing and software installed on computers located in an adjacent room.

Phase 1: demand

General procedure

At the start of a session, the white cue light was turned on above the left lever. This light flashed (0.5 s on/off) until the monkey pressed the left lever, at which point the light stopped flashing and turned solid white. While the left lever was being held down, the light correlated with the available drug (cocaine = red; remifentanil = blue) turned on above the operative lever. Responses on the operative lever counted toward the response requirement only if the left lever was being held. If at any time during the completion of the response requirement the monkey stopped holding the left lever, the light above the operative drug lever was turned off and the white light above the left lever flashed until the left lever was once again pressed and held down. For example, if cocaine was available by pressing the center lever, holding down the left lever turned on the red light above the center lever and the flashing white light above the left lever turned solid white. The monkey then could complete the response requirement on the center lever and earn a cocaine infusion. If remifentanil was available on the center lever, holding down the left lever turned on the blue light above the center lever, and completion of the response requirement on the center lever resulted in a remifentanil infusion. If a drug was available by responding on the right lever, holding down the left lever turned on the appropriate color light above the right lever, and responses on the right lever resulted in a drug infusion. The response requirement on the operative lever did not reset if the monkey stopped holding the left lever. and presses to either the center or right lever in the absence of holding the left lever had no consequence. This procedure was employed to ensure that monkeys could not press multiple levers simultaneously during the choice procedure in phase 2 of the experiment. With this procedure, the monkey’s left hand was occupied at all times while the right hand was used to complete a response requirement.

When a drug infusion was earned, the white light above the left lever and the red or blue light above the operative drug lever turned off and the light above the left lever turned green for the duration of the infusion. The following procedure ensured that two drugs could be delivered within the same session during phase 2 (the choice phase) of the experiment: Immediately before the start of the session, the pump tubing and catheter (external to the monkey’s body) were filled with 1 ml of saline. Thereafter, during each reinforcer delivery, the drug pump operated during the first 5-s infusion, pushing 1 ml of saline into the monkey. The second 5-s infusion (the 1-ml saline flush) pushed 1 ml of drug into the monkey. This resulted in a 10-s total infusion during each reinforcer. From the monkey’s perspective, each drug delivery was preceded by a 1-ml saline delivery. While this constituted a 5-s delay to reinforcement, the delay was bridged by the green stimulus light and the sound of the infusion pump. Research in our laboratory has indicated that this infusion procedure maintains comparable levels of responding to traditional drug-infusion preparations. For the remainder of this report, a reference to a single drug infusion or reinforcer will encompass both the 5-s drug and 5-s saline infusions.

Infusions were followed by a 45-s timeout in which all stimuli were turned off. Sessions lasted 2 h and were conducted twice daily at the same times each day (once in the morning and once in the afternoon), 7 days/week.

Demand procedure

Demand functions were obtained for two doses of cocaine (0.01 and 0.03 mg/kg/inf) and two doses of remifentanil (0.0001 and 0.0003 mg/kg/inf) across conditions. Only Monkey 20996 experienced both doses of each drug. Monkeys 3506, 3956, and 3953 responded for 0.03 mg/kg/inf cocaine, 0.0003 mg/kg/inf remifentanil, and 0.0001 mg/kg/inf remifentanil, with the conditions conducted in this order. Monkeys 2084, 3511, and 20996 responded for 0.03 mg/kg/inf cocaine, 0.0003 mg/kg/inf remifentanil, and 0.01 mg/kg/inf cocaine in this order. In addition, Monkey 20996 responded for 0.0001 mg/kg/inf remifentanil in a subsequent condition.

In determining each demand function, the response requirement for earning a drug infusion was increased after every other session resulting in two consecutive sessions at each requirement or ratio. The ratios tested included 10, 32, 100, 320, 562, and 1,000 in this order. After all ratios were tested, four additional sessions were conducted; two with FR 10 in effect, and two with FR 100 in effect.

For five of the six monkeys in the experiment, two demand functions were obtained for each dose of each drug tested. In obtaining one demand function, the center lever was operative (in addition to the left lever which had to be held down to turn on the operative drug lever). In obtaining the other, the right lever was operative. This was done to determine if any lever biases resulted from the left-lever holding procedure. For example, one possibility was that the monkeys would respond at a higher rate or earn more infusions of the drug available from responses to the center lever because of its proximity to the left lever (which had to be pressed to complete a ratio). Because no biases of this kind were observed, the data were collapsed across lever-type for each dose/drug tested, resulting in four to six determinations of consumption at each ratio. For Monkey 20996, cocaine was available by completing the response requirement on the center lever and remifentanil was available by responding on the right lever, resulting in two to four determinations of consumption at each ratio.

Phase 2: choice

General procedure

In phase 2, 0.03 mg/kg/inf cocaine and 0.0003 mg/kg/inf remifentanil were available concurrently. Cocaine was available upon completion of a response requirement on the center lever, and remifentanil upon completion of a response requirement on the right lever. Therefore, holding the left lever turned on the red light above the center lever and the blue light above the right lever (in addition to causing the flashing white light above the left lever to become solid), signaling that both alternatives were available. The monkey then could choose either drug by completing the ratio requirement on the appropriate lever. The ratio on one lever was not reset as the result of emitting a response on the alternative lever or as the result of releasing the left lever. The ratio for both drugs was reset only upon the delivery of an infusion, before the start of the next ratio. An infusion of either cocaine or remifentanil was signaled by the illumination of the green light above the left lever, as in phase 1. All other details regarding the general procedure were the same as in phase 1.

Choice procedure

Throughout the conditions of phase 2, one drug was available at an FR 32 (the fixed-price alternative) while the ratio for the alternative drug was increased across sessions (the manipulated-price alternative) according to the sequence of ratios used in phase 1 (10, 32, 100, 320, 562, and 1,000). In the first condition, cocaine was the fixed-price alternative (available at FR 32) and remifentanil was the manipulated-price alternative. Unlike phase 1, the ratio for the manipulated-price alternative was increased every session rather than every other session. Upon completion of the final ratios (i.e., cocaine at FR 32 and remifentanil at FR 1,000), the ratio for cocaine was reduced to 10 and the ratio for remifentanil was reduced to 32. This constituted the first session of the subsequent condition, in which remifentanil became the fixed-price alternative and cocaine became the manipulated-price alternative. Each set of conditions (i.e., cocaine as the manipulated-price alternative followed by remifentanil as the manipulated-price alternative) was replicated four to five times; only the last three determinations were considered in the analysis of the data.

Data analysis

Demand curves

A normalized demand-curve analysis (Hursh and Winger 1995) was used to describe the demand curves shown in Figs. 1 and 3. The first step in this analysis was to calculate a normalized dose (q) for each dose of each drug tested. This normalized dose, representing 1% of total session drug intake at the lowest response requirement, was calculated as 100 divided by the number of infusions earned at FR 10. This normalized dose was then used in unit price (FR/q) and consumption (infusions × q) calculations. Using the graphing software Prism (GraphPad Software, San Diego, CA), the log of the normalized consumption was plotted as a function of the log of the normalized dose and the following exponential equation (Wade-Galuska et al. 2007; Winger et al. 2006) was used to fit a function to each curve:

$${\text{log }}Q = \log {\left( L \right)}{\text{e}}^{{ - aP}} $$
Fig. 1
figure 1

Demand functions showing normalized consumption of cocaine and remifentanil as a function of normalized price in phase 1. The table shows the elasticity parameter a, Pmax, and r 2 for each function

Full size image

In this equation, Q represents normalized consumption, P represents normalized price, and L is the y-intercept. The parameter L provides an estimate of consumption with unrestricted access and, in normalizing the demand functions, is equated at 100. The rate of decay, which is the measure of the elasticity of demand for the commodity, is a.

To determine if differences between demand functions were statistically significant (p < 0.05), Prism used an F ratio to assess differences in the elasticity parameter (a). The null hypothesis was that the demand functions did not differ. If the null hypothesis was accepted, a single function described both data sets being compared. If the null hypothesis was rejected, a separate function was drawn for each data set. Note, however, one exception: In Fig. 1, although the difference between the cocaine and remifentanil demand functions was not statistically significant, a separate curve was fit to each drug for purposes of comparison.

Response rate

In calculating the response rates shown in Tables 1 and 2, the number of responses was divided by an adjusted session time. This adjusted session time did not include time spent in the reinforcement cycle (time spent receiving an infusion or in timeout) and only included the time during which the left lever was being held down and access to the schedules was available.

Table 1 Infusions, response output, and response rate (responses per second) averaged across monkeys for each ratio and drug dose in phase 1
Full size table
Table 2 Infusions, response output, and response rate (responses per second) averaged across monkeys for each ratio and drug in phase 2
Full size table

Results

Phase 1: demand

Figure 1 shows normalized demand curves for cocaine (0.01 and 0.03 mg/kg/inf) and remifentanil (0.0001 and 0.0003 mg/kg/inf). Also shown on Fig. 1 are the elasticity parameter a, Pmax, and r 2 for each function. The exponential equation adequately described the data, with at least 96% of the variance accounted for by each function. While a visual inspection of the elasticity parameter and Pmax revealed that cocaine consumption decreased at a lower rate than remifentanil consumption and that the Pmax was greater for cocaine, there was no statistically significant difference in the reinforcing effectiveness of cocaine and remifentanil, as measured by a normalized demand-curve analysis.

Table 1 shows the group average number of infusions, response output, and response rate obtained at each ratio in the demand sequence. At the lowest response requirement (FR 10), the average number of infusions was higher for the lower doses of each drug. As shown in Fig. 1, the number of infusions subsequently decreased as the response requirement increased in all cases.

The average response output, or average number of responses emitted in the session, initially increased with increases in the response requirement. The response output peaked at FR 100 for the lower dose of each drug and at FR 320 for the higher dose, and then proceeded to decline in conjunction with the number of infusions earned. Response rates were higher for 0.01 mg/kg/inf cocaine relative to 0.03 mg/kg/inf cocaine at all ratios and higher for 0.0001 mg/kg/inf remifentanil relative to 0.0003 mg/kg/inf remifentanil at ratios lower than 562. Apart from this, there were no systematic changes in rates with increases in the response requirement across the doses and drugs tested.

Phase 2: choice

Figure 2 shows the average number of infusions of the fixed-price alternative as a function of the ratio required to earn the manipulated-price alternative. Generally speaking, as the price of remifentanil was increased (Fig. 2, left panel), consumption of remifentanil decreased while consumption of cocaine increased. As the price of cocaine increased (Fig. 2, right panel), consumption of cocaine decreased while consumption of remifentanil increased. Data points on the abscissa represent instances in which no infusions were earned. In both conditions, monkeys chose both drugs about equally when the manipulated-price alternative was available at a FR 100 and then consumed more of the fixed-price drug (available at FR 32) when the manipulated-price alternative was available at an FR 320. These results show that cocaine and remifentanil functioned as substitutes and that this interaction was symmetrical.

Fig. 2
figure 2

The average number of cocaine and remifentanil self-administrations obtained as a function of the response requirement for the manipulated-price alternative. The error bars represent standard error. The left panel shows the condition in which remifentanil was the manipulated-price alternative and cocaine was the fixed-price alternative available at a constant FR 32. The right panel shows the condition in which cocaine was the manipulated-price alternative and remifentanil was the fixed-price alternative available at a constant FR 32

Full size image

The group average number of infusions, response output, and response rate obtained at each ratio for both the fixed-price and manipulated-price alternatives is shown in Table 2. As shown in Fig. 2, the number of infusions of the manipulated-price alternative decreased with increases in the response requirement. Concurrent with this decrease was an increase in infusions of the fixed-price alternative. Note that the monkeys stopped earning infusions of the manipulated-price alternative at a lower response requirement (FR 562) than when the drug was the sole alternative in phase 1.

While response output initially increased as the response requirement was increased for the manipulated-price alternative, it then decreased with subsequent price increases. Coinciding with the increase in infusions of the fixed-price alternative, response output increased and then leveled off when most of the responding was allocated toward this alternative. Response rates followed a similar pattern; generally, rates of responding for the manipulated-price alternative decreased as the price increased, while rates for the fixed-price alternative increased and then leveled off when the highest ratios were in effect for the manipulated-price alternative.

In the top left graph of Fig. 3, the average number of infusions of 0.03 mg/kg/inf cocaine from phase 1 is plotted along with the number of infusions of the same dose of cocaine in phase 2 (when remifentanil was concurrently available at an FR 32) as a function of the response requirement. Likewise, the number of infusions of 0.0003 mg/kg/inf remifentanil from phase 1 is shown with the number of infusions of the same dose of remifentanil in phase 2 (when cocaine was concurrently available at an FR 32; bottom left graph).

Fig. 3
figure 3

The left column shows the average number of cocaine infusions in phase 1 and in phase 2 when remifentanil was available as the fixed-price alternative (top graph) and the average number of remifentanil infusions in phase 1 and in phase 2 when cocaine was available as the fixed-price alternative (bottom graph). The error bars in these graphs represent standard error. The right column shows normalized cocaine consumption (top graph) and remifentanil consumption (bottom graph) as a function of normalized price in phase 1 when the drug was available alone and in phase 2 when a fixed-price alternative was concurrently available. The tables show the elasticity parameter a, Pmax, and r 2 for each function

Full size image

In both conditions, the availability of an alternative at a fixed price resulted in decreased consumption of the drug available at escalating prices. This effect was more pronounced with cocaine, where infusions were decreased at an FR 100 compared to FR 320 for remifentanil, suggesting that the availability of remifentanil as the fixed-price alternative had more of an effect on cocaine self-administration than the availability of cocaine did on remifentanil self-administration.

The elasticity of the demand functions shown in the left column of Fig. 3 is quantified in the right column of this figure, where a normalized demand-curve analysis was conducted for each condition. The parameters describing the normalized functions are shown on the graphs. The variance again was well accounted for by the exponential demand equation with r 2 ranging from 0.89 to 0.99.When remifentanil was the fixed-price alternative and the price of cocaine was manipulated (top graph), the elasticity parameter for cocaine was increased (F 1,8 = 11.51, p < 0.01), showing that cocaine became more elastic compared to when it was available alone. Likewise, when cocaine was the fixed-price alternative and the price of remifentanil was increased (bottom graph), elasticity of demand for remifentanil was increased (F 1,8 = 21.66, p < 0.01) but the effect was slightly smaller in magnitude. Again, this suggests that remifentanil was a better substitute for cocaine than cocaine was for remifentanil.

Discussion

The current study found that the opiate remifentanil and the stimulant cocaine were similar in their reinforcing effectiveness as measured using a demand-curve analysis, with no significant difference in demand elasticity. Winger et al. (2006) recently conducted a demand-curve analysis of cocaine and remifentanil self-administration in rhesus monkeys. They found that in two of three monkeys, demand for cocaine was slightly less elastic than demand for remifentanil (there was no statistically significant difference in demand for the third monkey), suggesting that cocaine was a more effective reinforcer. It is not clear why the results of the present studied differed from those obtained by Winger et al. Research conducted in our laboratory has shown that both remifentanil and cocaine are very effective reinforcers and that they are similar in measures of demand elasticity. If a difference in reinforcing effectiveness exists between these drugs, it is likely to be a small one. Moreover, our laboratory has observed relatively large differences in elasticity of drug demand (independent of drug type) across individual monkeys. Such variability among monkeys may make it difficult in an analysis of group data to detect the small differences in elasticity between remifentanil and cocaine, such as those reported by Winger et al. Due to the fact that demand was assessed for just one dose of one of the drugs (which varied depending on the monkey) in all monkeys but one, a meaningful comparison of demand for cocaine and remifentanil could not be conducted in individual monkeys in the current experiment.

The economic interaction between cocaine and remifentanil was first assessed by measuring the change in consumption of the fixed-price alternative with increases in the price of the manipulated-price alternative (cross-price changes in demand; Hursh 1984). This analysis revealed that cocaine and remifentanil functioned as economic substitutes: As the price of the manipulated-drug alternative was increased, consumption of the fixed-price alternative increased. This interaction was for the most part symmetrical; that is, consumption of the fixed-price alternative increased regardless of the drug designated as the fixed-price versus the manipulated-price alternative. Note that when the price of the manipulated alternative was FR 32 (see Fig. 2), more injections were earned when the fixed-price alternative was remifentanil than when it was cocaine. This provides some support for the conclusion that remifentanil was a better substitute for cocaine than cocaine was for remifentanil. One question that this observation raises, however, is why consumption of one drug in relation to the other (under circumstances in which the response requirement was FR 32 for both alternatives) was not the same regardless of which drug was the manipulated-price versus the fixed-price alternative. Recall that the ratio was increased for the manipulated-price alternative in each session resulting in a procedure that was not designed for measuring steady-state performance. Using the present procedure, monkeys responded primarily for the manipulated-price alternative when it was available at a FR 10. As the response requirement for this alternative was further increased, they sampled and eventually allocated their behavior to the fixed-price alternative exclusively. Therefore, when both alternatives were available at an FR 32, response allocation may have been largely the result of the conditions in effect during the previous session, in which the manipulated-price alternative was available at a cheaper price (FR 10), and did not necessarily represent what would have been observed in an assessment of steady-state choice.

The substitutability of cocaine and remifentanil also was supported by decreased demand for the manipulated-price alternative that occurred as the result of the availability of the fixed-price alternative (Fig. 3). The fact that demand elasticity increased more for cocaine when remifentanil was the fixed-price alternative than it did for remifentanil when cocaine was the fixed-price alternative showed again that remifentanil was a better substitute. This finding is probably due to the doses compared. It is likely that assessing the economic interaction between 0.03 mg/kg/inf cocaine and a lower dose of remifentanil would make the relation between the two drugs more symmetrical. Likewise, comparing a lower dose of cocaine with 0.0003 mg/kg/inf remifentanil should result in remifentanil being an even better substitute for cocaine than what was observed in the present experiment.

Some experiments have investigated the effects of the magnitude of an alternative reinforcer on drug consumption (Campbell and Carroll 2000; Comer et al. 1998; Higgins et al. 1994; Nader and Woolverton 1992; Petry and Bickel 1999). While the findings of this research generally showed that increasing the magnitude of an alternative (e.g., nondrug) reinforcer results in a corresponding decrease in consumption of a drug reinforcer, few studies have quantified these effects using a demand-curve analysis. In one study employing demand-curve analysis, Petry and Bickel (1999) allowed opioid-dependent patients in a substance-abuse treatment center to choose between buprenorphine and money. The patients had ten choices available to them in a session. The amount of buprenorphine received for each choice was held constant at 3 mg while the amount of concurrent money ranged from $0.30 to $20 across five conditions. Consumption of buprenorphine significantly decreased with increases in money, although the decrease in buprenorphine consumption was not proportionate to increases in the amount of concurrently available money (i.e., demand for buprenorphine was inelastic). The results of this study and the aforementioned research support the notion that the degree of substitutability between remifentanil and cocaine could be altered by manipulating the dose of either drug. More research needs to be conducted involving parametric manipulations of the magnitude of both the manipulated- and fixed-price alternatives to better determine the economic interactions among drugs (and among drug and nondrug reinforcers) under varied environmental circumstances.

Polydrug use involving opiates and cocaine is a relatively common phenomenon (see Leri et al. 2003 for a review), so it is useful to understand how changes in economic variables of either or both types of these drugs affect their use. Petry and Bickel (1998) assessed the economic relations between a variety of commonly abused drugs, including the opiate heroin and cocaine. Human participants who were current or former opiate addicts were allowed to choose between hypothetical drug alternatives. In one condition, they chose between heroin and cocaine. As the price of heroin was increased, choice for cocaine increased significantly, however, the effect was observed in only 23% of the subjects. On the other hand, the benzodiazepine diazepam was a strong substitute for heroin. As discussed by Petry and Bickel, this is consistent with the opinion that benzodiazepines reduce the symptoms of opiate withdrawal. Their experiment exemplifies how, in natural settings, the effects of the price of one drug (e.g., with price defined by limited availability, increased monetary cost, or increased risk associated with purchasing a drug) is likely to result in increased intake of a concurrently available alternative drug and also represents how understanding of both the economic and pharmacological relations between commonly abused drugs can help to predict how addicts will choose among the drugs concurrently available to them.

In addition to allowing for better prediction of polydrug abuse, behavioral economic procedures such those employed in the present study may aid in evaluating the effectiveness of pharmacotherapies in which a treatment drug functions as a substitute for a drug of abuse (e.g., Bickel et al. 1993). For example, Johnson et al. (2004) showed that nicotine gum and denicotinized cigarettes functioned as a substitute for nicotine-containing cigarettes. However if both nicotine gum and denicotinized cigarettes were available as fixed-price alternatives to nicotine-containing cigarettes, subjects increased their consumption of denicotinized cigarettes as a function of increases in the price of the nicotine-containing cigarettes but did not consume nicotine gum under these circumstances. This suggests that denicotinized cigarettes may be a better substitute for nicotine-containing cigarettes and may constitute a nonpharmacological treatment for smoking that is more effective than the pharmacological treatment offered by nicotine gum. Likewise, it has been shown that the availability of both buprenorphine and a saccharin-based reinforcer produced a greater decrease in cocaine self-administration by rhesus monkeys than when either of these commodities were the sole alternative, suggesting that a pharmacological and nonpharmacological commodity together may provide an effective treatment for cocaine abuse (Comer et al. 1996; Rodefer et al. 1997).

Some studies have suggested that addicts take stimulants and opiates in combination (commonly referred to as a “speedball”) because the combination functions as a more effective reinforcer than one or both of the constituent drugs (Duvauchelle et al. 1998; Mattox et al. 1997; Ranaldi and Munn 1998; Ranaldi and Wise 2000; Rowlett et al. 2005; Ward et al. 2005). While there are a number of studies that contradict or question the generality of this finding (Mello et al. 1995; Rowlett and Woolverton 1997; Rowlett et al. 1998; Ward et al. 2005; Winger et al. 2006), the conclusion that the combination is a better reinforcer might lead one to predict that the economic relation between stimulants and opiates is complementary. Another possibility, however, is that that the economic relation between a stimulant and an opiate depends on the context. For example, an additional observation made by Petry and Bickel (1998) was that, as the price of heroin was increased, patients in treatment for heroin addiction initially made fewer (hypothetical) purchases of both heroin and a fixed-price cocaine alternative. As the price of heroin was increased further (resulting in further decreases in the units of heroin purchased), cocaine purchases significantly increased. This suggests that while heroin and cocaine initially functioned as complements, cocaine was a substitute for heroin at higher heroin prices. Primarily due to the procedures used in the present experiment, observations of complementary relations at low prices were impossible to assess. Subjects consumed the manipulated-price alternative to the exclusion of the fixed-price alternative, preventing decreases in the latter at initial increases in price. Future experiments should assess more completely the circumstances in which multiple types of economic relations between stimulants and opiates might be observed.