Economic choice between remifentanil and food in squirrel monkeys

Self-administration studies have long served as the cornerstone of preclinical research for understanding drug reinforcement and searching for improved addiction therapies. Traditionally, the self-administration procedure has utilized rate-dependent measures of drug consumption, in the absence of alternative reinforcers. A behavioral approach with real-world parallels is one in which the consumption of drugs comes at the expense of alternative nondrug rewards, given that the choice to use drugs can result in the loss of a job, good family relations, and/or good health. In human laboratory studies, cocaine users choose sufficiently high monetary rewards over the option to self-administer cocaine, demonstrating that drug users are able to minimize their drug intake when behavioral alternatives are available. Contingency management, the therapeutic extension of this, is an effective nonpharmacological approach for the treatment of substance use disorders.

Behavioral economic theory and concurrent choice procedures in clinical settings have provided a framework for evaluating how individual variability in severity of addiction relates to distinct factors that could influence subjective valuation as reflected in choice between drug and nondrug rewards. The relationships between addiction severity or recovery success with differences in sensitivity to the effect of drug-associated stimuli, drug devaluation, or punishment on drug choice, have enabled conceptual conclusions about mechanisms of vulnerability and treatment efficacy. Though addiction severity or treatment success is not being modeled herein, those clinical applications highlight the value of an economic decision-making approach.

Nonhuman primate studies using macaques benefit from extensive prior use in behavioral pharmacology and neuroeconomics of reward-driven choice, but their complexities and costs bias toward small-N, within-subject designs. Rodent models are well-suited for between-group comparisons, and squirrel monkeys could provide a compromise with increased statistical power using nonhuman primate subjects. Establishing a statistically well-powered sample in nonhuman primates increases potential to explore novel addiction therapies, benefiting from closer similarity to humans. The present work establishes a squirrel monkey model of economic choice between the short-lived opioid remifentanil and a palatable food reward (diluted sweetened condensed milk) using a touchscreen apparatus varying reward magnitudes across trials, and evaluates pharmacological manipulations in terms of choice allocation rather than response rates. The aims were to: (1) establish a behavioral choice paradigm affording many trials and a wide parametric range; (2) demonstrate the economic nature of choices; and (3) evaluate the impact of pharmacological manipulation of remifentanil reward (μ-opioid receptor agonist and antagonist) on choice behavior.

Design and subjects: Ten adult male squirrel monkeys (Saimiri sciureus; 700–1100 g), individually housed under controlled conditions (12-h light/dark, lights on 7:00 a.m.), with ad libitum water and daily high-protein monkey diet plus fresh produce. Sessions ran ~5 days/week. All procedures were approved by the NIDA IRP IACUC.

Apparatus: Animals were seated in custom acrylic chairs within ventilated, sound-attenuating chambers. A 15-inch Elo touchscreen faced the subject. A syringe pump delivered 30% sweetened condensed milk to a drinking well beneath the screen; a second pump delivered intravenous remifentanil HCl (0.5 μg/ml in saline). Tasks were programmed and data collected with E-Prime Professional 3.0.

Touchscreen pretraining: Using milk reward only, monkeys learned to choose between two stimuli where 1–4 red circles within a white square indicated milk magnitude (75 μl/kg per circle). All learned to choose the larger reward (>80%).

Surgery: After milk training, a catheter (PVC, inner diameter 0.38 mm; outer 0.89 mm) was implanted in a jugular or femoral vein, exiting subcutaneously on the back, under isoflurane anesthesia and aseptic technique. Post-op recovery ≥1 week. Animals wore protective jackets. Catheters were flushed daily (Mon–Fri) and heparin locked on Fridays.

Remifentanil stimulus training: A new drug stimulus used 1–4 green triangles within a white square. Each triangle corresponded to an intravenous infusion unit dose of remifentanil (0.08–0.32 μg/kg/infusion). Initial sessions interleaved milk choice trials with single-stimulus drug trials; correct touches yielded the associated infusion. Next, drug choice trials (two drug stimuli differing in magnitude) replaced single-stimulus trials until animals discriminated larger drug rewards (>80% for three consecutive sessions). Finally, drug–milk choice trials were introduced, with one drug and one milk stimulus presented simultaneously.

Final task structure: Each session comprised 108 trials with a 60 s intertrial interval after reinforcement. There were 36 milk vs milk trials (ratios 2:3, 1:2, 1:3, 1:4) interleaved, followed by two consecutive remifentanil vs milk trials to limit drug accumulation and provide a drug-neutral trial type. For drug vs milk trials, the number of symbols for each option varied (1–4 per domain). Each drug:milk offer ratio was presented eight times, counterbalanced for left/right placement. Reward contrast was defined as (number of drug symbols − number of milk symbols) / (number of drug symbols + number of milk symbols), yielding values from −0.60 (DM 1:4) to +0.60 (DM 4:1). Choice allocation was plotted as % drug choice vs reward contrast. A logistic regression on this within-session dataset provided the indifference value (IndV), the reward contrast at which drug and milk choices were equally likely. In 17 of 377 curve fits where data did not span a full sigmoid (e.g., milk satiation or saline substitution), IndV estimates were bounded to the natural limits (±1). After initial baselines, in five animals the unit milk per symbol was doubled (from 0.075 ml/kg/circle to 0.15 ml/circle) to improve symmetry in choice distribution.

Testing schedule: Manipulations began on Mondays and continued through the week; analyses used the mean of the final three days (Wed–Fri). Each manipulation week was separated by at least one baseline week for contemporaneous comparison. Day-by-day shifts typically appeared on the first day (Supplementary figures).

Experimental manipulations:

• Reward devaluation: Milk satiation (free access to milk pre-session, up to 60 ml; typical consumption 30–60 ml vs ~20 ml during baseline sessions). Remifentanil devaluation by saline substitution (drug pump delivered saline).
• Pharmacological pretreatments: Naltrexone (0.1 or 0.3 mg/kg, i.v.) given 5 min pre-session via catheter; morphine sulfate (1.0 mg/kg, i.m.) given 15 min pre-session.

Response latency: Defined as time from stimulus onset to touch. To reduce outlier influence, per-session mean latencies were computed after trimming data above the 90th percentile of the response time distribution. Latencies were analyzed by reward contrast and treatment.

Exclusion criteria: Data from an animal were excluded for a treatment week if the subject failed to complete ≥3 sessions with ≥54 trials within a 3 h maximum session duration, typically due to nonspecific behavioral suppression (e.g., sedation). Occasionally, a week was missed due to catheter replacement.

Statistical analysis: Paired t-tests compared each manipulation to its preceding baseline (α=0.05). Mixed-model ANOVA with Holm–Sidak corrections compared treatment effects (e.g., naltrexone doses vs saline substitution). Response latencies across contrasts were assessed with one-way repeated-measures ANOVA (Holm–Sidak corrections). Baseline stability assessed via partial correlation between baseline IndV and session number, controlling for subject.

• Choice function: Across 10 monkeys, % drug choice vs reward contrast followed a sigmoidal function fit by logistic regression, enabling estimation of the indifference value (IndV).
• Response latencies: One-way RM ANOVA showed latencies varied with reward contrast (F(1.566,14.10)=4.82, p=0.032); drug-dominant contrasts had faster responses than the 1:1 contrast (Holm–Sidak p<0.01).
• Baseline stability: Partial correlation showed IndV was stable over months of testing with no significant relationship to cumulative remifentanil exposure/session number (partial r=−0.0880, p=0.19), controlling for subject.
• Milk satiation (n=9): Increased drug-choice probability and decreased IndV relative to baseline (paired t(8)=3.877, p=0.0047). Milk consumed during choice was reduced to 50±10% of baseline.
• Saline substitution (n=10): Decreased drug choice and increased IndV (t(9)=10.31, p≈1×10^-5). Despite prior experience, animals still selected the drug-associated stimulus (saline) at the highest drug offers, consistent with conditioned cue effects.
• Naltrexone pretreatment: 0.1 mg/kg i.v. increased IndV (t(9)=5.365, p=0.0011); 0.3 mg/kg i.v. produced a greater reduction in drug choice than 0.1 mg/kg. Mixed-model ANOVA showed a main effect of treatment (F(1.655,14.07)=18.18, p=0.0002); 0.3>0.1 mg/kg (Holm–Sidak p=0.0145); both naltrexone doses produced smaller effects than saline substitution (p=0.0024 and p=0.0142 for 0.1 and 0.3 mg/kg vs saline substitution, respectively).
• Morphine pretreatment (1.0 mg/kg i.m., n=5): No significant effect on choice allocation (t(4)=0.6697, p=0.5398), although it was behaviorally active (longer response times and reduced overall responding).
• Summary of shifts: Positive ΔIndV indicates decreased drug choice; saline substitution and both naltrexone doses yielded significant positive ΔIndV; milk satiation yielded significant negative ΔIndV.

The study establishes a high-throughput economic choice paradigm in squirrel monkeys, enabling goal-directed decisions between intravenous remifentanil and a palatable milk reward across a broad range of parametric offers within single sessions. Using the novel reward contrast metric and logistic regression, the indifference value (IndV) serves as a sensitive within-session measure of relative valuation. Choices were economic and flexible: selective devaluation of one reward shifted choice toward the alternative, demonstrating goal-directed control rather than habitual responding. Rapid remifentanil clearance minimizes pharmacokinetic carryover, allowing many independent trials and randomized left/right placement to reduce side biases.

Behavioral analyses showed faster responses for drug-heavy offers, indicating higher motivation for drug-dominant options in this species, differing from some macaque findings. Importantly, animals did not simply follow the number of stimulus elements; they evaluated both offers. For example, smaller drug doses at 1:1 could be chosen more than larger drug doses at 2:3 when the milk counteroffer increased, aligning with prior findings that drug choice depends on the nondrug counteroffer value.

Pharmacological manipulations altered relative valuation as predicted: milk satiation shifted choices toward drug (decreased IndV), while remifentanil devaluation via saline substitution or μ-opioid antagonism (naltrexone) shifted choices toward milk (increased IndV). Naltrexone effects were dose-dependent but did not match saline substitution, potentially reflecting antagonist effects on palatable food valuation or transient surmounting by remifentanil’s peak levels—distinctions that saline substitution cannot produce. Morphine pretreatment did not shift choice, consistent with prior findings that agonists alter choice primarily in dependence contexts, suggesting that modeling negative reinforcement may require incorporating physical dependence.

Overall, the paradigm provides a quantitative framework to detect treatment-induced shifts in opioid versus food valuation, suitable for evaluating candidate addiction therapies and probing neural circuits of drug choice in nonhuman primates with greater statistical power than typical macaque studies.

This work develops and validates an economic choice paradigm in squirrel monkeys that quantifies preference between remifentanil and palatable food using a within-session reward contrast metric and indifference value (IndV). The procedure yields robust sigmoidal choice functions, stable baselines, and sensitivity to selective reward devaluation and pharmacological manipulations (antagonist but not agonist). Advantages include rapid drug clearance enabling many independent trials, randomized offers to discourage habits, and a single-session metric for preference shifts. This model supports high-throughput assessment of pharmacotherapies and provides a platform for future systems neuroscience studies, including circuit-specific manipulations (e.g., optogenetic or chemogenetic approaches) to dissect mechanisms of opioid versus natural reward choice. Future research should extend the paradigm to dependence models, explore broader drug classes and doses, assess potential interactions of antagonists with palatable food valuation, and examine sex and individual differences.

• The study used nondependent monkeys; results may differ under opioid dependence/withdrawal conditions that can increase opioid choice via negative reinforcement mechanisms.
• Naltrexone may affect palatable food valuation (e.g., sweetness palatability), potentially confounding interpretation of reduced drug choice; no robust changes were observed in milk–milk trials, but subtle effects cannot be excluded.
• In some extreme manipulations (milk satiation, saline substitution), logistic fits did not span full sigmoids, requiring IndV bounding at ±1 for a minority of sessions (17/377), which may limit precision at extremes.
• Occasional exclusions occurred due to nonspecific behavioral suppression (e.g., sedation) or catheter issues; these could bias week-level analyses if systematic.
• The sample comprised only adult male squirrel monkeys; generalizability across sexes and species remains to be established.
• Morphine pretreatment reduced responding and increased latencies, indicating potential motoric effects even without shifting choice, which can complicate interpretation of some pharmacological manipulations.