The influence of naltrexone on dynamic threat learning

The influence of naltrexone on dynamic threat learning - TELOS

Status

Recruiting

Phases

Unknown

Study type

Interventional

Source

DRKS

Registry ID

DRKS00037064

Enrollment

101

Registered

2025-05-28

Start date

2025-03-10

Completion date

Unknown

Last updated

2025-12-08

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Healthy participants

Interventions

Group 1: Naltrexone: Blockade of opioid receptors before dynamic threat learning (by oral administration of 50 mg Naltrexone in capsule form). Task description To assess dynamic threat learning and b

see Nassar et al., 2019, 2021

Satti et al., 2025). In the predator task, participants defend themselves against predators with unpredictably varying attack directions along a circle (ranging from 1 to 360 degrees). Variation in th

Eligibility

Sex/Gender

All

Age

18 Years to 40 Years

Inclusion criteria

Inclusion criteria: - Healthy participants - Right-handed - Fluent in German, proficient in English - Average alcohol consumption below 15 units per week (1 unit = 1 glass of wine, beer, or similar, and 1 cl of stronger alcoholic beverages)

Exclusion criteria

Exclusion criteria: 1. Past or present illnesses - Brain or mind (incl. anxiety disorders, depression, schizophrenia, alcohol, drug or medication dependence, neurological disorders other than occasional headaches) - Heart or circulation (incl. hypertension) - Blood - Lungs - Liver - Kidneys - Thyroid gland - Eyes (incl. glaucoma, color blindness and myopia of more than -5 dioptres) - Skin - Gastrointestinal tract - Metabolism 2. Cancer 3. Allergy to Naltrexone 4. Current medication and medication in a period of 2 months before the experiment (except contraceptives, for example the so-called “pill”). Non-prescription medications are only allowed if their intake dates back more than 1 week before participation in the study 5. Other serious health problems or current severe mental or physical stress 6. The presence of irremovable metal objects in or on the body 7. Suspected pregnancy 8. Consumption of illegal drugs

Design outcomes

Primary

Measure	Time frame
Model-agnostic threat learning hypotheses •We expect higher estimation errors (EE) and prediction errors (PEs) for naltrexone vs. placebo. •We expect worse performance (fewer hits vs. misses) for naltrexone vs. placebo. Model-based threat learning hypotheses General dynamic threat learning Our general expectation regarding dynamic threat learning is that participants use a combination of fixed and adaptive learning rates. We further hypothesize that participants adjust their fixed learning rate according to the noise level and the predator type. In the regression model, we test these hypotheses as follows: Combined learning-rate model: •Main effect PE (fixed LR): ß_PE>0 •Interaction PE * a (adaptive LR): ß_a>0 •Interaction PE * noise: ß_noise?0 •Interaction PE * predator type: ß_predatorType?0 Deconstructed learning-rate model: •Main effect PE (fixed LR): ß_PE>0 •Interaction PE * CPP: ß_CPP>0 •Interaction PE * RU: ß_RU>0 •Interaction PE * noise: ß_noise?0 •Interaction PE * predator type: ß_predatorType?0 Pharmacological manipulation Regarding our pharmacological manipulation, our general hypothesis is that naltrexone leads to faster but less adaptive threat learning relative to placebo, measured as lower adaptive learning rates and higher fixed learning rates for naltrexone compared to placebo. Using the coefficients from the regression model, we test these hypotheses as follows: Combined learning-rate model: •Main effect PE (fixed LR): ß_Naltrexone>ß_Placebo •Interaction PE * a (adaptive LR): ß_Naltrexoneß_Placebo •Interaction PE * CPP: ß_Naltrexone ßPlacebo. Direct model fitting hypotheses While our overarching hypothesis is that opioid receptor blockade inflates fixed learning rates and subsequently reduces adaptive learning, we would like to gain more specific insight into what drives inflated learning rates. To investigate this, we will use direct fitting of the reduced Bayesian model to participant data to estimate free parameters of interest: uncert	—

Measure

Time frame

Model-agnostic threat learning hypotheses •We expect higher estimation errors (EE) and prediction errors (PEs) for naltrexone vs. placebo. •We expect worse performance (fewer hits vs. misses) for naltrexone vs. placebo. Model-based threat learning hypotheses General dynamic threat learning Our general expectation regarding dynamic threat learning is that participants use a combination of fixed and adaptive learning rates. We further hypothesize that participants adjust their fixed learning rate according to the noise level and the predator type. In the regression model, we test these hypotheses as follows: Combined learning-rate model: •Main effect PE (fixed LR): ß_PE>0 •Interaction PE * a (adaptive LR): ß_a>0 •Interaction PE * noise: ß_noise?0 •Interaction PE * predator type: ß_predatorType?0 Deconstructed learning-rate model: •Main effect PE (fixed LR): ß_PE>0 •Interaction PE * CPP: ß_CPP>0 •Interaction PE * RU: ß_RU>0 •Interaction PE * noise: ß_noise?0 •Interaction PE * predator type: ß_predatorType?0 Pharmacological manipulation Regarding our pharmacological manipulation, our general hypothesis is that naltrexone leads to faster but less adaptive threat learning relative to placebo, measured as lower adaptive learning rates and higher fixed learning rates for naltrexone compared to placebo. Using the coefficients from the regression model, we test these hypotheses as follows: Combined learning-rate model: •Main effect PE (fixed LR): ß_Naltrexone>ß_Placebo •Interaction PE * a (adaptive LR): ß_Naltrexoneß_Placebo •Interaction PE * CPP: ß_Naltrexone ßPlacebo. Direct model fitting hypotheses While our overarching hypothesis is that opioid receptor blockade inflates fixed learning rates and subsequently reduces adaptive learning, we would like to gain more specific insight into what drives inflated learning rates. To investigate this, we will use direct fitting of the reduced Bayesian model to participant data to estimate free parameters of interest: uncert

—

Secondary

Measure	Time frame
Skin conductance response hypotheses We will test the relationship between prediction errors (PE) and skin conductance responses (SCR) between drug groups (Stemerding et al., 2022). Comparison with common confetti cannon task Apart from learning in the predator task (which is the main focus of this study), we will also let participants perform a very similar predictive inference task that contains no threat of shock (confetti cannon task). We expect that opioid blockade primarily affects threat learning, rather than learning in general. Therefore, we hypothesize that there is no difference between naltrexone and placebo groups on learning in the confetti cannon task.	—

Measure

Time frame

Skin conductance response hypotheses We will test the relationship between prediction errors (PE) and skin conductance responses (SCR) between drug groups (Stemerding et al., 2022). Comparison with common confetti cannon task Apart from learning in the predator task (which is the main focus of this study), we will also let participants perform a very similar predictive inference task that contains no threat of shock (confetti cannon task). We expect that opioid blockade primarily affects threat learning, rather than learning in general. Therefore, we hypothesize that there is no difference between naltrexone and placebo groups on learning in the confetti cannon task.

—

Countries

Germany

Contacts

Public ContactFelix Klaassen

Institut für Systemische Neurowissenschaften

f.klaassen@uke.de+49 (0)40 7410 59711

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Feb 4, 2026