A neuroeconomic signature of opioid craving: How fluctuations in craving bias drug-related and nondrug-related value

How does craving bias decisions to pursue drugs over other valuable, and healthier, alternatives in addiction? To address this question, we measured the in-the-moment economic decisions of people with opioid use disorder as they experienced craving, shortly after receiving their scheduled opioid maintenance medication and ~24 h later. We found that higher cravers had higher drug-related valuation, and that moments of higher craving within-person also led to higher drug-related valuation. When experiencing increased opioid craving, participants were willing to pay more for personalized consumer items and foods more closely related to their drug use, but not for alternative “nondrug-related” but equally desirable options. This selective increase in value with craving was greater when the drug-related options were offered in higher quantities and was separable from the effects of other fluctuating psychological states like negative mood. These findings suggest that craving narrows and focuses economic motivation toward the object of craving by selectively and multiplicatively amplifying perceived value along a “drug relatedness” dimension.

Opioid craving and subjective valuation S3 Tastiness (snack foods only): In this part, we are going to show you all of the snacks, one at a time. We ask that you think about the TASTINESS of each snack to you, without regard for its healthiness, from ''Not at all tasty'' to ''Extremely tasty''.
Healthiness (snack foods only): In this part, we are going to show you all of the snacks, one at a time. We ask that you think about the HEALTHINESS of each snack, without regard for its tastiness, from ''Not at all healthy'' to ''Extremely healthy''.

Choice set algorithm (Day 1)
The subset of 12 consumer items/snack foods used in the task completed on Days 2 and 3 was individualized for each participant from a broader set of 40 items that were identical for all (see main text Fig. 2A and Fig. S1). These 40 items were selected by the investigators, in part informed by an informal pre-study survey conducted in the same methadone treatment program that inquired about everyday items that patients associated with opioid consumption and that they would pay any amount of money to purchase (even pennies). From this informal survey, we obtained items that had an evident linkage to opioid use such as hypodermic needle syringes. We also obtained consistent reports of items that may not be usually found in the list of stimuli of traditional cue reactivity studies, but that participants associated with their opioid consumption (e.g., 8 oz bottles of Poland Spring water). Such items may not traditionally be considered drugrelated (compared to others like syringes and lighters) and were as such included in the list of 40.
The individualization of the choice subset of 12 items was achieved by means of an algorithm that randomly and iteratively selected 12 items (6 snack food and 6 non-food items) from the broader list of 40 and computed the correlation between their drug-relatedness and their desirability as rated by that participant. The algorithm identified the set of 12 items with maximal standard deviation for drug relatedness and then computed the partial correlation between drug relatedness and desirability across that set controlling for item type (food vs. non-food). The algorithm also ran a paired t-test for both ratings by item type. Unless the significance level of the partial correlation and the t-test both exceeded 0.25 (critical P>0.25), equivalent to a critical correlation coefficient R<|0.2058|, the algorithm would continue through the remaining combinations of 12-item sets removing the current set from consideration. The fixed effect correlation between drug relatedness and general desirability for the final 12-item sets across all participants (P=0.267) confirmed orthogonalization procedure was successful.

In-task ratings (Days 2 and 3)
Opioid craving and subjective valuation S4 On Days 2 and 3, participants completed the willingness-to-pay task. The task included three trial types: bid trials, desire rating trials, and current mood/craving rating trials. In bid trials, participants were asked to indicate their willingness-to-pay in the current moment for each item in their 12-item choice set, offered one at a time in one of four quantities. In desire rating trials, participants were asked to indicate their desire in the current moment for each item, again offered in different quantities. In mood rating trials, participants were asked to report on their current feelings of boredom, stress, and happiness. In craving rating trials, current desire for heroin and methadone were used to evaluate overall opioid craving. The different mood and craving rating trials were identified by a picture cue corresponding to the rating type. All ratings scales were unit-less and used a continuous slider bar that recorded the cursor position in high-resolution (0.1% increments relative to the total bar length). All ratings data were transformed to 0-1 (reflecting the final position of the cursor relative to the max possible position). The specific instructions given to participants for these ratings are detailed below. […] On days 2 and 3 you will get bonus rewards. These rewards will be determined by your responses on bid trials. You are going to place bids for many items, and possibly more than once for each. You should always place yours bids based on how you feel in the current moment.
This because the task can end early. It can last anywhere from a few minutes to over an hour.
When the task ends, we are going to pick only ONE of your recent bids. We will then compare your bid for the item(s) shown on that one decision with a randomly chosen ''selling price''. If your bid is higher, you will get to buy the item(s) at the lower selling price and keep the change.

Your bonus will then be the item(s) plus the left over $. If your bid is lower, you won't get to buy the item(s). Your bonus will then be no item(s) plus $15.
Desire rating trials: In these trials, you will see an item and rate how much would like to have this item RIGHT  In the main data analyses, heroin and methadone craving ratings were averaged to obtain a single assessment of opioid craving. Given that each task block (lasting on average for 12-15 minutes) had only a single randomly interspersed craving rating per drug type and a single mood/affective state rating per mood/affective state type (see main text, Fig. 2B), grouping these into two composite measures provided a less noisy estimate of the aggregate level state of the person across the duration of the entire block. When we re-analyze the data separating the heroin and methadone craving ratings, we find (1) qualitatively similar results with heroin and methadone craving as separate measures, but (2) comparably worse overall model performance. Table S3, like the composite opioid craving measure (see Table 1), both heroin and methadone craving interact significantly with drug relatedness to influence valuation. This is observed when the two are used as predictors in separate models or in the same model, and when additionally controlling for study day (shortly after or 24 hours since last methadone dose).

As shown in
However, the relative contribution of between-person versus within-person effects differs somewhat between the two (with within-person effects having somewhat larger effect for heroin craving). Importantly, these models, which break out craving into heroin and methadone craving, do not provide a better fit to the data over the composite opioid craving model [and if anything, they provide worse fits, especially when accounting for study day: ∆BIC of -14.7 to -36.1 in favor of the composite model]. For parsimony, we therefore focused our main analyses on the composite opioid craving measure.

Bid realization for bonus payment (Days 2 and 3)
At the conclusion of the willingness-to-pay task at each task session (Days 2 and 3), a single recent (from the last completed block) bid trial was selected for realization. To determine a participant's bonus, we implemented a standard Becker-DeGroot-Marschak (BDM) auction whereby the selected bid b was compared against a randomly drawn selling price p. To determine p, chips numbered from $0 to $15 in $0.02 increments were drawn from a bag. If b ≥ p Opioid craving and subjective valuation S6 (the price on the chips), the item(s) offered on the selected trial could be purchased for the price p. If b < p, however, the item(s) could not be purchased. Thus, in the first case the bonus consisted of the item(s) offered plus any unspent money, and in the second case, no item(s) plus the full endowment. The BDM procedure is widely used in laboratory economic studies because it elicits a participant's "true" subjective value for each item, and in our case for a given moment in time. Its design ensures the best strategy for participants is to report their maximum willingness to-pay price, without over-or under-estimating this amount.
Opioid craving and subjective valuation S7 On Day 1, participants rated each of 40 common consumer items and snack foods on their subjective drug relatedness and general desirability. These ratings were used to sub-select a 12item choice set for each participant for the task completed on Days 2 and 3 (see main text and Fig. 2A). Shown here are the number of participants who had each item from the initial 40-item fixed set in their personalized 12-item sub-set (note that one participant inadvertently completed Opioid craving and subjective valuation S8 the task with 11 instead of 12 items, leading to 347 unique participant-item combinations: 28×12 + 1×11). Each item was represented in at least one participant's choice set. However, this could be for different reasons: the same item could belong to a choice set because it was rated as low or high on drug relatedness by a given participant. While for a few items there was consensus that the item was "especially" drug-related (e.g., rated as being >75% of the scale's max in drug relatedness by most participants), for most others there was high degree of idiosyncrasy in these subjective judgements.   . 3). In-task craving ratings correlate with the 14-item HCQ-Now multidimensional baseline assessment of craving (P=1.1×10 -14 ), suggesting this single-item measure well-approximates broader definitions of craving.