Using crowdsourcing to study the differential effects of cross-drug withdrawal for cigarettes and opioids in a behavioral economic demand framework.

Abstract

Smoking rates among those who use prescribed or recreational opioids are significantly higher than the general population. Hypothesized neuropharmacological interactions between opioids and nicotine may contribute to this pattern of polysubstance use, especially during withdrawal. However, little research has examined how the withdrawal of one substance may affect the consumption of the other (i.e., cross-drug withdrawal effects). Behavioral economic demand tasks (e.g., hypothetical purchase tasks) can be used to quickly assess the value of a drug. Crowdsourcing can be a convenient tool to gain preliminary insight into different processes in substance valuation that may otherwise be impossible or prohibitively difficult to study. The purpose of the present study was to provide a preliminary examination of the effects of hypothetical withdrawal of cigarettes and opioids on the consumption of those drugs among polysubstance users. Amazon Mechanical Turk workers who reported daily smoking and at least monthly opioid use completed a series of hypothetical purchase tasks for doses of opioids and cigarettes under various withdrawal conditions. Sensitivity to the price of both drugs decreased when under withdrawal for either, indicating a higher drug value of cigarettes and opioids due to effects of cross-drug withdrawal. Nicotine and opioid dependence severity, impulsive choice, and riskiness were also positively related to drug purchasing. Public Health SignificanceThe results of this study suggest that opioid withdrawal may increase the value of cigarettes and therefore difficulty in quitting smoking, while nicotine withdrawal may similarly increase the value of opioids and difficulty in quitting opioid use. Results of this study could inform treatment development by explaining difficulties in maintaining abstinence that may arise as a function of cross-drug withdrawal effects between opioids and nicotine.

Notes on demand: Conceptual and empirical benefits of applying Rachlin’s discounting equation to demand data.

“…Then, coefficients from that model were used as starts for the fixed effect of the mixed-effects models. This is similar to what was done in Rzeszutek, Gipson-Reichardt, et al (2022) and Koffarnus et al (2022) as it aids the mixed-effects algorithm with finding optimal parameter estimates. Parameters to be estimated were treated as fixed effects, and a random effect for each individual was estimated by treating participants as a random effect with a symmetrical correlation matrix.…”

“…For the exponentiated equation, k was set to 1.222925, which was the log 10 span of the observed data included for analysis. To aid in model fitting, parameters for both equations were fit in log 10 space, as this decreases convergence issues and decreases the minimum convergence tolerance when fitting data sets for the exponentiated equation (Kaplan et al, 2021; Rzeszutek, Gipson-Reichardt, et al, 2022). The likely reason for this is that scaling the parameters in such a way allows for the algorithm to identify changes in α that may otherwise be missed due to small step sizes if unscaled.…”

Interest in and perceived effectiveness of contingency management among alcohol drinkers using behavioral economic purchase tasks.

“…We also specified experimental condition and participant-level random effects to allow for individual Q 0 and α values to be predicted across conditions. As we have done in previous research, parameters Q 0 and α were expressed and fit in log space to minimize issues of convergence (Rzeszutek et al, 2022). Specifically, mixed-effect models assume the distributions of random effects are normally distributed.…”

Hyperbolic discounting and exponentiated demand: Modeling demand for cigarettes in three dimensions

“…Because of the complexity of the data and models, fitting demand data using log 10 -scaled parameters allows for achieving lower tolerances and better fits to the data (Kaplan et al, 2021;Rzeszutek, Franck et al, 2022;Rzeszutek, Gipson-Reichardt, et al, 2022;Traxler et al, 2022). This does not transform the response variable but only the scale in which the parameter is expressed (and thus, optimized).…”