Assessing measurement bias in substance withdrawal symptoms attributable to childhood adversity and genetic liability.

Abstract

Substance withdrawal results from the abrupt cessation or reduction of substance use in physically dependent individuals. Diagnosis of withdrawal requires endorsing multiple substance-specific symptoms, implicitly assuming that all symptoms are equally indicative of underlying withdrawal severity. However, some criteria may be more informative than others, and their endorsement may be impacted by genetic and environmental factors. We examined differential item functioning (DIF) of tobacco, alcohol, and opioid withdrawal symptoms as a function of substance use disorder (SUD) polygenic scores (PGS) and adverse childhood events (ACEs) in 8431 individuals (3987 of African-like ancestry and 4444 of European-like ancestry) from a large, multi-site genetic study of SUDs. Item response theory models were used to estimate difficulty and discrimination parameters for each withdrawal symptom, and a moderated nonlinear factor analysis was conducted to examine whether these parameters differed as a function of PGS and ACEs. Withdrawal symptoms varied in difficulty and discrimination, indicating that symptoms differ in how well they reflect underlying severity. We identified both uniform (i.e., differences in difficulty) and nonuniform (i.e., differences in discrimination) DIF by ACEs and PGS. Symptom difficulty was generally higher among individuals with greater ACEs exposure, while individuals with higher PGS were more likely to endorse certain affective withdrawal symptoms (e.g., restlessness and depressed mood) at the same level of withdrawal severity. These findings suggest that withdrawal symptom counts may not reflect equivalent levels of underlying severity across individuals, potentially leading to over- or underestimation of withdrawal severity depending on patient characteristics.