Polysubstance use is prevalent among individuals on opioid agonist treatment (OAT), yet past studies have focused primarily on distinct substances and their association with OAT retention. Data was collected from two prospective cohorts between 2005 and 2020 in Vancouver, Canada. Among 13,596 visits contributed by 1445 participants receiving OAT, we employed repeated measures latent class analysis using seven indicators and identified four longitudinal substance use classes. Using marginal structural Cox modeling, we found that compared to the primarily crack use class, the two opioid and stimulant use classes carried a higher risk of OAT discontinuation, while the primarily cannabis and crack use class had a lower OAT discontinuation risk. Our findings highlight the need for integrated treatment strategies to manage the co-use of opioids and stimulants during receipt of OAT and suggest future research should explore the potential of cannabis as a harm reduction strategy or adjunctive treatment to OAT.