Background. In cross-sectional studies, cocaine users generally display elevated levels of selfreported and cognitive impulsivity. To what extent these impairments are stable v. variable markers of cocaine use disorder, and, thus, are pre-existing or drug-induced, has not yet been systematically investigated. Method. We conducted a longitudinal study with cocaine users who changed or maintained their consumption intensity, measuring self-reported impulsivity with the Barratt Impulsiveness Scale (BIS-11), and cognitive impulsivity with the Rapid Visual Processing task (RVP), Iowa Gambling task (IGT), and Delay Discounting task (DD) at baseline and at 1-year follow-up. We assessed 48 psychostimulantnaive controls and 19 cocaine users with decreased, 19 users with increased, and 19 users with unchanged cocaine intake after 1 year as confirmed by hair analysis. Results. Results of linear multilevel modelling showed significant group × time interactions for the BIS-11 total score and the IGT total card ratio. Increasers showed a trend for elevated scores, whereas decreasers exhibited reduced self-reported impulsivity scores within 1 year. Surprisingly, increasers' IGT performance was improved after 1 year, whereas decreasers' performance deteriorated. By contrast, neither RVP response bias B nor DD total score showed substantial group × time interactions. Importantly, BIS-11 and DD revealed strong test-retest reliabilities. Conclusion. Self-reported impulsivity (BIS-11) and decision-making impulsivity (IGT) covary with changing cocaine use, whereas response bias and delay discounting remain largely unaffected. Thus, self-reported impulsivity and gambling decision-making were strongly state-dependent in a stimulant-using population and may be suitable to monitor treatment success, whereas delay of gratification was confirmed as a potential endophenotype of stimulant addiction.