Objective: To examine trends in the prevalence of problematic cocaine use in a large city and describe the problems encountered when estimating these trends using capture-recapture techniques based on secondary data Methods: We used clinical data on accident and emergency department episodes at four university hospitals in Barcelona (Spain) between 1999 and 2008 (3 capture periods per year). Users were categorized into two subgroups depending on concomitant heroin use (cocaine plus heroin, cocaine without heroin). Results: The mean age of users was 34 years and 25% were women (2008). The mean number of episodes per user differed between drug subgroups and over time. The estimated total number of cocaine users increased from 6,028 (95% confidence interval [95%CI]: 4,086-9,327) in 1999 to 22,640 (95%CI: 14,001-37,500) in 2006, but decreased thereafter. The prevalence of problematic use of cocaine plus heroin was stable throughout the study. Thus, trends in the prevalence of problematic cocaine use differed depending on concomitant heroin use. Conclusion: Our results are consistent with those provided by health surveys and treatment registries. They also highlight the pitfalls of applying the capture-recapture approach to secondary data, and the need for a better understanding of how information is collected and changes over time.