The epidemiologic case-crossover method is a powerful tool for research on suspected hazards of illegal drug use, the advantage being a subject-as-own-control approach that constrains stable individual-level susceptibility traits. Here, we use the case-crossover method to estimate the magnitude of excess occurrence of panic attacks during months of cocaine use vs. months of no cocaine use, motivated by a prior estimate that cocaine users have three-fold excess risk of panic attack. The self-report data on cocaine and panic are from assessments of a nationally representative sample of 1071 recent panic cases age 18 years or older identified as part of the National Household Surveys on Drug Abuse conducted in the United States during 1994-1997. Based on case-crossover estimates, cocaine use is associated with a three- to- four-fold excess occurrence of panic attack (estimated relative risk (RR) = 3.3, p = 0.049; 95% confidence interval: 1.0, 13.7). Year-by-year, the RR estimates from four independent yearly replicates (1994-1997) are 5.0, 2.0, 3.0, and 3.0. While there are several important limitations, this study adds new evidence about a previously reported suspected causal association linking cocaine use to occurrence of panic attacks, and illustrates advantages of the epidemiologic case-crossover approach and new directions in research on hazards of illegal drug use.