In this report, we summarize the strengths and problems of an observational data base approach to evaluating therapy and studying patient outcomes in long-term chronic disease. Because this approach includes a greater spectrum of patients than randomized clinical trials, it offers a definite advantage with regard to the elucidation of prognostic factors and the application of results to specific patients. The major difficulty with the observational data base approach is that the important prognostic factors must be known for treatment comparisons to be valid. Both the observational data base and randomized trial approaches are susceptible to criticism because the multiple comparisons and multiple experiments usually involved make the results of any one study not definitive. Either approach is useful in generating or confirming a hypothesis about particular subgroups. Regardless of the method used, proof that a particular therapy increases survival in any group or subgroup of patients with coronary artery disease usually requires confirmation by multiple studies. Finally, observational data base approach, because it capitalizes on data generated and paid for in the patient care process, offers the most feasible approach for evaluating whether changes in prognosis are occurring over time and whether such changes are independent of the mix of the patient population.