Hospitalization for heart failure (HF) is a clinical entity associated with high postdischarge morbidity and mortality, yet few therapies are available to improve outcomes in patients with this condition. In the past decade, large phase III studies of HF treatments have failed to demonstrate drug efficacy, safety, or both, despite encouraging results from preceding phase II trials. This Review is focused on this disconnect between the results of phase II and phase III trials of drugs for HF and discusses findings from five drug-development programs (for levosimendan, tezosentan, tolvaptan, rolofylline, and nesiritide) to shed light on common themes in clinical trials conducted in patients hospitalized for HF. In particular, the importance of selecting the 'right' patient population, drug, and clinical end points to optimize the trial design is discussed. Areas that require further investigation are highlighted and we suggest possible directions that will help to guide future clinical trials in these patients. Large, expensive phase III trials should not be initiated without adequate phase II evidence or on the basis of overly optimistic interpretation of phase II data. Additionally, drug development programs should be targeted not only to change short-term symptoms, but also to improve the postdischarge event rate.