VOLUME 69 NUMBER 3 MARCH 2001 novel therapies. 3 In the initial investigations of therapeutic candidates in humans, biomarkers can provide a basis for the selection of lead compounds for phase 3 clinical trials. 3-5 Biomarkers contribute knowledge about clinical pharmacology and provide a basis for the designing of clinical trials that expeditiously and definitively evaluate safety and efficacy. Biomarkers provide information for guidance in dosing and minimize interindividual variation in response. For example, rapid clearance of 99mTc-sestamibi, a substrate for P-glycoprotein that is associated with multidrug resistance, has been shown to predict lack of tumor response to adjuvant chemotherapy in some forms of breast cancer. 6 Biomarkers that represent highly sensitive and specific indicators of disease pathways have been used as substitutes for outcomes in clinical trials when evidence indicates that they predict clinical risk or benefit. Assessment of benefit and risk must be the goal of the development plan for all therapeutic interventions. The most reliable way to assess the clinical impact of a therapeutic intervention (eg, drug, device, surgery, vaccine, biologic agent, and behavioral modality) is through its effect on a well-defined clinical endpoint such as survival, myocardial infarction, stroke, bone fracture, or recurrence of cancer. However, this standard may be impractical for the evaluation of some long-term disease therapies because long periods are required for these clinical endpoints to be achieved and trials with large numbers of patients are needed for their evaluation. Biomarkers that can be reliable substitutes for clinical responses have the potential to improve the efficiency of clinical trials in which long-term disease interventions are evaluated. Biomarkers can be substituted reliably for clinical responses. In other cases, reliable biomarkers have been used as substitutes for clinical endpoints in decision-making situations when a devastating clinical outcome, such as death, represents an ethical dilemma. In recent years, this point has been illustrated in pharmacokinetic studies in which biomarkers such as human immunodeficiency virus (HIV) plasma viral load and CD4 cell counts were used as substitutes for clinical outcomes (for example, death and occurrence of opportunistic infections) in the evaluation of antiviral agents in patients with HIV infection. 7-9 Accompanying the increased knowledge about biomarkers is an increased appeal of the use of biomarkers as substitutes for clinical outcomes in other diseases. This interest is countered by concerns about the inherent limitations of biomarkers that have been shown by some dramatic failures. 10, 11 Among the most notable of these failures is the demonstration by the Cardiac Arrhythmia Suppression Trial (CAST) that suppression of ventricular arrhythmias is not a valid substitute for sudden death after myocardial infarctions. 12 This showed that considerable skepticism about conventional wisdom should accompany the adoption of biomarkers as a substitute for outcomes as the basis for approval of a novel therapy. When biomarkers are intended to be the basis for provisional evaluation and regulatory approval of a drug, they also must be a component of a predetermined strategy that recognizes inability of biomarkers to serve as final proof of clinical efficacy or long-term safety. One strategy for the requirement of a systematic approach with phase 4 trials has recently been proposed for cardiovascular therapies. 13 The reliance on biomarkers as substitutes for outcomes in clinical trials is best justified when adequate safety data are collected and when the results are viewed clearly as a …