End Points in Oncology Clinical Trials: Revisiting Overall Survival
When a promising investigational cancer drug enters a development area of great medical need, there is high pressure on the manufacturer from patients and from physicians to quickly secure patient access to this drug. This pressure can affect the trial design, regulatory requirements, and the trial results that are obtained before the drug is launched. For example, the drug may be approved based on data from phase 2 clinical trials that may have a single arm or a limited sample size or duration. These concessions are made as a result of the substantial unmet need for an effective therapy.
In clinical trials targeting a high unmet medical need, overall survival (OS) data may be immature, as a result of a long natural history of the disease or the limited follow-up period by the time the drug is being launched. In addition, patient crossover may confound the OS data. That is, for ethical reasons, when a patient’s disease progresses in the control arm of a phase 3 clinical trial, that patient should be switched to the best available therapy. However, in many cases when alternative therapies are limited, the options for best available therapies may include the experimental drug. This is due to the consequences of delaying treatment with an effective therapy—even an experimental therapy—in situations where the patient has a limited life expectancy with the use of currently available therapies. In such cases, many patients in the control arm ultimately receive the investigational drug, but only after failure of another therapy and, therefore, with a lag time relative to the population in the active-treatment arm.
This situation leads to a dilemma for drug makers, who have an ethical responsibility to provide access to investigational drugs. But exposure to the experimental drug in the control arm reduces the possibility of showing an OS benefit in the active-treatment arm, because patients in the control arm also gain the benefit of the investigational drug. Moreover, even when a clinical trial is designed to prohibit crossover to the investigational therapy, the study may be halted early as a result of positive efficacy results. This, in turn, reduces the amount of data available to demonstrate a significant OS benefit.
Such barriers to demonstrating OS benefit within phase 3 clinical trials can have important implications for the clinical acceptance and for reimbursement of a new and effective therapy. The reimbursement of novel,
potentially life-saving therapies is particularly challenging, because historically, an improvement in OS has been considered the clearest demonstration of clinical value. Improved OS indeed reflects the absolute goal of extending patient life and lends itself to health economic calculations of cost per life-year gained or cost per quality-adjusted life-year gained, without the need for the uncertainties of modeling based on surrogate end points. However, in cases where demonstration of an OS benefit may not be possible, or is underestimated, at the time of launch, the new drug may face reimbursement restrictions. Such restrictions have important implications for physicians and for patients, who may be unable to access the promising and potentially life-extending new drug.
Currently, methods for addressing limited OS data are gaining acceptance with payers. For example, in some cases, surrogate end points, such as response rate or progression-free survival, can be validated as predictors of OS. Similarly, when patients in the control group are crossed over to the investigational drug, statistical methods can be used to demonstrate an OS benefit by correcting for the efficacy of the experimental drug in the control arm.
In general, drug makers face a challenge in bridging the gap between the urgency for a new drug in a high-need area and the time required to collect data that can satisfy regulators and payers. Consequently, although OS remains an important end point that is relevant to patients, clinicians, and payers, there will continue to be situations where clinical need demands a drug with imperfect OS data. These instances require a collaborative approach among stakeholders, where drug makers collect the best possible data to demonstrate value, and payers acknowledge the limitations that may be necessary in terms of trial design (including end points).