Estimating a Treatment Effect by the Accelerated Hazards Model

In randomized clinical trials, when the outcome of interest is time-to-event, Cox’s proportional hazards model is often used to identify the treatment effect. This model usually assumes that the treatment effect is a constant of proportionality between the hazard function of the treatment and control arms. This assumption, however, is not applicable in certain trials. A simple alternative called the accelerated hazards model is then introduced when the treatment is believed to accelerate or decelerate the underlying risk progression over time. The treatment effect is therefore characterized by the hazard progression time ratio. Survival data from a randomized placebo-controlled trial (Brem, et al., 1995), which evaluates the effectiveness of biodegradable polymers with carmustine to treat malignant gliomas, is used to illustrate the model