Accelerated failure time model methods The methods in this section make use of accelerated failure time models, an alternative form of survi val model to the commonly used proportional hazards selleck chemicals Perifosine model. A proportional hazards model assumes that cov ariates multiply the hazard by a constant, whereas an AFT model assumes that a covariate multiplies the e 0 is often called the acceleration factor, the amount by which a patients expected time to event is increased by treatment. A value of e 0 1 indicates a beneficial treatment effect whereas e 0 1 suggests treatment has a detrimental effect, increasing the speed at which a patient moves towards their event. e 0 is perhaps easier to interpret Inhibitors,Modulators,Libraries than e ? 0 so results will be presented in this form.

By defining a binary process Xi which equals 1 when a patient is on experimental treatment and 0 otherwise, equation can be rewritten as predicted event time by a constant. These methods have been referred to Inhibitors,Modulators,Libraries as randomisa tion based efficacy Inhibitors,Modulators,Libraries estimators, as they com pare groups as randomised and therefore are intended to reduce biases which may be introduced by comparing groups as treated. The method of Loeys Goetghebeur described previously is also an RBEE as it preserves the randomisation balance and the significance level from an ITT analysis. Inhibitors,Modulators,Libraries Rank preserving structural failure time models Robins and Tsiatis describe the use of AFT models to estimate the true efficacy of a treatment. A patients observed event time is related to their counterfactual event time, that which would have been observed for a particular Inhibitors,Modulators,Libraries patient if they had not received any treatment.

sellckchem These models are referred to as rank preserving as they make the assumption that given two patients i and j, if i failed before j when both were on one treatment, then i would also fail before j if both patients took the same For a given value of, the hypothesis0 can be tested by first calculating Ui using equation. Z is then calculated as the test statistic for the hypothesis U R, i. e. a patients counterfactual event time is inde pendent of the treatment arm to which they were randomised. A number of different tests could be used to calculate Z. We considered four different statistical tests in this investigation the logrank, Cox, exponential and Weibull tests. The value of for which Z 0 is taken as the point estimate. This is the value for which U is balanced between treatment arms. The method has been extended and implemented in Stata by White et al as follows. Define Ci as the administrative censor ing time which corresponds to the end of follow up. Using equation, the censoring time for Ui is given by close to its value from the previous iteration, at which point the process is said to have converged.