Κανόνες διακοπής κλινικών δοκιμών φάσης ΙΙΙ: ανασκόπηση και σύγκριση
This review presents various methods to early terminate phase III and other large-scale clinical trials. Focus is given on the group sequential design which involves sequential patient enrollment and analyses of accumulating data at pre-defined inspection times. Efficacy evaluation is performed in the context of a repeated significance testing to allow for early stopping. In the first part, several approaches for the one-stage, two-arm design will be discussed. Treatment efficacy is being evaluated using either a continuous or a time-to-event endpoint. Emphasis is given in the mathematical background of each approach. Subsequently, various approaches for the multi-stage design will be discussed, as well as the seamless phase II/III design. The latter is widely used for time saving reasons (e.g. due to an urgent need for drug approval), for cost-reducing reasons and for minimising the sample size. Designs having two co-primary endpoints will also be discussed. In the second part, a comparison among various designs is attempted with regard to the total sample size needed to reach the expected statistical power, and their ability to detect a treatment difference, if any, without allowing a large number of participants to be engaged. Stopping rules constitute an integral part of the statistical design, primarily for ethical reasons. Early termination must take place in case of the new treatment proven to be inefficient (or not as efficient as the existing one) to prevent more patients from being exposed to ineffective drugs. Early stopping must also take place in case of the new treatment proven to be superior to the existing one to rapidly make it available to the public. Other reasons for stopping a trial, such as high toxicity rate, low accrual rate, arising breakthrough treatments, are out of scope of this review.