Limitations of randomized clinical trials. Proposed alternative designs.

Abstract

The classical two-period crossover and two-parallel-groups designs for clinical drug trials are unable to answer many current scientific questions, and are sometimes ethically or financially difficult. For example, the classical designs do not allow the study of the effects of combined treatments and their interactions. Also, the generation of parallel data that are no more than a repetition of previous research is ethically debatable and in a sense a waste of money. Objectives are to identify what a classical clinical trial cannot manage, to summarize and discuss alternative trial designs that are helpful for such purposes. A classical clinical trial cannot: (i) assess combined therapies, (ii) take historical data into account, (iii) safeguard ethics and efficacy during the course of long-term trials, (iv) study drugs, before well-established toxicity information is available, (v) account for the possibility of therapeutic equivalence between test and reference treatment, (vi) study multiple treatments in one trial, and (vii) adjust change scores for baseline levels. Alternative designs helpful for such purposes are respectively: (i) factorial designs, (ii) historical controls designs, (iii) group-sequential interim analysis designs, (iv) sequential designs for continuous monitoring, (v) therapeutic equivalence designs, (vi) multiple crossover-periods/multiple parallel-groups design, and (vii) increased precision designs through multivariate adjustment. Main problems include the increased risks of type I and type II errors and the loss of validity criteria. Non-classical trial designs are reviewed. They offer relevant scientific, ethical, and financial advantages. The increased risks of type I and type II errors should be accounted for in the design stage of the trial.