Patient-oriented Randomisation - a new clinical trial design

Abstract

In this thesis a new randomisation designs for clinical trials, called patient-oriented randomisation design, is introduced. This design was developed to counter problems of classical' randomised controlled trials comparing strategies (consisting of different treatments in each strategy) in presence of heterogeneity in patient-drug-interactions. The discrepancies between daily clinical perception and results of randomised controlled trials lead to the conviction that the methodological approach of classical' randomised controlled trials, such as the block randomisation design, is appropriate in this set-up. The patient-oriented approach of the CUtLASS' design reflects everyday clinical practice, by allowing for a patient-oriented choice of one treatment of each strategy. The allocation for a strategy is random. However, the results are highly dependent on the physicians' preferences. The goal of the design described here is to take an intermediate path between randomised controlled trials and the CUtLASS' design. The idea of the new trial design is to randomise two treatment pairs each consisting of one treatment of the one strategy and one treatment of the other strategy in a first step and subsequently, to involve the investigators in deciding for a pair most appropriate to the patients' needs and then to randomise the allocation to one treatment of that chosen pair. After a short introduction, in Chapter 2 basic definitions and notations are given. The considerations concentrate on the properties of the patient-oriented randomisation design which depend mainly on the number of treatments of each strategy. An advantage of the patient-oriented randomisation design (compared to classical randomisation designs) is that in some cases a clear patient-oriented treatment decision of the physician can be seen and investigated. The patient-oriented decision in the CUtLASS' and the patient-oriented randomisation design can lead to unbalanced treatment sample size in each strategy. Chapter 3 investigates the changes of the allocation probability of each treatment and determines the minimum and the maximum allocation probability in both designs. The investigation shows that the interval of possible values of allocation probabilities is lager in the CUtLASS' design. However, the patient-oriented randomisation design ensures that each treatment is in fact administered in the study. Hence, it is possible to compute the number of patients needed to avoid poorly represented treatments and poorly powered comparisons. After this, the practical implementation of the patient-oriented randomisation design are looked at and methods to estimate the probability of imbalance are presented. The block length of the patient-oriented randomisation design is large, hence, modification for the creation for the random list are possible and effective to reduce the probability of imbalance. Finally, Chapter 5 deals with a statistical model to compare two strategies consisting of different treatments. Basic definitions and notations are given for unbalanced one-way classification with fixed effects and the concept of contrasts. Additionally, the required sample size calculation is presented. The power of the associated contrast test depends particularly on the allocation of sample size in each treatment and the expected effect between both strategies. The final consideration shows that heterogeneity of patient-drug-interaction lead to no effect in the Block' design. Randomisation designs such as the CUtLASS' design and the patient-oriented randomisation design only work as well as the physicians selecting the treatments. In a patient-oriented randomisation design, it can be distinguished between patients receiving the treatment not only due to randomisation but due to a patient-oriented decision of the physician (patient-oriented treated patients) and patients receiving the treatment due to randomisation's reasons (randomised choice treated patients). Therefore, the patient-oriented randomisation design allows to test the difference in treatment means between patient-oriented treated patients and randomised choice treated patients in each treatment, each strategy, or overall. In particular, the possibility to analyse the difference between strategies for subgroups consisting of only randomised choice treated patients or only patient-oriented choice treated patients help to test the assumptions about the heterogeneity of patient-drug-interaction in the patient-oriented randomisation design. In the CUtLASS' design, it is not possible to answer this important question. The conclusion consists of a discussion of the results as well as important findings. An outlook on possible future work is also presented.