Randomisation design

Study design Randomised two period cross-over, 12-36 healthy volunteers. 

Subjects are randomised to receive A or P on different occasions in a crossover design. 

Table 4.3 ParaUel-study design with two alternating cohorts of eight subjects and a 6 2 randomisation to active drug (A) or placebo (P) - each subject receives either A on four occasions or P on four occasions...

A preliminary assessment of the effect of food on pharmacokinetics can generally be studied in a single-dose, two-arm, randomised, crossover design. Preliminary information can often be obtained by including a fed occasion in the first, dose-escalating study. This will be insuffi-cent for registration purposes, which require an adequately powered study performed with the final formulation, but the information should be sufficient to indicate whether there is need for restrictions on dosing relative to meals in repeat-dose studies in healthy volunteers and patient clinical trials. 

Crossover designs are susceptible to carry-over effects, that is, the treatment effect from the first period has not worn off at the time of conducting the second period. Tests of analysis can detect carry-over effects, but it is too late then to modify the design. Similarly, period effects may confound the interpretation of cross-over studies, that is, the order in which one treatment occurs in a sequence compared with another influences the response to early treatment. Randomisation usually, but not always, precludes the effect. 

In order to reduce the number of subjects on placebo in a clinical trial, some investigators employ an rmequal randomisation technique, whereby fewer subjects receive placebo than receive active comparator. For example, the ratio of 1 2, or 1 3 may be chosen in a large clinical study. These designs are not considered acceptable by aU statisticians... 

Many have argued that concerns for the individual intuitively lead to the use of adaptive designs in which randomisation is biased... 

However, there is a counter argument. In the ECMO trial, after the result of the nineth patient became known the RPW design requires that the patients be randomised to ECMO compared to the standard ventilator in the ratio 9 1. Clayton argues that if the one treatment is so much superior to the other that 90% of patients are allocated to it, it is unethical to withhold it from the remaining 10%. If we accept that argument, is it also true if the ratio is 8 1 or 7 1 How much information is sufficient to make us, ethically, refuse to randomise patients Such questions are not simple. 

Of course, the control in a crossover design is not contemporaneous because it occurs within different treatment periods. Nonetheless, such trials are randomised and the effect of differential period effects can be allowed for in the analysis and does not give rise to bias in treatment estimates. 

Palmer CR, Rosenberger WF. Ethics and practice alternative designs for phase III randomised clinical trials. Control Clin Trials 1999 20 172-86. 

Eulvestrant has been evaluated in two randomised phase III trials in postmenopausal women with advanced disease after progression on prior antiestrogen therapy. In both trials, fulvestrant was at least as effective as anas-trozole. In a prospectively designed combined analysis of the results from both trials, median time to progression (TTP) was 5.5 months for fulvestrant versus 4.1 months for anastrozole [172]. Eulvestrant and tamoxifen have been compared as first-line treatments in a trial including post-menopausal women with advanced breast cancer. In this study, the between-treatment difference was non-significant (median TTP 6.8 versus 8.3 months) [173]. 

Within the project there were essentially two trials and these clearly illustrated the effectiveness of the randomised controlled design. 

Randomisation is clearly a key element in the design of our clinical trials. There are two reasons why we randomise subjects to the treatment groups ... 

Unequal randomisation is sometimes needed as a result of these considerations. To achieve this, the randomisation list will be designed for the second example above with double the number of A and B allocations compared to placebo. 

In both the paired design and the cross-over design there is, of course, randomisation in the second paired design example above, it is according to which forearm receives A and which receives B and randomisation is to treatment order, A/B or B/A, in the cross-over design. 

In this context, each patient would be receiving each of the multiple treatments. In the cross-over trial with three treatments this would likely be a three-period, three-treatment design and patients would be randomised to one of the six sequences ABC, ACB, BAC, BCA, CAB or CBA. Although there are again ways of asking a simultaneous question relating to the equality of the three treatment means through an analysis of variance approach this is unlikely to be of particular relevance questions of real interest will concern pairwise comparisons. 

Peto R, Pike MG, Armitage P, Breslow NE et al. (1976) Design and analysis of randomised clinical trials requiring prolonged observation of each patient. 1. Introduction and design British Journal of Cancer, 34, 585-612... 

Other problems pointed out by Box et al. [20] are serially correlated errors, dynamic relations and feedback. All the above problems can be overcome by the use of properly designed statistical experiments that employ features such as randomisation, blocking and other suitable controls. 

The experimental order of this study was also changed from the ideal randomised and blocked design. This was justified for reasons of automation where column changes needed to be minimised. As for the aspirin study the validity of this compromise depends on the fact that the repeatability of the method, over a time span such as that required for the ruggedness test, had previously been established. 

---