Randomisation Methods

In general, randomisation methods can be classified into three types according to the restriction of the randomisation and the change in probability for randomisation with respect to the previous treatment assignments. The three types of randomisation methods are the complete randomisation, the permuted-block randomisation, and the adaptive randomisation. Randomisation can be performed either by random selection or by random allocation for methods of complete and permuted-block randomisation. Basically the adaptive randomisation consists of treatment and covariate and response adaptive randomisations.

Complete Randomisation

Simple randomisation such as the above mentioned complete randomisation is referred to as the procedure in which no restrictions are enforced on the nature of randomisation sequence except for the number of patients required for achieving the desired statistical power and ratio of patient allocation between treatments. For a clinical trial with N patients comparing a test drug and a placebo, the method of simple randomisation is called a simply complete randomisation.

Another type of simple randomisation that provides equal allocation of sample size is the random allocation. Random allocation is the simplest form of restricted randomisation. The method of random allocation randomly selects the N/2 out of a total of N patients without replacement and assign these N/2 patients to receive the test drug and the other half to receive the placebo. Since there are a total of N!/[(N/2)!]2 possible ways for the selection of N/2 patients, it is equivalent to generating a random permutation of numbers from 1 to 2N and assigning the first half to the test drug.

Permuted block randomisation

One of the major disadvantages of simple randomisation is that treatment imbalance can occur periodically. One resolution to this major disadvantage of simple randomisation is periodically to enforce a balance in the number of patients assigned to each treatment. In other words, we first divide the whole series of patients who are to enrol in the trial into several blocks with equal or unequal lengths. We then randomise the patients within each block. This method of randomisation is known as the permuted-block randomisation, and probably the most frequently employed method for the assignment of patients to treatments in clinical trials.

Adaptive randomisation

The method of complete randomisation can cause an imbalance in the patient allocation to treatments. On the other hand, the methods of random allocation and permuted-block randomisation are useful in forcing a balanced allocation of patients to treatments within either a fixed total sample size or a pre-specified block size. These methods of restricted randomisation can also maintain constant marginal probability for the assignment of patients to treatments. In practice, in addition to enforcing a balanced allocation among treatments to some degree, it is also of interest to adjust the probability of assignment of treatments during the study. This type of randomisation is called adaptive randomisation because the probability of a current patient being assigned is adjusted based on the assignment of previous patients. Unlike the other randomisation codes based on the method of adaptive randomisation cannot be prepared before the study begins. This is because that randomisation process is performed at the time a patient is enrolled in the study, whereas adaptive randomisation requires information on previously randomised patients. In clinical trials the method of adaptive randomisation is often applied with respect to treatment, covariates, or clinical response. Therefore the adaptive randomisation is also known as treatment adaptive randomisation, covariate adaptive randomisation, or responsive adaptive randomisation.

Treatment Adaptive Randomisation adjusts for the assigning probability of the current patient with respect to the number of patients who have been randomised to each treatment group. In certain diseases some of the prognostic factors are known to affect clinical outcomes of the treatment. Therefore it is desired to achieve a covariate balance with respect to these prognostic factors such as age and sex. For this purpose we may consider to employ the Covariate Adaptive Randomisation that is also known as minimisation method. Another method, Response Adaptive Randomisation, is to adjust for the assigning probability according to the success or failure of the treatment to which previous patients were assigned.

Unlike other randomisation methods, adaptive randomisation requires a much more complicated analysis compared to other methods of randomisation. This is because the adaptive randomisation method does not have an equal assigning probability for each patient, which needs to be calculated based on the assignment of previous patients.

Complete randomisation and permuted-block randomisation are quite often used in phase I, II and III studies, whereas adaptive randomisation is sometimes applied in phase III studies with large sample size and long study period.

Generation of Randomisation Code

Our assigned statistician uses a special program to generate randomisation code. For generating the randomization code a ‘seed’ is chosen in the program so that the randomisation code can be reproduced by changing the ‘seed’ can generate a different randomisation code.

The statistician generates the code according to the study design (such parallel design, crossover design, factorial design, Latin Square design, etc) and protocol requirements. Proper block size is used to ensure balanced distribution of treatment allocations. Once he generates the code he will also produce a frequency table to verify and ensure that the number of subjects and their distribution in the draft schedule is correct. Proper footnote, version number and time will be added to make the randomisation code easily understandable.

Draft randomisation code will be sent to the sponsor and reviewed by a sponsor. The reviewers will check that:

(1) Randomisation code is produced according to the protocol requirements
(2) The number of subjects in each treatment group or treatment sequence is correct

Any findings or comments on the draft version will be communicated back to the statistician.

The statistician will revise the program fully addressing those comments and suggestions. The statistician will then issue a new randomisation code with a different seed. This code is subject to the same checking procedure as the draft one by sponsor. If no further changes are required in the revised randomisation code, the statistician will use another seed to generate the final version.

Following finalisation, the statistician will export the final code into two electronic files: Excel file and PDF file. Those datasets are saved in a separate directory and declared master databases. This directory is then set to ‘read only’ using the directory permission settings. This ensures that the contents of the master database cannot be altered or accidentally deleted. The statistician will also save the final program used for the generation of the randomisation code into the final program directory.