Since patients to be randomized may present at unpredictable times, and the interval between their agreement to be randomized and the allocation to treatment should be kept to an absolute minimum, access to randomization should be maintained throughout normal clinic hours at least, and this often requires dedicated staff. Where this is impractical, or in emergency situations where treatment decisions must be made extremely rapidly, 'sealed envelope' randomization may be a practical, though not completely infallible, solution. This involves compiling a randomization sequence of treatment allocations and identification numbers, using simple or block randomization, for each centre taking part in the trial. Usually, for simplicity, patient characteristics are not taken into account, although this can be done. The central list is retained. Opaque envelopes are prepared - it is useful to print a copy of the registration form recording the necessary patient information on the envelope. A sheet containing the treatment allocation and identification number is placed in the envelope which is then sealed, the identification number is then written on the outside of the envelope so that they can be placed in order. When a patient is being considered for randomization, the registration sheet should be completed and the next envelope opened in sequence to obtain the treatment allocation, which in turn should be written onto the envelope, together with the date of randomization. The original envelope should be returned to the trials office immediately. The office can then check that the patient ID and treatment allocation tally, and that the dates of randomization are consistent with the envelopes being opened in sequential order. With rigorous checks, the potential for bias (opening envelopes until the treatment allocation 'desired' is reached) can be minimized - for example recording the time as well as date of randomization may provide data which can be checked against hospital records of, for example, operation times. This remains, however, theoretically open to abuse - for example as the envelopes must be prepared in advance, they could be opened in advance and operating lists arranged to suit. Random visits to centres to check that there are no opened but as yet unused envelopes can help, but in general, sealed envelope randomization is best used only as a method of last resort. The situation in which rapid telephone access to the randomization centre is truly impossible must be increasingly rare.
Was this article helpful?