An important part of trial design is deciding on the level of blinding. Generally it is preferable to use the highest level of blinding possible within the trial constraints.
Blinding is not possible or practical for many trials such as those testing medical devices or involving surgery. Trials of complex interventions such as health care policies or staff training methods are also not easily blinded.
Where blinding isn't possible it is important to have an objective outcome measure that can't be easily influenced by the observer. Alternatively it may be possible to use observers not involved in the patients' care who assess patient outcome without knowing the treatment group (blinded outcome assessment).
This usually refers to the patient being blinded to the treatment given but not the administering clinician. Trials of medical devices or policy in intensive care are usually of this nature because the patient is not fully conscious. Sometimes surgical trials can be single blinded if two different types of surgery are being compared or even if surgery is compared to no surgery (by making a placebo incision in the control group for instance).
Neither the patient nor clinician know which treatment the patient is randomised to. Placebo controlled drug trials are usually double blind. It may sometimes be possible for the clinician to guess which treatment the patient has received by blood markers or other side effects. If this is the case the simple randomisation code (see below) is not recommended because the clinician may break the blind on all patients by deducing the simple coding system from one patient.
Allocation concealment is a different concept to blinding. It means that the person randomising the patient does not know what the next treatment allocation will be. It is important as it prevents selection bias affecting which patients are given which treatment (the bias randomisation is designed to avoid).
Allocation concealment is possible with all types of trial, including unblinded trials, and is therefore universally recommended. The best way of ensuring allocation concealment is to use a centralised service, since this cannot be subverted by investigators and provides independent verification that it was not possible for the investigators to know the allocation sequence in advance. Read about the problems of not using centralised randomisation.
Note that using patient date of birth (e.g. odd/even) or alternating treatments as the randomisation scheme means that the allocation is not concealed but is open to all. The treatment allocation may then be subject to selection bias since patients expected to have a worse outcome may be selectively excluded from the active treatment group.
| Type of code | Description | Example |
|---|---|---|
| Explicit | The name of the treatment itself | surgery/no surgery |
| Simple | A code word for the treatment. Rarely used for double blind trials because breaking the code for one patient unblinds the entire trial. | A/B |
| Unique code | A unique number or code that is linked via a code list to the treatment | 1305 |
Double blind drug trials are often co-ordinated by the hospital pharmacy at each trial centre. One of the pharmacists will have responsibility for dispensing the trial drug to a patient based on the unique randomisation code for that patient. We provide the pharmacist with a code list for their centre that links up to the randomisation codes on our website system.
Alternatively the trial drugs may be labelled by the manufacturer or drug packager. We can either provide a list to be used for the labelling, or incorporate the list used by the packager into our system.
Read more about managing drug supplies …
