Systematic allocation

In the past, allocation of treatment to patients in clinical trials was often done systematically rather than randomly. If patients were being recruited sequentially, only the first participant would be allocated treatment at random. The remaining individuals were allocated treatments alternately. Other methods of systematic allocation (often misleadingly termed systematic randomization) were to assign patients according to date of birth or date of presentation (for example, odds days went to the treatment group, even days to the control group).

Such methods did have the advantage that treatment group sizes were the same or very similar. However, all systematic methods have the disadvantage of predictability, and so there is the possibility of bias when individuals are being allocated. Systematic allocation is therefore now seldom used and is emphatically not recommended.

An example

In the 1940s systematic assignment was considered to constitute adequate randomization for clinical trials. Wright et al. (1948) reported on a trial of anticoagulants for treatment of coronary thrombosis with myocardial infarction. Patients admitted on odd days of the month received anticoagulant whilst those admitted on even days did not. The resultant sizes of the two treatment groups are shown below:

{Fig. 1}

It can be seen that considerably more patients were recruited to the treated than to the control group. Unlike with random allocation, this should not be possible with systematic allocation which should produce (virtually) equal group sizes.

This means that either far more patients were admitted on odd days than on even days (which seems improbable) or that there was bias in the allocation of children to groups.....