Alternatives to randomization in surgical studies.

The purpose of randomization is to provide unbiased estimates of treatment effects and valid probability statements for hypothesis tests in comparative studies. Difficulties with randomized allocation of patients in surgical studies include: patient selection--due to protocol limitations, possible changes in referral patterns, the need for patient consent, and physician co-operation; disruption of the patient-physician relationship; static protocol not responsive to changes as surgical skill evolves and experience regarding patient selection is acquired; the larger number of patients required causes a longer study and more patients receiving what may be a worse therapy. Moreover, for evaluating a new heart valve, the primary purpose should be estimation rather than comparison with a current device. This can be accomplished faster, in a more generalizable way and with more concern for patient care by using non-randomized comparison groups.