Table 1 Theoretical risks associated with forced randomization and the corresponding risk mitigation strategies
Phenomenon | Example | Risks | Risk mitigation |
|---|---|---|---|
Temporary shortage of supply of a certain drug type | Suppose an experimental drug (E) is not available at any of the sites until the new batch is produced, but the study continues enrolling participants. If FR is allowed = > skipped allocations are known to belong to treatment E. When drug is re-supplied, massive backfillings to the skipped allocation numbers will be occurring. | Observer bias Selection bias Chronological bias | Option 1. Design the study to use scrambled allocation numbers to conceal that forced allocation occurred; this prevents the selection bias and the observer bias, but not the chronological bias associated with the time trend. Option2. If feasible, continue screening the participants but pause the enrollment until drug supplies become available on site. |
Potential unblinding in a study with interchangeable use of placebo run-in and active placebo kits | Suppose we have a two-period 1:1 RCT, where both arms have placebo run-in Period 1 followed by a randomized (Active: Placebo) Period 2 = > plenty of placebo supplies at the sites during randomization. If FR is allowed = > skipped allocations belong to Active and actual assignments are Placebos. | Selection bias | Use scrambled allocation numbers to prevent the knowledge of whether the participant was assigned according to the original randomization or using FR resolves the issue. |
Potential unblinding through FR in a study with a highly unequal allocation ratio | Suppose the initial drug supply to each site includes one block of kits (say, 5 + 1 drug kits in a study with 5:1 randomization, A:B). If it is known that some allocation has been skipped before 5 patients are randomized and before re-supply arrives at the site = > skipped allocation is known to be B. | Observer bias | Use scrambled allocation number to conceal occurrences of FR. Also, include more than one kit of drugs of the rare treatment group both in the initial and in the re-supply shipments |
Improper specification of the re-supply triggers | Suppose the study has unequal allocation ratio (e.g., 3:1, Active: Placebo) but uses equal re-supply triggers (e.g., request re-supply kits of a certain type if < 3 kits of this type are available). If at some point new shipments are delayed but the site continues randomizing new patients until the supplies are exhausted = > greater than desired allocation to Placebo. | Unblinding Selection bias | Carefully specify the drug resupply policy. Use scrambled allocation numbers to prevent the knowledge of whether the participant was assigned according to the original randomization or using FR. |
Site forgets to acknowledge the receipt of resupplies | IRT, not having an accurate picture of what is available at the site, force-allocates several subjects at the site to the same treatment. | Lack of within-center balance; possible bias | Ensure strong compliance with the drug receipt acknowledgement processes. |