Why a Checklist Is Needed
A common control group can avoid recruiting a separate control group for every treatment comparison. But a smaller planned participant count is not enough to justify sharing. The control participants must provide a meaningful comparison for each treatment, the allocation must be workable, and the saved recruitment must matter in the setting where the trial will run. This checklist brings those considerations together.
The question is not simply whether one control group is cheaper than several. It is whether the same control group is an appropriate comparator for all treatment arms, and whether the complete design is practical and worthwhile. This decision comes before the statistical planning of the comparisons themselves; sharing controls does not remove the need to plan those comparisons appropriately.
The Four-Part Checklist
Work through the four checks in order. A design that fails the comparability check should not be rescued by an attractive participant-count calculation. A design that passes the scientific check can still fail because of recruitment capacity, costs, or allocation constraints.
Ask whether every treatment and the control can be evaluated in the same target population, under compatible eligibility criteria, outcome definitions, follow-up, and trial procedures. Confirm that the control represents the relevant alternative for every treatment comparison.
Estimate the cost and effort for all participants, not just the number of control participants. Include treatment-specific costs, screening, sites, staff, monitoring, and any additional procedures required by the multi-arm protocol.
Verify that the planned numbers in each arm can be recruited and randomized under the available capacity, treatment supply, site constraints, and any practical or ethical requirements on allocation.
Compare the shared-control plan with a clearly specified alternative, such as separate two-arm trials, using sample sizes appropriate to the same objectives. Decide whether the reduction is large enough to matter after accounting for extra operational demands.
Check 1: Is the Control Comparable?
The same control group is useful only if its participants represent the control condition that would have been used in each treatment comparison. For example, if all arms enroll eligible patients with the same disease severity, use the same background care, and assess the same outcome at the same follow-up time, a common control may be a natural comparator.
Comparability can fail when treatments are studied in substantially different populations or under different procedures. Suppose one treatment is intended for early-stage disease and another for advanced disease, but the control response differs by disease stage. A pooled control group with a different mix of stages from either treatment group may not provide a reliable comparison for both. Likewise, controls recruited in a different calendar period may face different background care or outcome measurement. A shared control should ordinarily be concurrent with the treatments it is intended to compare.
Ask whether eligibility, treatment setting, background therapy, outcome definition, follow-up, and recruitment period are aligned. Some design differences can be handled by a prespecified, scientifically sound plan, but they should not be ignored. If the needed comparability is not credible, separate trials or a more carefully aligned design may be preferable.
Worked Example: Similar Treatments, Different Study Settings
A hypothetical team considers testing two treatments against usual care. Treatment A is planned for patients enrolled at hospital clinics during one year, with a 12-week outcome. Treatment B is planned for patients enrolled at community clinics the following year, with a 24-week outcome. The team proposes using one usual-care group of 100 people for both comparisons.
The arithmetic alone does not establish that these 100 controls are suitable. The recruitment settings and periods differ, and the outcome follow-up differs. If usual care changes between years, or clinic populations differ, the control outcomes may not represent the appropriate comparator for both treatments. A 12-week control outcome cannot directly serve as the 24-week outcome for Treatment B.
Checklist conclusion. The proposal fails the comparability check as stated. The team should align eligibility, recruitment periods, outcome definitions, and follow-up before considering a shared control. If these elements cannot be aligned, separate controls may be needed. No participant-saving calculation can repair an unsuitable comparison.
Check 2: What Is the Full Recruitment Cost?
Participant count and cost are related but not interchangeable. A shared-control design may recruit fewer people while requiring expensive treatment supplies, extra staff training, more complex randomization, or additional site procedures. Conversely, control recruitment can be costly, making it valuable to avoid duplicating control groups.
For a first-pass count comparison, suppose there are \(K\) treatment arms, each requiring \(n\) participants, and each comparison would use \(n_0\) control participants if run separately. The separate-trial total is \(K(n+n_0)\). If the same treatment and control targets are met with one shared group of \(n_0\) controls, the total is \(Kn+n_0\). The difference is:
This count calculation assumes that the stated arm targets are appropriate and that sharing does not require changing them. It is a comparison with a defined alternative, not a guarantee that a real shared-control plan will achieve the same targets. If allocation, precision, or recruitment constraints change the required numbers, recalculate using the actual proposed design.
Worked Example: Quantifying a Potential Participant Saving
In a hypothetical trial, three treatments are each planned to enroll 80 participants. If studied separately, each treatment comparison would also enroll 80 controls. The proposed shared-control design would enroll the same 80 participants in each treatment arm and one group of 80 controls.
With separate trials, the total is:
With one shared control group, the total is:
The projected saving is \(480-320=160\) participants, or \(160/480=0.333\), approximately 33.3% of the separate-trial total. This is a meaningful potential reduction if the 80-per-arm targets remain appropriate and the common control is comparable. The calculation does not show that costs also fall by 33.3%; treatment-arm costs and shared-trial operating costs must be assessed separately.
Check 3: Can the Allocation Be Delivered?
A mathematically attractive allocation can be operationally impossible. Check whether each site can recruit participants for all arms, whether treatment supplies and trained staff are available, and whether randomization can reliably deliver the planned numbers. Consider how long recruitment will take and whether the control arm can enroll at the rate required by the design.
Also assess whether the proposed control-to-treatment allocation is acceptable in practice. Participants and clinicians may be less willing to join if the chance of assignment to a particular arm is viewed as too low. In some settings, an allocation ratio may be constrained by equipoise, available treatment, or site capacity. The choice of allocation should therefore be reviewed against both the statistical plan and the realities of recruitment.
Worked Example: A Saving That Still Exceeds Recruitment Capacity
A hypothetical plan has three treatment arms. Each comparison requires at least 60 participants in its treatment arm and 60 in its control arm. Separate trials would therefore require \(3(60+60)=360\) participants. A shared-control plan with 60 controls would require \(3(60)+60=240\) participants. The available recruitment capacity for the study period is 210 participants.
The shared plan saves \(360-240=120\) participants relative to the separate-trial plan, a reduction of \(120/360=0.333\), or 33.3%. However, its own total of 240 exceeds capacity by \(240-210=30\) participants. The separate-trial plan also exceeds capacity, by \(360-210=150\) participants.
Checklist conclusion. Sharing is more feasible than running the separate trials, but neither plan is feasible as specified. The team must not silently lower the arm targets to fit the cap: those targets were stated as minimum requirements for the planned comparisons. It should revisit recruitment duration, sites, or the scientific objectives and then reassess the design.
Check 4: Are the Savings Worthwhile?
A useful comparison states the alternative clearly and reports both the absolute saving and its context. “Fewer participants” is incomplete unless the number is compared with a credible plan that addresses the same questions. When the sample sizes differ across arms, use the actual totals for each design rather than assuming equal allocation.
The value of a saving depends on what is scarce. Avoiding control recruitment may matter greatly when eligible participants are difficult to find or when control procedures are burdensome. A smaller percentage saving may still be worthwhile if it substantially reduces recruitment time. But a large nominal saving may have little practical value if shared-trial coordination adds considerable cost or if recruitment is not the limiting resource.
Worked Example: Separating Count Savings From Cost Savings
A hypothetical two-treatment study compares treatments A and B with usual care. Each treatment arm and each separate control group would enroll 50 participants. Suppose the estimated recruitment cost is 1 cost unit per control participant and 2 cost units per treatment participant. Ignore all other costs for this simple comparison.
Two separate trials would recruit \(2(50+50)=200\) participants. Their estimated recruitment cost is:
With one shared group of 50 controls, the total is \(2(50)+50=150\) participants. The estimated recruitment cost is:
The shared plan saves 50 participants, or \(50/200=25.0\%\), and 50 cost units, or \(50/300=16.7\%\), under the stated cost assumptions. The percentages differ because treatment participants cost more than control participants. Actual total costs could differ if multi-arm coordination, equipment, or site requirements add expenses; these must be included before concluding that the shared design is cheaper overall.
Putting the Checks Together
A suitable decision is conditional, not automatic. Use the following summary to record the evidence for each part of the checklist. A “no” or “uncertain” answer should prompt a design review rather than an assumption that sharing is beneficial.
| Check | Evidence to document | Warning sign |
|---|---|---|
| Comparability | Aligned population, control condition, outcomes, follow-up, and recruitment period | Control outcomes may not represent one or more treatment comparisons |
| Costs | Recruitment and operating costs for the full design | Only control recruitment costs are counted, while added multi-arm costs are omitted |
| Allocation | Arm targets, site capacity, treatment supply, and acceptable allocation | A required arm is unlikely to recruit or cannot be supplied as planned |
| Savings | Absolute and relative participant reduction against a defined alternative | The saving is presented without checking whether the shared plan itself is feasible |
The main alternatives are to use separate control groups in separate trials, or to revise the shared-control plan so its comparisons and operations are aligned. Separate trials may be preferable when populations, settings, or periods differ; they can also avoid some cross-arm operational complexity. A shared control may be preferable when the comparator is genuinely common, recruitment is aligned, and the resulting savings matter. Neither approach is universally superior.
Common Pitfalls and Limits
- Choosing sharing from the participant count alone. A smaller total does not make an unsuitable control scientifically comparable. Pass the comparability check first.
- Assuming fewer participants means lower total cost. Count treatment-specific and shared operational costs, not only control recruitment.
- Calling a design feasible because it saves relative to another design. Compare the proposed total with actual recruitment capacity. A smaller plan can still exceed it.
- Comparing against an unclear alternative. State whether the reference is separate trials or another planned design, and use targets appropriate to the same objectives.
- Treating the simple saving formula as universal. The expression \((K-1)n_0\) applies to the specified equal-target comparison. It does not account for changed targets, different allocation plans, or added costs.
Check Your Understanding
Answer each question, then compare your reasoning with the short solutions.
- Three separate comparisons each plan 40 treatment participants and 40 controls. A shared design keeps these targets and uses one group of 40 controls. How many participants are saved, and what is the shared-design total?
- Name two features of a trial that should be aligned when judging whether one control group is comparable across treatment arms.
- A shared-control plan saves 25% of participants compared with separate trials, but its total still exceeds the available recruitment capacity. Is it feasible as planned? Explain.
- Why can a 25% reduction in participant count differ from the percentage reduction in recruitment cost?
- When might separate trials be preferable even if a shared-control design has a lower nominal participant count?
Solutions. (1) Separate trials require \(3(40+40)=240\) participants; the shared plan requires \(3(40)+40=160\), so it saves \(240-160=80\) participants. (2) Examples include eligibility criteria, control condition, outcome definition, follow-up time, and recruitment period. (3) No. Relative savings do not establish feasibility; the proposed total must fit capacity or the plan must be revised. (4) Participants can have different recruitment costs, and a multi-arm protocol can add operating costs. (5) Separate trials may be preferable when populations, settings, periods, or control conditions are not sufficiently comparable, or when the shared design cannot be implemented reliably.