Why the Design Choice Is More Than a Participant Count
When several treatments are to be compared with a control, the comparisons can be organized as separate two-arm trials or as arms within one multi-arm protocol. A shared control can reduce the number of control participants recruited, as the earlier material on shared-control sample sizes establishes. But a lower total participant count does not, by itself, settle which arrangement is feasible. Recruitment rates, eligibility criteria, site capacity, treatment logistics, and the ability to coordinate one protocol also matter.
This comparison is most useful early in planning, when the team can still decide whether the treatment comparisons belong in one coherent study. The calculations below focus on recruitment and operations. They do not establish that either design is statistically preferable in every setting, and the illustrative sample sizes are not a substitute for a complete analysis plan.
A Planning Scenario: Two Treatments
Suppose a research group is considering two treatments, A and B, for the same condition. For illustration, assume the planning calculation for either treatment-versus-control comparison calls for 80 participants in the treatment arm and 80 in its control arm. These numbers are treated as given planning requirements; they would ordinarily depend on the outcome, target effect, variability, significance level, power, and analysis plan.
If the comparisons are run as separate trials, each trial recruits 80 treatment participants and 80 control participants. The two trials therefore recruit \(2(80+80)=320\) participants in total. In one multi-arm protocol, the study recruits 80 participants to control, 80 to treatment A, and 80 to treatment B, for \(80+80+80=240\) participants.
Worked Example: Comparing Randomized Participant Targets
A planning team proposes two treatment-versus-control comparisons. For each comparison, its assumed sample-size requirement is 80 participants per arm. Compare the total randomized participant targets under separate trials and one common-control protocol.
Separate trials. Each trial needs 80 treatment participants and 80 control participants, or \(80+80=160\) participants. Because the control groups are distinct, the total for two trials is:
One multi-arm protocol. The protocol needs 80 participants in each of its three arms: one common control arm, treatment A, and treatment B. Thus:
The arithmetic difference is \(320-240=80\) participants. Relative to the separate-trial total, that is \(80/320=0.25\), or a 25% reduction in randomized participants under these assumptions. The result counts participants, not screening contacts, staff effort, or calendar time.
Why the Participant-Count Difference Arises
The same count can be expressed for \(K\) treatment comparisons when every comparison requires \(n_C\) control participants and \(n_T\) participants in its treatment arm. Separate trials each recruit their own control group. A single protocol recruits one control group and one treatment group for each of the \(K\) treatments.
Subtracting the multi-arm count from the separate-trial count gives:
The treatment-arm contribution \(Kn_T\) appears in both totals and cancels. The count difference comes from recruiting \(K\) control groups separately instead of one shared control group. This is a bookkeeping justification, not a guarantee of a practical or statistical advantage. It assumes that the same control participants can appropriately serve all comparisons and that the required sample sizes remain as specified.
From Randomized Participants to Recruitment Work
A recruitment target is not the same as the number of people who must be approached or screened. If a proportion \(q\) of screened people are expected to be eligible and enrolled, a simple planning approximation is to divide the randomized target by \(q\). This approximation assumes a stable yield and does not account for uncertainty in the screening rate, losses after enrollment, or differences between arms.
Worked Example: Estimating Screening Targets
Use the two-treatment scenario above. Assume, hypothetically, that 80% of screened candidates are eligible and enroll, and that the same yield applies across the relevant groups. Estimate the number of people to screen for each design.
For separate trials, the total randomized target is 320. At a yield of \(q=0.80\), the approximate total screening target is:
For one multi-arm protocol, the randomized target is 240, giving:
Under the common-yield assumption, the multi-arm plan requires an estimated 100 fewer screening contacts. This estimate is not an exact prediction: if eligibility or consent differs by treatment arm, site, or protocol, a single 80% yield may be misleading. The team should use the yield relevant to its actual recruitment pathway and allow for uncertainty.
Calendar time also depends on how quickly eligible participants can be recruited, not just on the total target. Recruitment rates should be stated clearly: a rate per arm, per trial, or across the whole protocol will lead to different timeline calculations. If trials recruit concurrently, their durations should not be added as though they ran one after another.
Worked Example: Comparing Approximate Recruitment Durations
Suppose each separate trial in the two-treatment plan can recruit 8 participants per month in total. Assume the two trials recruit concurrently, and each must enroll 160 participants. The estimated duration for each trial is \(160/8=20\) months, so the pair of parallel trials also takes about 20 months to reach both targets.
Now suppose the single multi-arm protocol can recruit 10 participants per month across all three arms and needs 240 participants. Its estimated duration is:
Although the multi-arm protocol has the smaller participant target, it takes longer under these assumed rates: about 24 months rather than 20. This does not show that multi-arm recruitment is generally slower. It shows that a lower target does not guarantee a shorter timeline if the feasible accrual rate also changes. If the trials cannot run concurrently, or if the protocol has more sites or a different recruitment pool, the comparison must be recalculated.
Operational Requirements and Practical Comparisons
A single protocol can coordinate eligibility criteria, outcome measurement, data systems, and follow-up across treatment arms. It may also simplify oversight when the treatments can be studied in the same population, at the same sites, and during an overlapping period. Conversely, it requires the protocol and sites to manage several treatment pathways at once, including allocation, treatment-specific safety procedures, supply, and staff training.
Separate trials can be more practical when treatments require substantially different eligibility criteria, care settings, follow-up schedules, or operational teams. Each trial can be organized around its own treatment and population. The cost of that independence may include separate control recruitment, duplicated setup, and less opportunity to coordinate study procedures. Which side matters more depends on the actual setting; there is no universal operational winner.
| Planning issue | Separate trials | One multi-arm protocol |
|---|---|---|
| Control recruitment | A distinct control group for each comparison | One common control group, if appropriate for all comparisons |
| Protocol and oversight | Separate trial documents and processes may be needed | One coordinated protocol, but it must manage multiple treatment pathways |
| Eligibility and setting | Can be tailored to each comparison | Works best when comparisons can share a suitable population and setting |
| Recruitment timeline | Depends on whether trials run concurrently and on each trial's accrual rate | Depends on total accrual capacity across all arms |
| Staff and treatment logistics | May duplicate some infrastructure, while keeping procedures focused | May coordinate infrastructure, while increasing the number of procedures managed together |
Assumptions, Diagnostics, and Limits
A fair comparison starts by checking that the inputs refer to comparable plans. In particular, verify that the per-comparison sample-size requirements use compatible outcome definitions, target effects, follow-up, and analysis assumptions. The simple counts also assume no reuse of controls between separate trials, no additional losses, and no need to enlarge any arm beyond the stated target.
Before committing to one arrangement, examine whether the common control is genuinely suitable for each treatment comparison. Ask whether the eligibility criteria and control care can be aligned, whether all arms can recruit in the same period, and whether sites can deliver each treatment safely and consistently. Recruitment projections should be checked against prior screening yields, the number of eligible people in the catchment area, site capacity, and expected competition from other studies. If observed screening or accrual rates fall below assumptions, update the timeline and recruitment target rather than relying on the original estimate.
The count comparison has important limits. It does not price protocol development, site start-up, treatment supply, monitoring, or data management. It does not establish that the same control group is scientifically appropriate for every comparison. Nor does it account for changes to sample-size needs when the analysis plan requires them. In particular, the counts here are recruitment illustrations, not a complete confirmatory trial design or a treatment of multiplicity across comparisons.
Check Your Understanding
Answer each question using the assumptions given. Short solutions follow the questions.
- Three treatment comparisons each require 50 control participants and 70 participants in the treatment arm. How many randomized participants are planned for separate trials and for one common-control protocol?
- In Question 1, how many participants are saved by the shared-control count, and what is that saving as a percentage of the separate-trial total?
- A multi-arm protocol requires 240 randomized participants. If 75% of screened candidates are expected to enroll, what is the approximate screening target?
- Why can a multi-arm protocol have a smaller participant target but a longer recruitment duration?
- Name two operational conditions that could make separate trials more suitable than one multi-arm protocol.
Solutions. (1) Separate trials require \(3(50+70)=360\) participants. One common-control protocol requires \(50+3(70)=260\). (2) The count saving is \(360-260=100\). As a proportion of the separate-trial total, \(100/360\approx0.2778\), or about 27.8%. (3) The approximate screening target is \(240/0.75=320\) people. (4) The protocol may have a lower feasible accrual rate across all arms; for example, its sites may recruit 8 people per month, giving \(240/8=30\) months, while separate trials could recruit concurrently at rates that reach their targets sooner. (5) Examples include substantially different eligibility criteria, incompatible care settings or follow-up schedules, or sites that cannot safely or consistently deliver all treatments under one protocol.