Start With the Study’s Requirements
Selecting a crossover layout is not simply a matter of choosing the schedule with the most periods. A longer schedule can repeat treatments within participants, but it also requires more visits and more treatment transitions. The useful question is: what is the least burdensome schedule that meets the study’s scientific requirements and has credible assumptions?
For a two-formulation comparison, let T denote the test formulation and R the reference formulation. A sequence is the ordered schedule assigned to a participant. For example, TR means that the participant receives T in period 1 and R in period 2. A period is one treatment occasion; it is not the interval between treatments. A crossover schedule assigns each participant both formulations, in a planned order.
The schedule is one part of the design, not a guarantee of valid interpretation. Crossover comparisons rely on participants being suitable for repeated treatment periods, on measurements being collected comparably, and on period-to-period changes being manageable. In particular, the intended washout should make residual effects of a previous treatment negligible for the outcome being measured. If that is not plausible, adding periods or balancing sequences does not by itself remove the concern.
A Practical Selection Framework
First establish what the comparison requires. Is one observation per formulation per participant enough for the planned comparison, or must a formulation be repeated within participants? Is there a reason to repeat both formulations, or only R? How many periods can participants reasonably complete? Finally, consider whether the drug and outcome make negligible carryover after the planned washout plausible. These questions lead to a choice among layouts; they do not replace the protocol’s scientific justification.
Consider whether participants can receive both formulations in separate periods and whether an adequate washout is scientifically plausible. If not, do not treat a more elaborate sequence as a fix.
If each participant needs one exposure to each formulation, a two-period layout may suffice. If R must be repeated, consider a partial replicate; if both formulations must be repeated, consider a full replicate.
Count how often each formulation occurs in each period under the planned sequence allocation. Also inspect treatment transitions and the number assigned to each sequence.
More periods mean more visits and more opportunities for withdrawal or incomplete data. Choose the simplest layout that still meets the scientific and operational requirements.
Balance is a property of the planned schedule and allocation, not proof that all participants will complete it or that period effects are absent. An allocation plan should specify how participants are assigned across sequences; actual enrollment and completion should be monitored against that plan. The earlier material on enrollment balance and randomization lists covers those operational details.
Worked Layout Comparisons
Worked Example: A Two-Period Comparison With No Replication Requirement
Hypothetical scenario. A study needs each participant to receive T and R once. There is no stated need for within-participant replication, and a two-period schedule is practical. The team plans to enroll 24 participants, allocating 12 to TR and 12 to RT.
In sequence TR, each participant receives T in period 1 and R in period 2. In RT, the order is reversed. Across 12 participants in each sequence, period 1 has 12 T exposures and 12 R exposures; period 2 also has 12 T exposures and 12 R exposures. The arithmetic is \(12\) participants assigned TR plus \(12\) assigned RT, so each treatment occurs \(12\) times in each period.
Each participant has two periods and one exposure to each formulation. This layout meets the stated comparison requirement with the fewest periods among the layouts considered here. It does not provide repeated exposure to either formulation within a participant. The team should still justify the washout and assess whether the planned procedures are comparable across periods; equal sequence allocation does not establish either condition.
Decision. Given these requirements, the two-period layout is a reasonable choice. Adding replicate periods would increase participant burden without meeting a specified repetition requirement in this scenario.
Worked Example: Repeating the Reference Formulation
Hypothetical scenario. A study needs each participant to receive R twice and T once, but does not require repeating T. The team considers the three-period sequences TRR, RTR, and RRT, with 18 participants allocated equally: 6 per sequence.
The total number of treatment occasions is \(18 \times 3 = 54\). Each participant receives T once, so there are \(18 \times 1 = 18\) T occasions. Each receives R twice, so there are \(18 \times 2 = 36\) R occasions. The repetition requirement is met for every participant, not merely in the study total.
Check the period distribution. In period 1, TRR contributes 6 T and RTR plus RRT contribute \(6+6=12\) R. In period 2, RTR contributes 6 T and TRR plus RRT contribute \(6+6=12\) R. In period 3, RRT contributes 6 T and TRR plus RTR contribute \(6+6=12\) R. Thus, each period contains 6 T and 12 R occasions. Treatment proportions are the same in every period, although the raw counts differ because participants receive R more often.
The three sequences also distribute the adjacent transitions evenly in this allocation. TRR contains one TR and one RR transition; RTR contains one RT and one TR; RRT contains one RR and one RT. With 6 participants per sequence, the totals are 12 TR, 12 RR, and 12 RT transitions. This describes the planned schedule; it does not prove that carryover is absent.
Decision. The partial-replicate layout meets the requirement to repeat R, while using one fewer period than the four-period full-replicate layout. The team should choose it only if repeating R, rather than T, is the actual scientific need and the three-period burden is acceptable.
Worked Example: Repeating Both Formulations
Hypothetical scenario. A study requires each participant to receive T and R twice. The team plans a four-period full replicate using TRTR and RTRT, with 24 participants: 12 assigned to each sequence.
Each participant receives four treatments, so the study plans \(24 \times 4 = 96\) treatment occasions. Each receives T twice and R twice, giving \(24 \times 2 = 48\) occasions for each formulation. In every period, the TRTR sequence contributes 12 participants receiving the treatment scheduled there, and the RTRT sequence contributes 12 receiving the other treatment. Therefore, each period has 12 T and 12 R occasions.
The planned transitions are balanced as well. Each TRTR participant has two TR transitions and one RT transition; each RTRT participant has two RT transitions and one TR transition. Across the 12 participants per sequence, the totals are \(24+12=36\) TR transitions and \(12+24=36\) RT transitions.
Decision. The full replicate meets the requirement to repeat both formulations and balances treatment counts across periods under equal sequence allocation. It requires four periods, however, and its schedule cannot establish that washout is adequate. If repeated exposure to both formulations is not needed, the added periods may be an unjustified burden.
Worked Example: When Carryover Makes a Crossover Doubtful
Hypothetical scenario. A formulation has a long-lasting effect on the measured outcome, and the team cannot justify a washout that would make that effect negligible. The team considers enrolling 30 participants equally into TR and RT, or using a replicate schedule to add observations.
Equal allocation would give 15 participants per sequence. That balances treatment order in the planned two-period schedule, but it does not change the fact that a participant’s period-2 measurement could still reflect the period-1 treatment. A replicate schedule would create additional periods and transitions, so it would create more occasions on which residual effects must be considered; it would not make those effects disappear.
Decision. The central problem is not a shortage of periods or unequal sequence counts. It is the lack of a credible basis for assuming negligible carryover. The team should not select a crossover merely because a schedule is balanced; it should reconsider the design, including whether a parallel-group approach is more appropriate. That alternative has its own assumptions and is not a way to repair the crossover schedule.
Why Balance Helps—and What It Cannot Do
Period balance helps avoid a design in which one treatment is always given early and the other always given late. For example, if every participant received T in period 1 and R in period 2, a systematic period change could be entangled with the treatment contrast. With both TR and RT sequences represented, each formulation occurs in each period. This is a design-based protection against a simple treatment-period imbalance, not a guarantee that period effects are absent or that every possible source of bias is controlled.
Replication changes what a participant contributes. A partial replicate repeats one formulation; a full replicate repeats both. That can be relevant when a study needs repeated measurements within a participant, but repetition adds periods, treatment exposures, and opportunities for dropout. The design should therefore name the requirement it is intended to meet rather than assume that more observations are automatically better.
Carryover requires separate attention. A balanced set of sequences can distribute the order of treatments, but it cannot guarantee that a previous treatment has no residual effect. The team must judge the plausibility of negligible carryover from the drug, outcome, and washout plan before selecting the crossover. If that assumption is doubtful, simply adding replicate periods does not resolve it.
Common Mistakes and Diagnostics
- Mistake: Choosing the most replicated schedule by default. More periods add burden and may not serve a defined requirement. Correction: state which formulation must be repeated, and choose the least burdensome layout that meets that requirement.
- Mistake: Treating equal sequence numbers as proof against carryover. Balance describes allocation; carryover concerns residual treatment effects. Correction: justify washout independently and do not rely on schedule balance to establish negligible carryover.
- Mistake: Checking total treatment counts but not period counts. Equal overall numbers of T and R do not ensure that the treatments are distributed similarly across periods. Correction: make a period-by-treatment count for the proposed sequence allocation.
- Mistake: Assuming the planned balance will persist in the completed study. Withdrawals can leave fewer participants in one sequence. Correction: track enrollment and completion by sequence and assess the actual data before interpreting the study.
- Warning signs: one sequence is poorly represented; treatment counts differ sharply by period without a scientific reason; completion is low in later periods; or the washout justification does not address the outcome’s persistence. These findings call for review of the design or its conduct, not an automatic claim that the treatment comparison is invalid or valid.
The main alternatives here are the two-period, partial-replicate, and full-replicate crossover schedules. They differ in period count and within-participant repetition, not in whether crossover assumptions matter. A parallel-group design is a broader alternative when repeated treatment or carryover makes crossover unsuitable, but selecting it requires evaluating its own design and analysis requirements.
Check Your Understanding
Answer each question by linking the schedule to the requirement and its assumptions.
- A study needs every participant to receive T and R once, and there is no specified need for repeated exposure. Which layout is the simplest candidate?
- In a three-period TRR/RTR/RRT design with 9 participants per sequence, how many R treatment occasions are planned?
- For 12 participants in each of TR and RT, how many T exposures occur in period 1?
- Why does equal allocation to TR and RT not establish that carryover is negligible?
- A team needs repeated exposure to both formulations but wants to minimize periods among the layouts defined here. Which layout meets that requirement, and what trade-off remains?
Solutions. (1) A two-period crossover is the simplest candidate because each participant receives each formulation once. (2) There are 27 participants, each receiving R twice, so \(27 \times 2 = 54\) R occasions. (3) The 12 TR participants receive T in period 1, while the 12 RT participants receive R, so there are 12 T exposures in period 1. (4) Equal allocation balances treatment order, but it does not show that a prior treatment has no residual effect after washout. (5) The four-period full replicate repeats both T and R; the trade-off is the extra period and its associated participant burden and completion risk.