Why Assess Adaptations as a Package?
A trial may propose several changes at once: for example, narrowing the eligible population, changing the allocation ratio, increasing the enrollment target, and extending follow-up. Each change may sound reasonable in isolation. Together, however, they can create a plan with conflicting aims, missing decision steps, or demands that the trial cannot meet with its available staff and time.
Assessing adaptations together means examining the proposed changes as a connected plan rather than as a list of independent choices. The central questions are whether the package has a coherent purpose, whether each change can be implemented in the correct order, and whether the combined requirements fit the trial’s practical constraints. A feasibility calculation can reveal a shortfall under stated assumptions; it does not automatically establish that a target is impossible.
A useful assessment begins by linking changes that affect the same part of the trial. A population change can alter how quickly eligible participants are found. That change can then affect the enrollment target and the time needed to complete recruitment. A new allocation ratio affects the number assigned to each treatment, and may need to be coordinated with a sample-size change. A revised outcome or follow-up schedule can affect which data are available when an interim decision is due.
A Practical Review of the Combined Plan
For each proposed change, record its purpose, trigger, effective date, and consequences for participants already enrolled and those enrolled later. Then examine the links between changes. The aim is not to make every adaptation simple; it is to make the whole decision path explicit enough that the trial team can carry it out consistently.
Describe the problem the combined changes are meant to address. Check whether the changes support that purpose or quietly introduce different, competing aims.
Ask what information must be available before each change can occur, and whether one change alters the data or participants needed for another.
Identify who makes each decision, when the change takes effect, and how it applies to participants in each enrollment phase.
Compare recruitment, visits, data handling, staff time, and treatment supply with stated limits. Make sure the resulting plan can be explained consistently to the trial team and participants.
This review is different from checking each change separately. An individual change might be possible while the combination is not. For example, extending follow-up may be manageable at the original enrollment rate, and increasing enrollment may be manageable with the original visit schedule, but doing both together may exceed clinic capacity.
Worked Example: Population, Allocation, and Enrollment Changes
Hypothetical scenario. A trial plans to enroll 180 participants with a broad eligibility definition, allocate them equally to two treatments, and assess the primary outcome at week 12. A proposed interim package would restrict future enrollment to participants with a specified biomarker, use a 2:1 allocation favoring the experimental treatment for future participants, and raise the total enrollment target to 240. Week-12 outcome assessment would remain unchanged.
The package’s stated purpose is to focus future enrollment on the biomarker group while increasing the number assigned to the experimental treatment. The first review question is whether these changes fit together operationally. The biomarker restriction may reduce the proportion of screened people who qualify; therefore, raising the enrollment target at the same time increases the number of eligible participants the trial must recruit. The allocation change also applies only to future participants unless the protocol explicitly says otherwise.
Suppose 60 participants have already been randomized equally before the adaptation. They comprise 30 participants per treatment. If the remaining 180 participants are allocated 2:1, then the future allocation is 120 to the experimental treatment and 60 to the comparison treatment. The final totals would therefore be 150 and 90, respectively:
Thus, the 2:1 ratio describes allocation after the adaptation, not the final ratio across all 240 participants. If the intended rule is instead to achieve an overall 2:1 ratio, the future allocation would have to be specified differently. This distinction should be resolved in the protocol rather than left to staff to infer.
The package is not fully assessed until recruitment is considered as well. If the biomarker restriction lowers eligibility, a target of 240 may require more screening or a longer recruitment period. The team should estimate the eligible fraction using a stated source and check whether screening capacity and the available calendar time support the resulting plan. This example does not establish that the package is feasible: it identifies linked quantities that must be checked and a ratio definition that must be clarified.
Expected Recruitment Is Not a Guarantee
A common first check is to estimate how many eligible participants might be found among a planned number of people screened. Let \(I_i\) equal 1 if the \(i\)th screened person is eligible and 0 otherwise. If each of \(m\) screened people has eligibility probability \(p\), the expected number eligible is the expected value of the sum of these indicators:
The expectation calculation uses the same eligibility probability for each person; independence is not needed for the equality \(E[X]=mp\). If, in addition, eligibility outcomes are independent, then \(X\) has a binomial distribution, \(X\sim\operatorname{Binomial}(m,p)\). That stronger model can be used to calculate the probability of meeting a target, but it is a separate question from comparing the expected count with the target.
Worked Example: A Shortfall in Expected Eligible Participants
Hypothetical scenario. A trial expects to screen 120 people before a recruitment deadline. Based on preliminary screening information, the planning assumption is that 40% will meet a new eligibility criterion. The trial needs 60 eligible participants.
Using the expected-count calculation, \(m=120\) and \(p=0.40\), so:
The expected supply of 48 is 12 below the target of 60. Therefore, the plan does not meet the target on the expected-rate calculation. This result does not prove that fewer than 60 participants will be eligible: the realized count can differ from its expectation.
If the team wants to assess the probability of reaching at least 60, it must specify a probability model. Under the additional assumption that the 120 eligibility outcomes are independent and each has probability 0.40, \(X\sim\operatorname{Binomial}(120,0.40)\), and the probability in question is:
That probability should be evaluated under the stated model before making a probability-based feasibility claim. Alternatively, 150 screenings give an expected count of \(150(0.40)=60\), but that still does not guarantee 60 eligible participants. Recruitment planning should distinguish a target expected count from a desired probability of meeting the target.
Check Implementation Against Shared Constraints
The same principle applies to time and staffing. A package can change the number and timing of tasks, not just the enrollment total. Check whether interim data will be ready when the decision is scheduled, whether staff can deliver the revised visit schedule, and whether responsibilities are clear during the transition. These are checks on the combined plan: each adaptation may be administratively manageable alone but create a bottleneck when introduced with others.
Worked Example: Follow-Up Visits and a Treatment-Arm Change
Hypothetical scenario. At an interim review, a trial proposes to stop assigning new participants to one dose, change future allocation between the remaining dose and control to 2:1, and extend follow-up for every participant from 12 to 24 weeks. The clinics have capacity for 20 week-24 visits per month. Based on current enrollment and the proposed schedule, 24 week-24 visits per month are expected.
The arm change and allocation change must be described together: the revised ratio applies only to the treatments still open to new assignment, and the rule must state what happens to participants already assigned to the dose that will close. Separately, extending follow-up affects both existing and future participants if it applies to everyone. The package therefore needs a clear effective date and a clear account of which participants receive the additional visits.
The stated monthly workload exceeds the stated capacity by \(24-20=4\) visits. On these assumptions, the plan requires four more week-24 visits per month than the clinics can currently provide. It is not feasible within the stated capacity unless capacity is increased or the schedule is revised. That conclusion is conditional on the scenario’s workload and capacity figures; it is not a claim that the package could never be implemented with additional resources.
The review also exposes a communication issue: the participant materials and clinic instructions must not imply that participants already assigned to the closing dose will be reassigned if that is not the plan. A package can be operationally possible and still be poorly specified if its consequences for current participants are unclear.
Diagnostics, Alternatives, and Limitations
Several warning signs indicate that a package needs further work before it can be judged understandable and feasible:
- The stated purpose is vague, or different changes appear to serve unrelated goals.
- A trigger depends on data that will not be available by the proposed decision time.
- Two changes imply different effective dates, or the rules do not say which takes precedence.
- The protocol does not distinguish current participants from those enrolled after an adaptation.
- Recruitment or visit demand exceeds a stated expected rate or hard capacity limit, with no plan to address the gap.
- Different trial staff could reasonably interpret the same decision rule in different ways.
A one-change-at-a-time review is useful for understanding the purpose and consequences of each adaptation, but it can miss dependencies. A combined-package review is the appropriate next step: trace the links and calculate the total demands. For a complex package, a written decision table or process diagram can make the sequence easier to inspect. Formal statistical evaluation may also be needed for a trial’s inferential properties; an operational feasibility review does not replace that evaluation.
These checks have limits. Recruitment rates may change over time, capacity estimates may be uncertain, and a clear written rule does not ensure that implementation will go as planned. Record the assumptions behind each calculation, distinguish projected quantities from hard constraints, and identify what monitoring would reveal that the assumptions are no longer holding.
Common Mistake
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A proposed package narrows eligibility and increases the enrollment target. What recruitment quantity should be checked, and why is it not enough to inspect each change alone?
- A trial expects to screen 100 people, with an eligibility rate of 0.30. Its target is 40 eligible participants. Calculate the expected eligible count and state what the result does and does not establish.
- Before an adaptation, 24 participants are allocated equally to two treatments. Afterward, 90 additional participants are allocated 2:1. Calculate the final treatment totals and explain what the 2:1 ratio describes.
- A follow-up extension requires 18 visits per month, while staffed capacity is 16. State the feasibility conclusion supported by these figures and one possible response.
Solutions. 1. Check the number of eligible participants expected within the available screening capacity and recruitment time. Narrower eligibility may reduce the eligible fraction, while the increased target raises demand, so the combined requirement matters. 2. \(E[X]=100(0.30)=30\). The expected count is 10 below the target, so the plan does not meet the target on the expected-rate calculation; it does not prove that 40 cannot be reached. A probability-based conclusion would require a specified model and probability calculation. 3. The initial allocation is 12 per treatment. Of the 90 later participants, 60 go to the treatment favored by the 2:1 rule and 30 to the other. Final totals are 72 and 42. The 2:1 ratio describes post-adaptation allocation, not the overall totals. 4. Under the stated figures, demand exceeds capacity by \(18-16=2\) visits per month, so the plan does not fit the stated capacity unless it changes. One response is to add capacity; another is to revise the schedule, provided that revision is compatible with the trial plan.