Repeated Eligibility in Target Trial Emulation: When One Patient Quietly Enters the Trial Again
Anas H. Alzahrani, MD PhD MPH
Department of Preventive Medicine and Public Health
Faculty of Medicine, King Abdulaziz University
Many target trial emulations ask a clinically recurring question and then analyze it as if it happened only once. A patient is eligible at baseline, does not start treatment, remains eligible next month, and still faces the same decision. The temptation is obvious: why not let that patient enter the emulated trial again?
Sometimes that move is exactly right. Sometimes it is just a clever way to recycle the same trajectory into several pseudo-trials. The difference is not the software. It is whether later eligibility windows represent genuinely renewed treatment decisions with explicit rules for re-entry, carryover, and overlapping follow-up.
The Core Design Rule
Repeated eligibility is defensible only when each new entry can be described as a fresh trial decision in protocol language rather than a convenient second chance to count the same untreated time again.
Decision rule:
If you cannot say exactly who may re-enter, when they may re-enter, and why prior exposure does not contaminate the next window, you probably do not have repeated eligibility. You have repeated reuse.
The practical question is not whether a person appears twice in the data. It is whether the second appearance still represents a coherent comparison of strategies that a trial protocol could have stated in advance.
What Repeated Eligibility Is Trying to Rescue
Recurring treatment decisions
In chronic disease care, a patient may be eligible to start therapy at many visits, not just the first one.
Avoiding arbitrary baseline worship
Using only the first eligible moment can discard later clinically relevant decisions and overstate how special baseline really was.
Preserving the “initiate now” question
Sequential eligibility windows can better match the real policy question: start today versus defer today among patients eligible today.
That logic is strong. The danger is that researchers start loving the extra rows more than the extra design discipline.
A Concrete Clinical Example
Case
Monthly outpatient eligibility to start an SGLT2 inhibitor in heart failure
Imagine an EHR study asking whether patients with heart failure should start an SGLT2 inhibitor at a given outpatient month. A patient may qualify in January, remain untreated, qualify again in February, and still be clinically eligible in March. A single-baseline emulation throws away those later decisions.
Repeated eligibility may be reasonable here if each month represents a fresh initiation choice and the protocol says what happens after a patient starts treatment, stops treatment, or experiences an outcome. Without those rules, the study slides into a blur of start-now, start-later, and never-start without admitting which comparison it actually estimated.
Interactive repeated-eligibility triage
Decide whether you are emulating fresh treatment decisions or just recycling the same patient
This teaching tool does not replace a protocol. It forces the four design questions repeated-eligibility papers most often skip: how often the trial restarts, how long one emulated trial stays active, whether prior treatment carries forward, and what lets a patient re-enter.
Repeated eligibility can be defensible with explicit reset logic
Recurring eligibility can support a realistic initiation question when each new entry represents a clinically fresh decision and prior treatment history is handled explicitly.
Design move
State the cadence of trial emulation, define the washout or reset rule, cluster variance at the patient level, and show that carryover is clinically limited.
Reviewer cue
The key question is not whether the patient appears twice, but whether each appearance corresponds to a genuinely renewed treatment decision.
| Question | Why it matters |
|---|---|
| Can the same patient start a clinically fresh trial later? | Repeated eligibility is only honest when a later window corresponds to a renewed decision, not leftover consequences of an earlier one. |
| Does one trial’s follow-up overlap the next window? | Overlap can inflate apparent sample size while the same person contributes multiple correlated risk periods. |
| Can prior treatment still affect current outcome risk? | If yes, later “new” trials may inherit carryover unless the protocol defines washout or blocks re-entry. |
| Are standard errors clustered at the patient level? | Even a sensible repeated-eligibility design still reuses people. Precision claims should admit that dependence rather than pretending every row is new. |
Where Repeated Eligibility Usually Breaks
| Approach | Best use | Strength | Risk |
|---|---|---|---|
| Single baseline emulation | A one-time initiation decision, or a setting where later eligibility windows are clinically rare or not meaningfully fresh. | Simple estimand, one clear time zero, and less danger of overlapping follow-up. | Can waste information when patients remain untreated but truly face the same treatment decision again later. |
| Repeated eligibility with explicit reset rule | A recurring initiation question where later windows represent renewed clinical decisions and prior treatment history is handled explicitly. | Uses more of the observable decision process instead of pretending only the first eligible moment mattered. | Needs careful rules for re-entry, patient-level dependence, carryover, and the length of each outcome window. |
| Repeated eligibility without clear re-entry logic | Usually none beyond making the dataset look larger. | It can create the illusion of precision and realism. | The same patient gets recycled into pseudo-independent trials while prior exposure and overlapping follow-up contaminate the estimand. |
The Five Failure Modes Reviewers Should Hunt First
Failure mode 1
The paper treats every eligible month as a brand-new patient
Repeated rows from one person are described as if they were independent entrants into many small trials. That flatters sample size and hides within-person dependence.
What to demand instead
Report how many unique patients contributed more than one trial entry, and show how variance estimation or resampling respected clustering at the patient level.
Failure mode 2
Re-entry is allowed even though prior treatment can still matter
If treatment started in an earlier window can still influence later prognosis, later eligibility windows are not clean restarts. The new trial inherits carryover from the old one.
What to demand instead
Either block re-entry after initiation, define a washout or reset rule, or justify clinically why prior exposure cannot meaningfully affect the later outcome window.
Failure mode 3
Outcome windows overlap so one clinical trajectory counts multiple times
A 180-day outcome horizon plus monthly trial starts means several active follow-up periods can coexist for the same person. One hospitalization can end up influencing multiple emulated trials.
What to demand instead
Declare whether overlapping follow-up is allowed, and if it is, explain why the estimand still matches a sensible policy question. Otherwise shorten follow-up or restrict new entry while a prior trial window is still active.
Failure mode 4
Eligibility is repeated, but treatment strategies are not
Authors say they emulated “start now versus do not start now,” but in practice the comparison becomes a moving mix of defer-now, start-later, and never-start, with no protocol language to separate them.
What to demand instead
Write the treatment strategies in trial language for each window. If deferred initiation is allowed, say when and how that affects censoring, adherence, or the estimand.
Failure mode 5
Repeated eligibility is used because the first trial looked underpowered
Design complexity becomes a rescue mission for power rather than an answer to the real clinical decision structure.
What to demand instead
The protocol should make repeated eligibility defensible before looking at results. More rows are not an estimand.
What a Credible Protocol Should State Explicitly
- The cadence of trial emulation: every visit, every month, every quarter, or another predefined window.
- Who can re-enter after a prior trial window, including whether initiators are permanently ineligible afterward.
- Whether follow-up windows may overlap for the same person, and if so, why that still answers a coherent estimand.
- How prior treatment, discontinuation, or washout changes later eligibility.
- How inference respects repeated contributions from the same patient.
Those are not supplemental details. They are the design itself. A repeated-eligibility study that hides them is like a crossover trial that never explains the washout.
Reviewer Red Flags
- Does each later eligibility window represent a genuinely renewed treatment decision, or is the patient just lingering in the dataset untreated?
- Are re-entry rules explicit for both initiators and non-initiators, including whether prior treatment blocks later entry?
- Can one patient contribute overlapping follow-up windows, and if so, do the authors explain why that still answers a coherent policy question?
- Did the analysis handle patient-level dependence, rather than treating every emulated trial row as a new participant?
- Is the comparison really “initiate now versus defer now,” or has the paper blurred together immediate, delayed, and never-treated strategies?
The Practical Bottom Line
Repeated eligibility is useful because clinical decisions recur. It is dangerous because data rows recur even more easily. The right design does not ask whether a person could appear again in the spreadsheet. It asks whether a later appearance still corresponds to a trial that a clinician, patient, and reviewer would all recognize as genuinely new.
If your team needs help stress-testing whether a target trial emulation is answering a fresh decision question or just repackaging overlapping follow-up, Aqrab can help review the protocol logic before the analysis starts looking more precise than it really is. You can explore that in Aqrab.
Keep reading
Don't stop at one method.
Good methods judgment comes from contrast. Read the neighboring guides, see where the assumptions diverge, and avoid treating every observational problem like it needs the same hammer.
Grace Periods in Target Trial Emulation: Clinical Realism or Future Information in Disguise?
A practical guide to grace periods in target trial emulation for clinical researchers. Covers when a grace window is defensible, when it becomes immortal time in formalwear, and what reviewers should demand before trusting the result.
Informative Cluster Size: When the Biggest Sites Start Writing the Result
A practical guide to informative cluster size for clinical researchers. Covers why larger centers can quietly dominate treatment effects, how weighting changes the estimand, and what reviewers should demand before trusting clustered results.
Nested Case-Control Design: When Cheap Control Sampling Still Has to Respect Event Time
A practical guide to nested case-control design for clinical researchers. Covers incidence-density sampling, risk-set control selection, how it differs from case-cohort design, and what reviewers should demand before trusting the result.