Simulation-Based Coverage of Pediatric Age Groups

Why Assess Coverage Across Pediatric Ages?

A pediatric trial can enroll its planned total number of participants and still provide little usable evidence for an intended age group. The youngest children may be difficult to recruit, participants may age into another band before dosing or follow-up, and age-specific eligibility or dose-assignment rules may limit who contributes data. Assessing only total enrollment can conceal these weaknesses.

Simulation-based coverage assessment asks whether a proposed design is likely to produce the required information across its intended age range. It represents the trial’s operational steps and important sources of variation, then repeats the simulated trial many times. The result is a probability of meeting clearly stated age-specific evidence criteria under the assumptions used—not a guarantee that any particular trial will succeed.

The criteria should be chosen before comparing designs. “Adequate evidence” is not a simulation output until it is made operational. For example, a protocol team might require at least 12 evaluable participants in each of three age bands, with at least five receiving each of two planned dose levels in each band. Those thresholds are design criteria, not universal standards; their appropriateness depends on the scientific question and the evidence needed.

Specify the Trial Before Simulating It

A useful simulation follows each hypothetical participant through the decisions and events that determine whether their data count. The design specification should distinguish age at screening, enrollment, dosing, and analysis. It should also state whether age groups are assigned from age at enrollment or age at another prespecified time. If a participant crosses an age boundary, the rule for classification must be explicit.

At minimum, specify the intended age range and bands, recruitment assumptions by age, enrollment and dose-assignment rules, age-related transitions, follow-up duration, and the definition of an evaluable participant. Include plausible sources of variation such as recruitment delays, screening failures, discontinuation, missed visits, and unavailable measurements when these affect the success criteria. If dose assignment depends on safety review or another decision rule, simulate that rule rather than assuming every participant receives a fixed dose.

The simulation relies on assumptions. The age distribution and recruitment rates should be plausible for the population and sites; transition and attrition mechanisms should reflect the protocol; and the simulated participant outcomes should represent relevant variability. A simulation can be internally correct yet misleading if these inputs are poorly supported. Earlier material on pediatric age-stratified diagnostics is useful when deciding whether model behavior differs across age groups, but the coverage assessment itself evaluates the trial design and its operational consequences.

A Binomial Check for One Age Group

A simple calculation can provide a useful check when the question is only whether enough participants from one age group will enroll. Suppose each of \(n\) independent enrolled participants has probability \(p\) of belonging to that age group, and let \(X\) be the number who do. Under these assumptions, \(X\sim\mathrm{Binomial}(n,p)\). This approximation does not represent transitions, dose decisions, or follow-up unless those are incorporated into the definition of success or modeled separately.

Joint Coverage Across Age Groups

For a full design assessment, define trial-level success jointly. For example, if the design requires at least a specified number of evaluable participants in each of three age groups, success occurs only when all three counts meet their thresholds. The groups need not be statistically independent: they may compete for a fixed enrollment total, share recruitment constraints, or be linked by age transitions. Simulating the complete trial preserves such dependencies when they are represented in the model.

For \(R\) simulated trials, let \(I_r=1\) if replicate \(r\) meets every criterion and \(I_r=0\) otherwise. The estimated design coverage is the proportion of successful replicates. Since the simulations themselves are random, this estimate has Monte Carlo error.

The expression for \(SE_{\mathrm{MC}}\) follows from treating the independent replicate indicators as Bernoulli variables with success probability \(C\). Their variance is \(C(1-C)\), so the variance of their mean is \(C(1-C)/R\). Substituting \(\widehat{C}\) for the unknown \(C\) gives the estimated Monte Carlo standard error. This quantifies simulation precision; it does not measure uncertainty in the assumptions or model inputs.

Represent Dose Assignment and Follow-Up

Age-group counts at enrollment are often only an intermediate measure. A participant contributes to the intended evidence only if the protocol assigns a dose and the required follow-up or measurements are obtained. The simulation should apply the actual planned dose-assignment rule, including any age-dependent assignment, safety gate, or stopping condition. It should then apply the protocol’s follow-up and evaluability rules before counting participants toward success.

The next example illustrates the difference between enrollment coverage and evidence coverage. The simulated rates below are hypothetical design outputs; they are not observations from a real trial.

Diagnostics, Alternatives, and Limitations

A coverage estimate is most informative when accompanied by diagnostics. Inspect the distribution of evaluable counts by age group, the frequency of each failure mode, and age at enrollment and dosing. Check whether participants cross boundaries, whether age groups are classified consistently, and whether dose assignment or follow-up differs systematically by age. Compare simulated totals and transition rates with the assumptions used to generate them. Unexpectedly empty bands, extreme count distributions, or nearly identical results despite major changes in recruitment assumptions can signal coding or specification errors.

Use scenario analyses to examine uncertainty that is not resolved by more simulation replicates. For example, vary recruitment success in the youngest group, screening failure, transition timing, or follow-up completion over plausible ranges. A larger \(R\) reduces Monte Carlo error, but it cannot correct unrealistic inputs. Report the assumptions and scenarios alongside the estimated coverage so that readers can distinguish simulation precision from uncertainty about how the trial will actually operate.

Analytic binomial calculations are efficient and transparent for a narrowly defined enrollment count under independent, constant-probability assumptions. They make useful checks, as in the first example, but usually cannot capture linked age groups, changing eligibility, dose rules, and follow-up together. Deterministic scenarios can show how specified changes affect expected counts, but do not by themselves quantify the probability of meeting a criterion when trial outcomes vary. Simulation is more flexible, at the cost of requiring more assumptions, careful implementation, and Monte Carlo precision checks.

Several limitations should remain visible in interpretation. The estimated coverage applies to the simulated population, protocol, and model assumptions; it does not establish that the criteria themselves are scientifically sufficient. Misspecified recruitment, transition, or attrition mechanisms can bias the estimate. A single joint probability may hide which age group or process is limiting success, so report age-specific summaries as well. Finally, simulated evidence coverage is not a substitute for evaluating the scientific adequacy of the dose or the validity of a pharmacometric model.

Common Mistake

• Wrong reasoning: “The trial has a 95% chance of enrolling at least three children in the youngest group, so it has adequate coverage across the pediatric age range.”
• Correction: That binomial calculation concerns one group’s enrollment count under specific assumptions. Joint age-range coverage must be defined across all intended groups and must include the relevant dose-assignment and follow-up requirements. Estimate that joint probability using a simulation that represents those processes, and report the assumptions and failure patterns.