Why Scope Begins With Intended Use
A QSP model does not represent every patient, treatment, or clinical circumstance that could be associated with a disease. It represents a defined system for a defined purpose. The same biological disease can call for different model boundaries when the decision changes: estimating an effect in a trial-like population is not the same task as selecting a dose for patients with organ impairment or evaluating a therapy after prior treatment failure.
The earlier material on decision questions and model output boundaries established that the decision determines which outputs matter. The next step is to specify the patients, treatment strategies, and care circumstances to which those outputs are intended to apply. These boundaries are not just administrative labels. They determine which evidence is relevant and where a model’s conclusions should not be used.
Specify the Model’s Intended Domain
A useful scope statement ties the decision to three explicit questions: Which patients? Which treatment strategies? Under what care conditions? If any answer is left implicit, two analysts may build different models while believing they are answering the same question.
For precision, let \(D\) denote the set of decision scenarios for which conclusions are sought, and let \(M\) denote the set of scenarios the model is defined and supported to represent. A scenario includes the patient characteristics, treatment strategy, and care context relevant to the use. A basic scope requirement is:
This condition means that every intended decision scenario is within the model’s stated domain. It does not mean that every scenario in \(M\) is equally well supported by evidence, or that the model should cover every imaginable patient. Scope must be both sufficient for the decision and defensible from available knowledge. An intentionally narrow domain can be appropriate; a domain that excludes a patient group central to the decision cannot answer that decision.
The modeled population is not necessarily identical to the sample enrolled in a clinical study. A study sample is the group actually observed; a target population is the group to which an analysis seeks to apply; and the modeled population is the group represented by the model. Those groups may overlap without being identical. A model may use evidence from several sources, but it should state which patient characteristics are held within the intended range and which are outside it.
Patient boundaries should use characteristics that affect the intended application, not a long list of convenient descriptors. Depending on the question, relevant characteristics can include disease subtype or severity, age range, biomarker status, organ function, prior treatment, or comorbidity. Some characteristics may be represented as subgroups within the modeled population; others may be explicit exclusions. The decision determines which distinction is important.
A Practical Scope-Setting Sequence
Name what will be compared or selected and what the model output will inform. Avoid substituting a broad topic such as “understand the disease” for a decision use.
Specify defining disease features, treatment history, and patient characteristics that could change the decision or the applicability of the evidence.
State the regimens or strategies being evaluated, what background care accompanies them, and the clinical setting or treatment stage assumed.
Ask whether the intended scenarios fall within the model’s stated domain and whether evidence and assumptions support using the model there. Record exclusions and uncertainties rather than silently extending the scope.
The treatment boundary needs more than a drug name. Dose, schedule, route, duration, combination partners, and comparator can matter. “Treatment A versus usual care” is incomplete if usual care differs substantially by setting or if the intended question concerns add-on treatment rather than replacement. Likewise, the care context should identify relevant background therapy and treatment history when those conditions affect the interpretation of the comparison.
Worked Example: Add-On Therapy in a Defined Asthma Population
Hypothetical question: A development team wants to compare adding an investigational biologic with continuing standard controller therapy alone in adults with severe eosinophilic asthma whose disease remains uncontrolled. Which patients and care conditions belong in scope?
Step 1: Define the intended patients. The decision is explicitly about adults with severe eosinophilic asthma that remains uncontrolled despite controller treatment. Children, patients with mild asthma, and patients without the stated eosinophilic phenotype are not automatically part of this modeled population. They could become relevant under a different decision, but their inclusion would require a justified extension.
Step 2: Define the strategies. The comparison is investigational biologic added to controller therapy versus the same controller-therapy background without the biologic. It is not a comparison of biologic monotherapy with no treatment. Treating the comparator as “no therapy” would change the clinical question.
Step 3: Define the care context. The assumed setting is ongoing outpatient management, with background controller treatment maintained in both groups. The scope statement should also clarify any relevant rescue medication or other allowed concomitant care if those conditions differ between the intended use and the supporting evidence.
Result. A suitable scope statement is: “The model is intended to compare addition of the investigational biologic with no addition of that biologic, on a specified controller-therapy background, in adults with uncontrolled severe eosinophilic asthma in outpatient care.” The statement does not establish that the model is valid for every patient meeting these labels; it makes the intended domain explicit so support can be assessed.
Why a Changed Use Changes the Population
Population boundaries are purpose-dependent. If a model is initially developed for a controlled comparison in a selected trial-like group, that does not establish that its output applies unchanged to a broader routine-care population. Conversely, if the intended use is a population-level planning decision, a narrow trial-matched group may omit patient types that materially affect the decision. The correct boundary is therefore not “the broadest group available” or “the group in the most convenient data set.” It is the group required by the decision, with an explicit account of evidence limitations.
Worked Example: The Same Cancer Drug at Different Treatment Stages
Hypothetical question A: A model is being used to compare two first-line combination regimens in adults with a specified biomarker-positive cancer who have not received systemic treatment for advanced disease. The modeled population should reflect that disease and biomarker definition and the stated treatment history. The treatment boundary consists of the two first-line combinations, and the care context includes the background oncology care assumed for both.
Hypothetical question B: The team now wants to use a model output to consider a single-agent therapy after progression on a prior regimen. The intended patients have changed: they have progressed after prior treatment, and their treatment history is part of the decision scenario. The treatment strategies and care context have also changed. The first-line model’s scope cannot simply be relabelled to cover this use.
Scope check. Let \(D_A\) be the first-line decision scenarios and \(M_A\) the model domain defined for them. If the new post-progression scenarios are not represented in \(M_A\), then \(D_B \nsubseteq M_A\). The model may still help identify questions for the new use, but its existing output is not thereby evidence for the new population and treatment context. The appropriate response is to justify and evaluate an extension, or to state that the current model does not answer question B.
This distinction is not a claim that all later-line patients differ in every relevant respect. It is a scope safeguard: a changed treatment stage changes the scenario that must be justified, rather than presumed equivalent.
Care Context Is Part of the Boundary
A patient group and a regimen do not fully specify a clinical scenario. Standard care, concomitant treatments, treatment setting, and prior interventions can affect how a proposed comparison should be interpreted. For example, the same regimen evaluated on different background therapy is not necessarily the same treatment strategy. “Routine care” is especially important to define: practices can differ across regions, institutions, and calendar periods.
Care context should be described at the level needed to distinguish the intended use, not as an exhaustive catalogue of every clinical detail. If a feature is not expected to affect the decision or model application, it may not need to define a boundary. If it could alter the relevant comparison or make the evidence inapplicable, leaving it unspecified creates ambiguity.
Worked Example: Dose Selection in Renal Impairment
Hypothetical decision: A team wants to assess whether a lower starting dose should be considered for adults with moderate renal impairment receiving Drug Z for a chronic inflammatory disease. The intended care context is outpatient treatment with a named background therapy; dialysis patients are outside the question.
Suppose a proposed scope statement says only “adults with the inflammatory disease receiving Drug Z.” It fails to distinguish the group central to the decision (moderate renal impairment), does not state the background therapy, and does not make the dialysis exclusion visible. It therefore does not define the intended domain well enough to check applicability.
A clearer statement is: “The model evaluates the specified starting-dose strategies for adults with the defined disease and moderate renal impairment, receiving Drug Z in outpatient care on the stated background therapy; patients receiving dialysis are outside the current scope.” The treatment boundary must still specify the doses and schedules being compared.
Now consider a hypothetical set of 120 decision scenarios: 70 involve normal or mildly impaired renal function, 35 involve moderate impairment without dialysis, and 15 involve dialysis. If the intended decision is only for moderate impairment without dialysis, then the relevant subset is 35 scenarios, not all 120. The other 85 scenarios should not be treated as covered simply because they occur in the same disease population. This count illustrates classification by the declared boundary; it does not establish that the model has adequate evidence for the 35 included scenarios.
Diagnostics, Alternatives, and Common Failure Modes
A boundary is easier to evaluate when it can be tested against concrete scenarios. List representative patients and care situations, then ask whether each is in scope, out of scope, or uncertain. If reasonable reviewers cannot classify a scenario from the scope statement, the statement likely needs sharper definitions. Also ask whether each intended comparison is genuinely represented: a different comparator, treatment stage, dose schedule, or background therapy may be a different scenario, even if the drug and diagnosis are unchanged.
Three approaches are common, each with a different trade-off:
| Approach | Strength | Risk |
|---|---|---|
| Narrow, trial-matched scope | Clear alignment with a defined evidence base | May not cover the broader decision population |
| Broad, all-comers scope | Appears to cover many potential uses | Can conceal unsupported groups and important differences |
| Decision-specific scope with explicit subgroups | Makes the intended use and distinctions visible | Requires clear subgroup definitions and support for each intended application |
The decision-specific approach is often the most transparent: define the core population and treatment context, then identify any intended subgroups explicitly. It does not guarantee that the model is adequate, but it makes the adequacy question answerable. A broad scope is not a substitute for evidence, while a narrow scope is not automatically a limitation if it matches the decision.
A frequent mistake is to reason, “The model includes the disease, so it can be used for any patient with that disease.” This is wrong because disease identity alone does not establish equivalence across severity, biomarker status, treatment history, organ function, regimen, or care setting. The correction is to identify the intended decision scenario and check whether its defining features are within the model’s stated and supported domain.
- Overbroad boundary: A model described as applying to “all patients” can imply coverage of unconsidered subgroups. State relevant inclusion criteria and exclusions instead.
- Study sample mistaken for target: The participants in an evidence source do not automatically define the population for every use. Explain any intended difference and its support.
- Comparator left vague: “Standard care” can refer to different strategies. Specify the background care relevant to the comparison.
- Boundary changed without reassessment: Adding a new population or treatment stage is a substantive change in intended use, not merely a wording edit.
The scope statement has limits. It cannot prove that the model is correct, resolve missing evidence, or ensure that a proposed extension will work. Its purpose is to make the intended application explicit and expose mismatches early. When the decision changes, revisit the population, treatment, and care-context boundaries rather than assuming the old scope still applies.
Check Your Understanding
For each question, identify the scope issue and state a defensible response.
- A model compares two add-on regimens in adults with biomarker-positive disease. A team wants to use it for biomarker-negative patients. What must be clarified before treating this as an in-scope use?
- Why is “Drug A versus standard care” often an incomplete treatment-boundary statement?
- A model was defined for first-line therapy, but a proposed decision concerns treatment after progression. Explain why the existing scope cannot be assumed to cover the new use.
- For a dose decision in moderate renal impairment, list two details beyond the disease name that could help define the care scenario.
- What does \(D \subseteq M\) express in scope setting, and what does it not establish?
Solutions. (1) Clarify whether biomarker-negative patients belong to the intended population and whether evidence and assumptions support that extension; shared diagnosis alone is insufficient. (2) “Standard care” may differ in background treatment or clinical setting, changing the comparison; specify the relevant regimen or care conditions. (3) Prior progression and treatment history define a different patient and treatment scenario, so the team must justify and assess an extension rather than assume coverage. (4) Examples include the specific dose and schedule, dialysis status, background therapy, or outpatient versus inpatient setting. (5) It states that all intended decision scenarios are within the model’s defined domain; it does not prove that evidence is adequate, the model is correct, or every scenario in \(M\) is equally supported.