From One Disease Setting to Several Treatment Substudies
A conventional trial often asks whether one treatment works in a defined patient group. An umbrella trial starts from a different organizational problem: patients share a disease setting, but their tumors may have different molecular features that make different treatment hypotheses relevant. The protocol therefore needs a common way to screen patients and a controlled way to route eligible patients to the appropriate substudy.
Lung-MAP, also known as SWOG S1400, was designed for patients with advanced squamous non-small-cell lung cancer. Its central screening process used tumor genomic profiling to identify alterations that could qualify a patient for a biomarker-matched substudy. Substudies tested treatment hypotheses for particular molecular groups; examples from the program included pathways involving FGFR, PI3K, and CDK4/6. The specific substudies and eligibility rules depended on the protocol version in force.
The operational sequence matters. A patient first had to be considered for screening and provide suitable tumor material. The material then had to yield a reportable genomic result. The result was assessed against the protocol’s alteration definitions, and the patient’s clinical circumstances were checked against the substudy’s eligibility criteria. Only then could the patient be routed to that substudy. If the substudy randomized eligible patients, that randomization occurred after routing; patients were not generally randomized among every treatment in the umbrella protocol.
What Assignment Means—and What It Does Not
A biomarker result is not itself a treatment assignment. For example, a report that identifies an FGFR-related alteration does not automatically establish eligibility for an FGFR-directed substudy. The protocol must define which alteration types and thresholds qualify, how the result is verified, and whether other clinical criteria are met. A patient may have a potentially relevant alteration but fail a substudy’s clinical eligibility requirements. Conversely, a patient whose tumor has no qualifying alteration may not be eligible for a matched substudy, even if the patient is otherwise medically suitable for the overall screening protocol.
This distinction separates three decisions: screening asks whether a usable biomarker result can be obtained; routing asks which substudy, if any, the result and protocol rules support; and treatment allocation asks which intervention a substudy participant receives. Depending on the substudy design, treatment allocation may involve randomization. An assignment to a substudy is therefore not synonymous with receiving a particular investigational treatment.
Worked Example: Applying a Hypothetical Routing Rule
Suppose a fictional Lung-MAP protocol version defines a matched substudy for a specified FGFR alteration. A patient’s central report identifies that alteration, but the patient’s performance status falls outside the substudy’s stated eligibility range. What happens?
The report satisfies the molecular part of the rule, but not the full enrollment rule. In symbols, if \(M\) denotes a qualifying molecular result and \(C\) denotes clinical eligibility, substudy enrollment requires both:
Here \(M\) is true and \(C\) is false, so \(M \cap C\) is false. The patient is not enrolled in that substudy under the stated rules. The molecular finding is not discarded as a biological observation, but it cannot override the clinical eligibility criterion. This fictional example illustrates the logic of routing; it does not state an actual Lung-MAP eligibility threshold.
Screening Yield as a Sequence of Conditional Steps
The number of patients who reach a matched substudy depends on successive stages, not only on how often a target alteration occurs. Let \(A\) denote adequate tumor material, \(M\) a qualifying molecular match given adequate material, and \(C\) clinical eligibility given a match. Under this sequence, the probability of reaching the substudy is:
This is the multiplication rule for conditional probabilities. It does not require the stages to be independent: each probability is conditional on having passed the preceding stage. The formula assumes that the stages are defined consistently and that the recorded counts refer to the same screened population and protocol version. Its practical value is that it separates a low overall yield into stages that can be investigated.
Worked Example: Estimating the Number Routed to a Match
In a hypothetical screening cohort of 500 patients, 430 have adequate tumor material. Of those 430, 129 have a qualifying alteration. Of those 129, 103 also satisfy the relevant clinical eligibility criteria. Calculate the observed proportions at each stage and the final screening yield.
The adequate-material proportion is \(430/500=0.860\), or 86.0%. Among patients with adequate material, the match proportion is \(129/430=0.300\), or 30.0%. Among patients with a match, the clinically eligible proportion is \(103/129=0.798\), or 79.8%.
The overall proportion reaching a matched substudy is \(103/500=0.206\), or 20.6%. The same result follows from the conditional product, allowing for rounding:
Thus, 103 of the 500 screened patients reached a matched substudy in this hypothetical cohort. The other 397 did not reach that endpoint, but the counts alone do not identify a single cause: some lacked adequate material, some had no qualifying match, and some matched patients did not meet clinical criteria.
This decomposition also guides operational review. A low \(P(A)\) points toward tissue acquisition, specimen handling, or assay success. A low \(P(M\mid A)\) may reflect the prevalence of the defined targets in the screened population, the target definitions, or the range of substudies open at that time. A low \(P(C\mid A\cap M)\) points toward clinical eligibility or timing. These are different problems and call for different responses; changing a treatment hypothesis cannot repair a specimen-processing bottleneck.
Operational Demands of Shared Screening
A shared screening system reduces the need to begin an entirely separate molecular search for each substudy, but it creates coordination demands. Tissue must be requested, tracked, and tested in time to inform a treatment decision. A screening result must be linked to the correct patient and protocol version. Sites need to know which substudies are open, what results qualify, and how to act when a report contains multiple findings. Central laboratories, trial offices, investigators, and treating teams must have clear responsibilities for communicating results and confirming eligibility.
Turnaround time is clinically consequential. If a result arrives after a patient’s disease or health status changes, the molecular match may no longer translate into enrollment. Tissue can also be insufficient, unavailable, or unsuitable for the assay. In addition, a panel may report alterations that are scientifically interesting but do not meet the protocol’s prespecified definition. Recording these outcomes separately makes the screening pathway auditable and helps distinguish assay failure from absence of a qualifying target.
Worked Example: Planning Screening Capacity
A hypothetical site network expects to screen 240 patients over a planning period. Based on earlier operational experience, it anticipates that 90% will have adequate material, 25% of those will have a qualifying match, and 80% of matched patients will meet clinical eligibility. Estimate the expected number reaching a matched substudy.
Using the sequential probability calculation:
The planning expectation is about 43 patients reaching a matched substudy. This is an expected count, not a guarantee that exactly 43 will enroll. It assumes the rates are applicable to this population and protocol period, and that the stages occur in the stated order. If the actual count is much lower, the site network should examine observed counts at each stage rather than infer immediately that the treatment hypothesis is unpromising.
Routing When Results Are Complex
A tumor report may identify more than one potentially relevant alteration. A valid umbrella protocol therefore needs prespecified rules for such cases. Depending on the protocol, the rules might establish a priority order, define mutually exclusive biomarker groups, or specify another way to determine eligibility. The essential point is that sites should apply the same documented rule rather than choose a substudy informally. If no rule applies clearly, the case needs protocol-defined review rather than an improvised assignment.
The match definitions also need to be sufficiently precise for reproducibility. A general pathway label may cover several distinct alterations, but a substudy may accept only some of them. The report’s assay method, the protocol’s threshold, and any confirmation requirements must align. An apparently simple routing decision can otherwise vary across sites, weakening the interpretability of the enrolled group.
Worked Example: Two Findings on One Report
In a fictional case, a central report lists both a qualifying PI3K-pathway alteration and a qualifying FGFR alteration. The active protocol version contains two corresponding substudies, but the version’s routing rules specify a priority order when both are present. How should the site proceed?
First, the site confirms that the report is valid and that both findings meet the exact protocol definitions. Second, it applies the protocol’s stated priority rule. If that rule directs the patient to the PI3K-pathway substudy, the site evaluates clinical eligibility for that substudy; it does not select the FGFR substudy merely because that option appears more attractive. If clinical eligibility fails, the protocol’s rules determine whether another route is permitted. The site must not invent a fallback assignment.
This case illustrates why a matching algorithm is part of trial governance. Consistent prespecified routing protects against discretionary selection and makes the enrolled biomarker group interpretable. The exact priority rule in this example is hypothetical and is not asserted to be a Lung-MAP rule.
How the Umbrella Approach Compares
| Approach | How patients enter treatment evaluation | Main trade-off |
|---|---|---|
| Lung-MAP-style umbrella screening | A shared disease-setting screen routes patients to eligible molecular substudies. | Coordinates screening across hypotheses, but depends on tissue, timely results, and consistent routing. |
| Separate single-treatment trials | Each trial screens and recruits for its own eligibility criteria. | Can use a focused design, but may duplicate screening and compete for patients. |
| Unselected treatment trial | Patients enter without requiring a specific molecular match. | Can evaluate treatment in a broader group, but may dilute a treatment effect confined to a biomarker-defined subgroup. |
| Basket trial | A shared molecular feature can connect patients with different cancer types to treatment evaluation. | Can study a target across diseases, but does not organize multiple target-defined substudies within one disease setting in the same way. |
These approaches answer different questions. An umbrella design is useful when several molecularly defined hypotheses are being studied within a disease setting and a common screen is feasible. It does not automatically make rare subgroups large enough for precise estimates. Nor does a match establish that the biomarker predicts treatment benefit: that requires evidence from the treatment comparison and an analysis appropriate to the design.
Diagnostics, Limitations, and a Common Mistake
Useful operational diagnostics include the proportion of submitted specimens that are adequate, the time from submission to report, the proportion of reports that are informative, the frequency of qualifying alterations, the proportion of matched patients who meet clinical criteria, and the time from result to enrollment. These measures should use clear denominators. For example, “match rate” is ambiguous unless it states whether the denominator is all screened patients, patients with adequate specimens, or patients with reportable results.
A pathway can fail even when every component is scientifically reasonable. Substudy-specific eligibility may be restrictive; a target may be uncommon; or enrollment may lag because patients become ineligible while results are pending. Screening yields can also change as participating sites, available substudies, assay procedures, or protocol versions change. Comparing rates without checking those conditions can make an operational change look like a biological difference.
- Wrong reasoning: “Only a small fraction of screened patients entered a matched substudy, so the matched treatment was ineffective.”
- Correction: Entry yield is determined before treatment outcomes are observed and combines specimen adequacy, target frequency, and clinical eligibility. It cannot by itself estimate treatment efficacy.
- Wrong reasoning: “The tumor has a target, so the patient should receive the corresponding drug.”
- Correction: A target can support a substudy hypothesis, but protocol-defined molecular criteria, clinical eligibility, consent, and the substudy’s allocation procedure still apply.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- What is the difference between a biomarker match and enrollment in a matched substudy?
- In a hypothetical cohort, 300 patients are screened, 240 have adequate material, 72 have a qualifying match, and 54 of those meet clinical criteria. What proportion of all screened patients reaches a matched substudy?
- Why should a patient with two potentially qualifying alterations not be routed according to a site investigator’s informal preference?
- A screening report arrives after a patient’s clinical status changes. What two distinct issues should the study team assess?
- Why does a low matched-substudy enrollment count not, by itself, show that the treatment is ineffective?
Solutions. (1) A match identifies a molecular feature that may qualify for a substudy; enrollment additionally requires the protocol’s other criteria and completion of its enrollment process. (2) \(54/300=0.180\), so 18.0% of all screened patients reach the matched-substudy endpoint. (3) Prespecified routing rules ensure consistent assignment and reduce discretionary selection; the applicable protocol rule should govern. (4) The team should assess the result’s timing and whether the patient still meets clinical eligibility; a valid earlier match does not guarantee current eligibility. (5) Enrollment yield reflects specimen adequacy, target frequency, clinical eligibility, and operational timing, whereas treatment efficacy is assessed from outcome data under the substudy’s design.