Screening Yield and Treatment Access in Lung-MAP

From Screening to Matched Treatment

A molecularly matched platform can offer several investigational treatments, but an offer is not the same as access. A patient must first have usable tissue and an interpretable genomic result; the result must correspond to an open treatment option; the patient must meet that option’s clinical criteria; and the patient must proceed to enrollment. Each step can reduce the number who reach treatment. Separating these steps shows where the losses occur and prevents a single percentage from hiding distinct barriers.

The earlier material on Lung-MAP’s screening protocol and biomarker matching establishes the basic idea: patients with advanced squamous lung cancer undergo molecular screening, and treatment assignment depends on matching clinical and biomarker criteria to a substudy. Here the focus is the screening cascade: what fraction of those screened can actually reach a matched investigational treatment, and how should that fraction be measured?

For a simple cascade, let \(N_0\) be the number screened, \(N_1\) the number with adequate tissue, \(N_2\) the number with reportable genomic results, \(N_3\) the number with a target alteration matched to an available treatment, \(N_4\) the number meeting that treatment’s clinical eligibility criteria, and \(N_5\) the number enrolled. Then \(N_5/N_0\) is the overall enrollment yield among screened patients. The stage-specific proportions use the preceding stage as their denominator.

This is a telescoping identity when the stages are nested and counts refer to the same screened cohort. Each factor describes a conditional transition, such as the proportion with reportable results among those with adequate tissue. The identity does not assume that the stages are independent. In practice, the pathway can branch: a patient might have more than one relevant finding, a treatment option may close, or enrollment may occur through a route not captured by a single linear chain. The definitions and counts must then reflect that structure rather than force it into a misleading sequence.

Worked Examples of Yield and Access

What the Cascade Can—and Cannot—Show

The cascade is useful because the overall yield alone does not locate where access is limited. A low overall enrollment proportion could reflect inadequate tissue, nonreportable assays, few patients with a target alteration, restrictive clinical criteria, lack of an open matched option, or decisions not to enroll. These mechanisms suggest different operational responses, so reporting only \(N_5/N_0\) is rarely enough for evaluation.

The product decomposition is exact for nested counts, but it does not establish why a transition occurred. For instance, an association between a patient characteristic and lower enrollment does not show that the characteristic caused reduced access. Differences may reflect disease severity, tissue collection, site practices, timing, treatment availability, or other factors. The relevant comparisons and adjustment strategy depend on the question; a descriptive cascade is not itself a causal analysis.

A screening report should state the cohort and time period, define each stage, and make denominators explicit. It should distinguish at least:

• patients not successfully screened because tissue was unavailable or inadequate;
• patients with tissue but no reportable genomic result;
• patients with reportable results but no match to an open treatment;
• matched patients who do not meet clinical eligibility criteria; and
• eligible patients who do not enroll, while distinguishing non-consent, logistical barriers, and other recorded reasons where possible.

The distinction between “no match” and “no available matched treatment” is especially important. A genomic alteration can be detected without there being an open substudy that targets it at that time. Likewise, meeting biomarker criteria does not ensure clinical eligibility. Operational definitions should specify whether the endpoint is a potential match, an eligibility determination, consent, enrollment, or treatment initiation; these are not interchangeable.

Diagnostics, Alternatives, and Limitations

Several checks help detect a cascade that is incomplete or misclassified. First, verify that the stage counts are nested when the calculation assumes nesting: \(N_0\geq N_1\geq \cdots \geq N_5\). Second, reconcile losses: the number entering a stage should equal those progressing plus those with a documented reason for not progressing, with unresolved status shown separately. Third, check whether tissue failure, missing results, and no genomic match have been kept distinct. Fourth, review whether treatment openings and eligibility rules changed during the reporting period; a single pooled rate can conceal these changes.

A common alternative is to report only the fraction of patients with a target alteration among those with reportable genomic results. That answers a biomarker-frequency question, not the practical access question among everyone screened. Conversely, using all screened patients as the denominator for every transition obscures where losses occur. The cascade approach reports both the overall yield and conditional stage-specific proportions. A cohort flow diagram can communicate the same counts visually, while a table is often clearer for comparing stages or groups.

Stratifying the cascade by site, time period, or patient group can reveal variation in tissue success or enrollment. Such comparisons require consistent definitions and adequate counts. Small subgroups yield imprecise estimates, and repeated subgroup comparisons can invite chance findings. Moreover, rates may differ because the groups have different clinical or genomic profiles; a raw difference is descriptive and should not automatically be interpreted as an inequity or a causal effect.

Finally, a yield is tied to its target population and operational setting. A cohort with a different case mix, tissue-collection process, set of open treatments, or eligibility rules may have a different yield even if screening is carried out consistently. Enrollment is also not synonymous with receiving a dose, completing treatment, or achieving a clinical outcome. These endpoints answer distinct questions and should be counted separately when they are of interest.