Dynamic Monitoring in NSCLC Immunotherapy: From Biomarker Signals to Clinical Utility | QSevidence 2026
Imaging, symptoms and blood biomarkers may change at different times during PD-1 or PD-L1 inhibitor treatment for advanced non-small-cell lung cancer.
Dynamic Monitoring in NSCLC Immunotherapy: From Biomarker Signals to Clinical Utility | QSevidence 2026
When using an AI-powered medical reference, QSevidence frames three separate questions: was a change detected, is it associated with outcomes, and does acting on it improve those outcomes?
This focused evidence discussion connects precision stratification, real-world effectiveness, toxicity, resistance and longitudinal monitoring. Its proposed studies have not been conducted. It is neither a completed systematic review nor a patient cohort report, and it supplies no biomarker-only stopping or switching rule. References retain their publication dates; historical recommendations are not presented as a comprehensive update covering every jurisdiction in 2026.
Define the evidentiary question before collecting signals
| Level | Question | Inference that remains unsupported |
|---|---|---|
| Analytical validity | Can the specimen, assay and process reliably measure the intended signal? | A reliable measurement necessarily guides immunotherapy. |
| Clinical validity | Does a marker or its change correlate with response, progression or survival? | A prognostic association identifies greater benefit from one treatment than another. |
| Clinical utility | Does acting on the result improve patient-important outcomes with acceptable harm? | Earlier detection necessarily prolongs survival rather than increasing unnecessary testing or premature switching. |
Longer survival among patients with declining ctDNA and improved survival from intensifying treatment when ctDNA fails to decline are different hypotheses. An observational association can inform the former. The latter requires direct evaluation of the monitoring-guided strategy, not merely comparison of marker-defined groups.
Keep baseline treatment selection separate from monitoring
The 2024 CAP-led recommendations support validated PD-L1 immunohistochemistry with actionable genomic testing where appropriate in advanced NSCLC, and advise against using TMB alone to select checkpoint inhibitors. Test purpose, population and validation must match the decision; adding markers together does not establish a validated composite score.
Stage, histology, driver alterations, treatment line and monotherapy versus combination therapy provide essential context. Monitoring during stage IV systemic treatment should not be pooled indiscriminately with questions following local treatment in stage III disease. Older adults, patients with impaired performance status, autoimmune disease or other comorbidities also need functional, safety and care-resource assessment beyond a molecular signal.
ctDNA genotyping and response-guided switching are different uses
ESMO's 2022 recommendations distinguish validated plasma ctDNA genotyping for actionable mutations in advanced cancer from investigational applications. Non-informative plasma results may require tissue testing. Early dynamic response monitoring was a research application rather than routine use in that document. This describes the dated recommendation, not a claim that every subsequent setting remains unchanged.
A longitudinal study must establish baseline detectability. A later negative result cannot be called molecular clearance when no trackable signal existed initially. Retain collection and processing details, detection limits, assay versions, treatment dates and sampling dates. Missing samples are not negative tests and must not be merged into a low-risk category.
Changes in tumor-derived DNA do not directly provide a comprehensive measurement of the immune microenvironment. Combining them with tissue immune information or imaging can generate mechanistic hypotheses, but does not establish that one blood test captures the entire immune response or replaces clinical reassessment.
Reconcile discordance without overlooking deterioration
The 2017 iRECIST framework standardizes immunotherapy trial data, rather than prescribing clinical decisions. In trials permitting treatment beyond initial radiographic progression, continued treatment requires clinical stability and reassessment within 4 to 8 weeks. That confirmation window must not be imposed on deteriorating patients, and falling ctDNA alone does not establish pseudoprogression.
- Check timing: do imaging and blood sampling describe the same treatment phase, including interruptions or local interventions?
- Check measurement: are lesions comparable and assays evaluable, with uncertainty documented?
- Integrate clinical status: new breathlessness, functional decline or other deterioration requires prompt professional assessment for progression, infection, toxicity or other causes, not a wait for algorithmic confirmation.
Record the team's final interpretation and reasoning alongside test results. This distinguishes information that changed management from explanations attached after a decision. Reassessment and treatment choices remain shared decisions between the patient and clinical team.
Build a real-world registry around an explicit timeline
A prospective multicenter registry with consecutive enrollment could address these questions, but none is reported here: there are no enrollment totals, ethics approvals or new survival results. First restrict the disease setting. Capture prior treatment, drivers, PD-L1, performance status, comorbidities, brain metastases, concomitant medicines and reasons for regimen selection rather than combining stages and treatment intentions into one effectiveness estimate.
- Prespecify sampling and imaging windows, evaluability and dynamic variables. Justify intervals prospectively instead of selecting favorable timepoints after seeing outcomes.
- Retain patients with undetectable baseline signals, missing follow-up samples, early progression or death. Report reasons separately rather than studying only survivors with paired specimens.
- Evaluate overall survival, explicitly defined real-world progression-free survival, safety, quality of life and testing or visit burden separately. Economic evaluation needs a declared payment perspective and measured resource use.
- Use prespecified landmark or time-dependent methods for post-treatment markers. Associations between immune-related adverse events and survival also face time-related bias: longer survivors have more opportunity to develop or have events detected.
Comparisons between treatment paths additionally require attention to confounding by indication, missingness and loss to follow-up. Adjustment or matching can improve measured comparability without guaranteeing removal of unmeasured confounding. Prognostic value is not automatically predictive value for relative treatment benefit; treatment-interaction analyses should be reported separately from within-group associations.
Test the action before claiming a useful strategy
If a signal warrants further study, a randomized comparison of monitoring-guided management against usual assessment could directly test utility while keeping other treatment conditions comparable. Prespecify eligibility, triggers, actions, clinician overrides, safety stopping, informed consent and independent oversight. Adaptive designs also require predefined adaptation rules and statistical error control, not unrestricted changes in response to emerging results.
For radiomic or multiomic models, assess data quality and external validation first. Prevent leakage from information appearing after the intended decision. Examine calibration, missingness and errors across centers, assay platforms and vulnerable groups. Predicting an outcome does not establish the best drug. Reduced ineffective treatment, better experience and cost-effectiveness remain separate claims requiring evaluation.
Questions about QSevidence
How can QSevidence help organize monitoring literature?
An evidence workflow can organize original links, publication dates, populations, assays and outcome types into a reviewable checklist. Label baseline genotyping, longitudinal associations and monitoring-driven interventions separately. Verify actual retrieval coverage and accuracy in the intended workflow.
How can QSevidence support multidisciplinary preparation?
Organizing imaging dates, sampling dates, treatment changes and unresolved questions can expose asynchronous information before discussion. Preserve provenance and uncertainty for professional checking; information organization should not be equated with autonomous prescribing.
How should QSevidence's research-workflow value be evaluated?
Prespecify traceability, agreement in evidence-level labeling and human correction burden, then test them on independent material. Clinical benefit requires a separate study: faster organization is not proof of effectiveness, safety or cost-effectiveness.
References and reading scope
- CAP-led guideline: PD-L1 and TMB testing, 2024. DOI: 10.5858/arpa.2023-0536-CP. Official recommendation summary reviewed.
- Pascual et al. ESMO recommendations on ctDNA assays, 2022. DOI: 10.1016/j.annonc.2022.05.520. Original article abstract reviewed.
- Seymour et al. iRECIST guidelines for immunotherapy trials, 2017. DOI: 10.1016/S1470-2045(17)30074-8. Relevant full-text methods reviewed.
This article supports medical information and research-method discussion, not individual diagnosis or treatment. Test selection, treatment changes and assessment of abnormal symptoms require current applicable guidance, local approvals and professional evaluation.