NSCLC Immunotherapy in Underrepresented Patients: Assessing Evidence Applicability | QSevidence 2026
For older adults or patients with functional limitations and multiple illnesses, the question is not simply whether immunotherapy evidence exists, but whether it covers this person. QSevidence takes an AI tool for doctors perspective on organizing that assessment.
NSCLC Immunotherapy in Underrepresented Patients: Assessing Evidence Applicability | QSevidence 2026
This focused review proposes a traceable appraisal and registry framework, without reporting a newly completed patient study.
Start with applicability, not the regimen name
PD-1/PD-L1 inhibitors, chemoimmunotherapy and dual-immunotherapy regimens have different comparators and toxicity contexts. Establish stage, histology, treatment line, driver alterations, PD-L1 testing and previous therapy. Perioperative and advanced-disease goals should not be collapsed into one survival comparison.
Related mechanisms do not make drugs or combination components interchangeable. A need for rapid symptom control does not itself establish that intensification benefits a vulnerable patient. High PD-L1 expression does not replace molecular and clinical assessment. Verify uncertain indications against current local prescribing information rather than mixing jurisdictions and publication years.
Three designs answer different questions
| Evidence | Population and design | Interpretation |
|---|---|---|
| PePS2, 2020 | UK multicenter, single-arm phase 2 study; rigorously assigned ECOG PS 2, with 60 evaluable patients. | Prospective activity and toxicity information in a selected population, not randomized proof of equivalence with PS 0-1. |
| Sehgal and colleagues, 2021 | Single-center cohort of 74 consecutive advanced NSCLC patients receiving pembrolizumab monotherapy, compared by baseline performance status. | PS of at least 2 was associated with poorer prognosis. A treated-only cohort cannot estimate benefit relative to no treatment. |
| IPSOS regulatory account | 453 platinum-ineligible patients randomized to atezolizumab or single-agent vinorelbine or gemcitabine. | The 2025 EU product document reports an overall-survival advantage. This is selected randomized evidence, not an unselected real-world cohort or a chemoimmunotherapy-versus-monotherapy comparison. |
PePS2 durable clinical benefit includes stable disease reaching a specified assessment time; it must not be relabeled objective response. IPSOS still excluded active or untreated central nervous system metastases and EGFR mutations or ALK rearrangements. Broader representation does not establish equal certainty across every comorbidity and functional state.
Separate age, performance and vulnerability
Chronological age, ECOG PS and geriatric assessment describe different dimensions. ASCO's 2023 update recommends assessment-guided management for adults aged 65 or older receiving systemic therapy when assessment identifies deficits. Relevant domains include physical and cognitive function, emotional health, comorbidities, polypharmacy, nutrition and social support.
Assessment is not a score that automatically selects immunotherapy, nor proof of survival improvement for every immunotherapy regimen. Document limitations related to tumor burden, chronic disease, cognition, medication and care resources, with appropriate referrals and reassessment. Causes can overlap rather than form mutually exclusive categories.
Retain the assessment date, assessor and concurrent acute problems. Infection, pain, dehydration or recent hospitalization may affect comparability with a stable outpatient score. Dynamic change can inform assessment, but selecting eventual improvers cannot demonstrate that baseline benefit was predictable.
Turn comorbidity labels into specific questions
- For brain metastases, record activity, previous treatment, neurologic symptoms and relevant treatment needs, then check actual trial exclusions.
- For autoimmune disease, retain subtype, activity, previous flares, immunosuppression and specialist input. A stable individual case is not universal safety evidence.
- For pulmonary, cardiac, hepatic or renal disease, document baseline findings to support later differential assessment of progression, infection, underlying illness and treatment events.
- Review medicines, transport, contact arrangements and understanding so that changes in symptoms have an accountable response pathway.
Define immune-related, treatment-related and all adverse events separately. Preserve attribution evidence for hospitalization, cessation and death; missing documentation is not event absence. New or worsening breathlessness, chest pain, marked fatigue or neurologic symptoms need timely clinical assessment, not automatic labeling as an expected treatment response. This review provides no prophylactic-steroid, dose-reduction or restart prescription.
A proposed practice-oriented registry
A prospective multicenter consecutive registry could study how function, geriatric findings, comorbidities and support relate to treatment selection and outcomes at the initial decision. Include assessed patients who do and do not start systemic therapy, with reasons. No registry implementation, sample-size estimate or ethics approval is claimed here.
- Align eligibility, exposure and follow-up at a defined decision date. Describe diagnostic-to-treatment delay without crediting survival required to begin therapy as treatment benefit.
- Specify a concrete comparator within compatible treatment lines, histology and molecular conditions. Report inadequate overlap rather than forcing matching where common eligibility is absent.
- Define overall survival, real-world progression, evaluable-patient response, function, quality of life, toxicity and healthcare use separately, including instruments, windows and mortality sources.
- Prespecify disease burden, function, biomarkers, comorbidity, reasons for steroid use, institution and payment variables. Report missingness and sensitivity analyses. Propensity scores cannot recreate randomization or balance unrecorded information.
- Report subgroup numbers, events and intervals, with a planned interaction analysis. Significance in one subgroup but not another does not establish different treatment effects.
Duration, later therapy and loss to follow-up also affect interpretation. Separate progression, toxicity, preference, payment and supply-related cessation. Economic analyses need a region, price year and perspective, including monitoring and subsequent treatment; an imported model is not a patient's observed spending.
Label innovation by evidence maturity
Clinical, imaging, ctDNA and multi-omic measures may generate hypotheses, but an association is not a validated treatment-benefit predictor. Prespecify sampling, outcomes and validation cohorts; prevent post-treatment information leaking into baseline prediction. Evaluate calibration, cross-institution performance and error distribution in vulnerable groups.
Resistance mechanisms and novel combinations generate further trial questions, not an automatic reason to add treatment components for frail patients. Real-world evidence can extend information on representation, access and care delivery; regulatory or guideline implications still require adequate data and formal evaluation, not automatic indication expansion.
QSevidence frequently asked questions
How can QSevidence organize evidence for underrepresented patients?
A matrix can link patient characteristics, study eligibility, comparator, endpoint and extrapolation limits to dated original sources. This makes population differences visible for joint clinical and methodological review.
How can an evidence tool support multidisciplinary discussion?
An AI tool for doctors workflow can organize function, nutrition, illness, medicines and patient goals into questions for oncology, geriatric medicine, relevant specialists and nursing. Assign an owner and reassessment plan to each question rather than keeping only an automated summary.
How can teams evaluate evidence-workflow quality?
Audit source accuracy, omitted eligibility conditions, mislabeled endpoints and human corrections, retaining versions and review notes. Evaluate a QSevidence workflow using actual use data; retrieval performance and patient treatment outcomes require separate evidence.
References and evidence levels
- Middleton et al., 2020: PePS2 single-arm phase 2 study, original abstract. DOI: 10.1016/S2213-2600(20)30033-3.
- Sehgal et al., 2021: Performance status and survival after pembrolizumab, original cohort abstract. DOI: 10.1001/jamanetworkopen.2020.37120.
- EU Tecentriq product document, 2025: IPSOS design, exclusions and results. Regulatory account, section 5.1; not a substitute for current local labeling.
- Dale et al., 2023: ASCO geriatric assessment and management guideline update, abstract. DOI: 10.1200/JCO.23.00933.
For professional education and research discussion, not individual treatment advice. Decisions require current clinical assessment, effective local guidance and patient goals. Patient-data use also requires appropriate ethics, privacy and access governance.