Not yet independently audited

This page was written from verified sources but has not yet passed the independent citation audit. Treat its claims as provisional.

Lung adenocarcinoma — open questions

Last curated: 2026-10-30

Use. Tier 1 questions could change practice and can be tested with a feasible contemporary design. Tier 2 questions need assay, cohort, endpoint, or mechanistic development before a definitive practice-changing trial. IDs are stable; retire an ID rather than renumbering it.

Tier 1 — practice-changing and designable now

OQ-1 — Who needs first-line EGFR intensification?

Question. In untreated EGFR exon-19-deletion/L858R metastatic disease, which prespecified baseline clinical/genomic state identifies an OS or quality-adjusted-survival advantage from osimertinib plus chemotherapy or amivantamab-lazertinib over osimertinib alone?

Why open. Both intensification strategies improve PFS and now OS in aggregate but add cytotoxic or infusion/VTE/skin burden (grade ≥3 AEs 70% vs 34% and 80% vs 52%); no validated selector defines who needs it (FLAURA PMIDs: 29151359, 31751012; FLAURA2 PMIDs: 37937763, 41104938; MARIPOSA PMIDs: 38924756, 40923797).

2026-10-16 sweep. Sharpened, not answered. FLAURA2 and MARIPOSA have both reported significant OS gains over osimertinib monotherapy, HR 0.77 and HR 0.75 (PMIDs: 41104938, 40923797). The aggregate-survival argument against intensification is therefore gone, and the question is now purely one of selection. A co-mutation-restricted randomized trial (EGFR/TP53, PFS HR 0.44, OS immature; PMID 42574006) did not randomize TP53-wild-type patients, so it cannot show an interaction. A Japanese clinical score identifies long-term benefit on monotherapy but is explicitly prognostic (PMID 42767607). No trial has compared chemotherapy intensification with amivantamab-lazertinib.

OQ-2 — Can adjuvant osimertinib duration be individualized?

Question. Does tumor-informed ctDNA plus pathologic risk safely identify patients for shorter versus longer adjuvant osimertinib?

Why open. ADAURA used three years and improved OS, while postoperative ctDNA is prognostic but not yet treatment-directive (PMIDs: 32955177, 37272535, 28899864, 34844976).

2026-10-16 sweep. EVIDENCE-trial MRD data (adjuvant icotinib vs chemotherapy) add prognostic weight: landmark MRD+ DFS HR 4.44, longitudinal NPV 91.3%, and the 24-week timepoint the most informative (PMID 42706268). There is still no MRD-randomized duration trial. ADAURA's 8-year landmark (OS HR 0.52) does not address duration (PMID 42732874).

OQ-3 — What is the optimal post-adjuvant EGFR strategy?

Question. At relapse after adjuvant osimertinib, should treatment depend on time off drug, resistance genotype, disease site, or re-challenge response?

Why open. Metastatic resistance data largely arise after continuous first-line therapy, not the postoperative stop–relapse state (PMIDs: 32483558, 38871738).

OQ-4 — Alectinib alone or chemotherapy plus alectinib after resection?

Question. In resected ALK-positive stage II–IIIA NSCLC, does adding platinum chemotherapy to adjuvant alectinib improve cure enough to justify toxicity?

Why open. ALINA compared alectinib with chemotherapy, not chemotherapy followed by alectinib, and mature OS is pending (PMID 38598794).

2026-10-30 sweep. ELEVATE (ensartinib vs placebo after adjuvant chemotherapy; 24-month DFS HR 0.20 in stage II–IIIB; PMID 42418775) confirms that two years of a second-generation ALK TKI after chemotherapy delays recurrence. It still does not test chemotherapy-plus-alectinib against alectinib alone, and OS is immature. The question stands.

OQ-5 — Which ALK first-line lifetime sequence is best?

Question. Does first-line lorlatinib yield better lifetime CNS control, neurocognitive quality, and post-resistance survival than alectinib followed by genotype-directed later therapy?

Why open. ALEX and CROWN establish superiority to crizotinib but do not compare modern sequences head to head (PMIDs: 28586279, 33207094, 34139965).

2026-10-16 sweep. CROWN's 7-year update reports 7-year PFS of 55% on lorlatinib and no new intracranial progression after month 30 (PMID 42217582), which strengthens the CNS-protection argument for lorlatinib first. The comparator is still crizotinib, and OS has not reached its protocol-specified analysis, so the head-to-head sequence question stands.

OQ-6 — When can radiation be deferred for brain metastases?

Question. In asymptomatic measurable CNS disease with a targetable driver, does upfront CNS-active TKI with prespecified MRI rescue preserve neurocognition without compromising intracranial control versus immediate stereotactic radiation plus TKI?

Why open. CNS activity is strong for multiple TKIs, while symptomatic brain metastases still require local therapy; randomized deferral evidence is sparse (PMIDs: 30215676, 31838015, 38197815, 34932393).

2026-10-16 sweep. Partly addressed. STARLET, a pooled analysis of two randomized phase 2 trials (n=79), found no significant 12-month intracranial-PFS gain from adding upfront SRS to osimertinib in low-burden EGFR-mutant brain metastases. Its authors describe it as the first randomized evidence supporting osimertinib-first management, but it reported no neurocognitive endpoint and was EGFR-only (PMID 41534788). The question remains open for ALK/ROS1, for higher CNS burden, and for cognition-weighted outcomes.

OQ-7 — What should follow modern ROS1 inhibitors?

Question. Can prospective paired tissue/plasma genotyping allocate therapy after repotrectinib or entrectinib according to solvent-front, compound, and bypass resistance?

Why open. TRIDENT-1 addresses important earlier states, but post-new-generation resistance is heterogeneous and lacks a validated sequence (PMID 38197815).

2026-10-16 sweep. A single-centre phase 2 of type II inhibition (cabozantinib) gave 3 responses in 7 patients after type I TKIs, including repotrectinib and zidesamtinib, with MET-mediated escape at progression (PMID 42832369). This is mechanistic support for allocation by resistance, not a validated sequence.

OQ-8 — Can KRAS G12C move first line safely?

Question. Which KRAS G12C inhibitor combination improves OS over standard PD-L1-directed chemo-immunotherapy without excess hepatic, immune, or gastrointestinal toxicity?

Why open. Current randomized evidence is later-line with modest PFS benefit; first-line combinations are active in registries (PMIDs: 36764316, 35658005; NCT06793215 verified 2026-08-29).

2026-10-16 sweep. Still open. First-line single-arm data continue to accumulate (ifebemtinib + garsorasib, ORR 82% in 33 patients; PMID 42413526). Allele-agnostic RAS(ON) inhibition (daraxonrasib) and G12D inhibition (GFH375) now have early-phase lung data (PMIDs: 42685317, 42745055). NCT06793215 was rechecked on 2026-10-16 and is still recruiting.

2026-10-30 sweep. Later-line next-generation versus first-generation G12C inhibition now has randomized OS evidence: Krascendo 1, divarasib vs sotorasib/adagrasib, OS HR 0.56 (PMID 42929675). That does not answer the first-line combination-versus-chemo-IO question. NCT06793215 remains Recruiting.

OQ-9 — Are STK11 and KEAP1 predictive?

Question. In a biomarker-stratified randomized trial, do STK11 and/or biallelic KEAP1 states identify differential benefit from a specific chemo-immune or investigational regimen?

Why open. They associate with poor immune outcomes, and POSEIDON subgroup/translational evidence supports CTLA-4-containing chemo-immunotherapy as a resistance-mitigating strategy; dedicated prospective biomarker selection and independent validation remain incomplete (PMIDs: 29773717, 34740862, 32866624, 36526124, 39385035).

OQ-10 — HER2 TKI versus ADC sequence

Question. In HER2-mutant metastatic disease, should a HER2-selective TKI precede T-DXd to maximize CNS control and reduce cumulative ILD risk?

Why open. T-DXd has high response with clinically important ILD; zongertinib introduces an oral, mutation-selective alternative without randomized sequencing data (PMIDs: 34534430, 37694347, 40293180).

OQ-11 — Can MET resistance guide inhibitor class switching?

Question. Does prospective structural classification of D1228/Y1230 and bypass resistance improve outcomes from type-I/type-II MET inhibitor sequencing?

Why open. Preclinical sensitivity patterns exist, but capmatinib/tepotinib resistance has no validated routine sequence (PMIDs: 32877583, 32469185, 35690785).

OQ-12 — Is postoperative checkpoint therapy necessary?

Question. After neoadjuvant chemo-immunotherapy and complete resection, can patients with pCR and/or ctDNA clearance omit adjuvant checkpoint therapy without worse EFS/OS?

Why open. CheckMate 816 used neoadjuvant-only therapy, whereas KEYNOTE-671, AEGEAN, and CheckMate 77T combine pre- and postoperative therapy; cross-trial comparison cannot isolate the adjuvant component (PMIDs: 35403841, 37272513, 37870974, 38749033).

2026-10-16 sweep. Sharpened. AEGEAN's updated interim OS HR is 0.89 (95% CI 0.70–1.14) (PMID 42664473). A reconstruction of four postoperative-only PD-(L)1 trials estimated 2.3–6.1 extra patients per 100 without a DFS event at 36 months, with wide uncertainty (PMID 42843316). CheckMate 77T biomarker data show pre-surgical ctDNA clearance in 66% with nivolumab, and that every patient who converted to ctDNA-positive during adjuvant therapy recurred (PMID 42587156). None of this isolates the adjuvant component. S1914 adds that checkpoint therapy with SABR in inoperable early-stage disease does not improve OS (HR 1.04; PMID 42702214).

2026-10-30 sweep. KEYNOTE-867 is a second independent negative phase 3 of checkpoint therapy plus SBRT in unresected stage I–II disease (EFS HR 0.92, five treatment-related deaths; PMID 42950173). It does not isolate adjuvant IO after neoadjuvant response.

OQ-13 — Can MRD improve cure rather than prediction?

Question. Does ctDNA-triggered adjuvant escalation improve OS, or ctDNA-negative de-escalation preserve OS while reducing toxicity, versus stage/genotype-guided standard care?

Why open. Multiple cohorts establish recurrence prediction. A live PubMed search on 2026-08-29 found randomized intervention protocols in resected EGFR-mutant and advanced immunotherapy settings, but no mature randomized action result establishing survival utility (cohort PMIDs: 28899864, 28445469, 34844976, 34799585; protocol PMIDs: 39659920, 41628935).

2026-10-16 sweep. Still no randomized action result in PubMed or Europe PMC as of 2026-10-16. New prognostic data: EVIDENCE (PMID 42706268) and stage III CRT-durvalumab serial MRD, where detection during or after durvalumab, but not before it, predicted progression (PMID 42336205).

OQ-14 — Which never-smokers should receive LDCT?

Question. Can a risk model using age, family history, ancestry, air pollution, lung disease, and germline factors identify a never-smoker population with favorable mortality benefit, false-positive procedure burden, and overdiagnosis?

Why open. TALENT shows detection yield without randomized mortality benefit; smoking-based policies exclude never-smokers (PMIDs: 38042167, 33687470, 40604281).

2026-10-16 sweep. New candidate risk input: germline EGFR T790M, associated with lung-cancer risk in more than 3.3 million individuals, several-fold higher in never-smokers, with a Southern Appalachian founder origin (PMID 42752144). Never-smoker risk models could include it, but no screening trial has used it.

OQ-15 — Which screening selector improves equity?

Question. Does risk-model eligibility outperform categorical age/pack-year rules for deaths prevented per scan without worsening racial, sex, rural, or socioeconomic inequity?

Why open. Modelling favors risk-based strategies under some assumptions, while implementation disparities persist (PMIDs: 36745885, 37806735, 42038935).

2026-10-16 sweep. A blood-test-informed model (CPH-Lung) outperformed PLCOm2012 on discrimination in Danish ever-smokers, particularly for women (sensitivity 57.6% vs 39.2% at matched specificity; PMID 42863103). The equity question remains open because neither mortality nor deaths prevented per scan was measured.

OQ-16 — What is the best post-chemo-IO sequence?

Question. In driver-negative adenocarcinoma progressing after platinum-pemetrexed-checkpoint therapy, which biomarker-defined second-line strategy improves OS and patient-reported function versus docetaxel-based care?

Why open. First-line evidence is strong; post-combination resistance lacks a universal randomized sequence (KEYNOTE-189, PMID 29658856).

2026-10-30 sweep. PERSEE (PMID 42927854) is a first-line PD-L1 ≥50% comparison of pembrolizumab-chemotherapy versus pembrolizumab, not a post-chemo-IO sequence. Two China-only post-EGFR ADCs (PMIDs: 42933461, 42933457) enlarge the post-TKI menu for driver-positive disease and do not define a driver-negative second-line rule.

OQ-17 — Can oligoprogression be trial-defined?

Question. Does ablation of all progressing sites with continuation of an otherwise effective TKI improve time to next systemic therapy and OS versus immediate switching, stratified by driver and ctDNA state?

Why open. Small randomized oligometastatic trials predate modern driver-specific therapy and did not define molecular oligoprogression (PMIDs: 31067138, 28973074).

2026-10-30 sweep. A 16-RCT meta-analysis (1,343 patients) reports OS HR 0.68 overall and OS HR 0.52 in EGFR-mutant TKI-treated subgroups, with more grade ≥3 adverse events and pneumonitis (PMID 42955614). Numbers are from the abstract. The analysis still does not define molecular oligoprogression or isolate ablation-plus-continuation from an immediate systemic switch.

OQ-18 — How should frail patients be treated?

Question. In ECOG 2–3 or geriatric-vulnerable adenocarcinoma, which attenuated targeted, immune, or chemotherapy strategies preserve function and survival?

Why open. Pivotal systemic trials largely enrolled performance status 0–1, leaving major transportability uncertainty (PMIDs: 29658856, 27718847).

2026-10-30 sweep. A UK expert consensus on advanced NSCLC with comorbidity and frailty was indexed (22 of 23 statements reached ≥70% agreement; PMID 42956446). Statement text was not reviewed and is not imported as a treatment algorithm.

Tier 2 — enabling science and design work needed

OQ-19 — What initiates never-smoker adenocarcinoma?

Question. Which geographically varying mutational signatures correspond to causal, modifiable exposures rather than correlated geography or endogenous processes?

Why open. Sherlock-Lung maps signatures and driver differences, while PM2.5 experiments support a promoter mechanism; individual exposure attribution remains unresolved (PMIDs: 40604281, 37020004).

OQ-20 — Can pollution prevention reduce EGFR-driven incidence?

Question. Does sustained PM2.5 reduction lower EGFR-mutant adenocarcinoma incidence on a measurable lag, and through which inflammatory intermediates?

Why open. Epidemiology and IL-1β-dependent experimental promotion are aligned but intervention evidence is absent (PMIDs: 24911630, 37020004).

2026-10-16 sweep. Constituent-level evidence from the China Kadoorie Biobank implicates nitrate and ammonium in never-smoker lung cancer, with a nitrate × polygenic-risk interaction (PMID 42828982). Histology and EGFR status were not reported in the abstract, so the link to EGFR-mutant incidence is still inferred.

OQ-21 — What does histologic grade add after genomics?

Question. Does IASLC grade improve calibrated recurrence prediction beyond TNM, driver/co-mutation state, radiology, and ctDNA enough to change adjuvant treatment?

Why open. Grade is prognostic in validation/meta-analysis, but prospective treatment interaction is not established (PMIDs: 33905897, 38485464).

2026-10-16 sweep. A two-country cohort (n=1,745) showed grade and lymphovascular invasion add prognostic information to ninth-edition TNM (PMID 42055159), and frozen-section grading reached 79.9% concordance with final pathology (PMID 42119754). Neither tests a treatment-by-grade interaction.

OQ-22 — Is STAS biological or partly artifactual?

Question. Which standardized fixation, sectioning, and imaging rules distinguish true STAS from specimen artifact, and does a prospective STAS rule alter optimal resection extent?

Why open. STAS is prognostic but remains a descriptor rather than a TNM determinant (PMIDs: 30914285, 38508515).

2026-10-16 sweep. STAS-positive stage I tumors carry TP53 co-mutation, whole-genome doubling, and cell-cycle programs independent of EGFR status (PMID 42760303). This is evidence against pure artifact, but it is a single cohort. Note an unresolved discrepancy between abstracts: Saw 2026 describes STAS as incorporated into ninth-edition TNM, whereas Abolfathi 2026 states that only VPI is formally incorporated (PMIDs: 42760303, 42055159). The auditor should resolve this against the TNM-9 source papers (PMIDs: 38447919, 38885896).

OQ-23 — How much tumor must be sampled?

Question. What combination of multiregion tissue and plasma captures clinically actionable heterogeneity at diagnosis and progression without impractical biopsy burden?

Why open. TRACERx and mixed-response work show branched evolution and lesion discordance (PMIDs: 37046096, 38871738).

OQ-24 — Can spatial states become clinical assays?

Question. Can single-cell/spatial immune and stromal states be reduced to a locked, reproducible assay that improves treatment allocation beyond PD-L1 and genotype?

Why open. Atlases reveal rich cellular states but lack decision-utility validation (PMIDs: 28475900, 33972311, 41057692).

OQ-25 — What is a clinically meaningful plasma clone?

Question. Which variant-allele-frequency, error-model, and longitudinal rules separate actionable subclonal resistance from noise and clonal hematopoiesis?

Why open. Plasma complements tissue but low shedding and non-tumor variants complicate thresholds (PMIDs: 30988079, 34246791).

OQ-26 — Can AI improve a real pathology decision?

Question. Does an externally calibrated AI grading system improve interobserver agreement and treatment decisions, not only retrospective discrimination?

Why open. ANORAK and imaging classifiers report promising performance without prospective utility trials (PMIDs: 38200244, 37805451).

OQ-27 — Which biomarker captures productive immunity?

Question. Can antigen clonality, HLA competence, spatial T-cell state, and myeloid exclusion be combined into a robust selector outperforming PD-L1?

Why open. PD-L1 is clinically useful but incomplete; TMB and co-mutations show discordant immune biology (PMIDs: 27718847, 34648948, 32866624).

OQ-28 — How should chronic targeted-therapy survivorship be measured?

Question. Which longitudinal outcome set captures symptoms, cognition, work, adherence, caregiver burden, financial toxicity, and scan anxiety across driver groups?

Why open. Existing PRO and survivorship studies are broad lung-cancer cohorts and underdescribe years of oral therapy (PMIDs: 29110842, 22134070, 33555936).

2026-10-16 sweep. Driver-specific evidence is emerging. A French survey of people with ALK-positive disease found that professionals underestimated psychological, intimacy, and financial burden (PMID 42823962). SYMPRO-Lung links routine PRO monitoring to longer OS (post hoc HR 0.80; PMID 42727592).

OQ-29 — Can stigma interventions change behavior?

Question. Which clinician, media, and peer interventions reduce validated stigma scores and improve help-seeking, communication, trial discussion, or support use?

Why open. Stigma is measurable and associated with depression/delay, but intervention evidence is limited (PMIDs: 32721137, 30779396, 24769603).

OQ-30 — How should trial access be measured?

Question. Can molecular trials report a common screened → molecularly eligible → offered → consented → treated → retained cascade with patient time and travel burden?

Why open. Eligibility, referral, site, rural, and logistical barriers are repeatedly documented but denominators differ (PMIDs: 18650170, 22591607, 35451964).

2026-10-23 sweep — design refinement. For OQ-30, distinguish a trial not being offered from an offered trial being declined, and record navigator involvement, diagnostic-result handoffs, travel and time costs at each transition. Test navigation as a prespecified intervention rather than assume it improves enrollment; the new patient-navigation source entry is advocacy material, not an effectiveness study. No question is retired. Regional evidence remains incompletely assessed: SciELO and LILACS could not be searched (HTTP 403 access limitation); CENTRAL is also unreachable. Non-US registry attempts in this sweep yielded no usable current record, so ClinicalTrials.gov coverage must not be treated as exhaustive.

2026-10-30 sweep. No question retired. PERSEE (PMID 42927854) is a first-line PD-L1-high comparison, not a post-chemo-IO sequence, so it does not close OQ-16. Two China-only post-EGFR ADCs (PMIDs: 42933461, 42933457) and HARMONi-A OS (PMID 42307937) change the post-TKI menu but do not define a biomarker-directed second-line rule. SciELO and LILACS could not be searched (HTTP 403); CENTRAL is unreachable. WHO ICTRP served ChiCTR2000040590 as still Pending, last refreshed 15 March 2021.

Dots not yet connected

These are cross-domain junctions where both evidence bodies exist in this knowledge base but no completed study has joined them sufficiently to guide practice.

#	Dot A	Dot B	The missing junction	Powers
D1	Never-smoker geographic signatures (PMID 40604281)	Pack-year screening rules (PMID 33687470)	Prospective signature/exposure-informed risk selection with mortality endpoint	OQ-14, OQ-19
D2	PM2.5-driven clonal promotion (PMID 37020004)	EGFR-mutant incidence and prevention policy	Natural experiment linking pollution reduction to molecularly typed incidence	OQ-20
D3	IASLC grade/STAS prognosis (PMIDs: 38485464, 38508515)	Adjuvant targeted/immune therapy	Treatment-by-morphology interaction after genotype and stage	OQ-21, OQ-22
D4	Postoperative ctDNA prognosis (PMID 28899864)	ADAURA/ALINA duration (PMIDs: 37272535, 38598794)	MRD-randomized targeted-therapy duration	OQ-2, OQ-4, OQ-13
D5	CNS-active TKIs (PMIDs: 30215676, 38197815)	Neurocognitive cost of radiation	Randomized radiation-deferral design with CNS and cognition endpoints. STARLET (PMID 41534788) supplies EGFR-only intracranial-PFS evidence; the cognition endpoint is still missing	OQ-6
D6	TRACERx spatial evolution (PMID 37046096)	Oligoprogression local therapy (PMID 31067138)	Molecularly defined local-ablation versus switch trial	OQ-17, OQ-23
D7	STK11/KEAP1 immune exclusion and POSEIDON CTLA-4 subgroup signal (PMIDs: 29773717, 32866624, 39385035)	First-line chemo-IO benefit (PMID 29658856)	Dedicated biomarker-randomized validation of the CTLA-4 interaction or another rescue strategy	OQ-9
D8	HER2-ADC ILD (PMID 34534430)	HER2-selective oral TKI activity (PMID 40293180)	Randomized sequence with CNS, ILD, and QoL endpoints	OQ-10
D9	Perioperative pCR/ctDNA (PMID 35403841)	Postoperative checkpoint exposure (PMIDs: 37272513, 37870974)	Response-adapted omission of adjuvant IO	OQ-12, OQ-13
D10	Plasma whole-body sampling (PMID 30988079)	Transformation requiring tissue	Prospective triage rule for plasma-only versus mandatory biopsy	OQ-23, OQ-25
D11	Chronic TKI survival	Financial/caregiver trajectories (PMIDs: 33555936, 29476636)	Driver-specific longitudinal survivorship cohort	OQ-28
D12	Molecular patient communities	Trial access attrition (PMIDs: 18650170, 22591607)	Patient-led decentralized trial design with full cascade reporting	OQ-30

Retirement rule

An OQ is retired only when a sufficiently direct study answers the decision, not when a related biomarker association or single-arm response appears. The retirement entry should preserve the old ID, answer date, key source, remaining boundary, and pages changed.