Multifocal ground-glass opacities in lung cancer: a narrative review of genomic evolution, immune ecology, and biology-informed clinical management

Abstract

Background and Objective

Ground-glass opacity (GGO) is a computed tomography descriptor rather than a histologic diagnosis. Multifocal GGOs (mGGOs) are increasingly detected during lung cancer screening, yet their clonal origins, immune ecology, progression risk, and optimal management remain difficult to reconcile within conventional lung cancer frameworks. This narrative review evaluates genomic, immunologic, and clinical evidence relevant to mGGOs and examines how biological findings may inform surveillance and local treatment without encouraging overtreatment.

Methods

PubMed/MEDLINE, Web of Science, Google Scholar, and selected guideline and trial sources were searched through 29 July 2026 for English-language studies published mainly from January 2020 to July 2026, supplemented by earlier landmark reports. Search concepts included mGGO, subsolid nodule, synchronous multiple primary lung cancer (SMPLC), clonality, single-cell and spatial transcriptomics (ST), immune microenvironment, active surveillance, and parenchyma-sparing surgery.

Key Content and Findings

Many mGGO presentations are compatible with clonally independent synchronous primaries, although intrapulmonary metastasis (IPM) can occur and developmental mosaicism remains a single-study, selected-patient hypothesis. Early lesions may have low mutational burden and attenuated effector activity, but immunoediting, regulatory T-cell enrichment, dysfunctional CD8-positive T-cell states, and inhibitory non-classical antigen-presentation pathways can already be present in adenocarcinoma in situ or minimally invasive adenocarcinoma. We therefore use relative adaptive immune quiescence as a descriptive population-level term, not as evidence of absent immune pressure. Prospective ECTOP-1021 data support short-term feasibility of surveillance in a highly selected cohort, but do not establish comparative effectiveness.

Conclusions

A biology-informed framework can refine risk discussion for selected patients with mGGOs, provided that radiologic terms are not treated as histologic diagnoses and mechanistic associations are not converted into clinical recommendations. Surveillance, local treatment, and parenchyma preservation should remain multidisciplinary and evidence-calibrated; biomarker-guided triage and interception strategies require prospective GGO-specific validation.

Introduction

Epidemiologic context and clinical tension

The expansion of low-dose computed tomography (LDCT) screening has reshaped the clinical spectrum of early lung cancer. Randomized screening trials primarily enrolled people with substantial smoking exposure, whereas contemporary East Asian cohorts have documented frequent detection of subsolid and multifocal lesions among never-smokers, particularly women (1-3).

These populations differ in baseline risk, histologic spectrum, and competing harms, so evidence from risk-based randomized trials cannot be transferred uncritically to non-risk-based screening. Recent debate therefore centers not only on detection but also on downstream overdiagnosis, repeated procedures, and the absence of proven mortality benefit for screening average-risk never-smokers (1-3).

Many persistent subsolid nodules grow slowly or remain radiologically stable for prolonged periods, whereas a subset develops an enlarging solid component or other features associated with invasion (4,5). Multifocal disease therefore creates a management tension between cumulative loss of lung function from repeated treatment of indolent lesions and delayed recognition of progression in a clinically dominant lesion. Staging and treatment frameworks developed for solitary tumors or systemic metastasis do not fully resolve this tension (6).

Diagnostic dilemma: intrapulmonary metastasis (IPM) versus multiple primaries

For patients with multiple pulmonary nodules, a key clinical question is whether lesions represent IPM from a single clone (with major staging and treatment implications) or synchronous multiple primary lung cancers (SMPLCs) that are potentially curable with local therapies. Historically, clinicopathologic rules (e.g., Martini-Melamed) relied on anatomic distribution and histologic subtype. However, contemporary evidence supports an integrated approach in which comprehensive histologic assessment is complemented by molecular profiling, especially when morphology is similar across lesions.

Next-generation sequencing (NGS) studies have demonstrated that even histologically similar adenocarcinomas in different lobes frequently show distinct driver events, supporting independent clonal origins in a substantial fraction of cases (7,8). These observations are compatible with field susceptibility, in which a shared tissue environment can facilitate parallel tumor initiation, including through exposure-associated inflammatory mechanisms (9). Field susceptibility is not a patient-level clonality diagnosis, however, and does not exclude related tumors or IPM; lesion-level radiologic, pathologic, and molecular assessment remains necessary (7,8,10).

From imaging phenotype to integrated biology

Integrating multiregion genomics with single-cell and spatial profiling has changed the interpretation of multifocal ground-glass opacity (GGO). Many synchronous lesions share few somatic alterations, supporting independent origins or parallel evolution, although related lesions and intrapulmonary metastases remain possible (7,8). Early subsolid lesions also show heterogeneous immune states: some have low mutational burden and reduced effector activity, whereas others already display immunoediting, regulatory T-cell enrichment, or dysfunctional CD8-positive T-cell programs (8,11-13).

These findings support a biology-informed framework in which clonality is interpreted jointly with radiology and pathology, while immune and stromal observations are used to generate risk hypotheses rather than treatment rules. Contemporary guidance similarly emphasizes multidisciplinary classification and individualized management of multifocal lesions (5-7).

Accordingly, this narrative review addresses three linked questions: (I) what genomic evidence distinguishes independent primary lesions from IPM; (II) how immune, epithelial, and stromal states change from preinvasive or GGO-predominant lesions toward invasive disease; and (III) which biological observations can reasonably inform surveillance or local treatment, and which should remain investigational? We present this article in accordance with the Narrative Review reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1871/rc).

Methods

This narrative review synthesizes biological and clinical evidence on mGGOs. We prioritized studies linking molecular evolution, immune or stromal ecology, radiologic behavior, and clinical management. To calibrate Table 1, we assigned one of three narrative evidence-strength categories (moderate, low, or very low) by considering study design, directness to mGGOs, consistency, follow-up maturity, and principal limitations. These descriptive categories were not derived from a formal GRADE or risk-of-bias assessment.

Table 1. Narrative evidence synthesis for selected clinical scenarios involving multifocal GGOs.

Clinical scenario	Evidence-based interpretation	What current evidence supports (not a recommendation)	Narrative evidence strength and key limitation
Stable multiple pure GGOs or GGO-dominant lesions	Often indolent synchronous lesions; imaging alone does not establish histology or clonality	Surveillance as an option in selected patients, with individualized intervals and reassessment triggers	Low: selected prospective single-arm and longitudinal evidence plus consensus; no randomized mGGO comparison, direct evidence for stable pure-GGO populations is limited, and long-term risk remains incompletely defined (4,5,14)
Dominant lesion no more than 2 cm with CTR no more than 0.25	Lower short-term progression risk in selected cohorts; the threshold is not universally validated	Short-term surveillance feasibility within ECTOP-1021 eligibility boundaries, not comparative effectiveness	Low: one highly selected prospective single-arm cohort; median follow-up was 35.4 months, the 5-year risk set was limited, and external validity is restricted (4)
Growth, a new or enlarging solid component, or rising CTR	Higher likelihood of invasive progression; imaging cannot establish mechanism or pathology	Renewed multidisciplinary diagnostic and staging assessment; intervention is not determined by imaging alone	Moderate: consistent longitudinal evidence and expert consensus; imaging-histology discordance and the absence of mGGO-specific randomized comparisons remain (5,14)
Small peripheral lesion selected for surgery	Localized disease in which parenchyma preservation may be feasible.	Sublobar resection in appropriately staged small peripheral NSCLC or GGO-dominant tumors; the evidence does not address every mGGO lesion	Moderate: randomized trials in small peripheral NSCLC plus prospective single-arm GGO-dominant cohorts; multifocal applicability and the cumulative effects of repeated resection are indirect (15-18)
Discordant lesion behavior or possible intrapulmonary metastasis	Separate primaries and intrapulmonary metastasis remain competing explanations	Integrated imaging and histology, with molecular profiling when morphology is similar and staging would change	Low: retrospective molecular or pathologic cohorts and committee recommendations; classification depends on sampling, assay breadth, and evolutionary divergence (7,8,10)
Candidate biomarker-guided risk stratification	Liquid-biopsy or radiomic signals may complement imaging but are not validated for mGGO management	Trial development and hypothesis generation, not routine alteration of surveillance or surgery	Very low: for mGGO management, validation is mainly in broader lung-cancer or pulmonary-nodule cohorts; no prospective mGGO-specific clinical utility has been demonstrated (19-28)
Candidate metabolic, synthetic-lethal, or immune interception	Biologically plausible, with an uncertain benefit-harm balance in indolent disease	Preclinical and early-phase research only	Very low: evidence is preclinical or cross-tumor early clinical; efficacy and safety have not been demonstrated for mGGO interception (29-37)

The narrative evidence strength reflects study design, directness to mGGOs, consistency, follow-up maturity, and key limitations. Moderate indicates direct prospective or longitudinal evidence, or consistent evidence from closely related populations, with important design or applicability limitations; low indicates evidence dominated by observational or highly selected single-arm studies; very low indicates preclinical, cross-disease, or early-validation evidence without demonstrated mGGO-specific clinical utility. These categories are descriptive and are not formal GRADE certainty ratings or clinical recommendations. CTR, consolidation-to-tumor ratio; GGO, ground-glass opacity; mGGO, multifocal ground-glass opacity; NSCLC, non-small-cell lung cancer.

The search strategy is summarized in Table 2, with a reproducible PubMed example in Table S1. Given the cross-disciplinary scope, we selected studies for relevance, methodological quality, clinical importance, and recency, retaining pre-2020 landmark reports that established diagnostic, biological, or management principles. The primary literature search was conducted on 29 June 2026, and the search was updated on 29 July 2026.

Table 2. The search strategy summary.

Items	Specification
Date of search	Primary search: 29 June 2026; final update: 29 July 2026
Databases and other sources searched	PubMed/MEDLINE, Web of Science, Google Scholar, major cited guidelines, and prospective trials or cohort reports relevant to subsolid nodules and multifocal early lung cancer
Search terms used	Combinations of MeSH/free-text terms including: multifocal ground-glass opacity OR mGGO OR ground-glass nodule OR subsolid nodule; synchronous multiple primary lung cancer; intrapulmonary metastasis; lung adenocarcinoma; clonality; field cancerization; developmental mosaicism; EGFR; RBM10; tumor immune microenvironment; immunoediting; regulatory T cell; tissue-resident memory T cell; tertiary lymphoid structure; HLA-E; NKG2A; alveolar macrophage; FABP4; SPP1; single-cell RNA sequencing; spatial transcriptomics; active surveillance; ECTOP-1021; parenchyma-sparing surgery; segmentectomy; liquid biopsy; cfDNA; radiomics
Timeframe	Mainly January 2020 to July 2026, with earlier landmark classification, mechanistic, and clinical studies included when necessary
Inclusion and exclusion criteria	Included English-language original studies, prospective or retrospective clinical cohorts, translational studies, guidelines, consensus documents, and high-quality reviews addressing multifocal GGOs, subsolid nodules, early lung adenocarcinoma biology, or management. Excluded articles not focused on lung cancer/subsolid nodules, non-English articles, conference abstracts without full text, and studies without clear relevance to genomic, immune, radiologic, or clinical management questions
Selection process	Initial article screening and evidence assembly were performed by X.L., Y.L., and J.J. Disagreements about inclusion or interpretation were resolved through discussion with Z.X., B.Z., and N.Y. until consensus was reached
Additional considerations	This was a narrative review; therefore, no formal risk-of-bias scoring or quantitative evidence synthesis was performed. Mechanistic and biomarker findings were interpreted separately from clinically validated management evidence

cfDNA, cell-free DNA; ECTOP-1021, Early Lung Cancer With Multiple Ground-Glass Opacities Treated by Surgery or Active Surveillance; EGFR, epidermal growth factor receptor; FABP4, fatty acid-binding protein 4; GGO, ground-glass opacity; MeSH, Medical Subject Headings; mGGO, multifocal ground-glass opacity; RBM10, RNA binding motif protein 10; SPP1, secreted phosphoprotein 1.

During revision, OpenAI Codex (GPT-5; accessed 29 July to 1 August 2026) was used to assist with language editing, bibliographic organization, reference-metadata cross-checking, and document formatting. An OpenAI image-generation tool (accessed 29 July 2026) was used to assist with the visual rendering and refinement of author-designed schematic figures. The authors reviewed, edited, and verified all text, references, and figures and take full responsibility for the submitted manuscript.

Radiologic and pathologic terminology

The 2024 Fleischner Society glossary defines ground-glass as increased computed tomography (CT) attenuation that does not completely obscure underlying bronchial and vascular structures. A pure ground-glass nodule contains only ground-glass attenuation, whereas a part-solid nodule contains both ground-glass and soft-tissue components; both are subsolid nodules (38). These radiologic terms describe appearance, not histology, and can reflect benign, inflammatory, preinvasive, or invasive processes.

Radiologic-pathologic correlations are probabilistic. Ground-glass components often correspond to lepidic growth and solid components more often to invasion, but CT cannot by itself diagnose adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), or invasive adenocarcinoma (38). The 2021 World Health Organization (WHO) classification places atypical adenomatous hyperplasia and AIS under precursor glandular lesions, while MIA remains within adenocarcinoma (39). This distinction is material to discussions of screening-associated overdiagnosis: counting or treating every persistent GGO as established invasive cancer can inflate disease burden and expose patients to unnecessary procedures.

Genomic heterogeneity and clonal independence

Classification of synchronous mGGOs has shifted from anatomy- and morphology-based inference toward integrated histologic and molecular clonality assessment (7,8,10). Many ground-glass-dominant multifocal presentations are compatible with independent tumorigenesis rather than dissemination from one index lesion. Multifocality remains heterogeneous, however, and can include clonally related tumors, stochastic parallel evolution, and, in selected contexts, shared pre-neoplastic lineages that predate clinical detection (40,41).

Genomic evidence for clonal independence

High-throughput sequencing provides a framework to infer clonal relationships among multiple lesions and complements morphology and comprehensive histologic assessment (10). Phylogenetic inference integrates shared and private somatic alterations [including single nucleotide variants (SNVs)/indels and copy-number changes] to evaluate whether lesions are more consistent with dissemination versus parallel evolution, contingent on adequate sampling and appropriate analytical controls (42,43).

Across representative studies of synchronous lesions and TRACERx evolutionary analyses, a recurring observation is that synchronous GGOs frequently show limited overlap of somatic alterations within the same patient (8,10,43). In metastatic disease, primary and metastatic lesions typically share early truncal events, with divergence arising from later branch alterations acquired after dissemination (44). By contrast, many mGGO presentations demonstrate predominantly private passenger mutation spectra and divergent copy number variation (CNV) architectures, supporting separate clonal origins (8,10).

Importantly, shared involvement of a driver gene does not necessarily imply shared ancestry. When lesions harbor alterations in the same pathway (e.g., EGFR), deeper profiling may reveal distinct nucleotide-level events and/or divergent passenger and CNV architectures, consistent with convergent evolution under shared selective pressures (45,46). Accordingly, clonality interpretation should be integrated across radiology, pathology, and genomics, and conclusions should be framed probabilistically rather than as absolute (10).

From a management perspective, these findings support treating many synchronous mGGOs as separate early-stage entities when clinical, radiologic, and molecular features do not suggest dissemination, consistent with parenchyma-sparing surgical strategies in small peripheral non-small cell lung cancer (NSCLC) and GGO-dominant tumors (15,16). Nonetheless, related lesions can occur in a minority of cases; therefore, clonality assessment should remain individualized and multidisciplinary, particularly when radiologic patterns, interval kinetics, or nodal findings raise concern for spread (10).

Developmental mosaicism as a complementary model

Exposure-based field effects remain biologically plausible; for example, air-pollution-associated inflammation can promote expansion of EGFR-mutant lung epithelial clones (9). Such effects do not explain every multifocal presentation. In one lineage-inference study of selected patients with multiple EGFR-mutant tumors, Burr and colleagues identified anatomically separated tumors sharing a distant, non-germline ancestor compatible with an early developmental origin (40).

This observation provides a complementary model of distributed lineage susceptibility but does not establish that developmental mosaicism is common. The study did not demonstrate central immune tolerance, and developmental ancestry should not be used to guide treatment. We therefore treat mosaicism as hypothesis-generating; continued surveillance is justified by lesion multiplicity and natural history, not by inferred developmental origin (5,40).

Immune ecology of early ground-glass lesions

We use relative adaptive immune quiescence operationally to describe reduced productive effector activation in some early or subsolid lesions compared with more invasive counterparts. The term does not imply absent immune pressure. It is distinct from low immunogenicity, a tumor property related to antigen availability, and from T-cell exhaustion, an antigen-experienced dysfunctional state characterized in lung cancer (47). Available data instead support heterogeneous and evolving immune states across AIS, MIA, and invasive disease (8,13,48).

Mutational burden, immune pressure, and antigen presentation

The genomic architecture of some early lesions may reduce neoantigen availability, but immune pressure can still act before histologic invasion.

Low mutational burden and neoantigen availability

Whole-exome sequencing indicates that AIS/MIA and pure GGO-predominant lesions can have lower tumor mutational burden than invasive or solid adenocarcinomas (13). Fewer expressed nonsynonymous mutations may reduce neoantigen availability, but mutational burden alone does not determine immune recognition. Low neoantigen expression can impair T-cell priming in experimental colorectal cancer (49); this cross-tumor observation is mechanistically informative but cannot be assumed to explain every lung GGO.

Early immunoediting and functional antigen-presentation constraints

Multiregion sequencing has detected immunoediting and intratumor heterogeneity in AIS and MIA before acquisition of overt invasion (13), while spatial multi-omics has identified checkpoint-enriched immune interactions in lung adenocarcinoma precursors (50). Retention of classical human leukocyte antigen (HLA) class I alleles is therefore not evidence against immune pressure: antigen processing can be functionally constrained, and non-classical HLA-E signaling through the inhibitory NKG2A receptor can suppress CD8-positive T cells and natural killer cells in lung cancer (51).

Adaptive immune organization and suppression

Organized adaptive immunity should also be considered. In resected NSCLC, mature dendritic cells within tertiary lymphoid structures are associated with local type 1 and cytotoxic immune responses, and CD8-positive CD103-positive tissue-resident memory T cells can be tumor-reactive and prognostically favorable (52,53). These observations provide relevant lung-cancer context, but the prevalence and function of tertiary lymphoid structures and tissue-resident memory T cells in mGGOs remain incompletely defined.

Conversely, single-cell profiling of subsolid nodules shows increased regulatory T cells, reduced effector T-cell activity, and diminished natural killer or natural killer T-cell cytotoxic programs (48). In synchronous lesions, progression from preinvasive to invasive adenocarcinoma is accompanied by more regulatory T cells and exhausted CD8-positive T cells (8). These data support relative adaptive immune quiescence as a heterogeneous balance of reduced effector function and active suppression, not as immunological ignorance.

Macrophage states in early and invasive lesions

Single-cell studies implicate macrophage-state changes across early and invasive lung adenocarcinoma, but these associations are neither universal nor proven to be the initiating cause of radiologic progression (48,54,55).

FABP4-positive alveolar macrophages and metabolic support

Normal alveolar niches contain tissue-resident macrophages expressing FABP4 and MARCO, and experimental early-NSCLC models show that resident macrophages can support tumor initiation (54,56). In EGFR-driven lung adenocarcinoma models, tumor-derived granulocyte-macrophage colony-stimulating factor (GM-CSF) activates a peroxisome proliferator-activated receptor gamma (PPARγ)-linked metabolic program in alveolar macrophages that supports lipid handling and EGFR signaling (29). More broadly, nutrient partitioning in tumors is cell-type specific (57), but this general principle should not be cited as direct evidence for a GM-CSF/PPARγ mechanism in human mGGOs.

SPP1-positive macrophages and invasion-associated remodeling

Progression from pure or part-solid GGO toward invasive adenocarcinoma has been associated with loss of resident-like alveolar macrophage programs and enrichment of monocyte-derived SPP1-positive macrophage states (54,55). Broader studies implicate SPP1-CD44 or integrin signaling in macrophage-associated tumor and immune-cell communication and matrix-related programs (58,59). These data nominate SPP1-positive macrophages as correlates of an invasion-associated niche, not as a universal trigger or a validated clinical biomarker.

Macrophage checkpoint signaling

The CD47-SIRPα axis is a broadly studied inhibitor of macrophage phagocytosis in NSCLC (60). However, GGO-specific expression, causality, and clinical relevance have not been established. It should therefore be viewed as a candidate macrophage checkpoint rather than an established mechanism of mGGO persistence.

The immunological progression model is summarized in Figure 1.

Figure 1.

Proposed macrophage and stromal states across GGO-predominant progression. This evidence-calibrated schematic separates reported associations from candidate mechanisms. FABP4-positive alveolar macrophage and GM-CSF/PPARγ-linked metabolic support are derived mainly from experimental early-NSCLC or EGFR-driven models; CD47-SIRPα is shown as a candidate macrophage checkpoint without GGO-specific causal validation. SPP1-positive macrophage, stromal activation, and matrix remodeling are associated with invasive regions but are not established universal drivers. Original author-created schematic; no third-party material is reproduced. AIS, adenocarcinoma in situ; AM, alveolar macrophage; CAF, cancer-associated fibroblast; CD44, cluster of differentiation 44; CD47, cluster of differentiation 47; ECM, extracellular matrix; GM-CSF, granulocyte-macrophage colony-stimulating factor; GGO, ground-glass opacity; MIA, minimally invasive adenocarcinoma; MMP, matrix metalloproteinase; NSCLC, non-small cell lung cancer; PPARγ, peroxisome proliferator-activated receptor gamma; SIRPα, signal regulatory protein alpha; SPP1, secreted phosphoprotein 1; VEGF, vascular endothelial growth factor.

Single-cell epithelial states associated with indolent growth

Single-cell transcriptomics has identified regeneration-linked epithelial states in lung injury and cancer that may help explain lepidic growth, although no single cell state defines all radiologic GGOs (61,62).

KRT8-positive transitional epithelial states

Studies of alveolar injury and fibrosis have identified KRT8⁺/CLDN4⁺ transitional states, termed the KRT8⁺ alveolar differentiation intermediate (ADI) or damage-associated transient progenitor (DATP) (63,64). Related regeneration-linked programs have been observed across early lung adenocarcinoma, GGO-associated lesions, and lung-regeneration datasets (65-67), but these studies do not establish ADI/DATP cells as a specific cellular substrate of radiologic GGOs. During alveolar repair, alveolar epithelial type 2 cells (AT2) stem cells can transiently enter a KRT8⁺/CLDN4⁺ state before differentiating toward alveolar epithelial type 1 cells (AT1) cells (63,68).

In experimental models, oncogenic KRAS signaling can stabilize or co-opt transitional epithelial programs and promote early lung adenocarcinoma or lepidic growth (65,69). Some transitional cells express stress- or senescence-associated factors, including CTGF and AREG, that can remodel the local matrix (62,65,70). These models provide a plausible link between injury-response states and GGO-predominant adenocarcinoma, but they do not establish a universal differentiation arrest in human mGGOs.

Activator protein 1 (AP-1) and Yes-associated protein/transcriptional coactivator with PDZ-binding motif (YAP/TAZ) signaling

AP-1 factors can cooperate with YAP/TAZ during cellular transformation in breast models (71), while YAP/TAZ are broadly regulated by mechanical cues such as matrix stiffness across regenerative and disease contexts (72). Together, these studies provide a cross-context signaling hypothesis, not direct evidence that an AP-1/YAP/TAZ circuit maintains human GGOs or can be targeted therapeutically.

RNA-binding motif protein 10 (RBM10) loss and context-dependent cell motility

RBM10 loss-of-function occurs in a subset of lung adenocarcinomas, including EGFR-mutant tumors, but direct evidence of enrichment in GGO-predominant lesions is incomplete (73). Its clinical interpretation should therefore be separated from mechanistic findings in experimental lung adenocarcinoma systems.

In preclinical models, RBM10 loss alters splicing of cytoskeletal and extracellular-matrix transcripts. Inclusion of VCL exon 19 generates metavinculin and changes focal-adhesion dynamics, increasing cell motility and metastatic behavior (74).

These experiments do not establish that RBM10 loss causes lepidic spreading during the GGO phase, nor that it predicts progression in patients with subsolid nodules. RBM10 status is therefore a hypothesis-generating molecular feature rather than a clinical risk marker for mGGOs (73,74).

RBM10-deficient lung adenocarcinoma models also show replication stress and sensitivity to WEE1 inhibition (75). This is a preclinical vulnerability and should not be interpreted as a prevention or treatment option for GGO-predominant disease.

The epithelial-state and RBM10-associated mechanisms are summarized in Figure 2.

Figure 2.

Regeneration-linked epithelial states and preclinical RBM10 mechanisms. The KRT8-positive/CLDN4-positive and AP-1/YAP/TAZ panels summarize experimental models that may contribute to lepidic growth; they do not define all human GGOs. The RBM10 panel depicts splicing, motility, and WEE1-vulnerability findings from preclinical lung adenocarcinoma models. A causal link between RBM10 loss and GGO-phase lepidic spreading, clinical progression, or treatment response has not been established. Original author-created schematic; no third-party material is reproduced. ADI, alveolar differentiation intermediate; AP-1, activator protein 1; AT2, alveolar epithelial type 2 cell; CLDN4, claudin 4; DATP, damage-associated transient progenitor; ECM, extracellular matrix; GGO, ground-glass opacity; KRT8, keratin 8; mGGO, multifocal ground-glass opacity; RBM10, RNA-binding motif protein 10; SASP, senescence-associated secretory phenotype; TAZ, transcriptional coactivator with PDZ-binding motif; VCL, vinculin; WEE1, WEE1 G2 checkpoint kinase; YAP, Yes-associated protein.

Spatial organization of lepidic and invasive compartments

Spatial transcriptomics (ST) enables in situ delineation of transcriptional programs while preserving tissue context, thereby resolving spatially organized tumor-stroma-immune interactions that are not captured by dissociated profiling (76,77). In part-solid pulmonary nodules, region-resolved single-cell studies distinguish ground-glass and solid components (66). Complementary multiregion single-cell and spatial lung adenocarcinoma studies identify distinct lepidic/invasive histologic and multicellular neighborhoods (78-80). Considered together, these datasets support a compartmentalized model broadly separating a lepidic-aligned domain from an invasive-aligned front, but they should not be described as if all modalities were co-assayed in the same cohorts.

Lepidic-aligned compartment

The ground-glass component of a part-solid GGO typically corresponds to a lepidic-aligned compartment in which alveolar architecture is relatively preserved and stromal activation is limited compared with invasive regions (11,66,79). Spatially resolved and single-cell analyses indicate enrichment of epithelial stress/transitional programs (e.g., KRT8⁺/CLDN4⁺ injury-associated differentiation intermediates), consistent with a regeneration-linked epithelial state rather than fully established invasive phenotypes (66,67,79,80).

From an immune perspective, this compartment is often enriched for alveolar macrophage programs (e.g., FABP4⁺/MARCO⁺), resembling tissue-resident alveolar macrophage states observed in adjacent alveolar regions and in early/subsolid lesions (48,56,66,80). Notably, macrophage phenotypes in early-stage contexts may also adopt tumor-supportive and/or senescence-associated functional programs, suggesting that “resident-like” identity does not necessarily imply antitumor activity (81). Collectively, these features align with a relatively contained microenvironment in the lepidic-aligned region and may contribute to indolent clinical behavior in a subset of part-solid GGO lesions (12,48,56,66).

Invasive-aligned compartment

The solid component of a part-solid GGO corresponds to an invasive-aligned compartment characterized by intensified tumor-stroma coupling, including stromal remodeling and immune reconfiguration (66,78-80). Spatially informed analyses implicate activation of cancer-associated fibroblast (CAF) programs and collagen-rich matrix organization that can restrict effector lymphocyte access to tumor nests, supporting an immune-excluded phenotype (79,82,83).

Concurrently, invasive-associated regions show enrichment of SPP1+ macrophage states across mixed ground-glass nodule and lung adenocarcinoma datasets (55,79,80). In mixed-GGO specimens, macrophage ligand-receptor analyses identified SPP1-related interactions during malignant progression; broader pan-cancer and NSCLC studies implicate SPP1-CD44 signaling in macrophage-tumor or macrophage-T-cell communication (55,58,84). These observations support communication-associated immunosuppressive remodeling but do not establish that SPP1-CD44 directly drives extracellular-matrix deposition in mGGOs. A recent review summarizes the broader, context-dependent effects of osteopontin in the tumor microenvironment (85). T cells detectable within or adjacent to invasive compartments often exhibit dysfunctional or exhaustion-associated transcriptional features, including TIM-3 and LAG-3 programs (50,86). Together, these data support an association between invasive neighborhoods, CAF activation, SPP1⁺ macrophage states, immune exclusion, and T-cell dysfunction, while the causal sequence remains unresolved (50,79,82,83,86).

The spatial transition from lepidic- to invasive-aligned neighborhoods is summarized in Figure 3.

Figure 3.

Spatially associated lepidic and invasive compartments. The schematic summarizes associations reported by single-cell and spatial studies: lepidic-aligned regions often retain alveolar architecture and resident-like macrophage programs, whereas invasive-aligned regions can show SPP1-positive macrophages, fibroblast activation, extracellular-matrix remodeling, immune exclusion, and dysfunctional T-cell states. The arrow represents an inferred continuum rather than a proven universal sequence. Original author-created schematic; no third-party material is reproduced. CAF, cancer-associated fibroblast; CD44, cluster of differentiation 44; CLDN4, claudin 4; ECM, extracellular matrix; FABP4, fatty acid-binding protein 4; GGO, ground-glass opacity; KRT8, keratin 8; LAG-3, lymphocyte-activation gene 3; MARCO, macrophage receptor with collagenous structure; SPP1, secreted phosphoprotein 1; TIM-3, T-cell immunoglobulin and mucin-domain containing-3.

Clinical management: evidence and limits

Prospective evidence for surveillance of mGGOs is emerging but remains limited. Available studies and expert guidance support surveillance as a reasonable option for carefully selected patients; they do not establish superiority over immediate surgery or justify applying one protocol to all subsolid nodules (4,5,87).

Prospective evidence for active surveillance: ECTOP-1021

ECTOP-1021 (NCT06097910) is a prospective, multicenter, single-arm surveillance study in patients with at least three GGOs and a dominant lesion measuring no more than 2 cm with consolidation-to-tumor ratio (CTR) no more than 0.25 (88). The results report enrolled 406 patients across five centers and a median follow-up of 35.4 months. Five-year overall survival was estimated at 100%; because no deaths occurred, a two-sided 95% confidence interval was not estimable, and 89 patients remained at risk at 60 months. Radiologic progression occurred in 8.1% of patients, 1.5% developed new lesions, and eight underwent surgery; all resected lesions were stage IA1, with no nodal or distant metastases reported (4). These findings support short-term feasibility in a highly selected cohort. The single-arm design, limited 5-year risk set, median follow-up shorter than 5 years, and demographic concentration in younger female never-smokers preclude claims of validation or comparative effectiveness (4,5).

Surgical studies provide indirect reassurance that deferred local treatment can remain effective for selected small lesions. JCOG0804/WJOG4507L reported 99.7% 5-year relapse-free survival after sublobar resection for tumors no more than 2 cm with CTR no more than 0.25 (17), and JCOG1211 supports segmentectomy for selected GGO-dominant tumors no more than 3 cm (18). Randomized JCOG0802/WJOG4607L and CALGB/Alliance 140503 trials support sublobar resection for carefully staged small peripheral NSCLC, but these populations were not equivalent to untreated mGGOs (15,16). Thus, surgical evidence supports the availability of parenchyma-sparing treatment when progression triggers are met, not the safety of indefinite surveillance for every lesion (5,16,18).

Overdiagnosis and risk-adapted intervention

Detection of indolent subsolid lesions, particularly among never-smokers in East Asia, has intensified concern about overdiagnosis and overtreatment (89-92). A registry-based analysis of approximately 3.21 million residents in Shanghai estimated substantial increases in overdiagnosed lung adenocarcinoma after widespread LDCT use in average-risk populations, especially among women (93). Regional commentary and analyses from South Korea, Taiwan, and other East Asian settings likewise caution that stage shift and high survival do not by themselves establish a mortality benefit (89,91,92). A 2025 meta-analysis further showed that overdiagnosis estimates vary considerably with study design and follow-up duration (90).

These observations support risk stratification rather than uniform immediate resection for all detected nodules (5,87). Contemporary expert consensus for subsolid nodules recommends using dynamic radiologic markers—particularly growth kinetics and development or increase of a solid component (CTR change)—and emphasizes individualized decision-making in multifocal disease (5). Long-term natural history studies indicate that a subset of pure GGNs can grow even after prolonged stability (e.g., after 10 years), supporting continued but proportionate follow-up in selected patients (14).

When intervention is indicated, CTR-informed assessment can guide the extent of surgery and staging intensity. Radiologic studies continue to refine the prognostic and clinicopathologic utility of CTR categories in early-stage NSCLC (94). Contemporary surgical series report very low rates of nodal metastasis among small nodules containing a GGO component, supporting a tailored approach to lymph node evaluation in carefully selected GGO-predominant lesions (95). At the guideline level, recommendations continue to evolve; recent CHEST and ESMO updates emphasize evidence-based selection of local therapy and structured follow-up (87,96).

Table 1 summarizes what current evidence supports in selected clinical scenarios rather than issuing management recommendations. Its narrative grades are linked to representative sources and scenario-specific limitations and should not be interpreted as formal GRADE certainty ratings or substitutes for multidisciplinary judgment.

Emerging biomarkers and investigational interception

Management of mGGOs is increasingly informed by lesion biology. At present, the most actionable consequence is improved risk stratification and more disciplined selection of surveillance or local therapy; interception-oriented strategies should be viewed as investigational hypotheses that require prospective validation before clinical adoption.

Liquid-biopsy approaches

Mutation-focused ctDNA assays, including assays targeting EGFR variants, have limited sensitivity in small GGO-associated or early-stage lung cancers because tumor fraction and ctDNA shedding are often low (19,20). This constraint has motivated approaches that do not rely on detecting a single mutation, but instead capture genome-wide cell-free DNA (cfDNA) features (21-23).

Fragmentomics is one such strategy. Platforms such as DELFI infer tumor signal from fragmentation patterns across the genome, which can retain discriminatory information at low tumor fraction (21,23). Clinical validation studies support fragmentome assays for early lung cancer detection (21,22), while integrated imaging-cfDNA models have demonstrated pulmonary-nodule classification in broader screening-relevant cohorts (24). Methodological work has further established the analytical foundation for fragmentome-based detection (23). These findings provide proof of principle that fragment-based signals can complement imaging, but they do not establish clinical utility for mGGO surveillance.

cfDNA methylation profiling provides an orthogonal signal that may appear early in carcinogenesis and can be used for risk triage (25,26). Large studies have shown that models combining clinical variables, imaging features, and methylation markers can improve classification of pulmonary nodules (27). Additional multi-modal frameworks integrating methylation biomarkers with CT-based predictors have supported methylation as a complementary risk feature for early disease assessment (25). Work published in Lung Cancer has also reported methylation-based modeling approaches for non-invasive nodule assessment, underscoring growing clinical interest in this direction (26). Overall, these data are consistent with a broader shift in early-stage NSCLC liquid biopsy from mutation tracking to multi-feature cfDNA biology (fragmentation and methylation), particularly for low-shedding lesions (28).

Investigational intervention concepts

Early-lesion ecosystems suggest potential biological targets, but they do not support pharmacologic prevention or interception in routine practice. Metabolic, synthetic-lethal, and immune-based approaches remain preclinical or early clinical concepts; any trial would require stringent eligibility, low-toxicity interventions, clinically meaningful endpoints, and safeguards against treating lesions that might never cause harm.

Metabolic targets

Experimental data in EGFR-driven lung adenocarcinoma indicate that tumor-derived cues (including GM-CSF) can shape alveolar macrophage programs through a GM-CSF–PPARγ-linked axis, supporting tumor-associated metabolic states (29). This is biologically consistent with broader evidence that GM-CSF contributes to alveolar macrophage programming in the lung niche (30). These findings are mechanistically informative, but there is currently no validated metabolic intervention for patients with mGGOs; translation should proceed only through carefully designed early-phase or prevention-oriented studies (29,30).

Synthetic lethality-based targets

Genotype-defined vulnerabilities may eventually help refine treatment or prevention trials, but they are not ready for routine management of GGO-predominant disease. Functional genomics has identified WEE1 as a synthetic-lethal dependency in RBM10-deficient lung adenocarcinoma models, providing a mechanistic rationale for WEE1 inhibition in RBM10-altered disease (75). RBM10 alterations have also been linked to treatment response in EGFR-mutant settings (31). However, toxicity and the favorable natural history of many mGGOs (5,14) make prevention-like use of such agents inappropriate outside prospective research; pharmacologic targeting of RBM10 remains preclinical (32,75).

Immune-based prevention concepts

Precision-prevention frameworks are being developed, but no vaccine has demonstrated efficacy for mGGOs (33). A phase 1b/2 neoantigen-vaccine study in Lynch syndrome provides cross-tumor prevention feasibility rather than lung-specific evidence (34), while personalized vaccines in pancreatic cancer and melanoma show that neoantigen-directed T-cell responses can be induced in therapeutic settings (35,36). These findings cannot be used to guide surveillance or surgery for mGGOs; immune interception should remain confined to prospective trials (37).

A biology-informed surveillance framework, with investigational biomarker and interception directions clearly separated from current clinical management, is summarized in Figure 4.

Figure 4.

Evidence-calibrated surveillance and research framework. The surveillance criteria shown derive from selected cohorts and expert guidance and are not universal recommendations. ECTOP-1021 was single-arm and had limited mature follow-up; growth or increasing solid component should prompt multidisciplinary reassessment rather than automatic intervention. Fragmentomics, cfDNA methylation, radiomics, and metabolic, synthetic-lethal, or immune interception remain investigational. Original author-created schematic; no third-party material is reproduced. cfDNA, cell-free DNA; CT, computed tomography; CTR, consolidation-to-tumor ratio; GGO, ground-glass opacity; LDCT, low-dose computed tomography; mGGO, multifocal ground-glass opacity.

Strengths and limitations of this review

The principal strength of this review is its integration of genomic, single-cell, spatial, radiologic, surgical, and prospective surveillance evidence. Limitations include its narrative, non-systematic approach; heterogeneity in radiologic definitions, pathology, sampling, and molecular platforms; the predominance of retrospective or single-center datasets; and limited maturity of prospective surveillance data. Several mechanistic sections extrapolate from invasive NSCLC or experimental models rather than GGO-specific cohorts, and developmental mosaicism is based chiefly on one study. Evidence-strength labels in Table 1 are descriptive rather than formal GRADE certainty ratings and do not imply a guideline-development process. The associations summarized here should not be interpreted as causal or prescriptive.

Conclusions

mGGO is a radiologic presentation, not a single histologic or biological entity. Many lesions are compatible with synchronous primaries and parallel evolution, whereas related lesions and intrapulmonary metastases can occur; developmental mosaicism is a complementary, selected-patient hypothesis. Early immune states are likewise heterogeneous. Relative adaptive immune quiescence may coexist with early immunoediting, regulatory T-cell enrichment, dysfunctional CD8-positive T-cell programs, and inhibitory HLA-E/NKG2A signaling. Progression is associated with changes in epithelial state, SPP1-positive macrophages, fibroblasts, and extracellular matrix, but causal sequence and clinical utility remain uncertain. ECTOP-1021 supports short-term surveillance feasibility in a selected single-arm cohort, not comparative effectiveness. Clinical management should therefore combine longitudinal imaging, pathology, selective clonality assessment, patient preferences, and preservation of lung function. Liquid-biopsy and interception strategies remain investigational until prospectively validated in GGO-specific populations.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.