Reliable histological and immunohistochemical assessment of invasion in small lung adenocarcinoma: reproducibility and multicenter validation of the modified classification (IMPACTS Study – Independent validation of modified pulmonary adenocarcinoma classification on T stage)

Background: Accurate recognition and quantification of stromal invasion in small (≤3 cm) non-mucinous lung adenocarcinoma are pivotal for pathological T staging (AJCC/UICC 9th edition), yet reproducibility with WHO 2021 criteria is limited, especially in collapse-rich regions. A modified, anatomically anchored framework (Filipello et al.) combines morphology with CK7–elastin mapping to separate true invasion from collapse-related pseudo-invasion. Methods: IMPACTS is a two-round, blinded, international reproducibility study on digitized whole-slide images (Pathogate®). Thirty-eight pathologists completed Round 1 using WHO criteria; thirty-six completed Round 2 after structured training applying the modified framework. Sixty-four ≤3 cm, N0 resections were reviewed in Round 1 and 62 were evaluable in Round 2. The primary endpoint was invasion (Yes/No/Not sure); secondary endpoints included true invasion (unequivocal stromal/pleural/vascular infiltration), invasive size, histologic patterns, STAS, grading, and case-level diagnostic certainty. Agreement was assessed using Fleiss’ κ, Krippendorff’s α, and Lin’s concordance (ρc). Results: Invasion reproducibility improved from fair in Round 1 (κ=0.35) to moderate in Round 2 (κ=0.44), approaching near-perfect agreement in high-certainty cases (κ≈0.99). Agreement for true invasion reached κ=0.47. Tumor size concordance increased from ρc=0.79 to 0.86. Predominant pattern showed moderate agreement (κ=0.44), whereas secondary/tertiary patterns and recognition of biologic/iatrogenic collapse remained slight-to-fair. Case-level certainty shifted toward intermediate scores, consistent with more standardized CK7–elastin–guided interpretation. Conclusions: This third independent international validation indicates that the CK7–elastin–anchored modified classification yields higher reproducibility for invasion assessment than WHO 2021, with greatest gains for explicitly defined true invasion. Limitations include the retrospective digital design and self-reported training exposure. Outcome analyses are constrained by low event numbers and are exploratory (no consistent OS associations; limited RFS signals). Ongoing work will extend outcome linkage (≥60 months), prioritize targeted re-review of discordant cases, and meta-analyze aligned cohorts to clarify prognostic and practice impact.