An aggressive histological subtype. Micropapillary lung adenocarcinoma (MPA) is a distinct histological subtype of invasive lung adenocarcinoma characterized by a high propensity for lymphovascular invasion, distant dissemination, and early metastasis. Even among patients with stage I disease, the 5-year overall survival rate for those with an MPA component ranges from 54% to 79%, compared to 73% to 98% in those without MPA. MPA has been identified as a significant predictor of postoperative brain metastasis following surgical resection.
The clinical need for preoperative identification. Histopathological examination remains the gold standard for subtype classification, but intraoperative frozen section analysis has a sensitivity of only approximately 37% for detecting MPA, due to inconsistencies between frozen sections and formalin-fixed paraffin-embedded specimens. Because the presence of MPA influences surgical approach, extent of resection, and indication for adjuvant chemotherapy, reliable preoperative prediction of MPA would substantially improve treatment planning in early-stage disease.
Current gaps in predictive tools. While radiomics and artificial intelligence approaches integrating CT-derived features have shown promise in identifying high-grade adenocarcinoma components, there remains a lack of simple, interpretable, and clinically applicable predictive models specifically tailored for MPA. Existing tools are often complex, not externally validated, or not designed for the routine clinical setting where manual CT interpretation is standard.
Study objectives. This study retrospectively analyzed 680 patients with stage IA-IB lung adenocarcinoma, comparing those with and without MPA, to identify independent preoperative risk factors and construct a logistic regression nomogram with a corresponding online calculator for clinical use.
Single-center retrospective cohort. Patients diagnosed with lung adenocarcinoma at West China Hospital, Sichuan University between January 2011 and June 2022 were included. Two experienced pathologists (each with over 10 years of diagnostic experience) independently classified tumors as MPA or non-MPA based on the presence of a micropapillary component; cases of disagreement were adjudicated by a senior pathologist. Solid adenocarcinoma cases were excluded to ensure the non-MPA comparator group consisted only of lepidic, acinar, and papillary combinations.
1:1 matched cohort of 680 patients. Due to a significant disparity in patient numbers between MPA and non-MPA groups, 1:1 age and gender matching was performed, yielding 340 MPA patients and 340 non-MPA controls. Clinical, pathological, and imaging data were collected for both groups, with follow-up conducted through outpatient visits or telephone interviews. The primary outcome was disease-free survival.
CT imaging feature assessment. High-resolution chest CT scans were reviewed using standard lung window settings. Two pulmonologists independently evaluated each scan, with a third providing adjudication for discordant cases. Features assessed included: tumor size (maximum longitudinal diameter), nodule density (solid nodule, ground-glass opacity, or partially-solid nodule), lobulation, spiculation, vacuoles, and pleural retraction. Inter-observer consistency was evaluated using Cohen's kappa for categorical features and intraclass correlation coefficient for continuous measures.
Volume Doubling Time calculation. Volume doubling time (VDT) was calculated from two serial CT scans (minimum 2-week interval) using the formula VDT = (t x log2) / log(Vt/V0), where t is the interval between scans, V0 is initial volume, and Vt is preoperative volume. Tumor volume was computed from maximum diameters as V = (pi/6) x a x b squared. Missing VDT values were handled by multiple imputation.
Dramatically higher recurrence in MPA. Tumor recurrence occurred in 23.2% of MPA patients versus only 2.0% of non-MPA patients (p less than 0.001). The most common recurrence sites in MPA were intrapulmonary (33.3%), bone (24.6%), and brain (21.7%), reflecting the aggressive systemic dissemination pattern of this subtype. All-cause mortality was 4.7% in the MPA group versus 1.3% in the non-MPA group (p = 0.026), and all 13 cancer-specific deaths occurred in the MPA group.
Pathological features of MPA. Tumors in the MPA group were significantly larger (mean 23.2 vs 15.9 mm, p less than 0.001) and showed higher rates of spread through air spaces (STAS: 85.9% vs 16.5%), vascular invasion (8.5% vs 0.3%), and pleural involvement (46.8% vs 14.7%). The MPA group was predominantly composed of acinar and papillary predominant patterns rather than lepidic, and had a significantly lower degree of differentiation overall.
Imaging characteristics distinguishing MPA. Solid nodules were dramatically more prevalent in MPA (73.2% vs 12.6%), while ground-glass nodules predominated in non-MPA (50.0% vs 5.3%). Lobulation was present in 98.2% of MPA versus 72.4% of non-MPA cases, and spiculation in 95.0% versus 58.8%. Vacuoles appeared in 66.2% of MPA versus 44.1% of non-MPA. Volume doubling time was substantially shorter in MPA (median 268.7 days vs 545.9 days), indicating faster tumor growth.
MPA as an independent prognostic factor. Multivariable Cox regression analysis confirmed that the presence of an MPA component was an independent risk factor for worse prognosis (HR 2.921, 95% CI 1.007 to 8.470, p = 0.048), after adjusting for surgery type, pathological features (STAS, vascular invasion, pleural involvement), and radiological characteristics. MPA proportion (above vs below 20% of tumor) was not significantly associated with prognosis, suggesting that any MPA component confers risk regardless of extent.
Optimal cutoffs for continuous predictors. ROC curve analysis identified CT-diameter cutoff at 22.5 mm (identifying tumors above this threshold as higher-risk) and VDT cutoff at 310 days (identifying rapidly growing tumors). These thresholds were then used as binary predictors in subsequent logistic regression modelling.
Eight independent risk factors identified. Multivariate logistic regression identified eight independent predictors of MPA: smoking history (OR 1.866), CT-diameter 22.5 mm or above (OR 2.622), partially-solid nodule (OR 3.077), solid nodule (OR 25.604), lobulation (OR 5.194), spiculation (OR 2.184), vacuoles (OR 2.650), and VDT at or below 310 days (OR 5.250). Multicollinearity was negligible across all predictors by VIF analysis.
Nomogram construction and web calculator. A nomogram was constructed assigning each predictor points based on its logistic regression coefficient, with total score mapped to predicted MPA probability. An interactive online calculator was developed and published at https://mpariskassessment.shinyapps.io/MPA-risk-prediction/ to allow clinicians to input patient-specific values and receive real-time probability estimates. Using the Youden index, the optimal probability cutoff was set at 48.13%, yielding sensitivity 87.35%, specificity 86.18%, positive predictive value 86.34%, and negative predictive value 87.20%.
Model validation and calibration. Bootstrap-corrected calibration using 1,000 resamples showed the bias-corrected calibration curve closely overlapping the ideal reference line. The Hosmer-Lemeshow test returned p = 0.1575, confirming good calibration. Subgroup analysis showed consistent discrimination across patients with CT-diameter below 2 cm (AUC 0.899), above 2 cm (AUC 0.906), smokers (AUC 0.928), and non-smokers (AUC 0.917), demonstrating stable performance across important clinical subpopulations.
Risk stratification with direct prognostic consequence. Among high-risk patients (nomogram probability above 48.13%), 39.06% experienced recurrence within 2 years, while all recurrences in the low-risk group occurred more than 2 years after surgery (p = 0.023). High-risk patients were also more likely to develop recurrence involving more than two distant sites, indicating more aggressive systemic disease biology.
Lobectomy vs. sublobar resection in MPA. Among MPA patients, surgical approach did not produce statistically significant survival differences, but Kaplan-Meier analysis suggested that long-term outcomes after lobectomy tended to be inferior to those after sublobar resection. This counterintuitive finding aligns with prior research showing that segmentectomy preserves more lung function (with less reduction at 6 and 12 months) and may offer more benefit to MPA patients whose recurrence pattern involves distant sites rather than local control.
Adjuvant chemotherapy in stage IB MPA. Among stage IB MPA patients, adjuvant chemotherapy did not yield a statistically significant overall survival advantage, but a trend toward improved long-term prognosis was observed in the chemotherapy group. These trends support the potential benefit of adjuvant treatment in high-risk MPA patients, consistent with prior multicenter studies, though the statistical power in this single-center cohort was insufficient to confirm this definitively.
VDT as an underutilized biomarker. Volume doubling time provided unique predictive value beyond conventional CT morphological features. The median VDT in MPA cases (268.7 days) closely aligns with prior reports of solid or micropapillary predominant tumors (median 229 days), validating its inclusion. VDT is significantly associated with disease-free survival after resection, and its incorporation into the nomogram addresses the limitation of static single-timepoint CT interpretation.
Single-center retrospective design. The primary limitation is the single-center retrospective design at a high-volume tertiary center, which may limit generalizability. The patient cohort was drawn from one institution over a 10-year period, introducing potential selection bias from evolving CT scanning standards, surgical techniques, and pathological classification practice across that timeframe.
Treatment stratification limitations. The ability to demonstrate nomogram-based risk stratification guiding treatment decisions was limited by substantial imbalances in treatment distribution -- lobectomy versus sublobar resection was approximately 9:1, and adjuvant chemotherapy versus no chemotherapy was approximately 1:5. These imbalances, combined with relatively small subgroup sizes, precluded definitive conclusions about treatment effects by risk group.
Need for external validation and multicenter studies. The nomogram was internally validated with bootstrap resampling but has not yet been externally validated in an independent patient cohort from other institutions. Multicenter validation studies across different geographic populations and healthcare settings are required before widespread clinical implementation.
Clinical pathway forward. The nomogram and online calculator provide a practical preoperative tool to identify early-stage lung adenocarcinoma patients at high risk for MPA, supporting individualized decisions about extent of surgical resection and adjuvant therapy. Larger multicenter studies with more balanced treatment arms and longer follow-up are needed to confirm the treatment strategy implications of nomogram-based risk stratification.