PD-L1 guides immunotherapy decisions. When a lung cancer patient is diagnosed, oncologists test for PD-L1 - a protein that helps cancer cells evade immune detection. The level of PD-L1 expression determines which treatments are used: patients with high expression (above 50%) may receive immune checkpoint inhibitors alone, while those with lower expression receive different combinations. Those with no detectable PD-L1 may need a different combination strategy entirely.
Testing requires invasive biopsy. Measuring PD-L1 requires laboratory analysis of tumor tissue obtained through biopsy - a procedure that carries risks and is not always feasible. Patients with tumors in inaccessible locations, those who are medically frail, or those whose biopsy samples are insufficient may not receive PD-L1 testing. This leaves them without a key piece of information for treatment planning.
CT scans already exist. Nearly all lung cancer patients have already undergone CT scanning for diagnosis and staging before treatment decisions are made. If CT images could reveal something about PD-L1 expression levels, clinicians could extract additional clinical value from imaging already obtained - without any additional procedures.
What this study investigated. This study examined 116 patients with early-stage (pT1) lung adenocarcinoma to determine whether specific CT measurement parameters - particularly how much of a tumor appears solid versus hazy - could predict their PD-L1 expression levels. The focus on early-stage disease is especially relevant because these patients are increasingly receiving neoadjuvant immunotherapy before surgery.
What ground-glass opacity means. On a CT scan of the lungs, tumors can appear with different densities. A ground-glass nodule (GGN) looks like a hazy cloud - the underlying lung structure is still visible through it. This appearance often indicates less invasive tumor cells that have spread in a layered pattern along existing lung architecture rather than destroying it. Pure GGNs are often low-grade, slow-growing tumors.
What solid appearance means. A solid nodule appears as a dense, opaque mass on CT - the underlying lung structure is completely obscured. Solid appearance generally indicates more invasive cancer cells with features like stromal invasion, vascular invasion, or fibroblastic proliferation. Solid tumors tend to be more aggressive and are more frequently associated with worse prognosis.
Part-solid nodules are a middle ground. Many lung adenocarcinomas are part-solid - having both a solid inner component and surrounding ground-glass haze. The ratio between the solid and total tumor components (measured by diameter or volume) has become an important parameter in lung cancer staging because larger solid fractions indicate greater invasiveness and correlate with outcomes.
AI-driven volumetric analysis. Modern CT workstations equipped with deep learning algorithms can automatically segment tumors in three dimensions, separately measuring the total tumor volume and the solid component volume. This enables calculation of the solid component volume rate - a more precise and reproducible measurement than manual diameter estimates - which was the key quantitative parameter evaluated in this study.
116 patients with early-stage lung adenocarcinoma. The study enrolled 116 patients (65 male, 51 female; mean age 71.6 years) with pathologically confirmed pT1 lung adenocarcinoma who underwent preoperative CT between January 2017 and December 2021 at Yamaguchi University Hospital. All patients had PD-L1 testing performed on their surgical specimens. The PD-L1 results were: 9.5% with high expression (TPS 50% or above), 38.8% with intermediate expression (TPS 1 to 49%), and 51.7% with no expression (TPS below 1%).
Multiple CT measurements obtained. Two radiologists independently measured each tumor, then averaged their results. Measurements included total tumor diameter, longest diameter of the solid component, and their ratio (rate of solid component diameter). An AI-powered CT workstation then automatically analyzed the same scans to produce volumetric measurements: total tumor volume, solid component volume, and their ratio (rate of solid component volume).
Strong interobserver agreement. Agreement between the two radiologists was substantial to almost perfect: kappa of 0.64 for tumor morphology classification, intraclass correlation of 0.81 for tumor diameter, and 0.74 for solid component diameter. This level of agreement indicates the CT measurements are reproducible enough to be clinically meaningful.
Statistical approach. The researchers compared CT parameters across PD-L1 expression groups, then built ROC curves to identify optimal cutoff values for predicting high PD-L1 expression. Multiple linear regression was used to identify which CT parameters independently predicted PD-L1 levels after accounting for all other factors including age and smoking history.
Solid nodules dominate high PD-L1 expression. Among patients with high PD-L1 expression (TPS 50% or above), 81.8% had solid nodules on CT. In contrast, among patients with no PD-L1 expression (TPS below 1%), only 10.0% had solid nodules - and 81.7% had part-solid nodules. This strikingly different distribution confirms a strong visual association between tumor appearance and immune checkpoint expression.
Solid fraction ratios are most discriminating. While absolute tumor size and volume did not differ significantly between PD-L1 groups, the proportion of the tumor that appears solid was highly discriminating. Median rate of solid component volume was 96.1% in the high PD-L1 group, 52.1% in the intermediate group, and just 28.2% in the no-expression group (p less than 0.001). This near-complete separation across three groups points to solid fraction as a reliable surrogate marker.
Rate of solid component volume: best single predictor. In ROC analysis, the rate of solid component volume achieved the highest accuracy for predicting high PD-L1 expression (TPS 50% or above), with an AUC of 0.876. At a cutoff of 60% solid component volume, it achieved 100% sensitivity and 72% specificity. For predicting any PD-L1 expression (TPS 1% or above), the rate of solid component volume achieved AUC of 0.735.
Independent predictor in multivariate analysis. Multiple linear regression, accounting for age, smoking history, tumor diameter, and all other CT parameters simultaneously, confirmed that rate of solid component volume was the only independent predictor of both high PD-L1 expression (p less than 0.001) and any PD-L1 expression (p=0.048). This independence confirms it is not merely a proxy for other known risk factors.
Solid component reflects invasiveness. On CT, solid areas within a lung adenocarcinoma correspond to histological features of invasiveness including stromal invasion, vascular invasion, collapsed alveolar spaces, and fibroblastic proliferation. More invasive tumors have higher rates of genomic instability and immune interactions - mechanisms that drive upregulation of PD-L1 as a defense against immune attack.
PD-L1 as an immune evasion response. Cancer cells upregulate PD-L1 when they are under attack by immune cells. This makes biological sense: tumors that have invaded surrounding tissue and lymphatics are more likely to be detected by the immune system, and consequently more likely to upregulate PD-L1 as a defense mechanism. Pure ground-glass tumors, which grow slowly along existing structures without destroying them, face less immune pressure and express less PD-L1.
AI volumetry outperforms manual measurement. While manual diameter-based measurements showed associations with PD-L1, the AI-derived volumetric solid fraction achieved higher AUC values and remained significant in multivariate analysis when the diameter-based rate did not. This advantage reflects the superiority of three-dimensional volume measurement over two-dimensional diameter estimation - volumes capture the true three-dimensional extent of the solid component rather than relying on a single measurement in one plane.
Application to neoadjuvant immunotherapy. The authors specifically highlight that predicting PD-L1 from CT is important for patients receiving neoadjuvant immunotherapy - immunotherapy given before surgery to shrink the tumor. In this setting, knowing which patients are likely to have high PD-L1 expression could inform treatment selection before biopsy results are available or when tissue is insufficient for testing.
Core finding. Quantitative CT analysis - specifically the proportion of a lung adenocarcinoma that appears solid in three-dimensional volumetric analysis - is a useful non-invasive predictor of PD-L1 expression. High rates of solid component volume are associated with higher PD-L1 expression in pT1 lung adenocarcinoma.
Practical implications. Because CT scanning is universally performed in lung cancer workup, this finding could allow preliminary PD-L1 risk stratification from imaging that already exists before any biopsy results return. Patients with predominantly solid tumors may be prioritized for urgent PD-L1 testing or considered as candidates for neoadjuvant immunotherapy.
Limitations to consider. The study was retrospective and single-center with only 116 patients, limiting statistical power - particularly for the small group with high PD-L1 expression (only 11 patients with TPS 50% or above). The study focused exclusively on early-stage (pT1) surgically resected tumors, while immunotherapy is more commonly used for unresectable advanced disease. Validation in larger, multicenter cohorts and in advanced-stage patients is needed before clinical deployment.
One of multiple predictive modalities. The authors acknowledge that CT-based PD-L1 prediction is complementary to, not a replacement for, other non-invasive approaches including PET/CT metabolic parameters and radiomics models. Integration of these modalities could potentially achieve better prediction than any single approach alone.