What is PD-L1 heterogeneity and why does it matter in squamous cell lung cancer?

PD-L1 is a protein expressed on tumor cell surfaces. Its expression varies significantly across different regions of the same tumor, a phenomenon called intratumoral heterogeneity. In squamous cell lung cancer, published research has documented this variation at small, medium, and large scales across tumor tissue. A biopsy from one region may return a different tumor proportion score than a biopsy from an adjacent region in the same patient, which creates uncertainty when TPS thresholds are used to guide treatment decisions.

Why does PD-L1 heterogeneity affect immunotherapy response prediction?

Checkpoint inhibitors targeting the PD-1/PD-L1 pathway are prescribed partly based on TPS thresholds, typically 1% and 50%. If a biopsy sample underestimates or overestimates overall tumor PD-L1 expression due to spatial heterogeneity, the TPS result may place a patient in a different treatment category than their actual tumor biology warrants. This interpretive uncertainty is greatest when TPS scores fall close to a clinical threshold.

Does PD-L1 expression in squamous NSCLC differ from adenocarcinoma?

The oncology community continues to study whether PD-L1 TPS has the same predictive value in squamous NSCLC as in adenocarcinoma. Research suggests the relationship between TPS score and clinical outcomes may differ across histological subtypes. Squamous tumors are frequently sampled via small bronchoscopic biopsies, which cover a limited geographic area and may be more susceptible to sampling bias than resection specimens used in other histologies.

Can AI tools solve the PD-L1 heterogeneity problem in lung cancer pathology?

AI-assisted pathology tools can reduce inter-observer scoring variability among pathologists, which addresses one source of uncertainty in TPS results. However, AI scoring systems assess the tissue that has already been sampled - they do not resolve the underlying biological heterogeneity across the whole tumor. Better scoring consistency and better spatial sampling are two distinct problems, and current AI tools address only the former.

What data should oncology clinics capture alongside each PD-L1 test result?

Useful attributes to record with each PD-L1 result include the biopsy date, the anatomical site sampled, the assay platform used, and the specific antibody clone. Different clones such as 22C3 and SP142 produce results that may not be directly comparable. When a clinic stores this context alongside the TPS number, serial results across a patient's treatment course become substantially more interpretable.

PD-L1 Heterogeneity in Squamous Lung Cancer | Rucja

The Core Problem: A Single Score for a Moving Target

Checkpoint inhibitor immunotherapy changed how doctors treat squamous cell lung cancer (SqCLC). PD-L1 expression, measured as tumor proportion score (TPS), is the main biomarker used to predict whether a patient will respond to anti-PD-1 and anti-PD-L1 treatments. Clinics use TPS thresholds (usually 1% and 50%) to decide on treatment.

The problem: PD-L1 expression varies across a tumor. A study in Modern Pathology looked at PD-L1 staining in tissue cores from 79 squamous cell lung cancers and 71 adenocarcinomas. Researchers found big differences in PD-L1 levels between different cores from the same tumor in both types. When a single threshold decides whether a patient gets treatment, this variation matters clinically.

Intra-Tumoral Heterogeneity: What the Data Show

A study published via NIH PubMed Central looked at how biopsy location affects PD-L1 prediction in NSCLC. Results showed that 78% of tumors had small-scale differences within the tumor, 50% had medium-scale differences, and 46% had large-scale differences. In 53% of cases, PD-L1 levels also differed between the main tumor and lymph node metastases.

These numbers matter. If more than half of patients have real PD-L1 differences between their main tumor and involved lymph nodes, a single biopsy may miss the full picture. The study concluded that taking multiple samples reduces, but does not eliminate, this problem.

Why Squamous Histology Complicates Prediction

Squamous cell lung cancer has a different immune profile than adenocarcinoma. Research at ASCO and in the Journal of Clinical Oncology shows that PD-L1 TPS may predict immunotherapy response differently in squamous versus non-squamous NSCLC. The link between TPS score and clinical benefit is not straightforward in squamous tumors.

This matters because SqCLC is usually found with small bronchoscopic biopsies or cytology rather than surgery. These samples cover only a small part of the tumor. Based on the heterogeneity data above, the TPS result may show one area while nearby areas have very different PD-L1 levels.

Small Biopsy, Large Variance

A Nature study on SqCLC addressed whether small biopsies work for PD-L1 testing. The point is simple: a biopsy result comes from one place at one moment. In a tumor with varying PD-L1, that result may overestimate or underestimate overall PD-L1.

This is not an argument against biopsy testing. Biopsy is still the standard, and no other method is proven for routine SqCLC testing. Rather, clinicians should treat biopsy TPS results as estimates with known limits - especially when a score is near a treatment decision point.

Spatial Heterogeneity and the Tumor Microenvironment

The tumor microenvironment (TME) partly controls PD-L1 expression. In SqCLC, immune cell patterns vary by tumor region. Areas with many T-cells may increase PD-L1 locally through interferon-gamma signaling, while nearby regions with fewer immune cells may express little PD-L1. This variation is not random; it reflects real biological differences across the tumor.

Studies of PD-L1 in different tumor regions in advanced NSCLC, published via PMC, suggest that where PD-L1 is located in the main tumor may matter for immune checkpoint treatment outcomes. This research is early, but it suggests that single-point TPS scoring may be combined with multi-region testing in the future.

AI-Assisted TPS Scoring: One Response to Observer Variability

To handle differences between pathologists in TPS scoring, researchers developed AI-assisted pathology tools. A study in JCO Precision Oncology tested an AI system trained on 393,565 labeled tumor cells from 802 whole-slide images to measure PD-L1 TPS in NSCLC. The goal was to cut variation between pathologists, which adds uncertainty on top of the biological differences already in the tumor.

AI scoring does not solve the spatial heterogeneity problem. It makes measurements more consistent within the tissue sample available. The two issues are different. Consistent scoring of an unrepresentative sample still gives a result with limited predictive value.

The Clinic Data Dimension

For oncology clinicians and clinic operations teams, PD-L1 heterogeneity is a data management issue. Patients with SqCLC often get repeat biopsies during treatment - at baseline, at progression, or when switching treatments. Each biopsy may return a different PD-L1 result. Without a record of how TPS changed over time, a clinician cannot properly interpret a new result.

Systems that gather biomarker results over time - pulling PDF pathology reports, extraction outputs, and lab data into one patient record - let oncologists see trends instead of single numbers. Our overview of EGFR, ALK, and PD-L1 testing in NSCLC clinic workflows explains how to manage multiple biomarker tests at once, including how quickly results come back.

Getting PD-L1 data from PDF pathology reports - which most labs still use - is hard. Our piece on AI lab extraction in clinical work explains this. A TPS result on page three of a PDF is less useful than one that automatically appears in a clinician's patient timeline during a review appointment.

What Clinic Teams Should Track

There is no standard protocol for repeat PD-L1 testing in SqCLC. The current approach uses baseline biopsy TPS for first-line treatment decisions. In real clinics, data quality questions come up:

When a patient progresses on a checkpoint inhibitor, is the repeat PD-L1 result stored with the original baseline in the same patient record, with dates and location noted?
When a biopsy comes from a different site - like a lymph node instead of the main tumor - is that noted with the TPS result?
Is the specific PD-L1 antibody used recorded? Different assay types (22C3, 28-8, SP142, SP263) are not the same, and different assays add another source of variation.

These are data-quality questions as much as clinical ones. A clinic that keeps PD-L1 results as static numbers in a PDF archive cannot answer them. A clinic that organizes biomarker data over time - with location, assay, and date for each result - can find patterns that would otherwise stay hidden in a patient's treatment history.

The Operational Takeaway

Research into PD-L1 heterogeneity in SqCLC is ongoing. Researchers are testing multi-region biopsies, liquid biopsies, and computational pathology tools to capture a fuller picture of tumor PD-L1. None are standard treatment options today.

What exists today is better data systems. Making sure every PD-L1 result is captured, dated, location-labeled, and visible to the clinician at decision time is something clinic leaders can do now, regardless of how the science evolves.

See how this works with a 30-minute demo. We show how Rucja surfaces biomarker trends - including serial PD-L1 results - on real hospital data. Contact us for a demo.

See Rucja on your own hospital data.

Demos take 30 minutes. We'll walk you through the bits of this article in your live workflow.

Book a demo