What Is PD-L1 IHC and How Does It Affect Cancer Treatment?

PD-L1 IHC is a lab test performed on tumor tissue that measures how much of a protein called PD-L1 sits on the surface of cancer cells and nearby immune cells. The result helps oncologists decide whether a patient is likely to benefit from immunotherapy drugs known as checkpoint inhibitors. In cancers like lung, breast, gastric, and bladder cancer, the PD-L1 score can determine which drugs are offered, in what order, and sometimes whether immunotherapy is offered at all. The test is straightforward in concept but surprisingly complex in practice, with the choice of testing kit, the scoring method used, and even how the tissue sample was handled all capable of shifting a patient’s result.

How PD-L1 Helps Tumors Hide from the Immune System

Your immune system has built-in brakes. One of the most studied is a receptor on T cells called PD-1. When PD-1 locks onto its partner protein, PD-L1, it sends a “stand down” signal that keeps the T cell from attacking. Under normal conditions this prevents the immune system from damaging healthy tissue. Tumors exploit this safety mechanism by coating themselves with PD-L1, effectively hanging a “do not attack” sign that shuts down the immune cells trying to destroy them.¹ The PD-1/PD-L1 pathway is one of the most studied immune checkpoints in oncology, and its role in allowing tumors to evade immune surveillance has been confirmed across many cancer types.²

Checkpoint inhibitor drugs like pembrolizumab, nivolumab, and atezolizumab work by blocking either PD-1 or PD-L1, preventing that handshake from happening. With the brake released, T cells can recognize and attack the tumor again. The question for any individual patient is whether their tumor relies heavily on PD-L1 to survive. That is what the IHC test tries to answer.

What Happens in the Lab

Immunohistochemistry, or IHC, is a long-established technique in pathology. A thin slice of tumor tissue is placed on a glass slide and exposed to specially designed antibodies that latch onto a target protein. A chemical staining step then makes the antibody visible under a microscope, so the pathologist can see where and how much of that protein is present.³ IHC is widely used for diagnosing cancer and identifying specific tumor markers, and the technique involves both slide preparation and the pathologist’s interpretation of the staining pattern.⁴

For PD-L1 testing specifically, the antibody used is designed to bind to the PD-L1 protein on cell surfaces. The pathologist then estimates what percentage of the relevant cells stain positive. This sounds simple, but in practice the percentage that counts and the cells that count depend on which scoring system is being used and which cancer is being evaluated.

Scoring Systems and What the Numbers Mean

Three main scoring systems are used in PD-L1 IHC, and understanding which one applies matters because the same piece of tissue can produce different scores depending on the method.

• TPS (Tumor Proportion Score): the percentage of viable tumor cells showing partial or complete membrane staining. This is the most common score used in lung cancer. Key thresholds are typically 1% and 50%.
• CPS (Combined Positive Score): the number of PD-L1-staining cells of any type (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100. CPS is used in gastric, esophageal, cervical, and triple-negative breast cancer, among others. Common thresholds are CPS ≥1 and CPS ≥10.
• IC (Immune Cell score): the proportion of tumor area occupied by PD-L1-positive immune cells. This method is used primarily with the atezolizumab-associated SP142 assay and has been applied in breast and bladder cancer.

These are not interchangeable labels for the same measurement. A tumor could have a TPS below 1% yet a CPS above 10 if immune cells in and around the tumor stain strongly. A study comparing pathologist agreement across these systems found that while TPS scoring showed excellent agreement between two pathologists, CPS agreement was considerably weaker, with only fair correlation at both the 1% and 10% thresholds.⁵ The reason CPS is harder to score consistently is that it forces the pathologist to identify and count immune cells alongside tumor cells, a task with more room for judgment calls.

Different Assays, Different Answers

One of the most frustrating aspects of PD-L1 testing is that several different commercial test kits exist, each developed alongside a specific immunotherapy drug. The 22C3 assay was co-developed with pembrolizumab, the 28-8 assay with nivolumab, the SP142 assay with atezolizumab, and the SP263 assay with durvalumab. Each uses a different antibody clone, runs on a specific staining platform, and was validated for specific cancer types.

A major international effort called the Blueprint project compared these assays head to head. In phase 1, three of the four assays (22C3, 28-8, and SP263) produced comparable tumor cell staining, but the SP142 assay consistently stained fewer tumor cells.⁶ In practical terms, swapping one assay for another could change a patient’s PD-L1 classification in more than a third of cases. A larger follow-up study across multiple tumor types confirmed the pattern: when all three assays were run on the same tissue, only about 60% of specimens received the same PD-L1 status from all three, and among the cases that were positive on at least one assay, the SP142 assay flagged less than a third of them compared to over 94% detected by 22C3.⁷

This is not just an academic concern. In triple-negative breast cancer, the SP142 assay classified roughly 46% of patients as PD-L1-positive at the IC ≥1% threshold, while the SP263 and 22C3 assays classified about 75% and 73% as positive using the same threshold.⁸ That gap means a patient whose tissue is tested with one assay may be deemed eligible for immunotherapy while the same tissue tested with another might not.

How PD-L1 Scores Guide Treatment in Specific Cancers

The clinical weight of a PD-L1 score varies dramatically by cancer type. In non-small cell lung cancer (NSCLC), PD-L1 testing has been most firmly embedded into treatment algorithms. For patients without a targetable gene mutation, a TPS of 50% or higher can qualify them for single-agent pembrolizumab as their first treatment. Roughly 30% of NSCLC patients fall into this high-expression group. Patients with PD-L1 below 50% are generally started on platinum-based chemotherapy, sometimes in combination with immunotherapy.⁹ The 22C3 companion diagnostic assay was validated alongside pembrolizumab’s clinical trials, where patients with TPS ≥50% had an overall response rate of about 41%, double the rate seen in unselected patients.¹⁰

In advanced triple-negative breast cancer, the scoring shifts to CPS, with thresholds of CPS ≥1 and CPS ≥10. A large trial of pembrolizumab plus chemotherapy used both thresholds as primary endpoints for evaluating overall survival, with the CPS ≥10 group showing the clearest benefit.¹¹

In gastric and gastroesophageal cancers, an estimated 55% to 66% of advanced cases express PD-L1, and clinical trials have consistently shown that PD-L1-positive patients are more likely to respond to checkpoint inhibitors than negative ones. PD-L1 and microsatellite instability status have become the two most widely adopted biomarkers for selecting candidates for anti-PD-1 therapy in these cancers.¹²

The distinction between a “companion diagnostic” and a “complementary diagnostic” matters here. A companion diagnostic is required before prescribing a drug; the treatment is restricted to patients who test positive. A complementary diagnostic provides useful information that can improve the benefit-to-risk ratio but does not prevent access to the drug.¹³ PD-L1 IHC serves as a companion diagnostic for some drug-cancer combinations (pembrolizumab monotherapy in NSCLC, for instance) and as a complementary diagnostic for others.

When PD-L1 Is Low or Negative

A low or negative PD-L1 result does not automatically mean immunotherapy is off the table. In NSCLC, pooled data from randomized trials show that combining a checkpoint inhibitor with chemotherapy improves both overall survival and progression-free survival compared to chemotherapy alone, even in patients whose tumors express less than 1% PD-L1.¹⁴ A separate meta-analysis of first-line trials confirmed that patients benefited from the combined approach regardless of PD-L1 expression level.¹⁵

The implication for patients is worth stressing: PD-L1 is a useful enrichment biomarker, meaning higher levels generally predict a stronger response, but it is not an on-off switch. Some PD-L1-negative patients respond well, and some PD-L1-high patients do not. The test helps tilt the odds and guide sequencing decisions rather than definitively predict outcome.

The Tumor Heterogeneity Problem

PD-L1 expression is not uniform across a tumor. A sample taken from one region may look very different from a sample taken a few centimeters away. In NSCLC, the majority of tumors show intra-tumoral heterogeneity: one study found it at the small scale in 78% of cases, and at the large scale in 46%. When the primary tumor was compared to lymph node metastases, expression differed in over half the cases, sometimes enough to flip the patient across a clinically meaningful threshold.¹⁶

A separate analysis of lung adenocarcinoma and squamous cell carcinoma found substantial inconsistencies in PD-L1 staining percentages across different cores from the same tumor. Geographic variation in PD-L1 expression within a single tumor explained the discordant results, and the sensitivity of a single small biopsy for detecting PD-L1 expression at standard clinical thresholds could drop as low as 85%.¹⁷

A computer modeling study quantified the risk more precisely. When a tumor’s true PD-L1 positivity hovered near a cutoff value (between 40% and 60% positive cells), the error rate for a single biopsy reached about 33%. By contrast, tumors with very low (under 20%) or very high (over 80%) expression had error rates below 1%. Smaller biopsies correlated with higher error rates across the board.¹⁸ For patients whose true PD-L1 level sits near a threshold, the single-biopsy result they receive may be a coin flip.

Pathologist Disagreement and Pre-Analytical Pitfalls

Even when tissue and assay are held constant, pathologists can disagree about the score. In gastric and gastroesophageal cancers evaluated with CPS, a study of multiple pathologists found only fair agreement both before and after specialized training. The underlying reasons included difficulty distinguishing positively staining stromal cells from tumor cells, faint or variable staining intensity, and tissue artifacts from surgical handling like cautery and crush damage.¹⁹ TPS scoring in lung cancer tends to be more reproducible because counting only tumor cell membranes is a simpler visual task than distinguishing every PD-L1-positive cell type, as CPS requires.

Before a pathologist even looks through the microscope, tissue handling can alter the result. Leaving a specimen in formalin fixative for too long degrades PD-L1 staining quality. Fixation beyond 72 hours significantly increased the proportion of suboptimally stained samples, and at 96 hours the SP142 antibody was more affected than SP263.²⁰ Cold ischemia time, meaning the delay between tissue removal from the body and placement in fixative, poses an even sharper risk. False-negative results hit 25% in samples exposed to six hours of cold ischemia and climbed to 34% at 12 or 24 hours, again with SP142 showing less tolerance than SP263.²¹ These findings underscore that a patient’s PD-L1 result can be influenced by something as mundane as how quickly the operating room staff gets a specimen into the jar.

Turnaround Time and Practical Logistics

From a patient’s perspective, waiting for biomarker results can feel agonizing when treatment decisions hang in the balance. PD-L1 IHC is one of the faster tests in the biomarker toolkit. Results typically come back within two to five days.²² Real-world data from labs running both 28-8 and 22C3 assays confirmed turnaround times between two and four days with low test failure rates.²³ The catch is that PD-L1 is often just one of several biomarkers ordered simultaneously. Molecular profiling panels that look for targetable gene mutations can take weeks, and treatment decisions in NSCLC increasingly depend on the full panel, not PD-L1 alone.

Why Resistance Happens Despite a High Score

A high PD-L1 score is not a guarantee that immunotherapy will work, because tumors have many ways to evade the immune system beyond the PD-1/PD-L1 axis. Resistance to checkpoint inhibitors can be primary (the drug never works) or acquired (the tumor initially responds and then escapes). The mechanisms are varied and often overlap within a single patient.²⁴

On the tumor’s side, genetic and epigenetic changes can shut down antigen presentation, essentially removing the molecular flags that T cells use to identify cancer cells. Loss of certain proteins involved in displaying antigens on the cell surface impairs the ability of cytotoxic T cells to recognize the tumor even after the PD-L1 brake has been released. Alterations in key signaling pathways and disruptions to interferon signaling also contribute to resistance.²⁵ In other words, PD-L1 measures only one dimension of a tumor’s immune relationship. A tumor can be plastered with PD-L1 and still have escape routes that checkpoint inhibitors cannot block.

Beyond Tissue Biopsies and Toward Liquid Biomarkers

Researchers are exploring whether PD-L1 status can be assessed without a traditional tissue biopsy. Liquid biopsy approaches look for circulating tumor cells (CTCs) in the blood and measure PD-L1 on their surface. In patients receiving checkpoint inhibitors, the presence of PD-L1-positive CTCs at early or late time points after starting treatment was associated with shorter progression-free survival, suggesting that tracking PD-L1 in the blood over time could offer real-time monitoring that a single tissue biopsy cannot.²⁶

A dual-biomarker approach combining PD-L1 on CTCs with PD-1 on circulating immune cells has shown promise in advanced NSCLC, with initial associations between this combined measurement and both treatment response and survival. That work remains retrospective and needs prospective validation, but it illustrates the direction the field is moving: away from a single snapshot of one protein on a tiny tissue sample and toward dynamic, multi-marker profiling.²⁷

Newer Approaches That May Outperform Standard PD-L1 IHC

Standard PD-L1 IHC evaluates a single protein in a relatively crude way. Multiplex immunohistochemistry and immunofluorescence (mIHC/IF) stain for multiple proteins simultaneously on the same tissue slide, revealing not just whether PD-L1 is present but which cell types express it and how those cells are spatially organized relative to the tumor. A systematic review and meta-analysis comparing different biomarker approaches found that mIHC/IF was a significantly better predictor of checkpoint inhibitor response than standard PD-L1 IHC, tumor mutational burden (TMB), or gene expression profiling.²⁸ A separate study in melanoma reported that comprehensive immune profiling had higher sensitivity for predicting response than either PD-L1 IHC or TMB alone.²⁹

These advanced approaches are not yet standard of care in most settings. They require specialized equipment, trained personnel, and validated scoring criteria that do not yet exist at the scale needed for routine clinical use. But they point to a future where the binary “PD-L1 positive or negative” question is replaced by a richer picture of the tumor’s immune landscape.

Artificial Intelligence in PD-L1 Scoring

Given the variability between pathologists, automated scoring using artificial intelligence is an active area of development. In diffuse large B-cell lymphoma, an AI algorithm provided quantitative PD-L1 scores that aligned closely with the average pathologist score, achieving an intraclass correlation of 0.96 with manual scoring, which was actually higher than the 0.94 correlation among pathologists themselves.³⁰ Across NSCLC and other solid tumors, AI-based assessment has shown high correlation with median pathologist scores and variability that falls within the normal range of pathologist disagreement.³¹

Perhaps most practically, when untrained pathologists were given AI-assisted pre-reads, they achieved higher consistency and spent less time than trained pathologists scoring without AI assistance.³² This suggests a near-term role not as a replacement for pathologists but as a calibration tool, especially for the CPS scoring method where human agreement has been weakest. It could also level the playing field between major academic centers with PD-L1 scoring expertise and community hospitals where pathologists see fewer cases.