For more than a decade, immune checkpoint inhibitors have transformed cancer treatment, producing durable responses across multiple malignancies. Despite these advances, one of the most important challenges in modern oncology remains unanswered: which patients will truly benefit from immunotherapy?
Two patients with the same cancer type, stage, and treatment may experience dramatically different outcomes. One may achieve a durable response lasting years, while the other experiences primary resistance despite receiving the same immune checkpoint inhibitor. Understanding this difference has become one of the central goals of precision immuno-oncology.
PD-L1 expression, tumor mutational burden (TMB), and microsatellite instability-high/deficient mismatch repair (MSI-H/dMMR) remain the only clinically validated predictive biomarkers used routinely across multiple tumor types. Each is supported by a strong biological rationale, validated companion diagnostics, and landmark clinical trials. However, none consistently distinguishes responders from non-responders, highlighting the complexity of tumor–immune interactions that extend far beyond a single molecular marker.
The challenge is becoming increasingly important as immune checkpoint inhibitors continue to expand into earlier disease stages, perioperative treatment, and combination strategies. Selecting the right patient is no longer simply a scientific question—it has become a critical component of treatment quality, helping maximize clinical benefit while avoiding unnecessary toxicity and healthcare costs.
How can clinicians improve patient selection when currently approved biomarkers provide only part of the picture?
This question will be addressed at the OncoDaily Community Oncology Global Congress 2026, where international experts and frontline oncology professionals will discuss the current limitations of established biomarkers, emerging predictive technologies, and practical strategies for integrating next-generation biomarkers into routine oncology practice. The virtual congress will take place from August 28 to 30, 2026.
Looking Beyond Tumor Cells
If current biomarkers provide only part of the picture, where should researchers look next?
Increasingly, the answer lies within the tumor microenvironment (TME).
Rather than focusing exclusively on malignant cells, the tumor microenvironment represents a dynamic ecosystem composed of immune cells, stromal cells, blood vessels, extracellular matrix, and numerous soluble mediators. Together, these components determine whether an effective antitumor immune response develops—or whether the tumor successfully evades immune surveillance.
Growing evidence suggests that the tumor microenvironment may predict response to immune checkpoint inhibitors more accurately than tumor-intrinsic biomarkers alone. Tumors enriched with activated CD8⁺ cytotoxic T cells, dendritic cells, and interferon-γ signaling generally demonstrate greater sensitivity to checkpoint blockade. In contrast, tumors dominated by regulatory T cells, myeloid-derived suppressor cells, tumor-associated macrophages, and cancer-associated fibroblasts often create an immunosuppressive environment that limits T-cell activation and promotes immune escape.
Equally important is the spatial organization of immune cells within the tumor. The presence of lymphocytes deep within the tumor parenchyma or concentrated at the invasive margin often reflects a pre-existing antitumor immune response that can be reactivated by immune checkpoint inhibition. Conversely, tumors demonstrating immune exclusion—where lymphocytes remain confined to the surrounding stroma—are frequently associated with primary resistance despite favorable conventional biomarkers.
Another important consideration is that the tumor microenvironment is highly dynamic. Immune-cell composition evolves during disease progression and under therapeutic pressure, meaning that a single pretreatment biopsy may not fully represent the biological state that ultimately determines treatment response. Rather than functioning as a single biomarker, the TME is increasingly viewed as a comprehensive framework integrating multiple biological signals that together provide a more accurate picture of immunotherapy sensitivity.
Advances in multiplex immunohistochemistry, digital pathology, spatial transcriptomics, and single-cell technologies are allowing investigators to characterize these complex cellular interactions with unprecedented resolution. As these technologies continue to mature, the tumor microenvironment is becoming one of the most promising sources of next-generation predictive biomarkers.
Why Are Tertiary Lymphoid Structures Receiving So Much Attention?
Among all emerging biomarkers, tertiary lymphoid structures (TLSs) have generated some of the strongest evidence for predicting response to immune checkpoint inhibitors.
Unlike scattered immune-cell infiltration, TLSs are highly organized lymphoid aggregates that resemble secondary lymphoid organs. They develop within or adjacent to tumors in response to chronic immune stimulation and function as local sites of antigen presentation, T-cell activation, B-cell maturation, and antibody production. Rather than representing passive collections of immune cells, TLSs actively support the generation and maintenance of antitumor immunity within the tumor microenvironment.
Across multiple tumor types, mature TLSs have consistently been associated with higher response rates, longer progression-free survival, and improved overall survival following immune checkpoint blockade. Their predictive value frequently extends beyond PD-L1 expression, with favorable clinical outcomes observed even in patients whose tumors demonstrate low PD-L1 expression but contain abundant mature TLSs. These findings suggest that TLSs may better reflect the presence of a functional pre-existing immune response that can be reactivated by checkpoint inhibition.
Despite these encouraging findings, important challenges remain before TLS assessment can become part of routine clinical practice. There is currently no universally accepted method for defining TLS maturity, density, or anatomical location. Assessment varies considerably between conventional histopathology, multiplex immunohistochemistry, and advanced spatial imaging technologies, limiting reproducibility between institutions.
Standardized scoring systems, prospective validation studies, and harmonized pathological criteria will therefore be essential before TLS evaluation can be routinely incorporated into clinical decision-making. Rather than replacing established biomarkers, TLSs are increasingly viewed as one component of a broader immune landscape that, when combined with complementary biomarkers, may substantially improve prediction of immunotherapy benefit.
Can Blood-Based Biomarkers Transform Patient Selection?
Although tissue remains the cornerstone of biomarker testing, it provides only a single snapshot of an evolving disease. As tumors respond to treatment or develop resistance, their biological characteristics—and the host immune response—continue to change. This has fueled growing interest in blood-based biomarkers, which offer the opportunity to monitor these changes in real time.
Unlike tissue biopsies, liquid biomarkers can be obtained repeatedly through minimally invasive blood sampling, allowing clinicians to follow both tumor evolution and systemic immune responses throughout treatment. Rather than replacing tissue-based biomarkers, they provide complementary information that may help refine treatment decisions over the course of therapy.
Among circulating biomarkers, circulating tumor DNA (ctDNA) has generated the greatest clinical interest. Several studies have shown that early clearance of ctDNA during immune checkpoint inhibition is associated with improved clinical outcomes, whereas persistent or increasing ctDNA levels may identify patients with primary resistance before radiographic progression becomes apparent. This raises the possibility that ctDNA could become an important tool not only for predicting response but also for monitoring treatment efficacy and detecting minimal residual disease.
Researchers are also investigating numerous additional blood-based biomarkers, including peripheral immune-cell populations, circulating cytokines, soluble immune checkpoint molecules, and inflammatory markers. Each reflects a different aspect of systemic immune activation, providing insights that cannot be obtained from tumor tissue alone.
However, several challenges continue to limit routine clinical implementation. Analytical methods vary considerably between laboratories, clinically relevant thresholds have not been standardized, and prospective validation remains limited for most circulating biomarkers. As a result, blood-based assays are currently viewed as complementary rather than replacement biomarkers, with their greatest value likely to emerge when integrated with tissue pathology, imaging, and molecular profiling into comprehensive predictive models.
How Will Artificial Intelligence Change Biomarker Development?
As the number of potential biomarkers continues to grow, another challenge becomes increasingly apparent: how can clinicians integrate such complex biological information into routine treatment decisions?
Artificial intelligence (AI) is emerging as one of the most promising solutions.
Rather than identifying a single predictor of response, AI has the potential to integrate enormous volumes of clinical, pathological, radiological, genomic, transcriptomic, and immunological data into predictive models capable of capturing the complexity of tumor–immune interactions. This reflects an important shift in precision oncology—from relying on individual biomarkers toward understanding how multiple biological factors work together to influence treatment response.
One of the earliest clinical applications of AI has been digital pathology. Machine learning algorithms can quantify immune-cell infiltration, evaluate the spatial distribution of lymphocytes, identify tertiary lymphoid structures, and detect subtle histopathological features that may be difficult to recognize using conventional microscopy. Similar advances in radiomics allow routine CT, MRI, and PET imaging to reveal imaging signatures associated with immune activation and clinical response, creating opportunities for non-invasive biomarker assessment.
Perhaps the greatest promise of artificial intelligence lies in its ability to integrate multiple biomarker platforms into a single predictive framework. Early studies suggest that combining PD-L1 expression, tumor mutational burden, genomic alterations, tumor microenvironment characteristics, circulating biomarkers, microbiome composition, and clinical variables consistently outperforms models based on any individual biomarker alone. Rather than replacing established biomarkers, AI provides a powerful tool for interpreting them together and generating a more comprehensive estimate of immunotherapy benefit.
Despite rapid progress, important challenges remain before AI-guided biomarker selection can become part of routine clinical practice. Most predictive algorithms have been developed using retrospective datasets, external validation remains limited, and many models continue to function as “black boxes” with limited biological interpretability. Standardization of data collection, prospective validation, regulatory oversight, and transparent model development will all be essential before AI can be fully integrated into everyday oncology practice.
For now, artificial intelligence should be viewed not as a replacement for clinical judgment, but as a tool capable of bringing together multiple complementary biological signals. As precision immunotherapy continues to evolve, AI may become the key to transforming complex biomarker data into practical clinical decision-making.
COMPASS: Can AI Predict Immunotherapy Response Across Multiple Cancers?
Which Questions Remain Unanswered?
Despite rapid progress, several important questions continue to limit the clinical implementation of next-generation immunotherapy biomarkers.
Which Biomarker or Combination of Biomarkers Best Predicts Response?
PD-L1 expression, tumor mutational burden, tumor-infiltrating lymphocytes, tertiary lymphoid structures, circulating tumor DNA, gene-expression signatures, and microbiome profiles each provide valuable but incomplete information.
No individual biomarker consistently identifies all patients who will respond to immune checkpoint inhibition or excludes all patients who will not benefit. The central challenge is therefore no longer simply discovering additional biomarkers, but determining which combinations provide clinically meaningful improvements over existing approaches.
Integrated models may offer the most promising path forward. By combining tumor genomics, immune-cell composition, spatial organization, circulating biomarkers, imaging, and clinical characteristics, researchers hope to generate a more complete picture of both tumor biology and host immunity. However, whether these composite models will outperform individual assays in prospective clinical practice remains under investigation.
How Should Biomarker Testing Be Standardized?
Technical variability remains one of the greatest barriers to widespread implementation.
Different PD-L1 antibody clones, scoring systems, sequencing platforms, spatial imaging technologies, and computational methods can produce different results from similar biological samples. Thresholds used to define biomarker positivity may also vary between tumor types, laboratories, and clinical trials.
Without harmonized methodologies, validated cutoffs, and standardized reporting systems, even biologically promising biomarkers may prove difficult to reproduce across institutions. Standardization will therefore be essential before emerging biomarkers can reliably guide treatment decisions outside specialized research centers.
Should Biomarkers Be Reassessed During Treatment?
Most current treatment decisions are based on a single biomarker assessment performed before therapy begins. However, tumor biology is not static.
The tumor microenvironment evolves during treatment, immune-cell populations change under therapeutic pressure, and resistant tumor clones may emerge over time. Circulating biomarkers such as ctDNA may also change weeks or months before progression becomes visible on conventional imaging.
These observations raise an important question: should biomarker assessment become a dynamic process rather than a one-time test?
Repeated tissue biopsy is not always feasible, but liquid biopsy, functional imaging, and computational analysis may allow clinicians to monitor biological changes during therapy. Whether this information should be used to escalate treatment, discontinue ineffective therapy, or introduce new combinations remains an active area of investigation.
Can Artificial Intelligence Improve Clinical Decision-Making?
Early studies suggest that AI-based models integrating pathology, radiology, molecular profiling, and clinical data may predict immunotherapy response more accurately than individual biomarkers alone.
However, accuracy is not the only requirement for clinical adoption. Predictive models must also be externally validated, transparent, reproducible, and applicable across diverse patient populations and healthcare systems.
An algorithm developed using data from a single academic center may not perform equally well in community practices, lower-resource settings, or populations underrepresented in the original dataset. Prospective validation and clear regulatory standards will therefore be essential before AI-guided biomarker selection becomes part of routine care.
Why This Matters for Community Oncology
The development of increasingly complex biomarkers creates both opportunities and challenges for community oncology.
Major academic centers may have access to broad genomic sequencing, spatial transcriptomics, digital pathology, liquid biopsy, molecular tumor boards, and sophisticated computational platforms. Many community practices, however, operate with limited tissue, restricted reimbursement, delayed testing, and limited access to specialized interpretation.
The future of precision immunotherapy must therefore address not only which biomarkers are scientifically informative, but also which can be implemented reliably, affordably, and equitably in routine practice.
A biomarker that cannot be accessed, interpreted, or linked to a practical treatment decision has limited clinical value. Similarly, increasingly complex predictive models must not widen existing disparities between patients treated at major academic institutions and those receiving care in community or resource-limited settings.
Practical implementation will require standardized assays, clear reporting systems, regional testing networks, multidisciplinary collaboration, clinician education, and evidence that biomarker-guided treatment decisions improve meaningful patient outcomes.
Join the Discussion at the Community Oncology Global Congress
Predicting response to immune checkpoint inhibitors will be among the important precision-oncology challenges discussed at the OncoDaily Community Oncology Global Congress 2026.
The discussion will move beyond the limitations of PD-L1, TMB, and MSI-H/dMMR to examine how emerging biomarkers—including the tumor microenvironment, tertiary lymphoid structures, circulating tumor DNA, spatial biology, and artificial intelligence—may improve patient selection for immunotherapy.
International experts and frontline oncology professionals will explore how these technologies can be validated, standardized, and integrated into routine clinical practice, including outside major academic centers.
The virtual congress will take place from August 28 to 30, 2026.
The Bottom Line
Predicting response to immune checkpoint inhibitors remains one of the central unresolved challenges in modern oncology.
PD-L1, TMB, and MSI-H/dMMR have transformed treatment selection, but each reflects only one component of a much more complex tumor–immune interaction. Patients may respond despite negative conventional biomarkers, while biomarker-positive tumors may demonstrate primary resistance.
The next generation of predictive tools is therefore moving beyond isolated molecular tests.
The tumor microenvironment provides information about immune-cell composition and spatial organization. Tertiary lymphoid structures may identify tumors with pre-existing functional immunity. Circulating tumor DNA offers the possibility of monitoring response dynamically. Artificial intelligence may help integrate these signals with pathology, imaging, genomic data, and clinical characteristics.
None of these approaches is likely to serve as a universal biomarker alone.
The future of precision immunotherapy will probably depend on integrated predictive models that combine multiple complementary biological signals. The central goal is not simply to identify more biomarkers, but to determine which information meaningfully improves treatment selection, can be standardized across institutions, and is practical enough to reach patients in everyday oncology care.