Trend AnalysisMedicine & Health

Predicting Immunotherapy Response: Beyond PD-L1 and TMB

Immune checkpoint inhibitors (ICIs) have transformed oncology — anti-PD-1/PD-L1 and anti-CTLA-4 antibodies produce durable responses in a subset of patients across many cancer types. But only 20–40% o...

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

The Question

Immune checkpoint inhibitors (ICIs) have transformed oncology — anti-PD-1/PD-L1 and anti-CTLA-4 antibodies produce durable responses in a subset of patients across many cancer types. But only 20–40% of patients benefit, and current biomarkers (PD-L1 expression, tumour mutational burden) are imperfect predictors. PD-L1 expression is heterogeneous within tumours and between biopsies; TMB thresholds vary across cancer types. Can machine learning models trained on routine clinical data — blood counts, liver function tests, LDH — predict ICI response better than current molecular biomarkers?

Landscape

Yoo et al. (2025) in Nature Medicine, developed SCORPIO — a machine learning system that predicts ICI efficacy using only routine blood tests and clinical data, without requiring tumour tissue or advanced genomic assays. Trained on 1,628 patients at MSKCC and validated across 10 phase 3 trials (4,447 patients) and an external cohort (1,159 patients), SCORPIO outperformed both PD-L1 and TMB as a standalone predictor. This finding has profound access implications: routine blood tests are available everywhere, while genomic testing requires specialised laboratories.

X. Wang et al. (2024) in Annals of Oncology, reviewed TMB as a predictive biomarker, documenting both its utility and limitations. TMB predicts ICI response across cancer types, but the optimal cut-off varies (10 mut/Mb for pembrolizumab's pan-cancer approval, but different thresholds may be appropriate for different tumour types). Technical variability between sequencing panels further complicates clinical implementation.

Di Federico et al. (2024) revealed a critical clinical problem: intrapatient variation in biomarker measurements. PD-L1 showed only moderate concordance between paired samples (rho=0.53), with major increases or decreases in approximately 9–10% of cases. TMB concordance was higher (rho=0.80) but PD-L1 variation alone can alter treatment decisions. This spatial heterogeneity fundamentally limits single-biopsy biomarker approaches, particularly for PD-L1.

Yin et al. (2024) reviewed emerging biomarkers beyond PD-L1 and TMB, including the tumour immune microenvironment (TIME), microsatellite instability (MSI), HLA diversity, and ctDNA-based approaches. Their analysis concluded that multivariable predictive models integrating multiple features will likely outperform any single marker.

Key Claims & Evidence

<
ClaimEvidenceVerdict
Routine blood tests predict ICI response as well as or better than PD-L1/TMBSCORPIO ML model outperforms PD-L1 and TMB across cancer types (Yoo et al. 2025)Supported; paradigm-shifting if validated prospectively
TMB's optimal threshold varies by cancer typeDifferent tumour types have different TMB-response correlations (X. Wang et al. 2024)Confirmed; universal cut-off is inappropriate
Intrapatient PD-L1 heterogeneity limits single-biopsy approachesPD-L1 concordance moderate (rho=0.53); TMB more consistent (rho=0.80) (Di Federico et al. 2024)Confirmed; particularly problematic for PD-L1-based treatment decisions
Multi-parameter models outperform single biomarkersIntegration of TIME, genomic, and clinical features improves prediction (Yin et al. 2024)Supported by multiple studies; clinical implementation challenging

Open Questions

  • Prospective validation of SCORPIO: Can routine-blood-test-based prediction replicate in prospective, interventional trials where treatment decisions are guided by the model?
  • Dynamic biomarkers: Biomarker status changes during treatment. Can serial monitoring (ctDNA dynamics, circulating immune cells) improve prediction during therapy, not just at baseline?
  • Combination biomarkers: What is the optimal combination of molecular (TMB, MSI), tissue (PD-L1, immune infiltrate), and blood-based (SCORPIO, NLR) biomarkers for each cancer type?
  • Cost-effectiveness: At what cost threshold does advanced biomarker testing become cost-effective compared to empirical ICI therapy with early response assessment?
  • Referenced Papers

    • [1] Yoo, S.-K. et al. (2025). SCORPIO: predicting ICI efficacy from routine blood tests. Nature Medicine. DOI: 10.1038/s41591-024-03398-5
    • [2] Wang, X. et al. (2024). TMB for prediction of PD-(L)1 blockade efficacy. Annals of Oncology. DOI: 10.1016/j.annonc.2024.03.007
    • [3] Di Federico, A. et al. (2024). Intrapatient variation in PD-L1 and TMB in NSCLC. Annals of Oncology. DOI: 10.1016/j.annonc.2024.06.014
    • [4] Yin, X. et al. (2024). Predictive biomarkers in antitumor immunotherapy. Frontiers in Oncology, 14, 1483454. DOI: 10.3389/fonc.2024.1483454
    • [5] Shao, M.-M. et al. (2024). TMB as a predictive biomarker in NSCLC treated with ICI. Clinical and Translational Oncology. DOI: 10.1007/s12094-023-03370-8

    References (5)

    Yoo, S., Fitzgerald, C. W., Cho, B. A., Fitzgerald, B. G., Han, C., Koh, E. S., et al. (2025). Prediction of checkpoint inhibitor immunotherapy efficacy for cancer using routine blood tests and clinical data. Nature Medicine, 31(3), 869-880.
    Wang, X., Lamberti, G., Di Federico, A., Alessi, J., Ferrara, R., Sholl, M., et al. (2024). Tumor mutational burden for the prediction of PD-(L)1 blockade efficacy in cancer: challenges and opportunities. Annals of Oncology, 35(6), 508-522.
    Di Federico, A., Alden, S., Smithy, J., Ricciuti, B., Alessi, J., Wang, X., et al. (2024). Intrapatient variation in PD-L1 expression and tumor mutational burden and the impact on outcomes to immune checkpoint inhibitor therapy in patients with non-small-cell lung cancer. Annals of Oncology, 35(10), 902-913.
    Yin, X., Song, Y., Deng, W., Blake, N., Luo, X., & Meng, J. (2024). Potential predictive biomarkers in antitumor immunotherapy: navigating the future of antitumor treatment and immune checkpoint inhibitor efficacy. Frontiers in Oncology, 14.
    Shao, M., Xu, Y., Zhang, J., Mao, M., & Wang, M. (2024). Tumor mutational burden as a predictive biomarker for non-small cell lung cancer treated with immune checkpoint inhibitors of PD-1/PD-L1. Clinical and Translational Oncology, 26(6), 1446-1458.

    Explore this topic deeper

    Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

    Click to remove unwanted keywords

    Search 8 keywords →