Comparative Predictive Value Of Immunotherapy Biomarkers Across Cancers

Key Takeaways

• ctDNA ranked highest for sensitivity and overall discriminative ability in the pooled comparison.
• PD-L1 performance depended on threshold, with higher cutoffs aligning with specificity and lower cutoffs aligning with sensitivity.
• MSI was most specific, TMB showed a middle-range balance, and the authors said performance varied by cancer type and clinical context.

A systematic review and network meta-analysis compared 13 biomarkers for immunotherapy benefit across cancers in 54,634 patients from 194 clinical studies worldwide. ctDNA had the highest sensitivity at 0.82 (95% CI 0.72-0.89) and the strongest overall discriminative ability, with a DOR of 9.75 (95% CI 5.20-16.73) and an AUC of 0.769. In the pooled analysis, PD-L1 performance also shifted meaningfully by cutoff, with a tradeoff between sensitivity and specificity. No single biomarker led every metric, and performance differed by biomarker type.

The investigators synthesized biomarkers used to predict benefit from immune checkpoint inhibitors across tumor types, including ctDNA, PD-L1 thresholds, TMB, MSI, irAEs, PLR, NLR, LIPI, and LDH. Searches spanned multiple databases and trial registries from inception to 1 September 2025, providing a broad comparative evidence base. The analysis emphasized pooled diagnostic metrics across a mixed literature rather than a single cancer-specific setting. Subgroup analyses were performed for various cancers, and the investigators also assessed heterogeneity and publication bias. Within that framework, threshold choices and marker class shaped different strengths across the pooled rankings.

For PD-L1, the 50% or higher cutoff had the highest specificity at 0.78 (95% CI 0.73-0.81), diagnostic accuracy with DOR 2.60 (95% CI 1.86-3.52) and AUC 0.661, but the lowest sensitivity at 0.42 (95% CI 0.36-0.49). By contrast, the 1% or higher cutoff had the highest sensitivity at 0.68 (95% CI 0.65-0.71) and the lowest specificity at 0.48 (95% CI 0.45-0.51). TMB showed a moderate balance, with sensitivity of 0.56 (95% CI 0.50-0.60) and specificity of 0.69 (95% CI 0.65-0.73), while MSI reached the highest specificity overall at 0.89 (95% CI 0.85-0.93) but sensitivity of 0.36 (95% CI 0.27-0.46). irAEs showed sensitivity of 0.69 (95% CI 0.60-0.77) and specificity of 0.59 (95% CI 0.50-0.67), PLR slightly exceeded NLR by AUC 0.623 versus 0.613, and LIPI and LDH had AUCs of 0.585 and 0.544. Across markers, different biomarkers led sensitivity, specificity, and overall discrimination rather than one marker leading uniformly.

The authors identified ctDNA, high-threshold PD-L1 such as 50% or higher, and TMB as leading predictors, while suggesting that combinations could optimize performance. They also concluded that biomarker performance varied by cancer type and clinical context, limiting how evenly any single ranking applies across settings. The PubMed abstract called for better handling of heterogeneity and stronger standardization, but did not detail subgroup findings or bias results. Overall, the comparative rankings were informative but context-dependent, rather than establishing one universally dominant biomarker.