The Biomarker Value Problem: Immune checkpoint inhibitors have transformed NSCLC treatment, but not all patients benefit equally. Multiple predictive biomarkers exist - PD-L1 IHC, tumor mutational burden (TMB), gene expression profiling - each with costs, limitations, and different predictive accuracy profiles.
What Is Early Health Technology Assessment: Early HTA is a modeling approach that estimates the cost-effectiveness of a medical technology before it enters widespread clinical use, using available data and assumptions to project value. It helps decision-makers prioritize which technologies warrant formal clinical trials and economic evaluation.
Study Objective: This study used early HTA modeling to assess whether adding predictive biomarker testing to standard PD-L1 scoring improves cost-effectiveness in NSCLC immunotherapy decisions, and if so, in which patient subgroups biomarker testing provides the greatest value.
Healthcare System Perspective: The analysis was conducted from a health system payer perspective, comparing strategies of treat-all with immunotherapy, treat based on PD-L1 alone, or treat based on additional composite biomarker testing, using UK NHS cost thresholds as the reference frame.
Decision Tree and Markov Model: The model combined a short-term decision tree (representing the testing and treatment selection phase) with a long-term Markov model (tracking health states over time including progression-free survival, post-progression, and death). This structure captures both immediate testing costs and long-term survival effects.
Test Performance Inputs: For each biomarker test, sensitivity and specificity for predicting immunotherapy response were sourced from clinical trial and validation datasets. These parameters determined how accurately each test identified patients who would benefit versus those who would not.
Cost and Utility Inputs: Drug costs for first-line immune checkpoint inhibitors and chemotherapy regimens were taken from NHS reference costs. Health utilities (quality-adjusted life year weights) for progression-free and post-progression states were sourced from published lung cancer utility studies.
Incremental Cost-Effectiveness Ratio: The primary outcome was the incremental cost-effectiveness ratio (ICER) for each testing strategy compared to no additional testing. Strategies with ICERs below pound 30,000 per QALY gained were considered cost-effective by NHS standards.
Cost-Effective in PD-L1 Less Than 50% Patients: For patients with PD-L1 TPS below 50%, additional biomarker testing (such as TMB or gene expression signatures) was cost-effective. In this group, standard PD-L1 testing does not definitively guide treatment, so additional tests that refine selection provide genuine value by directing effective therapy to likely responders.
Not Cost-Effective for PD-L1 50% or Higher Patients: In patients with PD-L1 TPS >=50%, pembrolizumab monotherapy is already indicated and produces strong response rates. Adding further tests in this group generated incremental costs without proportionate benefit, pushing ICERs above acceptable thresholds.
Value of Information Analysis: A value of information analysis quantified how much uncertainty in test performance parameters drives ICER uncertainty. The analysis showed that reducing uncertainty about test sensitivity for immunotherapy response - by running larger clinical studies - would be most valuable for the economic case for biomarker testing.
Testing Cost Threshold: The model identified a maximum acceptable price per test (approximately several hundred pounds) at which additional biomarker testing crosses the cost-effectiveness threshold. Tests priced above this threshold would require stronger evidence of clinical benefit to be considered cost-effective.
TMB Testing: Tumor mutational burden testing showed moderate cost-effectiveness in PD-L1 intermediate patients. However, its clinical utility is complicated by lack of standardized cutoffs across platforms and limited prospective validation in first-line treatment settings.
Multi-Gene Expression Signatures: RNA expression-based immune signatures (such as T-cell inflamed gene expression profiles) showed favorable cost-effectiveness when test costs were assumed moderate. Their predictive accuracy for both progression-free and overall survival contributed to better ICER performance.
Composite Testing (PD-L1 plus TMB plus GEP): A hypothetical composite test combining PD-L1, TMB, and gene expression profiling achieved the lowest ICER among all strategies tested, reflecting the additive predictive accuracy of combining orthogonal biomarkers - though at higher test cost.
Sensitivity Analyses: The cost-effectiveness conclusions were robust to variations in drug pricing and utility assumptions but were sensitive to test sensitivity and specificity inputs. This reinforces the need for high-quality biomarker validation studies to reduce parameter uncertainty.
Guidance for Test Coverage Decisions: Health technology assessment bodies (NICE in the UK, HAS in France, etc.) can use this model as a framework to evaluate biomarker test reimbursement. The findings suggest conditional coverage of TMB or GEP tests in the PD-L1 <50% population could be economically justified.
Trial Design Recommendations: The value of information analysis identifies where clinical trial investment would most reduce ICER uncertainty. Specifically, head-to-head trials of testing strategies in the PD-L1 intermediate group (TPS 1%-49%) would most meaningfully improve economic certainty.
Price Negotiation Lever: The model provides a quantitative maximum test price for cost-effective use, which health systems can use in negotiations with diagnostics manufacturers. Tests priced above the threshold face a clear evidence burden to justify their cost.
International Applicability: While calibrated to UK NHS thresholds, the model structure can be adapted to other healthcare systems by substituting country-specific drug costs, test prices, and willingness-to-pay thresholds, making it a generalizable tool for global health policy.
Early HTA Uncertainty: As an early HTA, the model relies heavily on assumptions and indirect evidence for test accuracy parameters. Formal cost-effectiveness analyses using real-world outcomes data from large clinical registries would provide more reliable estimates.
Biomarker Landscape Evolution: The immunotherapy biomarker field evolves rapidly. New tests (ctDNA, microbiome signatures, radiomics) and new indications are emerging continuously, meaning this model requires regular updating to remain relevant.
Treatment Line Specificity: The model focused on first-line treatment decisions. Biomarker testing value may differ substantially in second-line or maintenance settings, where patient populations, response rates, and treatment options are different.
Patient Preferences: Economic models optimize population-level cost per QALY but may not capture individual patient preferences around certainty, testing burdens, or the psychological value of knowing one's predicted treatment response. Patient preference studies would complement the health economic analysis.