Clinical Utility of the EpiSwitch CiRT Test to Guide Immunotherapy Across Solid Tumors: Interim Results from the PROWES Study

Abstract

Background: Immunotherapy has revolutionized oncology care, but clinical response to immune checkpoint inhibitors (ICIs) remains unpredictable, and treatment carries substantial risks and costs. The EpiSwitch^® CiRT blood test is a novel 3D genomic assay that stratifies patients by probability of ICI benefit using a binary, blood-based classification: high (HPRR) or low (LPRR) probability of response. Methods: This interim analysis of the ongoing PROWES prospective real-world evidence study evaluates the clinical utility of CiRT in 205 patients with advanced solid tumors. The primary endpoint was treatment decision impact, assessed by pre-/post-test physician surveys. Secondary endpoints included treatment avoidance, time to ICI initiation, concordance with clinical response, early discontinuation rates, and exploratory health economic modeling. Longitudinal use, resistance monitoring, and equity analysis by social determinants of health (SDoH) were also explored. Results: CiRT results influenced clinical decision-making in a majority of cases. LPRR status was associated with higher rates of treatment avoidance and early discontinuation due to immune-related adverse events (IrAEs). In contrast, HPRR patients experienced greater clinical benefit and longer ICI exposure. CiRT classification was not associated with short-term imaging-based response outcomes, supporting its role as an independent predictor. Given that ICI therapy and supportive care can cost more than $850,000 per patient, CiRT offers potential value in avoiding ineffective treatment and associated toxicities. Conclusions: CiRT demonstrates meaningful clinical utility as a non-invasive, predictive tool for guiding immunotherapy decisions across tumor types. It enables more precise treatment selection, improves patient outcomes, and supports value-based cancer care.

Keywords: 3D genome conformation; EpiSwitch CiRT assay; blood-based cancer diagnostics; checkpoint inhibitors; immunotherapy response; precision oncology; predicting checkpoint inhibitor response; real-world evidence.

Figure 1

Study schema for patient enrollment, CiRT testing, decision impact, and evaluable outcome subsets.

Figure 2

This heatmap displays pairwise correlations and variance inflation factors (VIF, on the diagonal) across 205 patient baseline features. Strong negative correlation is observed between sex and height, while ICI therapy status (current or past) shows collinearity with line of therapy and baseline response classification. Variables with VIF > 5, such as “ICI therapy current/past,” suggest moderate-to-high collinearity and were evaluated for exclusion or adjustment in multivariate modeling to reduce redundancy and overfitting.

Figure 3

SHAP (SHapley Additive exPlanations) values derived from a logistic regression model predicting CiRT classification. Each dot represents a patient, with position on the x-axis indicating the impact (SHAP value) of that feature on the model output. Features are ranked by average importance. Color denotes the actual feature value (yellow = high, purple = low). Female sex, lower comorbidity burden, earlier cancer stage, and earlier line of therapy were associated with CiRT High classification. These results support the biological plausibility and clinical alignment of CiRT calls with real-world indicators of immunologic fitness.

Figure 4

SHAP log-odds plots for top predictive clinical features. Panel (A) shows SHAP values for line of therapy, colored by cancer stage. Patients receiving first-line therapy with stage II disease showed the strongest association with a CiRT High classification, while later lines of therapy and stage III disease were more often associated with a Low classification. Panel (B) displays SHAP values for resection status, colored by line of therapy. Patients who had not undergone prior surgical resection and were receiving first-line therapy were more likely to be classified as CiRT High. Higher SHAP values indicate greater model contribution toward a High classification, reinforcing the role of clinical indicators of immune readiness in predicting likelihood of response to ICI therapy.

Figure 5

Linear Discriminant Analysis (LDA) of baseline clinical features stratified by CiRT classification and sex. The X-axis shows LD1 scores from an LDA model trained to predict CiRT classification using all available baseline clinical variables, excluding subject ID, sex, and weight. The Y-axis groups individuals by a combination of CiRT call (high or low probability of response) and sex, which was not used in model training. Each dot represents a patient, and navy rhombus indicate the group mean. Clear separation between CiRT High and Low groups is observed, particularly among male patients, suggesting that CiRT captures a composite immune readiness signal derived from the full clinical profile.

Figure 6

Treatment decision algorithm for patients with unresectable hepatocellular carcinoma (uHCC) incorporating EpiSwitch^® CiRT results. EpiSwitch CiRT results help stratify patients by probability of response to immune checkpoint inhibitors (ICIs). High-probability CiRT results support selection of ICI-based regimens such as durvalumab + tremelimumab or atezolizumab + bevacizumab, while low-probability results may prompt consideration of kinase inhibitors (lenvatinib or sorafenib). Adapted from clinical decision modeling aligned with findings from Ouf et al., where CiRT status was associated with progression-free survival and treatment response in uHCC patients receiving ICIs.