Circulating Immune Protein Signature Is Validated as OS Prognostic Tool in Recurrent Ovarian Cancer

A previously developed prognostic model combining 4 baseline circulating immune-oncology proteins and 4 clinical variables retained its ability to stratify overall survival (OS) risk when applied to an independent validation cohort of patients with recurrent ovarian cancer, according to data from the phase 3 ANITA/ENGOT-OV41/GEICO 69-O trial (NCT03598270) presented during the 2026 ESMO Gynaecological Cancers Congress.¹

In the validation cohort (n = 54), the model's concordance index for OS was 0.61 (95% CI, 0.50-0.72), with an area under the curve (AUC) of 0.70 (95% CI, 0.51-0.89) at 6 months and 0.69 (95% CI, 0.50-0.89) at 12 months. Patients classified as high risk had shorter OS than those classified as low risk (HR, 2.17; 95% CI, 1.02-4.63; log-rank P = .049), a finding consistent with the original 332-patient cohort in which the model was developed (HR, 2.65; 95% CI, 1.99-3.52; log-rank P < .0001).

"Liquid biopsy-based protein signatures may improve risk stratification in recurrent ovarian cancer and guide future translational and clinical research in this setting," Antonio González-Martín, MD, the principal investigator of the Translational Oncology Group, the director of the Cancer Center Clínica Universidad de Navarra and of the Department of Medical Oncology, and a medical oncology specialist at Clínica Universidad de Navarra, Madrid, Spain, said during the presentation.

What were the background and design of this analysis?

In the ANITA trial, adding atezolizumab (Tecentriq) to a platinum doublet (without bevacizumab [Avastin]) followed by niraparib (Zejula) maintenance did not improve progression-free survival (PFS), the primary end point, or OS in patients with recurrent ovarian cancer.² An exploratory objective of the trial was to evaluate the relationship between clinical outcomes and selected immune-oncology proteins assessed in plasma before and during treatment using multiplexing immunoassays across four Luminex (ProcartaPlex) panels.¹

Investigators previously reported a prognostic model combining 4 baseline circulating immune-oncology proteins, VISTA (B7-H5), CA125, CD96 (Tactile), and VEGF-A, with 4 clinical variables: BRCA status, PD-L1 status, treatment-free interval from last platinum (TFIp), and initial FIGO stage. That model demonstrated a concordance index of 0.69 (95% CI, 0.65-0.73) and a 6-month ROC AUC of 0.87 (95% CI, 0.85-0.91). The current analysis reports external validation of that model along with exploratory longitudinal biomarker analyses.

The original OS model was based on multivariable Cox models adjusted for prespecified covariates and validated internally with 10-fold cross-validation. For external validation, the frozen model coefficients were applied to an independent cohort of 54 additional patients to generate individual risk scores. Associations between dynamic changes in protein levels by z-scores (baseline to cycle 4) and efficacy were analyzed using LASSO-penalized multivariable Cox models adjusted for PD-L1 status, BRCA status, and TFIp (n = 310). Pre-maintenance protein levels were assessed for association with OS using the same modeling approach, adjusted for PD-L1 and BRCA status (n = 207).

What did the pre-maintenance and dynamic biomarker analyses show?

In the pre-maintenance circulating marker analysis (n = 207), 4 proteins were associated with favorable OS: ICOS ligand (HR, 0.49; 95% CI, 0.37-0.63; P < .0001), uPAR (HR, 0.51; 95% CI, 0.35-0.75; P < .001), PD-L1 positivity (HR, 0.66; 95% CI, 0.44-0.99; P = .043), and CD31 (HR, 0.71; 95% CI, 0.54-0.94; P = .017). Five markers were associated with unfavorable OS: CA125 (HR, 1.18; 95% CI, 1.03-1.35; P = .014), TIMD4 (HR, 1.28; 95% CI, 1.03-1.59; P = .0017), EpCAM (HR, 1.35; 95% CI, 1.08-1.68; P = .0077), CTLA-4 (HR, 1.43; 95% CI, 1.11-1.85; P = .0061), and Nectin-2 (HR, 1.56; 95% CI, 1.17-2.09; P = .0025), along with non-mutated BRCA status (HR, 2.34; 95% CI, 1.26-4.35; P = .0069). This pre-maintenance model, adjusted for PD-L1 and BRCA status, achieved a concordance index for OS of 0.73 (95% CI, 0.69-0.77; likelihood ratio test [LRT] P < .0001), with AUCs ranging from 0.77 to 0.91 across the 6- to 36-month landmarks assessed.

In the analysis of dynamic protein changes from baseline to cycle 4 (n = 310 paired samples), a decrease in VEGF-A was associated with favorable outcomes for both PFS (HR, 0.83; 95% CI, 0.75-0.93; P = .001) and OS (HR, 0.83; 95% CI, 0.79-0.92; P = .0006). Decreases in E-cadherin (HR, 0.83; 95% CI, 0.72-0.96; P = .010) and calreticulin (HR, 0.81; 95% CI, 0.66-1.00; P = .049) were also associated with favorable OS, while an increase in CD73 was associated with unfavorable OS (HR, 1.23; 95% CI, 1.05-1.44; P = .012). This dynamic-change model achieved concordance indices of 0.65 (95% CI, 0.61-0.70; LRT P < .0001) for OS and 0.61 (95% CI, 0.57-0.65; LRT P < .0001) for PFS.

“We have confirmed the robustness of our prognostic model for OS based on the determination of some circulating immune oncology proteins at baseline, in addition to 4 clinical variables on treatment,” González-Martín said. “Protein chains, especially those involving anti-angiogenic pathway, were associated with OS and may reflect treatment response and tumor biology. The difference between the circulating immune oncology proteins identified in the OS model at baseline and the OS model before starting maintenance may reflect the biology changes and clone selection induced by the chemotherapy.”

References

1. González-Martín A, Tavira B, González-Gomariz J, et al. Validation of a prognostic circulating immune protein signature and exploratory dynamic biomarker analyses from the phase 3 ANITA (ENGOT-Ov41/GEICO 69-O) trial in recurrent ovarian cancer. Presented at: 2026 ESMO Gynaecological Cancers Annual Congress; June 17, 2026. Abstract 1RO.
2. González-Martín A, Rubio Pérez MJ, Heitz F, et al. Final results from the ENGOT-Ov41/GEICO 69-O/ANITA randomised phase III trial of atezolizumab (atezo), platinum-based chemotherapy (CT) and maintenance niraparib (nira) for late-relapsing recurrent ovarian cancer (rOC). Ann Oncol. 2025;36(suppl 2):S670-S671. doi: 10.1016/j.annonc.2025.08.1706