A novel prognostic tool leveraging tumor fraction (TF) from liquid biopsies shows promise in refining patient selection for early-phase cancer trials. The tool, which integrates TF with other key clinical markers, aims to improve survival prediction and reduce early mortality in patients entering Phase I oncology studies.
Currently, selecting appropriate patients for Phase I oncology trials remains a challenge. While guidelines suggest a life expectancy exceeding three months, a significant 15-20% of patients succumb within 90 days of enrollment. Existing scoring systems like the Royal Marsden Hospital (RMH) and Gustave Roussy Immune (GRIM) scores offer some guidance, but their precision is limited.
Tumor Fraction as a Prognostic Biomarker
Liquid biopsies are increasingly used in advanced cancer care, with circulating tumor DNA (ctDNA) analysis providing valuable insights. Measuring tumor fraction (TF) within ctDNA offers a quantitative assessment of tumor burden. Prior studies have indicated that higher TF often correlates with poorer prognosis across various cancer types.
This study investigated whether TF, derived from commercial liquid biopsy tests, could provide substantial prognostic data to improve patient selection for early-phase trials. The research focused on determining if TF could serve as a reliable biomarker to predict survival outcomes.
Study Design and Patient Cohorts
The study involved two cohorts: a development cohort (BIP, n=965) and a validation cohort (STING, n=947), both comprising patients with advanced solid tumors enrolled between December 2020 and December 2021. The median follow-up was 17.1 months for the BIP cohort and 9.9 months for the STING cohort. Median overall survival (OS) was 11.5 months and 11.8 months, respectively. A correlation between TF and OS was observed in both cohorts.
Model Development and Validation
Multivariate analysis within the BIP cohort identified TF, albumin, LDH, metastatic site count, and NLR as significantly associated with OS. The refined model incorporated elevated TF, reduced albumin, presence of three or more metastatic sites, and heightened NLR as indicators of decreased OS. The 3-month AUROC curve registered at 79.87% (95%CI: 75.40-84.34), with a Brier score of 0.091, indicating good accuracy.
Validation in the STING cohort yielded a 3-month AUROC of 76.11% (95%CI: 69.31-82.90) and a Brier score of 0.093, confirming the model's predictive capability.
Nomogram for Survival Prediction
A nomogram was constructed to predict three-month survival probabilities, integrating TF, albumin levels, metastatic sites, and NLR. Each variable was weighted based on its prognostic significance. For example, a patient with high TF (≥10%), low albumin, more than two metastatic sites, and a high NLR would accumulate a high score, correlating with a lower three-month survival probability.
Independent Prognostic Value of TF
Further analyses demonstrated that TF was independently predictive of OS, irrespective of tumor volume. The prevalence of high TF did not significantly differ between patients with high versus low tumor burden (44.4% vs. 40.5%, p=0.65), suggesting that ctDNA shedding is more influenced by intrinsic tumor biology than by tumor volume.
