Combining imaging- and gene-based hypoxia biomarkers in cervical cancer improves prediction of chemoradiotherapy failure independent of intratumor heterogeneity

Purpose: Emerging biomarkers from medical imaging or molecular characterization of tumor biopsies open up for combining the two and exploiting their synergy in treatment planning. We compared pretreatment classification of locally advanced cervical cancer patients by two previously validated imaging- and gene-based hypoxia biomarkers, appraised the influence of intratumor heterogeneity, and investigated the benefit of combining them in prediction of chemoradiotherapy failure. Experimental Design: Hypoxic fraction, determined from dynamic contrast enhanced (DCE)-MR images, and an expression signature of 6 hypoxia-responsive genes were used as imaging- and gene-based biomarker, respectively, in 118 patients. Intratumor heterogeneity was assessed by variance analysis. The biomarkers were combined using a dimension reduction procedure. Results: The two biomarkers classified 75% of the patients with the same hypoxia status. Inconsistent classification in some cases was not related to imaging-defined intratumor heterogeneity in hypoxia, and hypoxia status of the slice covering the biopsy region was representative of the whole tumor. Hypoxia assessed by gene expression was independent on tumor cell fraction in the biopsies and showed minor heterogeneity across multiple samples in 9 tumors. Inconsistent classification was therefore rather caused by a difference in the hypoxia phenotype reflected by the biomarkers, providing a rational for combining them into a composite score. This score showed improved prediction of treatment failure (HR:7.3) compared to imaging (HR:3.8) and genes (HR:3.0) and significant prognostic impact in multivariate analysis with clinical variables. Conclusion: Combining our imaging- and gene-based biomarkers enables more precise and informative assessment of hypoxia-related chemoradiotherapy resistance in cervical cancer.