Integrative Analysis of DCE-MRI and Gene Expression Profiles in Construction of a Gene Classifier for Assessment of Hypoxia-Related Risk of Chemoradiotherapy Failure in Cervical Cancer.

Purpose: A 31-gene expression signature reflected in dynamic contrast enhanced (DCE)-MR images and correlated with hypoxia-related aggressiveness in cervical cancer was identified in previous work. We here aimed to construct a dichotomous classifier with key signature genes and a predefined classification threshold that separated cervical cancer patients into a more and less hypoxic group with different outcome to chemoradiotherapy.
Experimental Design: A training cohort of 42 patients and two independent cohorts of 108 and 131 patients were included. Gene expression data were generated from tumor biopsies by two Illumina array generations (WG-6, HT-12). Technical transfer of the classifier to a reverse transcription quantitative PCR (RT-qPCR) platform was performed for 74 patients. The amplitude ABrix in the Brix pharmacokinetic model was extracted from DCE-MR images of 64 patients and used as an indicator of hypoxia.
Results: Classifier candidates were constructed by integrative analysis of ABrix and gene expression profiles in the training cohort and evaluated by a leave-one-out cross-validation approach. On the basis of their ability to separate patients correctly according to hypoxia status, a 6-gene classifier was identified. The classifier separated the patients into two groups with different progression-free survival probability. The robustness of the classifier was demonstrated by successful validation of hypoxia association and prognostic value across cohorts, array generations, and assay platforms. The prognostic value was independent of existing clinical markers, regardless of clinical endpoints.
Conclusions: A robust DCE-MRI-associated gene classifier has been constructed that may be used to achieve an early indication of patients' risk of hypoxia-related chemoradiotherapy failure. Clin Cancer Res; 22(16); 4067-76. ©2016 AACR.
(©2016 American Association for Cancer Research.)