Abstract
BACKGROUND AND PURPOSE: Gene expression signatures are often used to identify hypoxic tumors. However, intratumoral heterogeneity raises concern that multiple biopsies may be necessary to assess global hypoxia status. The objective of this study was to compare the impact of heterogeneity on the discriminative capacity of several previously described hypoxia gene signatures and determine if a single biopsy is sufficient to obtain a reliable estimate of hypoxia in cervical cancer. MATERIALS AND METHODS: Multiple biopsies (33) were obtained from 11 locally advanced (FIGO IB to IVB) cervical cancers prior to treatment. Ten hypoxia gene signatures were analyzed. Variance component analysis was used to determine the ratio of within-tumor variability to total-tumor variability when one to five biopsies are available for analysis (W/T(1-5)). The mean standardized error in the signature scores was estimated by comparing the score using one biopsy randomly selected from each tumor to the 'global' score using all available biopsies. RESULTS: The ten hypoxia signatures were comprised of 6-99 genes each. The W/T(1) ratios for individual genes commonly found in the signatures ranged from 0.17 to 0.73. W/T(1) ratios for the signatures were generally lower (0.21-0.45), implying greater capacity to discriminate among tumors. With additional biopsies, the signature W/T ratios (ie W/T(2-5)) decreased further. The mean error in the signature scores varied from 0.27 to 0.40 of one standard deviation, suggesting high capacity to discriminate among tumors with different global hypoxia scores. CONCLUSIONS: Compared with individual probes, hypoxia gene expression signatures are generally more consistent across multiple biopsies from different regions of a tumor and more tolerant of intratumoral heterogeneity.