We investigated whether microarray-based gene expression profiling of primary tumor biopsy material could be used to predict lymph node (LN) metastasis in patients with uterine squamous cell carcinoma by comparing this approach with magnetic resonance imaging.

Forty three primary cervical cancer samples (16 with LN metastasis and 27 without LN metastasis) from radical hysterectomy with pelvic LN dissection were obtained, RNA was isolated, and oligonucleotide gene chips (Macrogen, Seoul, Korea) were hybridized. The samples were randomly divided into training (31 samples) and test (12 samples) sets. A prediction model for LN metastasis from the training set was developed by support vector machine methods using a 10-fold cross-validation and it was tested for its prediction accuracy by applying it to the test set. We evaluated pelvic LN status by MRI with newly designed criteria in these patients and compared the accuracy of MRI with microarray. In addition, we created a new approach by a combination of both.

The "LN prediction model" derived from the signature of 156 distinctive genes had a prediction accuracy of 83% when applied to the independent test set. MRI showed an accuracy (69%) for the prediction of LN metastasis. The combination model with MRI findings and microarray improved prediction accuracy over MRI alone but the improvement was not statistically significant (74% and 69%, respectively; p=0.688).

Current data show that the prediction of LN metastasis can be allowed by DNA microarray of the primary tumor biopsy, alone or in combination with MRI.