Abstract
Distant metastasis of papillary thyroid cancer occur in 5-10 % patients. No prognostic factors are known until now. Microarray technique investigating gene expression profile in tumor tissues seems to be a new diagnostic tool which may be applied for this goal.
Aims: The aim of the study was to assess the feasibility of genomic prediction of papillary thyroid cancer metastasis risk.
Methods: We assessed gene expression profile of 49 PTC samples taken intraoperatively (11 from patients with distant metastases, 38 from patients with no disease dissemination) by HG-U133A oligonucleotide microarrays. Both patients with distant metastases present at diagnosis as well recurrent distant disease (follow-up up to 11 years) were included while local relapse was not considered. Microarray dataset was pre-processed by GC-RMA method, class prediction was carried out by linear discriminant analysis with 0.632+-bootstrap. Data were validated on independent set of 85 PTC samples (18 patients with distant metastases and 68 patients with no dissemination of disease) by real-time quantitative PCR, with class prediction by linear discriminant analysis.
Results: We found out that 108 genes were significantly associated with the risk of distant relapse (non-corrected p<0.001, false discovery rate for the list approx. 20%), although each alone was a weak predictors of the risk. We assessed an index of risk, calculated from the best 10 genes in a cross-validation procedure. The estimated negative predictive value (NPV) of the method was high (81%), with much weaker positive predictive value (PPV) of 33%.
In the validation on an independent set of 85 samples, the index built from 28 genes chosen from microarray study confirmed the initial study, with NPV 85% and PPV 32%. Similar results were obtained by reduction of the classifier size to 10 genes.
Conclusion: Prediction of low risk of PTC dissemination is feasible, with clinically relevant negative prognostic value of multi-gene classifier over 80%.