Expression profiling to predict the clinical behaviour of ovarian cancer fails independent evaluation

BMC Cancer

Table 2 Comparison with published gene sets. Accuracy of all published gene set models on the independent data set both when predicting stage and platin resistance ranked by stage accuracy. Gene sets have been named after the first author of the publication followed by a description of its relationship to patient outcome. References have been used when the same first author had multiple publications

Gene set first author	Description	Stage accuracy (%)	Platin accuracy (%)
Ouellet	low malignant potential/borderline disease vs. invasive disease: tumour tissue	97.96	55.56
Hibbs	Disease vs. normal or other tissues	95.92	66.67*
Spentzos	Residual disease vs. complete response at second look surgery	93.88	37.78
Lu	Disease vs. normal	93.88	48.89
Helleman	Platin sensitivity vs. platin resistance: differential expression	93.88	44.44
Ouellet	low malignant potential/borderline disease vs. invasive disease: primary cultures	91.84	53.33
Zhu	Clear cell vs. serous histology	91.84	66.67*
Lancaster [14]	Short-term vs. long-term survival	91.84	44.44
Helleman	Platin sensitivity vs. platin resistance: 16-gene predictive model	91.84	46.67
Berchuck	Short-term vs. long-term survival	91.84	55.56
Schwartz	Clear cell vs. other histological types	91.84	46.67
Hartmann	Early vs. late relapse after platin based chemotherapy	91.84	66.67*
Spentzos	Short-term vs. long-term survival	87.76	66.67*
Lancaster [14]	Disease vs. normal	79.59	66.67*
Roberts	Platin sensitivity vs. platin resistance	75.51	42.22
Lancaster [20]	Ovarian cancer tissue vs. metastatic tissue: 27-gene predictive model	71.43	60.00#
Lancaster [20]	Ovarian cancer tissue vs. metastatic tissue: differential expression	57.14	66.67*

*models predicting only the majority class (platin sensitive patients) on the independent data set
#best platin model

ISSN: 1471-2407