Figure 2
From: Decision theory for precision therapy of breast cancer
Selection of samples with definite hormone receptor status. Selection started from 2802 samples HER2−, see Fig. 1. If an IHC estimate for estrogen (ER) was available, ER was considered defined (ER def., 2365). Otherwise the ODDS method57 was applied, yielding either a safe result (imp. ER, 414) or resulting uncertain (imp. ER unc.), contributing to the set of 249 samples shown in beige. ODDS was applied to the set of 414 samples to impute progesterone status, either safely (imp. PGR, 332, green) or ending up uncertain (imp. PGR unc.), contributing to the beige set of 249. Finally, out of the 2365 samples with IHC available, IHC for progesterone was either available (IHC PGR def., 1714, green), safely imputed by ODDS (imp. PGR, 513, green) or uncertain (imp. PGR unc.), again contributing to the beige set. All in all, IHC receptor status was available for 1714 samples and safely imputed for 513 + 332 = 845 samples. This resulted in 2559 samples with definite hormone receptor status, out of which 1432 were positive (red) and 1127 negative (blue). These constituted the input set for DST, as shown in Fig. 7. Note that no samples exist with PGRIHC defined and ERIHC imputed.
