The removal of multiplicative, systematic bias allows integration of breast cancer gene expression datasets – improving meta-analysis and prediction of prognosis

Table 2 Published breast cancer datasets used in this study.

Datasets	No. Tumours	Array express/GEO ID	GeneChip	ER+	Age	Tumour Size (cm)	FU (years)	Reference
Chin et al. 2006	114	E-TABM	U133AA	67%	51	2.3	6.1	[16]
Desmedt et al. 2007	198	GSE7390	U133A	68%	47	2.0	13.6	[17]
Farmer et al 2005	49	GSE1561	U133A	58%	-	-	-	[11]
Ivshina et al. 2006	249	GSE4922	U133A	85%	63	2.0	9.9	[18]
Loi et al. 2007	119, 87	GSE6532	U133A, U133 plus2.0	100%	65, 62	2.4, 2.1	5.2, 11.4	[32]
Minn et al. 2007	58	GSE5327	U133A	0%	-	-	7.2	[33]
Pawitan et al. 2005	159	GSE1456	U133A	83%	58^$	2.2^$	7.1	[19]
Richardson et al.	40	GSE3744	U133 plus2.0	38%	-	-	-	[10]
Sotiriou et al. 2006	101*	GSE2990	U133A	71%	60	2.0	5.8	[20]
Wang et al. 2005	286	GSE2034	U133A	73%	52	-	7.2	[52]

Continuous variables (age, size and follow up) are given as median values, except where indicated ^$ the mean was given. The follow up (FU) endpoints for the datasets Loi et al, Pawitan et al. and Sotoriou et al were recurrence-free survival, for datasets Desmedt et al. and Ivshina et al. it was disease-free survival and for datasets Minn et al. and Wang et al. it was distant metastasis-free survival. *The full dataset of Sotiriou et al. includes 189 tumours, but 88 of the Uppsala tumours are included in dataset Ivshina et al.

ISSN: 1755-8794