Table 5.

Association of recurrence or death with changes in Annexin stained PS over CRT after adjustment for confounding by one other predictor

Confounder	RECURRENCE				DEATH
	Slope of line for Ann + ve PS during CRT		Positive slope for Ann + ve PS		Slope of line for Ann + ve PS during CRT		Positive slope for Ann + ve PS
	HR (90 % CI)	p value	HR	p value	HR (90 % CI)	p value	HR	p value
None	4.19 (1.7, 10.3)	0.005	6.40 (2.0, 20.1)	0.001	4.45 (1.6, 12.6)	0.009	3.50 (0.9, 14.0)	0.057
ECOG pre-op	2.89 (1.1, 7.6)	0.036	4.04 (1.0,16.2)	0.049	3.55 (1.1–11.5)	0.038	1.50 (0.3, 8.3)	0.347
N:L ratio at start CRT	3.34 (1.2, 9.5)	0.029	4.88 (1.4, 17.5)	0.021	5.06 (1.14, 22.4)	0.037	1.92 (0.3, 11.5)	0.274
Neutrophil count at start CRT	1.23 (1.0, 1.6)	0.082	4.33 (1.3, 14.5)	0.023	1.45 (1.0, 2.1)	0.056	1.88 (0.4, 9.4)	0.259
Pre-op KPS	3.51 (1.3, 9.2)	0.016	5.43 (1.6, 18.3)	0.011	4.24 (1.4, 13.2)	0.018	2.88 (0.6,13.3)	0.128
EGFRvIII	4.13 (1.4, 12.4)	0.017	6.89 (1.9, 24.9)	0.007	5.9 (1.1, 32.1)	0.041	2.42 (0.5, 1.0)	0.167

Each row represents four models with each adjusted for the confounder. For comparison, the first row shows the results from the univariate model in Table 3. Changes are presented as either a linear change over CRT or the presence of a positive versus a negative slope. As in Table 3, for the linear change, the HR (90 % CI) is presented per one interquartile range (IQR) i.e. the 75th versus the 25th percentile, as shown in Table 2. p Values are one-sided and are only shown for the MV variable

Potential confounders were included here if they achieved an a priori significance level of p < 0.15 in individual models (Table 4)