Table 5.

Trade-offs between strength and precision of cut-point/time-to-death association and number of patients with low or high expression, by stage distribution and 5-year risk of death, for multiple cut points of E-cadherin weighted average expression

	Cut point 0.52b		Cut point 0.60		Cut point 0.85
Cox model estimatea	HR 2.57	95 % CI (CLR) 1.10, 6.03 (5.48)	HR 2.40	95 % CI (CLR) 1.29, 4.49 (3.48)	HR 1.75	95 % CI (CLR) 0.99, 3.08 (3.11)
Stage	Low	High	Low	High	Low	High
Local (n = 99)	6	93	16	83	56	43
Regional (n = 66)	5	61	11	55	41	25
Distant (n = 23)	0	23	1	22	11	12
Mortality risk
Died within 5 years (n = 62) (x)	7	55	14	48	42	20
Total at risk (n = 188) (y)	11	177	28	160	108	80
5-year risk of death post-surgery (%) (x/y)	64	31	50	30	39	25

Mortality Risk Effect estimatesc	Estimate	95 % CI	Estimate	95 % CI	Estimate	95 % CI
Risk difference (%)	33	3, 62	20	0, 40	14	1, 27
Risk ratio	2.05	1.25, 3.37	1.67	1.07, 2.59	1.56	0.99, 2.43

CLR Confidence limit ratio (upper limit/lower limit)

E-cadherin measured as a weighted average (weighted by area analyzed) of continuous tumor core average intensities (0–3) prior to dichotomization. Of 190 subjects, 2 had missing data for E-cadherin, but multiple imputation enabled retention of these 2 in Cox models

^aFor effect of dichotomous E-cadherin expression (low versus high) on time-to-death, adjusted for age (continuous), TNM stage (local/regional/distant), neoadjuvant chemotherapy (yes/no), and neoadjuvant radiation treatments (yes/no)

^bStatistically-optimal cut point by best model fit

^cUnadjusted comparisons of low versus high expression status