Diagnostic accuracy and prediction increment of markers of epithelial-mesenchymal transition to assess cancer cell detachment from primary tumors

Table 3 Distributions of predicted probabilities for individuals of all-cause mortality within 5 years of diagnosis among colorectal cancer patients for prediction models with and without primary tumor E-cadherin measurements (n = 188)

		Predicted Mortality Probability as Percentage
Model	Range	10th percentile	25th percentile	50th percentile	75th percentile	90th percentile	Mean	SD
Lymph Node Evaluation + Radiologic Imaging (Base Model)	22–69	22	22	22	38	69	33	14
Base Model + Continuous E-cadherin	9–87	15	21	29	41	60	33	17
Base Model + E-cadherin Dichotomized at 0.52	20–79	20	20	35	35	69	33	17
Base Model + E-cadherin Dichotomized at 0.60	19–93	19	19	33	40	67	33	17
Base Model + E-cadherin Dichotomized at 0.85	15–83	15	25	28	45	58	33	17

Base Model includes standard diagnostic tests of lymph node evaluation and radiologic imaging (each coded as dichotomous positive versus negative). Each of the other models includes standard diagnostic tests and cellular membrane E-cadherin expression measured by immunohistochemistry in primary tumor cancer cells on a continuous average intensity scale (0–3), then modeled as continuous or dichotomized at the indicated cut point (if dichotomized, then coded as dichotomous EMT positive versus EMT negative). For all dichotomous predictors, a positive result means evidence supporting detachment of cancer cells from the primary tumor and a negative result means no evidence of detachment. Each model is a Cox proportional hazards models of time from cancer diagnosis to all-cause mortality, censored at 5 years after diagnosis. 62 subjects died within 5 years of diagnosis

ISSN: 1471-2407