Table 3 Univariate and multivariate logistic regression analysis

From: DNA demethylation in normal colon tissue predicts predisposition to multiple cancers

Factor Univariate logistic regression Multivariate logistic regression Stepwise logistic regression
  Coefficients P-value Coefficients P-value Coefficients P-value
(a)       
Age 0.102 0.0262 0.071 0.1476 0.082 0.0793
Gender (M vs F) 0.087 0.8980 0.662 0.4154 Excluded  
Location (C+A vs T) 1.118 0.1060 0.983 0.2096 Excluded  
Stage (I+II vs III) 0.276 0.7470 0.253 0.8018 Excluded  
MSI (MSI vs MSS) −0.271 0.7161 −0.456 0.5930 Excluded  
RDL in NCM 0.299 0.0104 0.306 0.0254 0.276 0.0284
(b)       
Age 0.064 0.0367 0.052 0.1753 0.061 0.0992
Gender (M vs F) −0.034 0.9432 0.121 0.8387 Excluded  
Location (C+A vs T) 0.716 0.1550 0.848 0.1836 Excluded  
Stage (I+II vs III) 1.051 0.0434 1.115 0.0937 1.050 0.1057
MSI (MSI vs MSS) 0.254 0.6549 0.059 0.9306 Excluded  
RDL in NCM 0.442 1.9E-05 0.484 5.8E-05 0.472 4.9E-05
  1. Regression coefficients and P-values of three different logistic regression models to study several factors as predictors of the probability of developing a metachronous tumor (a) or a double tumor, either metachronous or synchronous (b). The regression models in Table a include the data from 10 patients that developed a metachronous tumor and 69 patients that did not develop the tumor in the follow-up period. The regression models in Table b include the data from 75 patients with a single tumor and 24 with double tumors: 14 with synchronous tumors at the moment of the initial resection, and 10 that developed a metachronous tumor in the follow-up period. In the univariate logistic regression analyses, the predictive value of each factor was studied individually, introducing only one factor at a time into the model. In the multivariate logistic regression analyses, all factors were simultaneously introduced into the model. In the stepwise logistic regression the multivariate model containing all the factors was optimized by eliminating and adding factors in a stepwise automatic process until the best-fitting model with the lowest Akaike's Information Criterion (AIC) was obtained. Age and RDL in NCM were introduced as numerical continuous variables, while gender, stage, location and MSI status were introduced as dichotomous categorical variables.
  2. Location. C, A and T: cecum, ascending and transverse colon, respectively.
  3. In bold type significant values (P<0.05).