How can I use glmfit to reduce the number of independent variables in a Logistic Regression model?

I am using GLMFIT to perform Logistic Regression on data. I would like to use it to eliminate the independent variables that are statistically unimportant to the model. Is there any technical documentation on the statistical parameters (structure stats) that is the output of GLMFIT that would help me to do this model-fitting? Any help or pointers would be highly appreciated.