fitglm with binary response and reference level only 0's

Dear all,

I am using a logistic regression model of a binomal response with one categorical predictor comprised of 4 levels. When using the reference level, which for all 89 cases has 0's, the iteration does not work, producing meaningles. When substitung one 0 by 1, the iteraion works and the result seem to make sense. Is there a workaround solution?

best wishes

Christoph