Missing data on covariates occurs in all regression modelling problems. In this talk I discuss how to deal with missing data in a proportional hazard regression model for the duration of breast-feeding in West Australia. The particular issue of substantive interest is the role of important explanatory variables with a high proportion of missing data which are omitted in the complete case analysis, so as not to lose too much of the sample. Complete case analysis is in most cases inefficient, and I describe some of the current attempts to remedy this problem. I give in detail an analysis based on multiple imputation.