As an alternative to proportional hazards model, the additive hazards model has been widely used.To capture the nonlinear effect of some specific covariate, we consider the partially linear additive hazards regression model.The main statistical challenge for this model is that the profile likelihood approach cannot be directly used to eliminate the nuisance functions in estimating the slope parameter efficiently.In the present paper, we proposed a two-stage estimating procedure for the slope parameters.The estimator is shown to be consistent and semiparametrically efficient.Extensive simulation studies provide evidence supporting the theory.The method is illustrated with a real data.