Modelling the effect of risk factors on the time to pregnancy using regression models for survival data

Abstract

Information on the time of pregnancy(TTP) was available for a sample of Danish couples in the age range of 20-35 years who were planning to discontinue contraception to achieve a pregnancy. The information was collected by follow-up over six complete menstrual cycles. The aim was to investigate the effect of risk factors such as smoking status, sperm concentration, alcohol consumption and intake of caffeine on the time to pregnancy. for this purpose, a proportional hazards model, an additive hazards model and a proportional hazards model with time varying effects were fitted. We performed goodness-of-fit for each model which indicates which model is appropriate to model the data. All the models revealed a similar result with regard to the significance of covariates. Alcohol consumptions in women and sperm concentration were found to be significantly associated with time to pregnancy. The result obtained from all models showed a negative impact of alcohol consumption of women on the monthly probability of conception. This means that the hazard of time to pregnancy decreases if the number of alcoholic drinks increases for the female. The study also revealed a significant positive effect of sperm concentration on the hazard of time to pregnancy. Checking the goodness-of-fit for all models showed there is a problem in the functional form of the covariate sperm concentration (M.ZKON0). A logarithmic transformation of this covariate resulted in a better fit for the additive hazards model.