Analysis of postpartum lactational amenorrhoea in relation to breast-feeding: some methodological and practical...

A large multicentre epidemiological study was carried out by WHO between 1991 and 1995 to analyse the duration of lactational amenorrhoea in relation to breast-feeding. The main results of this analysis, which used classical statistical modelling, have been already published. However, some specific aspects of the postpartum fertility amenorrhoea and breast-feeding covariates, and more specifically the observed progressive exhaustion of the breast-feeding inhibitory effect on the reproductive axis, may justify a closer look at the validity of the statistical tools. Indeed, as has already been emphasized, analysis of large longitudinal data sets in reproduction often faces three difficulties: (i) the precise determination of the event of interest, (ii) the way to handle the time evolution of both the studied variables and their effect on the event of interest and (iii) the often discrete nature of the data and the associated problem of tied events. The first objective of the present work was to give additional insights into the estimation and quantification of the dynamics of the effect of breast-feeding over time, considering this covariate either as fixed or time-dependent. The second objective was to show how to perform the analyses using corresponding adapted procedures in widely available statistical packages, without the need for acquiring particular programming skills.