Statistical Methodology
Quantitative measurements are given by median and ranges of observed values. The median follow-up times are calculated by the Kaplan–Meier estimate of potential follow-up. The probabilities of death or to develop a given event at various points in time after the diagnosis of PSC were investigated by a competing risk analysis, considering that a subject can experience more than one type of event and that the occurrence of one type of event may hinder the occurrence of other types. Owing to the presence of competing risks, non-parametric cumulative incidence (CI) estimates are used instead of Kaplan–Meier estimates to construct the 'survival' curves. In case that dates were only known up to the year or month accurate, an exact date is randomly imputed, drawn from a uniform distribution. All analyses have been performed using r software, version 2.9.2. The cuminc function (package cmprisk) of the statistical software r was used to obtain the non-parametric CI functions. The kly function was used for the pairwise comparison of the non-parametric CI functions pertaining to the same competing risk analysis. The Mann–Whitney U-test was used as a two-sample test.