Josephine van Dongen

Chapter 2 30 was based on likelihood ratio test (LRT). Propor tional hazard assumption was checked using the Schoenfelds’ residuals. We applied multiple imputation to account for missings, as described in supplement. As statistical software we used: SPSS IBM version 25, http://openepi.com and RStudio version 1.1.456. Ethical approval This study was reviewed by the Institutional Review Board of the University Medical Center Utrecht, which has declared it does not involve the Medical Research Involving Human Subjects Act. Figure 1. Study flowchar t. Abbreviations: FU = follow up, SD = standard deviation, N = number.