SIRS, The management of steroid-refractory severe ulcerative colitis is still a difficult terrain for the clinical gastroenterologist. The decision between insisting on medical management and indicating colectomy remains a delicate one. 2 The CYSIF study recently showed, in a prospective multicentre European series, that infliximab and ciclosporin were equivalent in this setting, with similar success and colectomy rates. Recently, a paper by Croft et al. seems to dispute these findings. The authors found infliximab clearly superior to ciclosporin as rescue therapy in steroid-refractory severe ulcerative colitis. However, we feel that some aspects of their work deserve a comment that could shed a different light on their conclusions. First and foremost, this paper does not report on a randomised clinical trial, but on the outcomes observed in a real-life clinical series. Treatment allocation was thus not random, and the preferences of the managing physician may have influenced the choice of drug. Also, there is a chronological difference between ciclosporinand infliximab-treated patients: at the beginning of the period considered, only ciclosporin was available, and other local factors may have been at work in those early days, that explain the difference in the results of both drugs. Besides, this work represents a single-centre experience, which may have introduced local preferences and preconceptions in the therapeutic decisions. Possibly, ciclosporin use was not optimized in this study, as reflected by the fact that patients failing to reach therapeutic levels were excluded, instead of being managed with higher doses. However, the most important factor that seriously influences the authors’ findings is that the definition of treatment failure is not clearly stated at all, being instead purely subjective and not predetermined. Therefore, the most important variable was left at the managing clinician’s discretion, without predefined endpoints that could have been applied similarly to both patient populations. As the authors state, the ongoing CONSTRUCT trial will probably add new data to this fascinating clinical situation, but in the meantime, we must insist the following: in this clinical series, treatment allocation was not random, management of both patient populations was not necessarily similar and, most importantly, the decision of declaring medical treatment failure was open, not predefined and apparently left to subjectivity.