Results: Forty-seven patients were treated in the first interval and 53 in the second. There was no difference in patient demographics, clinical or walled-off necrosis characteristics and laboratory parameters between the groups, apart from a higher proportion of women and Caucasians in the later period. The treatment success rate was higher for the algorithmic approach compared with conventional treatment (91 versus 60 per cent respectively; P <0·001). On multivariable logistic regression, management based on the algorithm was the only predictor of treatment success (odds ratio 6·51, 95 per cent c.i. 2·19 to 19·37; P = 0·001).