Nomogram development and validation for predicting minimally invasive step-up approach failure in infected
Gang Li1,2, Shuai Li1, Longxiang Cao1
1Department of Critical Care Medicine, Center of Severe Acute Pancreatitis (CSAP), Jinling Hospital, Medical School of Nanjing University.
Background:
Previous studies have shown that minimally invasive treatment for infected necrotizing pancreatitis (INP) may be safer and more effective than open necrosectomy (ON), but ON is still irreplaceable in a portion of INP patients. Furthermore, there is a lack of tools to identify INP patients at risk of minimally invasive step-up approach failure (eventually received ON or died), which may enable appropriate treatment for them. Our study aims to identify risk factors that can predict minimally invasive step-up approach failure in INP patients and to develop a nomogram for early prediction.
Methods:
Multivariate logistic regression was performed to evaluate the association between minimally invasive step-up approach failure and factors regarding demographics, disease severity, laboratory index, and the location of extrapancreatic necrotic collections. A novel nomogram was developed, and its performance was validated both internally and externally by its discrimination, calibration, and clinical usefulness.
Results:
There were 267, 89, and 107 patients in the training, internal, and external validation cohorts, respectively. Multivariate logistic regression demonstrated that the computed tomography severity index (CTSI) greater than 8 points, Acute Physiology and Chronic Health Evaluation II (APACHE II) score of 16 points or more, early spontaneous bleeding, fungi infection, granulocyte and platelet decrease within 30 days of acute pancreatitis onset, and extrapancreatic necrosis collection located in small bowel mesentery were independent risk factors for minimally invasive step-up approach failure. The area under the curve and coefficient of determination ( R2 ) of the nomogram constructed from the above factors were 0.920 and 0.644, respectively. The Hosmer-Lemeshow test showed that the model had good fitness ( P =0.206). In addition, the nomogram performed well in both the internal and external validation cohorts.
Conclusions:
The nomogram had a good performance in predicting minimally invasive step-up approach failure, which may help clinicians distinguish INP patients at risk of minimally invasive step-up approach failure early.
