Letter to the Editor: Omission of key predictors of sepsis undermines the validity of risk modelling for post-endoscopic retrograde cholangiopancreatography biliary infection

Abstract

Post-endoscopic retrograde cholangiopancreatography biliary infection remains a significant cause of morbidity in malignant perihilar biliary obstruction. A recent model by Wang et al, published in the recent issue of the World Journal of Gastroenterology, which identified hypokalaemia, Bismuth-Corlette classification, and aspartate aminotransferase as predictors, yet omitted critical sepsis determinants. The absence of key variables, including baseline inflammatory biomarkers, biliary colonization, stent characteristics, antibiotic protocols, and particularly the degree of biliary drainage, such as hepatic volume decompressed and undrained ducts after failed cannulation, significantly limits validity. Drainage adequacy is central to infection pathogenesis and outcome. Comprehensive models incorporating these parameters are essential for accurate risk prediction and meaningful clinical application.

Key Words: Endoscopic retrograde cholangiopancreatography; Infection; Drainage; Sepsis; Stents; Biliary; Ducts

Core Tip: Post-endoscopic retrograde cholangiopancreatography biliary infection is a significant risk in malignant obstruction. Current risk models are incomplete, omitting critical factors like drainage adequacy (hepatic volume decompressed) and inflammatory biomarkers. Comprehensive models are essential for accurate postoperative biliary infection risk prediction and better clinical outcomes.

TO THE EDITOR

We read with great interest the recent multicentre retrospective study by Wang et al[1] published in the World Journal of Gastroenterology. The authors should be commended for applying both logistic regression and artificial neural network (ANN) approaches to predict postoperative biliary infection (PBI) following endoscopic retrograde cholangiopancreatography (ERCP) in malignant perihilar biliary obstruction (MPHBO). However, we want to raise a key methodological concern that may undermine the predictive validity and generalizability of the model. The variables evaluated Bismuth Corlette classification, hypokalaemia and aspartate aminotransferase although statistically significant, miss one of the most established determinants of post ERCP sepsis which is adequacy of biliary drainage. This omission poses a particular problem in ANN-based prediction because ANNs are very sensitive to variable selection. The omission of mechanism predominant predictors like drainage adequacy will disproportionately invalidate the model by compelling the network to learn incomplete causal pathways. The amount of hepatic volume to which decompression is successfully achieved, the proportion of undrained ducts, and the technical success of cannulation are the factors that directly determine biliary sterility and risk of infection. The adequacy of drainage must be measured in objective terms such as: (1) The amount of liver mass that was drained (preferably ≥ 50 of total hepatic volume); (2) Whether one or both sides was drained in lesions of the Bismuth type III-IV; and (3) Any remaining undrained hepatic mass should be reported after the operation.

Kühl et al[2], in finding the main factor in determining infectious sequelae following ERCP, demonstrated that patients with residual obstruction had significantly increased rates of cholangitis and bacteremia. Similar results were drawn by Tarnasky and Kedia[3] who found that the key predictive factor of post-ERCP cholangitis is incomplete biliary drainage, and the central modifiable factor that determines the outcome of infections is the adequacy of the procedures. Decades ago, Deviere et al[4] showed that septicaemia after endoscopic internal drainage of hilar malignant strictures was associated with undrained intrahepatic ducts and partial decompression. Additional important determinants that were not included in the analysis are stent characteristics (type and size), and underlying aetiology. Self-expandable metal stents provide superior patency and reduced rates of stent occlusion and cholangitis compared with plastic stents in malignant obstruction and larger or multiple stents enhance time to dysfunction; conversely, small-diameter plastic stents have higher reintervention rates factors that materially affect post-procedural infectious risk and need to be included in predictive models[5-7]. Moreover, the type of primary tumor (cholangiocarcinoma, gallbladder carcinoma, hepatocellular carcinoma, or metastatic disease) affects the ductal anatomy, the probability of multifocal obstruction, the response to biliary decompression and stent patency thereafter, all of which affect the risk of infection and clinical outcome[2,3,7].

The extent of these parameters relating to drainage is limited, and the modelling is restricted from matching the recommendations of the Tokyo Guidelines 2018, which emphasize complete biliary decompression as the cornerstone of sepsis prevention[8]. Such critical procedural data are likely to be omitted in the environment of MPHBO where the biochemical factors are inadvertently overstressed due to the exclusion of anatomical complexity in many cases.

In the future, predictive models should thus incorporate objective measures of biliary drainage sufficiency, hepatic volume decompressed, and remnant obstruction to enhance the mechanistic validity of clinical application of PBI risk prediction. This will ensure that the models are more realistic in terms of the procedural determinants of the infection in the real-world and enable risk-adjusted intervention planning that is evidence-based.