The current rating systems for predicting mortality in patients with acute pancreatitis have limited accuracy and limited clinical value, according to a systematic review of the evidence published online July 26 in the Annals of Internal Medicine.
"[D]espite the many decades since prediction models of mortality in [acute pancreatitis] were proposed, evidence establishing their prognostic accuracy is incomplete. Furthermore, the available data do not provide clear guidance on which models should be used in speciﬁc patient populations and how they should direct speciﬁc therapy to achieve improved clinical outcomes or enhance the cost-effectiveness of care," Meng-Yang Di, MD, PhD, from Brown University, Providence, Rhode Island, and colleagues write.
Acute pancreatitis varies widely in severity, from mild to serious illness with systemic involvement and multiorgan failure. Serious pancreatitis can affect about 25% of patients and carries a mortality rate between 20% and 30%. The prognosis, however, can be challenging to make on presentation or early in the disease course.
About 20 scoring systems exist for acute pancreatitis. Studies have suggested these systems vary widely in test characteristics and prognostic accuracy. Some assess a large number of variables, making them unwieldy for daily use. Even though several clinical guidelines recommend using a scoring system, they differ on which system to use, and how to use it.
In the study, researchers searched Ovid MEDLINE and EMBASE from inception until May 2016 for studies in any language that evaluated the prognostic accuracy, incremental predictive value (the value of adding one scoring system to another), and clinical utility (effect on patient outcomes) of scoring systems for predicting mortality in acute pancreatitis. They also looked at cost-effectiveness. For the analysis, they used the Quality in Prognosis Studies tool, which has been used in past reviews of predictive models.
The analysis included 94 studies (91 cohort studies, three randomized controlled trials) that assessed 18 scoring systems in 53,547 patients. All studies evaluated prognostic accuracy, although most scores had low prognostic accuracy. Only six studies assessed incremental predictive value, but none did so as a primary outcome. No studies evaluated clinical utility or cost-effectiveness. All studies received the rating of fair to low quality resulting from high to moderate risk for bias.
The Acute Physiology and Chronic Health Evaluation (APACHE) II score and the Ranson criteria were studied most broadly for prognostic accuracy in predicting mortality. The APACHE II score had a median sensitivity of 100% (range, 68% - 100%) and a median specificity of 63% (range, 21% - 96%). The Ranson criteria had a median sensitivity of 90% (range, 0% - 100%) and a median specificity of 67% (range, 14% - 97%).
Limitations included wide clinical heterogeneity and inadequate methodological and reporting quality, which precluded a meta-analysis.
The study did not evaluate scoring system performance in patients with mild acute pancreatitis, so these findings do not apply to this patient population.
"Our ﬁndings suggest that there is an opportunity for further research to clarify how best to use the existing prediction scores and assess their effect, as well as to develop new prediction tools. Some suggestions apply to both the validation of existing scores and the development of new ones," the authors conclude.
In a linked editorial, Chris E. Forsmark, MD, from the University of Florida, Gainesville, and Dhiraj Yadav, MD, MPH, from the University of Pittsburgh, Pennsylvania, disagreed about the need for further research on prediction models of mortality in acute pancreatitis.
"Regarding research, we believe that developing new multiple-factor clinical scoring systems will not increase accuracy in prediction, and that all current systems are inadequate to guide clinical judgment," they write.
They point out that widely available, simple clinical and laboratory evaluations such as serum hematocrit, blood urea nitrogen, and the presence of systemic inﬂammatory response syndrome may be "as accurate" as scoring systems in predicting severe acute pancreatitis.
Dr Forsmak and Dr Yadav highlighted the recently revised Atlanta classification system, which received international consensus. It simplifies decision making and provides "much clearer and more quantiﬁable deﬁnitions of severe [acute pancreatitis than scoring systems]," they write. The Atlanta classification system separates acute pancreatitis into three categories: severe (persistent organ failure), moderate (transient organ failure, or local systemic complications without persistent organ failure), and mild (no organ failure or local or systemic complications).
"Future studies should focus on the translational research needed to identify better biological markers of the primary determinants of outcome—that is, organ failure and pancreatic necrosis—as well as host-related factors that inﬂuence the severity of the systemic response to pancreatic necrosis," they conclude, "Such an approach might help in understanding the mechanisms of severe disease, identifying at-risk patients soon after or before pancreatic injury occurs, and developing therapies to mitigate disease severity."
The study was funded by the Global Scholarship Programme for Research Excellence for 2014 to 2015, The Chinese University of Hong Kong. The authors have disclosed no relevant financial relationships. Dr Yadav reports consulting for AbbVie and receiving royalties from Up-To-Date.
Medscape Medical News © 2016 WebMD, LLC
Send comments and news tips to email@example.com.
Cite this: Pancreatitis Mortality Prediction Models Lack Evidence - Medscape - Jul 28, 2016.