Abstract
Background/Aim: The diagnostic accuracy (DA) for patients with acute pancreatitis (AcPa) has been rarely evaluated. Patients and Methods: In the AcPa study group, there were 22 patients versus 1,311 patients in the non-AcPa group. The clinical history-taking variables (CHT) (n=22), clinical signs & tests details (CST) (n=14), and laboratory analyses (n=3) were recorded in each patient. Meta-analytical techniques were used to detect the summary sensitivity (Se) and specificity (Sp) estimates for each data set; CHT, CST, and diagnostic scores (DS). Results: In receiver operating characteristic (ROC) analysis, the area under curve (AUC) values for i) CHT, ii) CST, and iii) DS were as follows: i) AUC=0.640 (95%CI=0.550-0.730); ii) AUC=0.588 (95%CI=0.520-0.656), and iii) AUC=0.943 (95%CI=0.910-0.976). The differences between these AUC values (roccomp analysis) are as follows: i) versus ii) p=0.155; i) versus iii) p<0.0001; ii) versus iii) p<0.0001. Conclusion: The new DS introduced in this study proved to be far superior to both symptoms and signs & tests in its DA for AcPa, as demonstrated by HSROC analysis.
Acute pancreatitis (AcPa) is one of the most common diseases leading to hospitalization for gastroenterological reasons in the Western world (1-3). AcPa is characterized by a sudden inflammation of the pancreas, focal necrosis of the pancreatic fat, and hemorrhage. The clinical picture of AcPa ranges from mild to a severe life-threatening disease. The main causes for AcPa are excess alcohol use and/or migration of small gallstones with obstruction of bile or pancreatic enzyme flow. However, in up to 20% of AcPa patients the etiology of the disease is unknown. AcPa is diagnosed by using clinical history-taking (CHT) and clinical signs & tests (CST). Imaging studies are used to identify organ failure, offer prognostic information, and detect if fluid replacement is in balance.
Upper abdominal pain, radiating to the back in 50% of patients, is usually the pre-dominant clinical symptom in AcPa following poor appetite, nausea, and vomiting (2, 3). The most common clinical signs of AcPa are fever, tachycardia, and positive clinical tests showing peritoneal irritation. In severity grading of AcPa, there are several prognostic scoring models that have been used as predictors of organ failure, complications, and survival in AcPa patients (2-6). The American College of Gastroenterology (ACG) (7) and World Society of Emergency Surgery (WSES) (8) have produced guidelines for the diagnosis and treatment of AcPa. These guidelines are mostly based on classic risk assessment tools such as Ranson criteria, which can predict disease severity and suggest treatment in AcPa, but they have a significant limitation, because many of these criteria used are not obtainable at presentation to the emergency, which makes the prediction how AcPa will progress difficult for emergency physicians (9). Therefore, due to the lack of diagnostic accuracy (DA) studies on AcPa patients, we assessed the performance of i) CHT, ii) CST, and iii) the diagnostic score (DS) in confirming AcPa.
Patients and Methods
The AcPa study group comprised 22 patients with clinically confirmed AcPa and 1,311 patients in the non-AcPa group. The CHT (n=22), and CST variables (n=14) as well as laboratory analyses (n=3) were recorded in each patient. The diagnosis of AcPa was confirmed by considering all CHT and CST variables and results of the laboratory tests together and following the diagnostic criteria of acute abdominal pain (AAP) and AcPa.
DS models. A multivariate logistic (stepwise) regression analysis (SPSS Statistics 6.0.0.1; IBM, Armonk, NY, USA) was used to disclose the variables with an independent predictive value. All the variables of CHT and CST presented in Table I and Table II were included in the analysis as binary data, e.g., AcPa=1 and other diagnosis of AAP=0. Using the coefficients of the regression model, a DS was built and its DA for AcPa was studied. The coefficient of the multivariate analysis shows the relative risk of a patient with a given CHT and CST variable to have AcPa.
The clinical history of acute pancreatitis (AcPa) patients versus other cause of acute abdominal pain.
Diagnostic score (DS) for acute pancreatitis (AcPa) patients DS model. The DS model shown at six different cut-off levels of symptoms, signs, and tests. Cut-off levels: DS I=0.0017, DS II=0.0165, DS III=0.048, DS IV=0.055, *DS V=0.0165, and *DS VI=0.0046.
The DS formula for AcPa was: DS I-IV=2.60×Sex (male=1, female=0)+1.15×Nausea (PE=1, NE=0)-0.84×Vomiting (PE=1, NE=0)+1.57×Appetite (PE=1, NE=0)+3.06×Use of alcohol (PE=1, NE=0)+1.96×Tenderness (PE=1, NE=0)-9.43, where PE=positive endpoint and NE=negative endpoint.
The diagnostic score without nausea and vomiting was: DS V-VI=2.48×Sex (male=1, female=0)+1.58×Appetite (PE=1, NE=0)+2.97×Use of alcohol (PE=1, NE=0)+1.94×Tenderness (PE=1, NE=0)-8.90. PE=positive endpoint and NE=negative endpoint.
Statistical analysis. STATA/SE version 17.1 (StataCorp, College Station, TX, USA) was used for analysis. The statistical tests presented were two-sided, and p-value <0.05 was considered statistically significant. Using 2×2 tables, sensitivity (Se) and specificity (Sp) with 95% confidence intervals (95%CI) for each CHT and CST variables were determined. The study protocols of hierarchical receiving operating characteristic (HSROC) analysis and meta-analytical technique (metaprop; STATA) used in this study were detailed in previous reports by Meklin et al. (10-12).
Results
The study cohort. In the AcPa study group, 22 patients were included, and the non-AcPa group consisted of 1,311 patients, with the following AAP diagnoses: non-specific abdominal pain (NSAP) (n=614), acute appendicitis (AA) (n=271), acute cholecystitis (n=124), acute renal colic (n=59), acute small bowel obstruction (ASBO) (n=53), non-organic dyspepsia (n=50), and other AAP patients (n=138). The mean (SD) age was 37.5 (21.7) years.
The CHT in AcPa. The overall Se of the CHT for detecting AcPa was 60% (95%CI=47-73%) (Figure 1). The Se was higher than 60% for 10 of the symptoms. The five most sensitive CHT variables (vertigo, appetite, jaundice, micturition, sex) showed 91-100% Se in the diagnosis of AcPa (Figure 1). The overall Sp of the CHT for detecting AcPa was only 58% (95%CI=43-71%) (Figure 2). Altogether, 13 symptoms showed Sp higher than 58%. The five most specific CHT details of AcPa (intensity of pain, previous indigestion, drugs for abdominal pain, previous abdominal diseases, use of alcohol) showed 79-96% Sp (Figure 2).
Sensitivities of history-taking in acute pancreatitis (AcPa) (random-effects model). ES: Estimated sensitivity; CI: confidence interval.
Specificities of history-taking in acute pancreatitis (AcPa) (random-effects model). ES: Estimated specificity; CI: confidence interval.
The CST in AcPa. The overall Se of the CST for AcPa was 53% (95%CI=34-71%) (Figure 3), and 8 signs & tests had Se exceeding 53%. The six most accurate CST variables (scar, tenderness, mass, guarding, rectal digital tenderness, urine) showed 73-100% Se (Figure 3). The overall Sp of the CST variables was 60% (95%CI=42-76%) (Figure 4), while 9 CST variables showed Sp higher than 60%. The five most specific CST variables (colour, abdominal movement, distension, Murphy’s positive, bowel sounds) showed 86-93% Sp (Figure 4).
Sensitivities of signs and tests in acute pancreatitis (AcPa) (random-effects model). ES: Estimated sensitivity; CI: confidence interval.
Specificities of clinical signs and tests in acute pancreatitis (AcPa) (random-effects model). ES: Estimated specificity; CI: confidence interval.
The DS in AcPa. The most significant predictors of AcPa in multivariate analysis were sex, nausea, vomiting, appetite, use of alcohol, and tenderness (Table II). The best diagnostic level for DS model [DS II; Se=91%, Sp=90%, efficiency (Eff)=90%] was reached at cut-off level 0.0165 for DS (Figure 5 and Figure 6, Table II). The DS models were tested at six different cut-off levels to disclose the highest DA (Figure 5 and Figure 6). The overall Se and Sp of these six DS models was 88% (95% CI=79-96%) and 82% (95% CI=64-95%), respectively (Figure 5 and Figure 6). Four of these models showed Se ≥91% and four models had Sp ≥90% (Table II).
Sensitivities of diagnostic scores at six different cut-off levels (DS I-VI). ES: Estimated sensitivity; CI: confidence interval.
Specificities of diagnostic scores at six different cut-off levels (DS I-VI). ES: Estimated specificity; CI: confidence interval.
HSROC and AUC values. HSROC curves were used to visualise the pooled DA of the CHT (Figure 7), and CST variables (Figure 8) and different DS models (Figure 9) in detecting AcPa. In SROC analysis, the AUC values for i) CHT ii) CST iii) DS were as follows: i) AUC=0.640 (95%CI=0.550-0.730); ii) AUC=0.588 (95%CI=0.520-0.656), and iii) AUC=0.943 (95%CI=0.910-0.976). The differences between these AUC values (roccomp analysis) were as follows: i) versus ii) p=0.155; i) versus iii) p<0.0001; ii) versus iii) p<0.0001.
Hierarchical summary receiver operating characteristic (HSROC) curve of the history-taking in acute pancreatitis (AcPa).
Hierarchical summary receiver operating characteristic (HSROC) curve of the clinical signs and tests in acute pancreatitis (AcPa).
Hierarchical summary receiver operating characteristic (HSROC) curve of the six diagnostic scores (DSs).
Discussion
The difficulty of AcPa diagnosis among the AAP patients and the lack of DA studies with HSROC analysis, motivated us to assess the DA of i) CHT, ii) CST variables, and iii) the DS in AcPa.
Clinical history-taking. Due to the clinical complexity, it is difficult to recall all of the symptoms at once, and therefore it is advisable to set out the most important variables in making the differential diagnosis between AcPa and non-AcPa. A special attention is paid to the location of pain, which in AcPa tends to begin in the upper half of the abdomen (UAbd) and often remains in the UAbd until the AcPa patient presents to hospital. In our study, 77% of the AcPa patients had location of initial pain at UAbd and 73% of the patients had location of pain at diagnosis at UAbd. The pain of AcPa is often quite severe and in our study 36% of AcPa patients had intolerable pain. One feature useful in distinguishing AcPa from peritonitis or acute perforated peptic ulcer is that the pain in AcPa is sometimes colicky or intermittent as in 45% of patients in our study, whereas this type of pain is very rare in peritonitis. In 23% of AcPa patients, the pain is aggravated by movement, coughing, and respiration and this feature does not distinguish AcPa from the lower abdominal diseases such as AA, ASBO, and acute diverticulitis. Other gastrointestinal symptoms are quite often helpful in the diagnosis of AcPa. In particular, this applies to nausea and vomiting, both usually favouring AcPa. However, the DA of these variables has not been shown in detail before. In our study, 77% of the AcPa patients had nausea and 50% had vomiting. Because the pain in AcPa is usually quite short in duration, most patients with AcPa do not notice any abnormality in bowel habits (bowels normal in 73%, Table I) or micturition on admission (micturition normal in 100%, Table I). It is worthwhile enquiring about previous use of alcohol, because AcPa tends to be associated with regular drinking of alcohol. In this series, 59% of the AcPa patients had positive clinical history of regular use of alcohol.
Clinical signs and tests. The patient with AcPa is unlikely to have a flat board-like totally immobile abdomen at physical examination. This happened only in about one out of 10 patients with AcPa in our study (abdominal movement normal 91%, Table III). Regarding tenderness at abdominal palpation, about three quarters of AcPa patients had UAbd tenderness (73%), whereas rebound was present in half (50%) of the AcPa patients. Rigidity was present in a third (32%) of the AcPa patients only, and this is a point of distinction from perforated peptic ulcer or peritonitis, where rigidity is a rule rather than exception. At abdominal auscultation, about a third of the AcPa patients in our study had abnormal bowel sounds and the two thirds of AcPa patients had bowel sounds, which could be distinguished from normal.
The clinical signs and investigations of acute pancreatitis (AcPa) patients versus other cause of acute abdominal pain.
In the disease severity grading of AcPa, there are several prognostic scoring models that have been used as predictors of organ failure, complications, and survival in AcPa patients (2-6). Chatzicostas et al. (4) assessed the DA of Ranson and acute physiology and chronic health evaluation (APACHE) II and III models in predicting the severity of AcPa. All three models correlated with length of stay and disease severity. AUC for Ranson score was far higher than that for APACHE II or III models (0.817 versus 0.618, and 0.676, respectively). These data suggest that Ranson score is a slightly better prognostic formula than the more complicated APACHE II and III scoring systems (3).
Triantopoulou et al. (5) compared APACHE II model, C-reactive protein levels, and computed tomography index (Balthazar, CTSI) in DA of AcPa. They found a significant correlation between CTSI model and local complications in AcPa, whereas no correlation was available between CTSI model and survival or hospitalization in the intensive care unit. Although the pancreatic necrosis detected on contrast-enhanced CT is considered as a risk factor for poor survival in AcPa patients, the reported results are still widely controversial, as shown in the Triantopoulou et al.’s study. These authors concluded that the CTSI model is not an efficient tool in predicting survival in AcPa.
Papachristou et al. (6) proposed a bedside index for severity in AcPa (BISAP) intended for early identification of the patients at risk for in-hospital mortality. The aim of their study was to compare BISAP [blood urea nitrogen >25 mg/dl, impaired mental status, systemic inflammatory response syndrome (SIRS), age>60 years, and pleural effusions] with the “traditional” multifactorial scoring systems: Ranson’s, APACHE-II, and the CTSI in predicting severity, pancreatic necrosis (PNec), and mortality in a prospective cohort of patients with AP. The AUC values for BISAP, Ranson’s, APACHE-II, and CTSI in predicting SAP were 0.81, 0.94, 0.78, and 0.84, respectively. The authors concluded that the prognostic accuracy of BISAP is similar to that of the other AcPa scoring systems (6).
Vannier E et al. (13) conducted a retrospective study at Caen University Hospital between January 2014 and December 2017, including patients hospitalized for AcPa. They developed an Admission Severe Acute Pancreatitis (ASAP) model based on parameters associated with a severe AcPa and validated a model in a cohort of 80 patients. Hypothermia, low oxygen saturation or albumin levels, and high creatinine levels were significantly associated with severe AcPa. The ASAP model showed notable DA with an AUC value of 0.82, which was significantly higher than that of the Sequential Organ Failure Assessment (SOFA model) and Balthazar score using the −2.1742 threshold; the ASAP score had a sensitivity and specificity of 74% and a negative predictive value of 95%. These predictive performances for ASAP score were confirmed in the validation cohort.
ROC analysis has become popular to evaluate the DA of various clinical methods and tests (10-12). Figure 7 shows the ROC curve for the CHT in AcPa; the curve closely parallels the diagonal reference line (AUC=0.5) with low AUC value (AUC=0.640; 95%CI=0.550-0.730). The DA of the CST was even lower than that of the CHT (AUC=0.588; 95%CI=0.520-0.656, Figure 8).
Figure 9 shows the ROC curve for the DS, with the curve moved towards the upper left corner, showing significantly better DA in the AcPa than that of the CHT or CST. In this study, however, the DA of the DS (AUC=0.943; 95%CI=0.910-0.976) was significantly higher for the AcPa patients than that of CHT or CST. Although, this simple DS for AcPa performed well, the present AcPa cohort was quite small for final evidence and possibly a new clinical trial with a larger AcPa patient group is needed to determine the validity of this DS and to provide additional data for possible revisions.
Conclusion
The new DS introduced in this study proved to be far superior to both symptoms and CST in its diagnostic accuracy for AcPa, as demonstrated by HSROC analysis. Unfortunately, we could not perform direct comparisons to earlier studies on AcPa, because this is the first study to provide evidence that DS could significantly improve the difficult diagnosis of AcPa.
Acknowledgements
The study was funded by the Päivikki and Sakari Sohlberg Foundation, Helsinki, Finland.
Footnotes
Authors’ Contributions
All Authors contributed to the collection and analysis of data, drafting, and revising the manuscript, and read and approved the final article.
Conflicts of Interest
The Authors report no conflicts of interest or financial ties in relation to this study.
- Received June 12, 2022.
- Revision received July 12, 2022.
- Accepted July 13, 2022.
- Copyright © 2022, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
