G Model YDLD-2572; No. of Pages 6
ARTICLE IN PRESS Digestive and Liver Disease xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Liver, Pancreas and Biliary Tract
Predictors of severe and critical acute pancreatitis: A systematic review Catherine J. Yang, Joseph Chen, Anthony R.J. Phillips, John A. Windsor, Maxim S. Petrov ∗ Department of Surgery, University of Auckland, Auckland, New Zealand
a r t i c l e
i n f o
Article history: Received 20 November 2013 Accepted 26 January 2014 Available online xxx Keywords: Acute pancreatitis Severity Prediction Persistent organ failure Infected pancreatic necrosis
a b s t r a c t Background: Persistent organ failure and infected pancreatic necrosis are major determinants of mortality in acute pancreatitis, but there is a gap in the literature assessing the best available predictors of these two determinants. The purpose of this review was to investigate the utility of predictors of persistent organ failure and infected pancreatic necrosis in patients with acute pancreatitis, both alone and in combination. Methods: We performed a systematic search of the literature in 3 databases for prospective studies evaluating predictors of persistent organ failure, infected pancreatic necrosis, or both, with strict eligibility criteria. Results: The best predictors of persistent organ failure were the Japanese Severity Score and Bedside Index of Severity in Acute Pancreatitis when the evaluation was performed within 48 h of admission, and blood urea nitrogen and Japanese Severity Score after 48 h of admission. Systemic Inflammation Response Syndrome was a poor predictor of persistent organ failure. The best predictor of infected pancreatic necrosis was procalcitonin. Conclusions: Based on the best available data, it is justifiable to use blood urea nitrogen for prediction of persistent organ failure after 48 h of admission and procalcitonin for prediction of infected pancreatic necrosis in patients with confirmed pancreatic necrosis. There is no predictor of persistent organ failure that can be justifiably used in clinical practice within 48 h of admission. © 2014 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Prediction of disease severity is one of the most important clinical goals in the management of patients with acute pancreatitis. An accurate marker of severity allows early identification of those patients who require transfer to a referral centre, treatment in an intensive care unit (ICU), and/or specific interventions [1–3]. The literature is replete with studies investigating predictors for the severity of acute pancreatitis, but these studies are limited by the great variation in their definitions of severity (i.e. what they aim to predict) [4]. Definitions include local and/or systemic complications (as defined by the 1992 Atlanta symposium) [5], infectious pancreatic complications alone, admission to an ICU, length of ICU or hospital stay, complications requiring intervention, grading by radiological imaging [6], and multifactorial prognostic clinico-biochemical scores [7]. It has been argued that improvements in predicting severity of acute pancreatitis can hardly
∗ Corresponding author at: Department of Surgery, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand. Tel.: +64 9 923 2776; fax: +64 9 377 9656. E-mail address: [email protected] (M.S. Petrov).
be achieved without agreement on which endpoints should be used [2]. Recent advances in the understanding of the drivers of severity of acute pancreatitis has drawn attention to the importance of persistent organ failure (POF), as well as infected pancreatic and/or peripancreatic necrosis (IPN), as determinants of mortality in patients with acute pancreatitis [8,9]. As a result, the determinant-based international multidisciplinary classification has been developed [10] and validated [11–13]. One of the features of the new classification is a new category of acute pancreatitis severity called “critical” acute pancreatitis. This category incorporates patients with both POF and IPN. The new international multidisciplinary classification has also updated the definitions of “severe” acute pancreatitis. This category now incorporates patients with either POF or IPN [10]. Given that POF and/or IPN are the major determinants of mortality in acute pancreatitis, there is an urgent clinical need to determine the best available predictors of these two determinants of severity. To the best of our knowledge, no study has systematically addressed this important clinical question. Thus, the aim of this study is to conduct a systematic literature review to determine the utility of predictors of POF and IPN, both alone and in combination.
1590-8658/$36.00 © 2014 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.dld.2014.01.158
Please cite this article in press as: Yang CJ, et al. Predictors of severe and critical acute pancreatitis: A systematic review. Dig Liver Dis (2014), http://dx.doi.org/10.1016/j.dld.2014.01.158
G Model YDLD-2572; No. of Pages 6
ARTICLE IN PRESS C.J. Yang et al. / Digestive and Liver Disease xxx (2014) xxx–xxx
2
2. Methods
reference tests, and cut off levels were extracted from the source papers by one author (CJY).
2.1. Study identification A search of EMBASE, Cochrane Reviews, and PubMed databases from 1950 to January 2013 was carried out. The search strategy for predictors of POF included “persistent organ failure/ OR multiple organ failure” AND “acute pancreatitis” AND “Predicting/ OR predictor/ OR predict/ OR identify/ OR indicator/ OR indicate/ predictors/ OR sign”. The search strategy for predictors of IPN included “infected pancreatic necrosis/ OR infective pancreatic necrosis/ OR infected necrotizing pancreatitis/ OR infectious peripancreatic complication/ OR acute peripancreatic fluid collection/ OR acute post necrotic pancreatic collection/ OR infective pancreatic necrosum” AND “predicting/ OR predictor/ OR predict/ OR identify/ OR indicator/ OR indicate/ predictors/ OR sign”. Language restrictions were not applied. Additionally, the citations found in primary and review articles were hand searched for relevant publications that were not selected by the above searches. The Internet search engines Google Scholar and Bing were also used with the above search terms to further identify relevant articles. Titles and abstracts of all citations were independently screened by 2 authors (CJY, JC) and any disagreement between them was resolved by the senior author (MSP). 2.2. Selection criteria The study inclusion criteria were patients with acute pancreatitis, a prospective clinical study design, and an ‘index test’ comprising any single predictor or combination of predictors. The index test had to be collected prior to the diagnosis of POF or IPN. The reference standard had to be either Sequential Organ Failure Assessment or Multiple Organ Dysfunction Score for the predictors of POF, and surgery, autopsy, percutaneous drainage, fine needle aspiration, or gas bubbles on computed tomography for the predictors of IPN. Organ failure was defined as reported by the authors of the original papers. POF was defined as either organ failure for more than 48 h in duration, or failure of at least 2 organ systems. IPN was defined as the presence of bacteria and/or fungi within areas of diffuse or focal necrotic tissue. We were very careful to exclude those papers with poor or unacceptable definitions. We also excluded patients who were transferred before collection of predictor data, retrospective studies, studies with suboptimal reference standards, and papers examining reference standards as their test predictor (i.e. selffulfilling predictors). In particular, in the published papers on predictors of IPN, commonly the reference standard was not only intra-operative or autopsy findings, but could be clinical follow-up alone. Having such imprecise reference standard was a criterion for exclusion. In addition, we also excluded studies using predictive criteria (e.g. Atlanta classification, Japanese Severity Scale) as the reference standard to define severity, as their use is invariably associated with a 30–40% misclassification error [1,2]. Studies were excluded if the index test on predictors of IPN was fine needle aspiration and gas bubbles on computed tomography, as well as if sensitivity and specificity of the predictor were not reported, or the original paper provided no sufficient data to derive such parameters. The strict study eligibility criterion is one of the important features of this literature review. 2.3. Data acquisition The number of patients, study design, clinical setting, study population, prevalence of POF or IPN and their definitions with
2.4. Quality assessment Study quality was assessed using the QUADAS tool for assessing diagnostic studies [14]. The parameters evaluated for inclusion were: the study was judged to be of a representative spectrum; the study satisfied the defined selection criteria; the time between standard and reference tests was acceptable; the population received the reference standard regardless of index test result; the index test was independent of the reference standard; there was a sufficient description of the index test and a sufficient description of the reference standard; the interpretation of the reference standard was conducted while blinded to the outcome of the index test; the authors reported on the interpretable results and explained any withdrawals from the studies. For the POF section, the inclusion criteria were the same as above, except that the question related to the availability of clinical data when interpreting the index test results was omitted in this analysis, as that information is not applicable to testing for organ failure. For the IPN section, the question related to the verification using reference standard was omitted, as it would be unethical to perform a fine needle aspiration in patients without symptoms. This gave a maximum score of 11 points for predictors of POF and 12 points for predictors of IPN.
2.5. Statistical analysis The data for predictors of POF and IPN were analysed separately according to 2 commonly used clinical assessment periods after hospital admission for acute pancreatitis (10, conclusive increase in the likelihood of study endpoint [17]. Sensitivity was defined as the proportion of patients with POF or IPN that had a positive result on the index test. Specificity was defined as the proportion of patients without POF or IPN who had a negative index test result. The positive likelihood ratio is the ratio of the true positive rate to the false positive rate. The negative likelihood ratio is the ratio of the false negative rate to the true negative rate. DOR measures the effectiveness of a diagnostic test: it is the ratio of the odds of the test being positive in the presence of the disease, relative to the odds of the test being positive in the absence of the disease. This study met the criteria for waiver of ethical approval. No consent was required from patients involved in the reviewed studies.
Please cite this article in press as: Yang CJ, et al. Predictors of severe and critical acute pancreatitis: A systematic review. Dig Liver Dis (2014), http://dx.doi.org/10.1016/j.dld.2014.01.158
G Model YDLD-2572; No. of Pages 6
ARTICLE IN PRESS C.J. Yang et al. / Digestive and Liver Disease xxx (2014) xxx–xxx
3
Fig. 1. PRISMA diagram of literature searches for predictors of persistent organ failure and infected pancreatic necrosis.
3. Results
predictors: Ranson’s, BISAP, Japanese Severity Scale, Pancreatitis Outcome Prediction (POP), Systemic Inflammation Response Syndrome (SIRS), and blood urea nitrogen (p < 0.05). However, there was no significant difference between the sensitivity of APACHE II and those of creatinine, Glasgow, Harmless Acute Pancreatitis Score (HAPS), and Panc3. The Japanese Severity Scale with a cut-off of 1 had the best specificity (0.90) and this was significantly better than the specificities of all the other studied predictors (p < 0.05). The Japanese Severity Scale also had the best positive likelihood ratio (5.19) and this was significantly better than the positive likelihood ratios of creatinine, HAPS, and Panc3 (p < 0.05) but there was no significant difference between the positive likelihood ratio of the Japanese Severity Scale and that of APACHE II, BISAP, Ranson’s, Glasgow, POP, SIRS, and blood urea nitrogen. POP with a cut-off of 9 had the best negative likelihood ratio (0.61), and this was significantly better than the negative likelihood ratio of APACHE II (p < 0.05), while there was no significant difference between POP and other predictors. APACHE II with a cut-off of 7 had the best DOR (13.71) and this was significantly better than the DOR of HAPS and Panc3 (p < 0.05) but there was no significant difference between the DOR of APACHE II and other predictors. Among predictors evaluated at or after 48 h of admission, the accuracy of the Ranson’s score was the most frequently evaluated (in 3 cohorts). The remaining 11 predictors were evaluated in only 1 or 2 cohorts. The data on a total of 11 predictors were suitable for meta-analysis (Supplementary Table S1). When comparing predictors after 48 h of admission, sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and DOR were considered separately. APACHE II with a cut off of 7 had the best sensitivity (0.84). This was significantly better than the specificities of Ranson’s
3.1. Predictors of POF A total of 526 studies were assessed for eligibility and 7 studies examining a total of 15 predictors were included in the final review. The flow chart of the selection process is outlined in Fig. 1 (PRISMA diagram). Of the 15 predictors included, 11 were evaluated in more than 1 study. The methodological quality of the included studies was excellent, as 6 of 7 studies reached the maximum score of 11 points (Table 1). Table 1 lists the characteristics of the studies examining predictors of POF. Overall, the studies enrolled a total of 1589 patients (on average, 199 patients per study) with the incidence of POF calculated as 17% (273 of 1498 reported). Timing of index test was
ARTICLE IN PRESS Digestive and Liver Disease xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Liver, Pancreas and Biliary Tract
Predictors of severe and critical acute pancreatitis: A systematic review Catherine J. Yang, Joseph Chen, Anthony R.J. Phillips, John A. Windsor, Maxim S. Petrov ∗ Department of Surgery, University of Auckland, Auckland, New Zealand
a r t i c l e
i n f o
Article history: Received 20 November 2013 Accepted 26 January 2014 Available online xxx Keywords: Acute pancreatitis Severity Prediction Persistent organ failure Infected pancreatic necrosis
a b s t r a c t Background: Persistent organ failure and infected pancreatic necrosis are major determinants of mortality in acute pancreatitis, but there is a gap in the literature assessing the best available predictors of these two determinants. The purpose of this review was to investigate the utility of predictors of persistent organ failure and infected pancreatic necrosis in patients with acute pancreatitis, both alone and in combination. Methods: We performed a systematic search of the literature in 3 databases for prospective studies evaluating predictors of persistent organ failure, infected pancreatic necrosis, or both, with strict eligibility criteria. Results: The best predictors of persistent organ failure were the Japanese Severity Score and Bedside Index of Severity in Acute Pancreatitis when the evaluation was performed within 48 h of admission, and blood urea nitrogen and Japanese Severity Score after 48 h of admission. Systemic Inflammation Response Syndrome was a poor predictor of persistent organ failure. The best predictor of infected pancreatic necrosis was procalcitonin. Conclusions: Based on the best available data, it is justifiable to use blood urea nitrogen for prediction of persistent organ failure after 48 h of admission and procalcitonin for prediction of infected pancreatic necrosis in patients with confirmed pancreatic necrosis. There is no predictor of persistent organ failure that can be justifiably used in clinical practice within 48 h of admission. © 2014 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Prediction of disease severity is one of the most important clinical goals in the management of patients with acute pancreatitis. An accurate marker of severity allows early identification of those patients who require transfer to a referral centre, treatment in an intensive care unit (ICU), and/or specific interventions [1–3]. The literature is replete with studies investigating predictors for the severity of acute pancreatitis, but these studies are limited by the great variation in their definitions of severity (i.e. what they aim to predict) [4]. Definitions include local and/or systemic complications (as defined by the 1992 Atlanta symposium) [5], infectious pancreatic complications alone, admission to an ICU, length of ICU or hospital stay, complications requiring intervention, grading by radiological imaging [6], and multifactorial prognostic clinico-biochemical scores [7]. It has been argued that improvements in predicting severity of acute pancreatitis can hardly
∗ Corresponding author at: Department of Surgery, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand. Tel.: +64 9 923 2776; fax: +64 9 377 9656. E-mail address: [email protected] (M.S. Petrov).
be achieved without agreement on which endpoints should be used [2]. Recent advances in the understanding of the drivers of severity of acute pancreatitis has drawn attention to the importance of persistent organ failure (POF), as well as infected pancreatic and/or peripancreatic necrosis (IPN), as determinants of mortality in patients with acute pancreatitis [8,9]. As a result, the determinant-based international multidisciplinary classification has been developed [10] and validated [11–13]. One of the features of the new classification is a new category of acute pancreatitis severity called “critical” acute pancreatitis. This category incorporates patients with both POF and IPN. The new international multidisciplinary classification has also updated the definitions of “severe” acute pancreatitis. This category now incorporates patients with either POF or IPN [10]. Given that POF and/or IPN are the major determinants of mortality in acute pancreatitis, there is an urgent clinical need to determine the best available predictors of these two determinants of severity. To the best of our knowledge, no study has systematically addressed this important clinical question. Thus, the aim of this study is to conduct a systematic literature review to determine the utility of predictors of POF and IPN, both alone and in combination.
1590-8658/$36.00 © 2014 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.dld.2014.01.158
Please cite this article in press as: Yang CJ, et al. Predictors of severe and critical acute pancreatitis: A systematic review. Dig Liver Dis (2014), http://dx.doi.org/10.1016/j.dld.2014.01.158
G Model YDLD-2572; No. of Pages 6
ARTICLE IN PRESS C.J. Yang et al. / Digestive and Liver Disease xxx (2014) xxx–xxx
2
2. Methods
reference tests, and cut off levels were extracted from the source papers by one author (CJY).
2.1. Study identification A search of EMBASE, Cochrane Reviews, and PubMed databases from 1950 to January 2013 was carried out. The search strategy for predictors of POF included “persistent organ failure/ OR multiple organ failure” AND “acute pancreatitis” AND “Predicting/ OR predictor/ OR predict/ OR identify/ OR indicator/ OR indicate/ predictors/ OR sign”. The search strategy for predictors of IPN included “infected pancreatic necrosis/ OR infective pancreatic necrosis/ OR infected necrotizing pancreatitis/ OR infectious peripancreatic complication/ OR acute peripancreatic fluid collection/ OR acute post necrotic pancreatic collection/ OR infective pancreatic necrosum” AND “predicting/ OR predictor/ OR predict/ OR identify/ OR indicator/ OR indicate/ predictors/ OR sign”. Language restrictions were not applied. Additionally, the citations found in primary and review articles were hand searched for relevant publications that were not selected by the above searches. The Internet search engines Google Scholar and Bing were also used with the above search terms to further identify relevant articles. Titles and abstracts of all citations were independently screened by 2 authors (CJY, JC) and any disagreement between them was resolved by the senior author (MSP). 2.2. Selection criteria The study inclusion criteria were patients with acute pancreatitis, a prospective clinical study design, and an ‘index test’ comprising any single predictor or combination of predictors. The index test had to be collected prior to the diagnosis of POF or IPN. The reference standard had to be either Sequential Organ Failure Assessment or Multiple Organ Dysfunction Score for the predictors of POF, and surgery, autopsy, percutaneous drainage, fine needle aspiration, or gas bubbles on computed tomography for the predictors of IPN. Organ failure was defined as reported by the authors of the original papers. POF was defined as either organ failure for more than 48 h in duration, or failure of at least 2 organ systems. IPN was defined as the presence of bacteria and/or fungi within areas of diffuse or focal necrotic tissue. We were very careful to exclude those papers with poor or unacceptable definitions. We also excluded patients who were transferred before collection of predictor data, retrospective studies, studies with suboptimal reference standards, and papers examining reference standards as their test predictor (i.e. selffulfilling predictors). In particular, in the published papers on predictors of IPN, commonly the reference standard was not only intra-operative or autopsy findings, but could be clinical follow-up alone. Having such imprecise reference standard was a criterion for exclusion. In addition, we also excluded studies using predictive criteria (e.g. Atlanta classification, Japanese Severity Scale) as the reference standard to define severity, as their use is invariably associated with a 30–40% misclassification error [1,2]. Studies were excluded if the index test on predictors of IPN was fine needle aspiration and gas bubbles on computed tomography, as well as if sensitivity and specificity of the predictor were not reported, or the original paper provided no sufficient data to derive such parameters. The strict study eligibility criterion is one of the important features of this literature review. 2.3. Data acquisition The number of patients, study design, clinical setting, study population, prevalence of POF or IPN and their definitions with
2.4. Quality assessment Study quality was assessed using the QUADAS tool for assessing diagnostic studies [14]. The parameters evaluated for inclusion were: the study was judged to be of a representative spectrum; the study satisfied the defined selection criteria; the time between standard and reference tests was acceptable; the population received the reference standard regardless of index test result; the index test was independent of the reference standard; there was a sufficient description of the index test and a sufficient description of the reference standard; the interpretation of the reference standard was conducted while blinded to the outcome of the index test; the authors reported on the interpretable results and explained any withdrawals from the studies. For the POF section, the inclusion criteria were the same as above, except that the question related to the availability of clinical data when interpreting the index test results was omitted in this analysis, as that information is not applicable to testing for organ failure. For the IPN section, the question related to the verification using reference standard was omitted, as it would be unethical to perform a fine needle aspiration in patients without symptoms. This gave a maximum score of 11 points for predictors of POF and 12 points for predictors of IPN.
2.5. Statistical analysis The data for predictors of POF and IPN were analysed separately according to 2 commonly used clinical assessment periods after hospital admission for acute pancreatitis (10, conclusive increase in the likelihood of study endpoint [17]. Sensitivity was defined as the proportion of patients with POF or IPN that had a positive result on the index test. Specificity was defined as the proportion of patients without POF or IPN who had a negative index test result. The positive likelihood ratio is the ratio of the true positive rate to the false positive rate. The negative likelihood ratio is the ratio of the false negative rate to the true negative rate. DOR measures the effectiveness of a diagnostic test: it is the ratio of the odds of the test being positive in the presence of the disease, relative to the odds of the test being positive in the absence of the disease. This study met the criteria for waiver of ethical approval. No consent was required from patients involved in the reviewed studies.
Please cite this article in press as: Yang CJ, et al. Predictors of severe and critical acute pancreatitis: A systematic review. Dig Liver Dis (2014), http://dx.doi.org/10.1016/j.dld.2014.01.158
G Model YDLD-2572; No. of Pages 6
ARTICLE IN PRESS C.J. Yang et al. / Digestive and Liver Disease xxx (2014) xxx–xxx
3
Fig. 1. PRISMA diagram of literature searches for predictors of persistent organ failure and infected pancreatic necrosis.
3. Results
predictors: Ranson’s, BISAP, Japanese Severity Scale, Pancreatitis Outcome Prediction (POP), Systemic Inflammation Response Syndrome (SIRS), and blood urea nitrogen (p < 0.05). However, there was no significant difference between the sensitivity of APACHE II and those of creatinine, Glasgow, Harmless Acute Pancreatitis Score (HAPS), and Panc3. The Japanese Severity Scale with a cut-off of 1 had the best specificity (0.90) and this was significantly better than the specificities of all the other studied predictors (p < 0.05). The Japanese Severity Scale also had the best positive likelihood ratio (5.19) and this was significantly better than the positive likelihood ratios of creatinine, HAPS, and Panc3 (p < 0.05) but there was no significant difference between the positive likelihood ratio of the Japanese Severity Scale and that of APACHE II, BISAP, Ranson’s, Glasgow, POP, SIRS, and blood urea nitrogen. POP with a cut-off of 9 had the best negative likelihood ratio (0.61), and this was significantly better than the negative likelihood ratio of APACHE II (p < 0.05), while there was no significant difference between POP and other predictors. APACHE II with a cut-off of 7 had the best DOR (13.71) and this was significantly better than the DOR of HAPS and Panc3 (p < 0.05) but there was no significant difference between the DOR of APACHE II and other predictors. Among predictors evaluated at or after 48 h of admission, the accuracy of the Ranson’s score was the most frequently evaluated (in 3 cohorts). The remaining 11 predictors were evaluated in only 1 or 2 cohorts. The data on a total of 11 predictors were suitable for meta-analysis (Supplementary Table S1). When comparing predictors after 48 h of admission, sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and DOR were considered separately. APACHE II with a cut off of 7 had the best sensitivity (0.84). This was significantly better than the specificities of Ranson’s
3.1. Predictors of POF A total of 526 studies were assessed for eligibility and 7 studies examining a total of 15 predictors were included in the final review. The flow chart of the selection process is outlined in Fig. 1 (PRISMA diagram). Of the 15 predictors included, 11 were evaluated in more than 1 study. The methodological quality of the included studies was excellent, as 6 of 7 studies reached the maximum score of 11 points (Table 1). Table 1 lists the characteristics of the studies examining predictors of POF. Overall, the studies enrolled a total of 1589 patients (on average, 199 patients per study) with the incidence of POF calculated as 17% (273 of 1498 reported). Timing of index test was
