- Email: [email protected]
The effect of gastrointestinal bleeding on outcomes of patients with acute pancreatitis: A national population-based study
Pancreatology 12 (2012) 331e336
Contents lists available at SciVerse ScienceDirect
Pancreatology journal homepage: www.elsevier.com/locate/pan
Original article
The effect of gastrointestinal bleeding on outcomes of patients with acute pancreatitis: A national population-based study Hsiu-Nien Shen a, *, Chin-Li Lu b, c, Chung-Yi Li b a
Department of Intensive Care Medicine, Chi Mei Medical Center, No. 901 Chung-Hwa Road, Yong-Kang Dist., Tainan city, Taiwan Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan c Department of Medical Research, Chi Mei Medical Center, Tainan, Taiwan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 March 2012 Received in revised form 24 June 2012 Accepted 1 July 2012
Objectives: To investigate the adverse effect of gastrointestinal bleeding (GIB) in patients with acute pancreatitis (AP), accounting for the status of organ failure (OF). Methods: We analyzed 107,349 patients with first-attack AP from the Taiwan National Health Insurance Research Database between 2000 and 2009. Patients were categorized into four groups according to the status of GIB and OF, the effect of which was assessed using multivariable analyses with generalized estimating equations models. Primary outcomes were 14-day and hospital mortality. Secondary outcomes were septic complication and prolonged hospital stay (>18 days). Results: The covariate-adjusted odds ratio for 14-day mortality, hospital mortality, septic complication, and prolonged stay all significantly increased at 4.63 (95% confidence interval [CI] 3.80e5.63), 4.22 (95% CI 3.66e4.87), 3.52 (95% CI 3.03e4.08), and 1.27 (95% CI 1.20e1.35), respectively for the patients with OF only (n ¼ 88,561). The corresponding figures for the patients with GIB only (n ¼ 5184) were lower but still significant at 1.44 (95% CI 1.09e1.91), 1.42 (95% CI 1.15e1.75), 1.54 (95% CI 1.19e2.00), and 1.38 (95% CI 1.28e1.48). The co-existence of GIB in patients with OF (n ¼ 1663) showed little additional risk of all adverse outcomes. Results of sensitivity analyses (enrolling only patients with principal diagnosis of AP) showed similar findings except that septic complication was not seen for GIB only. Conclusions: OF poses greater adverse effects than GIB on outcomes of AP patients. Nevertheless, GIB still modestly increased the risks of prolonged stay and death in AP patients without OF. Copyright Ó 2012, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Keywords: Acute pancreatitis Gastrointestinal bleeding Organ failure Outcome
1. Introduction The clinical features of acute pancreatitis (AP) range from mild self-limited symptoms to a severe complicated course, leading to organ failure (OF) and death [1,2]. While the case-fatality rate has been decreasing over the decades [3,4], severe cases still carry a high mortality (20%e50%) and consume nearly half of the resources and costs incurred by all patients with AP [5,6]. Accordingly, many efforts have been made to identify correlates of disease severity and predictors for mortality in patients with AP [7e9]. Severe AP has been defined by the presence of several clinical parameters, including OF, local complication, Ranson score 3 and acute physiology and chronic health evaluation [APACHE] II score 8 [1]. However, among subsequent guidelines on AP, discrepancies exist with respect to the inclusion and definition of the
* Corresponding author. Tel.: þ886 6 2812811x57106; fax: þ886 6 2828928. E-mail address: [email protected] (H.-N. Shen).
parameters, especially OF and gastrointestinal bleeding (GIB) [10,11]. According to the practice guidelines published by the American College of Gastroenterology [11], GIB, which is included in the Atlanta classification [1], “will undoubtedly” be deleted in the future revision because it “is rarely encountered in AP”. We argue that this suggestion may not be appropriate and need to be reconsidered for two reasons. First, GIB, albeit relatively infrequent, is not rare; it occurs in 0%e6.4% of all AP [3,12e16] and in 18.0%e 28.6% of severe cases [3,5,16e19]. Second, GIB is not innocent; it has been recently shown to be associated with an increased risk of death in patients with severe AP [19], although the effect is relatively small as compared to that resulting from other OFs. The effect of GIB on the risk of death in patients with AP has not been specifically examined. Earlier studies that assessed the potential adverse effects of GIB were limited because of certain problems in the design and analysis [5,12,14e18,20]. First, most studies were conducted in single centers, had relatively small sample sizes [5,12,18], were limited to a specific subgroup [18,20] or included other non-bleeding gastrointestinal complications [5,17].
1424-3903/$ e see front matter Copyright Ó 2012, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved. http://dx.doi.org/10.1016/j.pan.2012.07.012
332
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
Second, these studies usually included only severe cases who usually simultaneously suffered from both OF and GIB and failed to examine the adverse effect of each of the two conditions separately or jointly [5,16,18,20]. In other words, the effect of GIB on outcomes may vary according to the existence of OF. To overcome these limitations, studies with a sufficient number of study subjects and stratified analysis are needed. Therefore, we re-analyzed data from a recent population-based study [3] to investigate the effects of GIB and OF separately and jointly on the risk of death and selected adverse outcomes in patients with first-attack AP. 2. Methods 2.1. Database The database, National Health Insurance Research Database (NHIRD) [21], and the patients were described in a recent study on the epidemiology of first-attack AP in Taiwan from 2000 through 2009 [3]. The National Health Insurance System in Taiwan is compulsory and covers all citizens except prisoners. The NHIRD was released for research purposes by the National Health Research Institute [22] and is one of the largest and most comprehensive databases in the world. Information included in the inpatient database incorporated sex, date of birth, encrypted patient identification numbers, residential or work area, dates of admission and discharge, medical institutions providing the services, the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes of diagnoses (up to five) and procedures (up to five), outcome at hospital discharge (recovered, died or transferred out), order codes and hospital charges. The approval from the human subjects institutional review board and informed consent from the patients were not needed because the study used an encrypted administrative database.
[3,6], ICU admission, local complications, cholecystectomy [3,6], resource use (including total parenteral nutrition [TPN], hemodialysis and use of vasopressors and mechanical ventilation [MV] and hospital length of stay [LOS]). The Charlson comorbidity index is a weighted summary measure of clinically important concomitant diseases that has been adapted for use with ICD-9-CM coded administrative databases [24,25]. 2.4. Outcomes Primary outcomes were 14-day and hospital mortality [3]. Secondary outcomes included septic complication and prolonged LOS. Septic complication, defined as bacteremia or septicemia (ICD9-CM codes 038, 790.7, 117.9, 112.5, 112.81) [26], was included because patients with GIB would be more likely to receive transfusion which is associated with an increased risk of infection [27,28]. Prolonged LOS was defined as a hospital stay more than 18 days, which was the 90th percentile of all patients with AP [29]. 2.5. Statistics
AP was defined by ICD-9-CM code 577.0 in any position of the five diagnoses. We validated the diagnostic code previously, showing a positive predictive value of 90.0% (95% confidence interval [CI], 79.2%e96.2%) [3]. Severity criteria of AP was defined primarily according to the Atlanta classification scheme [1], but modified by the presence of intensive care unit (ICU) admission (as a surrogate of APACHE II score 8), OF (see Appendix 1), GIB (see Appendix 1) or the requirement of drainage for pancreatic abscesses or cysts (defined by order codes for the procedure) [3,6]. The process of patient enrollment was described previously [3]. Briefly, between 1996 and 2009 (inclusive), we identified 241,699 hospitalizations having a discharge diagnosis of AP from the NHIRD [3]. To ensure the inclusion of the first-attack AP, we excluded 99,438 readmissions and 33,547 patients who had ever been hospitalized for AP in 1996e1999. Patients with a diagnosis of chronic pancreatitis (ICD-9-CM code 577.1) (n ¼ 1315) or missing data on sex (n ¼ 50) were also excluded. The remaining patients (n ¼ 107,349) admitted for the first time between 2000 and 2009 were included in the analysis. Among them, 6847 (6.4%) had GIB and 13,604 (12.7%) had OF. The frequencies of the ICD-9-CM codes for GIB are shown in Appendix 2.
We hypothesized that GIB is associated with an increased risk of the adverse outcomes and that the extent of which is related to the co-existence of OF. First, the patients were categorized into four groups according to the status of GIB and OF, that is without GIB and OF, GIB only, OF only, and with both GIB and OF. Continuous variables were presented as median (inter-quartile range) owing to a skewed distribution; discrete ones as count or percentage. Formal statistical tests among groups were considered unnecessary because of a very large sample size from the entire population. Then, the respective effect of GIB and OF was analyzed using logistic Generalized Estimating Equations methods to account for a hospital cluster effect [30]. Outcomes were regressed with a logistic link function. Both univariable and multivariable analyses were performed to yield the crude and adjusted risk estimates, referenced to the group of patients without GIB and OF. We performed two consecutive multivariable analyses to assess the influences of baseline characteristics (model 1) and the provision of definitive and life-support treatments (model 2). In model 1, the baseline covariates used for adjustment included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of AP (categorized into biliary, alcohol-related, both or others) and another two severity criteria (i.e., ICU admission and local complication, as independent binary variables) (model 1). In model 2, additional treatment covariates included cholecystectomy, TPN, vasopressors, hemodialysis and MV. As we simultaneously included a number of sociodemographic and clinical conditions in the multivariable regression models, a potential numerical problem concerned collinearity between covariates, rendering estimated regression coefficients invalid. We assessed such potential problem by examining the estimated slope coefficients and standard errors of the mean, and found no indication of collinearity. Finally, we did a sensitivity analysis by enrolling only patients with a principal diagnosis of AP [3] to assess the robustness of the estimates. Data were analyzed with SPSS for Windows, version 17.0. (SPSS Inc., Illinois, U.S.A.). A two-tailed p value of <0.05 was considered significant.
2.3. Measurements
3. Results
Patient characteristics included age, sex, year of admission, urbanization (including urban, suburban and rural area) [3,23], hospital level (including medical center [>500 beds], regional [250e500 beds] and district hospitals [20e249 beds]), Charlson Comorbidity Index [24,25], causes of AP (biliary or alcohol-related)
3.1. Patient characteristics
2.2. Definitions and patients
Table 1 shows characteristics of the 107,349 patients with firstattack AP according to the status of GIB and OF. In patients without OF, those with GIB were older, had a higher proportion of males,
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336 Table 1 Characteristics of the patients with first-attack acute pancreatitis according to the status of gastrointestinal bleeding (GIB) and organ failure (OF). Variables
GIB () OF () (n ¼ 88,561)
Hospital level, % Medical center 28.2 Regional hospital 46.2 District hospital 25.5 Patient characteristics Median age 51 (IQR), yr (39e68) Male sex, % 64.3 Urbanization, % Urban 53.4 Suburban 34.2 Rural 12.4 Causes, % Biliary stone 27.0 Alcohol abuse 4.6 Charlson comorbid index, % 0 41.2 1 37.4 2 14.1 3 7.3 Peptic ulcer, % 17.8 Additional severity criteria, % Intensive care 5.2 unit admission Local 1.0 complications a 6.2 Cholecystectomy , % Resource use Total parenteral 2.1 nutrition, % Vasopressor, % 1.4 Hemodialysis, % 2.0 Mechanical 1.3 ventilation, % Median hospital 6 (4e9) LOS (IQR), d
GIB (þ) OF () (n ¼ 5184)
GIB () OF (þ) (n ¼ 11,941)
GIB (þ) OF (þ) (n ¼ 1663)
15.9 43.9 40.2
32.9 43.1 24.0
25.0 34.8 40.2
53 (40e71) 72.1
59 (43e74) 63.8
59 (43e74) 73.7
48.3 37.2 14.5
52.1 33.9 14.0
42.6 40.4 17.0
15.2 3.1
20.7 4.1
9.1 3.1
13.6 46.3 26.1 14.0 69.0
26.4 38.6 18.4 16.6 9.0
13.3 31.0 36.0 19.7 64.5
9.2
50.5
53.4
1.5
5.4
6.0
2.7
6.9
3.8
2.9
13.1
15.2
2.8 3.2 2.9
35.8 11.8 42.8
40.3 11.9 46.3
10 (5e21)
10 (4e25)
7 (4e11)
IQR: interquartile range; LOS: length of stay. a The surgery was performed during admission for first-attack acute pancreatitis.
were more likely to be treated in district hospitals and to live in suburban or rural areas, had fewer biliary causes, had more comorbidities (especially peptic ulcer) and had more other severe criteria, including ICU admission and local complications. In addition, they were less likely to undergo cholecystectomy during hospitalization, but were more likely to receive life-support treatments, including TPN, vasopressors, hemodialysis and MV. The hospital stay was also slightly longer in patients with GIB than those without such condition. When OF existed, similar differences were observed in the aforementioned characteristics (except age, hemodialysis and hospital LOS) between patients with and without GIB. Irrespective of the GIB status, patients with OF were older, received more intensive care and had more local complications and resource uses. 3.2. Outcomes Among patients with GIB only (n ¼ 5184, 4.8%), 1.0% died in 14 days and 1.9% died at hospital discharge (Table 2). Compared to the control group, patients with GIB only also had more septic complications and prolonged hospital stays (Table 3). After adjusting for covariates in model 1, patients with GIB only had a 73% and 63% higher risk of death in 14 days and at hospital discharge, respectively, as compared to those free from GIB and OF (Table 2). The corresponding figures for risks of septic complication
333
and prolonged hospital stay were also significant at 53% and 62% (Table 3). Risks of these adverse outcomes slightly attenuated, but remained significant, after adjusting for additional covariates (model 2 in Tables 2 and 3). With the exception of prolonged hospital stay, the magnitude of the adverse effects associated with GIB was much lower than those associated with OF (Tables 2 and 3). The covariates-adjusted odds ratios (ORs) for the patients with OF only ranged from 1.27 for prolonged hospital stay to 4.63 for 14-day mortality. Besides, when OF was present, the co-existence of GIB added little to the risks of selected adverse outcomes. 3.3. Sensitivity analysis A total of 69,280 patients (64.5%) with a principal diagnosis of AP were enrolled in the sensitivity analysis. The covariate-adjusted ORs and 95% confidence interval (CI) for the patients with GIB only were 1.43 (95% CI 0.77e2.65) for 14-day mortality, 1.36 (95% CI 0.86e2.15) for hospital mortality, and 1.77 (95% CI 1.53e2.06) for prolonged hospital stay. The estimate for septic complication was not assessed because there was no such outcome in patients with GIB only. 4. Discussion In this study, we found that OF posed greater adverse effects than GIB in patients with AP. Despite that, the occurrence of GIB, in the absence of OF, was still significantly associated with an increased risk of adverse outcomes, including prolonged stay and death, suggesting a need of re-consideration to retain GIB as a criteria in classifying severity in AP at least in those without OF. Many prior studies failed to detect the significant adverse effect of GIB in patients with AP mainly because of a limited sample size. Compared to the rates of organ failure, the rate of GIB in patients with AP was in a range from 0% to 6.4% [12e16], and was considered low. Our data showed that 6.4% of AP patients were suffering from GIB and the OR of selected adverse effects associated with GIB was also considered small at only 1.5. With a sufficiently large sample, our study was able to perform detailed stratified analyses that examined the adverse effects posed by OF and GIB separately without compromising the statistical power. Removal of a clinically significant GIB (>500 ml/24 h) from the severity criteria of AP [11] is not well-matched to our experiences in daily clinical practice. Although the adverse effect of GIB on the risk of septic complication was not seen in the sensitive analysis, the study findings clearly demonstrated that GIB alone may pose a greater risk of prolonged hospital stay and death in AP patients. A recent study showed a similar finding, suggesting a higher risk of death in patients with severe AP accompanied by GIB (odds ratio 10, 95% CI 4e28); despite that, the effect size is relatively small as compared to those contributed by other OFs [19]. The results of this study have two important clinical implications. First, because the occurrence of GIB was associated with moderately increased risks of adverse outcomes, it should not be entirely deleted from the severity criteria of the future revision of the Atlanta classification, as suggested by a recent guideline on AP [11]. In the absence of OF, the hospital mortality in our AP patients with GIB is close to that found in a recent study on severe AP excluding OF (1.9% vs. 2.0%) [19]. Both are also associated with a longer hospital stay. Therefore, patients with GIB but without OF should be classified as having moderate AP [12,19,31], which has been recently suggested to include “those with local (peri)pancreatic complications but no persistent systemic complications [31]”. The other implication is that it may be interesting for future studies to investigate whether prophylaxis of GIB in patients with AP can reduce the adverse outcomes. Two recent studies in Asian
334
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
Table 2 Effects of gastrointestinal bleeding (GIB) with and without organ failure (OF) on the risks of 14-day and hospital mortality in patients with first-attack acute pancreatitis (Total n ¼ 107,349). Outcomes
Predictors GIB
No. (%) with outcome
Odds ratio (95% confidence interval) Crudea
OF
Adjusteda,b Model 1
Model 2
14-day mortality No Yes No Yes
No No Yes Yes
397 51 1789 270
(0.4) (1.0) (15.0) (16.2)
1.00 2.28 (1.73e3.01) 38.54 (32.99e45.03) 46.33 (32.61e65.82)
1.00 1.73 (1.33e2.27) 16.19 (13.41e19.54) 15.86 (11.67e21.55)
1.00 1.44 (1.09e1.91) 4.63 (3.80e5.63) 4.18 (3.15e5.56)
No Yes No Yes
No No Yes Yes
679 96 2879 463
(0.8) (1.9) (24.1) (27.8)
1.00 3.00 (2.35e3.84) 46.99 (40.21e54.92) 66.28 (47.77e91.98)
1.00 1.63 (1.36e1.95) 12.16 (10.77e13.72) 13.81 (11.02e17.29)
1.00 1.42 (1.15e1.75) 4.22 (3.66e4.87) 4.29 (3.40e5.41)
Hospital mortality
a
Multivariable logistic regression models with considering the cluster effect of hospitals using Generalized Estimating Equations models. Covariates included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of acute pancreatitis (categorized into biliary, alcohol-related, both and others), intensive care unit admission and local complication (model 1). In model 2, additional covariates included cholecystectomy, total parenteral nutrition, vasopressors, hemodialysis and mechanical ventilation. b
populations showed that acute gastrointestinal mucosal lesions (AGML) and peptic ulcer disease are common in patients with AP, occurring in 65% and 52.6%, respectively [32,33]. We also found that peptic ulcer was present in nearly 70% of AP patients with GIB, suggesting that there might be a close link among them. Besides, although the presence of AGML is not associated with severity of AP [32], the occurrence of peptic ulcer can be predicted by a severe attack defined by an APACHE II score 6 [33]. Therefore, after excluding the relatively uncommon sources of serious bleeding from varices or pseudoaneurysms, patients who have a severe attack and are at risk of peptic ulcer disease and GIB might be potential candidates for the bleeding prophylaxis, which has not been addressed by current guidelines [11,34]. Some limitations deserve comments. First, the definitions of AP, OF and GIB in this study rely solely on the coding instead of the clinical criteria. Although we had validated the code for AP, information on the causes and severity of GIB is not available in the database. Therefore, the impact of various causes of GIB, such as peptic ulcer disease, stress induced ulceration, varices and visceral artery pseudoaneurysms, could not be assessed. Further studies are needed. Although some patients with less severe bleeding (500 ml/24 h) might be included, the observed risks of the adverse outcomes would be more likely to be underestimated by the inclusion of these patients. Besides, a hemoconcentration-
dilution scenario, that is, early hemoconcentration followed by fluid resuscitation for AP, may cause a marked drop in hematocrit, leading to an “over-diagnosis” (or misclassification) of GIB. This potential misclassification is likely because we found that patients with GIB were more likely to be treated in district hospitals and to live in suburban or rural areas. However, hemocencentration has been shown to be a poor predictor of a worse outcome in patients with AP [35e37], which implies that the hemoconcentrationdilution scenario alone (i.e., without associated OF) is less likely to be associated with an increased risk of adverse outcomes. Therefore, the misclassification of GIB due to the scenario should not be a valid argument for the increased risk of adverse outcomes in AP patients with GIB-only since it will tend to under-estimate rather than over-estimate the observed effect of GIB. Moreover, the misclassification of GIB due to the scenario may also help to explain the lack of additional risk posed by GIB in patients who already had OF. Second, the low yield in retrieving the causes may limit the adjustment in biliary and alcohol-related AP. However, the extent of the residual confounding, albeit unknown, seems not likely to change the conclusion. Third, we were unable to know which patients had necrotizing or interstitial pancreatitis. Given that the overall prevalence of necrotizing pancreatitis is less than 10%, it may be possible that most GIB (overall incidence 6.4% in this study) was seen in patients with necrotizing pancreatitis. Besides,
Table 3 Effects of gastrointestinal bleeding (GIB) with and without organ failure (OF) on the risks of septic complication and prolonged hospital stay in patients with first-attack acute pancreatitis (Total n ¼ 107,349). Outcomes
Predictors GIB
No. (%) with outcome
Odds ratio (95% confidence interval) Crudea
OF
Adjusteda,b Model 1
Model 2
Septic complication No Yes No Yes
No No Yes Yes
707 68 746 71
(0.8) (1.3) (6.2) (4.3)
1.00 1.52 (1.10e2.09) 8.28 (7.07e9.71) 5.69 (3.78e8.57)
1.00 1.53 (1.18e1.98) 3.80 (3.21e4.35) 2.60 (1.98e3.41)
1.00 1.54 (1.19e2.00) 3.52 (3.03e4.08) 2.41 (1.82e3.20)
No Yes No Yes
No No Yes Yes
5713 572 3532 530
(6.5) (11.0) (29.6) (31.9)
1.00 2.44 (2.10e2.85) 7.11 (6.23e8.11) 9.71 (7.01e13.46)
1.00 1.62 (1.46e1.80) 1.89 (1.75e2.05) 2.17 (1.77e2.65)
1.00 1.38 (1.28e1.48) 1.27 (1.20e1.35) 1.35 (1.15e1.60)
Prolonged stayc
a Multivariable logistic regression models with considering the cluster effect of hospitals using Generalized Estimating Equations (GEE) models. However, the estimates for the risk of septic complication were made only by multivariable logistic regressions because of non-convergence in GEE models. b Covariates included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of acute pancreatitis (categorized into biliary, alcohol-related, both and others), intensive care unit admission and local complication (model 1). In model 2, additional covariates included cholecystectomy, total parenteral nutrition, vasopressors, hemodialysis and mechanical ventilation. c Prolonged stay was defined as a hospital stay more than 18 days, which was the 90th percentile of all patients with acute pancreatitis.
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
the septic complications in AP patients with GIB may be due to dysfunction of gut barrier caused by pancreatitis instead of transfusions, especially in those with severe AP. Nevertheless, this study is strengthened by a large number of patients from a nationwide database, which can provide an unbiased selection and enhance its generalizability. In conclusion, the results of this study suggest that GIB still accounts for an appreciable number of patients with AP and may significantly predict various adverse outcomes associated with AP even in the absence of OF. Further studies are needed to see whether prophylaxis of GIB, after excluding variceal or pseudoaneurysmal bleeding, in high risk patients can reduce the associated adverse outcomes. Acknowledgments The study was performed in Chi Mei Medical Center. HNS received a research grant CMFHR10025 from the hospital. CLL and CYL have no conflicts of interest or funding to disclose. The study was based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health or National Health Research Institutes.
335
Appendix 2. Distribution of various ICD-9-CM codes for gastrointestinal (GI) bleeding in patients with acute pancreatitis (AP).
ICD-9-CM code
Diagnosis
531.x (exclude 531.3, 531.7, 531.9) 532.x (exclude 532.3, 532.7, 532.9) 533.x (exclude 533.3, 533.7, 533.9) 534.x (exclude 534.3, 534.7, 534.9) 535.x1
Gastric ulcer with bleeding
31.6
28.4
Duodenal ulcer with bleeding
19.4
15.3
Peptic ulcer with bleeding
8.2
7.3
Gastrojejunal ulcer with bleeding
1.0
1.0
Gastritis/duodenitis with bleeding Hematemesis Blood in stool Unspecified GI bleeding e
11.4
16.8
0.3 0.5 27.6
0.3 0.4 30.5
100.0
100.0
578.0 578.1 578.9 Total
All patients with AP (n ¼ 6847), %
Patients with principal diagnosis of AP (n ¼ 2985), %
ICD-9-CM: International Classification of Diseases, Ninth Revision, Clinical Modification.
Appendix 1. Organ failure and gastrointestinal bleeding codes.
Organ failure
Codes
Code description
Cardiovascular
458.0 458.8 458.9 785.5 785.51 785.59 796.3 518.81 518.82 518.85 786.09 799.1 96.7 (96.71, 96.72) 96.04 93.90 584.x 570 572.2 573.3 573.4 293 348.1 348.3 780.01 780.09 89.14 286.2 286.6 286.9 287.3e5 790.92 276.2 531.x (exclude 531.3, 532.x (exclude 532.3, 533.x (exclude 533.3, 534.x (exclude 534.3, 535.x1 578.0, 578.1, 578.9
Hypotension, postural Hypotension, specified type, not elsewhere classified Hypotension, arterial, constitutional Shock Shock, cardiogenic Shock, circulatory or septic Hypotension, transient Acute respiratory failure Acute respiratory distress syndrome (ARDS) ARDS after shock or trauma Respiratory insufficiency Respiratory arrest Ventilator management Endotracheal intubation (emergency procedure) Continuous positive airway pressure Acute renal failure Acute hepatic failure or necrosis Hepatic encephalopathy Hepatitis (septic & not elsewhere classified) Hepatic infarction Transient organic psychosis Anoxic brain injury Encephalopathy, acute Coma Altered consciousness, unspecified Electroencephalography Disseminated intravascular coagulation Purpura fulminans Coagulopathy Thrombocytopenia, primary, secondary or unspecified Abnormal coagulation profile Acidosis, metabolic or lactic Gastric ulcer with bleeding Duodenal ulcer with bleeding Peptic ulcer with bleeding Gastrojejunal ulcer with bleeding Gastritis/duodenitis with bleeding Non-specific gastrointestinal bleeding
Respiratory
Renal Hepatic
Neurologic
Hematologic
Metabolic Gastrointestinal bleeding
531.7, 532.7, 533.7, 534.7,
531.9) 532.9) 533.9) 534.9)
336
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
References [1] Bradley 3rd EL. A clinically based classification system for acute pancreatitis. Summary of the international symposium on acute pancreatitis, Atlanta, Ga, September 11 through 13, 1992. Arch Surg 1993;128:586e90. [2] Acute Pancreatitis Classification Working Group. Revision of the Atlanta classification of acute pancreatitis (3rd revision). Available at: www. pancreasclub.com/resources/AtlantaClassification.pdf [accessed 20.02.12]. [3] Shen HN, Lu CL, Li CY. Epidemiology of first-attack acute pancreatitis in Taiwan from 2000 through 2009: a nationwide population-based study. Pancreas 2012;41:696e702. [4] Yadav D, Lowenfels AB. Trends in the epidemiology of the first attack of acute pancreatitis: a systematic review. Pancreas 2006;33:323e30. [5] Flint R, Windsor J, Bonham M. Trends in the management of severe acute pancreatitis: interventions and outcome. ANZ J Surg 2004;74:335e42. [6] Shen HN, Lu CL. Incidence, resource use, and outcome of acute pancreatitis with/without intensive care: a nationwide population-based study in Taiwan. Pancreas 2011;40:10e5. [7] Papachristou GI, Muddana V, Yadav D, O’Connell M, Sanders MK, Slivka A, et al. Comparison of BISAP, Ranson’s, APACHE-II, and CTSI scores in predicting organ failure, complications, and mortality in acute pancreatitis. Am J Gastroenterol 2010;105:435e41. [8] Gravante G, Garcea G, Ong SL, Metcalfe MS, Berry DP, Lloyd DM, et al. Prediction of mortality in acute pancreatitis: a systematic review of the published evidence. Pancreatology 2009;9:601e14. [9] Petrov MS, Shanbhag S, Chakraborty M, Phillips ARJ, Windsor JA. Organ failure and infection of pancreatic necrosis as determinants of mortality in patients with acute pancreatitis. Gastroenterology 2010;139:813e20. [10] Bollen T, Van Santvoort H, Besselink M, Van Leeuwen M, Horvath K, Freeny P, et al. The Atlanta classification of acute pancreatitis revisited. Br J Surg 2008; 95:6e21. [11] Banks PA, Freeman ML. Practice guidelines in acute pancreatitis. Am J Gastroenterol 2006;101:2379e400. [12] De-Madaria E, Soler G, Martinez J, Gomez-Escolar L, Sanchez-Fortun C, Sempere L, et al. Update of the Atlanta classification of severity of acute pancreatitis: should a moderate category be included? Pancreatology 2009;9:433e4. [13] Lankisch PG, Pflichthofer D, Lehnick D. No strict correlation between necrosis and organ failure in acute pancreatitis. Pancreas 2000;20:319e22. [14] Sharma M, Banerjee D, Garg PK. Characterization of newer subgroups of fulminant and subfulminant pancreatitis associated with a high early mortality. Am J Gastroenterol 2007;102:2688e95. [15] Sharma PK, Madan K, Garg PK. Hemorrhage in acute pancreatitis: should gastrointestinal bleeding be considered an organ failure? Pancreas 2008;36:141e5. [16] Company L, Saez J, Martinez J. Factors predicting mortality in severe acute pancreatitis. Pancreatology 2003;3:144e8. [17] Mole DJ, Olabi B, Robinson V, Garden OJ, Parks RW. Incidence of individual organ dysfunction in fatal acute pancreatitis: analysis of 1024 death records. HPB 2009;11:166e70. [18] Gardner TB, Vege SS, Chari ST, Pearson RK, Clain JE, Topazian MD, et al. The effect of age on hospital outcomes in severe acute pancreatitis. Pancreatology 2008;8:265e70.
[19] Vege SS, Gardner TB, Chari ST, Munukuti P, Pearson RK, Clain JE, et al. Low mortality and high morbidity in severe acute pancreatitis without organ failure: a case for revising the Atlanta classification to include “moderately severe acute pancreatitis”. Am J Gastroenterol 2009;104:710e5. [20] Xin MJ, Chen H, Luo B, Sun JB. Severe acute pancreatitis in the elderly: etiology and clinical characteristics. World J Gastroenterol 2008;14:2517e21. [21] National Health Insurance Research Database. Available at: http://w3.nhri.org. tw/nhird//en/index.htm [accessed 20.02.12]. [22] Bureau of National Health Insurance, Department of Health, Executive Yuan. Available at: http://www.nhi.gov.tw/webdata/webdata.aspx?menu¼17&menu_ id¼661&WD_ID¼689&webdata_id¼3351 [accessed 20.02.12]. [23] Liu C, Hung Y, Chuang Y, Chen Y, Weng W, Liu J, et al. Incorporating development stratification of Taiwan townships into sampling design of large scale health interview survey. J Health Manag 2006;4:1e22. [24] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373e83. [25] Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613e9. [26] Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med 2003;348:1546e54. [27] Sarani B, Dunkman WJ. Transfusion of fresh frozen plasma in critically ill surgical patients is associated with an increased risk of infection. Crit Care Med 2008;36:1114e8. [28] Claridge JA, Sawyer RG, Schulman AM, McLemore EC, Young JS. Blood transfusions correlate with infections in trauma patients in a dose-dependent manner. Am Surg 2002;68:566e72. [29] Singla A, Csikesz NG, Simons JP, Li YF, Ng SC, Tseng JF, et al. National hospital volume in acute pancreatitis: analysis of the nationwide inpatient sample 1998e2006. HPB (Oxford) 2009;11:391e7. [30] Hanley JA, Negassa A. Statistical analysis of correlated data using generalized estimating equations: an orientation. Am J Epidemiol 2003;157:364e75. [31] Petrov MS, Windsor JA. Classification of the severity of acute pancreatitis: how many categories make sense? Am J Gastroenterol 2009;105:74e6. [32] Chen TA, Lo GH, Lin CK, Lai KH, Wong HY, Yu HC, et al. Acute pancreatitisassociated acute gastrointestinal mucosal lesions: incidence, characteristics, and clinical significance. J Clin Gastroenterol 2007;41:630e4. [33] Lee KM, Paik CN, Chung WC, Yang JM. Association between acute pancreatitis and peptic ulcer disease. World J Gastroenterol 2011;17:1058e62. [34] Nathens AB, Curtis JR, Beale RJ, Cook DJ, Moreno RP, Romand J-A, et al. Management of the critically ill patient with severe acute pancreatitis. Crit Care Med 2004;32:2524e36. [35] Lankisch PG, Mahlke R, Blum T, Bruns A, Bruns D, Maisonneuve P, et al. Hemoconcentration: an early marker of severe and/or necrotizing pancreatitis? A critical appraisal. Am J Gastroenterol 2001;96:2081e5. [36] Remes-Troche JM, Duarte-Rojo A, Morales G, Robles-Díaz G. Hemoconcentration is a poor predictor of severity in acute pancreatitis. World J Gastroenterol 2005;11:7018e23. [37] Gardner TB, Olenec CA, Chertoff JD, Mackenzie TA, Robertson DJ. Hemoconcentration and pancreatic necrosis: further defining the relationship. Pancreas 2006;33:169e73.
Contents lists available at SciVerse ScienceDirect
Pancreatology journal homepage: www.elsevier.com/locate/pan
Original article
The effect of gastrointestinal bleeding on outcomes of patients with acute pancreatitis: A national population-based study Hsiu-Nien Shen a, *, Chin-Li Lu b, c, Chung-Yi Li b a
Department of Intensive Care Medicine, Chi Mei Medical Center, No. 901 Chung-Hwa Road, Yong-Kang Dist., Tainan city, Taiwan Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan c Department of Medical Research, Chi Mei Medical Center, Tainan, Taiwan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 March 2012 Received in revised form 24 June 2012 Accepted 1 July 2012
Objectives: To investigate the adverse effect of gastrointestinal bleeding (GIB) in patients with acute pancreatitis (AP), accounting for the status of organ failure (OF). Methods: We analyzed 107,349 patients with first-attack AP from the Taiwan National Health Insurance Research Database between 2000 and 2009. Patients were categorized into four groups according to the status of GIB and OF, the effect of which was assessed using multivariable analyses with generalized estimating equations models. Primary outcomes were 14-day and hospital mortality. Secondary outcomes were septic complication and prolonged hospital stay (>18 days). Results: The covariate-adjusted odds ratio for 14-day mortality, hospital mortality, septic complication, and prolonged stay all significantly increased at 4.63 (95% confidence interval [CI] 3.80e5.63), 4.22 (95% CI 3.66e4.87), 3.52 (95% CI 3.03e4.08), and 1.27 (95% CI 1.20e1.35), respectively for the patients with OF only (n ¼ 88,561). The corresponding figures for the patients with GIB only (n ¼ 5184) were lower but still significant at 1.44 (95% CI 1.09e1.91), 1.42 (95% CI 1.15e1.75), 1.54 (95% CI 1.19e2.00), and 1.38 (95% CI 1.28e1.48). The co-existence of GIB in patients with OF (n ¼ 1663) showed little additional risk of all adverse outcomes. Results of sensitivity analyses (enrolling only patients with principal diagnosis of AP) showed similar findings except that septic complication was not seen for GIB only. Conclusions: OF poses greater adverse effects than GIB on outcomes of AP patients. Nevertheless, GIB still modestly increased the risks of prolonged stay and death in AP patients without OF. Copyright Ó 2012, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Keywords: Acute pancreatitis Gastrointestinal bleeding Organ failure Outcome
1. Introduction The clinical features of acute pancreatitis (AP) range from mild self-limited symptoms to a severe complicated course, leading to organ failure (OF) and death [1,2]. While the case-fatality rate has been decreasing over the decades [3,4], severe cases still carry a high mortality (20%e50%) and consume nearly half of the resources and costs incurred by all patients with AP [5,6]. Accordingly, many efforts have been made to identify correlates of disease severity and predictors for mortality in patients with AP [7e9]. Severe AP has been defined by the presence of several clinical parameters, including OF, local complication, Ranson score 3 and acute physiology and chronic health evaluation [APACHE] II score 8 [1]. However, among subsequent guidelines on AP, discrepancies exist with respect to the inclusion and definition of the
* Corresponding author. Tel.: þ886 6 2812811x57106; fax: þ886 6 2828928. E-mail address: [email protected] (H.-N. Shen).
parameters, especially OF and gastrointestinal bleeding (GIB) [10,11]. According to the practice guidelines published by the American College of Gastroenterology [11], GIB, which is included in the Atlanta classification [1], “will undoubtedly” be deleted in the future revision because it “is rarely encountered in AP”. We argue that this suggestion may not be appropriate and need to be reconsidered for two reasons. First, GIB, albeit relatively infrequent, is not rare; it occurs in 0%e6.4% of all AP [3,12e16] and in 18.0%e 28.6% of severe cases [3,5,16e19]. Second, GIB is not innocent; it has been recently shown to be associated with an increased risk of death in patients with severe AP [19], although the effect is relatively small as compared to that resulting from other OFs. The effect of GIB on the risk of death in patients with AP has not been specifically examined. Earlier studies that assessed the potential adverse effects of GIB were limited because of certain problems in the design and analysis [5,12,14e18,20]. First, most studies were conducted in single centers, had relatively small sample sizes [5,12,18], were limited to a specific subgroup [18,20] or included other non-bleeding gastrointestinal complications [5,17].
1424-3903/$ e see front matter Copyright Ó 2012, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved. http://dx.doi.org/10.1016/j.pan.2012.07.012
332
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
Second, these studies usually included only severe cases who usually simultaneously suffered from both OF and GIB and failed to examine the adverse effect of each of the two conditions separately or jointly [5,16,18,20]. In other words, the effect of GIB on outcomes may vary according to the existence of OF. To overcome these limitations, studies with a sufficient number of study subjects and stratified analysis are needed. Therefore, we re-analyzed data from a recent population-based study [3] to investigate the effects of GIB and OF separately and jointly on the risk of death and selected adverse outcomes in patients with first-attack AP. 2. Methods 2.1. Database The database, National Health Insurance Research Database (NHIRD) [21], and the patients were described in a recent study on the epidemiology of first-attack AP in Taiwan from 2000 through 2009 [3]. The National Health Insurance System in Taiwan is compulsory and covers all citizens except prisoners. The NHIRD was released for research purposes by the National Health Research Institute [22] and is one of the largest and most comprehensive databases in the world. Information included in the inpatient database incorporated sex, date of birth, encrypted patient identification numbers, residential or work area, dates of admission and discharge, medical institutions providing the services, the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes of diagnoses (up to five) and procedures (up to five), outcome at hospital discharge (recovered, died or transferred out), order codes and hospital charges. The approval from the human subjects institutional review board and informed consent from the patients were not needed because the study used an encrypted administrative database.
[3,6], ICU admission, local complications, cholecystectomy [3,6], resource use (including total parenteral nutrition [TPN], hemodialysis and use of vasopressors and mechanical ventilation [MV] and hospital length of stay [LOS]). The Charlson comorbidity index is a weighted summary measure of clinically important concomitant diseases that has been adapted for use with ICD-9-CM coded administrative databases [24,25]. 2.4. Outcomes Primary outcomes were 14-day and hospital mortality [3]. Secondary outcomes included septic complication and prolonged LOS. Septic complication, defined as bacteremia or septicemia (ICD9-CM codes 038, 790.7, 117.9, 112.5, 112.81) [26], was included because patients with GIB would be more likely to receive transfusion which is associated with an increased risk of infection [27,28]. Prolonged LOS was defined as a hospital stay more than 18 days, which was the 90th percentile of all patients with AP [29]. 2.5. Statistics
AP was defined by ICD-9-CM code 577.0 in any position of the five diagnoses. We validated the diagnostic code previously, showing a positive predictive value of 90.0% (95% confidence interval [CI], 79.2%e96.2%) [3]. Severity criteria of AP was defined primarily according to the Atlanta classification scheme [1], but modified by the presence of intensive care unit (ICU) admission (as a surrogate of APACHE II score 8), OF (see Appendix 1), GIB (see Appendix 1) or the requirement of drainage for pancreatic abscesses or cysts (defined by order codes for the procedure) [3,6]. The process of patient enrollment was described previously [3]. Briefly, between 1996 and 2009 (inclusive), we identified 241,699 hospitalizations having a discharge diagnosis of AP from the NHIRD [3]. To ensure the inclusion of the first-attack AP, we excluded 99,438 readmissions and 33,547 patients who had ever been hospitalized for AP in 1996e1999. Patients with a diagnosis of chronic pancreatitis (ICD-9-CM code 577.1) (n ¼ 1315) or missing data on sex (n ¼ 50) were also excluded. The remaining patients (n ¼ 107,349) admitted for the first time between 2000 and 2009 were included in the analysis. Among them, 6847 (6.4%) had GIB and 13,604 (12.7%) had OF. The frequencies of the ICD-9-CM codes for GIB are shown in Appendix 2.
We hypothesized that GIB is associated with an increased risk of the adverse outcomes and that the extent of which is related to the co-existence of OF. First, the patients were categorized into four groups according to the status of GIB and OF, that is without GIB and OF, GIB only, OF only, and with both GIB and OF. Continuous variables were presented as median (inter-quartile range) owing to a skewed distribution; discrete ones as count or percentage. Formal statistical tests among groups were considered unnecessary because of a very large sample size from the entire population. Then, the respective effect of GIB and OF was analyzed using logistic Generalized Estimating Equations methods to account for a hospital cluster effect [30]. Outcomes were regressed with a logistic link function. Both univariable and multivariable analyses were performed to yield the crude and adjusted risk estimates, referenced to the group of patients without GIB and OF. We performed two consecutive multivariable analyses to assess the influences of baseline characteristics (model 1) and the provision of definitive and life-support treatments (model 2). In model 1, the baseline covariates used for adjustment included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of AP (categorized into biliary, alcohol-related, both or others) and another two severity criteria (i.e., ICU admission and local complication, as independent binary variables) (model 1). In model 2, additional treatment covariates included cholecystectomy, TPN, vasopressors, hemodialysis and MV. As we simultaneously included a number of sociodemographic and clinical conditions in the multivariable regression models, a potential numerical problem concerned collinearity between covariates, rendering estimated regression coefficients invalid. We assessed such potential problem by examining the estimated slope coefficients and standard errors of the mean, and found no indication of collinearity. Finally, we did a sensitivity analysis by enrolling only patients with a principal diagnosis of AP [3] to assess the robustness of the estimates. Data were analyzed with SPSS for Windows, version 17.0. (SPSS Inc., Illinois, U.S.A.). A two-tailed p value of <0.05 was considered significant.
2.3. Measurements
3. Results
Patient characteristics included age, sex, year of admission, urbanization (including urban, suburban and rural area) [3,23], hospital level (including medical center [>500 beds], regional [250e500 beds] and district hospitals [20e249 beds]), Charlson Comorbidity Index [24,25], causes of AP (biliary or alcohol-related)
3.1. Patient characteristics
2.2. Definitions and patients
Table 1 shows characteristics of the 107,349 patients with firstattack AP according to the status of GIB and OF. In patients without OF, those with GIB were older, had a higher proportion of males,
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336 Table 1 Characteristics of the patients with first-attack acute pancreatitis according to the status of gastrointestinal bleeding (GIB) and organ failure (OF). Variables
GIB () OF () (n ¼ 88,561)
Hospital level, % Medical center 28.2 Regional hospital 46.2 District hospital 25.5 Patient characteristics Median age 51 (IQR), yr (39e68) Male sex, % 64.3 Urbanization, % Urban 53.4 Suburban 34.2 Rural 12.4 Causes, % Biliary stone 27.0 Alcohol abuse 4.6 Charlson comorbid index, % 0 41.2 1 37.4 2 14.1 3 7.3 Peptic ulcer, % 17.8 Additional severity criteria, % Intensive care 5.2 unit admission Local 1.0 complications a 6.2 Cholecystectomy , % Resource use Total parenteral 2.1 nutrition, % Vasopressor, % 1.4 Hemodialysis, % 2.0 Mechanical 1.3 ventilation, % Median hospital 6 (4e9) LOS (IQR), d
GIB (þ) OF () (n ¼ 5184)
GIB () OF (þ) (n ¼ 11,941)
GIB (þ) OF (þ) (n ¼ 1663)
15.9 43.9 40.2
32.9 43.1 24.0
25.0 34.8 40.2
53 (40e71) 72.1
59 (43e74) 63.8
59 (43e74) 73.7
48.3 37.2 14.5
52.1 33.9 14.0
42.6 40.4 17.0
15.2 3.1
20.7 4.1
9.1 3.1
13.6 46.3 26.1 14.0 69.0
26.4 38.6 18.4 16.6 9.0
13.3 31.0 36.0 19.7 64.5
9.2
50.5
53.4
1.5
5.4
6.0
2.7
6.9
3.8
2.9
13.1
15.2
2.8 3.2 2.9
35.8 11.8 42.8
40.3 11.9 46.3
10 (5e21)
10 (4e25)
7 (4e11)
IQR: interquartile range; LOS: length of stay. a The surgery was performed during admission for first-attack acute pancreatitis.
were more likely to be treated in district hospitals and to live in suburban or rural areas, had fewer biliary causes, had more comorbidities (especially peptic ulcer) and had more other severe criteria, including ICU admission and local complications. In addition, they were less likely to undergo cholecystectomy during hospitalization, but were more likely to receive life-support treatments, including TPN, vasopressors, hemodialysis and MV. The hospital stay was also slightly longer in patients with GIB than those without such condition. When OF existed, similar differences were observed in the aforementioned characteristics (except age, hemodialysis and hospital LOS) between patients with and without GIB. Irrespective of the GIB status, patients with OF were older, received more intensive care and had more local complications and resource uses. 3.2. Outcomes Among patients with GIB only (n ¼ 5184, 4.8%), 1.0% died in 14 days and 1.9% died at hospital discharge (Table 2). Compared to the control group, patients with GIB only also had more septic complications and prolonged hospital stays (Table 3). After adjusting for covariates in model 1, patients with GIB only had a 73% and 63% higher risk of death in 14 days and at hospital discharge, respectively, as compared to those free from GIB and OF (Table 2). The corresponding figures for risks of septic complication
333
and prolonged hospital stay were also significant at 53% and 62% (Table 3). Risks of these adverse outcomes slightly attenuated, but remained significant, after adjusting for additional covariates (model 2 in Tables 2 and 3). With the exception of prolonged hospital stay, the magnitude of the adverse effects associated with GIB was much lower than those associated with OF (Tables 2 and 3). The covariates-adjusted odds ratios (ORs) for the patients with OF only ranged from 1.27 for prolonged hospital stay to 4.63 for 14-day mortality. Besides, when OF was present, the co-existence of GIB added little to the risks of selected adverse outcomes. 3.3. Sensitivity analysis A total of 69,280 patients (64.5%) with a principal diagnosis of AP were enrolled in the sensitivity analysis. The covariate-adjusted ORs and 95% confidence interval (CI) for the patients with GIB only were 1.43 (95% CI 0.77e2.65) for 14-day mortality, 1.36 (95% CI 0.86e2.15) for hospital mortality, and 1.77 (95% CI 1.53e2.06) for prolonged hospital stay. The estimate for septic complication was not assessed because there was no such outcome in patients with GIB only. 4. Discussion In this study, we found that OF posed greater adverse effects than GIB in patients with AP. Despite that, the occurrence of GIB, in the absence of OF, was still significantly associated with an increased risk of adverse outcomes, including prolonged stay and death, suggesting a need of re-consideration to retain GIB as a criteria in classifying severity in AP at least in those without OF. Many prior studies failed to detect the significant adverse effect of GIB in patients with AP mainly because of a limited sample size. Compared to the rates of organ failure, the rate of GIB in patients with AP was in a range from 0% to 6.4% [12e16], and was considered low. Our data showed that 6.4% of AP patients were suffering from GIB and the OR of selected adverse effects associated with GIB was also considered small at only 1.5. With a sufficiently large sample, our study was able to perform detailed stratified analyses that examined the adverse effects posed by OF and GIB separately without compromising the statistical power. Removal of a clinically significant GIB (>500 ml/24 h) from the severity criteria of AP [11] is not well-matched to our experiences in daily clinical practice. Although the adverse effect of GIB on the risk of septic complication was not seen in the sensitive analysis, the study findings clearly demonstrated that GIB alone may pose a greater risk of prolonged hospital stay and death in AP patients. A recent study showed a similar finding, suggesting a higher risk of death in patients with severe AP accompanied by GIB (odds ratio 10, 95% CI 4e28); despite that, the effect size is relatively small as compared to those contributed by other OFs [19]. The results of this study have two important clinical implications. First, because the occurrence of GIB was associated with moderately increased risks of adverse outcomes, it should not be entirely deleted from the severity criteria of the future revision of the Atlanta classification, as suggested by a recent guideline on AP [11]. In the absence of OF, the hospital mortality in our AP patients with GIB is close to that found in a recent study on severe AP excluding OF (1.9% vs. 2.0%) [19]. Both are also associated with a longer hospital stay. Therefore, patients with GIB but without OF should be classified as having moderate AP [12,19,31], which has been recently suggested to include “those with local (peri)pancreatic complications but no persistent systemic complications [31]”. The other implication is that it may be interesting for future studies to investigate whether prophylaxis of GIB in patients with AP can reduce the adverse outcomes. Two recent studies in Asian
334
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
Table 2 Effects of gastrointestinal bleeding (GIB) with and without organ failure (OF) on the risks of 14-day and hospital mortality in patients with first-attack acute pancreatitis (Total n ¼ 107,349). Outcomes
Predictors GIB
No. (%) with outcome
Odds ratio (95% confidence interval) Crudea
OF
Adjusteda,b Model 1
Model 2
14-day mortality No Yes No Yes
No No Yes Yes
397 51 1789 270
(0.4) (1.0) (15.0) (16.2)
1.00 2.28 (1.73e3.01) 38.54 (32.99e45.03) 46.33 (32.61e65.82)
1.00 1.73 (1.33e2.27) 16.19 (13.41e19.54) 15.86 (11.67e21.55)
1.00 1.44 (1.09e1.91) 4.63 (3.80e5.63) 4.18 (3.15e5.56)
No Yes No Yes
No No Yes Yes
679 96 2879 463
(0.8) (1.9) (24.1) (27.8)
1.00 3.00 (2.35e3.84) 46.99 (40.21e54.92) 66.28 (47.77e91.98)
1.00 1.63 (1.36e1.95) 12.16 (10.77e13.72) 13.81 (11.02e17.29)
1.00 1.42 (1.15e1.75) 4.22 (3.66e4.87) 4.29 (3.40e5.41)
Hospital mortality
a
Multivariable logistic regression models with considering the cluster effect of hospitals using Generalized Estimating Equations models. Covariates included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of acute pancreatitis (categorized into biliary, alcohol-related, both and others), intensive care unit admission and local complication (model 1). In model 2, additional covariates included cholecystectomy, total parenteral nutrition, vasopressors, hemodialysis and mechanical ventilation. b
populations showed that acute gastrointestinal mucosal lesions (AGML) and peptic ulcer disease are common in patients with AP, occurring in 65% and 52.6%, respectively [32,33]. We also found that peptic ulcer was present in nearly 70% of AP patients with GIB, suggesting that there might be a close link among them. Besides, although the presence of AGML is not associated with severity of AP [32], the occurrence of peptic ulcer can be predicted by a severe attack defined by an APACHE II score 6 [33]. Therefore, after excluding the relatively uncommon sources of serious bleeding from varices or pseudoaneurysms, patients who have a severe attack and are at risk of peptic ulcer disease and GIB might be potential candidates for the bleeding prophylaxis, which has not been addressed by current guidelines [11,34]. Some limitations deserve comments. First, the definitions of AP, OF and GIB in this study rely solely on the coding instead of the clinical criteria. Although we had validated the code for AP, information on the causes and severity of GIB is not available in the database. Therefore, the impact of various causes of GIB, such as peptic ulcer disease, stress induced ulceration, varices and visceral artery pseudoaneurysms, could not be assessed. Further studies are needed. Although some patients with less severe bleeding (500 ml/24 h) might be included, the observed risks of the adverse outcomes would be more likely to be underestimated by the inclusion of these patients. Besides, a hemoconcentration-
dilution scenario, that is, early hemoconcentration followed by fluid resuscitation for AP, may cause a marked drop in hematocrit, leading to an “over-diagnosis” (or misclassification) of GIB. This potential misclassification is likely because we found that patients with GIB were more likely to be treated in district hospitals and to live in suburban or rural areas. However, hemocencentration has been shown to be a poor predictor of a worse outcome in patients with AP [35e37], which implies that the hemoconcentrationdilution scenario alone (i.e., without associated OF) is less likely to be associated with an increased risk of adverse outcomes. Therefore, the misclassification of GIB due to the scenario should not be a valid argument for the increased risk of adverse outcomes in AP patients with GIB-only since it will tend to under-estimate rather than over-estimate the observed effect of GIB. Moreover, the misclassification of GIB due to the scenario may also help to explain the lack of additional risk posed by GIB in patients who already had OF. Second, the low yield in retrieving the causes may limit the adjustment in biliary and alcohol-related AP. However, the extent of the residual confounding, albeit unknown, seems not likely to change the conclusion. Third, we were unable to know which patients had necrotizing or interstitial pancreatitis. Given that the overall prevalence of necrotizing pancreatitis is less than 10%, it may be possible that most GIB (overall incidence 6.4% in this study) was seen in patients with necrotizing pancreatitis. Besides,
Table 3 Effects of gastrointestinal bleeding (GIB) with and without organ failure (OF) on the risks of septic complication and prolonged hospital stay in patients with first-attack acute pancreatitis (Total n ¼ 107,349). Outcomes
Predictors GIB
No. (%) with outcome
Odds ratio (95% confidence interval) Crudea
OF
Adjusteda,b Model 1
Model 2
Septic complication No Yes No Yes
No No Yes Yes
707 68 746 71
(0.8) (1.3) (6.2) (4.3)
1.00 1.52 (1.10e2.09) 8.28 (7.07e9.71) 5.69 (3.78e8.57)
1.00 1.53 (1.18e1.98) 3.80 (3.21e4.35) 2.60 (1.98e3.41)
1.00 1.54 (1.19e2.00) 3.52 (3.03e4.08) 2.41 (1.82e3.20)
No Yes No Yes
No No Yes Yes
5713 572 3532 530
(6.5) (11.0) (29.6) (31.9)
1.00 2.44 (2.10e2.85) 7.11 (6.23e8.11) 9.71 (7.01e13.46)
1.00 1.62 (1.46e1.80) 1.89 (1.75e2.05) 2.17 (1.77e2.65)
1.00 1.38 (1.28e1.48) 1.27 (1.20e1.35) 1.35 (1.15e1.60)
Prolonged stayc
a Multivariable logistic regression models with considering the cluster effect of hospitals using Generalized Estimating Equations (GEE) models. However, the estimates for the risk of septic complication were made only by multivariable logistic regressions because of non-convergence in GEE models. b Covariates included age (as a continuous variable), sex, year of admission, Charlson Comorbidity Index (categorized as 0,1,2 and 3), urbanization, hospital level, causes of acute pancreatitis (categorized into biliary, alcohol-related, both and others), intensive care unit admission and local complication (model 1). In model 2, additional covariates included cholecystectomy, total parenteral nutrition, vasopressors, hemodialysis and mechanical ventilation. c Prolonged stay was defined as a hospital stay more than 18 days, which was the 90th percentile of all patients with acute pancreatitis.
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
the septic complications in AP patients with GIB may be due to dysfunction of gut barrier caused by pancreatitis instead of transfusions, especially in those with severe AP. Nevertheless, this study is strengthened by a large number of patients from a nationwide database, which can provide an unbiased selection and enhance its generalizability. In conclusion, the results of this study suggest that GIB still accounts for an appreciable number of patients with AP and may significantly predict various adverse outcomes associated with AP even in the absence of OF. Further studies are needed to see whether prophylaxis of GIB, after excluding variceal or pseudoaneurysmal bleeding, in high risk patients can reduce the associated adverse outcomes. Acknowledgments The study was performed in Chi Mei Medical Center. HNS received a research grant CMFHR10025 from the hospital. CLL and CYL have no conflicts of interest or funding to disclose. The study was based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health or National Health Research Institutes.
335
Appendix 2. Distribution of various ICD-9-CM codes for gastrointestinal (GI) bleeding in patients with acute pancreatitis (AP).
ICD-9-CM code
Diagnosis
531.x (exclude 531.3, 531.7, 531.9) 532.x (exclude 532.3, 532.7, 532.9) 533.x (exclude 533.3, 533.7, 533.9) 534.x (exclude 534.3, 534.7, 534.9) 535.x1
Gastric ulcer with bleeding
31.6
28.4
Duodenal ulcer with bleeding
19.4
15.3
Peptic ulcer with bleeding
8.2
7.3
Gastrojejunal ulcer with bleeding
1.0
1.0
Gastritis/duodenitis with bleeding Hematemesis Blood in stool Unspecified GI bleeding e
11.4
16.8
0.3 0.5 27.6
0.3 0.4 30.5
100.0
100.0
578.0 578.1 578.9 Total
All patients with AP (n ¼ 6847), %
Patients with principal diagnosis of AP (n ¼ 2985), %
ICD-9-CM: International Classification of Diseases, Ninth Revision, Clinical Modification.
Appendix 1. Organ failure and gastrointestinal bleeding codes.
Organ failure
Codes
Code description
Cardiovascular
458.0 458.8 458.9 785.5 785.51 785.59 796.3 518.81 518.82 518.85 786.09 799.1 96.7 (96.71, 96.72) 96.04 93.90 584.x 570 572.2 573.3 573.4 293 348.1 348.3 780.01 780.09 89.14 286.2 286.6 286.9 287.3e5 790.92 276.2 531.x (exclude 531.3, 532.x (exclude 532.3, 533.x (exclude 533.3, 534.x (exclude 534.3, 535.x1 578.0, 578.1, 578.9
Hypotension, postural Hypotension, specified type, not elsewhere classified Hypotension, arterial, constitutional Shock Shock, cardiogenic Shock, circulatory or septic Hypotension, transient Acute respiratory failure Acute respiratory distress syndrome (ARDS) ARDS after shock or trauma Respiratory insufficiency Respiratory arrest Ventilator management Endotracheal intubation (emergency procedure) Continuous positive airway pressure Acute renal failure Acute hepatic failure or necrosis Hepatic encephalopathy Hepatitis (septic & not elsewhere classified) Hepatic infarction Transient organic psychosis Anoxic brain injury Encephalopathy, acute Coma Altered consciousness, unspecified Electroencephalography Disseminated intravascular coagulation Purpura fulminans Coagulopathy Thrombocytopenia, primary, secondary or unspecified Abnormal coagulation profile Acidosis, metabolic or lactic Gastric ulcer with bleeding Duodenal ulcer with bleeding Peptic ulcer with bleeding Gastrojejunal ulcer with bleeding Gastritis/duodenitis with bleeding Non-specific gastrointestinal bleeding
Respiratory
Renal Hepatic
Neurologic
Hematologic
Metabolic Gastrointestinal bleeding
531.7, 532.7, 533.7, 534.7,
531.9) 532.9) 533.9) 534.9)
336
H.-N. Shen et al. / Pancreatology 12 (2012) 331e336
References [1] Bradley 3rd EL. A clinically based classification system for acute pancreatitis. Summary of the international symposium on acute pancreatitis, Atlanta, Ga, September 11 through 13, 1992. Arch Surg 1993;128:586e90. [2] Acute Pancreatitis Classification Working Group. Revision of the Atlanta classification of acute pancreatitis (3rd revision). Available at: www. pancreasclub.com/resources/AtlantaClassification.pdf [accessed 20.02.12]. [3] Shen HN, Lu CL, Li CY. Epidemiology of first-attack acute pancreatitis in Taiwan from 2000 through 2009: a nationwide population-based study. Pancreas 2012;41:696e702. [4] Yadav D, Lowenfels AB. Trends in the epidemiology of the first attack of acute pancreatitis: a systematic review. Pancreas 2006;33:323e30. [5] Flint R, Windsor J, Bonham M. Trends in the management of severe acute pancreatitis: interventions and outcome. ANZ J Surg 2004;74:335e42. [6] Shen HN, Lu CL. Incidence, resource use, and outcome of acute pancreatitis with/without intensive care: a nationwide population-based study in Taiwan. Pancreas 2011;40:10e5. [7] Papachristou GI, Muddana V, Yadav D, O’Connell M, Sanders MK, Slivka A, et al. Comparison of BISAP, Ranson’s, APACHE-II, and CTSI scores in predicting organ failure, complications, and mortality in acute pancreatitis. Am J Gastroenterol 2010;105:435e41. [8] Gravante G, Garcea G, Ong SL, Metcalfe MS, Berry DP, Lloyd DM, et al. Prediction of mortality in acute pancreatitis: a systematic review of the published evidence. Pancreatology 2009;9:601e14. [9] Petrov MS, Shanbhag S, Chakraborty M, Phillips ARJ, Windsor JA. Organ failure and infection of pancreatic necrosis as determinants of mortality in patients with acute pancreatitis. Gastroenterology 2010;139:813e20. [10] Bollen T, Van Santvoort H, Besselink M, Van Leeuwen M, Horvath K, Freeny P, et al. The Atlanta classification of acute pancreatitis revisited. Br J Surg 2008; 95:6e21. [11] Banks PA, Freeman ML. Practice guidelines in acute pancreatitis. Am J Gastroenterol 2006;101:2379e400. [12] De-Madaria E, Soler G, Martinez J, Gomez-Escolar L, Sanchez-Fortun C, Sempere L, et al. Update of the Atlanta classification of severity of acute pancreatitis: should a moderate category be included? Pancreatology 2009;9:433e4. [13] Lankisch PG, Pflichthofer D, Lehnick D. No strict correlation between necrosis and organ failure in acute pancreatitis. Pancreas 2000;20:319e22. [14] Sharma M, Banerjee D, Garg PK. Characterization of newer subgroups of fulminant and subfulminant pancreatitis associated with a high early mortality. Am J Gastroenterol 2007;102:2688e95. [15] Sharma PK, Madan K, Garg PK. Hemorrhage in acute pancreatitis: should gastrointestinal bleeding be considered an organ failure? Pancreas 2008;36:141e5. [16] Company L, Saez J, Martinez J. Factors predicting mortality in severe acute pancreatitis. Pancreatology 2003;3:144e8. [17] Mole DJ, Olabi B, Robinson V, Garden OJ, Parks RW. Incidence of individual organ dysfunction in fatal acute pancreatitis: analysis of 1024 death records. HPB 2009;11:166e70. [18] Gardner TB, Vege SS, Chari ST, Pearson RK, Clain JE, Topazian MD, et al. The effect of age on hospital outcomes in severe acute pancreatitis. Pancreatology 2008;8:265e70.
[19] Vege SS, Gardner TB, Chari ST, Munukuti P, Pearson RK, Clain JE, et al. Low mortality and high morbidity in severe acute pancreatitis without organ failure: a case for revising the Atlanta classification to include “moderately severe acute pancreatitis”. Am J Gastroenterol 2009;104:710e5. [20] Xin MJ, Chen H, Luo B, Sun JB. Severe acute pancreatitis in the elderly: etiology and clinical characteristics. World J Gastroenterol 2008;14:2517e21. [21] National Health Insurance Research Database. Available at: http://w3.nhri.org. tw/nhird//en/index.htm [accessed 20.02.12]. [22] Bureau of National Health Insurance, Department of Health, Executive Yuan. Available at: http://www.nhi.gov.tw/webdata/webdata.aspx?menu¼17&menu_ id¼661&WD_ID¼689&webdata_id¼3351 [accessed 20.02.12]. [23] Liu C, Hung Y, Chuang Y, Chen Y, Weng W, Liu J, et al. Incorporating development stratification of Taiwan townships into sampling design of large scale health interview survey. J Health Manag 2006;4:1e22. [24] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373e83. [25] Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613e9. [26] Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med 2003;348:1546e54. [27] Sarani B, Dunkman WJ. Transfusion of fresh frozen plasma in critically ill surgical patients is associated with an increased risk of infection. Crit Care Med 2008;36:1114e8. [28] Claridge JA, Sawyer RG, Schulman AM, McLemore EC, Young JS. Blood transfusions correlate with infections in trauma patients in a dose-dependent manner. Am Surg 2002;68:566e72. [29] Singla A, Csikesz NG, Simons JP, Li YF, Ng SC, Tseng JF, et al. National hospital volume in acute pancreatitis: analysis of the nationwide inpatient sample 1998e2006. HPB (Oxford) 2009;11:391e7. [30] Hanley JA, Negassa A. Statistical analysis of correlated data using generalized estimating equations: an orientation. Am J Epidemiol 2003;157:364e75. [31] Petrov MS, Windsor JA. Classification of the severity of acute pancreatitis: how many categories make sense? Am J Gastroenterol 2009;105:74e6. [32] Chen TA, Lo GH, Lin CK, Lai KH, Wong HY, Yu HC, et al. Acute pancreatitisassociated acute gastrointestinal mucosal lesions: incidence, characteristics, and clinical significance. J Clin Gastroenterol 2007;41:630e4. [33] Lee KM, Paik CN, Chung WC, Yang JM. Association between acute pancreatitis and peptic ulcer disease. World J Gastroenterol 2011;17:1058e62. [34] Nathens AB, Curtis JR, Beale RJ, Cook DJ, Moreno RP, Romand J-A, et al. Management of the critically ill patient with severe acute pancreatitis. Crit Care Med 2004;32:2524e36. [35] Lankisch PG, Mahlke R, Blum T, Bruns A, Bruns D, Maisonneuve P, et al. Hemoconcentration: an early marker of severe and/or necrotizing pancreatitis? A critical appraisal. Am J Gastroenterol 2001;96:2081e5. [36] Remes-Troche JM, Duarte-Rojo A, Morales G, Robles-Díaz G. Hemoconcentration is a poor predictor of severity in acute pancreatitis. World J Gastroenterol 2005;11:7018e23. [37] Gardner TB, Olenec CA, Chertoff JD, Mackenzie TA, Robertson DJ. Hemoconcentration and pancreatic necrosis: further defining the relationship. Pancreas 2006;33:169e73.