Epidemiology and risk factors for mortality in critically ill patients with pancreatic infection

Background The AbSeS-classification defines specific phenotypes of patients with intra-abdominal infection based on the (1) setting of infection onset (community-acquired, early onset, or late-onset hospital-acquired), (2) presence or absence of either localized or diffuse peritonitis, and (3) severity of disease expression (infection, sepsis, or septic shock). This classification system demonstrated reliable risk stratification in intensive care unit (ICU) patients with intra-abdominal infection. This study aimed to describe the epidemiology of ICU patients with pancreatic infection and assess the relationship between the components of the AbSeS-classification and mortality. Methods This was a secondary analysis of an international observational study (“AbSeS”) investigating ICU patients with intra-abdominal infection. Only patients with pancreatic infection were included in this analysis (n=165). Mortality was defined as ICU mortality within 28 days of observation for patients discharged earlier from the ICU. Relationships with mortality were assessed using logistic regression analysis and reported as odds ratio (OR) and 95% confidence interval (CI). Results The overall mortality was 35.2% (n=58). The independent risk factors for mortality included older age (OR=1.03, 95% CI: 1.0 to 1.1 P=0.023), localized peritonitis (OR=4.4, 95% CI: 1.4 to 13.9 P=0.011), and persistent signs of inflammation at day 7 (OR=9.5, 95% CI: 3.8 to 23.9, P<0.001) or after the implementation of additional source control interventions within the first week (OR=4.0, 95% CI: 1.3 to 12.2, P=0.013). Gram-negative bacteria were most frequently isolated (n=58, 49.2%) without clinically relevant differences in microbial etiology between survivors and non-survivors. Conclusions In pancreatic infection, a challenging source/damage control and ongoing pancreatic inflammation appear to be the strongest contributors to an unfavorable short-term outcome. In this limited series, essentials of the AbSeS-classification, such as the setting of infection onset, diffuse peritonitis, and severity of disease expression, were not associated with an increased mortality risk. ClinicalTrials.gov number: NCT03270345


Introduction
Acute pancreatitis is an acute abdominal condition which is self-limiting and uncomplicated in most cases of edematous pancreatitis.[4] Severe acute necrotizing pancreatitis is characterized by high mortality due to a fulminant inflammatory process early in the course of the disease and the development of pancreatic and extra-pancreatic necrosis, infection, and multisystem organ failure (MOF) at a later stage. [ 5 ]The characterization of these pathways has improved the understanding of the disease, leading to the identification of potential molecular targets, including the mechanisms of calcium signaling and acinar cell injury/death.Additionally, the protective role of the unfolded protein response and the toxic role of unsaturated fatty acids and has possible therapeutic implications.Further research, however, is needed to translate these insights into clinical benefits. [ 6 ]Despite these achievements, the outcome of severe acute necrotizing pancreatitis has not improved significantly over the past years. [ 2 ]Among other serious intraabdominal infections, pancreatic infection complicating severe acute pancreatitis frequently poses a challenge in the intensive care unit (ICU) setting. [ 7 ]or this group of patients, multidisciplinary and wellorganized management is crucial, including close monitoring, appropriate use of imaging, proper fluid resuscitation, and use of antibiotics, nutritional and organ support.Additionally, the appropriate type and timing of endoscopic procedures, radiological interventions, and surgical interventions are critical. [8]Surgical and interventional endoscopic source control are crucial and are both therapeutic and diagnostic.These interventions may include debridement of necrotic tissues (including transgastric necrosectomy), drainage of abscesses, control of the source of infection, restoration of anatomy and function, and, although infrequently performed, abdominal cleansing and irrigation.Often several surgical interventions are required at different stages of the disease. [ 9 , 10 ]hile a dysregulated immune response, together with hemodynamic and coagulation abnormalities, may lead to early onset organ failure (within 24-72 h), a second deterioration phase may be triggered by bacterial or fungal superinfection of the necrotic tissue (i.e., infected pancreatic necrosis), and may, together with exacerbation of the pre-existing inflammation, result in late-onset organ failure (beyond 2 weeks). [ 2 , 11 ]Infection may be suspected in patients with a new, otherwise unexplained fever, (peri)pancreatic gas on imaging, bacteremia or candidemia in addition to worsening or persistent leukocytosis and other inflammatory parameters, and overall clinical deterioration. [ 8 , 12 ]5][16] Compared with sterile necrosis of the pancreas, pancreatic infection has a much worse prognosis, with mortality rates between 30% and 70%. [ 8 , 17-25 ]s in any intra-abdominal infection, the microorganisms involved include a broad spectrum of Gram-positive and Gramnegative bacteria, as well as anaerobic bacteria and in pancreatic infection fungi.Multidrug-resistant (MDR) bacteria are more commonly observed in patients with recent exposure to broad-spectrum antibiotics, underlying medical conditions, and prior hospitalization for more than 5 days. [ 10 , 26 ]In addition to patient-related factors, the risk of infection with MDR pathogens differs by the geographical region, local (in-hospital) ecology, and timing of appearance -depending on the disease severity and the intervention methods. [ 10 , 27 ]iven the broad spectrum of presentation, the appropriate classification of intra-abdominal infection and sepsis has long been debated.The international AbSeS study aimed to demonstrate the relationship between a risk classification tool and mortality, independent of the type of intra-abdominal infection. [ 7 ]he classification is based on three generic features of infection leading to specific clinical phenotypes of intra-abdominal infection.These features include (1) the setting of infection acquisition defined as either community, healthcare-associated, or early onset hospital-acquired (i.e., ≤ 7 days of hospital admission), or late-onset hospital-acquired infection ( > 7 days); (2) the absence or presence of anatomical barrier disruption, resulting either in localized or diffuse peritonitis; [ 7 , 28 ] and (3) the severity of the disease manifestation, defined according to the Sepsis-3 criteria (i.e., infection, sepsis, or septic shock). [ 29 ]The components of the AbSeS -classification were independently associated with mortality in the entire cohort, including all patients with intra-abdominal infection in the ICU and in the subgroup of patients with secondary peritonitis. [ 7 , 30 ]lthough the AbSeS -classification allowed adequate risk adjustment of different types of intra-abdominal infections, such as secondary peritonitis, its risk prediction value in other specific types of intra-abdominal infections remains uncertain.We aimed to assess the predictive value of the components of the AbSeS-classification in ICU patients with pancreatic infection.Therefore, the aim of this study was to assess risk factors for mortality in a subset of critically ill patients with pancreatic infection using the AbSeS risk classification.

Study design
This is a secondary analysis of data from the AbSeS study: a multinational observational cohort study of ICU patients with intra-abdominal infection. [ 7 ]The AbSeS cohort included 2621 patients from 309 ICUs in 42 countries.For the present study, only patients with pancreatic infection were considered ( n = 165, Figure 1 ).Approval was granted by established national, regional, or local institutional review boards.The study is registered at ClinicalTrials.gov(NCT03270345).More detailed descriptions of inclusion and exclusion criteria, definitions, methods, and collection of data are described elsewhere. [ 7 ]

Variables and definitions
Pancreatic infection was defined according to the International Sepsis Forum Consensus Conference on Definitions of Infection in the Intensive Care Unit. [ 31 ]Patients were eligible for inclusion if they had (1) microbiologically confirmed pancreatic infection (i.e., positive cultures directly from the pancreas or the surrounding structures by percutaneous aspiration or direct visualization and culture at the time of surgery or from the bloodstream), (2) probable infected pancreatitis (i.e., the presence of surgical or radiographic evidence of an abnormal collection of an inflammatory focus within the substance of the pancreas or the surrounding structures with a positive Gram stain from the pancreatic collection without evidence from culture), or (3) possible infected pancreatitis (i.e., radiographic or direct surgical inspection with evidence suggestive of pancreatic abscess or other types of infection originating from the pancreas).
The following data were selected from the AbSeS database: demographics (sex, age, and body mass index [BMI]), comorbidities (cardiovascular disorders, chronic pulmonary disease, neurologic disease, diabetes, acquired immunodeficiency syndrome [AIDS], liver disease, chronic renal failure, immunosuppression, malnutrition, and obesity), lifestyle risk factors (tobacco, drug, and alcohol abuse), risk factors for MDR pathogens (nursing home resident, out-of-hospital parenteral/nutrition or vascular access, chronic dialysis, recent hospitalization, and antibiotics use in the past 6 months), and components of the AbSeSclassification as mentioned above.
The severity of acute illness was assessed using the Simplified Acute Physiology Score (SAPS II score) at the time of ICU admission [ 32 ] and the Sequential Organ Failure Assessment (SOFA score) at the time of diagnosis. [ 33 ]The AbSeS protocol included a source control evaluation on day 7. Source control failure was defined as either the presence of persistent inflammation or the need for an additional intervention following the initial source control approach.However, source control evaluation is challenging in pancreatic infection.Persistent inflammation may be a sign of ongoing pancreatitis, although the infection is wellcontrolled.Additionally, recurrent interventions may be a strategy of damage control in acute infected pancreatitis as removal of necrotic material is often not done in a single procedure and can, therefore, not be used as a proxy of source control failure.Consequently, we decided to keep the clinical evaluation on day 7 in the study analysis but did not define it as either successful or failed source control.The clinical evaluation on day 7 was then categorized as "stable , " "unstable as evidenced by persistent signs of inflammation , " or "unstable with need for recurrent interventions ." In this evaluation, having "persistent signs of inflammation " may reflect either source control failure or ongoing pancreatitis.Likewise, the "need for recurrent interventions " may reflect either source control failure or a complicated case with the need for a stepwise damage control strategy.For patients who died before day 7, the source control evaluation was carried out on the last day alive.Outcome data included ICU length of stay and ICU mortality within 28 days of observation for patients discharged earlier from the ICU.
To assess microbial etiology and cultures derived from intraoperative sampling, trans-abdominal fine-needle aspiration, abdominal drains ( ≤ 24 h post-surgery), and blood cultures were considered.The basic empiric antimicrobial coverage (targeting Gram-positive, Gram-negative or anaerobic bacteria, and fungi) was evaluated.EUCAST breakpoints are used as antimicrobial resistance thresholds. [ 34 ]Antimicrobial resistance was defined as extended-spectrum beta-lactamase production (ESBL), carbapenem-resistance, and fluoroquinolone-resistance in Gram-negative microorganisms [ 35 ] and methicillin-resistance for Staphylococcus aureus (MRSA) or vancomycin-resistance in enterococci (VRE) for Gram-positive bacteria.

Statistical analyses
Continuous variables were described with median values and 25th to 75th percentiles (interquartile range [IQR]), and frequency and percentage (%) were used to describe categorical variables.The Mann-Whitney U test was used for comparisons between continuous variables, and the chi-squared or Fisher's exact tests were used to compare categorical variables.
Logistic regression analysis was performed to assess relationships with mortality.The following variables were considered in the model (irrespective of their relationship in univariate analysis): demographics, comorbidities, day-7 clinical evaluation, and elements of the AbSeS-classification (i.e., setting of infection acquisition, anatomical disruption, and the severity of disease expression).Feature selection was performed using the stepwise approach to remove covariates with P > 0.15.The final model contained, preferably, a maximum of one independent covariate per ten outcome events (dead).Results are reported as odds ratio (OR) and 95% confidence interval (CI).The SOFA score was not considered as it substantially overlaps with the severity of disease expression in the AbSeS-classification. Survival analysis of patients according to source control outcome was assessed using the Kaplan-Meier method, and the log-rank test was used to compare the survival curves.Statistical significance was defined as P < 0.05.Statistical analyses were performed using SPSS Statistics version 28, and all tests were two-tailed.
The overall mortality was 35.2% ( n = 58).Table 1 summarizes patients' characteristics for survivors and non-survivors.Overall, underlying conditions and lifestyle risk factors were more frequent in the non-survivor group.Non-survivors were older, had higher SOFA scores, and experienced more septic shock compared to survivors.Peritonitis (either localized or diffuse) and failure of source control were more common among nonsurvivors.The non-survivors presented more commonly with late-onset hospital-acquired infection.No difference between the groups was observed regarding the ICU length of stay.No clinically relevant differences in microbial etiology were observed between survivors and non-survivors.

Independent relationships with mortality
The objective of the logistic regression analysis was to assess relationships between the AbSeS -classification and mortality, alongside other risk factors for death.For that reason, we planned to "force " the AbSeS -classification components in the model irrespective of its relationship with the outcome.Due to the limited number of deceased patients ( n = 58), however, the final logistic regression model could not contain more than six independent covariates.The setting of the infection onset had no association with mortality.Therefore, this component of the AbSeS -classification was excluded from the model.The final model of the logistic regression analysis identified localized peritonitis, older age, and a clinical evaluation on day 7, indicating persistent inflammation or the need for repeated interventions for source/damage control as independent risk factors for death in critically ill patients with infected pancreatitis ( Table 3 ).Other parameters, such as microbial etiology, comorbidities, and lifestyle risk factors, were not associated with mortality and were excluded ( P > 0.15).NA: not applicable.In this group, 58 patients had no anatomical barrier disruption (71.6%), 10 had localized peritonitis (12.3%), and 13 had diffuse peritonitis (16.0%).The orange line represents the need for additional intervention following the initial approach.These patients had a median age of 60 years (46-71).In this group, 16 patients had no anatomical barrier disruption (57.1%), 3 had localized peritonitis (10.7%), and 9 had diffuse peritonitis (32.1%).The red line represents a condition with persistent signs of inflammation.These patients had a median age of 59 years (48-75).In this group, 37 patients had no anatomical barrier disruption (66.1%), 8 had localized peritonitis (14.3%), and 11 had diffuse peritonitis (19.6%).The P -value for the Log-rank test for differences between the curves was < 0.001.
Because of the significant relationship between the intermittent evaluation on day 7 and the outcome, we further analyzed the survival path stratified by the clinical status one week after diagnosis.These unadjusted survival curves are shown in Figure 2 .The cumulative survival curve for a clinical condition with persistent signs of inflammation is significantly lower, with overall higher mortality.

Discussion
This secondary analysis of the AbSeS study aimed to explore the association between mortality and independent risk factors in ICU patients with pancreatic infection.Our study is the first to report the specific implementation of the AbSeS-classification components for grading intra-abdominal infection in ICU pa-tients with pancreatic infection.Mortality in ICU patients with pancreatic infection was high (35.2%),significantly greater than the overall mortality in the AbSeS cohort (i.e., 29.1%). [ 7 ]The main finding of this study is that older age, localized peritonitis, and unstable clinical condition on day 7 (either due to persistent inflammation or the need for recurrent surgical intervention) are independently associated with mortality.The AbSeSclassification components, with previously recognized risk factors, such as late-onset hospital-acquired infection, diffuse peritonitis, and sepsis or septic shock, appeared to have no predictive value in this specific population.
In infected pancreatitis, localized peritonitis was associated with mortality, whereas diffuse peritonitis was not.This observation is counterintuitive, and this may be because diffuse peritonitis is rare in the cohort.Pancreatic infection rarely results in diffuse peritonitis being a retroperitoneal disease.The association with localized peritonitis should also be interpreted with caution as well due to the small numbers involved.A clinical picture of diffuse peritonitis may trigger earlier surgical intervention with a positive prognostic impact: studies have shown that delayed intervention is beneficial. [ 8 , 12 , 36-40 ]In this regard, it is challenging to assess the timing of surgical intervention as the onset of infection on top of a chronic inflammatory process is impossible to determine.
Based on the entire cohort of the AbSeS study, Blot et al. [ 7 ] identified the following variables as independent risk factors for mortality in intra-abdominal infections: higher SOFA scores, diffuse peritonitis, sepsis and septic shock, older age, malnutrition, diabetes mellitus, liver failure, and congestive heart failure.With the exception of older age, these aspects reflect the severity of the acute illness.In contrast, in the present study, the severity of acute illness and chronic underlying conditions did not prove to be significant determinants of mortality.Again, the limited study power must be considered, as it is highly uncommon for clinical aspects such as septic shock not to be associated with an increased risk of mortality.This association has also been reported by Zhu et al. [ 41 ] in the specific context of patients with severe acute pancreatitis, albeit the outcome window of 24 h may have boosted the relative importance of septic shock as a risk factor for death.
These findings reiterate that adequate source control is the cornerstone in the management of pancreatic infection in ICU patients. [ 7 , 16 , 24 , 39 ]A minimally invasive step-up approach is preferable, starting with the least invasive technique and gradually escalating in case of treatment failure. [ 23 , 40 , 42 , 43 ]The present data indicate that recurrent interventions are associated with poor outcomes.However, this probably reflects complicated cases requiring multiple procedures in the pursuit of damage control.Invasive procedures also carry a risk of superinfection and may worsen the prognosis in patients with sterile necrosis. [ 24 ]In pancreatic infections, the risk of superinfection is inferior to the beneficial impact of source control -antimicrobials as a single treatment appear to contribute less to infection control, primarily due to the persistence of infection in the necrotic tissues.
Delaying an intervention in stable patients is commonly recommended to allow for the development of a fibrous wall around the necrosis tissue. [ 8 , 12 , 36-40 ]This strategy is associated with fewer complications, such as iatrogenic intraoperative injury, bleeding, postoperative septic and systemic inflammatory response, and reduced mortality. [ 36 ]] Contrary to our conclusion, other studies conclude that conservative treatment with antibiotics as a single approach can resolve pancreatic infections and prevent the need for surgical intervention [48][49][50][51][52] or even reduce the risk of mortality. [ 53 , 54 ]Few antibiotics (e.g., carbapenems, quinolones, and metronidazole) are believed to penetrate sufficiently into the necrotic infected focus. [ 8 , 12 ]Tian et al. [ 55 ] suggested restricting the use of broadspectrum antibiotics and proposed more targeted drug administration based on specimen collection.For most of our patients, a conservative approach based on antimicrobials and organ support alone was probably not possible, given the high prevalence of sepsis or septic shock ( Table 1 ).
Microbiologically, the spectrum of pathogenic microorganisms was generally monomicrobial, with a predominance of Gram-negative bacteria (35.2%), consistent with previous studies. [ 7 , 10 , 55 ]While positive microbiology is as common as in the entire AbSeS cohort (71.5% and 75.6%, respectively), the AbSeS study reports a strikingly higher number of Gramnegative bacteria (58.6%) from clinical samples.Noor et al. [ 56 ] claimed that there is a shift from Gram-negative to Grampositive bacteria caused by the progression of pancreatitis and the longer hospital stay.Interestingly, a low number of patients ( n = 7, 4.2%) had MDR bacteria.The overall prevalence of antimicrobial resistance in the AbSeS study was higher (26.3%), [ 7 ] which may be explained by a high incidence of nosocomial infections, especially late-onset hospital-acquired infections in an exclusively ICU cohort. [ 55 ]Other studies also described the rising emergence of MDR bacteria. [ 7 , 27 , 57 , 58 ]An explanation for the low MDR rate present in our cohort may be the current policy of withholding the initial empiric broadspectrum antibiotics, despite ongoing inflammation and organ failure, until the most appropriate antimicrobial therapy is available based on culture results later in the course of the disease, with a secondary deterioration due to documented or probable infection.With this strategy, infection will develop without prior antimicrobial selective pressure toward MDR pathogens.
Infection of necrotic pancreatic tissue increases mortality and the probability of MOF, which is also an important determinant of death. [ 18 , 20 , 22 , 25 , 39 , 59-61 ]The co-existence of both complications has a synergistic effect that doubles the risk of death. [ 22 , 24 ]hese determinants and the need for ICU admission represent the most severe forms of the disease. [ 62 ]As our cohort exclusively contains ICU patients with pancreatic infection, they are considered at high risk for extra-pancreatic complications. [ 5 , 55 ]s previously shown, age has a significant impact on mortality in acute pancreatitis. [ 63 ]Similar to the study by Tian et al., [ 55 ] the mean age of patients in the non-survival group was significantly higher than in the survival group.Compared to the overall cohort of the AbSeS study, the mean age in the present cohort is significantly lower (66 years and 58 years, respectively), indicating that acute pancreatitis in the ICU is associated with high mortality rates irrespective of age.
First, per protocol, the AbSeS database only included patients with infection, making it impossible to assess the impact of infection in a subset of patients with acute pancreatitis.Second, some observations are counterintuitive, indicating that our study might be underpowered.In fact, with only 58 deceased patients, the number of covariates allowed in a logistic regression model to avoid overfitting is quite low.Therefore, larger studies are necessary to elucidate further the impact of the variables in the AbSeS-classification on mortality in ICU patients with pancreatic infection.Third, variables potentially influencing the outcome might not have been collected (e.g., the etiology of pancreatitis and location and extent of necrosis). [ 64 ]Fourth, no specific data were collected regarding the classification of acute pancreatitis, according to the revised Atlanta classification by Banks et al. [ 65 ] Yet, we assume that the included cases had severe pancreatic infections according to the definitions of the study protocol.Fifth, acute pancreatitis is a complicated disease, as it involves triggering a severe inflammatory process, distinguishing it from other intra-abdominal infections.Therefore, these cases are more difficult to categorize or predict with the AbSeS -classification, especially concerning the setting of the infection onset, where the majority are late complications: in contrast to secondary peritonitis, where the infection is part of the disease process from the beginning.Furthermore, the timing of surgical intervention as from infection onset proved impossible as it is unclear when the case of acute pancreatitis became complicated with infection.Finally, as already outlined in the methods section, in infected pancreatitis, source control evaluation on day 7 as per the AbSeS protocol proved unusable.

Conclusions
Older age, localized peritonitis, and a clinical condition reflecting persistent inflammation at 7 days, as well as the need for recurrent interventions within a week in pursuit of source or damage control, were the strongest predictors of mortality in critically ill patients with pancreatic infection.The AbSeSclassification system did not predict mortality in this sample of patients.

Figure 1 .
Figure 1.Included study population of pancreatic infection based on the total AbSeS study cohort.

Figure 2 .
Figure 2. Survival curves according to clinical status at day 7 in critically ill patients with pancreatic infection.The green line represents a stable clinical status (i.e., without overt intra-abdominal inflammation or the need for additional interventions for either source control or damage control).These patients had a median age of 59 years(44-72).In this group, 58 patients had no anatomical barrier disruption (71.6%), 10 had localized peritonitis (12.3%), and 13 had diffuse peritonitis (16.0%).The orange line represents the need for additional intervention following the initial approach.These patients had a median age of 60 years(46-71).In this group, 16 patients had no anatomical barrier disruption (57.1%), 3 had localized peritonitis (10.7%), and 9 had diffuse peritonitis (32.1%).The red line represents a condition with persistent signs of inflammation.These patients had a median age of 59 years(48-75).In this group, 37 patients had no anatomical barrier disruption (66.1%), 8 had localized peritonitis (14.3%), and 11 had diffuse peritonitis (19.6%).The P -value for the Log-rank test for differences between the curves was < 0.001.

Table 1
Characteristics of ICU patients with pancreatic infection according to survival status.
Continuous variables are presented as median (interquartile range).Categorical variables are presented as n (%).ICU: Intensive care unit; NA: Not applicable; SAPS: Simplified Acute Physiology Score; SOFA: Sequential Organ Failure Assessment.

Table 2
Source control characteristics and clinical outcomes at day 7.

Table 3
Independent relationships with mortality in critically ill patients with pancreatic infection.