Impact of prior antibiotics on infected pancreatic necrosis microbiology in ICU patients: a retrospective cohort study

Severe acute pancreatitis is a common reason for intensive care unit (ICU) admission and is associated with long hospital stays and high morbidity and mortality rates [1, 2]. Infected pancreatic necrosis (IPN) develops in about one-third of patients, who are then at increased risk for death [3]. International guidelines about the management of IPN [4‐8] recommend a step-up strategy, based on findings from randomized controlled trials [9, 10]. The first step is minimally invasive (percutaneous or endoscopic) drainage combined with broad-spectrum antibiotic therapy. When this fails, minimally invasive endoscopic or surgical necrosectomy is performed. Microbiological studies are performed on all drainage and necrosectomy samples.

Prophylactic antibiotic therapy was not helpful in preventing IPN in several randomized controlled trials, and is therefore not recommended [11‐13]. However, most international guidelines [4, 5, 14, 15] fail to provide recommendations about using antibiotics in patients with IPN, perhaps due to the scarcity of relevant published data. Canadian guidelines [16] recommend reserving antibiotics for patients with confirmed IPN and tailoring the antibiotic regimen to the species and sensitivities of the bacteria recovered from necrotic tissue samples. However, these guidelines also suggest that empirical antibiotic therapy may be considered before the culture results are available.

For sepsis and septic shock, in contrast, there is a clear recommendation to start antibiotics as early as possible [17]. Because sepsis and septic shock are possible presentations of IPN, many patients with IPN therefore receive antibiotics before the diagnosis of IPN is made or before a pancreatic sample can be collected. Furthermore, patients with IPN may require immediate antibiotic therapy for a co-existing infection at another site. The penetration of antibiotics within foci of pancreatic necrosis has been poorly evaluated but may be limited [18‐26]. Thus, whether antibiotic exposure before the collection of pancreatic samples sterilizes the microbiological findings is unclear. Importantly, studies have documented a rise in the proportion of patients with pancreatic samples positive for multidrug-resistant bacteria (MDRB) up to values of 52% [27] and 63% [28]. Limiting the development of bacterial resistance by restricting the use of antibiotics is clearly a major public health goal [29‐31]. If antibiotic exposure before pancreatic sampling hinders the identification of causative organisms, then exposed patients might not only have poorer outcomes, but also require longer and/or broader antibiotic treatments at higher risk for selecting resistant strains, and negative samples may not be sufficient reason to stop the antibiotics.

Our primary objective was to investigate whether the microbiology of pancreatic samples from critically ill patients with suspected IPN was sterilized according to whether antibiotics were started before or only after sample collection. The secondary objective was to describe the bacterial species recovered and their resistance patterns in the groups previously exposed vs. unexposed to antibiotics.

Organ failure is common in patients with IPN [34]. The recommendation when following the step-up strategy is to wait at least 4 weeks if possible before performing a pancreatic procedure. During this period, antibiotics are often given due to sepsis or septic shock or to a documented infection outside the pancreas (e.g., nosocomial pneumonia, catheter-related infection, or bacteraemia). Negative samples collected while the patient was receiving antibiotics may indicate absence of an infection, presence of an infection that is responding to antibiotics, or a false-negative result. In our study, most of the patients with suspected IPN had positive cultures, with no difference between the exposed and unexposed groups. Thus, pancreatic sample cultures seem to preserve their diagnostic value even in patients on antibiotics and, in addition, provide valuable antimicrobial susceptibility information. If the cultures are negative, the diagnosis of IPN should be reappraised. Moreover, as in previous studies [9, 10], our results showed a high proportion of patients (86%) with a confirmed IPN by positive sample culture collected during drainage procedure or necrosectomy. These findings are consistent with international recommendations [4, 7, 8, 15, 16] advocating against the routine use of fine needle aspiration to confirm IPN.

In a study from England and Wales, antibiotics were given before the first pancreatic sample was collected in 439/712 (62%) patients [35], compared to 59% in our study. However, the inclusion criterion was acute pancreatitis, as opposed to suspected IPN in our study, and the patients were not all critically ill, with only 60/439 having sepsis. Furthermore, IPN was the reason for the first, second, and third courses of antibiotics in only 8, 0, and 1 patients, respectively, whereas in our study, a third of the patients had septic shock. Piperacillin/tazobactam was the most common antimicrobial used (34%), similar to our cohort (30%). In a retrospective study of a prospective database in the US of 182 consecutive symptomatic patients who had undergone pancreatic procedures for walled-off necrosis, 41% had culture-proven IPN [36]. Of these patients with positive cultures, 70% received antibiotics within 14 days before sample collection; and of the patients exposed to antibiotics before first sample collection, 70% had proven IPN. These results are consistent with ours, as are reports that about a third of samples contain more than one identified organism and that Gram-negative bacteria predominate [27, 37‐39].

About a third of our patients had MDRB or XDRB identified in their pancreatic samples. Highly resistant strains of several bacterial species are becoming increasingly common in ICU patients [31]. In addition, among patients with acute pancreatitis, 50% to 85% have been found to have highly resistant bacteria [27, 28, 39]. Infections due to MDRB and XDRB are associated with considerably higher morbidity, mortality, and healthcare costs compared to those due to less resistant bacteria [39]. Acute pancreatitis with IPN runs a prolonged course that is often punctuated by a series of complications. Thus, many patients have long ICU and hospital stays associated with a high risk of cross-transmission [29], receive multiple courses of antibiotics, and require more than one pancreatic procedure. After a drainage procedure, more than half the patients require necrosectomy [9, 10]. The median number of procedures per patient in our study was 2. This complicated course may increase the opportunities for developing highly resistant bacteria.

Candida was identified in only 7% of our patients. In previous studies, this proportion varied widely, from 5% to 68.5%, perhaps in part due to differences in patient selection criteria. Prophylactic antibiotic therapy increases the risk of fungal infections which are associated with increased morbidity and mortality rates [40‐42] and is not recommended in international guidelines. The absence of prophylactic antibiotic therapy in our patients may explain the low frequency of fungal infections.

In our study, antibiotic therapy given before sample collection failed to sterilize the pancreatic samples, even when broad-spectrum agents were given for prolonged periods. One possible explanation is poor penetration of the antibiotics into the fluid collections and necrotic foci. Very few data are available on this point. In two studies, after a single intravenous dose of ertapenem or imipenem, concentrations in pancreatic tissue and juice, respectively, were low but above the minimum inhibitory concentrations for the main pathogens responsible for intra-abdominal infections [22, 43, 44]. Several studies of animal models of acute pancreatitis [24‐26] showed penetration within necrotic tissue to levels above the minimal inhibitory concentrations of common pathogens, and a single study [23] demonstrated good cefepime concentrations in pancreatic pseudocyst fluid or pancreatic resection specimens for cancer. However, pancreatic necrotic tissue during acute pancreatitis was not studied. If penetration is poor in necrotic tissue, then the sharp rise in highly resistant bacteria among patients with acute pancreatitis supports the use of antibiotic-sparing strategies in patients with IPN [37]. Antibiotic de-escalation based on antimicrobial susceptibility test results is needed to reduce the development of resistance [29, 30]. The cornerstone of IPN management is the removal of infected fluids and necrotic tissues according to the step-up strategy. Systemic antibiotics alone are not recommended for IPN [4, 5, 15, 16].

Blood cultures were positive in about a third of our patients, in keeping with earlier findings [9]. Antibiotic therapy before blood sample collection from patients with severe sepsis has been reported to markedly decrease the likelihood of positive blood cultures, even when time from antibiotic initiation to sampling was short [45]. Thus, obtaining pancreatic samples is crucial in patients with suspected IPN. For patients in septic shock (about one-third in our study), antibiotics must be started immediately. For the other patients, the question is whether to withhold antibiotics until the causative agent is identified or to start antibiotics then discontinue them promptly if the microbiological tests are negative.

Moreover, the low frequency of fungal infections in our study suggests that antifungal agents may be appropriate only in patients with a documented fungal infection, and in those with septic shock provided the treatment is stopped if the samples are negative for fungi.

Regarding the recovered microorganisms, the exposed group had more Enterobacter cloacae complex and non-fermenting Gram-negative bacteria and the unexposed group more Gram-negative anaerobic bacteria. Candida also seemed more frequent in the exposed group (n = 5 versus n = 1) but the difference was not significant, probably due to the small sample size. These disparities in microbial findings can be explained by exposure to antibiotics, which had a median duration of 7 days before bacteriological sampling in the exposed group. Other patient outcomes in our study were comparable in the exposed and unexposed groups, except for a higher overall number of antibiotics in the exposed group. Similarly, in a 2019 retrospective cohort study of ICU patients with acute pancreatitis, mortality was not different between the groups with vs. without antibiotic therapy at ICU admission [46].

The main limitation of our study is the retrospective single center design, with no standardization of the antibiotic treatment strategy. No prophylactic antibiotics were used in any patients. However, we did not have information on antibiotic exposure before ICU admission, on the reason for the timing of interventions, or on the reasons for discontinuing antibiotic therapy. Few guidelines exist about when to start and stop antibiotics or which antibiotics to use. Empirical therapy with antibiotics active on gastrointestinal organisms followed by adjustments based on antimicrobial susceptibility tests has been suggested [1, 7, 16]. We also had no information on nosocomial infections at extra-abdominal sites. In patients with multiple pancreatic fluid collections and/or necrotic foci, samples were not routinely obtained from all foci. This fact may have resulted in false-negative culture findings. Furthermore, patients categorized in the unexposed group may have received previous antimicrobial therapy for more than 24 h that may have affected the microbiological findings. However, given that cultures were positive in 86% of both patients exposed and unexposed to antibiotics within 24 h before sampling, in keeping with earlier data [9, 10], we are confident that previous antimicrobial therapy exposure is unlikely to have sterilized the samples. No information was available on intestinal carriage or known colonization of these patients, or on the mechanisms of acquired antibiotic resistance in the patients with highly resistant strains. Data on inflammatory status including body temperature, leukocytosis, and procalcitonin levels were also lacking, as they were not reliably recorded in the medical files. However, these data are of limited relevance in critically ill patients. Despite these limitations, our study provides the first evidence on the impact of prior antibiotics on pancreatic sample microbiology in patients managed with the currently recommended step-up strategy for suspected IPN and no prophylactic antibiotics. We collected a vast amount of information on the types and timing of antibiotic treatments, culture results, and antimicrobial susceptibilities.

In patients admitted to the ICU with suspected IPN, starting antibiotic therapy more than 24 h before collection of the first pancreatic sample did not seem to sterilize microbiological cultures and had a limited impact on microbiological findings. The cultures were positive in most patients. It therefore seems appropriate to start empirical antibiotics immediately in patients with sepsis, septic shock, or a new organ failure. A pancreatic sample should then be obtained as soon as possible to identify the organism and assess the susceptibility profile. Given the high prevalence of bacteria exhibiting resistance to many agents, in other situations, antibiotic therapy may be best reserved for those patients with positive pancreatic samples may be appropriate, in order to limit the unnecessary use of antibiotics.