Metagenomic next-generation sequencing, instead of procalcitonin, could guide antibiotic usage in patients with febrile acute necrotizing pancreatitis: a multicenter, prospective cohort study

Backgrounds: The effectiveness of procalcitonin-based algorithms in guiding antibiotic usage for febrile acute necrotizing pancreatitis (ANP) remains controversial. Metagenomic next-generation sequencing (mNGS) has been applied to diagnose infectious diseases. The authors aimed to evaluate the effectiveness of blood mNGS in guiding antibiotic stewardship for febrile ANP. Materials and methods: The prospective multicenter clinical trial was conducted at seven hospitals in China. Blood samples were collected during fever (T ≥38.5°C) from ANP patients. The effectiveness of blood mNGS, procalcitonin, and blood culture in diagnosing pancreatic infection was evaluated and compared. Additionally, the real-world utilization of antibiotics and the potential mNGS-guided antimicrobial strategy in febrile ANP were also analyzed. Results: From May 2023 to October 2023, a total of 78 patients with febrile ANP were enrolled and 30 patients (38.5%) were confirmed infected pancreatic necrosis (IPN). Compared with procalcitonin and blood culture, mNGS showed a significantly higher sensitivity rate (86.7% vs. 56.7% vs. 26.7%, P<0.001). Moreover, mNGS outperformed procalcitonin (89.5 vs. 61.4%, P<0.01) and blood culture (89.5 vs. 69.0%, P<0.01) in terms of negative predictive value. Blood mNGS exhibited the highest accuracy (85.7%) in diagnosing IPN and sterile pancreatic necrosis, significantly superior to both procalcitonin (65.7%) and blood culture (61.4%). In the multivariate analysis, positive blood mNGS (OR=60.2, P<0.001) and lower fibrinogen level (OR=2.0, P<0.05) were identified as independent predictors associated with IPN, whereas procalcitonin was not associated with IPN, but with increased mortality (Odds ratio=11.7, P=0.006). Overall, the rate of correct use of antibiotics in the cohort was only 18.6% (13/70) and would be improved to 81.4% (57/70) if adjusted according to the mNGS results. Conclusion: Blood mNGS represents important progress in the early diagnosis of IPN, with particular importance in guiding antibiotic usage for patients with febrile ANP.


Introduction
Acute pancreatitis (AP) is one of the leading causes of emergency admissions among gastrointestinal disorders, resulting in significant utilization of healthcare resources [1] .The disease is characterized with local and systemic inflammatory response and has a varying clinical course.Approximately 20% of patients develop acute necrotizing pancreatitis (ANP), which involves the necrosis of pancreatic or peripancreatic tissue.Infected pancreatic necrosis (IPN) occurs in 20-40% of ANP and is associated with a high mortality (30-40%) [2] .Early diagnosis of IPN remains a great challenge to the clinicians since sterile ANP and IPN may share similar clinical manifestations.Therefore, the indiscriminate use of antibiotics in ANP is quite common, which inevitably results in the emergence of antimicrobial resistance, unnecessary drug adverse events, and higher healthcare expenses [3][4][5] .Moreover, it has been established that infection caused by multidrug-resistant organisms is associated with adverse outcomes and increased mortality among patients with IPN [6] .
Thus, it is imperative to identify a simple and effective method for diagnosing pancreatic infection and guiding appropriate antibiotic usage.Fine needle aspiration (FNA) has been previously used as the procedure of choice to establish the diagnosis of pancreatic infection.Unfortunately, the demand for high standard technique and personal experience as well as the potential risk of iatrogenic complications do not render FNA an ideal approach [7,8] .Although air bubble sign within the necrotic collection presented on computed tomography is considered proof of infection, it occurs only in a minority of cases (20% or less) with infected necrosis [9] .In recent years, great expectations have been placed on procalcitonin which acts as a sensitive biomarker of bacterial infections [10][11][12] .The results of a single-center, patient-blinded, randomized controlled trial (PROCAP) indicate that procalcitonin-guided care may reduce the use of antibiotics without increasing the risk of infection in patients with AP [13] .However, there is no statistically significant risk difference of antibiotic use in patients with moderate or severe AP who had significantly higher risk of being infected [14,15] .Thus, there is an urgent need for an effective and easily accessible infection guide that can accurately identify patients who develop pancreatic infections.Such a guide would significantly improve the outcomes of ANP.
Metagenomic next-generation sequencing (mNGS) is a culture-independent analysis that uses sequencing technologies to study a mixture of microbial genomes.Numerous published case reports and clinical studies have demonstrated the successful application of mNGS in various sample types, including blood, urine, feces, cerebrospinal fluid, respiratory secretions, and tissues.In theory, this high-throughput technique has the potential to identify all pathogens present in a clinical sample through a single sequencing run, providing valuable evidence for guiding treatment options and improving antibiotic stewardship.Significantly, mNGS offers notable advantages in diagnosing infections that are rare or difficult to detect, as well as those involving novel or complex pathogens.Additionally, it has shown efficacy in diverse infectious disorders, such as pancreatic infections, and is rapidly transitioning from research settings to clinical laboratories [16][17][18] .Our prospective single-center study has shown the potential value of blood mNGS in diagnosing IPN with affordable cost and shorter turnaround time [18] .mNGS of pancreatic fluid aspiration has also been investigated in a small cohort of suspected IPN [19] .However, there is no prospective multicenter study to provide a more thorough insight into the implications of mNGS in the diagnosis of IPN and especially the guidance of antibiotics usage.
We designed this prospective multicenter trial specifically targeting patients with febrile ANP who were usually administered antibiotics indiscriminately in clinical practice.In this study, the effectiveness of blood mNGS, procalcitonin, and blood culture in diagnosing pancreatic infection was evaluated and compared.Additionally, the real-world utilization of antibiotics was also analyzed.Our hypothesis was that mNGS, instead of procalcitonin, could guide antibiotics usage in patients with febrile ANP.

Study design
This study was a prospective, multicenter, observational clinical trial that was conducted at seven hospitals in China from May 2023 to October 2023.The cohort included clinical, radiological, microbiological, and follow-up data.The study protocol was reviewed and approved by the ethics committee of all the participating hospitals.Written informed consent was obtained from all patients or their representative for the study participation.This study has been reported in line with the STROCSS (Strengthening the Reporting of cohort, cross-sectional and casecontrol studies in Surgery) Guideline [20] (Supplemental Digital Content 1, http://links.lww.com/JS9/B880).The study was performed in accordance with the Declaration of Helsinki.

Definition and patient selection
The diagnosis and classification of AP and ANP were based on the Revised Atlanta Classification [21] and the American Gastroenterological Association guideline [7] .Persistent organ failure (POF) was defined as that at least one of the three organ systems (pulmonary, circulatory, and renal) failure lasted for 48 h or more.IPN was defined as the presence of positive culture in pancreatic or peripancreatic necrotic specimens obtained during the initial drainage procedure.Sterile pancreatic necrosis (SPN) was defined as negative culture in pancreatic or peripancreatic specimens during the initial surgical intervention, or as the absence of infection signs upon discharge or follow-up without the necessity of any further surgical intervention.Antibiotic susceptibility was defined as reported by the local microbiology laboratories.When a susceptibility report was missing, the susceptibility was additionally interpreted by a clinical

HIGHLIGHTS
• Blood metagenomic next-generation sequencing showed superior diagnostic performance for infected pancreatic necrosis compared with procalcitonin.• Real-world antibiotic utilization in febrile acute necrotizing pancreatitis was not ideal.• The metagenomic next-generation sequencing-guided antimicrobial therapy could potentially improve the rate of correct antibiotic usage to 81.4% for patients with febrile acute necrotizing pancreatitis.
microbiologist according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines [22] .Patients with febrile ANP, which were defined as clinical manifestation of sepsis with temperature ≥ 38.5°C at least once during disease course without prior abdominal or retroperitoneal invasive procedure, were enrolled in this study.The exclusion criteria included age under 14 years old and confirmed extrapancreatic infectious complications at screening, such as biliary infection, urinary tract infection, and pneumonia.

Sample collection and data acquisition
Blood samples were collected simultaneously for mNGS, procalcitonin, culture, and other routine biomarkers such as C-reaction protein during fever (T ≥ 38.5°C).For patients who met surgical indications, peripancreatic specimens were collected for microbial culture during the initial surgical intervention.
The mNGS testing and bioinformatic analysis were provided by BGI China.The mNGS protocol has been described previously [18] .After quality control, ≥ 1 reads of bacteria or fungi detected by mNGS were considered positive.Procalcitonin assay and microbial cultures were performed according to the standard protocols in each study center.Procalcitonin was analyzed by a semi-automated chemoluminescent immunoassay (LUMITEST-PCT, BRAHMS Diagnostica AG).In accordance with Siriwardena's study [13] , the threshold for positive procalcitonin in this study was 1.0 ng/ml.The fully automated blood culture apparatus was BACTECFX (Becton, Dickinson and Company).Bacterial identification was performed by using the VITEK MS detection system (BioMérieux, Marcyl'Étoile).A positive culture was defined as the detection of a specific pathogen such as bacteria or fungi in the specimen, while a negative culture was defined as the absence of any organisms growing during the incubation period.
The data of clinical manifestations, laboratory tests, imaging results, antibiotic usage, and clinical outcomes were extracted from electronic medical records.

Treatment modality
This study was a real-world observational trial.The results of mNGS were blinded to the multidiscipline teams consisting of pancreatic surgeons, intensive care physicians, gastroenterologists, and interventional radiologists.Broad-spectrum antibiotics with the ability to penetrate pancreas were administered to patients with febrile ANP when the body temperature was greater than 38.5°C.If the antibiotics treatment failed, such as persistent fever without resolution or development of POF, a step-up surgical approach consisting of percutaneous catheter drainage, minimal access retroperitoneal necrosectomy, and/or open necrosectomy would be attempted.

Statistical analysis
Continuous variables were presented as mean SD and median (interquartile range, IQR) depending on the distribution.Categorical variables were described as n (%).χ 2 test or Fisher exact test was used to compare the distribution of categorical variables, and Student t-test or Mann-Whitney U test for continuous variables.The primary endpoint was the diagnosis of either IPN or SPN.The secondary endpoint was the mortality.Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic accuracy (the sum of true positives and true negatives divided by total) were calculated and used to examine the performance of blood mNGS, procalcitonin, and blood culture for diagnosing IPN using the Pearson χ 2 test.In the univariate analysis, Student t-test, Fisher exact test, Mann-Whitney U test, and χ 2 test were used for bivariate comparisons.The significant variables were subsequently included in the multivariate analysis to identify independent risk factors.This was achieved using binary logistic regression analysis.Odds ratio (OR) with 95% CI was calculated.Data analyses were performed using SPSS statistics, version 22 (IBM Corp) and GraphPad Prism 10 (GraphPad Software).A two-sided P value <0.05 was considered statistically significant.

Clinical characteristics and outcomes
Between May 2023 and October 2023, a total of 81 patients with febrile ANP were prospectively and consecutively enrolled from a pool of 276 ANP patients.Three patients were excluded from the analysis due to confirmed biliary infection (n = 1), urinary infection (n = 1), and pulmonary infection (n = 1), resulting in a final sample size of 78 patients.The flow chart for patient selection was described in Figure 1.Baseline characteristics, major complications, interventions, and clinical outcomes of the patients were summarized in Table 1.The median age was 45.0 years (33.5-51.5 years) with 58 (74.4%) males.The median time from onset to the manifestation of fever (also the blood sample collection) was 10.5 days (6.0-17.8days).IPN was confirmed in 30 out of 78 patients (38.5%), with a median onset time of 15 days (9.3-19.8days).The peripancreatic cultures in the patients with IPN revealed polymicrobial infections in ten patients (33.3%).Among the 30 patients with IPN, gram-negative bacteria were isolated in 22 patients (73.3%) by culture, with Acinetobacter baumannii being the most frequently reported pathogen (n = 9, 30.0%).Gram-positive bacteria were isolated in 12 patients (40.0%), with Enterococcus faecium being the most frequent species (n = 10, 33.3%).Fungal infections were found in five patients (16.7%), with Candida albicans being the most frequently reported fungus species (n = 2, 6.7%).Eight patients (10.3%) who died of irreversible POF before any surgical intervention could be attempted were classified undetermined, since FNA was never used to confirm IPN in participating hospitals.The remaining 40 patients (51.3%) were classified as SPN, among which 31 patients (77.5%) did not undergo any surgical intervention.During hospitalization, the overall mortality rate of the entire cohort was 20.5% (16/78).
Diagnostic performance of blood mNGS, procalcitonin, and blood culture for differentiating IPN from SPN The sensitivity, specificity, PPV, and NPV of blood mNGS, procalcitonin, and blood culture for diagnosing IPN were illustrated in Figure 2A.Compared with procalcitonin and blood culture, mNGS showed a significantly higher sensitivity rate (86.7% vs. 56.7% vs. 26.7%,P < 0.001).Moreover, mNGS outperformed procalcitonin (89.5 vs. 61.4%,P < 0.01) and blood culture (89.5 vs. 69.0%,P < 0.01) in terms of NPV.However, there were no statistically significant differences in terms of specificity and PPV among the three methods.The overall diagnostic accuracy of blood mNGS, procalcitonin, and blood culture was presented in Figure 2B.Notably, blood mNGS exhibited the highest accuracy (85.7%) in diagnosing IPN and SPN, significantly superior to both procalcitonin (65.7%) and blood culture (61.4%).

Undetermined patients
Eight patients (10.3%) who died of irreversible POF before any surgical intervention could be attempted were classified undetermined.Peripancreatic specimens were not available to pathogenically confirm the diagnosis of IPN or SPN.The median survival time in this subgroup was 18 days (18.0-20.5 days).Among them, six cases (75.0%) were positive of blood mNGS, and three cases were positive of both blood mNGS and blood culture.The median procalcitonin level was as high as 18.5 ng/ml (6.3-47.4ng/ml).

The potential mNGS-guided antimicrobial strategy in febrile ANP
This was a real-world observational study and the results of mNGS were blinded to the multidisciplinary teams.Thus, the rates of correct use, overuse, and misuse of antimicrobial therapy were assumed according to the results of mNGS (Fig. 3).Thirtyfour SPN patients with negative blood mNGS would not receive antibiotics (correct use) and the remaining six SPN patients with positive blood mNGS would be prescribed antibiotics (overuse).Among the 30 patients with IPN, antibiotics which could provide coverage according to the recommendations from antimicrobial drug usage card by BGI China would be prescribed in 23 patients (correct use).Three patients would be prescribed antibiotics without any coverage to the microorganisms (misuse) and four patients would not be prescribed any antibiotics due to negative mNGS results (misuse).Overall, the rate of correct use of antibiotics would be improved to 81.4% (57/70).The detailed consistency of pathogens between blood mNGS and peripancreatic culture was illustrated in Figure 4.

Discussion
To our knowledge, this is the first prospective multicenter study to systemically evaluate the effectiveness of blood mNGS in guiding antimicrobial therapy for ANP.This study demonstrated the superiority of blood mNGS over the widely used biochemical 'gold standard' procalcitonin as a distinct tool for diagnosis of IPN and initiation of antibiotics in febrile ANP patients.These findings strongly support an important role of mNGS use in helping antibiotic stewardship in AP and may lead to modifications of current practice and even recommendations in guidelines.
In the present study, patient recruitment was restricted to cases with ANP when temperature was more than 38.5°C.In such situation, nearly all the physicians would collect blood samples for culture and inflammatory biomarkers, including procalcitonin, to help establishing or excluding pancreatic infection.Procalcitonin looked like a good solution and seemed to have sound evidence from PROCAP trial [13] .The single-center, patient-blinded, randomized controlled trial demonstrated that procalcitonin-guided care could effectively reduce antibiotic usage in patients with AP without increasing the risk of infection [13] .However, there was no statistically significant risk difference of antibiotic use in patients with severe or febrile ANP who had the higher risk of infective complications [14,15,23] .Rau et al. [12] demonstrated in a prospective international multicenter study that procalcitonin did not allow the prediction of pancreatic infections in general, though procalcitonin proved to be an excellent variable to assess overall prognosis of AP, which was also observed in the present study.Thus, in critically ill patients with ANP, procalcitonin measurements needed to be carefully interpreted.In addition, procalcitonin was a nonspecific marker of sepsis and did not provide any information about the underlying microbes of infection.Whether antibiotics ought to be initiated and what kinds of antibiotics should be used were still unanswered in febrile ANP.Previous studies have observed an excessive and inappropriate utilization of antibiotics in patients with ANP [4,24] .Similar results were also observed in the present study.As high as 92.3% of patients with febrile ANP were administered empirical antibiotics, but only 18.6% were correct use.Early diagnosis of IPN remained to be a significant challenge [25] .
mNGS was an innovative technique that has been utilized for diagnosing various infectious diseases, particularly in cases of rare infection, complicated infection, and critical infection [16,26] .It could provide sensitive and accurate pathogen results in a rapid manner and has been proven to reduce morbidity in acutely ill infants and lower the cost of hospitalization [27] .Our previous study conducted at a single-center with a relatively small sample size showed that blood mNGS had a good potential to be an effective method for early diagnosis of pancreatic infections [18] .However, the clinical effectiveness of mNGS-guided antimicrobial therapy in ANP has not yet been established.One of the main aims of the present study was to compare blood mNGS and the conventional workup including procalcitonin and blood culture in terms of effectiveness of diagnosing IPN in patients with febrile ANP.The results demonstrated that blood mNGS had a better diagnostic efficacy compared with its counterparts.Compared with procalcitonin and blood culture, not only did mNGS show a significantly higher sensitivity rate, but also outperformed in terms of NPV.The overall accuracy in diagnosing IPN and SPN was also promoted from 65.7% (procalcitonin) and 61.4% (blood culture) to 85.7% (blood mNGS).Since infection sources other than the abdomen such as  pulmonary, urinary tract, or catheter infections were frequently observed in critically ill patients, a positive mNGS result needed to be carefully interpreted.However, 81.3% of PPV in the current series, in general, meant in the absence of an abdominal septic focus, other sites of infections had by far a lesser influence on blood mNGS.
The most important strength of the present study was that the usefulness of blood mNGS in guiding antibiotic usage for patients with febrile ANP was assumed and evaluated.With high NPV of 89.5%, blood mNGS showed excellent performance in diagnosing SPN and might help avoid unnecessary antibiotics usage.Moreover, with good concordance of pathogen identification, blood mNGS could greatly facilitate antibiotics choices and decrease antibiotics misuse.Even with great advantages mentioned above, it was noteworthy that due to false negative results of mNGS, four cases of IPN in the study would miss the time window of initiating antibiotics.
According to our previous study [18] , despite the higher cost, mNGS only accounted for 1.8% of the average treatment cost of IPN patients and 3.6% of the average treatment cost of SPN patients.Moreover, the implication of the technique would lead to early diagnosis of infection and timely initiation of infection control strategies including antibiotics and surgical drainage and/ or debridement.Consequently, overall hospitalization costs might be reduced accordingly.Furthermore, as the mNGS technique continues to develop and popularize, cost reduction will be anticipated.
There are several limitations to consider.First, the observational nature of this study has inherent limitation in the ability to ascertain the role of mNGS in guiding antibiotic usage in ANP.Randomized controlled clinical trial is warranted in the future.Secondly, there may be IPN patients who have recovered just on antibiotics without any intervention and would have been branded as SPN.Ideal criteria for SPN or IPN population are to obtain culture through FNA of the necrosis.Unfortunately, in the present study, unless patients undergo drainage or surgery, it is not clear who have real pancreatic infection.Given the fact that a fraction (~20-30%) of patients respond to antibiotics alone [28,29] , IPN patients who recovered with antibiotics may have been misclassified as SPN.In addition, we believe that in the undetermined group, most of the patients have a high likelihood of being infected, as indicated by the fact that 75% of patients in this subgroup have positive blood mNGS results and 37.5% are positive of both blood mNGS and blood culture.Thirdly, the risk of false negative of mNGS cannot be ignored.Four cases of IPN in this series presenting with negative blood mNGS may miss the time window of lifesaving antimicrobial therapy if following mNGS-guided antimicrobial therapy alone.Thus, it is important to emphasize that mNGS is no substitute for careful history and clinical examination of the individual patient.Evidence generated from this study supports that mNGS measurement should be added into the armamentarium of guiding antimicrobial therapy for ANP.
In conclusion, it is a prospective multicenter real-world study that highlights applying blood mNGS to early pathogenic diagnosis of IPN.It also emphasizes the significant role of blood mNGS in guiding antibiotic usage for patients with febrile ANP.Further randomized controlled trials are warranted to gain a more comprehensive and definitive conclusion.

Table 1
Clinical characteristics and outcomes in 78 patients with febrile ANP.