Advanced chronic liver disease creates a state of relative immunocompromise,1 and infections are a key precipitant of clinical decline in patients with cirrhosis.2–4 Specifically, infections of any kind are a common cause of acute on chronic liver failure (ACLF)4,5 and complicate 20%–40% of hospital admissions of patients with cirrhosis.2,6–9 Although other pathogens are also important, the vast majority of infections in patients with cirrhosis are of bacterial origin.10 In other high-risk hospitalized patient populations, early initiation of antibiotics has been shown to reduce the need for intensive care unit admission and short-term mortality.11–15 Unfortunately, the presentation of infection in patients with cirrhosis is protean and may be subtle. Diagnosis may be delayed by the absence of classic signs of infection, such as fever and absolute leukocytosis.16 Vital diagnostic studies, such as paracentesis17 or other cultures (blood, urine, or sputum), may be delayed or may not yield results until days after presentation. Delaying antibiotics until such studies result is inadvisable in patients with infection; however, empiric administration of antibiotics to all patients with cirrhosis admitted to the hospital would likely result in waste and increased risk of antibiotic resistance, Clostridium difficile infection, and other adverse effects. Sub-groups of patients with cirrhosis are already known to benefit from empiric/prophylactic antibiotics, including those with acute upper gastrointestinal hemorrhage18,19 and those with an increased risk of spontaneous bacterial peritonitis.20,21 Other sub-groups, however, have not demonstrated survival benefit with antibiotic prophylaxis, including those with severe acute alcohol-associated hepatitis and acute liver failure.22,23In addition to these readily identified sub-groups, the risk of infection increases with more advanced cirrhosis, as marked by a higher Model for End-stage Liver Disease-Sodium (MELD-Na) score.4,6,24 However, efficient identification of a potential subgroup of patients with cirrhosis most at risk remains challenging. There are also biomarkers of systemic infection that, although not specific, may be used to identify patients more likely to have an active infection and thus allow more targeted use of empiric antibiotics.25 Specifically, an increase in serum C-reactive protein (CRP, > 25 mg/L)26–32 and procalcitonin (PCT, > 0.5 ng/mL)26,33 are associated with acute bacterial infection and increased short-term mortality.28,29,34,35These factors may permit the identification of patients who would benefit from prophylactic/empiric antibiotics. There has, however, been little prospective investigation of the effect of prophylactic antibiotics on hospitalized patients with advanced cirrhosis outside the sub-populations described above.In this pilot study, we aimed to assess the feasibility of subject identification and data collection, including specimen processing, as well as the rate of enrollment and events. These data were intended to inform the design of larger studies of the effect of empiric antibiotics for patients with advanced cirrhosis admitted to the hospital without an existing indication for antibiotic treatment or new prophylaxis (eg, patients with variceal hemorrhage). Specifically, we assessed the incidence of infection after the time of enrollment and associated outcomes.
METHODS
We pre-screened all adult patients presenting for hospital admission to the inpatient Hepatology ward service with a diagnosis of cirrhosis based on a composite of available histology, laboratory, radiology (including elastography), and historical evidence as soon as possible after admission. To enrich for risk of infection and minimize unhelpful exposure, only patients with MELD-Na ≥ 18 were considered eligible. Patients were excluded if they had a known history of allergy to ceftriaxone (CTX) or recurrent C. difficile infection; were receiving new antibiotics for the treatment or prevention of possible infection (eg, prophylaxis in the setting of gastrointestinal hemorrhage); were receiving systemic immunosuppression; were known to have an infection based on available clinical data; or had enrolled in the study on a prior admission. If not done as part of the standard of care, a chest radiograph, urinalysis and urine culture, blood culture, and diagnostic paracentesis (cell counts and culture) were performed to screen for infection. Patients receiving antibiotics as part of the continuation of outpatient prophylaxis for spontaneous bacterial peritonitis were eligible. The study intervention was stopped in the event of clinical requirement for antibiotic therapy comparable to CTX in the spectrum of activity. These subjects were still followed to observe outcomes. Informed consent was provided by subjects or legally authorized representatives in cases of incapacity.Blood samples were collected from each subject before the initiation of the study intervention and frozen for future processing. CRP and PCT concentrations were measured using ELISA (Abcam Cat ab99995 and ab100630, Waltham, MA) following protocols provided in the kit. Each sample was measured twice. Samples with variance between duplicates more than 10% of the mean were re-measured. Samples with measurements falling beyond the highest concentration in the standard curve were further diluted and measured again.Subjects were randomly assigned to receive either 1 g/day of IV CTX or an equivalent amount (100 mL) of normal saline for a maximum of 7 days. Randomization was performed in blocks of 4 subjects (2 each of placebo and CTX) and stratified based on the presence or absence of (either of) serum creatinine ≥ 2.0 mg/dL or diuretic-refractory ascites. Random sequences were generated using a Mersenne Twister algorithm. Investigators and subjects were blinded to the study intervention until the study was completed. Research pharmacy staff generated and maintained the final blinding sequence and created identical-appearing doses of placebo or CTX. Unblinding was performed for data analysis after the completion of enrollment. The sample size was chosen based on available resources to permit a preliminary assessment of effect size and assess capacity for enrollment for this feasibility study.This study was approved by the Beth Israel Deaconess Medical Center Institutional Review Board and registered through ClinicalTrials.gov (NCT04218695). All research was conducted in accordance with both the Declarations of Helsinki and Istanbul, and all subjects or a legally authorized representative provided written informed consent before participation. Funding for the study was from the AASLD Foundation.The overall primary outcome of interest was the feasibility of the study design, as detailed above. This included the ability to identify and enroll potential subjects, as well as the rate of enrollment and observation of the rate of infection and any potentially related adverse events.The primary subject-level outcome of interest was incident bacterial infection up to 7 days after enrollment identified using standard of care. Radiograph and culture results were reviewed when performed as part of standard clinical care, and the clinical team’s diagnosis defined the presence of infection. See Supplemental information, http://links.lww.com/HC9/A725 for definitions. Subjects were followed until hospital discharge, liver transplantation, or death.Secondary subject-level outcomes of interest were pre-specified based on overall clinical relevance to patient condition or plausible relation to CTX use: 30-day mortality (Vital status at 30 days was confirmed through review of the medical record or phone calls for those without follow-up at our institution.), change in ACLF score by Chronic Liver Failure Consortium36 and MELD-Na (≥3 points) between enrollment and end of treatment), length of stay subsequent to enrollment, intensive care unit admission after enrollment, and incident variceal hemorrhage. For assessment of key potential adverse effects of antibacterial therapy: C. difficile infection and other CTX-resistant infections, as well as fungal infection, were recorded.Covariates of interest included age, sex, etiology of cirrhosis, history of decompensation, initial laboratory values (Table 1), and use of other antibiotics during hospitalization [ie, for spontaneous bacterial peritonitis (SBP) prophylaxis].Baseline characteristicsNotes: Categorical variables presented as n (%).Continuous variables presented as median (IQR).Abbreviations: CLIF-C ACLF Score, Chronic Liver Failure Consortium acute on chronic liver failure score; INR, International Normalized Ratio; MELD-Na, Model for End-Stage Liver Disease-Sodium; SBP, Spontaneous Bacterial Peritonitis; WBC, white blood cell count (peripheral blood).Comparisons between groups were made using the Fisher exact test for categorical variables and the Wilcoxon-rank sum for continuous variables. Outcomes were analyzed by intention to treat for all subjects receiving any number of study treatments.
RESULTS
After screening 276 patients with cirrhosis admitted to the Hepatology ward service, we enrolled 31 subjects overall between August 2020 and July 2021 (Figure 1). One subject developed an infection after randomization but before receiving any study intervention and was thus excluded from analyses, leaving 15 subjects each receiving CTX or placebo. Most subjects were men (Table 1) with an overall median age of 58 years. Alcohol-associated liver disease was the most common etiology. Overall, the 2 groups were similar, notably with comparable starting MELD-Na and ACLF scores as well as levels of CRP and PCT (Table 1).Study enrollment. Abbreviations: MELD-Na, Model for End-Stage Liver Disease-Sodium.Of 30 subjects receiving any amount of study treatment, 16 (8 CTX and 8 placebo) completed a full 7-day course. Of the 14 not completing all 7 days of treatment, 5/7 on CTX and 4/7 on placebo improved and were discharged before completion of 7 days of intervention. One in seven on CTX and 3/7 on placebo required new standard of care antibiotics for presumed infection with comparable bacterial coverage to CTX and thus, the study intervention was stopped. One subject elected to stop the study treatment early due to adverse reactions (nausea and pruritus); this subject was subsequently found to have received CTX. As expected with this population, adverse events were frequent, including serious adverse events. However, no serious adverse events potentially related to CTX exposure were observed. A single subject diagnosed with C. difficile infection within 30 days after enrollment received the placebo intervention. Among the 30 subjects beginning therapy, the proportion of subjects with newly diagnosed bacterial infections was lower among those receiving CTX (1/15) compared to placebo (3/15). This difference was not statistically significant in this pilot sample (OR 3.4, 95% CI: 0.23–196, p = 0.6) (Table 2). The overall number of infections occurring through the end of treatment was also lower among subjects receiving CTX (1 infection) compared to those receiving placebo (5 infections). Thirty-day mortality was also not different between groups (2/15 CTX vs. 5/15 placebo, OR 3.1, 95% CI: 0.40–39, p = 0.39). Of all subjects discharged before study day 7, none were re-admitted within 7 days. Of all subjects discharged before study day 30, only 1 was re-admitted within 30 days (due to urinary tract infection). Upon review, this subject had received placebo.OutcomesNotes: Categorical variables are presented as n (%).Continuous variables are presented as median (IQR).Abbreviations: CLIF-C ACLF Score, Chronic Liver Failure Consortium acute on chronic liver failure score; EOT, end of treatment course; MELD-Na, Model for End-Stage Liver Disease-Sodium.The median reduction in MELD-Na at the end of treatment was greater among those receiving CTX (−2.0) compared to placebo (0.0); however, again, this difference did not meet statistical significance in this pilot cohort. Improvement in MELD-Na by 3 or more points was observed with similar frequency in the 2 groups (5/15 CTX and 4/15 placebo, OR 0.74, 95% CI: 0.11–4.6, p = 1). The post-treatment median change in chronic liver failure Consortium acute on chronic liver failure score was favorable in the CTX group: 0.31 decrease compared to placebo: 0.26 increase; however, this was not statistically significant (p = 0.74). The median LOS after enrollment was similar: 11 and 10 days in CTX and placebo groups, respectively (p = 0.89). No cases of variceal hemorrhage were observed in either group. 0/15 CTX and 1/15 placebo-treated subjects required intensive care unit care during the 7 days of potential treatment (p = 0.1).We did not detect a difference in median CRP at the start of treatment between those who developed an infection and those who did not [9.47 (6.8) vs. 7.5 (12.4), p = 0.36]. We also did not detect a difference in median PCT at the start of treatment between those with infection observed and those without [0.26 (0.2) vs. 0.27 (0.27), p = 0.95]. Of subjects with CRP > 25 g/mL, 1/1 developed an infection during the study period. This subject received a placebo. Of subjects with PCT > 0.5 ng/mL, 0/5 developed infection during the study period. 4/5 of these subjects received placebo.
DISCUSSION
In this pilot study, we enrolled hospitalized adult patients with cirrhosis felt to be at increased risk of infection based upon elevated MELD-Na (≥18) to receive either CTX or placebo for up to 7 days. The primary subject-level outcome of interest was incident bacterial infection, which occurred more often among those receiving placebo than CTX. In this small sample selected for feasibility assessment, this difference did not reach statistical significance. The overall rate of infection observed here is similar to other observations in this patient population.24,37As expected, based on a wealth of clinical experience with CTX, the treatment was well-tolerated. Only 1 subject discontinued due to adverse reaction (nausea and pruritus, which resolved completely within 1 day after stopping treatment). The rate of C. difficile colitis was low overall, with only 1 subject (in the placebo arm) with this diagnosis. Similarly, results from a study of protocolized use of amoxicillin-clavulanate in patients with severe alcohol-associated hepatitis did not show an increased incidence of C. difficile.22 We did not observe any proven cases of CTX-resistant infection in this cohort. No fungal infections were observed in this study. We did not identify any concerns regarding unacceptable interference with the standard of care from the clinical teams caring for these patients.CRP and PCT were measured before initiation of the study intervention based on prior data suggesting that they may aid in assessing the likelihood of the development of infection. This was intended to assess their possible utility in subject selection for future study to minimize exposure to prophylactic antibiotics at a lower likelihood of benefit. However, although incident infection was not rare, the presence of CRP > 25 g/mL was uncommon in this cohort, limiting its utility for discrimination of very short-term infection risk. Alternatively, elevated PCT (>0.5 ng/mL) was relatively common (5/30 subjects); however, this did not associate with incident infection in our sample as none of these 5 subjects developed infection. Of note, only 1/5 of these subjects received CTX, suggesting that any potential beneficial effect of prophylaxis was not responsible for the lack of association between PCT and short-term infection incidence. The observed lack of utility of these markers is discordant with multiple prior studies suggesting they are associated with risk of infection; however, these studies typically reviewed the risk of infection among ambulatory patients over a longer period of exposure, for example, 3–6 months.28,35,38 The risk of infection in the hospital may depend on different factors than the risk among ambulatory patients, and the strength of predictive ability for these markers may be limited over such a short period of observation.Protocol-based study of changes in the gut microbiome may add to our understanding of the risk and mechanism of benefit; however, this does not currently reflect the standard of care overall and exceeds the scope of this pilot investigation. Other studies suggest that gut microbiota differences are associated with the risk of nosocomial infection in this patient population, and targeted intestinal treatments may merit study as an alternative to systemic antibiotic exposure.While the sample size here does not permit assessment of the true effect of antibiotic prophylaxis in this setting, there was a general tendency toward the potential protective effect of antibiotics (infection and death were numerically less frequent and change in laboratory markers, eg, Chronic Liver Failure Consortium acute on chronic liver failure score and MELD-Na were generally more favorable). This is consistent with experience among patients with variceal hemorrhage and alcohol-associated hepatitis in which antibiotics administered in the absence of a known infection have been shown to reduce the subsequent rate of infections.18,19,22 Based on our observations, it would require a sample of approximately 116 subjects in each arm to yield 80% power to detect a difference in incidence of infection of the magnitude observed in this pilot cohort. We observed a very small rate of attrition to study enrollment (1 subject did not receive any study treatment, and a second subject stopped treatment due to a nonsevere AE). Even assuming a greater frequency of dropout of 20%, 140 subjects per arm would be required. Enrollment was somewhat hindered during this pilot study due to coincidence with the COVID-19 pandemic. Future, larger studies with sample sizes designed to determine the efficacy of such a regimen may require multiple centers to expedite enrollment, which may also serve to better support the generalizability of findings.Strengths of this study include a pragmatic design intended to mimic standard clinical practice, randomized, double-blinded design, and reasonably broad inclusion criteria. Limitations include single-center enrollment and a small sample size intended for feasibility purposes, which precludes assessment of the effect of CTX prophylaxis in this population. It is possible that some infections could have been missed due to the absence of a protocol for longitudinal monitoring; however, infections of clinical importance should have been identified through standard of care. Even with pre-specified criteria, infectious may be covert and their definition and diagnosis can, at times, remain challenging. Providers may elect to treat patients with antibiotics when there is uncertainty to avoid a delay in potentially time-sensitive treatment. This could lead to overestimation of the incidence of true infections. Blinded, external adjudication may be required to optimize the assessment of this outcome in future studies. Changes in inflammatory markers over time and changes in microbiota may also be of interest but were not collected for this study.
CONCLUSIONS
We conducted a pilot study of the effect of CTX on prophylaxis of bacterial infection among hospitalized patients with cirrhosis. We found this to be associated with a relatively low rate of adverse events, and in this small sample, there was a suggestion of a potential for lower incidence of infection as well as several other favorable outcomes (eg, reduction in MELD-Na and ACLF score). However, this pilot study requires confirmation of these potential benefits in broader study with close monitoring for adverse effects. As with other settings for antibiotic prophylaxis, for example, variceal hemorrhage, and primary prophylaxis of SBP, it is likely there is a subgroup of patients at the highest risk of infection in whom antibiotic prophylaxis has the best benefit-to-risk ratio. Identification of this group remains a challenge, and attempts to optimally target antibiotic prophylaxis should continue. We conclude there is sufficient equipoise here with regard to the relative paucity of related adverse events observed and potential benefits such that a larger, definitive study should be undertaken.
SN
Dr. Sebastian Niezen
Gastroenterology · General GastroenterologyBeth Israel Deaconess Medical Centersebastiann