key: cord-0933188-5drcl59z authors: Zhao, Lina; Gao, Yanxia; Guo, Shigong; Lu, Xin; Yu, Shiyuan; Ge, Zengzheng; Zhu, Huadong; Li, Yi title: Prognosis of Patients with Sepsis and Non-Hepatic Hyperammonemia: A Cohort Study date: 2020-12-29 journal: Med Sci Monit DOI: 10.12659/msm.928573 sha: 5f327ec0caeedd64dea5ecbfbc8e3ca61fb75fbd doc_id: 933188 cord_uid: 5drcl59z BACKGROUND: Hyperammonemia has been reported in some critically ill patients with sepsis who do not have hepatic failure. A significant proportion of patients with non-hepatic hyperammonemia have underlying sepsis, but the association between non-hepatic hyperammonemia and prognosis is unclear. MATERIAL/METHODS: Information about patients with sepsis and non-hepatic hyperammonemia was retrieved from the Medical Information Mart for Intensive Care-III database. Survival rates were analyzed using the Kaplan-Meier method. Multivariate logistic regression models were employed to identify prognostic factors. Receiver operating characteristic (ROC) curve analysis was used to measure the predictive ability of ammonia in terms of patient mortality. RESULTS: A total of 265 patients with sepsis were enrolled in this study. Compared with the non-hyperammonemia group, the patients with hyperammonemia had significantly higher rates of hospital (59.8% vs. 43.0%, P=0.007), 30-day (47.7% vs. 34.8%, P=0.036), 90-day (61.7% vs. 43.7%, P=0.004), and 1-year mortality (67.3% vs. 49.4%, P=0.004). In the survival analysis, hyperammonemia was associated with these outcomes. Serum ammonia level was an independent predictor of hospital mortality. The area under the ROC curve for the ammonia levels had poor discriminative capacity. The hyperammonemia group also had significantly lower Glasgow Coma Scale scores (P=0.020) and higher incidences of delirium (15.9% vs. 8.2%, P=0.034) and encephalopathy (37.4% vs. 19.6%, P=0.001). Intestinal infection and urinary tract infection with organisms such as Escherichia coli may be risk factors for hyperammonemia in patients who have sepsis. CONCLUSIONS: Higher ammonia levels are associated with poorer prognosis in patients with sepsis. Ammonia also may be associated with sepsis-associated encephalopathy. Therefore, we recommend that serum ammonia levels be measured in patients who are suspected of having sepsis. Sepsis is a serious medical condition responsible for approximately 19 .77% of all deaths worldwide [1, 2] . The mortality is a result of the systemic inflammation and end-organ dysfunction associated with these infections [3] . The rate of mortality in patients diagnosed with sepsis is 30%, and 50% in individuals with severe sepsis. In patients in whom the disease progresses to septic shock, the mortality rate can rise to as high as 80%. As an individual's infection worsens, the risk of mortality gradually increases [4] . Sepsis-associated encephalopathy (SAE) can be found in up to 70% of patients with severe sepsis and it is a common neurological complication [5] , with a mortality rate of up to 70% [6] . Ammonia is a major factor in the pathogenesis of hepatic encephalopathy and it crosses the blood-brain barrier readily, resulting in significant neurotoxicity [7] . Disorders of ammonia metabolism can lead to hyperammonemia, which usually is a consequence of hepatic failure. Hyperammonemia also can occur in critically ill patients who do not have hepatic disease [8] , including individuals with sepsis, gastrointestinal bleeding, kidney failure, elevations in sodium, and exposure to valproate [8, 9] . In recent reports, serum ammonia has been suggested as a possible predictor of 28-day mortality and hospital stay in patients with sepsis. While elevation in ammonia level has been reported as a novel biomarker for sepsis [10, 11] , its roles in long-term prognosis and as a risk factor for non-hepatic hyperammonemia in patients with sepsis are unclear. The relationship between serum ammonia and the development of sepsis and its prognosis in patients with the condition remains under-explored. The aim of this study was to determine the significance of elevated serum ammonia levels to both the short-and long-term prognosis of patients with sepsis. We also explored risk factors for non-hepatic hyperammonemia in sepsis and the association between non-hepatic hyperammonemia and SAE. Database This was a retrospective study based on information recorded in the publicly available Medical Information Mart for Intensive Care (MIMIC-III) database between 2001 and 2012. Use of the database was approved by the Massachusetts Institute of Technology (Cambridge, Massachusetts, U.S.A.) and the Institutional Review Board of Beth Israel Deaconess Medical Center (Boston, Massachusetts, U.S.A.). Individual patient consent was not required because the study was a retrospective review of publicly available, anonymized data and the analysis did not affect the care of individual patients. The raw data were extracted using structure query language (SQL) with Navicat and further processed with R software. Inclusion criteria for the study were as follows: (1) a diagnosis of sepsis, severe sepsis, or septic shock according to International Classification of Diseases, Ninth Revision (ICD-9) codes; (2) age ³18 and £89 years; (3) admission for >24 hours in the intensive care unit (ICU); and documentation of blood ammonia levels. A blood ammonia level >35 μmol/L was defined as hyperammonemia in the MIMIC-III database. Exclusion criteria for the study were as follows: (1) a history of acute or chronic liver disease, including hepatitis, hepatic cirrhosis, hepatic encephalopathy, hepatorenal syndrome, hepatic injury, and other chronic liver disease, according to ICD-9 diagnosis codes on patient discharge (Supplementary Table 1) ; and (2) no documentation of vital signs or ICD-9 diagnostic codes. R statistical software (R Foundation for Statistical Computing, Vienna, Austria) was used to collect data on baseline characteristics information such as age, sex, and vital signs and laboratory parameters during the first 24 hours of ICU admission. The maximum value for ammonia during each patient's ICU stay also was retrieved. Infection type (Supplementary Table 2 ), microbiology type (Supplementary Table 3) , and patient comorbidities (Supplementary Table 4 ) were determined according to the primary ICD-9 codes, as documented in each patient's discharge summary. We retrieved the SQL scripts from the GitHub website (https://github.com/MIT-LCP/mimic-code/ tree/master/concepts/severity-scores) and used them to calculate the severity scores. Simplified Acute Physiology Score (SAPSII), Sequential Organ Failure Assessment (SOFA) score, and Glasgow Coma Scale (GCS) ratings also were recorded during the first 24 hours of each patient's ICU stay. Outcomes of patient conditions such as delirium, encephalopathy, mechanical ventilation, renal replacement therapy (Supplementary Table 5) , and survival status were recorded. Relevant information was obtained about patients who were diagnosed with "sepsis," "severe sepsis," and "septic shock" on discharge, according to ICD-9 codes (Supplementary Table 6 ). Patients were assigned to the hyperammonemia and non-hyperammonemia groups based on serum ammonia levels. They were also divided into conscious (GCS=15), sub-coma (GCS 9-14), and deep coma groups (GCS 3-8) based on GCS scores. The statistical analysis compared the hyperammonemia and non-hyperammonemia groups. Data distribution was tested using the Shapiro-Wilk test. Continuous variables were expressed as means with standard deviation for normal distributed data, and for non-normally distributed data, medians (interquartile range [IQR]) were expressed. Categorical variables were represented as frequencies with percentage and compared using a chi-square test. Variables with missing data are relatively common in the MIMIC-III database and we replaced them with median values (Supplementary Material 1). A non-parametric test (Mann-Whitney U or Kruskal-Wallis) was used for comparisons between the baseline characteristics and outcomes in the hyperammonemia and non-hyperammonemia groups and the relationship between serum ammonia and consciousness. Kaplan-Meier curves were analyzed using log-rank tests for comparison of hospital mortality between the hyperammonemia and non-hepatic hyperammonemia groups. A Cox regression model was used to screen for variables associated with hospital mortality in survivors versus non-survivors. A 2-tailed P<0.05 was considered statistically significant. All statistical analyses were performed with R software (version 3.4.3). The patient inclusion flowchart is shown in Figure 1 . A total of 2159 patients were tested for blood ammonia according to information in the MIMIC-III database. Using the inclusion and exclusion criteria, 1051 patients were identified for further screening. Of those patients, 265 were diagnosed with "sepsis," "severe sepsis," or "septic shock" on discharge, according to ICD-9 codes, and were enrolled in the study. The incidence of non-hepatic hyperammonemia was 40.4% with a 67.3% rate of 1-year mortality. Information on the patients' baseline characteristics, vital signs, laboratory parameters, infection type, microbiology type, and comorbid diseases is summarized in Table 1 . There were 107 patients in the hyperammonemia group and 158 patients in the non-hyperammonemia group. Patients in the hyperammonemia group had significantly more intestinal infections (23.4% vs. 13.3%, P=0.034) and urinary tract infections (UTIs) (45.8% vs. 24.7%, P<0.001) than patients in the non-hyperammonemia group. Patients with hyperammonemia were more likely to be infected with Escherichia coli (42.1% vs. 22.8%, P=0.001). Patients in the hyperammonemia group had lower GCS scores than patients in the non-hyperammonemia group (P=0.020). No correlation was found between ammonia levels and respiratory infection, gastrointestinal bleeding, heart failure, kidney failure, or infection in other tissues by E. coli. In addition, there were no significant differences in SAPSII or SOFA scores between the 2 groups. Table 2 shows the outcomes in the hyperammonemia and nonhyperammonemia groups. As illustrated, a greater proportion of patients in the hyperammonemia group were diagnosed with delirium (15.9% vs. 8.2%, P=0.034) and encephalopathy (37.4% vs. 19 .6%, P=0.001). Patients with hyperammonemia also had higher rates of short-and long-term mortality Excluded the following patients (n=1108) Combined liver disease (n=989) Hepatitis (n=464) Alcoholic cirrhosis (n=248) Cirrhosis without mention of alcohol (n=131) Biliary cirrhosis (n=S) Hepatic encephalopathy (n=63) Liver cancer (n=O ) Other liver disease (n=79) More than 18 years old and less than 89 years old (n=80) Lack of diagnoses lCD (n=37) Lack of vitals signs (n=2) 2159 Patients were tested for blood ammonia MIMIC III database 1051 Patients for further selection A total of 265 patients were diagnosed as "sepsis, "severe sepsis" or "septic shock" Hyperammonemia group n=107 Non-hyperammonemia group n=158 Hyperammonemia group n=107 Patients in the hyperammonemia group had worse survival rates (in-hospital, 90-day, and 1-year mortality) ( Figure 2 ). Furthermore, univariate and multivariate Cox analysis was performed of baseline variables (age and sex) and results of laboratory tests (alanine aminotransferase, aspartate aminotransferase, creatinine, blood urea nitrogen, hemoglobin, platelet count, partial thromboplastin time, international normalized ratio, prothrombin time, white blood cell count, and ammonia). The factors significantly correlated with survival were adjusted for in the multivariate analysis. The analysis revealed that ammonia remained an independent prognostic factor in patients with sepsis. (P<0.01 or P<0.05) ( Table 3 ). Receiver operating characteristic curves of ammonia indices for predicting mortality To further confirm the reliability of ammonia, we plotted the area under the receiver operating characteristic (ROC) curve for 90-day and 1-year survival, and in-hospital mortality. The discriminative ability of ammonia levels based on the ROC curve analysis was 0.625 for in-hospital mortality, 0.620 for 90-day survival, and 0.624 for 1-year survival ( Figure 3 ). Patients were divided into conscious (n=109), sub-coma (n=112), and deep coma groups (n=44) based on GCS score. As shown in Figure 4 , patients with lower GCS scores had higher serum ammonia levels. The serum ammonia levels were highest in the deep coma group, compared with the other 2 groups (P<0.001), and they were significantly higher in the subcoma group than in the conscious group (P<0.001) (Figure 4 ). Our study demonstrated that the incidence of non-hepatic hyperammonemia is 40.4% in patients with sepsis and the incidence of sepsis with encephalopathy in patients with non-hepatic hyperammonemia is 37.4%. Serum ammonia level may be a predictor of mortality in patients with sepsis who do not have hepatic disease. In addition, we found that intestinal infection, UTI, and infections in other tissues caused by E. coli were risk factors for nonhepatic hyperammonemia in patients with sepsis. We also found that the rate of hospital mortality in patients with sepsis who had non-hepatic hyperammonemia was 59.8%, which was significantly higher than in patients with sepsis who had normal serum ammonia levels (46.4%) [1] . A higher serum ammonia level may be a risk factor for mortality. Our results are consistent with the findings of Zhao et al., which showed that in patients with sepsis, an increased serum ammonia level on admission to the emergency department was correlated with an increased rate of mortality at 28 days. Our study explored mortality levels for up to 1 year, and we found that serum ammonia is an independent risk factor for long-term prognosis in patients with sepsis. In a case series, McEwan et al. suggested that higher serum ammonia levels are related to adverse clinical outcomes, which correlates with our findings. However, Zhao et al. showed that serum ammonia levels had a robust ability to predict the 28-day mortality rate in patients with sepsis, with an area under the ROC curve of 0.813, which is in contrast to our findings. That discrepancy may be attributable to differences in basic patient characteristics between the 2 studies. It suggests that serum ammonia level may be a new prognostic marker for patients with sepsis. An interesting finding in our study is that non-hepatic hyperammonemia may be associated with an increased risk of SAE [12] . SAE is mainly characterized by symptoms of delirium with changes in a patient's consciousness, and it also can lead to coma [13] . Our study demonstrated that patients with hyperammonemia had lower GCS scores. In the absence of previous cerebrovascular and encephalopathic brain disease, SAE is more likely to occur as the serum ammonia level increases. SAE is a diffuse brain dysfunction that occurs secondary to sepsis in the body without overt infection of the central nervous system. Its pathogenesis is multifaceted and is attributed to a combination of astrocyte swelling, an increase in glutamine synthesis, and a disproportionate ratio of aromatic amino acids to branched chain amino acids [14] [15] [16] . Based on our study results, we hypothesize that non-hepatic hyperammonemia may be associated with SAE. Unfortunately, in our study, some primary brain diseases (such as cerebral hemorrhage and cerebral infarction) and some secondary brain diseases (such as metabolic encephalopathy and pulmonary encephalopathy) were not excluded. The association between non-hepatic hyperammonemia and SAE needs to be validated in future well-designed experimental trials. Intestinal infection, UTI, and infection of other tissues by E. coli may be risk factors for non-hepatic hyperammonemia in patients with sepsis. Our results showed that the incidence of intestinal infections in the hyperammonemia group was 23.4% higher than in the non-hyperammonemia group. This is consistent with research by Wang et al., which found that in patients with infection-induced hepatic encephalopathy, levels of plasma ammonia were significantly higher in association with intestinal tract infection compared with other sites of infection. Their results, along with our findings, support the notion that intestinal infection is related to hyperammonemia [17] . A possible explanation for the link between intestinal infection and non-hepatic hyperammonemia is intestinal flora. Colonic bacteria have been known to produce ammonia from amino acid deamination or via urease, the hydrolysis of urea into carbon dioxide and ammonia [18] . When the body develops sepsis, the composition of intestinal microbes changes, due to factors such as antibiotic usage, systemic inflammation, and intestinal leakage [19] . In the patient's feces, the composition of the microbial components changes rapidly, the microbial diversity is largely lost, and the proportion of anaerobic bacteria significantly reduced and of Enterobacteriaceae increased [20] . Ammonia production is increased by converting nitrate to nitrite, and subsequently to ammonia [21] . Our results are consistent with previous studies, in which an increase in ammonia was associated with higher rates of infection by Enterobacteriaceae [3, 13, 22] . Therefore, serum ammonia should be measured when risk factors are present, such as intestinal infection or infection by E. coli. Our study showed that UTI is significantly associated with non-hepatic hyperammonemia in patients with sepsis, which is in line with the literature [23] [24] [25] . The possible explanation for the link between non-hepatic hyperammonemia and UTI is urease-producing bacteria and distal renal tubular acidosis [26] . With the entry of urea into the urinary tract, urease-producing bacteria form "ammonia," which results in alkalinization of the urine. The pH of the urine, when relatively high compared with that of the blood, enhances the diffusion of "ammonia" into the bloodstream [27, 28] . Another plausible explanation for the linkage between hyperammonemia and UTI is distal renal tubular acidosis. Severe UTIs occasionally are accompanied by altered distal renal tubular function, which results in reduced bicarbonates, and in turn, leads to increased renal "ammonia" production [29] . The last explanation could be urinary retention associated with a neurogenic bladder. As the pressure in the bladder increases, the area of the bladder expands and promotes drainage of more ammonia directly into the inferior vena cava via the internal iliac veins [30] . Therefore, in patients with UTIs, serum ammonia levels should be closely monitored and timely measures taken to reduce them. Several limitations of the present study must be acknowledged. First, the result suggests a link between higher serum ammonia levels and lower GCS scores. Because of the nature of the retrospective analysis, the onset times of coma were not always available or documented, and some patients with primary and secondary encephalopathy in this study were not excluded. Therefore, whether there is a causal relationship between ammonia and SAE cannot be determined based on our results. Second, due to the limitations of the database, information was missing on some clinical variables, such as bilirubin, albumin, and intravenous nutrition. Inclusion of those data may have led to a more comprehensive understanding of the role of other biomarkers in sepsis with non-hepatic hyperammonemia. Third, our cohort study used ICD-9 diagnostic codes for sepsis, severe sepsis, and septic shock, but the e928573-8 concept of severe sepsis was eliminated in Sepsis 3.0, which may have led to bias in our research results. Non-hepatic hyperammonemia is associated with mortality in patients with sepsis. The present study was essentially a pilot that requires validation. We recommend that serum ammonia levels be measured in patients who have risk factors, such as intestinal infection, UTI, and E. coli infection. Infection caused by E. coli is a potential biomarker for sepsis in patients who have non-hepatic hyperammonemia. Our study also demonstrated a correlation between non-hepatic hyperammonemia and an increased risk of SAE. Global, regional, and national sepsis incidence and mortality, 1990-2017: Analysis for the Global Burden of Disease Study Efficacy and safety of corticosteroids for septic shock in immunocompromised patients: A cohort study from MIMIC An engineered E. coli Nissle improves hyperammonemia and survival in mice and shows dose-dependent exposure in healthy humans Assessing available information on the burden of sepsis: Global estimates of incidence, prevalence and mortality Sepsis-associated encephalopathy: from delirium to dementia? Sepsis-associated encephalopathy: A vicious cycle of immunosuppression Hepatic encephalopathy: A critical current review Characteristics and outcomes of critically ill patients with severe hyperammonemia Features of adult hyperammonemia not due to liver failure in the ICU Serum ammonia levels on admission for predicting sepsis patient mortality at D28 in the emergency department: A 2-center retrospective study Ammonia vs. lactic acid in predicting positivity of microbial culture in sepsis: The ALPS pilot study Sepsis-associated encephalopathy is septic All that glitters is not gold': Wen hyperammonaemia is not from hepatic aetiology Mechanistic insight, diagnosis, and treatment of ammonia-induced hepatic encephalopathy Ammonia levels and hepatic encephalopathy in patients with known chronic liver disease Prevention and treatment of hepatic encephalopathy A study on the position and etiology of infection in cirrhotic patients: A potential precipitating factor contributing to hepatic encephalopathy Ammonia production by intestinal bacteria The intestinal microenvironment in sepsis The role of the gut microbiota in sepsis Nitrate reduction to nitrite, nitric oxide and ammonia by gut bacteria under physiological conditions A novel Two-domain architecture within the amino acid kinase enzyme family revealed by the crystal structure of Escherichia coli glutamate 5-kinase Relapsing hypoglycemia associated with hypocarnitinemia following treatment with cefcapene pivoxil in an elderly man Noncirrhotic hyperammonemia after deceased donor kidney transplantation: A case report Hyperammonemia associated with distal renal tubular acidosis or urinary tract infection: A systematic review A case of hyperammonemia with obstructive urinary tract infection by urease-producing bacteria A Case of hyperammonemia caused by urinary tract infection due to urease-producing bacteria Hyperammonemia in distal renal tubular acidosis: A new case and review of the literature Hyperammonaemic coma in ureterosigmoid urinary diversion None. The MIMIC-III database (version 1.4) is publically available from https://mimic.physionet.org/. Any researcher who adheres to the data use requirements is permitted access to the database. The use of the database was approved by the Massachusetts Institute of Technology (Cambridge, Massachusetts, U.S.A.) and the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, Massachusetts, U.S.A.). ICD-9 -International Classification of Diseases, Ninth Revision.