www.elsevier.com/locate/ajem

American Journal of Emergency Medicine (2013) 31, 8–15

Original Contribution

Prognostic values of blood ammonia and partialpressure of ammonia on hospital arrival inout-of-hospital cardiac arrests☆,☆☆,☆☆☆,★,★★

Chih-Hao Lin MDa,⁎, Chih-Hsien Chi MD, PhDa, Shyu-Yu Wu MSb,Hsiang-Chin Hsu MDa, Ying-Hsin Chang MDa, Yao-Yi Huang MDa, Chih-Jan Chang MDa,Ming-Yuan Hong MDa, Tsung-Yu Chan MDa, Hsin-I Shih MDa

aDepartment of Emergency Medicine, National Cheng Kung University Hospital, College of Medicine,National Cheng Kung University, Tainan 70403, TaiwanbDepartment of Electrical Engineering, National Sun Yat-Sen University, Kaohsiung 80424, Taiwan

Received 4 March 2012; revised 4 April 2012; accepted 28 April 2012

AbstractPurposes: Outcome prediction for out-of-hospital cardiac arrest (OHCA) is of medical, ethical, andsocioeconomic importance. We hypothesized that blood ammonia may reflect tissue hypoxia in OHCApatients and conducted this study to evaluate the prognostic value of ammonia for the return ofspontaneous circulation (ROSC).Methods: This prospective, observational study was conducted in a tertiary university hospital betweenJanuary 2008 and December 2008. The subjects consisted of OHCA patients who were sent to theemergency department (ED). The primary outcome was ROSC. The prognostic values were calculatedfor ammonia levels and the partial pressure of ammonia (pNH3), and the results were depicted as areceiver operating characteristics curve with an area under the curve.Results: Among 119 patients enrolled in this study, 28 patients (23.5%) achieved ROSC. Ammonialevels and pNH3 in the non-ROSC group were significantly higher than those in the ROSC group (167.0μmol/L vs 80.0 μmol/L, P b .05; 2.61 × 10−5 vs 1.67 × 10−5 mm Hg, P b .05, respectively). Thepredictive capacity of area under the curve for ammonia and pNH3 for non-ROSC was 0.85 (95%confidence interval, 0.75-0.95) and 0.73 (95% confidence interval, 0.61-0.84), respectively. Themultivariate analysis confirmed that ammonia and pNH3 are independent predictors of non-ROSC. Theprognostic value of ammonia was better than that of pNH3. The cutoff level for ammonia of 84 μmol/Lwas 94.5% sensitive and 75.0% specific for predicting non-ROSC with a diagnostic accuracy of 89.9%.

☆ Disclosures of conflict of interest: The authors disclose no conflicts.☆☆ Ethical adherence: The study procedures were in accordance with the ethical standards and were approved by the institutional review board in the

hospital.☆☆☆ Funding and support: No financial support was granted.★ Writing assistance: None.★★ Author contributions: CHL conceived and supervised the study. All authors were involved in acquisition of data. CHL and SYW interpreted the data

and performed statistical analysis. CHL drafted the manuscript, and all authors contributed substantially to its revision. CHL and CHC offered administrative andtechnical supports. CHL is the corresponding author who takes responsibility for the manuscript as a whole.

⁎ Corresponding author. Tel.: +886 6 2353535x2237, +886 932989778; fax: +886 6 2359562.E-mail addresses: emergency.lin@gmail.com (C.-H. Lin).

0735-6757/$ – see front matter © 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.ajem.2012.04.037

9Prognostic values of blood ammonia and partial pressure of ammonia

Conclusions: Hyperammonemia on ED arrival is independently predictive of non-ROSC for OHCApatients. The findings may offer useful information for clinical management.© 2013 Elsevier Inc. All rights reserved.

1. Introduction

Outcome prediction for out-of-hospital cardiac arrest(OHCA) is of medical, ethical, and socioeconomic impor-tance. The early management of OHCA is complicated bythe lack of readily available prognostic predictors of thepossible spontaneous return of circulation (ROSC). Highersurvival rates have been observed in patients withwitnessed ventricular tachycardia/ventricular fibrillation,bystander cardiopulmonary resuscitation (CPR), and earlydefibrillation [1-4]. Several assessment tools or biomarkers,including Glasgow Coma Score [5], brain stem reflex [6],somatosensory-evoked potentials [7], end-tidal carbondioxide [8,9], S-100 protein [10], neuron-specific enolase[11], serum glucose [12], serum lactate [13], and brainnatriuretic peptide [14], have been used to evaluate theoutcome of cardiac arrest after ROSC. However, no singlebiomarker has been identified to reliably predict whetherthe cardiac arrest patient can or cannot achieve ROSC.

Hepatic dysfunction is a common finding in criticallyill patients and is associated with longer intensive careunit stays and increased hospital mortality [15]. Hypoxichepatitis, also known as ischemic hepatitis or shock liver,is caused by the insufficient uptake of oxygen by thehepatocytes. Hypoxic hepatitis causes several complica-tions, such as spontaneous hypoglycemia, respiratoryinsufficiency, and hyperammonemia. Blood ammonia ispredictive of mortality for hospitalized patients with acutehepatic failure and can be used for risk stratification[16,17]. Recent studies have also suggested that bloodammonia levels at hospital arrival predict the neurologicoutcome of patients with OHCA after ROSC [18]. Wehypothesized that blood ammonia levels may reflect tissuehypoxia and that blood ammonia levels can thus be usedto predict the likelihood of ROSC achievement forOHCA patients.

Ammonium (NH4+), also known as ammonium ion, is an

ionized form of ammonia (NH3). Ammonia exists predom-inantly as ammonium in the blood, whereas a small pH-dependent fraction exists in a nonionized form (b2%). Thenonionized form of ammonia, reflected by the partialpressure of ammonia (pNH3) [19], permeates cell mem-branes and is more toxic than the ionized form of ammonia.Because acidemia is one of the causes of cardiac arrests [20]and a decrease in pH may influence the value of pNH3,pNH3 should also be considered a predictive tool for ROSC.Blood ammonia and blood gas analysis are diagnostic teststhat are available in most emergency department (ED)settings. We conducted this study to evaluate the relationship

between ROSC and blood ammonia as well as pNH3 inOHCA patients.

2. Materials and methods

2.1. Study setting

This prospective, observational study was conducted inthe ED of a tertiary university hospital in Taiwan betweenJanuary 2008 and December 2008. The subjects of the studyconsisted of consecutive OHCA patients sent to the ED.Patients with a known pregnancy, severe hypothermia(defined as a body temperature b30°C), a valid do-not-attempt-resuscitation order, or obvious signs of irreversibledeath and patients who were younger than 18 years old wereexcluded from the study. The signs of irreversible deathincluded decapitation, hemicorporectomy, dependent lividi-ty, rigor mortis, decomposition, and thermal carbonizationwithout detectable vital signs. There were no rules concerningthe termination of resuscitation in prehospital settings inTaiwan at the time of our study. All patients with witnessedand unwitnessed cardiac arrests assessed by emergencymedical technicians (EMTs) were sent to hospitals unlessobvious signs of irreversible death were present.

Demographic data were obtained from a prospectiveregistry database. Basic and advanced life support measureswere carried out by EMTs at the scene and/or emergencyphysicians in the ED according to standard protocols. Thedata comprise all of the information required for theinternational Utstein-style criteria [21,22], such as thepatient's history, cardiac risk factors, response intervals,initial cardiac rhythms, and the extent and amount ofemergency care. The definitive cause of cardiac arrest wasdocumented on discharge from the hospital or after thepatient's death in the hospital.

Emergency medical technicians may place intravenouslines, use advanced airway management devices, administerdoses of epinephrine, and apply automated externaldefibrillators for OHCA patients before ED arrival. Basicand advanced cardiovascular life support was providedaccording to the 2005 American Heart AssociationGuidelines for Cardiopulmonary Resuscitation and Emer-gency Cardiovascular Care [23] during the study period.Resuscitation was terminated if the patient exhibitedpersistent asystole after 30 minutes of advanced cardiovas-cular life support. All patients with sustained ROSC wereadmitted to the intensive care units and received standardintensive care treatment.

10 C.-H. Lin et al.

The study procedures were in accordance with the ethicalstandards and were approved by the institutional reviewboard of the hospital.

2.2. Assessment of the blood ammonia levelsand the pNH3

Blood samples were collected from the femoral vessels ofOHCA patients within 5 minutes after ED arrival. The bloodsample of each patient was divided into several tubes orsyringes for analysis.

The blood samples in tubes containing lithium heparinwere cooled in ice-cold water before analysis and weretransported to the clinical laboratory immediately. Ammonialevels were measured using Vitros NH3 DT Slides withVitros Fusion 5.1 FS Automated Chemistry Analyzer (OrthoClinical Diagnostics, Johnson & Johnson Co., NJ). TheVitros NH3 DT Slide is a multilayered, analytical elementthat is coated on a polyester support. Water and nonprotei-naceous components travel to the underlying bufferedreagent layer, and the ammonium ions are converted togaseous ammonia. The semipermeable membrane allowsonly ammonia to pass through, which prevents buffer orhydroxyl ions from reaching the indicator layer and thusavoids the influence of possible acidemia. The detectionrange for blood ammonia in the assay was 1 to 500 μmol/L,and the institutional reference range was 9 to 33 μmol/L.Blood ammonia levels of 500 μmol/L were consideredbeyond the upper limits of detection.

The blood gas samples in Luer slip syringes containingheparin were analyzed to obtain the pH value. The bloodgas analysis was determined using i-STAT EG7+ cartridgeswith i-STAT System (Abbott Point of Care Inc., NJ). Thedetection range for blood pH was 6.5 to 8.2. Combining themass balance equation and Henry's law [24], the partialpNH3 can be calculated as follows: p[NH3] = (KH ×[NH3]T)/(1 + [H+]/Ka), where [NH3]T = total bloodammonia, KH for ammonia = 1.75 × 10−3, Ka = 9.8 ×10−10, and [H+] is derived from the pH values as theantilog (−pH).

The patients who did not follow the aforementioned bloodsampling procedures were excluded. Patients whose labora-tory results were not available were also excluded. Thephysicians who provided medical care during the studyperiod had no knowledge of the study results. Therapeuticdecisions were not biased by assay measurements.

2.3. Outcome measures

The primary outcome was ROSC. Return of spontaneouscirculation was defined as an organized electrocardiogramwith a palpable pulse. Patients with the return of a palpablepulse for less than 5 minutes were considered to not haveachieved ROSC.

2.4. Statistical analysis

Data were collected by the designated registered nursesand were entered into an Excel (Microsoft Corp,Redmond, WA) spreadsheet. The study physicians per-formed manual checks on the accuracy of data entrymonthly for further analysis.

The skewness of data distribution was checked beforethe analysis. Categorical variables are shown as a numberwith a percentage, and quantitative data are shown as amedian value with an interquartile range (IQR). TheROSC achievement group and the non-ROSC group werecompared using the Mann-Whitney U test for continuousvariables and the χ2 test or Fisher test for discretevariables, wherever applicable. The sensitivity and spec-ificity of the prediction of non-ROSC were calculated fordifferent cutoff values of blood ammonia levels and thepNH3, and the results were depicted as a receiveroperating characteristics (ROC) curve with an area underthe curve (AUC).

We carried out univariate analysis of the clinical andbiochemical variables that could influence ROSC. Thevariables available on ED admission were considered forinclusion in the logistic regression. The significant variableswere dichotomized using discriminant values derived byconstructing ROC curves for each variable. The odds ratiowith a 95% confidence interval (95% CI) was calculated foreach variable. The predicted probability of non-ROSC wasderived from the variables found to be significant afterlogistic regression analysis.

Two-tailed P b .05 was considered significant. We usedthe SPSS software (version 17; SPSS Inc, Chicago, IL) forthe statistical analysis.

3. Results

A total of 175 consecutive OHCA patients sent to the EDduring the study period were evaluated for inclusion. Fifty-six patients were excluded because of an age of younger than18 years (n = 4), obvious signs of irreversible death (n = 12),severe hypothermia (n = 2), valid do-not-attempt-resuscita-tion orders (n = 16), inconsistent blood sampling procedures(n = 14), or unavailable laboratory data (n = 8). A final totalof 119 OHCA patients (60 men and 59 women) wereincluded in this study. The median age of the includedpatients was 74.0 years.

Of the 119 patients, 28 patients (23.5%) achieved ROSC,and 91 (76.5%) did not. Table 1 compares the baselinecharacteristics between the 2 groups. The 10 clinical andbiochemical variables that proved to be significantlydifferent between the 2 groups are as follows: the etiologyof cardiac arrest, initial cardiac rhythm, the presence ofbystander CPR, ammonia, pNH3, alanine aminotransferase(ALT), creatinine, potassium, pH, and PCO2.

Table 1 Clinical and biochemical parameters among patients with OHCA

Total OHCA Non-ROSC group ROSC group P

Total 119 91 28Age (y) 74 (61.0-80.5) 74.0 (59.0-81.0) 73.5 (63.5-79.0) .856Sex (male) 60 (50.4%) 47 (51.6%) 13 (46.4%) .927Etiology (nontrauma) 94 (79.0%) 70 (76.9%) 24 (85.7%) b.05Witness cardiac arrest 49 (41.1%) 33 (36.3%) 16 (57.1%) .054Bystander CPR 10 (8.4%) 8 (8.79%) 2 (7.14%) b.05Initial cardiac rhythm(shockable)

28 (23.5%) 18 (15.1%) 10 (35.7%) b.05

Ammonia (μmol/L) 150.0 (108.0-187.5) 167.0 (120.0-190.0) 80.0 (47.5-96.0) b.05pNH3 (10−5 mm Hg) 2.38 (1.67-3.47) 2.61 (1.88-3.81) 1.67 (0.89-2.31) b.05White blood counts(10−3/μL)

10.7 (7.8-14.2) 10.5 (7.6-13.8) 11.7 (8.8-17.2) .319

Hemoglobin level (g/dL) 11.3 (8.6-13.6) 11.2 (8.6-13.4) 11.5 (8.5-14.7) .694AST (U/L) 81 (42.5-208.0) 84 (44.0-270.5) 52 (34.5-111.0) .06ALT (U/L) 39 (19-131.5) 42 (21.5-151.0) 24 (15.0-63.0) b.05Creatinine (mg/dL) 1.7 (1.1-2.6) 1.6 (1.1-2.3) 2.4 (1.4-4.2) b.05Sodium (mmol/L) 141.0 (137.0-144.0) 142.0 (138.0-145.0) 140.5 (134.5-142.5) .110Potassium (mmol/L) 5.6 (4.6-7.3) 5.8 (4.6-7.5) 4.9 (4.3-6.2) b.05Glucose (mg/dL) 148.0 (89.5-211.5) 142.0 (82.0-212.0) 165.0 (118.0-202.0) .497pH 7.0 (6.9-7.2) 7.0 (6.8-7.2) 7.1 (7.0-7.2) b.05PO2 (mm Hg) 20.0 (13.0-38.7) 20.0 (12.0-37.0) 21.9 (14.5-71.0) .145PCO2 (mm Hg) 67.7(49.2-89.7) 71.6(55.0-100.0) 55.3(44.0-69.9) b.05Bicarbonate (mmol/L) 19.5 (12.7-23.9) 20 (13.1-24.3) 18.6 (12.4-21.5) .34BE (mmol/L) −10 (−19.0 to −6.0) −9.7 (−19.5 to −6.0) −12 (−17 to −5.6) .86

Categorical variables were given as number (percentage), whereas quantitative data were given as median (interquartile range). P value comparingnon-ROSC achievement group and ROSC achievement group less than .05 was considered significant. AST indicates aspartate aminotransferase; ALT,alanine aminotransferase; BE, base excess.

11Prognostic values of blood ammonia and partial pressure of ammonia

3.1. Ammonia levels and pNH3 values at ED arrival

The results of blood gas analysis and ammonia levels wereobtained within 2 and 15 minutes after blood sampling,respectively. The median ammonia level for OHCA patientsat ED arrival was 150 μmol/L (IQR, 108-187.5 μmol/L). Theblood ammonia levels of 115 patients (96.6%) exceeded theupper institutional normal limit of 33 μmol/L.

Ammonia levels and pNH3 values in the non-ROSCgroup were significantly higher than the levels in the ROSCachievement group (P b .05), as shown in Fig. 1. Themedian ammonia level was 167 μmol/L (IQR, 120-190μmol/L) in the non-ROSC group and 80 μmol/L (IQR,47.5-96.0 μmol/L) in the ROSC achievement group. Themedian pNH3 value was 2.61 × 10−5 mm Hg (IQR, 1.88 ×10−5-3.81 × 10−5 mm Hg) in the non-ROSC group and1.67 × 10−5 mm Hg (IQR, 0.89 × 10−5-2.31 × 10−5 mmHg) in the ROSC achievement group.

3.2. The prognostic value of ammonia levels

An ROC curve was plotted with the ammonia levels ofeach patient as the independent variable and non-ROSC asthe outcome variable, as shown in Fig. 2A. The AUC was0.85 (95% CI, 0.75-0.95). The optimal cutoff level forammonia was 84 μmol/L and was found to be 94.5%

sensitive and 75.0% specific for predicting non-ROSC with adiagnostic accuracy of 89.9%. The pretest odds ratio was 3.2,and the posttest odds ratio was 12.3.

3.3. Prognostic value of pNH3

The ROC curve for pNH3 values and non-ROSC is shownin Fig. 2B, with an AUC of 0.73 (95% CI, 0.61-0.84). Theoptimal cutoff pNH3 value of 1.89 × 10−5 mm Hg was foundto be 74.7% sensitive and 71.4% specific for predicting non-ROSC with a diagnostic accuracy of 73.9%. The posttestodds ratio was 8.5.

3.4. Comparison of the non-ROSC and the ROSCachievement group

The univariate analysis showed that 10 clinical andbiochemical variables significantly influenced ROSCachievement. The results are shown in Table 1. Thesevariables were analyzed using logistic regression analysiswith non-ROSC as the dependent variable. These variableswere dichotomized for best discrimination between the non-ROSC andROSC achievement groups using the cutoff valuesderived from the construction of individual ROC curves, asshown in Table 2. The following 5 variables were found to beindependent predictors of non-ROSC: blood ammonia,

Fig. 1 Comparison of blood ammonia levels (A) and the partial pressures of ammonia (B) of the non-ROSC group and the ROSC group.

12 C.-H. Lin et al.

pNH3, ALT, potassium, and PCO2. The remaining variables,which include the etiology of the cardiac arrest, the presenceof bystander CPR, the initial cardiac rhythm, creatininelevels, and pH values, either had no effect or lost theirindependent predictive capacity in the multivariate model.

The blood ammonia level, with an assigned cutoff valueof 84 μmol/L, had a much higher odds ratio (51.6; 95% CI,14.9-178.8) compared with the other variables.

Fig. 2 A ROC curve with an AUC for non-ROSC was depicted for b

3.5. Comparative prognostic value of ammonialevels and pNH3

The prognostic value of blood ammonia and pNH3 at EDarrival was analyzed in the 119 OHCA patients. The findingsfrom the ROC curves constructed with non-ROSC as theoutcome variable showed that the AUC was greater forammonia levels than for pNH3 values (0.85 vs 0.73). The

lood ammonia levels (A) and the partial pressure of ammonia (B).

Table 2 Clinical and biochemical variables significantly influencing achievement of ROSC were dichotomized using discriminantvalues derived by constructing ROC curves for each variable

Non-ROSC group (n = 91) ROSC group (n = 28) P Odds ratio (95% CI)

Etiology .32 3.2 (1.0-10.3)Nontrauma 70 (76.9%) 24 (85.7%)Trauma 21 (23.1%) 4 (14.3%)Bystander CPR .78 1.3 (0.3-6.3)Absent 83 (91.2%) 26 (92.6%)Present 8 (8.8%) 2 (7.4%)Initial cardiac rhythm .08 0.4 (0.2-1.1)Nonshockable 73 (80.2%) 18 (64.3%)Shockable 18 (19.8%) 10 (35.7%)Ammonia (μmol/L) b.05 51.6 (14.9-178.8)b84.0 5 (5.5%) 21 (75.0%)≥84.0 86 (94.5%) 7 (25.0%)pNH3 (10

−5 mm Hg) b.05 7.4 (2.9-19.0)b1.89 23 (25.3%) 20 (71.4%)≥1.89 68 (74.7%) 8 (28.6%)ALT (U/L) b.05 3.1 (1.3-7.6)b36.5 37 (40.7%) 19 (67.9%)≥36.5 54 (59.3%) 9 (32.1%)Creatinine (mg/dL) .61 1.4 (0.3-6.0)b0.85 7 (7.7%) 3 (10.7%)≥0.85 84 (92.3%) 25 (89.3%)Potassium (mmol/L) b.05 2.9 (1.2-7.2)b5.5 38 (42.8%) 19 (67.9%)≥5.5 53 (58.2%) 9 (32.1%)pH .33 N/Ab7.36 88 (96.7%) 28 (100%)≥7.36 3 (3.3%) 0 (0%)PCO2 (mm Hg) b.05 11.6 (2.6-52.0)b78.4 48 (52.7%) 26 (92.9%)≥78.4 43 (47.3%) 2 (7.1%)

Odds ratio with 95% CI for each variable was listed. N/A indicates not applicable.

13Prognostic values of blood ammonia and partial pressure of ammonia

diagnostic accuracy and posttest odds ratio of ammonialevels for non-ROSC were also better than those of pNH3

(89.9% vs 73.9% and 12.3 vs 8.5, respectively). Themultivariate analysis confirmed that the odds ratio ofammonia for non-ROSC was much higher than that ofpNH3 (51.6 vs 7.4), as shown in Table 2.

4. Discussion

We found that the ammonia levels at ED arrival weresignificantly higher among non-ROSC patients than amongROSC patients. The optimal cutoff level for ammonia thatwas predictive for non-ROSC in this study was 84 μmol/L.

Our findings are consistent with the results of previousstudies. Nagamine [25] found that a blood ammonia level ofless than 180 μg/dL (105.6 μmol/L) was predictive for fullneurologic recovery in witnessed OHCA patients. Yanagawaet al [26] showed that ammonia levels were significantlyhigher among the survivors with higher cerebral performancecategory (CPC) scores 1 month after cardiac arrest thanamong survivors with lower CPC scores (327 vs 124 μg/dL,

P = .001). Shinozaki et al [18] also reported that a bloodammonia level above 170 μg/dL (99.8 μmol/L) wasindependently predictive of higher CPC scores 6 months ofcardiac arrest.

The pKbh of ammonia, which is approximately 8.90 at37 °C [24], is close to the normal blood pH of 7.4. Anychange in blood pH would significantly affect the ratio ofnonionized to ionized ammonia. A greater percentage ofammonia may enter the brain as the blood pH rises becauseof an increase in the amount of ammonia present in itsnonionized form. Bhatia et al [17] demonstrated that anarterial ammonia level greater than over equal to 124 μmol/L had a diagnostic accuracy of 77.5% for predictingmortality in patients with acute hepatic failure. However,their study showed no significant difference in prognosticvalues between the total blood ammonia level and pNH3. Inour study, we found that the blood ammonia level wassuperior to pNH3 for predicting non-ROSC among OHCApatients. The blood ammonia level alone may then sufficefor predictive purposes.

Hyperammonemia may result from an increased pro-duction of ammonia. In mammals, skeletal muscle and

14 C.-H. Lin et al.

intestinal mucosa are major contributors to ammoniaproduction [27]. Protein is broken down, and glutamate istransaminated to form glutamine in the skeletal muscles.Glutamine is the temporary storage form of waste nitrogenand is a major source of ammonia if deaminated byglutaminase. In the intestinal mucosa, ammonia is producedafter the uptake of amino acids because of glutaminedeamination. 5′-adenosine monophosphate deaminase,which can destroy adenosine [28,29] and cause systolicarrest or arrhythmia [30], generates ammonia fromadenosine monophosphate during muscle contraction andrigor mortis [31]. Nagamine [25] found that the bloodammonia level of OHCA patients at hospital arrival waspositively correlated with the time elapsed from confirma-tion of cardiac arrest to hospital arrival. Previous studieshave also shown a strong correlation between the processesof blood lactate and ammonia production in anaerobicexercise [32,33]. Ishida et al [34] reported that thedevelopment of respiratory and metabolic acidosis duringcardiac arrest may induce the release of ammonia from redblood cells. Because of these findings, we hypothesize thathyperammonemia may result from an increase in theproduction of ammonia when the tissues are in a state ofoxygen depletion due to cardiac arrest. The amount ofammonia production may correlate with the time elapsedfrom cardiac arrest. The blood ammonia level can then bepredictive of non-ROSC of OHCA patients.

Hyperammonemia may also result from a decrease inthe elimination of ammonia. The liver and kidney areresponsible for the detoxification of portal ammonia andthe export of ammonia. In the kidney, ammonia isformed from glutamine deamination. However, renalammonia production mainly contributes to the bufferingof H+ ions, whereas the excretion of ammonia in urineplays only a minor role in the overall ammoniadetoxification [27]. The liver plays a vital role in thedetoxification of ammonia because the urea cycle locatedin periportal hepatocytes detoxifies a vast amount ofsurplus nitrogen [35]. Pneumatosis intestinalis and hepaticportal venous gas, which most commonly occur second-ary to intestinal ischemia and necrosis, have beenobserved in patients after CPR [36,37] and are associatedwith grave outcomes [38]. The poor mesenteric and portalperfusion in patients during cardiac arrest even with CPRmay cause hypoxic hepatocytic necrosis and thus impairammonia elimination. Pneumatosis intestinalis may affectthe intestinal microorganisms with enzymes, which mayenable protein and urea degradation and lead tohyperammonemia [39].

The pathogenesis of hyperammonemia may includeimpaired bioenergetics, altered neurotransmission, gluta-mate-mediated excitotoxicity, electrophysiologic derange-ments, oxidative and nitrosative stress, and mitochondrialdysfunction [40-43]. Ammonia is mainly toxic to the brain,and most of the complications associated with hyperammo-nemia are neurologic. Hypothermia has been reported to

decrease cerebral edema and improve encephalopathy inpatients with acute hepatic failure through reducing cerebraluptake of ammonia [40]. Hyperammonemia was alsoreported to be predictive of poor neurologic outcomes innontraumatic OHCA patients [18,25].

We found that hyperammonemia was independentlypredictive of non-ROSC in OHCA patients. The findingsof this study may offer useful information on the clinicalmanagement of OHCA patients. Physicians should considerterminating resuscitation efforts for OHCA patients if theinitial ammonia levels are extremely high. Further studies arealso needed to clarify the pathophysiologic significance ofammonia and the mechanism of hyperammonemia inpatients with cardiac arrest.

4.1. Limitations

The blood samples were collected from the femoralvessels of OHCA patients on ED arrival. However, we couldnot be sure whether the blood was arterial or venous due tothe technical difficulties of blood sampling during resusci-tation. To our knowledge, the difference in the ammonialevels of arterial and venous blood during resuscitation hasnot been determined.

The blood samples in this study were obtained frompatients within 5 minutes after ED arrival. We did notanalyze the correlation of the blood ammonia level with thetime elapsed from confirmation of cardiac arrest to EDarrival. Our study included both witnessed and unwitnessedcardiac arrest patients, and thus, the time between arrest andlaboratory sampling was unknown in some patients.

An awareness of the laboratory reports of biomarkers inthis study was not blinded. However, therapeutic decisionswere not biased because the physicians in charge of OHCApatients could not know the optical cutoff values until thestudy was completed. The results of ammonia levels wereobtained within 15 minutes in this study. Using ammonialevels to predict ROSC would be more feasible and practicalif the results were obtained sooner.

5. Conclusions

A blood ammonia level on ED arrival greater than 84μmol/L was 94.5% sensitive and 75.0% specific forpredicting non-ROSC in OHCA patients with a diagnosticaccuracy of 89.9%. The predictive value of the bloodammonia level for non-ROSC is better than that of the pNH3.

References

[1] Eisenberg MS, Cummins RO, Larsen MP. Numerators, denominators,and survival rates: reporting survival from out-of-hospital cardiacarrest. Am J Emerg Med 1991;9:544-6.

15Prognostic values of blood ammonia and partial pressure of ammonia

[2] Pepe PE, Levine RL, Fromm Jr RE, et al. Cardiac arrest presentingwith rhythms other than ventricular fibrillation: contribution ofresuscitative efforts toward total survivorship. Crit Care Med1993;21:1838-43.

[3] Weston CF, Wilson RJ, Jones SD. Predicting survival from out-of-hospital cardiac arrest: a multivariate analysis. Resuscitation 1997;34:27-34.

[4] Mashiko K, Otsuka T, Shimazaki S, et al. An outcome study of out-of-hospital cardiac arrest using the Utstein template—a Japaneseexperience. Resuscitation 2002;55:241-6.

[5] Booth CM, Boone RH, Tomlinson G, et al. Is this patient dead,vegetative, or severely neurologically impaired? Assessing outcomefor comatose survivors of cardiac arrest. JAMA 2004;291:870-9.

[6] Edgren E, Hedstrand U, Kelsey S, et al. Assessment of neurologicalprognosis in comatose survivors of cardiac arrest. BRCT I StudyGroup. Lancet 1994;343:1055-9.

[7] Madl C, Grimm G, Kramer L, et al. Early prediction of individualoutcome after cardiopulmonary resuscitation. Lancet 1993;341:855-8.

[8] Levine RL, Wayne MA, Miller CC. End-tidal carbon dioxide andoutcome of out-of-hospital cardiac arrest. N Engl J Med 1997;337:301-6.

[9] Ahrens T, Schallom L, Bettorf K, et al. End-tidal carbon dioxidemeasurements as a prognostic indicator of outcome in cardiac arrest.Am J Crit Care 2001;10:391-8.

[10] Bottiger BW, Mobes S, Glatzer R, et al. Astroglial protein S-100 is anearly and sensitive marker of hypoxic brain damage and outcome aftercardiac arrest in humans. Circulation 2001;103:2694-8.

[11] Fogel W, Krieger D, Veith M, et al. Serum neuron-specific enolase asearly predictor of outcome after cardiac arrest. Crit Care Med 1997;25:1133-8.

[12] Mullner M, Sterz F, Binder M, et al. Blood glucose concentration aftercardiopulmonary resuscitation influences functional neurologicalrecovery in human cardiac arrest survivors. J Cereb Blood FlowMetab 1997;17:430-6.

[13] Mullner M, Sterz F, Domanovits H, et al. The association betweenblood lactate concentration on admission, duration of cardiac arrest,and functional neurological recovery in patients resuscitated fromventricular fibrillation. Intensive Care Med 1997;23:1138-43.

[14] Sodeck GH, Domanovits H, Sterz F, et al. Can brain natriuretic peptidepredict outcome after cardiac arrest? An observational study.Resuscitation 2007;74:439-45.

[15] Kramer L, Jordan B, Druml W, et al. Incidence and prognosis of earlyhepatic dysfunction in critically ill patients—a prospective multicenterstudy. Crit Care Med 2007;35:1099-104.

[16] Clay AS, Hainline BE. Hyperammonemia in the ICU. Chest 2007;132:1368-78.

[17] Bhatia V, Singh R, Acharya SK. Predictive value of arterial ammoniafor complications and outcome in acute liver failure. Gut 2006;55:98-104.

[18] Shinozaki K, Oda S, Sadahiro T, et al. Blood ammonia and lactatelevels on hospital arrival as a predictive biomarker in patients with out-of-hospital cardiac arrest. Resuscitation 2011;82:404-9.

[19] Warren KS, Nathan DG. The passage of ammonia across the blood-brain-barrier and its relation to blood pH. J Clin Invest 1958;37:1724-8.

[20] Neumar RW, Otto CW, Link MS, et al. Part 8: adult advancedcardiovascular life support: 2010 American Heart AssociationGuidelines for Cardiopulmonary Resuscitation and EmergencyCardiovascular Care. Circulation 2010;122:S729-67.

[21] Cummins RO, Chamberlain DA, Abramson NS, et al. Recommendedguidelines for uniform reporting of data from out-of-hospital cardiacarrest: the Utstein Style. Task Force of the American HeartAssociation, the European Resuscitation Council, the Heart and StrokeFoundation of Canada, and the Australian Resuscitation Council. AnnEmerg Med 1991;20:861-74.

[22] Jacobs I, Nadkarni V, Bahr J, et al. Cardiac arrest and cardiopulmonaryresuscitation outcome reports: update and simplification of the Utsteintemplates for resuscitation registries. A statement for healthcareprofessionals from a task force of the international liaison committeeon resuscitation (American Heart Association, European ResuscitationCouncil, Australian Resuscitation Council, New Zealand ResuscitationCouncil, Heart and Stroke Foundation of Canada, InterAmerican HeartFoundation, Resuscitation Council of Southern Africa). Resuscitation2004;63:233-49.

[23] 2005 American Heart Association Guidelines for CardiopulmonaryResuscitation and Emergency Cardiovascular Care. Circulation 2005;112:IV1-203.

[24] Manning RT. A nomogram for estimation of pNH3. J Lab Clin Med1964;63:297-8.

[25] Nagamine K. Does blood ammonia level at time of initial treatmentpredict the outcome of patients in cardiopulmonary arrest on arrival?J Jpn Assoc Acute Med 2005;16:283-8.

[26] Yanagawa Y, Sakamoto T, Sato H. Relationship between laboratoryfindings and the outcome of cardiopulmonary arrest. Am J Emerg Med2009;27:308-12.

[27] Haberle J. Clinical practice: the management of hyperammonemia. EurJ Pediatr 2011;170:21-34.

[28] Baer HP, Drummond GI, Duncan EL. Formation and deamination ofadenosine by cardiac muscle enzymes. Mol Pharmacol 1966;2:67-76.

[29] Clarke DA, Davoll J, Philips FS, et al. Enzymatic deamination andvasodepressor effects of adenosine analogs. J Pharmacol Exp Ther1952;106:291-302.

[30] Askari A, Rao SN. Regulation of AMP deaminase by 2,3-dipho-sphoglyceric acid: a possible mechanism for the control of adeninenucleotide metabolism in human erythrocytes. Biochim Biophys Acta1968;151:198-203.

[31] Bendall JR, Davey CL. Ammonia liberation during rigor mortis and itsrelation to changes in the adenine and inosine nucleotides of rabbitmuscle. Biochim Biophys Acta 1957;26:93-103.

[32] Itoh H, Ohkuwa T. Ammonia and lactate in the blood after short-termsprint exercise. Eur J Appl Physiol Occup Physiol 1991;62:22-5.

[33] Itoh H, Ohkuwa T, Yamazaki Y, et al. Human blood lactate andammonia levels after supramaximal uphill and downhill running.Nagoya J Med Sci 1996;59:135-42.

[34] Ishida H, Mastuoka T, Yokota J, et al. The clinical significance of anincrease in the blood concentration of ammonia in cases withcardiopulmonary arrest on arrival. J Jpn Soc EmergMed 2002;5:490-5.

[35] Haussinger D. Nitrogen metabolism in liver: structural and functionalorganization and physiological relevance. Biochem J 1990;267:281-90.

[36] Lai CF, ChangWT, Liang PC, et al. Pneumatosis intestinalis and hepaticportal venous gas after CPR. Am J Emerg Med 2005;23:177-81.

[37] Reuter H, Bangard C, Gerhardt F, et al. Extensive hepatic portalvenous gas and gastric emphysema after successful resuscitation.Resuscitation 2011;82:238-9.

[38] Lien WC, Chang WT, Huang SP, et al. Hepatic portal venous gasassociated with poor outcome in out-of-hospital cardiac arrest patients.Resuscitation 2004;60:303-7.

[39] Windmueller HG. Glutamine utilization by the small intestine. AdvEnzymol Relat Areas Mol Biol 1982;53:201-37.

[40] Felipo V, Butterworth RF. Neurobiology of ammonia. Prog Neurobiol2002;67:259-79.

[41] Norenberg MD. Oxidative and nitrosative stress in ammonianeurotoxicity. Hepatology 2003;37:245-8.

[42] Norenberg MD, Rama Rao KV, Jayakumar AR. Signaling factors inthe mechanism of ammonia neurotoxicity. Metab Brain Dis 2009;24:103-17.

[43] Rose C, Michalak A, Pannunzio M, et al. Mild hypothermia delays theonset of coma and prevents brain edema and extracellular brainglutamate accumulation in rats with acute liver failure. Hepatology2000;31:872-7.