INFECTIOUS DISEASE/BRIEF RESEARCH REPORT

Serum Lactate as a Predictor of Mortality in Emergency Department Patients With Infection

Nathan I. Shapiro, MD, MPH Michael D. Howell, MD Daniel Talmor, MD, MPH Larry A. Nathanson, MD Alan Lisbon, MD Richard E. Wolfe, MD J. Woodrow Weiss, MD
From the Department of Emergency Medicine (Shapiro, Wolfe, Nathanson), the Department of Medicine, Division of Pulmonary and Critical Care Medicine (Howell, Weiss), and the Department of Anesthesia and Critical Care (Talmor, Lisbon), Beth Israel Deaconess Medical Center, Boston, MA.

Study objective: Little is known about risk-stratication biomarkers in emergency department (ED) patients with suspected infection, and lactate is a biologically plausible candidate. We determine whether a serum venous lactate is associated with an increased risk of death in ED patients with infection. Methods: This was a prospective cohort study in an urban, academic medical center with 50,000 annual ED visits. A total of 1,278 consecutive patient visits met enrollment criteria between July 24, 2003, and March 24, 2004, and all patients were enrolled. Inclusion criteria were age 18 years or older, serum lactate level obtained, and admission to the hospital with an infection-related diagnosis. The main outcome measure was all-cause 28-day inhospital mortality and death within 3 days of presentation. Results: Among 1,278 patient visits, there were 105 (8.2%) deaths during hospitalization, with 55 (4.3%) of 1,278 deaths occurring in the rst 3 days. Mortality rates increased as lactate increased: 43 (4.9%) of 877 of patients with a lactate level between 0 and 2.5 mmol/L died, 24 (9.0%) of 267 patients with a lactate level between 2.5 and 4.0 mmol/L died, and 38 (28.4%) of 134 patients with a lactate level greater than or equal to 4.0 mmol/L died. Lactate level greater than or equal to 4.0 mmol/L was 36% (95% condence interval [CI] 27% to 45%) sensitive and 92% (95% CI 90% to 93%) specic for any death; it was 55% (95% CI 41% to 68%) sensitive and 91% (95% CI 90% to 93%) specic for death within 3 days. Conclusion: In this cohort of ED patients with signs and symptoms suggestive of infection, our results support serum venous lactate level as a promising risk-stratication tool. Multicenter validation, as well as comparison of the lactate level with clinical predictors, needs to be done before widespread implementation. [Ann Emerg Med. 2005;45:524-528.]

0196-0644/$-see front matter Copyright 2005 by the American College of Emergency Physicians. doi:10.1016/j.annemergmed.2004.12.006

INTRODUCTION
Bacteremia and sepsis are responsible for signicant morbidity, mortality, and costs to patients in our health care system. There are an estimated 751,000 cases of severe sepsis per year in the United States,1 with a hospital case mortality rate between 30% and 50%.2-4 The therapies of activated protein C and early goal-directed therapy have been recently shown to reduce mortality in patients with severe sepsis.2,4,5 Although these therapies are effective, early goal-directed therapy is resource intensive, activated protein C is expensive, and all 524 Annals of Emergency Medicine

aggressive therapies carry risk. Thus, it is important to select patients who may benet most. Additionally, the early presentation of sepsis can be subtle and difcult to detect, which also represents a challenge to the clinician. A number of laboratory tests and markers have been proposed to assist in risk stratication of patients with infection. Leukocytosis, bandemia, thrombocytopenia, elevated C-reactive protein, elevated sedimentation rates, and elevated procalcitonin have all been suggested; however, there is no single reliable test. Serum lactate offers theoretical promise as a risk-stratication
Volume 45, no. 5 : May 2005

Shapiro et al

Lactate as a Predictor of Mortality Despite its routine use in ICU patients already identied to be seriously ill, measurement of a serum lactate as a riskstratication tool in patients with infection is relatively limited in most emergency departments (EDs). Given its ability to detect patients with tissue hypoxia and increased metabolic demands resulting from sepsis, we hypothesized that the serum lactate level is a useful screen to identify patients in the ED with increased risk of death. Accordingly, the objective of this study is to determine the ability of a serum lactate level to predict death in ED patients with infection.

Editors Capsule Summary What is already known on this topic Elevated serum lactate levels have been associated with mortality among ICU patients and have been used as an inclusion criterion in a trial of emergency department (ED) patients with severe septic shock who demonstrated a mortality benet from early goal-directed hemodynamic therapy. What question this study addressed What is the association of venous lactate levels with inhospital mortality among ED patients who have suspected infection and are admitted to the hospital? What this study adds to our knowledge Death within 3 days occurred in 22.4% of ED patients admitted with presumed infection who had lactate levels greater than 4 mmol/L versus 1.5% with levels less than 2.5 mmol/L; lactate greater than 4 mmol/L, which was present in 10.5% of patients, was 55% sensitive and 91% specic for this outcome. Clinical predictors of outcome were not assessed. How this might change clinical practice Assessing lactate levels in the ED among potentially infected patients may be helpful in identifying patients for more intensive therapy and monitoring, but its usefulness compared with that of clinical predictors is unknown. Research wed like to see Further research is needed to evaluate the independent predictive role of lactate compared with other routinely obtained clinical and laboratory parameters and to determine whether it provides any unique information, particularly among ED patients without hypotension or sepsis-induced organ dysfunctions.

METHODS
Study Design This was a prospective, observational, cohort study of consecutive ED patients with infection who had a serum lactate level obtained. All adult patients (aged 18 years or older) who presented to the ED between July 24, 2003, and March 24, 2004, were eligible. The ED, with 50,000 annual visits, serves an urban, academic teaching hospital with 490 beds. The study was approved by the institutional review board, and a waiver of consent was obtained. The study inclusion criteria were age 18 years or older, serum lactate level obtained, and admission to the hospital with an infection-related diagnosis. Before study initiation, we had instituted an ED-wide clinical practice protocol whereby a serum lactate level was to be obtained in patients suspected of having a clinically signicant infection. This practice was operationalized through physician didactics, nursing in-services, bedside posters, e-mail reminders, quality assurance feedback, and a practice of routinely testing for serum lactate level when a blood culture was ordered. This study therefore represents a population of adult ED patients who were admitted to the hospital and for whom a serum lactate was obtained in the ED as part of a standardized practice measuring lactate in patients with clinically suspected signicant infection. The lactate was drawn during the course of the patient evaluation, usually on initial evaluation. It was always from the ED, and we included only admitted patients; however, the draw often took place before the decision to admit or discharge was known. We assessed peripheral venous rather than arterial or central venous lactate level because venous blood is more practical to obtain in the ED setting. Previous work has shown equivalence between these measures,6,7 and others have used venous lactate to risk stratify different ED patient populations.6,8 To identify patients who were admitted to the hospital with infection, the hospital admitting diagnosis was collected. An investigator blinded to lactate level and subsequent hospital course categorized each admitting diagnosis as infection related or noninfection related. A similarly blinded second investigator independently reviewed a 20% random sample to assess the reproducibility. The k value for determination of infection was 0.9. This categorization was done on the basis of admitting diagnosis and chief complaint. Although it was not done in real time, it was done with only the information available at admission (ie, if someone was admitted with an ED diagnosis of
Annals of Emergency Medicine 525

tool in patients with infection. Impaired tissue oxygenation is a hallmark of shock, but blood pressure, pulse rate, and urine output are insensitive markers for this disturbance, and many patients have clinically unapparent shock. However, tissue hypoxia causes cellular metabolism to shift to the anaerobic pathway, increasing lactate production; at the same time, sepsis impairs lactate clearance, which may lead to an increase in circulating lactate levels, even when vital signs are not yet compromised. In practice, determination of circulating lactate levels is technically feasible, commonly used, and clinically available with rapid turnaround times. Finally, lactate has been evaluated as a prognostic tool in a number of critical-care patient populations, including patients with septic shock, recent surgery, burns, and trauma, and has been found to predict mortality among these very ill patients.
Volume 45, no. 5 : May 2005

Lactate as a Predictor of Mortality pneumonia that an inpatient evaluation revealed to be congestive heart failure, it was still classied as pneumonia). Initial serum lactate level was collected, along with the leukocyte count, band percentage, and platelet count. The primary outcome was 28-day inhospital mortality. Patients who were discharged alive in fewer than 28 days were considered alive for the purposes of this analysis. A secondary outcome of early death was dened as death within 3 days of hospital admission. Statistical Methods Descriptive statistics for lactate level, age, leukocyte count, platelet count, and band percentage are reported. The lactate results were stratied a priori into 3 risk groups based on clinical practice and thresholds previously used in the medical literature.2,9 The groups were dened as low (\2.5 mmol/L; n=877), medium (2.5 to 3.99 mmol/L; n=267), and high (R4 mmol/L; n=134). Analysis was performed using SAS software (version 8, SAS Institute, Inc., Cary, NC), and univariate comparisons were made using t test, Wilcoxon rank-sum test, and c2 test, with a equal to 0.05. To assess for independent predictors and compare the ability of lactate level to predict death compared with other laboratory predictors, a multiple logistic regression model was built, and the C-statistic is reported to assess the performance of the model. The model controlled for age, whereas WBC count greater than 15,000/ mm3, platelet count less than 150,000 cells/mm3, and band percentage greater than 5% were all eligible dichotomous covariates dened a priori on the basis of optimal thresholds from previous studies. Covariates that reached a signicance of P value less than .05 were included in the model. The receiver operating characteristic (ROC) area under the curve for lactate level as a predictor of death and early death was calculated and reported.

Shapiro et al

Figure 1. Histogram of lactate results.

RESULTS
A total of 1,278 patients met criteria, and all were enrolled. There were 105 (8.2%) of 1,278 inhospital deaths, of which 55 (4.3%) of 1,278 patients died within 3 days of admission (early death). The mean lactate level of 4.0 mmol/L (95% condence interval [CI] 3.4 to 4.7 mmol/L) in those who died was higher than 2.2 mmol/L (95% CI 2.1 to 2.3 mmol/L) in those who lived (Figure 1). The mortality rates for the low, medium, and high groups were 43 of 877 (4.9%; 95% CI 3.5% to 6.3%), 24 of 267 (9.0%; 95% CI 5.6% to 12.4%), and 38 of 134 (28.4%; 95% CI 21% to 36%) (Figure 2). Thus, a serum lactate level greater than or equal to 2.5 mmol/L was 59% (95% CI 50% to 68%) sensitive and 71% (95% CI 69% to 74%) specic for death. Lactate greater than or equal to 4 mmol/L was 36% (95% CI 27% to 45%) sensitive and 92% (95% CI 90% to 93%) specic for death. The ROC area under the curve for lactate level as a predictor of death was 0.67. To obtain the adjusted estimate of lactate, a multiple logistic regression model was constructed, which, when controlled for age, showed lactate level greater than or equal to 4 mmol/L (odds ratio [OR] 4.9; 95% CI 3.1 to 7.9), band percentage 526 Annals of Emergency Medicine

Figure 2. Lactate as a predictor of mortality. For 28-day inhospital mortality, 105 (8.2%) of 1,278 of patients died; death within 3 days occurred in 55 (4.3%) of 1,278 patients. Reprinted from Academic Emergency Medicine, Vol. 12, Shapiro NI, Howell MD, Talmor D. A blueprint for a sepsis protocol, 2005, with permission from The Society for Academic Emergency Medicine.16

greater than 5% (OR 1.9; 95% CI 1.1 to 3.1), and platelet count less than 150,000 cells/mm3 (OR 2.3; 95% CI 1.1 to 1.5) to be signicant independent predictors of death in patients with suspected infection. Patients with a WBC count greater
Volume 45, no. 5 : May 2005

Shapiro et al than 15 cells/mm3 were at no signicant increased risk of death (OR 0.8; 95% CI 0.5 to 1.3). The C-statistic (area under the curve) for the model was 0.78. More detailed results including patient stratication are available in Appendix E1 (available at http://www.mosby.com/AnnEmergMed). The early death rates (within 3 days) for the low, medium, and high groups were 13 of 877 (1.5%; 95% CI 0.7% to 2.3%), 12 of 267 (4.5%; 95% CI 2.0% to 7.0%), and 30 of 134 (22.4%; 95% CI 15% to 29%; P\.01) (Figure 2). Thus, a serum lactate level greater than or equal to 2.5 mmol/L was 76% (95% CI 65% to 88%) sensitive and 71% (95% CI 68% to 73%) specic for early death. Lactate levels greater than or equal to 4.0 mmol/L were 55% (95% CI 41% to 68%) sensitive and 91% (95% CI 90% to 93%) specic for early death. A multiple logistic regression model controlling for age showed lactate levels greater than or equal to 4.0 mmol/mm3 to have a 9.5 (95% CI 5.5 to 17) odds of death within 3 days, whereas a bandemia greater than 5% had an OR of 3.7 (95% CI 2.0 to 6.6). Neither WBC count of 15,000 cells/mm3 nor platelet count of less than 150,000 cells/mm3 were signicant independent predictors. The C-statistic for the model was 0.84. The ROC area under the curve for lactate level as a predictor of early death was 0.80.

Lactate as a Predictor of Mortality ment (SOFA) scores? How does lactate perform in a population stratied by severity or in a non-ICU population? The present study does not and cannot answer these questions, but we believe that it substantiates the dedication of resources to future investigations.

DISCUSSION
Our data demonstrate that in a large cohort of ED patients with signs and symptoms of infection, a single measurement of venous lactate provides risk-stratication information about inhospital mortality and mortality within 3 days. We measured lactate in a systematic fashion and used the objective endpoint of all-cause mortality. The mortality rate of 28% for patients with a lactate level greater than or equal to 4.0 mmol/mm3 clearly demonstrates that a lactate level may be helpful in identifying a high-risk patient population in the ED. Additionally, the serum lactate level was also useful to identify patients who were at increased risk of death within a 3-day period; thus, this population certainly deserves rapid attention and perhaps aggressive intervention. Previous studies of the prognostic utility of lactate level have focused on the critically ill ICU population. Seminal work by Broder and Weil10 studied 56 patients in shock from a variety of causes and found a mortality rate of 89% when a single lactate value was greater than or equal to 4 mmol/L. Aduen et al11 investigated lactate levels in a cross-section of 180 undifferentiated ICU patients. They found that lactate level was signicantly higher in nonsurvivors, regardless of blood pressure. Other authors have focused on patients specically with septic shock. Bakker et al12 focused on 48 ICU patients with septic shock and found that initial and serial blood lactate levels were superior to pulmonary arterial catheterderived data in predicting mortality. Vincent et al,13 Bakker et al,14 and Falk et al15 found that serial lactate levels predicted mortality in ICU patients with septic shock but that initial levels had less of a prognostic ability. The present study substantially extends previous investigations about the prognostic utility of lactate levels. First, we applied it to a broad population of patients who were admitted to the hospital through the ED and had a clinically suspected signicant infection, a group in which we believe that a biochemical risk-stratication tool would be very useful. We did not limit it to patients with septic shock, which is a population that other investigators have previously evaluated in detail. Rather, we studied lactate level in the broader population that an emergency physician would see as patients present for care, which would enable the clinician to identify patients without clinically apparent septic shock but who are still at a substantial risk for mortality. Our results are biologically plausible and concordant with ndings in critically ill patients. A risk-stratication biomarker such as lactate level may aid in triage decisions for patients with clinically suspected infection because patients who are more severely ill should be considered for closer monitoring, aggressive intervention, and placement in an ICU. Lactate level
Annals of Emergency Medicine 527

LIMITATIONS
This study is a preliminary examination and has several important limitations. Because this was a preliminary study, we collected no data about hemodynamics, comorbidities, or other clinical conditions. We did not describe our patient population in terms of proportions of patients with conditions such as systemic inammatory response syndrome, severe sepsis, or septic shock. It is also possible that lactate is a covariate of some other measurable clinical factor. There are also patients in our population who may have had an infection but for whom a lactate level may not have been obtained. Additionally, a serum lactate level was not obtained for patients when the clinician may have missed an occult infection and outpatients were not included in this investigation; thus, the results apply to ED patients admitted to the hospital with suspected infection and may not be generalizable to all ED patients, which represents a practical problem of a busy ED but is consistent with clinical practice. However, as discussed above, we believe that this more closely parallels the manner in which patients initially present for care and the way in which any potential riskstratication biomarker will be used. Future studies will be required to address some important clinical questions before widespread implementation. Some of these questions include: Is this study generalizable outside of our institution? How does serum lactate level perform in patients with occult infections not apparent to the clinician? How does it perform in the population of patients without hypotension? How does it compare with traditional risk stratication scoring systems such as the Acute Physiology and Chronic Health Evaluation (APACHE-II), Multiple Organ Dysfunction Syndrome (MODS), and Sepsis-Related Organ Failure AssessVolume 45, no. 5 : May 2005

Lactate as a Predictor of Mortality may be helpful in deciding which patients could benet from these aggressive treatment strategies. In conclusion, in patients admitted to the hospital who present to an ED with a clinically suspected signicant infection, a single measurement of venous lactate provides substantial prognostic information about inhospital mortality and 3-day mortality. Our results support serum lactate as a promising risk-stratication tool. Multicenter validation, as well as combining the lactate level with other predictors, needs to be done before widespread implementation.
Author contributions: NIS, MDH, and DT were involved in all phases of idea conception, study design, data collection, analysis, and manuscript preparation. LAN was involved in study design, data collection, and manuscript review. AL, REW, and JWW were involved in study design and manuscript preparation. NIS, MH, and DT take responsibility for the paper as a whole. Funding and support: The authors report this study did not receive any outside funding or support. Publication dates: Received for publication July 9, 2004. Revisions received November 3, 2004, and November 16, 2004. Accepted for publication November 29, 2004. Available online March 22, 2005. Address for reprints: Nathan I. Shapiro, MD, MPH, Department of Emergency Medicine, Beth Israel Deaconess Medical Center, 1 Deaconess Road, CC2-W, Boston, MA 02215; 617-7542343, fax 617-754-2350; E-mail nshapiro@bidmc.harvard.edu. REFERENCES
1. Angus DC, Linde-Zwirble WT, Lidicker J, et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001;29: 1303-1310. 2. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med. 2001;345:1368-1377.

Shapiro et al
3. Shapiro NI, Wolfe RE, Moore RB, et al. Mortality in Emergency Department Sepsis (MEDS) score: a prospectively derived and validated clinical prediction rule. Crit Care Med. 2003;31: 670-675. 4. Bernard GR, Vincent JL, Laterre PF, et al. Efcacy and safety of recombinant human activated protein C for severe sepsis. N Engl J Med. 2001;344:699-709. 5. Annane D, Sebille V, Charpentier C, et al. Effect of treatment with low doses of hydrocortisone and udrocortisone on mortality in patients with septic shock. JAMA. 2002;288: 862-871. 6. Lavery RF, Livingston DH, Tortella BJ, et al. The utility of venous lactate to triage injured patients in the trauma center. J Am Coll Surg. 2000;190:656-664. 7. Weil MH, Michaels S, Rackow EC. Comparison of blood lactate concentrations in central venous, pulmonary artery, and arterial blood. Crit Care Med. 1987;15:489-490. 8. Schmiechen NJ, Han C, Milzman DP. Emergency department use of rapid lactate to evaluate patients with acute chest pain. Ann Emerg Med. 1997;30:571-577. 9. Levy B, Sadoune LO, Gelot AM, et al. Evolution of lactate/ pyruvate and arterial ketone body ratios in the early course of catecholamine-treated septic shock. Crit Care Med. 2000;28: 114-119. 10. Broder G, Weil MH. Excess lactate: an index of reversibility of shock in human patients. Science. 1964;143:1457-1459. 11. Aduen J, Bernstein WK, Khastgir T, et al. The use and clinical importance of a substrate-specic electrode for rapid determination of blood lactate concentrations. JAMA. 1994;272: 1678-1685. 12. Bakker J, Coffernils M, Leon M, et al. Blood lactate levels are superior to oxygen-derived variables in predicting outcome in human septic shock. Chest. 1991;99:956-962. 13. Vincent JL, Dufaye P, Berre J, et al. Serial lactate determinations during circulatory shock. Crit Care Med. 1983;11:449-451. 14. Bakker J, Gris P, Coffernils M, et al. Serial blood lactate levels can predict the development of multiple organ failure following septic shock. Am J Surg. 1996;171:221-226. 15. Falk JL, Rackow EC, Leavy J, et al. Delayed lactate clearance in patients surviving circulatory shock. Acute Care. 1985;11: 212-215. 16. Shapiro NI, Howell MD, Talmor D. A blueprint for a sepsis protocol. Acad Emerg Med. 2005;12.

528 Annals of Emergency Medicine

Volume 45, no. 5 : May 2005