Journal of Critical Care (2013) 28, 832837

Early lactate clearance in septic patients with elevated lactate levels admitted from the emergency department to intensive care: Time to aim higher?,,,
Craig A. Walker MB, ChB, FCEM , David M. Griffith MB, ChB, FRCA, Alasdair J. Gray MB, ChB, MD, FRCS, FCEM, Deepankar Datta MB, ChB, BSc(Hons), Alasdair W. Hay MB, ChB, FRCA, DICM, MSc
The Royal Infirmary of Edinburgh, 51 Little France Crescent, Edinburgh, Scotland, United Kingdom EH16 4SA

Keywords:
Lactate clearance; Sepsis; Predict; Mortality

Abstract Purpose: Septic patients with hyperlactatemia have increased mortality rates, irrespective of hemodynamic and oxygen-derived variables. The aims of the study are the following: (1) to ascertain whether lactate clearance (LC) (percentage change in lactate over unit time) predicts mortality in septic patients admitted to intensive care directly from the emergency department and (2) to calculate the optimal cut-off value for mortality prediction. Methods: Three-year retrospective observational study of consecutive patients with severe sepsis and septic shock admitted to intensive care from the emergency department of a tertiary UK hospital. We calculated 6hour LC, performed receiver operating characteristic analyses to calculate optimal cut-off values for initial lactate and LC, dichotomized patients according to the LC cut-off, and calculated hazard ratios using a Cox proportional hazards model. Results: One hundred six patients were identified; 78, after exclusions. Lactate clearance was independently associated with 30-day mortality (P b .04); optimal cut-off, 36%. Mortality rates were 61.1% and 10.7% for patients with 6-hour LC 36% or less and greater than 36%, respectively. Hazard ratio for death with LC 36% or less was 7.33 (95% confidence interval, 2.17-24.73; P b .001). Conclusions: Six-hour LC was independently associated with mortality, and the optimal cut-off value was 36%, significantly higher than previously reported. We would support further research investigating this higher LC as a distinct resuscitation end point in patients with severe sepsis and septic shock. 2013 Elsevier Inc. All rights reserved.

We would like to thank Dr Stephanie Lewis, Deputy Medical Director at the Edinburgh Medical Research Centre Hub for Trials Methodology Research for her assistance with the statistical analysis of our data. The authors declare that they have no competing interests. There were no study sponsors. CW conceived the study, performed data collection, assisted with the statistical analyses, drafted the initial manuscript, and participated in all manuscript revisions. CW is the corresponding author. DG participated in the study design, performed the statistical analyses, drafted the statistical methods and results sections, and participated in all manuscript revisions. DD performed data collection and participated in all manuscript revisions. AG joint supervised the study and participated in all manuscript revisions. AH joint supervised the study, assisted with the statistical analyses, and participated in all manuscript revisions. There was no financial research support. Corresponding author. E-mail addresses: craigawalker@nhs.net (C.A. Walker), d.griffith@nhs.net (D.M. Griffith), Alasdair.Gray@luht.scot.nhs.uk (A.J. Gray), deepankardatta@nhs.net (D. Datta), Alasdair.Hay@luht.scot.nhs.uk (A.W. Hay).

0883-9441/$ see front matter 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcrc.2013.02.004

Early LC in septic patients with elevated lactate levels

833 consecutive adults ( 16 years) with sepsis admitted directly from the ED to the ICU of a tertiary UK hospital. The ICU admits more than 1000 level 3 patients per annum [10].

1. Introduction
Sepsis is an increasingly common reason for hospital admission. A prospective observational study of 198 intensive care units (ICUs) across 24 European countries in 2002 revealed that 24.7% had a diagnosis of sepsis on unit admission [1]. The mortality rate was 27% for patients with sepsis, increasing to 54.1% in patients with septic shock. Epidemiological data from the United States demonstrate that sepsis is a major and growing public health problem. Over a 22-year study period, Martin and colleagues demonstrated that sepsis accounted for more than 10 million hospital admissions, rising steadily from approximately 164 000 in 1979 to 660 000 in 2000 [2]. Recent studies have shown that outcome from sepsis may be improved with optimal resuscitation. When oxygen demand exceeds oxygen delivery, anaerobic metabolism ensues and results in increased lactate production. Septic patients with high initial serum lactate have a higher mortality rate, and this is independent of other organ failure indicators [3-5]. Bakker et al [3] identified, over 2 decades ago, that lactate was superior to oxygen-derived variables in predicting outcome in patients with septic shock. In 2001, the single-center randomized controlled trial of Rivers et al [6] demonstrated that early goal-directed therapy (EGDT) improved outcome in severe sepsis. Early goal-directed therapy has now become routine practice and is endorsed in the Surviving Sepsis Guidelines [7]. There are currently 3 international multicenter trials recruiting septic patients to EGDT. There is ongoing debate regarding optimal resuscitation targets. Although Rivers et al [8] used central venous saturations (ScvO2), others have advocated targeting lactate clearance (LC) as a resuscitation end point. In 2004, Nguyen et al [9] demonstrated that LC at 6 hours was an independent predictor of survival in severe sepsis and septic shock; an LC of more than 10% at 6 hours was reported as the best cut-off value to predict survival. EMShockNet Investigators subsequently used a target of 10% LC at 6 hours in an EGDT trial comparing LC and ScvO2 [8]. We believe that this LC target may be too low. Nguyen et al [9] identified a mean LC of 12% in nonsurvivors, which suggests that a large proportion of these nonsurvivors had an LC greater than 10%. In addition, in a separate EMShockNet trial, only 9% (15/166) of septic patients failed to attain 10% LC at 6 hours [5]. The objectives of our study were (1) to ascertain whether LC predicts mortality in our population of septic patients admitted to the ICU directly from the emergency department (ED) and (2) to calculate the optimal cut-off value for LC using receiver operating characteristic (ROC) analysis.

2.2. Inclusion and exclusion criteria

The Scottish critical care tracking and audit program, Ward Watcher [11], was interrogated to identify the following: (i) patients admitted between January 1, 2008, and December 31, 2010, (ii) admitted directly from the ED, or (iii) with an admission ICU diagnosis or primary diagnosis (the latter being completed on discharge from ICU for all patients) of infection or sepsis (Fig. 1). The hospital records were obtained and searched manually to confirm that the presentation diagnosis of sepsis was correct. Patients were excluded if (i) there was no record of arterial lactate measurement in the ED or (ii) the confirmed diagnosis on review of medical notes was not sepsis or infection or (iii) the hospital written records were unobtainable (for N 6 months).

2.3. Data collection

The Ward Watcher database, hospital, and ED records were interrogated, and the following data collected: patient age and sex, infection type, initial ED lactate, 6-hour lactate, Acute Physiology and Chronic Health Evaluation (APACHE) II scores at 24 hours after hospital admission, and 30-day survival.

2.4. Lactate clearance definition

Percentage LC was calculated as shown below to allow comparison with previous studies [5,8,9]: Lactate clearance LactateED presentation Lactate6 hours 100 LactateED Presentation

2.5. Statistical analysis

Using logistic regression, the relationship between LC and death was investigated. Using univariate analysis, the potential confounders of age, sex, APACHE II score, and initial lactate were tested to determine if they were related to 30-day mortality. Those factors found to be significant (P b .05) on univariate testing were included in a multivariate logistic regression model with LC to determine if LC is significantly associated with death after adjusting for these confounders. To test for interaction between initial lactate and LC, an interaction term was included in the model. For patients who had an abnormal initial lactate concentration ( 2 mmol/L), ROC curves were constructed for initial lactate and LC to test the ability to predict mortality at 30 days. To facilitate ROC analysis, LC was converted

2. Methods
2.1. Design and setting
We performed a 3-year (January 1, 2008, to December 31, 2010) single-center retrospective observational study of

834

C.A. Walker et al.

Date admitted to ICU between 01/01/2008 and 01/01/2011 AND Primary Unit Diagnosis = Any of: GI tract; Renal tract; Respiratory; Source not specified Bacteraemia / Septicaemia; Meningococcal infection; Multiple abscess formation; Systemic fungal infection; VRE; Other infection. Chest Infection (Atypical; Bacterial; Clinical (culture negative); Fungal; PCP; TB; Viral; Epiglottitis; Lung abscess; Mediastinitis; Other chest infection Infective endocarditis Cerebral abscess; Encephalitis; Meningitis; Other CNS infection Acute cholecystitis; Clostridium difficile; GI perforation; Gastroenteritis; Hepatic abscess; Other GI infection UTI Cellulitis; Necrotising fasciitis; Osteomyelitis; Perioral abscess; Superficial abscess; Wound infection

Septic shock Infection / Immunocompromise

Respiratory

Cardiovascular CNS GI Renal Musculoskeletal/Skin

Fig. 1

Search criteria for Ward Watcher database.

into lactate non-clearance by converting the percentage LC into a proportion and subtracting from 1. Receiver operating characteristic curves were constructed for initial lactate for (i) all patients and (ii) the subgroup of patients with abnormal initial lactate concentration ( 2 mmol/L) to allow direct comparison with the LC ROC curves. The area under each of these curves (AUC) was calculated. Optimal cut-off points were determined by maximizing Youden's index [12]. Kaplan-Meier survival curves were constructed for patients with an abnormal initial lactate dichotomized into those with an LC above and below the optimal cut-off value at 6 hours. A Cox proportional hazards model was constructed to determine a hazard ratio for mortality. All statistical analyses were carried out using the open source statistical package R and were reviewed and approved by a statistician [13]. Ethical approval was granted by The South East Scotland Research Ethics Service.

testing, APACHE II (P b .001), initial lactate (P b .001), and age (P = .01) but not sex (P = .13) were associated with 30-day mortality and were included in the multivariable model. After adjustment for those confounders found to be significant on univariate testing, LC was independently associated with 30day mortality (P = .04; Table 2). The interaction term for LC and initial lactate did not reach statistical significance (P = .30). Sixty-four patients had abnormal lactates at presentation and were included in the ROC analysis (the other 14 patients all survived). The AUC for initial lactate as a predictor of 30day mortality (in all patients) was 0.68 (95% CI, 0.57-0.80); the optimal initial lactate cut-off calculated by maximizing Youden's index was 4.2 mmol/L. The AUC for initial lactate as a predictor of 30-day mortality in patients with an abnormal initial lactate ( 2 mmol/L) was 0.57 (95% CI, 0.43-0.71). The ROC curve for lactate nonclearance (Fig. 3) yielded an AUC of 0.79 (95% CI, 0.68-0.90). The optimal cut-off was 0.64, corresponding to an LC of 36%. Using this threshold clearance of 36%, patients were divided into 2 groups. Patients with an LC 36% or less had a

3. Results
One hundred six patients were initially identified. Seventy-eight patients were available for analysis after exclusion criteria were identified (Fig. 2). Characteristics of the included patients are summarized in Table 1. The 11 patients excluded because hospital written records were not available after 6 months from initial request were similar to the study population (10 were male; age range was 36-82 years; median APACHE II score was 19 [18.5-23.0]; and 30day, 90-day, and ICU mortality rates were all 27%). Median initial lactate values were 3.4 mmol/L among survivors (interquartile range [IQR], 1.8-6.4) and 6.0 mmol/ L (4.2-13.3) among nonsurvivors. Six-hour LC was 37.2% among survivors (95% confidence interval [CI], 1.4-55.0) and 10.5% (95% CI, 0.7 to 29.5) among nonsurvivors. As percentage LC increased, mortality reduced (odds ratio for each unit change 0.99; 95% CI, 0.98-1.00). On univariate
Patients assessed for eligibility n = 106

Excluded (n = 28) - 11 notes not available after 6 months - 6 not sepsis - 6 not via ED - 4 lactates not performed or not recorded - 1 venous lactate only

Available for analysis n = 78

Fig. 2

Study flow diagram.

Early LC in septic patients with elevated lactate levels

Table 1 Patient characteristics Participants Age, median (IQR) Female, n (%) APACHE II, mean (95% CI) Initial lactate, median (IQR) LC, median (IQR) 30-d mortality 56 (40-66) 34 (43%) 24.6 (22.5-26.7) 4.9 (2.1-7.8) 26.9% ( 0.1 to 50.6) 32.1% (25/78)

835

mortality rate of 61.1% (22/36), whereas those with an LC greater than 36% had a mortality rate of only 10.7% (3/28). Lactate clearance at 6 hours of 36% or less predicted mortality with sensitivity of 88.0%, specificity of 64.1%, positive predictive value of 61.1%, and negative predictive value of 89.3%. The Kaplan-Meier survival analysis for LC is shown in Fig. 4. Patients with an LC of 36% or less were much more likely to die (hazard ratio 7.33 [95% CI, 2.17-24.73]; P b .001).

Optimal cut-off 0.64 (x) corresponds to lactate clearance of 36% (1-x)

Fig. 3 Lactate clearance ROC curve. Optimal cut-off 0.64 (x) corresponds to lactate clearance of 36% (1-x).

4. Discussion
In patients with severe sepsis identified in the ED and admitted directly to the ICU, our data confirm that both initial lactate and 6-hour LC are independently associated with 30-day mortality. An AUC value of 0.79 suggests that early LC performs remarkably well as a predictor of 30-day mortality and is a more discriminant test than initial lactate (AUC, 0.57). Optimal cut-off values were 4.2 mmol/L for initial lactate (for all patients) and 36% for 6-hour LC. Mortality was significantly worse in patients with LC 36% or less compared with patients with LC greater than 36% (61.1% vs 10.7%; P b .001). The association between LC and outcome in sepsis is well established [3,5,9,14-16]. Unlike many other factors associated with mortality in sepsis, LC is a dynamic measure calculated during initial resuscitation. There is evidence that LC may be a marker of the quality of resuscitation rather than purely a fixed measure of patient response to treatment. Since the publication of the Rivers study in 2001 demonstrating benefit for EGDT [6], there has been interest in whether LC could be used as a resuscitation end point to guide therapy. In 2011, Nguyen et al [16] demonstrated in a

multicenter quality improvement study that the addition of an LC target to a sepsis bundle improved outcome. To date, there have been 3 EGDT randomized controlled trials, which have investigated resuscitation guided by an LC target. The EMShockNet Investigators performed a noninferiority study comparing a target LC of greater than 10% at 6 hours with an ScvO2 target of greater than 70% at 6 hours and found comparable mortality rates (17% in the LC group vs 23% in the ScvO2 group) [8]. Jansen et al [17] compared

Table 2 Logistic regression model: variables and mortality odds ratios Variable LC (per % change lactate) Initial lactate (per mmol/L) Age (per year) APACHE II score (per point) Odds ratio (95% CI) 0.99 1.21 1.04 1.06 (0.97-1.00) (1.06-1.42) (1.00-1.09) (1.02-1.19) P .04 .01 .05 .02
Time (Days)
Hazard Ratio for death in patients with lactate clearance 36% compared to those with lactate clearance >36% = 7.33 (95% CI 2.2 to 24.7, P < .001).

Fig. 4 Kaplan-Meier survival curves. Hazard ratio for death in patients with LC 36% or less compared with those with LC greater than 36% = 7.33 (95% CI, 2.2-24.7; P b .001). Time (days).

Probability of Survival

836 resuscitation with and without a target LC of 20% every 2 hours in a heterogenous group of ICU patients but demonstrated no significant outcome benefit for the intervention. Tian et al [18] randomized septic shock patients to 6 hours of either target LC greater than 30%, target LC greater than 10%, or standard EGDT. For the 62 patients analyzed, 28-day mortality rates were significantly lower in the 30% LC group (28.6%) and the 10% group (36.4%) than the control group (63.2%); P b .05. With the varying outcomes and targets in these studies, it is difficult to make firm conclusions on the use of LC as a target for resuscitation. It could be argued that the EMShock Net study demonstrates that LC can be used as an equivalent or alternative to targeting ScvO2. On the other hand, in the Janson study, there was a high LC in both the intervention and control arms suggesting that aggressive resuscitation may provide good LC without the need for an explicit LC target. Tian's study supports targeting high LC but suffers from a number of methodological limitations (exclusion of the 11 patients who failed to meet LC or EGDT targets and the lack of an intention-to-treat analysis, for example). The utility of LC to guide the resuscitation of septic patients may depend on the target set. A variety of LC targets has been suggested and used varying from 10% at 6 hours to 20% in 2 hours [9,15,17,18]. The article of Nguyen et al in 2004 was the first to report an LC target of 10% at 6 hours as the best cut-off value to predict survival in severe sepsis and septic shock [9]. However, in the same article, they also demonstrated that, for every 10% increase in LC at 6 hours, the mortality improved. The LC value which best discriminates between survivors and nonsurvivors in an observational study is not necessarily the best target. If we hypothesize that LC is a marker of the quality of resuscitation rather than a measure of a patient's intrinsic ability to respond to treatment then, in the context of a trial, we should target the LC with the best outcome. Our logistic regression model also demonstrates that there is an incremental improvement in survival with increasing clearance, suggesting that a higher target may be set. Indeed, the updated Surviving Sepsis Campaign (2012) guidelines advocate targeting resuscitation to normalize lactate in patients with elevated lactate levels [7]. There are a number of strengths in our study design. We included extremely sensitive search criteria and, therefore, reduced the likelihood of overlooking appropriate patients. The Ward Watcher database is a tightly governed and accurate data set with less than 6% of patients incorrectly assigned as having sepsis. We subsequently searched ED and hospital notes manually to maximize data completion. A limitation is our relatively low sample size and the retrospective sampling. Numbers are, however, comparable with previous reported studies [3,9,16]. An explanation for the low sample size is that it only included patients admitted directly from the ED to the ICU; this was deemed necessary to ensure that aggressive resuscitation was provided throughout the first 6-hour period and also because repeat

C.A. Walker et al. lactate measurements do not currently form part of routine clinical care in the general wards and are thus not reported on the laboratory blood gas analyzers in our hospital; we felt that including such patients would not have significantly increased our data set. We emphasize that our derived optimal LC relates only to those patients with elevated initial serum lactate levels and support continuing to resuscitate patients as per current EGDT targets while further research into LC is undertaken. Although we feel that further research needs to be performed to ascertain how best to incorporate early LC into current EGDT for severe sepsis and septic shock, our study reaffirms the important independent relationship that this index has with mortality. In addition, we are the first to clearly demonstrate the superiority of LC as a predictor of mortality when compared with baseline blood lactate concentration.

5. Conclusions
In ED patients admitted to intensive care with severe sepsis or septic shock, we found that 6-hour LC was a significantly better predictor of mortality than initial lactate and that this effect was independent of other markers of organ dysfunction. Lactate clearance, therefore, provides important additional information over and above other measures. We found that the best cut-off value for LC was 36%. This is a significantly higher LC than previously reported, and we would support further research targeting a higher LC of 36% as a distinct resuscitation end point in patients with severe sepsis and septic shock.

Acknowledgments
We would like to thank Dr Stephanie Lewis, Deputy Medical Director at the Edinburgh Medical Research Centre Hub for Trials Methodology Research for her assistance with the statistical analysis of our data.

References
[1] Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, et al. Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 2006;34(2):344-53. [2] Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the united states from 1979 through 2000. N Engl J Med 2003;348(16):1546-54. [3] Bakker J, Coffernils M, Leon M, Gris P, Vincent JL. Blood lactate levels are superior to oxygen-derived variables in predicting outcome in human septic shock. Chest 1991;99(4):956-62. [4] Mikkelsen ME, Miltiades AN, Gaieski DF, Goyal M, Fuchs BD, Shah CV, et al. Serum lactate is associated with mortality in severe sepsis independent of organ failure and shock. Crit Care Med 2009;37(5): 1670-7.

Early LC in septic patients with elevated lactate levels

[5] Arnold RC, Shapiro NI, Jones AE, Schorr C, Pope J, Casner E, et al. Emergency Medicine Shock Research Network (EMShockNet) Investigators. Multicenter study of early lactate clearance as a determinant of survival in patients with presumed sepsis. Shock 2009;32(1):35-9. [6] Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001;345(19):1368-77. [7] Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, et al. Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med 2013;41(2): 580-637. [8] Jones AE, Shapiro NI, Trzeciak S, Arnold RC, Claremont HA, Kline JA. Emergency Medicine Shock Research Network (EMShockNet) Investigators. Lactate clearance vs. central venous oxygen saturation as goals of early sepsis therapy: a randomized clinical trial. JAMA 2010;303(8):739-46. [9] Nguyen HB, Rivers EP, Knoblich BP, Jacobsen G, Muzzin A, Ressler JA, et al. Early lactate clearance is associated with improved outcome in severe sepsis and septic shock. Crit Care Med 2004;32(8):1637-42. [10] Scottish Intensive Care Society Audit Group (SICSAG). Audit of Critical Care in Scotland 2011, Reporting on 2010. http://www.sicsag. scot.nhs.uk/Publications/SICSAG-report_2011-webversion.pdf. [11] Millar B. Ward Watcher (Version 11.81) [Computer Program]. Scotland: Critical Care Audit Group Ltd; 2010.

837
[12] Youden WJ. An index for rating diagnostic tests. Cancer 1950;3:32-5. [13] Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2010. ISBN 3-900051-07-0, URL http://www.R-project.org/. [14] Bakker J, et al. Serial blood lactate levels can predict the development of multiple organ failure following septic shock. Am J Surg 1996;171(2):221-6. [15] Yang CS, Qiu HB, Huang YZ, Xie JF, Mo M, Liu SQ, et al. Prospective research on the prognosis of septic shock based on the change of lactate concentration in arterial blood. Chung Hua Wai Ko Tsa Chih 2009;47(9):685-8. [16] Nguyen HB, Kuan WS, Batech M, Shrikhande P, Mahadevan M, Li CH, et al. Asia Network to Regulate Sepsis Care (ATLAS) Investigators. Outcome effectiveness of the severe sepsis resuscitation bundle with addition of lactate clearance as bundle item: a multinational evaluation. Crit Care 2011;15:R229. [17] Jansen TC, van Bommel J, Schoonderbeek FJ, Sleeswijk Visser SJ, van der Klooster JM, Lima AP, et al. Early lactate-guided therapy in intensive care unit patients: a multicenter, open-label, randomized controlled trial. Am J Respir Crit Care Med 2010 Sep 15;182(6): 752-61. [18] Tian H, Han S, Lv C, Wang T, Li Z, Hao D, et al. The effect of early goal lactate clearance rate on the outcome of septic shock patients with severe pneumonia. Zhongguo Wei Zhong Bing Ji Jiu Yi Xue 2012; 24(1):42-5.