a b s t r a c t

Background: Gastrointestinal hemorrhage (GIH) is a common complaint seen in the emergency department (ED) and carries a small but significant mortality rate. The principal purpose of this investigation was to determine whether an ED venous lactate as part of initial laboratory studies is predictive of mortality in patients admitted to the hospital for GIH.

Methods: Retrospective cohort study for 6 years at an urban tertiary referral hospital included all ED patients with the charted diagnosis of acute GIH. Serum lactate was drawn at the bedside as part of patient care after arrival to the ED at the discretion of the clinical team. Clinical parameters and inpatient mortality were collected from the medical record. Optimal cut points for lactate were derived using receiver operating characteristics curves and imputed into a multivariable logistic regression model.

Results: Of the 2834 medical records that had GIH diagnoses, 1644 had an ED lactate recorded. A lactate greater than 4 mmol/L conferred a 6.4-fold increased odds of in-hospital mortality (94% specificity, P b .001). Controlling for age, initial hematocrit, and heart rate, every 1-point increase in lactate conferred a 1.4-fold increase in the odds of mortality.

Conclusions: Elevated initial lactate drawn in the ED can be associated with in-hospital mortality for ED patients with acute GIH. Prospective validation studies are warranted.

(C) 2014

Introduction

Background

Nonhemorrhoidal gastrointestinal hemorrhage (GIH) is a common complaint seen in the emergency department (ED) and is associated with significant morbidity and mortality. In the United States, between 250 000 and 300 000 people are hospitalized annually for upper GIH alone [1], with more than $2.5 billion [1] spent every year for the care of these patients. Despite the large sum of money dedi- cated to treatment, 15 000 to 30 000 will die annually from this diagnosis [1]. Over time, we have improved our abilities to provide quality care to patients with GIH, but it is not always easy to ascertain which patients are likely to be at high risk for the 6% to 12% mortality rate of upper GIH or the 10% to 20% mortality rate of lower GIH [2-8]. Unfortunately, risk stratification for in-hospital morbidity and mor- tality remains difficult.

? Disclosures: None.

* Corresponding author. Icahn School of Medicine at Mount Sinai, Box 1620, New York, NY 10029.

E-mail address: [email protected] (M. Chisolm-Straker).

Importance

Existing GIH Mortality prediction tools include the following: the Glasgow-Blatchford Bleeding Score, which is only specific to upper gastrointestinal (GI) bleeding and aims to determine which patients are “low risk” and candidates for outpatient management; the Acute Physiology and Chronic Health Evaluation II scoring system, which predicts all comers at high risk for mortality but requires multiple factors and at least 24 hours in an intensive care unit (ICU); the Rockall cardiopulmonary resuscitation, which stratifies patients with upper GI bleeding about their risk for rebleeding and death but requires endoscopic diagnosis, which is not available to ED providers early in a patient’s presentation [1,7,8]. Some authors believe these to be overly cumbersome or imprecise, and ED providers unable to accurately access which patients are likely to become critically ill, overadmit for observation, serial laboratories, and vital signs surveillance [1]. Lactate has been used to predict the severity of illness and risk of mortality in many disease processes including poisoning, sepsis, trauma, cancer, and pediatric cardiac disease [9-15]. The previous literature on the use of lactate and its correlation with morbidity and mortality is compelling, particularly in septic or ICU patients [13,16-20]. Studies have indicated that lactate can be used to

http://dx.doi.org/10.1016/j.ajem.2014.02.010

0735-6757/(C) 2014

predict the outcome of patients in ICUs, who have a variety of pathologies [14]. Other studies demonstrate that lactate predicts multisystem organ failure and death in septic persons [17,18,20]. Shapiro et al in 2005 [17] and Howell et al in 2007 [19] both found significantly increased mortality in patients with infection who had lactates greater than 4 mmol/L, compared with those with lactates less than 4 mmol/L. Lactate has been reported to be associated with increased bleeding in moderate upper GI bleeding in ED patients [21], but until now, there has been no guidance pertaining to the routine use of lactate specifically for risk stratification of patients with GIH regardless of location of hemorrhage.

Goals of this investigation

The aim of this investigation was to study whether ED lactate predicts in-hospital mortality in ED patients admitted to the hospital with GIH. We hypothesized that an elevated serum lactate at ED presentation on initial laboratory studies would predict GIH-related fatality.

Methods

Study design

This was a retrospective medical record review conducted be- tween January 1, 2004, and December 31, 2009. Institutional review board (IRB) approval was obtained from the participating institution with waiver of informed consent.

Setting

This is a single-center study that took place at an urban, tertiary care, academic ED with an annual census of approximately 100 000 visits. Gastroenterologists and general surgeons are available 24 hours daily, 7 days a week, for our patients who require emergent endoscopy or surgical intervention.

Selection of participants

To identify all potential patients with GIH, the electronic medical record was searched by diagnosis for all medical records between January 1, 2004, and December 31, 2009, that contained any of the following strings: GI bleed, intestinal b-, diverticulo, rectal b-, anal hemorr-, melen-, esophageal v-, varic-, hemorrhage of rectum, hemor- rhage, esophageal hem-, angiod-, mallor-, and tarry. Most medical records (70%) featured the diagnosis of “GI Bleed” as the system prompted providers to pick from a list when entering the final diagnosis. An additional 246 medical records were identified after the study was completed with the diagnosis of “Acute Gastroint Hemorr” but were not included because the IRB-approved study period was complete.

Data collection and processing

The principal investigator (A.S.) trained 5 research assistants (RAs) to manually abstract the medical records in the components of stan- dardized data collection to guarantee consistency between abstrac- tions according to standard guidelines [22]. One investigator (A.S.) formally reviewed 5% of the RAs medical record reviews, as part of their training and required at least 95% agreement to progress to study participation. The RAs were not blind to the study purpose or hypothesis; in addition, RAs were intermittently observed abstracting data and randomly checked medical records to evaluate for accu- racy. No data were found to be erroneous during the random checks. Four medical records were found to have missing data but were retrieved by the principal investigator prior to data analysis. The medical records were retrospectively queried for age, sex, medical history, procedures, need for surgery, length of stay, ICU days, survival

to discharge, and blood products consumed. Triage vitals and labo- ratory values pertaining to the hospitalization episode of interest were examined: ED venous lactate, hemoglobin/hematocrit, and in- ternational normalized ratio values were specifically isolated by re- viewing the ED medical record.

Outcome measures

The primary outcome, mortality, was determined by discharge status of “expiration” or as noted in the hospital discharge summary. The cause and timing of death were not reviewed for this study.

Methods of measurement

The venous lactate was drawn, at the discretion of the ED health care provider (resident, physician assistant, or attending) at the bedside in the ED and measured by a substrate-specific electrode on a Radiometer ABL 300; Radiometer, Denmark. Lactates were not adjusted for liver disease. Serum was analyzed using ampero-metric electrodes with enzymatic membranes and run using 1 of 3 analyzers: (1) Radiometer ABL 300, (2) NovaCCX; Nova Biomedical, Waltham, MA, or

(3) GEM Premier 3000; Instrumentation Laboratory, Bedford, MA. According to the manufacturer of all 3 machines, the range of normal values for venous serum lactate concentration is 1.0 to 2.5 mmol/L.

Primary data analysis

Categorical and continuous clinically relevant variables were assessed using the ?² and t test, respectively, with 2-tailed ? equal to

.05. The variables were analyzed for goodness of fit. receiver operating characteristic curves were created to determine the Diagnostic test characteristics and the optimal cut point for serum lactate con- centration. The optimal cut point was defined as the serum lactate concentration that maximized the sum of sensitivity plus specificity, rounded to the nearest integer. Controlling for age, heart rate, and initial hematocrit, computer analysis was performed using stepwise multi- variable logistic regression analysis, via SPSS version 17 software (IBM, Chicago, IL). The authors did not assess for effect modification.

Sample size and power

Assuming that the control group would have a mean serum lactate concentration of 2 mmol/L, we estimated the need to enroll 1500 patients. This would yield 90% power to detect a 0.2-mmol/L difference in mean serum lactate concentration using the t test. The study was powered to prevent overfitting of the model, based on a 3% outcome prevalence.

Results

Enrollment

Of the 2834 medical record visits screened with GIH diagnoses during the study period, 1644 visits had an ED lactate recorded. Of those, no medical records were excluded for missing data. The initial lactate drawn in the ED is that which is being reviewed in this study.

Baseline characteristics

Of this population, the mean age was 56.8 (13-102) years, 47% were female, and the mean hematocrit was 30.65 (95% confidence interval [CI], 30.26-31.04). Demographics and medical history are summarized in Table 1.

Study outcomes

Among the 1644 visits surveyed whose lactate levels were measured in the ED, 3% died (95% CI, 2%-4%) during the hospital admission for acute GI bleed.

Fig. Lactate outperforms age and delta hematocrit (Hct).

Table 1 Demographics and medical history of study sample
	Total	Deaths in group
Sex Female	775	18
Male	869	30
Age (y) 13-35	119	1
36-55	345	13
56-65	308	6
>=66	872	28
Medical history Hypertension	740	8
Diabetes mellitus	215	9
Chronic anemia	118	1
End-stage renal disease	204	4
Atrial fibrillation	145	4
Hepatitis	161	1
Cirrhosis	118	4
end-stage liver disease	46	4
Cancer	251	11
HIV	66	2
Cerebral vascular accident	149	5
GI bleed or hemorrhage	125	0
Thrombosis or polycythemia vera (PCV)	64	3
Bleeding dyscrasia	34	1
Coagulopathy	10	1
inflammatory bowel disease	35	1
GI ulcer	61	1

Lactate data

Mean (SD) lactate for all subjects was 1.95 (0.04) mmol/L and was higher in patients with the study outcome (t test, P b .001); the mean lactate for those who survived was 1.9 mmol/L, whereas the mean lactate for those who died was 4.6 mmol/L. Controlling for age, initial hematocrit, and heart rate, as possible confounders, for every point increase in lactate, there is a 1.4-fold increase in the odds of a patient’s mortality (see Table 2). A lactate level greater than 4 mmol/L conferred a 6.4-fold increased odds of mortality (see Table 2). Mean lactates associated with ED diagnoses are included in Table 3.

Receiver operating characteristic analysis

Based on ROC curve analysis (c statistic = 0.74, P b .001), the optimal lactate cut point was 4.0 mmol/L, which had a 94% specificity and a 33% sensitivity, and conferred 6.4 times increased odds of mortality (odds ratio [OR], 7.4; CI, 3.3-12.577). Thirty-three percent who died had a lactate level of 4 mmol/L or higher. Unlike lactate, initial hematocrit (P = .236) and heart rate (P = .811) were not significantly associated with mortality (see Fig.).

To determine whether venous lactate was associated with mor- tality, we constructed a multivariate regression model. Confounding variables included hematocrit, age, and heart rate. Even adjusting for these factors, lactate was associated with mortality (OR, 1.4; CI, 1.289- 1.551). The results are summarized in Table 2.

Multivariable logistic regression

lactate, hematocrit, blood pressure, and age. The lactate coefficient was the only result that was significant, indicating that an elevated lactate was independently correlated with death. To control for confounding variables, we then performed stepwise multivariable logistic regression analysis. The model was constructed using potential confounders determined a priori: age, hematocrit, heart rate, and lactate. Use of a lactate cut point of 4 mmol/L for the model was determined using ROC curve analysis. The adjusted OR for lactate greater than 4 mmol/L was

6.4 (95% CI, 3.3-12.4) when controlling for confounders. Adjusted and unadjusted ORs for use of lactate in GIH are summarized in Table 2.

Discussion

The main result of this study is that lactate was a useful biomarker to risk stratify for mortality in adult ED patients with GIH. We found that the optimal cut point for risk stratification using lactate was 4 mmol/L. Surprisingly, we also found that a low hematocrit (requiring transfu- sion), change in hematocrit, and rapid heart rate at presentation were not associated with death.

Importantly, some patients had elevated lactates without abnor- mal hematocrit or heart rates. Therefore, lactate may be helpful even

Table 3

GI bleed/hemorrhage type and mean lactate

Diagnosis % (n) Mean lactate (mg/dL)

We first analyzed the correlation between lactate and death; then we performed a linear regression examining death as a function of								Diverticulosis esophageal varices/hemorrhage	1.8 (29) 1.3 (21)	1.2 2.6
								Gastritis	0.36 (6)	1.1
								GI bleed/hemorrhage	81.3 (1339)	2.01
Table 2								Hematemesis	0.18 (3)	2.47
Adjusted and unadjusted ORs for use of lactate in GIH								Hemorrhage of rectum and anus	1.0 (17)	1.24
								Hemorrhoids	0.18 (3)	1.9
Lactate Unadjusted Adjusteda								Lower GI bleed	4.3 (71)	1.71
	OR	95% CI		OR	95% CI		Melena Rectal anal hemorrhage		3.3 (54) 0.18 (3)	1.84 2.77
Lactate b4 mg/dL	1.44	1.37-1.51		1.41	1.29-1.55		Rectal bleeding		8.2 (135)	1.55
Lactate N 4 mg/dL	7.81	4.1-14.7		6.44	3.3-12.6		Ulcers		0.55 (9)	0.5
a Adjusted model controlled for age (in years), initial heart rate (in beats per Upper GI bleed 2.7 (45) 2.64

minute), and hematocrit (in percent). Some patients received multiple diagnoses.

in seemingly stable patients with GIH in the ED. Early identification of high-risk patients may provide a better assessment of who requires more rapid interventions (eg, endoscopy). Prompt recognition of potentially ill patients may allow providers to prevent poor outcomes. Hyperlactatemia in the setting of GIH may occur from many mech- anisms, all of which suggest major metabolic insult and systemic compromise. Lactate is produced intracellularly when anaerobic energy production takes place. Anaerobic activity occurs when aerobic energy production is insufficient or unavailable. This happens in many conditions of severe illness, for example, in low-volume states, when the peripheral vasculature vasoconstricts to allow for perfusion of key, end-organs. The areas from which blood and, thus, oxygen, are shunted still require ATP to function and generate lactate as a by- product of efforts to maintain homeostasis. This can occur even before

outwardly obvious signs of hypoperfusion, like hypotension.

Other studies have indicated the usefulness of early lactate for prognostication purposes. In 2008, Jansen et al [15] found the mortality rate for patients presenting to the ED with Abnormal vital signs or a lowered Glasgow Coma Scale. In this study, the authors found that a lactate cut point of 3.5 mmol/L correlated with a mortality rate of 41% to 47% compared with those with a lactate level of less than 3.5 mmol/L who had a mortality rate of 12% to 15%. Shapiro et al [17] found that in ED patients with infection and a lactate level of 4 mmol/L or higher was 36% sensitive and 92% specific for predicting mortality; it was 55% sensitive and 91% sensitive for death within 3 days of presentation. In our study, we also found that a lactate of level 4 mmol/L had similar implications, supporting the use of this cut point for risk stratification.

This is a retrospective, single-center, cohort study. Although the findings are in line with prior work, a prospective, multicenter, blinded investigation examining the usefulness of early lactate in risk stratification will aid in validating our findings. If similar results are found, practitioners can confidently identify and stratify patients with GI hemorrhage at higher risk for death.

Limitations

A large subset of patients were excluded due to absence of ED serum lactate, which may have biased the lactate cut point data; however, this would probably bias toward the null hypothesis because clinicians are typically more likely to draw lactate in more severely ill patients. Furthermore, 3 different lactate machines were used in clinical practice, and thus, the reference ranges may differ slightly, changing the accuracy and precision of the data. Importantly, this study took place at only one site, perhaps limiting its gene- ralizability; however, this urban ED is located in one of the most diverse zip codes in the nation, according to most recent Census data. As mentioned in the Methods section, some ED visits were missed by the original search terms and not included in the study because IRB approval was already completed. It is possible that additional GIH medical records were not captured by our search strategy. This might bias our results if certain diagnoses are asso- ciated with higher mortality but not lactate rise. However, given our large sample size, the results are not likely to be different, unless a large number of GIH deaths were missed. This is unlikely because all ED death records were subsequently reviewed, and no additional

GIH diagnoses were identified.

Another possible limitation of this study is that the ED final Diagnosis codes were used to determine which medical records were eligible for review. The initial diagnoses at presentation may not have been the actual disease process present in a particular patient. In addition, the abstractors were not blind to the study hypothesis and may have been subject to reviewer bias, although this seems less

likely given the nature of the objective outcome measured. We did not calculate formal ? values for interrater reliability; however, we assessed 5% of medical records for more than 95% agreement as part of the abstraction training process. Finally, time to the obtaining serum lactate concentration, as well as whether the source was arterial or venous, may also limit interpretation of our data. However, our data represent real-world scenario, which adds credence to the overall concept of generalizability.

Conclusion

This study demonstrates that initial lactate drawn in the ED can be used to identify patients with higher in-hospital mortality with acute GIH. More research is needed to validate the findings described here. For patients presenting with a complaint or diagnosis of GIH, an initial ED lactate cut point of at least 4 mmol/L can help with early-risk stratification.

References

1. Imperiale TF, Dominitz JA, Provenzale DT, Boes LP, Rose CM, Bowers JC, et al. Predicting poor outcome from acute Upper gastrointestinal hemorrhage. Arch Intern Med 2007;167:1291-6.
2. Cagir B, Kate J, editors. lower gastrointestinal bleeding. Medscape; 2011.
3. Fallam MA, Prakash C, Edmundowicz S. Acute gastrointestinal bleeding. Med Clin North Am 2000;84(5):1183-208.
4. Gilbert DA. Epidemiology of upper gastrointestinal bleeding. Gastrointest Endosc 1990;36(Suppl):58-513.
5. Johanson JF. Curbing the costs of GI bleeding. Am J Gastroenterol 1998;93:1384-5.
6. Quirk DM, Barry MJ, Aserkogg B. Physician specialty and variations in the cost of treating patients with Acute upper gastrointestinal bleeding. Gastroenterology 1997;113:1443-8.
7. Cooper GS, Chak A, Way LE, Hammer pJ, Harper DL, Rosenthal GE. Endoscopic practice for upper gastrointestinal hemorrhage; differences between major teaching and community-based hospitals. Gastrointest Endosc 1998;48:348-53.
8. Rockall TA, Logan RF, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut 1996;38(3):316-21.
9. Manini AF, Kumar A, Olsen D, Vlahov D, Hoffman RS. Utility of serum lactate to predict drug-overdose fatality. Clin Toxicol (Phila) 2010;48(7):730-6.
10. Sobhian B, Kropfl A, Holzenbein T, Khadem A, Redl H, Bahrami S. Increased circulating D-lactate levels predict risk of mortality after hemorrhage and surgical trauma in baboons. Shock 2012;37(5):473-7.
11. Hajjar A, Nakamura RE, de Almeida JP, Fukushima JT, Hoff PM, Vincent JL, et al. Lactate and Base deficit are predictors of mortality in critically ill patients with cancer. Clinics (San Paolo) 2011;66(12):2037-42.
12. Maarslet L, Moller MB, Dall R, Hjortholm K, Raun H. Lactate levels predict mortality and need for Peritoneal dialysis in children undergoing congenital heart surgery. Acta Anaesthesiol Scand 2012;56(4):459-64.
13. Neville AL, Nemtsev D, Manasrah R, Bricker SD, Putnam BA. Mortality risk stratification in elderly trauma patients based on initial arterial lactate and base deficit levels. Am Surg 2011;77(10):1337-41.
14. Nichol A, Bailey M, Egi M, Pettila V, French C, Stachewski E, et al. Dynamic lactate indices as predictors of outcome in critically ill patients. Crit Care 2011;15(5): R242.
15. Jansen TC, van Bommel J, Mulder pg, Rommes JH, Schieveld SJM, Bakker J. The prognostic value of Blood lactate levels relative to that of vital signs in the prehospital setting: a pilot study. Crit Care 2008;12:R160. http://dx.doi.org/ 10.1186/cc7159.
16. Bakker J, Gris P, Coffernils M, Kahn RJ, Vincent JL. Serial blood lactate levels can predict the development of multiple organ failure following septic shock. Am J Surg 1996;171(2):221-6.
17. Shapiro NI, Howell MD, Talmor D, Nathanson LA, Lisbon A, Wolfe RE, et al. Serum lactate as a predictor of mortality in emergency department patients with infection. Ann Emerg Med 2005;45(5):524-8.
18. Nguyen HB, Loomba M, Yang JJ, Jacobesen G, Shah K, Otero RM, et al. Early lactate clearance is associated with biomarkers of inflammation, coagulation, apoptosis, organ dysfunction and mortality in severe sepsis and septic shock. J Inflamm (Lond) 2010;7:6.
19. Howell MD, Donnino M, Clardy P, Talmor D, Shapiro NI. Occult hypoperfusion and mortality in patients with suspected infection. Intensive Care Med 2007;33 (11):1892-9.
20. Bakker J, Jansen TC. Don’t take vitals, take a lactate. Intensive Care Med 2007;33 (11):1863-5.
21. Wada T, Hagiwara A, Yahagi N, et al. Lactate clearance is a predictor of sustained bleeding in emergency room patients with moderate upper gastrointestinal bleeding. Abstracted in Critical care 2013;17(Suppl 2: 239).
22. Gilbert EH, Lowenstein SR, Koziol J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1996;27:305-8.