American Journal of Emergency Medicine (2012) 30, 1667-1673

Original Contribution

Impact of metformin use on the prognostic value of lactate in sepsis^?,??

Jeffrey P. Green MD ^a,?, Tony Berger MD, MS ^a, Nidhi Garg MD ^b, Alison Suarez MD ^b, Yolanda Hagar PhD ^c, Michael S. Radeos MD, MPH ^b, Edward A. Panacek MD, MPH ^a

^aDepartment of Emergency Medicine, UC Davis School of Medicine, Sacramento, CA 95817, USA

^bDepartment of Emergency Medicine, Weill Cornell Medical College, New York Hospital Queens, Flushing, NY 11355, USA

^cDepartment of Public Health Sciences and Biostatistics, University of California, Davis, Davis, CA 95616, USA

Received 13 December 2011; revised 11 January 2012; accepted 16 January 2012

Abstract

Objective: The objective of this study is to determine if metformin use affects the prevalence and prognostic value of hyperlactatemia to predict mortality in septic adult emergency department (ED) patients.

Methods: This is a single-center retrospective cohort study. Emergency department providers identified study subjects; data were collected from the medical record.

Patients: Adult ED patients with suspected infection and 2 or more systemic inflammatory response syndrome criteria were included. The outcome was 28-day mortality. The primary risk variable was serum lactate (b2.0, 2.0-3.9, >=4.0 mmol/L) categorized by metformin use; covariates: demographics, Predisposition, Infection, Response, Organ Dysfunction score and metformin use contraindications.

Setting: The study was conducted at an urban teaching hospital; February 1, 2007 to October 31, 2008. Results: A total of 1947 ED patients were enrolled; 192 (10%) were taking metformin; 305 (16%) died within 28 days. Metformin users had higher median lactate levels than nonusers (2.2 mmol/L [interquartile range, 1.6-3.2] vs 1.9 mmol/L [interquartile range, 1.3-2.8]) and a higher, although nonsignificant, prevalence of hyperlactatemia (lactate >=4.0 mmol/L) (17% vs 13%) (P = .17). In multivariate analysis (reference group nonmetformin users, lactate b2.0 mmol/L), hyperlactatemia was associated with an increased adjusted 28-day mortality risk among nonmetformin users (odds ratio [OR], 3.18; P b .01) but not among metformin users (OR, 0.54; P = .33). In addition, nonmetformin users had a higher adjusted mortality risk than metformin users (OR, 2.49; P b .01). These differences remained significant when only Diabetic patients were analyzed.

^? Preliminary results previously presented in abstract form at 2010 SAEM Annual Meeting.

^?? Financial/nonfinancial disclosures: This publication was made possible in part by grant number 2UL 1RR024146 from the National Center for Research Resources, a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research.

* Corresponding author. Department of Emergency Medicine, UC Davis Health System, Sacramento, CA 95817, USA. Tel.: +1 646 942 8290; fax: +1 916

734 7950.

E-mail addresses: [email protected], [email protected] (J.P. Green), [email protected] (T. Berger), [email protected] (N. Garg), [email protected] (A. Suarez), [email protected] (Y. Hagar), [email protected] (M.S. Radeos), [email protected] (E.A. Panacek).

0735-6757/$ - see front matter (C) 2012 http://dx.doi.org/10.1016/j.ajem.2012.01.014

Conclusions: In this study of adult ED patients with suspected sepsis, metformin users had slightly higher median lactate levels and prevalence of hyperlactatemia. However, hyperlactatemia did not predict an increased mortality risk in patients taking metformin.

(C) 2012

Introduction

Each year in the United States, approximately 2.3 million adult patients present to emergency departments (EDs) for suspected severe sepsis, resulting in more than 200 000 deaths [1,2]. Accurate risk stratification of this population is essential to optimize treatment and focus limited resources on high-risk patients. elevated lactate levels are associated with increased mortality risk in severe sepsis and are the most common serologic test used for risk stratification [3-5]. In addition, hyperlactatemia (lactate >=4.0 mmol/L) is an enrollment criterion for early goal-directed therapy , a protocol believed to reduce short-term mortality in sepsis [6-8]. Although potentially beneficial in high-risk patients, EGDT is also resource consumptive and requires invasive procedures with their associated complications. Hyperlactatemia may be caused by conditions other than sepsis [9]. Even in sepsis, the degree of lactate elevation may be affected by multiple factors, which could impact its prognostic value [10]. Because lactate levels routinely impact vital Clinical decisions, it is essential to identify factors such as metformin use that may cause hyperlactatemia indepen- dent of sepsis severity.

Metformin is an oral antihyperglycemic agent in the biguanide class. Biguanides are known to interfere with mitochondrial metabolism and inhibit hepatic uptake of serum lactate [11,12]. Unlike other biguanides, metformin use does not appear to increase the prevalence of hyperlacta- temia [13]. However, during periods of physiologic stress such as sepsis, metformin use is thought to increase the likelihood of developing hyperlactatemia. Because of this concern, the Food and Drug Administration recommends withholding metformin in patients with probable sepsis [14]. Given this knowledge, the prevalence of hyperlactatemia among septic metformin users requires quantification.

Because metformin use could affect lactate levels, it could also interfere with the prognostic value of lactate in sepsis. A metformin user who becomes septic could accumulate lactate as a direct result of sepsis, from the impact that diabetes and metformin use have on cellular metabolism, or through a combination of these physiologic mechanisms [15]. Hyper- lactatemia in sepsis has consistently been shown to be associated with increased mortality risk [3-5], but the association of hyperlactatemia with mortality due to metfor- min use is less clear [13]. If hyperlactatemia’s association with mortality risk in sepsis is affected by metformin use, then risk stratification using lactate in these patients would be less useful. The objective of this study was to determine if metformin use affects the prevalence and prognostic value of

hyperlactatemia to predict mortality in adult ED patients with suspected sepsis.

Methods

Study design

We performed a single-center retrospective cohort study of adult patients hospitalized from a single, urban ED for suspected sepsis. Findings from a portion of this cohort have been reported previously [16]. The institutional review board of New York Hospital Queens Hospital approved the study with a waiver of informed consent.

Study setting and population

A protocol was in place during the study to routinely test a serum lactate level and other markers of organ dysfunction on adult patients having laboratory studies in the ED for a suspected infection, as recommended by Consensus guidelines [17]. All patients with lactate testing during the study were evaluated for inclusion. Patients were enrolled in the study if they were adults (>=21 years), had a serum lactate level tested in the ED, had a provider-suspected infection as reported in the electronic medical record (EMR) and 2 or more Systemic Inftammatory Response Syndrome (SIRS) Criteria (using initial ED vital signs and laboratory studies). The SIRS criteria include body temperature less than 96.8?F or greater than 100.4?F, heart rate more than 90 beats per minute, respiratory rate more than 20 breaths per minute, and a white blood cell count less than 4000 cells/mm³ or greater than 12 000 cells/mm³ or greater than 10% immature neutrophils (band forms).

The study was entirely performed at New York Hospital Queens Hospital, a 450-bed urban teaching hospital with an annual ED census of 95 000 patients. For patients with repeat ED visits during the study period, only the initial visit was used. Patients were enrolled from February 1, 2007, to October 31, 2008.

Study protocol

All data were collected from the EMR. Trained research associates abstracted the medical records of all patients with lactate levels tested in the ED during the study period. Published recommendations for quality chart abstraction were followed [18]. Specifically, data abstractors were trained in advance, used standardized data abstraction sheets,

were routinely audited, and were blinded to the study hypothesis. Ten percent of subjects had all variables collected by a second blinded abstractor to confirm reliability of the results (? >= 0.80 for all variables).

Serum lactate (micromoles per liter) levels were measured using a serum-based immunoassay (Unicel Synchron; Beckman Coulter, Inc, Brea, CA). Arterial or venous lactate testing was allowed to improve protocol compliance, similar to other ED lactate screening studies [3-5]. Previous studies have demonstrated that venous lactate levels correlate with arterial lactate levels as well as with short-term mortality risk, in adult ED patients with suspected sepsis [3,19]. Only Initial serum lactate levels, tested before hospital admission, were used. Determination of whether patients were taking metformin on ED arrival was made by review of patient-reported medication use for the initial ED visit. Patients reporting use of any medication containing metformin were considered metfor- min users for study purposes. The primary outcome was 28-day mortality. For patients who were discharged alive before 28 days from the Initial ED encounter, the Social Security Death Index (SSDI) was queried (more than 1 year after the initial ED evaluation) to confirm whether the patient survived to 28 days [20].

Data analysis

Summary and descriptive statistics were generated for the study cohort. The prevalence of hyperlactatemia among metformin and nonmetformin users was investigated using ?2. To quantify associations between metformin use, lactate levels, and mortality, a logistic regression model was deve- loped, with the outcome of 28-day mortality. The primary measure of interest was a metformin/lactate level interaction variable that represented the 6 possible combinations of 3 serum lactate categories (b2.0, 2.0-3.9, >=4.0 mmol/L), categorized by metformin use. The lactate cutoffs correspond to levels previously reported to be associated with different mortality risks in septic adult patients [3,21]. For study pur- poses, only a lactate level 4.0 mmol/L or higher was con- sidered to be hyperlactatemia because it represents a cutoff associated with significant mortality risk that is commonly used to determine need for aggressive interventions [17].

Patient demographics were evaluated for possible inclusion in the model. In addition, to adjust for illness severity, the Predisposition, Infection, Response, Organ Dysfunction (PIRO) score was included [22]. This scoring system was chosen because it has been internally and externally validated in similar populations of adult ED patients with suspected sepsis and includes variables that could be abstracted from the EMR [22]. The PIRO score was developed for the outcome of mortality in sepsis using adjusted odds ratios (ORs) of categorized Serum lactate levels as well as 16 other clinical variables routinely available for septic ED patients. For this analysis, adjusted ORs of categorized lactate levels for mortality were deter-

mined, adjusted by the 16 other PIRO clinical variables. This analysis allowed for the determination of lactate’s association with mortality risk, as modified by other available markers of illness severity.

In addition, the Food and Drug Administration has de- fined relative contraindications to metformin use based on the increased likelihood of lactic acidosis from certain con- ditions (liver and renal dysfunction, serum acidosis, and hypoxia) [14]. However, published reports have demon- strated that physicians frequently prescribe metformin in spite of these contraindications [23] with no apparent impact on the incidence of lactic acidosis or patient outcomes [24,25]. Such patients were, therefore, included in the study, but potential confounding caused by these contraindications was adjusted for by inclusion in the logistic regression model. A medical history of diabetes mellitus was expected to exhibit significant collinearity with metformin use, which raises the possibility that any observed impact of metformin use in the model could actually be caused by the associated medical history of diabetes. To account for this concern, the model was repeated only for the subgroup of patients with a known history of diabetes mellitus.

We quantified the effects of covariates on mortality rates

by including them individually in separate logistic regression models that already included the metformin/lactate level interaction variable. Covariates significant at the P = .05 level were simultaneously included in a multivariate logistic regression model, and those remaining significant at the P = .05 level were retained in the final model. The final model was confirmed using stepwise Akaike Information Criterion methods, which compares the fit of different regression models [26].

Metformin use has not previously been shown to effect mortality in sepsis, but metformin has antioxidant and vasoactive properties that could be protective in sepsis [27-29]. This study was not designed to identify a protective effect of metformin use in sepsis. However, to distinguish any association of metformin use with mortality from met- formin’s impact on lactate’s prognostic use, contrasts of the final model were performed. Contrasts allowed for compar- isons of the different metformin/lactate groups to each other, in addition to reference group comparisons. In this way, the relative impact of metformin use on mortality risk for the entire cohort and among different lactate stratum could be analyzed. Calculations were performed in SAS v9.2 (SAS Institute, Inc, Cary, NC) and R (http://www.r-project.org/).

Results

Lactate levels were tested on 2650 adult patients in the ED during the study period. Of these patients, 1947 (73%) had 2 or more SIRS criteria and an ED provider-suspected infection. These patients made up the study population. Baseline characteristics are reported in Table 1. A total of 1755 subjects were not taking metformin, whereas 192

Table 1 Patient demographics and clinical characteristics for all subjects stratified by metformin use and 28-day mortality

Variable	All subjects	Nonmetformin users	Metformin users	Survivors	Nonsurvivors
n (%)	1947	1755 (90)	192 (10)	1642 (84)	305 (16%)
Age (y)	72 +- 17.4	72 +- 17.7	71 +- 13.7	70 +- 17.6	81 +- 12.5
Sex (male), n (%)	938 (48%)	846 (48%)	92 (48%)	801 (49%)	137 (45%)
Race
White	928 (48%)	851 (48%)	77 (40%)	755 (46%)	173 (57%)
Asian	384 (20%)	342 (19%)	42 (22%)	323 (20%)	61 (20%)
Black	230 (12%)	199 (11%)	31 (16%)	202 (12%)	28 (9%)
Hispanic	242 (12%)	217 (12%)	25 (13%)	224 (14%)	18 (6%)
Other	161 (8%)	144 (8%)	17 (9%)	136 (8%)	25 (8%)
Not reported	2 (0.1%)	2 (0.1%)	0 (0%)	2 (0.1%)	0 (0%)
Vital signs
Temperature (?F) a	100.3 +- 2.3	100.3 +- 2.3	100.6 +- 2.1	101 +- 2.2	99 +- 2.5
Heart rate (beats per minute) a	105 +- 21.5	105 +- 21.5	105 +- 21.0	105 +- 20.9	104 +- 24.3
Respiratory rate (breaths per minute) a	22 +- 5.2	22 +- 5.2	22 +- 5.2	21 +- 4.9	23 +- 6.1
Systolic blood pressure (mm Hg) a	126 +- 31.1	125 +- 30.9	133 +- 31.8	128 +- 30.4	112 +- 31.0
Oxygen saturation (%) a
Median (IQR)	97 (95-99)	97 (95-99)	96 (94-98)	97 (95-99)	97 (93-100)
Comorbidities
Diabetes mellitus	535 (27%)	343 (20%)	192 (100%)	456 (28%)	79 (26%)
Chronic pulmonary disease	479 (25%)	432 (25%)	47 (24%)	382 (23%)	97 (32%)
Chronic renal insufficiency	316 (16%)	297 (17%)	19 (10%)	237 (14%)	79 (26%)
Congestive heart failure	425 (22%)	392 (22%)	33 (17%)	320 (19%)	105 (34%)
Sepsis categories
Suspected sepsis	387 (20%)	350 (20%)	37 (19%)	367 (22%)	20 (7%)
Severe sepsis b	1334 (69%)	1196 (68%)	138 (72%)	1125 (69%)	209 (69%)
Septic shock c	226 (12%)	209 (12%)	17 (9%)	150 (9%)	76 (25%)
Laboratory data
White blood cell count (103/mm3)	13 +- 8.0	14 +- 8.3	12 +- 5.4	13 +- 7.1	16 +- 11.6
Immature neutrophils (103/mm3)
Median (IQR)	1 (0-13)	1 (0-13)	1 (0-14)	1 (0-13)	2 (0-17)
Platelets (103/mm3)	258 +- 129	258 +- 129	262 +- 120	255 +- 123	277 +- 152
Glucose (mg/dL)	158 +- 105	153 +- 102	210 +- 117	158 +- 107	159 +- 89
Creatinine (mg/dL)	1.6 +- 1.4	1.6 +- 1.5	1.3 +- 0.7	1.5 +- 1.4	1.9 +- 1.5
Bilirubin, total (mg/dL)	1.2 +- 0.9	1.2 +- 0.9	1.1 +- 0.5	1.1 +- 0.8	1.4 +- 1.4
Anion gap (mEq/L)	10.3 +- 3.8	10.2 +- 4.0	10.5 +- 3.3	10.0 +- 3.5	11.5 +- 4.8
Lactate (mmol/L)
Median (IQR)	1.9 (1.3-2.9)	1.9 (1.3-2.8)	2.2 (1.6-3.2)	1.8 (1.2-2.7)	2.5 (1.7-4.5)
Categorical lactate
0-1.9 mmol/L	994 (51%)	919 (52%)	75 (39%)	891 (54%)	103 (34%)
2.0-3.9 mmol/L	691 (35%)	606 (35%)	85 (44%)	578 (35%)	113 (37%)
>=4.0 mmol/L	262 (13%)	230 (13%)	32 (17%)	173 (11%)	89 (29%)
PIRO score	11 +- 5.0	11 +- 5.0	11 +- 4.7	10 +- 4.8	14 +- 4.8
Metformin use	192 (10%)	-	-	177 (11%)	15 (5%)
28-d mortality	305 (16%)	290 (17%)	15 (8%)	-	-
Continuous, parametric values presented as mean +- SD; continuous, nonparametric data presented as median (IQR); and categorical data presented as frequency (percentage) among groups. a Triage or first ED vital sign recorded. b Defined as sepsis with objective evidence of organ dysfunction [22]. c Defined as persistent hypotension (systolic blood pressure, b90 mm Hg) following an isotonic ftuid bolus.

subjects (10%) were taking metformin. Subjects taking met- formin had a higher median lactate level than nonmetformin users (2.2 mmol/L [interquartile range, or IQR, 1.6-3.2] vs 1.9 mmol/L [IQR, 1.3-2.8]). In addition, the prevalence of hyperlactatemia was higher among metformin users than nonusers, although this finding did not achieve statistical

significance (17% vs 13%; P = .17). Twenty-eight-day all- cause mortality for the cohort was 16% (95% confidence interval [CI], 14.0%-17.3%) (Table 1).

Among all subjects, elevated lactate levels were associated with higher 28-day mortality (Table 1). In addition, metformin users had a significantly lower

A. Model for all patients in the study cohort (n = 1947)
Nonmetformin users	b2.0 mmol/L	919	Reference a
	2.0-3.9 mmol/L	606	1.27	0.92-1.78	.15
	>=4.0 mmol/L	230	3.18	2.12-4.66	b.01
Metformin users	b2.0 mmol/L	75	0.59	0.21-1.70	.32
	2.0-3.9 mmol/L	85	0.77	0.34-1.78	.54
	>=4.0 mmol/L	32	0.54	0.15-1.83	.33
B. Model including only diabetic patients within the study cohort (n = 535)
Nonmetformin users	b2.0 mmol/L	168	Reference a
	2.0-3.9 mmol/L	127	1.41	(0.72-2.78)	.32
	>=4.0 mmol/L	48	3.68	(1.62-8.38)	b.01
Metformin users	b2.0 mmol/L	75	0.60	(0.19-1.88)	.38
	2.0-3.9 mmol/L	85	0.81	(0.32-2.07)	.66
	>=4.0 mmol/L	32	0.60	(0.16-2.25)	.45

unadjusted 28-day mortality risk than nonmetformin users (8% vs 17%; P b .01). multivariate regression models showed that compared with the reference group of nonmetformin users with lactate levels less than 2.0 mmol/L, nonmetformin users with lactate levels 4.0 mmol/L or higher had a higher adjusted mortality risk (OR, 3.18; P b .01). There was no significant difference in mortality risk between the reference group and any of the metformin users, regardless of categorized lactate levels (Table 2A). The modified PIRO score, patient age, and creatinine level greater than 1.4 mg/dL were also found to be significant predictors of mortality and were included in the model. Sex, Serum bilirubin, race, diabetes mellitus, anion gap, and percent oxygen saturation were not significant predictors of mortality at the P = .05 level. Results of the multivariate analysis were similar when restricted to only diabetic patients in the study population (Table 2B).

Table 2 Multiple logistic regression demonstrating adjusted OR for 28-day inpatient mortality categorized by metformin use and categorized lactate

Parameter Lactate stratum n OR 95% CI P

The model was adjusted for patient age (OR, 1.04 for every year older than 21; P b .01), modified PIRO score (OR, 1.11 for each point increase in scoring system; P b .01), and serum creatinine greater than 1.4 mg/dL (OR, 1.77; P = .01).

^a Reference group: 21-year-old subject with a lactate less than 2.0 mmol/L not taking metformin.

Contrasts were performed to determine the effect of metformin use on mortality risk. Among all study subjects, after adjustment for lactate level, modified PIRO score, age, and serum creatinine level, individuals not taking metformin were 2.5 times more likely to die within 28 days when compared with individuals taking metformin (OR, 2.49; 95% CI, 1.38-4.73; P b .01). These results were similar when the analysis was restricted only to diabetic patients in the study cohort (OR, 2.62; 95% CI, 1.32-5.17; P b .01). Among patients with low to moderate lactate elevation (b4.0 mmol/L), there was no difference in adjusted mortality risk when stratified by metformin use. In contrast, patients with lactate levels 4.0 mmol/L or higher not taking metformin had more than a 5 times greater adjusted mortality risk than metformin users (OR, 5.95; 95% CI, 1.68-21.10; P = .01). This finding

was similar when only diabetic patients were included in the analysis (OR, 6.16; 95% CI, 1.52-25.02; P = .01).

Discussion

In this study of adult ED patients with suspected sepsis, metformin use was associated with slightly higher median lactate levels and prevalence of hyperlactatemia, although this difference was not statistically significant. As demon- strated in previous studies [3,4], 28-day mortality risk was higher in this cohort with increasing categorized lactate levels. However, among metformin users, we did not observe the same association between elevated lactate levels and increased mortality. Indeed, we found no association whatsoever between hyperlactatemia and mortality in patients taking metformin in the study population.

The lack of association between hyperlactatemia and mortality risk for metformin users could be caused by several factors. Metformin users could have been less ill at study entry than nonusers, but in this study, metformin users had similar illness severity to nonmetformin users. In addition, when adjusted for illness severity, hyperlactatemia was still not associated with mortality risk among metformin users.

Metformin use could lead to lactate accumulation by a mechanism distinct from that of sepsis. This accumulated lactate could affect the prevalence of hyperlactatemia among metformin users, resulting in a different association of lactate with mortality risk in these patients. Metformin users had a slightly higher prevalence of hyperlactatemia in this study, but hyperlactatemia had a much lower associated mortality risk in these patients. In addition, contrasts demonstrated that the impact of metformin use on mortality was most

pronounced at high lactate levels, implying that lactate accu- mulation may have partially occurred by a different mechanism in these patients. Causality cannot be determined from this study design, but further investigation of this find- ing is warranted.

Alternatively, metformin use could be protective in the setting of sepsis, which could impact the association of lactate with mortality risk in these patients. In this study, nonmetformin users had higher 28-day mortality risk than metformin users, even when adjusted for lactate levels and other important covariates. This study was not designed to identify a protective effect of metformin use in sepsis, but metformin does have potentially beneficial physiologic ef- fects in sepsis, including vasoactive and antiinftammatory properties [27,28,30]. In addition, metformin has been found to inhibit expression of lipopolysaccharide-induced nitric oxide synthase in an Experimental model, which could decrease systemic vasodilatation in sepsis [29]. These phy- siologic characteristics of metformin use could have a pro- tective effect in septic patients, but further study is needed to determine this association.

Previous research has demonstrated that diabetic patients

may have equivalent or better outcomes from critical illness than nondiabetic patients, even when adjusted for illness severity [31,32]. Diabetic patients could also have a different prevalence and prognostic value of lactate levels in sepsis. As expected, metformin use was strongly associated with a history of diabetes in this study. To account for the possi- bility that the observed impact metformin use had on lactate levels and mortality could actually have been related to an associated medical history of diabetes, the analysis was re- peated among only diabetic patients. Results of this analysis were similar to that of the overall study population. This finding demonstrates that the apparent effect of metformin use on the prognostic value of hyperlactatemia in sepsis is more likely associated with metformin use than with a his- tory of diabetes mellitus.

This study had multiple limitations. First, providers in the ED prospectively identified patients, but all data were collected from the EMR. The risk of misclassification bias was mitigated by following previously published guidelines for medical record abstraction [18] and by restricting the analysis to objective clinical findings (demographics, comor- bidities, laboratory results, and vital signs). Metformin use was also collected from the EMR, and that information could have been reported in error. However, during the study period, a protocol was in place whereby the ED triage nurse recorded the patient’s current medication use and the treating ED provider confirmed the list, decreasing the likelihood of misclassification error.

In addition, hyperlactatemia is an enrollment criterion for EGDT. Performance of EGDT was recommended during the study period for patients with persistent hypo- tension (systolic blood pressure b90 mm Hg or lactate

>=4.0 mmol/L), but compliance with the protocol was

limited. A quality assurance review of EGDT protocol com-

pliance during a portion of the study period found that less than 20% of EGDT candidates had the 6-hour therapeutic bundle completed as recommended. Enrollment in this clinical intervention was not measured for the entire cohort, but the relatively large size of the study and low protocol compliance make it less likely that EGDT use would be significantly imbalanced between groups.

A further limitation is that patients were not followed up prospectively after discharge from the hospital. Use of the SSDI to determine survival outcomes of discharged patients has previously been validated [20], but it is possible that some patients discharged before 28 days from the initial ED encounter may have died and were not entered into the SSDI. Finally, the study was entirely performed at 1 hospital. The external validity of these results is unknown, and further study, in other settings, may be warranted.

We found that ED patients hospitalized for suspected sepsis who were on metformin had a slightly higher pre- valence of hyperlactatemia than nonmetformin users. In addition, in this cohort, categorized serum lactate levels did not demonstrate useful prognostic utility in adult ED patients with suspected sepsis who were actively taking metformin. Alternative Prognostic markers for mortality risk should be considered in these patients.

Acknowledgments

The authors acknowledge Drs James Ryan, Diane Sixsmith, and David Barlas for their contribution to the development of this project; the research associates at New York Hospital Queens for their tireless data collection; and Dr James Holmes for his review of the manuscript.

References

1. Wang HE, Shapiro NI, Angus DC, et al. National estimates of severe sepsis in United States emergency departments. Crit Care Med 2007; 35(8):1928-36.
2. 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(7):1303-10.
3. Mikkelsen ME, Miltiades AN, Gaieski DF, 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.
4. Howell MD, Donnino M, Clardy P, et al. Occult hypoperfusion and mortality in patients with suspected infection. Intensive Care Med 2007;33(11):1892-9.
5. Trzeciak S, Dellinger RP, Chansky ME, et al. Serum lactate as a predictor of mortality in patients with infection. Intensive Care Med 2007;33(6):970-7.
6. 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(19):1368-77.
7. Shapiro NI, Howell MD, Talmor D, et al. Implementation and outcomes of the Multiple Urgent Sepsis Therapies (MUST) protocol. Crit Care Med 2006;34(4):1025-32.
8. Nguyen HB, Corbett SW, Steele R, et al. Implementation of a bundle of quality indicators for the early management of severe sepsis and

septic shock is associated with decreased mortality. Crit Care Med 2007;35(4):1105-12.

1. Juneja D, Singh O, Dang R. Admission hyperlactatemia: causes, incidence, and impact on outcome of patients admitted in a general medical intensive care unit. J Crit Care 2011;26(3):316-20.
2. Levy B. Lactate and shock state: the metabolic view. Curr Opin Crit Care 2006;12(4):315-21.
3. Correia S, Carvalho C, Santos MS, et al. Mechanisms of action of metformin in type 2 diabetes and associated complications: an overview. Mini Rev Med Chem 2008;8(13):1343-54.
4. Lalau JD. Lactic acidosis induced by metformin: incidence, manage- ment and prevention. Drug Saf 2010;33(9):727-40.
5. Salpeter SR, Greyber E, Pasternak GA, et al. Risk of fatal and nonfatal lactic acidosis with metformin use in type 2 diabetes mellitus. Cochrane Database Syst Rev 2010(1):CD002967.
6. Distributed by Bristol-Meyers Squibb Company. Glucophage (met- formin hydrochloride) [package insert]. Princeton (NJ): Bristol-Myers Squibb; 2009.
7. Scale T, Harvey JN. Diabetes, metformin and lactic acidosis. Clin Endocrinol (Oxf) 2011;74(2):191-6.
8. Green JP, Berger T, Garg N, et al. Serum lactate is a better predictor of short-term mortality when stratified by C-reactive protein in adult emergency department patients hospitalized for a suspected infection. Ann Emerg Med 2011;57(3):291-5.
9. Dellinger RP, Levy MM, Carlet JM, et al. Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2008. Crit Care Med 2008;36(1):296-327.
10. Panacek EA. Performing chart review studies. Air Med J 2007;26(5): 206-10.
11. Gallagher EJ, Rodriguez K, Touger M. Agreement between peripheral venous and arterial lactate levels. Ann Emerg Med 1997; 29(4):479-83.
12. Quinn J, Kramer N, McDermott D. Validation of the Social Security Death Index (SSDI): an important readily-available outcomes database for researchers. West J Emerg Med 2008;9(1):6-8.
13. Nichol AD, Egi M, Pettila V, et al. Relative hyperlactatemia and hospital mortality in critically ill patients: a retrospective multi-centre study. Crit Care 2010;14(1):R25.
14. Howell MD, Talmor D, Schuetz P, et al. Proof of principle: the predisposition, infection, response, organ failure sepsis staging system. Crit Care Med 2011;39(2):322-7.
15. Calabrese AT, Coley KC, DaPos SV, et al. Evaluation of prescribing practices: risk of lactic acidosis with metformin therapy. Arch Intern Med 2002;162(4):434-7.
16. Holstein A, Stumvoll M. Contraindications can damage your health- is metformin a case in point? Diabetologia 2005;48(12):2454-9.
17. Schmitt-Honl B. Metformin. Contraindications in heart failure reconsidered. Med Monatsschr Pharm 2008;31(4):148-9.
18. Aikaike H. A new look at statistical model identification. IEEE Trans Inf Theory 1974;19(6):716-23.
19. Detaille D, Guigas B, Chauvin C, et al. Metformin prevents high- glucose-induced endothelial cell death through a mitochondrial permeability transition-dependent process. Diabetes 2005;54(7): 2179-87.
20. Faure P, Wiernsperger N, Polge C, et al. Impairment of the Antioxidant properties of serum albumin in patients with diabetes: protective effects of metformin. Clin Sci (Lond) 2008;114(3):251-6.
21. Kato Y, Koide N, Komatsu T, et al. Metformin attenuates production of nitric oxide in response to lipopolysaccharide by inhibiting MyD88-independent pathway. Horm Metab Res 2010; 42(9):632-6.
22. Carter AM, Bennett CE, Bostock JA, et al. Metformin reduces C-reactive protein but not complement factor C3 in overweight patients with type 2 diabetes mellitus. Diabet Med 2005;22(9):1282-4.
23. Graham BB, Keniston A, Gajic O, et al. Diabetes mellitus does not adversely affect outcomes from a critical illness. Crit Care Med 2010;38(1):16-24.
24. Egi M, Bellomo R, Stachowski E, et al. Blood glucose concentration and outcome of critical illness: the impact of diabetes. Crit Care Med 2008;36(8):2249-55.