Академический Документы
Профессиональный Документы
Культура Документы
DOI 10.1007/s00134-007-0563-9
Stephen Trzeciak
R. Phillip Dellinger
Michael E. Chansky
Ryan C. Arnold
Christa Schorr
Barry Milcarek
Steven M. Hollenberg
Joseph E. Parrillo
ORIGINAL
Introduction
Elevation in serum lactate has been associated with an
increased risk of death in emergency department (ED)
patients with infection [1] and in heterogeneous populations of patients with sepsis [26]. Current guidelines
from the Surviving Sepsis Campaign (SSC) [7, 8] advocate lactate measurement in patients with infection and
971
Interventions
In the 3 months (JanuaryMarch 2004) immediately preceding the study period we conducted a multidisciplinary
hospital-wide education program [13] that was based on
the recently released SSC guidelines [14]. We instructed
clinicians to measure lactate in patients with infection
and suspected sepsis to help identify high risk of death,
and we tested clinicians on this concept [13]. Next we
implemented a hospital-wide clinical practice algorithm
for sepsis resuscitation (adopted from Rivers et al. [9])
in which an initial step of the algorithm was to measure
lactate in patients with infection and suspected sepsis
in accordance with the SSC guidelines. Details of the
ED-specific elements of this implementation process are
published elsewhere [15].
This post-hoc analysis of a prospectively compiled registry was carried out in an urban academic medical center
(Cooper University Hospital, Camden, NJ, USA). It was
approved by the institutional review board, which waived
the need for informed consent. The subjects were 1,177
patients aged at least 18 years with primary or secondary
Gender
Male
Female
Age (years)
49
5065
6675
75
Locationa
ED
ICU
Non-ICU ward
Primary source of infection
Lung
Urinary tract
Intra-abdominal
Bloodstream
Skin/soft tissue
Other
a At
591
586
50
50
271
331
231
344
23
28
20
29
707
261
209
60
22
18
403
285
222
83
81
103
34
24
19
7
7
9
Data source
We utilized a prospectively compiled lactate registry for
this study. Our performance improvement committee for
critical care commissioned the creation of the lactate registry in March 2004 as a tool to aid in the tracking of patients for assessment of clinician adherence to applicable
SSC guidelines. The registry was created by a two-step
automated process: (a) whenever lactate was measured in
our central laboratory, a lactate report was automatically
generated and electronically transmitted to a data manager
who entered the new case into a dedicated database; and
(b) the registry database was linked to our institutions central administrative database allowing automated importing
of data fields for demographics, diagnoses, and outcomes.
972
Measurements
Secondary analyses
Data analysis
We stratified the lactate values into three a priori defined groups (low, 0.02.0; intermediate, 2.13.9; high,
4.0 mmol/l or above) consistent with the cutoff values
previously utilized by other authors [2, 5]. We tested
for linear associations with mortality risk by one-way
analysis of variance with Bonferronis correction for
multiple comparisons between groups. The proportionality of differences in survival was tested by the log
rank statistic using the KaplanMeier method accounting
for censoring. Using a lactate value of 4.0 mmol/l or
higher as a cutoff for a positive test, we determined the
sensitivity/specificity, receiver operating characteristic,
odds ratios, and likelihood ratios for death.
Results
Figure 1 displays the distribution of patients by initial
lactate value and the mortality in each lactate group.
Results of the one-way analysis of variance are presented
in the Electronic Supplementary Material (S. T1, S. T2).
The number of deaths within 3 days and in-hospital deaths
increased significantly and linearly with increasing range
of initial lactate values. The overall statistical significance
of linear differences between lactate groups (log-rank
statistic) was p < 0.0001 with one degree of freedom.
Bayesian analysis
Bayesian statistics combine pretest probability estimates
with new data (in this case a lactate value 4 mmol/l)
to yield new updated posttest probabilities. First, we
generated a full range (0.010.99) sample of hypothetical
clinician pretest probability estimates for death on the
grounds that this would permit demonstration of the
impact of a lactate level higher than 4 mmol/l on the
full range of possible pretest probability estimates that
a clinician could derive from all other clinical parameters
that are available. Second, we converted these pretest
probabilities into pretest odds and multiplied them by
the likelihood ratios for death from our data. Third, we
converted these posttest odds back to probabilities to
yield the full range of posttest probabilities. Lastly, we
graphically represented the relationships between pre- and
posttest probabilities through the full range of possible
experience.
973
974
Table 3 Use of an initial lactate value of 4.0 mmol/l or higher to predict mortality: subgroup analysis stratified by hospital location (brackets
95% confidence interval, ED emergency department, ICU intensive care unit)
Patients alive
Patients dead
Dead 3 days
Dead > 3 days
Lactate values
4.0 mmol/l
< 4.0 mmol/l
Acute-phase death ( 3 days)
Sensitivity (%)
Specificity (%)
Odds ratio
Likelihood ratio positive
Likelihood ratio negative
In-hospital death
Sensitivity (%)
Specificity (%)
Odds ratio
Likelihood ratio positive
Likelihood ratio negative
ED
(n = 707, 60%)
ICU
(n = 261, 22%)
Non-ICU ward
(n = 209, 18%)
621 (88%)
86 (12%)
21 (24%)
65 (76%)
151 (58%)
110 (42%)
33 (30%)
77 (70%)
180 (86%)
29 (14%)
9 (31%)
20 (69%)
70 (10%)
637 (90%)
38 (15%)
223 (85%)
4 (2%)
205 (98%)
24 [1144]
91 [9091]
3.0 [1.18.3]
2.6 [1.15.1]
0.84 [0.610.99]
46 [3259]
90 [8892]
7.4 [3.316.5]
4.5 [2.67.3]
0.61 [0.440.78]
22 [737]
99 [9899]
28.3 [4.3188.0]
22.2 [4.1118.0]
0.79 [0.630.95]
22 [1530]
92 [9193]
3.2 [1.85.8]
2.8 [1.74.3]
0.85 [0.750.93]
20 [1525]
89 [8693]
2.1 [1.14.2]
1.9 [1.13.4]
0.90 [0.810.99]
6 [212]
99 [9899]
6.6 [1.139.1]
6.2 [1.134.6]
0.94 [0.890.99]
graphically represents how lactate measurement can affect any possible value for pretest probability for acutephase death and in-hospital death as determined by a clinicians bedside assessment (derived from clinical parameters available prior to lactate measurement).
Secondary analyses
In the subset of patients who developed severe sepsis,
the median time from meeting severe sepsis criteria to
lactate measurement was 75 min (range 01,220). Table 3
presents the subgroup analysis stratifying patients by
hospital location. For the majority (60%) of patients in
the study who had initial lactate measured in the ED,
a lactate 4.0 mmol/l was associated with similar odds
ratios for both acute-phase and in-hospital death (3.0
and 3.2 respectively). In the ICU population, an initial
lactate 4.0 mmol/l was most useful in the prediction of
acute-phase death (odds ratio 7.4). The highest odds ratios
for death in the sample were observed in the minority
(18%) of patients from a non-ICU ward, although lactate
values above 4 mmol/l were relatively uncommon in this
subgroup.
Discussion
In patients with infection [1] and in heterogeneous populations of patients with sepsis [25] elevation in serum
lactate concentration has been associated with increased
risk of death. Guidelines from the SSC endorsed by 11
medical professional societies including the European
975
Limitations
We recognize limitations in the interpretation of our findings. First, the timing of measuring lactate in relation to the
time that a clinician first identified the presence of an acute
infection was not available for all subjects in this lactate
registry and could only be inferred from a sample of high
severity patients who also developed severe sepsis criteria. This represents an important limitation that could have
potentially introduced some heterogeneity into the sample
with respect to acuity of the infectious process at the time
lactate was measured. However, this concern is attenuated
by the following: (a) the median time from meeting severe sepsis criteria to lactate measurement in the secondary
analysis was only 75 min, and (b) our multidisciplinary education program conducted immediately prior to the study
period instructed all clinicians to obtain a lactate measurement promptly as soon as the possibility of sepsis was first
considered [14].
Second, the fact that this study only included patients
in whom lactate was measured by a clinician represents an
inherent bias in the sample. Although we recognize that
the decision to measure lactate was necessarily dependent
upon a variable level of proficiency with sepsis identification among individual clinicians, we point out that varying clinician proficiencies are an inherent factor in transfer of all guidelines to the bedside. Therefore we would
suggest that this real world experience [23] is the most
informative type of data with regard to implementation of
the SSC lactate guideline. We also recognize that using lactate data that was the product of a clinical decision rather
than measured at a fixed time point in the hospital course
(e.g., hospital admission) may be responsible in part for
a contrast between our results and the results of previous studies that found lactate values upon initial presentation were nondiscriminatory between sepsis survivors and
nonsurvivors [24, 25]. Rather than providing support for
the utility of lactate measurement at a fixed time point in
the hospital course, our data supports the utility of lactate
measurement at the time of initial clinical suspicion of sepsis.
Fourth, this registry did not capture comprehensive
clinical information. Specifically, we do not have data on
the presence of systemic inflammatory response syndrome
criteria at the time of lactate measurement. Therefore
it is unclear how many subjects would have met the
classical consensus definition of sepsis (i.e., infection
plus two or more criteria for the systemic inflammatory
response syndrome (SIRS)) [26]; rather, we utilized the
clinicians action of obtaining a lactate measurement
as a surrogate for clinical suspicion of sepsis for these
patients with a confirmed infection. The registry also
did not capture data on comorbidities, advance directive
status, hemodynamics, organ dysfunction, or therapies
administered, and thus a multivariate analysis could not
be performed. Therefore it is possible that patients with
high lactate values would have also been identified as
having high risk of death by other clinical indicators
independently of lactate measurement. That is why we
performed a Bayesian analysis (Fig. 3) to show how the
utilization of lactate measurement may impact any pretest
probability estimate that can be derived from all other
conventional clinical parameters that are available.
Lastly, we did not capture clinicians estimates for
probability of death prior to obtaining the lactate measurements. This would be a necessary element of a future study
976
specifically designed to confirm the accuracy and practical ment of mortality risk. Specifically, an initial lactate level
utility of the Bayesian approach to mortality prediction of 4 mmol/l or higher can have a substantial impact on
pretest probability of acute-phase ( 3 days) death.
and its capacity to impact clinical practice decisions.
Acknowledgements. The authors thank Tara Haag for her assistance in the preparation of this manuscript. Dr. Trzeciak is supported
Conclusions
by a Scientist Development Grant from the American Heart AssociThere was no outside source of funding for the project costs
When implemented in a multidisciplinary hospital-wide ation.
associated with this study. None of the authors have financial confashion, routine measurement of lactate in patients with flicts of interest to disclose. The authors had full control of the data
infection and possible sepsis can impact clinical assess- and the decision to publish.
References
1. Shapiro NI, Howell MD, Talmor D,
Nathanson LA, Lisbon A, Wolfe RE,
Weiss JW (2005) Serum lactate as
a predictor of mortality in emergency
department patients with infection. Ann
Emerg Med 2 45:524528
2. Aduen J, Bernstein WK, Khastgir T,
Miller J, Kerzner R, Bhatiani A,
Lustgarten J, Bassin AS, Davison L,
Chernow B (1994) The use and clinical
importance of a substrate-specific
electrode for rapid determination of
blood lactate concentrations. JAMA
272:16781685
3. Bakker J (2001) Lactate: may I have
your votes please? Intensive Care Med
27:611
4. Bakker J, Coffernils M, Leon M, Gris P,
Vincent JL (1991) Blood lactate levels
are superior to oxygen-derived variables
in predicting outcome in human septic
shock. Chest 99:956962
5. Bakker J, Gris P, Coffernils M, Kahn RJ,
Vincent JL (1996) Serial blood lactate
levels can predict the development of
multiple organ failure following septic
shock. Am J Surg 171:221226
6. Varpula M, Tallgren M, Saukkonen K,
Voipio-Pulkki LM, Pettila V (2005)
Hemodynamic variables related to
outcome in septic shock. Intensive Care
Med 31:10661071
7. Dellinger RP, Carlet JM, Masur H,
Gerlach H, Calandra T, Cohen J,
Gea-Banacloche J, Keh D, Marshall JC,
Parker MM, Ramsay G, Zimmermann JL, Vincent JL, Levy MM
(2004) Surviving Sepsis Campaign
guidelines for management of severe
sepsis and septic shock. Crit Care Med
32:858873
977