Introduction
Although the incidence, associated morbidity, attributable mortality, and overall costs related to sepsis vary widely, it is generally agreed that sepsis is ubiquitous and adversely affects patients [
1‐
4]. Early identification of sepsis and aggressive restoration of peripheral perfusion are the cornerstones of management. To improve outcomes, sepsis bundles focusing on early resuscitation and early antibiotics have been implemented. Serial measures of lactate—a byproduct of anaerobic metabolism used as a surrogate marker for suboptimal perfusion—are increasingly advocated as a target of resuscitation in sepsis bundles. The 2016 Surviving Sepsis Campaign Guidelines endorse measuring lactate levels within the first hour of resuscitation and every two to four hours thereafter if the initial level is > 2.0 mmol/L. [
5]
Increased reliance on lactate levels to guide clinical decision-making in sepsis is driven by studies associating elevated lactate with increased mortality in sepsis. One retrospective study found that lactate levels greater than 2.5 mmol/L correlated with a 28-day mortality of 16.9% [
6]. The prognostic value of lactate was further emphasized when lactate-directed resuscitation led to lower Sequential Organ Failure Assessment (SOFA) scores, earlier cessation of inotropes, earlier weaning from mechanical ventilation, earlier intensive care unit (ICU) discharge, and a 9.6% absolute reduction in hospital mortality when compared to usual care. [
7] Rising levels of lactate are also associated with increasing mortality, regardless of the presence/absence of shock. [
8] Taken collectively, these data suggest that elevated lactate is a deleterious proxy for tissue hypoperfusion and anaerobic metabolism in sepsis. However, there are other etiologies for hyperlactatemia including reduced clearance (i.e., liver dysfunction in the setting of hypoperfusion) and medication administration (i.e., large volumes of lactated Ringer’s solution, excessive albuterol) [
9‐
12].
An alternative viewpoint is that some of the hyperlactatemia produced during sepsis may be an adaptive response. Sepsis-related increases in metabolic rate and catecholamine levels increase glycolysis, glycogenolysis, gluconeogenesis, and reduce insulin release. This cascade ultimately results in increased pyruvate production with shunting to lactate production which can be used as an alternate energy source. The net effect is increased lactate production in the absence of tissue hypoxia—a form of Type B2 lactic acidosis [
13,
14]. This line of thinking, espoused in several popular online blogs and podcasts, suggests that because not all of the etiologies for increased lactate seen are harmful, clinicians should parse out adaptive lactate signaling an intact stress response from lactate caused by hypoperfusion and cellular ischemia [
15‐
17]. One clinical trial supports this hypothesis as administration of esmolol—pharmacologic blunting of the catecholamine response—reduced lactate concentrations in patients with septic shock [
18]. Another study showed that increasing serum lactate in the earliest phases of sepsis, presumably due to sympathetic activation, was associated with reduced mortality [
19].
Given these diametrically opposed positions regarding the deleterious effects of elevated serum lactate, we conducted this retrospective study to explore whether elevations of lactate in septic patients were driven primarily by abnormal blood pressure or by increased catecholamine activity. Because catecholamine levels are not measured directly during routine clinical care, we used patients’ presenting heart rates as a coarse proxy for their degree of inherent sympathetic activity.
Discussion
Lactate kinetics are multifaceted and often over-simplified in critical illness. An ongoing challenge to the primacy of lactate in sepsis has been the assertion that some degree of hyperlactatemia in septic patients results from catecholamine stimulation of ß2 receptors and that the resulting lactate (so-called Type B lactic acidosis) confounds the predictive utility of lactate in sepsis. Accordingly, our study attempted to explore this issue using data from a large cohort of sepsis patients and a relatively simplistic design. Our intent was to assess whether the effects of elevated lactate level in septic patients were affected by varying levels of systemic catecholamine activation (evidenced by higher presenting HR). Although our results support the hypothesis that patients with higher levels of sympathetic activity have higher levels of lactate, mortality was directly linked to elevations in lactate and was not affected by the degree of sympathetic activity.
Using accepted cut-points from the SIRS criteria for MAP and HR, we categorized septic patients as normotensive (MAP ≥ 65 mm Hg) or hypotensive (MAP < 65 mm Hg) and as having low sympathetic drive (HR < 90 bpm) or high sympathetic drive (HR ≥ 90 bpm). This allowed us to put septic patients into four groups: (1) hypotensive with low sympathetic drive; (2) hypotensive with high sympathetic drive; (3) normotensive with low sympathetic drive; and (4) normotensive with high sympathetic drive. Our hypothesis was that if both hypotension and sympathetic independently drive increased lactate levels, patients with both hypotension and high sympathetic drive would have the highest lactate levels, patients with hypotension or high sympathetic drive alone would have intermediate lactate levels, and patients with neither hypotension nor high sympathetic drive would have the lowest lactate levels. Our data precisely showed this relationship. Further, when holding MAP constant, we found that patients presenting with higher HR consistently had higher lactate levels compared to patients with lower HR (Fig.
2). These findings suggest that intrinsic sympathetic activity drive increases lactate levels. This relationship between HR and lactate was not affected by age, temperature, hepatic dysfunction, heart failure, and/or diabetes: when these factors were considered as covariates in statistical models, all inferences and the observed magnitude of the effects seen did not change.
Conventional wisdom regarding sepsis has focused on elevated lactate as a fundamental marker of tissue hypoxia due to distributive and/or hypovolemic shock. This has led to multiple disproven therapeutic recommendations including blood transfusions to maintain the hemoglobin > 10 mg/dL and using inotropes to drive the mixed venous oxygen saturation to > 70% among others. It is increasingly understood that in septic patients, excess lactate production through multiple avenues
and reduced lactate clearance likely come into play—and that lactate production does not always signal physiologic aberrancy due to dysoxia [
20‐
23].
However, any incremental lactate elevations due to increased catecholamines remained associated with an increased risk of death and were not less harmful as some have postulated. Consistent with numerous other studies, lactate was the strongest predictor of in-hospital death in this cohort of septic patients. Because we did not find a statistically significant interaction between HR and either lactate or MAP when looking at mortality or ICU discharge, our data do not suggest that increased HR (sympathetic drive) is protective. This is not entirely surprising and one could posit that our analysis is another example of the ‘squeezing the balloon’ analogy: any reduction in lactate production caused by an increase in MAP is replaced by a comparable amount of lactate due to the increased catecholamines needed to increase the HR to improve the MAP. Another alternate explanation suggests that lactate etiology (aerobic or anaerobic) is irrelevant and that lactate is a marker of a more severely dysregulated inflammatory response—a known maladaptive response [
24]. Regardless, each of these hypotheses is consistent with the observation that elevated lactate levels are ominous even in the absence of shock [
8]. Our study lends external validity to this association between lactate and mortality.
Our study supports the concept that hyperlactatemia in sepsis is more complex than simple anaerobic metabolism and that catecholamine stimulation can affect both MAP and lactate levels. One could propose that because lactate is not solely a marker of tissue hypoperfusion, intensivists should take pause in using lactate in isolation to guide resuscitation—an idea supported by a recent large, randomized, controlled trial which did not show benefit to using lactate kinetics to guide resuscitation. [
25] As suggested by others, this subject deserves significant further study given the magnitude of the problem posed by septic shock coupled with the trend whereby lactate levels have evolved into a quality metric. [
26] However, our finding that mortality was associated with increased lactate—and the lack of moderation/mediation of this relationship by HR—lactate appears to be an appropriate marker of severity of illness without consideration of the degree of sympathetic activity.
The biggest strengths of our study are the large sample size and the use of data from multiple hospitals, the majority of which were community-based. To maximize standardization despite using a retrospective dataset, all abstracted variables were objective measurements made at the time of admission. Because baseline clinical data and labs were used for analysis, we obviated the potential confounding effects of the most clinically important medications (i.e., beta-adrenergic agonists or blockers, sedatives) and other therapeutic interventions (i.e., mechanical ventilation) that could affect heart rate. We were also purposeful in including all patients with sepsis—not just those admitted to the ICU—as current sepsis alerts, treatment guidelines and/or quality metrics often include these individuals.
However, our study is not without significant limitations that must be considered when interpreting our results. The biggest barrier to drawing definitive conclusions from this study is our use of the HR as a surrogate for sympathetic activity as opposed to directly measuring catecholamine levels. Admittedly, HR is a very coarse measure of sympathetic activity and is potentially affected by multiple other factors including degree of fluid resuscitation, stroke volume, exogenous catecholamines, tachyarrhythmias, home medications (beta blockers, amiodarone, midodrine), fever, and comorbid illness (chronic hypertension). However, HR is uniformly available in real-time and is routinely used by clinicians at the bedside as a surrogate marker for sympathetic activation. Although we could effectively minimize the effects of some of these variables (fluid resuscitation, exogenous catecholamines) by collecting data at presentation prior to any therapy, other variables (tachyarrhythmia, comorbidity) could not be accounted for.
Another limitation was our reliance upon MAP as the sole marker of perfusion. We used MAP as our index of perfusion as it is a key component of the Surviving Sepsis Campaign and the ESICM guidelines. However, MAP does not consistently correlate with tissue perfusion at the microcirculatory level [
27]. Other markers of tissue perfusion such as urine output, capillary refill, and central venous oxygen saturation were not reliably available across facilities. In addition to the biases inherent with a retrospective study design (selection bias, measurement error, confounding), we did not adjust for severity of illness via acuity scores (e.g., APACHE, SOFA) for two reasons: (1) APACHE score subsumes MAP and HR which precluded evaluation of the primary independent variables in this study and (2) acuity scores were not calculated for most patients cared for in the four community-based hospitals.
A definitive resolution of this issue would ultimately require further investigation—ideally prospective data with direct measurement of serum catecholamines, an assessment of presenting severity of illness, delineation of concurrent medications, and controlling for medical comorbidities. However, such an extensive endeavor should only be undertaken if we suspect that the effort might meaningfully alter clinical practice. Although imperfect, our data do not suggest any effect of sympathetic activity on the association between lactate and mortality. Accordingly, efforts to delineate this issue further do not appear justified.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.