Lactate Clearance vs Central Venous Oxygen Saturation as Goals of Early Sepsis Therapy:  A Randomized Clinical Trial FREE

The rate of severe sepsis hospitalizations has doubled during the last decade with estimates indicating that at least 750 000 persons are affected annually in the United States.^1- 3 Approximately, 500 000 patients with severe sepsis in the United States annually are initially treated in emergency departments.⁴ The Surviving Sepsis Campaign international consensus guidelines recommend protocol-driven treatment that uses quantitative resuscitation for emergency department patients with severe sepsis and septic shock.⁵

Quantitative resuscitation refers to the use of an explicit protocol that targets predefined physiological or laboratory goals to be achieved within the first several hours. This concept was pioneered by Shoemaker et al⁶ to treat high-risk surgical patients. Results of a recent meta-analysis indicated a survival benefit associated with the use of an early and quantitative resuscitation strategy applied to heterogeneous populations of patients with sepsis.⁷

The optimal goals for quantitative resuscitation of sepsis remain uncertain. It is generally accepted that hemodynamic targets should include some measure of the adequacy of cardiac preload, such as central venous pressure, and perfusion pressure, such as mean arterial pressure.⁸ A more controversial issue is the method of determining tissue oxygen delivery. Citing a single-center study, the Surviving Sepsis Campaign guidelines recommend the use of central venous oxygen saturation (ScvO₂) or mixed venous oxygen saturation to assess the balance of tissue oxygen delivery and consumption⁹; however, since its publication in 2001 a substantial amount of controversy about this single-center study has been generated in the scientific community.^10- 12 Additionally, recently published practice surveys have indicated that the time, expertise, and specialized equipment required to measure ScvO₂ collectively pose a major barrier to the implementation of protocol-driven quantitative resuscitation programs.^13- 14 In contrast, lactate clearance, derived from calculating the change in lactate concentration from 2 blood specimens drawn at different times, potentially represents a more accessible method to assess tissue oxygen delivery.^15- 16

To address the potential utility of lactate clearance as a substitute for ScvO₂, we conducted a multicenter, randomized trial among patients presenting to the emergency department with severe sepsis and septic shock, with the primary hypothesis that early resuscitation targeting lactate clearance as the marker of adequacy of oxygen delivery was noninferior to the currently recommended ScvO₂ monitoring for the outcome of in-hospital mortality.

This study was a prospective randomized, parallel group, nonblinded, clinical trial designed to assess the noninferiority of lactate clearance vs ScvO₂ as the protocol goal that evaluated adequacy of oxygen delivery during early quantitative resuscitation of severe sepsis and septic shock. The trial took place from January 2007 to January 2009 in the emergency departments of 3 large urban medical centers in the United States. The research protocol was approved by the local institutional review boards and performed in accordance with Good Clinical Practice guidelines.

Patients with severe sepsis or septic shock were assessed for inclusion, which required that patients be older than 17 years with confirmed or presumed infection, have 2 or more systemic inflammatory response criteria,¹⁷ and have hypoperfusion evidenced by either a systolic blood pressure lower than 90 mm Hg after a minimum of 20 mL/kg rapid volume challenge or a blood lactate concentration of at least 36 mg/dL (4 mmol/L). The criteria for exclusion from the study were pregnancy, any primary diagnosis other than sepsis, suspected requirement for immediate surgery within 6 hours of diagnosis, an absolute contraindication to chest or neck central venous catheterization, cardiopulmonary resuscitation, transfer from another institution with a sepsis-specific resuscitative therapy underway, and advanced directive orders that would restrict the study procedure. Using a 24-hour day, 7-day-week method that was previously established for the routine clinical care of sepsis patients at each of the participating institutions, an alert was sent to inform clinical care resources when patients were identified as candidates for early aggressive resuscitation. This alert was also received by study staff who responded and screened the patients for study enrollment. Each enrolled patient or the patient's legally authorized next of kin provided written informed consent prior to collection of data. Patients or family members self-identified their race.

Patients were randomly assigned to 1 of 2 groups (eFigure). Each group received structured quantitative resuscitation while in the emergency department. The ScvO₂ group was resuscitated by sequentially providing therapy needed to meet thresholds of central venous pressure, followed by mean arterial pressure, and then ScvO₂. The lactate clearance group had similarly targeted thresholds in central venous pressure, followed by mean arterial pressure, and then lactate clearance instead of ScvO₂. Standard measures were used to ensure appropriate concealment of group assignment until after informed consent was obtained. The group assignment sequence was generated by an independent statistician using a parallel design, balanced randomization schedule (1:1 ratio of cases and controls), using the PROC PLAN function in SAS incorporating a sample size of 300, block size equal to 10, with a seed of 6 457 149 (SAS Institute Inc, Cary, North Carolina). After written informed consent was obtained, study staff opened an opaque sealed envelope containing the randomization assignment. Study staff then enforced the study protocol. By design, the clinical staff in the emergency departments could not be blinded to group assignment; however, the clinical staff (physicians and nurses) who assumed subsequent care of the patients in the intensive care units (ICUs) were unaware of group assignment. Prior to entry in the study, patients were cared for by emergency physicians, who provided basic care processes according to participating institutional standards.

Appropriate specimens were taken for culture, and antibiotics were administered as soon as practical. Blood pressure was monitored by either noninvasive automated cuff sphygmomanometer or arterial catheter according to the clinical team's preference. All patients received chest or neck central venous catheter capable of measuring continuous ScvO₂ (PreSep, Edwards Lifesciences, Irvine, California).

Patients randomized to the ScvO₂ group had their central venous catheters connected to a computerized spectrophotometer (Edwards Lifesciences) that displayed continuous ScvO₂ readings. The patients were then cared for according to the prespecified treatment plan. First, isotonic crystalloid was administered in boluses to achieve a central venous pressure of 8 mm Hg or higher. Second, the mean arterial pressure goal of 65 mm Hg or higher, if not achieved with fluid administration, was targeted by initiating and titrating vasopressors (dopamine or norepinephrine) to achieve this desired blood pressure goal. Finally, the ScvO₂ goal of 70% or higher was targeted after central venous and mean arterial pressure goals were met. If the ScvO₂ was lower than 70% and the hematocrit was lower than 30%, packed red blood cells were transfused to achieve a hematocrit of at least 30%. If the ScvO₂ remained lower than 70% after the hematocrit was 30% or higher, dobutamine was initiated and titrated in attempts to achieve an ScvO₂ of at least 70%.

Patients randomized to the lactate clearance group received an identical central venous catheter capable of measuring continuous ScvO₂. The primary intervention consisted of the act of not connecting the catheter to the computerized spectrophotometer thus preventing display of ScvO₂ at any time in the emergency department. These patients were cared for according to an identical prespecified treatment plan for central venous and mean arterial pressure targets as outlined for the ScvO₂ group. However, in the lactate clearance group, clinicians used lactate clearance instead of ScvO₂ as the last resuscitation goal in the protocol and targeted a lactate clearance of at least 10%.^15- 16 The lactate clearance was defined by the equation [(lactate_initial −lactate_delayed)/lactate_initial] × 100%, for which lactate initial was the measurement at the start of the resuscitation and lactate delayed was another measurement after a minimum of 2 hours after resuscitation was initiated. If the lactate clearance was not at least 10% at the first delayed measurement and the hematocrit was less than 30%, packed red blood cells were transfused to achieve a hematocrit of at least 30%. If the lactate clearance remained lower than 10% after the hematocrit was at least 30%, dobutamine was initiated and titrated in attempts to achieve a lactate clearance of at least 10%. When treatment was continued due to lactate clearance less than 10%, subsequent lactate measurements were performed at a minimum of 1-hour intervals and repeat lactate clearance calculated. Lactate measurements were performed using venous whole blood samples using US Food and Drug Administration–approved devices performed either at the point of care or in the central hospital laboratory, according to participating institutional standards.¹⁸ The lactate clearance goal was met by a lactate clearance of at least 10% or if both the initial and delayed lactate concentrations were not elevated (≤18 mg/dL [2 mmol/L]).

Study patients were treated in the emergency department during the entire study treatment period, from randomization to either of the 2 study termination criteria: all treatment goals were achieved or 6 hours had elapsed. Patients were then transferred to an ICU where the critical care physicians, unencumbered by the study protocol in any way, assumed the care of all patients. The study investigators did not provide care for the patients or influence their care in the ICU.

As a safety measure, clinical physicians could elect study group crossover. For patients assigned to the ScvO₂ group, clinicians could order a second lactate concentration to calculate the lactate clearance. In the lactate clearance group, the clinician could request to connect the central venous catheter to monitor ScvO₂. To execute 1 of these options, the clinician was required to indicate clinical deterioration based on 1 of the following criteria: (1) falling systolic blood pressure or inadequate urine output (<0.5 mL/kg per hour); (2) worsening ventilatory status based on either clinical (respiratory rate, oxygen saturation, or oxygen requirement), arterial blood gas, or mechanical ventilator parameters; or (3) worsened mental status. Patients for whom the clinicians enacted the study group crossover methods were assessed using an intent-to-treat analysis of the original group assignment.

During the study resuscitation treatment period, the patient's physiological parameters were measured routinely. All data needed to calculate the Simplified Acute Physiology Score (SAPS) II,¹⁹ Sequential Organ Failure Assessment (SOFA) score,²⁰ and the Mortality in Emergency Department Sepsis²¹ were collected. After ICU admission, patients were assessed daily for 72 hours and detailed data were collected. Patients were followed up until hospital discharge or death.

The primary end point was absolute in-hospital mortality rate. Secondary end points were ICU length of stay, hospital length of stay, ventilator-free days, and new onset multiple organ failure. Other end points assessed were the number of resuscitative goals achieved, administered treatments, and predefined protocol-related serious adverse events.

The sample size calculation and primary analytical plan centered on the hypothesis of noninferiority of lactate clearance vs ScvO₂. For the sample size estimate, previously published data suggested a 25% primary outcome event rate for the ScvO₂ group; we predicted no difference between the groups for the primary outcome event rate, and we set the noninferiority margin (Δ) at −10%.^22- 24 The −10% margin was chosen because it represented a two-thirds proportion of the active comparator's (ScvO₂ group) established superiority in a similar clinical trial⁹ and large databases indicate that the mortality rate for severe sepsis can be expected to vary up to 10% between both state-of-the art medical facilities and regions of the United States.²⁵ Using a 1-sided test of noninferiority, assuming a control group mortality rate of 25% and α = .05, a sample size of 150 per group gave 71% power to determine the intervention did not increase mortality by more than 10%.

When appropriate, categorical data were compared using a χ² test or Fisher exact test and continuous variables were compared with a Mann-Whitney U test or unpaired t test, 2-sided with P < .05 considered significant (StatsDirect v 2.7.7, Cheshire, England). Two preplanned blinded interim safety analyses were performed after one-third and two-thirds of the participants were enrolled, and these results were reviewed by an independent safety monitor who had the authority to terminate the study for safety concerns. No sample size adjustments were planned during the interim analyses. We planned both intent-to-treat and per-protocol analyses after study completion.

Of 452 patients who underwent screening for eligibility, 300 underwent randomization (Figure). Of these 300 patients, the same pattern of protocol noncompletion was observed in each group: 2 because a chest or neck central venous catheter could not be placed and 1 because family members had decided to withdraw support before the research procedure was completed. One patient in the lactate clearance group had the crossover enacted by a physician after all study goals had been met. The ScvO₂ value in this patient was normal, so no treatment changes were made. Because the study treatment period had ended, this patient was analyzed in the lactate clearance group for both the intent-to-treat and per-protocol analyses. None were lost to follow-up or voluntarily withdrew from the study, leaving 150 patients in each group for the intent-to-treat analysis and 147 in each group for the per-protocol analysis.

The baseline characteristics of the 2 groups are shown in Table 1 and in Table 2. There were no significant differences between the groups in demographics, comorbid conditions, severity of illness scores, or suspected site of infection. The lungs were the most common source of infection, and 38% of patients had a blood specimen that yielded growth of bacteria, whereas 84% had at least 1 culture specimen that was positive. The median time from emergency department triage to eligibility was 111 minutes (interquartile range [IQR], 56-192 minutes) in the lactate clearance group and 105 minutes (IQR, 60-175 minutes) in the ScvO₂ group (P = .67); the median time from eligibility to study entry was 14 minutes (IQR, 1-48 minutes) in the lactate clearance group and 13 minutes (IQR, 1-55 minutes) in the ScvO₂ group (P = .72). Prior to enrollment in the study, the mean (SD) amount of intravenous fluid administered was 2.3 L (1.4 L) in the lactate clearance group and 2.4 L (1.4 L) in the ScvO₂ group (P = .37).

Table 3 shows the physiological and severity of illness variables during the first 72 hours of hospitalization. During the first 24 hours, both groups of patients tended to show a trend toward slightly worsening severity of illness in the form of lower systolic blood pressures and higher SOFA scores. The mean initial lactate concentrations were 35.1 mg/dL (3.9 mmol/L) in the lactate clearance group and 37.8 mg/dL (4.2 mmol/L) in the ScvO₂ group (P = .39). The mean (SD) lactate concentration measured at 2 hours in the lactate clearance group was 23.4 (23.3) mg/dL (2.6 [2.59] mmol/L) and the median lactate clearance at 2 hours was 40% (IQR, 18%-64%). After the initial 24 hours, survivors in both groups manifested improvements in their physiological and severity of illness scores. We observed that lactate measurements did not worsen over the initial 72 hours (Table 3), suggesting that the initial lactate levels were often the most abnormal.

There were no differences in the administered treatments through the initial 72 hours of hospitalization as shown in Table 4. During the emergency department–based 6-hour resuscitation period, patients received approximately 4.5 L of crystalloid, 221 patients (74%; 95% confidence interval [CI], 68%-79%) required vasopressors for hypotension, and 79 patients (26%; 95% CI, 21%-32%) required mechanical ventilation. Notably, only 29 patients (10%; 95% CI, 7%-14%) required either dobutamine infusion or packed red blood cell transfusion during the initial 6 hours of treatment. Activated protein C administration was only administered in 5 patients (2%; 95% CI, 1%-4%).

Among the 294 (147 per group) patients included in the per-protocol analysis, the central venous pressure goal was achieved in 133 patients (91%; 95% CI, 85%-95%) in the lactate clearance group and 133 (91%; 95% CI, 85%-95%) in the ScvO₂ group (P = .99); the mean arterial pressure goal was achieved in 142 patients (97%; 95% CI, 92%-99%) in the lactate clearance group and 142 (97%; 95% CI, 92%-99%) in the ScvO₂ group (P = .99); and the lactate clearance goal was met in 139 patients (95%; 95% CI, 90%-98%) in the lactate clearance group and the ScvO₂ goal was met in 136 (93%; 95% CI, 87%-96%) patients in the ScvO₂ group (P = .67). The median time from patient triage in the emergency department to first antibiotic administration was 115 (IQR, 62-180) minutes in the lactate clearance group and 115 (IQR, 66-170) minutes in the ScvO₂ group (P = .98).

The primary and secondary study outcome analysis is outlined in Table 5. In the intent-to-treat analysis the in-hospital mortality rate was 17% (25 of 150 [95% CI, 11%-24%]) in the lactate clearance group compared with 23% (34 of 150 [95% CI, 17%-30%]) in the ScvO₂ group. The difference in these mortality rates was 6% (95% CI, −3% to 15%). The lower limit of this CI is well above the −10% predefined noninferiority threshold, confirming the primary hypothesis of noninferiority between the lactate clearance and ScvO₂ groups for in-hospital mortality. These results did not change substantially in the per-protocol analysis.

There were no differences in the observed rates of predefined protocol-related serious adverse events between the lactate clearance (9 of 150 [6%; 95% CI, 3%-11%]) and ScvO₂ (11 of 150 [7%; 95% CI, 4%-13%]) groups (P = .81).

The results of this large multicenter randomized controlled trial of 2 resuscitation protocols for early sepsis resuscitation indicate that a protocol targeting lactate clearance of at least 10% as evidence of adequate tissue oxygen delivery produces a similar short-term survival rate as a protocol using ScvO₂ monitoring. Patients in the group resuscitated to a lactate clearance of 10% or higher had 6% lower in-hospital mortality than those resuscitated to an ScvO₂ of at least 70% (95% CI for this difference, −3% to 15%) exceeding the −10% predefined noninferiority threshold. These data support the substitution of lactate measurements in peripheral venous blood as a safe and efficacious alternative to a computerized spectrophotometric catheter in the resuscitation of sepsis. To our knowledge, this is the largest randomized trial of emergency department–based early quantitative resuscitation for sepsis conducted to date and the first such trial to investigate the relative value of different goals of early, emergency department–resuscitation strategies.

The physiological basis for lactate clearance presumes that circulatory shock causes inadequate oxygen delivery, resulting in mitochondrial hypoxia. Under hypoxic conditions, mitochondrial oxidative phosphorylation fails, and energy metabolism becomes dependent on anaerobic glycolysis.²⁶ Anaerobic glycolysis sharply increases the production of cellular lactate, which diffuses into the blood during prolonged cell hypoxia. In patients with a clinical picture of severe infection, the blood lactate concentration varies in proportion to the ongoing deficit in tissue oxygenation, and the ability of the patient to reduce the blood lactate concentration indicates restoration of oxygen delivery with resuscitation.²⁷ Previous work has found that a lactate clearance of 10% or more predicts survival from septic shock, providing the rationale for this goal.^15- 16 In addition, we constructed the protocol to consider 2 normal lactate levels (≤18 mg/dL [2 mmol/L]) at least 2 hours apart as evidence of ongoing adequate tissue oxygenation. The rationale for including this criterion was that clinically it would make no sense to attempt to clear a value that is already normal and that 2 normal values provide a reasonable clinical signal that effective resuscitation has prevented worsening of tissue oxygenation and anaerobic metabolism.

We have previously documented that many clinicians perceive a significant degree of technical difficulty associated with the use of computerized spectrophotometric catheters to monitor ScvO₂.²⁸ These devices require equipment and expertise that are not available in many tertiary care emergency departments.¹³ Use of ScvO₂ monitoring catheters requires preplanned training and real-time calibration and troubleshooting that can divert attention from the patient.²⁸ We thus submit that the need exists for a simpler and more generalizable method to monitor the adequacy of tissue oxygen delivery as a research imperative in the treatment of patients with severe infection. Our results address this unmet need by providing data that justify the use of lactate clearance instead of continuous ScvO₂ monitoring.

The sample size assumed a mortality rate of 25% in the active comparator (ScvO₂) group. The observed mortality was 23% in the ScvO₂ group, indicating that we maintained approximately 80% power to detect a true difference. Although our observed overall mortality rate (20%) is lower than the 37% overall rate reported in the sentinel emergency department–based resuscitation study,⁹ it is nearly identical to mortality rates that both individual investigators from our group^22- 24 and others²⁹ have previously reported in effectiveness studies conducted in more heterogeneous populations. We believe that the mortality rate represents a contemporaneous and accurate estimate of the true mortality rate for patients with severe sepsis and septic shock treated with an early quantitative resuscitation protocol in the emergency department.

Several limitations of our study warrant discussion. By its design, the groups could not be blinded, allowing for possible treatment bias. Our protocol was designed with safeguards to minimize this potential effect. For example, every participant received identical central venous catheters so that group assignment would not be easily identifiable. Also, investigators involved with the study were not allowed to provide care for the participants or influence their care in the ICU. Second, we did not have a method to assess whether an indicated therapeutic action was performed in response to a parameter below the intended goal (eg, if central venous pressure was 4 mm Hg, we did not record whether a fluid bolus was given). Rather, we only assessed for compliance with individual treatment goals during the study treatment period. Third, this study was conducted at 3 institutions that had established emergency department–based quantitative resuscitation programs for sepsis prior to initiation of the study. Therefore, our results may not be generalizable to centers that do not routinely perform early quantitative resuscitation. Fourth, if other influences on care were ignored, it could be suggested that the potential difference in protocol actions directly attributable to using lactate clearance vs ScvO₂ was small, because only 10% of patients went on to receive dobutamine or packed red blood cell transfusion. Fifth, we did not have a mechanism to query ICU admission for potentially missed cases; thus, we may have missed patients who met criteria because a clinical alert was not activated. Finally, it is possible that knowledge of the study prompted clinical care providers to have a heightened awareness and provide differential treatment patterns (ie, a Hawthorne-like effect).

In conclusion, in this randomized trial, we found no difference in mortality for patients with severe sepsis and septic shock resuscitated with a protocol that used lactate clearance compared with a protocol that used ScvO₂ as the method of measuring total body oxygen metabolism.