Editorial|February 17, 2019
How Best to Resuscitate Patients With Septic Shock?
Derek C. Angus
JAMA. 2019;321(7):647-648. doi:10.1001/jama.2019.0070
What is the best approach for resuscitation of a patient with septic shock? Despite considerable investigation over several decades, this important question still has no clear answer. There is agreement that resuscitation should proceed quickly, for the longer the delay, the greater the physiologic stress and end-organ injury. However, treatment options are principally intravenous fluids and vasoactive agents, which have the capacity to both help or harm the patient. Thus, these therapies must be titrated in response to markers of the adequacy of resuscitation. The 2016 Surviving Sepsis Guidelines advocate that the principal marker of the adequacy of resuscitation is serum lactate.1 Serum lactate level is elevated when the body relies on anaerobic metabolism. As such, an elevated serum lactate level is a reflection of tissue hypoperfusion and thus considered axiomatic of shock. However, serum lactate level can be elevated in conditions other than shock and yet may not always be elevated in some shock states. Furthermore, lactate clearance is often considered too slow to provide timely feedback to clinicians regarding the consequences of their treatment decisions. Moreover, not all patients with shock can be cared for in settings with access to rapid-turnaround serum lactate assays.
An alternative approach is to monitor compensatory physiological mechanisms. For example, in response to shock, the body will attempt to restrict blood flow to peripheral vascular beds. Thus, one of the simplest bedside clinical examination findings in shock is delayed capillary refill in the skin. The problem with delayed capillary refill, however, is that the examination can be difficult to standardize. Furthermore, there has been little empirical evaluation of resuscitation strategies based on this finding.
In this issue of JAMA, Hernández and colleagues2 report the results of a clinical trial in which 424 adults with early septic shock were randomized to 2 alternative 8-hour resuscitation strategies: one based on serial measurements of serum lactate levels and the other on peripheral perfusion, assessed by serial capillary refill time (CRT) examinations. The trial was conducted among patients in the intensive care unit with early septic shock, defined as patients with suspected or confirmed infection who were recognized within 4 hours to have elevated serum lactate level (>2 mmol/L) and require vasopressors to maintain blood pressure despite a bolus of 20 mL/kg of intravenous crystalloids.
The lactate level–based strategy was designed to reflect best current practice and involved titrating resuscitation to serum lactate level, measured every 2 hours, with adequacy of resuscitation defined as a greater than 20% decrease in serum lactate level across the 2-hour epoch until lactate level fell within normal range. The 2-hour frequency was based on lactate clearance kinetics. The peripheral perfusion–targeted strategy represented an experimental approach in which resuscitation was guided by a standardized assessment of CRT, applying a microscope slide to the finger pulp for 10 seconds of blanching and then counting via a chronometer the time to reperfusion. The CRT test was assessed every 30 minutes and resuscitation was judged adequate if CRT was less than 3 seconds. The 30-minute frequency was chosen because physiologic compensation occurs more quickly than lactate clearance.
In both groups, all other actions and assessments were common. Thus, the study protocol provided a standard approach to assess fluid responsiveness, provide boluses of crystalloids when patients were still fluid responsive, titrate vasopressors, and conduct trials of inodilators as needed. Furthermore, all sites were encouraged to follow all other aspects of critical care management for septic shock in keeping with the Surviving Sepsis Guidelines.
The primary outcome was 28-day all-cause mortality, and the study was powered to demonstrate a large improvement from 45% to 30% (a relative risk reduction of one-third). Secondary outcomes included a variety of measures of recovery of organ dysfunction. The cohort was enrolled briskly, with all patients enrolled in 1 year. Follow-up was excellent, with complete follow-up for the primary end point. The cohort was well-balanced across the 2 groups, with features typical of septic shock: pneumonia and intra-abdominal sepsis were the most common sources of infection, patients had moderately high severity-of-illness scores and evidence of multiorgan dysfunction, and the mean dose of norepinephrine at enrollment was high. Adherence with the 2 treatment protocols was generally high, with lack of adherence occurring in 29 patients (13.7%) in the CRT-guided group and 23 (10.8%) in the lactate level–guided group.
Processes of care during resuscitation were similar across both groups, although serum lactate and CRT perhaps improved faster in the CRT-guided group, as evidenced by better values at some but not all points over the first 72 hours, whereas the lactate level–guided group received more fluid at 8 hours (mean difference, −408 mL [95% CI, −705 to −110]; P = .01). At 28 days, the mortality rates were 43.4% in the lactate level–guided group and 34.9% in the CRT-guided group, with a hazard ratio of 0.75 (95% CI, 0.55 to 1.02) in favor of the CRT-guided group, which was not statistically significant (P = .06). There was less organ dysfunction at 72 hours in the CRT-guided group, as evidenced by a mean 72-hour Sequential Organ Failure Assessment (SOFA) score of 5.6 vs 6.6 (mean difference, −1.00 [95% CI, −1.97 to −0.02]; P = .045), but none of the other secondary outcomes showed a significant difference. A predefined test of heterogeneity of treatment effect by degree of organ dysfunction at presentation was significant, with a finding that the hazard ratio was significantly lower for the CRT-guided group in patients with a presenting SOFA score less than 10, but no difference in those with a SOFA score of 10 or greater.
Several issues are raised by this study. First, the primary result just missed the traditional threshold for statistical significance. Does this mean resuscitation based on serial CRT examinations could be superior to resuscitation based on lactate level? Perhaps. This study is a failed test of superiority. There are several potential explanations for this finding, the first of which is that the study was underpowered. The mortality rate in the control group was similar to the initial assumptions used in the power calculations, but many clinicians would likely consider a minimal clinically important difference to be considerably less than the 15% absolute reduction required by this investigation. In retrospect, it may have been helpful if the study size was planned to be larger. Nonetheless, proof of superiority would still require a confirmatory trial. An alternative explanation is that the CRT-guided therapy is not in fact superior. That raises the issue of whether this approach is equivalent to, or worse than, a lactate level–guided approach. The study was not designed to test for equivalence or noninferiority. Nonetheless, adverse events were rare in both study groups, and there was little obvious evidence of harm. Furthermore, the confidence intervals around the difference in mortality make it highly unlikely that the true effect of the CRT-guided approach would be to worsen outcome. In other words, it seems quite likely that the CRT-guided approach is, at a minimum, not inferior to the lactate level–guided approach.
There are also important issues regarding some study design choices. The study could not be blinded, and clinicians were capable of ordering serum lactate levels or conducting CRT tests in either treatment group as well as making a wide variety of diagnostic and therapeutic decisions while knowing treatment assignment. Thus, the steering committee had to rely on an approach that consisted of intense protocols to drive intervention-specific treatment decisions, together with in-depth training in their use and ongoing monitoring. This approach is typical for studies of complex treatment protocols but ultimately relies on strong commitment at each site and, even with good intentions, may be susceptible to bias.
Another choice that some clinicians may find arbitrary is to frame the monitoring of resuscitation as using either CRT or serum lactate level. Why not both? Both diagnostic tests are relatively straightforward and inexpensive, and perhaps their use could be complementary. Furthermore, clinicians may argue that the decisions they make for a given lactate level or CRT assessment vary depending on other patient features not captured in the resuscitation algorithms used by the authors. Such criticisms have some validity but also represent a potential rabbit hole: how can every possible set of diagnostic tests and treatment interventions be reduced to a manageable number of actions and protocols that are likely enforceable and testable? The investigators followed a number of steps to develop their algorithms, including relying on best evidence, conducting pilot studies, and engaging in extensive discussion with participating sites. This approach is state of the art but does not exclude the possibility that neither algorithm adequately represents the care decisions other clinicians might make.
The investigators provided considerable detail on the course of resuscitation, which suggested that resuscitation appeared similar in both study groups. There are 2 broad ways of thinking about these data. First, the fact that the distributions of interventions, such as mean fluid volumes or amount of vasopressor use, appeared similar would support an argument that the 2 approaches are in fact similar and should yield similar, not different, outcomes. Alternatively, the distributions of interventions may appear similar at the group level, but the decisions to use these interventions may have been made differently for different individuals. For example, one algorithm may have prompted more fluid for a given patient but less for another, relative to the other algorithm, but with no obvious difference at the group level. Such heterogeneity at the individual level is interesting to think about but difficult to study. The investigators could have attempted to catalog what decisions would have been prompted for each patient had that patient been assigned to the other group, but the authors have not conducted that analysis, and its interpretation would not be straightforward.
In conclusion, the report by Hernández et al represents a well-conducted study of 2 highly complex interventions for an important clinical problem. Had the study been larger, the findings may have been easier to interpret. It is nonetheless intriguing to consider that assessment of capillary refill can be conducted in a simple, standardized way and may have a role in the titration of fluids and vasopressors during the care of patients with septic shock. This may be particularly important in research-limited settings. Furthermore, the ability to conduct high-quality, safe assessments of such interventions in this setting is encouraging when considering the significant treatment uncertainties that persist. Hopefully, capillary refill, as well as other diagnostic and treatment modalities, will continue to be evaluated prospectively with the eventual goal of optimal resuscitation for all patients with septic shock.
