Editorial 

Capillary Refill Time in Sepsis—Searching for the Holy Grail

Flavia R. Machado, Matthew W. Semler

JAMA Published Online: October 29, 2025

doi: 10.1001/jama.2025.20518

Almost 25 years ago, a landmark trial of early goal-directed therapy (EGDT) by Rivers and colleagues reshaped the care of septic shock by introducing the concept that “early” management was crucial.¹ Although implementation has varied across high- and low-resource settings, the importance of early sepsis recognition and treatment is now nearly universally accepted. In contrast, the specific manner by which goal-directed therapy for septic shock should be provided remains highly controversial. Conceptually, goal-directed therapy is a resuscitation protocol involving a goal to be achieved, serial assessments of the patient’s physiologic status, and administration of therapies that the assessments indicate will help achieve the goal. But what are the right goals, assessments, and therapies to improve outcomes for patients with early sepsis?

The overarching goal of hemodynamic management for early septic shock is to rapidly correct the hypoxia that arises in the tissue beds of vital organs secondary to the hypoperfusion that is axiomatic of “shock.” However, tissue hypoxia is hard to measure directly. Thus, resuscitation protocols rely on indirect surrogates, such as central venous oxygen saturation (Scvo₂), blood lactate levels, venous-to-arterial carbon dioxide difference, or clinical markers of perfusion, such as skin temperature, mottling, or capillary refill time. The original EGDT protocol targeted a goal of Scvo₂ of at least 70%, used assessments of central venous pressure and mean arterial pressure, and administered the therapies of intravenous fluid, vasopressors, inotropes, and packed red blood cells.¹Central venous oxygen saturation theoretically reflects the overall balance of oxygen delivery and consumption but fails to detect uncoupling between the macrocirculation and the microcirculation. Multiple recent large, randomized trials found that the original EGDT protocol did not improve outcomes for patients with early sepsis.²Lactate levels have been proposed as an alternative goal in early sepsis³ but may be abnormal despite normal tissue oxygenation because of increased lactate production from aerobic glycolysis and decreased lactate metabolism and may change too sluggishly over time.⁴ Both these approaches also require resources that are not available in many settings where patients present with septic shock. The limitations of Scvo₂ and lactate led clinicians and researchers worldwide to search for a sensitive, reliable, available, and inexpensive perfusion marker to serve as a target goal in early sepsis.

Some experts believe that capillary refill time may be this “holy grail” for assessing early sepsis. For patients in shock, blood flow is diverted away from the skin to prioritize the perfusion of vital organs. The delivery of oxygen to the skin may be impaired not only by reduced perfusion pressure but also by altered microvascular function, endothelial activation, and microthrombosis.⁵ Thus, bedside clinical signs such as skin temperature, skin mottling, and capillary refill times are receiving increasing attention as markers of perfusion and microcirculatory function. Capillary refill time is assessed by applying pressure to the distal phalanx for several seconds and measuring the time after release required for color to return, which indicates the restoration of blood flow to the compressed area. Rapid improvement in capillary refill times after resuscitation implicates inadequate systemic blood flow as the cause, whereas persistence after resuscitation may indicate microvascular dysfunction or uncoupling of the macrovasculature and microvasculature.⁵ Given that the aim of administering fluid and vasopressors in sepsis is frequently to improve perfusion, a normal capillary refill time may identify patients for whom the risks of additional fluid and vasopressor administration may outweigh the benefits.

Over the past decade, use of capillary refill time to assess peripheral perfusion has gained increasing interest. Observational studies reported variable accuracy of capillary refill time in predicting mortality, with better results when high-quality measurements are performed.⁶ The ANDROMEDA-SHOCK randomized trial compared the goal of normalizing capillary refill time vs normalizing lactate levels among 424 patients with septic shock.⁷ Patients in the capillary refill time group received less intravenous fluid, experienced less organ dysfunction, and had a numerically lower incidence of death. In a post hoc bayesian analysis, the posterior probability that resuscitation targeting capillary refill time was superior to using lactate exceeded 90%.⁸ These provocative findings led to calls for a large, confirmatory trial.

A new study published in JAMA, the ANDROMEDA-SHOCK-2 trial, was a nonblinded, parallel-group, multicenter randomized trial comparing the use of a personalized hemodynamic resuscitation protocol targeting capillary refill time vs usual care.⁹ The trial was conducted in 86 intensive care units across 19 countries and enrolled adults within 4 hours of developing septic shock, as defined by a suspected or confirmed infection, a serum lactate level of 2.0 mmol/L or higher, and receipt of norepinephrine to maintain a mean arterial pressure of at least 65 mm Hg after receiving at least 1000 mL of crystalloids. At enrollment, patients had received a median of 1500 mL of intravenous fluid (approximately 22 mL/kg) and were receiving a median dose of norepinephrine of 0.22 μg/kg/min, with 1 in 5 patients receiving a second vasopressor.

In the intervention group, trained study personnel administered a 6-hour protocol to participants in which the goal was to achieve a capillary refill time of 3 seconds or less (assessed after each intervention or, if normal, hourly); the physiologic assessments when capillary refill time was abnormal could include measurements of pulse pressure (the difference between the systolic and diastolic blood pressures), fluid responsiveness, mean arterial pressure, and ventricular function on echocardiography; and the therapies administered could include intravenous fluid, vasopressors, or inotropes. Of the 720 patients in the intervention group, 36% started with a normal capillary refill time and did not receive additional assessments or therapies, 14% ended the 6 hours with a capillary refill time that was still prolonged despite any assessments and therapies, and the remaining patients had a capillary refill time that normalized with treatments (approximately 33% after fluid bolus administration, 6% after treatment of ventricular dysfunction or inotrope therapy, 5% after vasopressors to increase mean arterial pressure, and only 3% after vasopressors to increase diastolic blood pressure). Treating clinicians (some of whom were also trained study personnel) determined and delivered the care in the usual care group.

At 6 hours after enrollment, the intervention and usual care groups differed with respect to the mean volume of intravenous fluid received (595 mL intervention vs 847 mL usual care), the percentage of patients receiving norepinephrine (94.7% vs 91.4%), the percentage of patients receiving dobutamine (12.3% vs 5.3%), and the percentage of patients with a capillary refill time of 3 seconds or less (85.9% vs 61.7%), without clinically meaningful differences in central venous pressure (9.1 mm Hg vs 9.8 mm Hg), lactate level (3.2 mmol/L vs 3.5 mmol/L), or Scvo₂ (74.4% vs 72.4%).

The primary outcome was a hierarchical composite of death, duration of organ support (defined as the final cessation of mechanical ventilation, vasopressors, or kidney replacement therapy), and length of hospital stay, all assessed at 28 days. The primary outcome was compared between trial groups using a win ratio, stratified for patients’ baseline severity of illness. Among 1467 patients, there were 131 131 (48.9%) wins in the intervention group vs 112 787 (42.1%) in the usual care group, yielding a win ratio of 1.16 (95% CI, 1.02-1.33; P = .04). The incidence of death by 28 days was 27% in both groups. The mean time to hospital discharge did not differ significantly between groups (15.3 vs 16.2 days). Thus, the difference in the primary outcome was primarily driven by a difference in the median duration of organ support between the intervention group (3.0 days) and the usual care group (4.0 days). The number of days alive and free of each individual supportive therapy did not differ between groups.

The investigators, networks, and approach to global collaboration involved in ANDROMEDA-SHOCK-2 deserve commendation. The trial enrolled a severely ill patient population and delivered a time-sensitive intervention with limited missing data for clinical outcomes despite conduct in variable resource settings. The performance of capillary refill time was highly standardized. Some parts of the protocol (eg, measurement of capillary refill time and pulse pressure) are simple, inexpensive, and widely available, facilitating implementation in resource-limited settings. Although a less important outcome than death, a shorter duration of organ support may be meaningful to patients, families and clinicians. Importantly, the shorter duration of organ support could translate into savings and improve equity by redirecting resources to other pressing priorities.

The assessment of separation between groups in a trial evaluating a personalized intervention is challenging. The magnitude of the average differences between groups in fluid therapy (approximately 250 mL less fluid in the intervention group) and vasopressors (approximately 3 percentage points more vasopressor receipt in the intervention group) in this trial was less than the differences in other large trials that did not observe differences in death or duration of organ support.^10,11 In addition to the care provided to patients by intensive care unit clinicians in both trial groups, in the capillary refill time group additional dedicated, trained personnel were at the bedside during the first 6 hours facilitating the delivery of the protocolized care. It is impossible to know whether the difference in outcomes observed resulted from protocolized hemodynamic management, the additional resources and reassessment involved in delivering the intervention, or both. Also, the trial was unblinded, and cessation of organ support is subject to clinician judgment, raising the possibility that awareness of treatment allocation unintentionally influenced decisions.

While the initial steps of the protocol are widely available, other parts of the protocol are not (eg, echocardiography) or are not based on strong evidence (eg, administering vasopressors to raise diastolic blood pressure). The assessment of the protocol as a bundle prevents understanding which individual elements are helpful, harmful, or ineffective. Targeting capillary refill time might have led to a faster resolution of tissue hypoperfusion as a result of these multiple interventions. However, this may simply reflect optimized fluid management: withholding fluids in patients with normal capillary refill time who might otherwise have been treated for elevated serum lactate, and administering fluids to those with prolonged capillary refill time if fluid responsive. Together, these 2 scenarios accounted for 69% of enrolled patients. Some protocol components, such as administering vasopressors to increase diastolic blood pressure, were less convincing or even disappointing.

So, is capillary refill time the holy grail for guiding hemodynamic management in early sepsis? The critical care community may choose not to fully embrace the complex assessments and interventions of this new approach to EGDT tested in ANDROMEDA-SHOCK-2, but rather to integrate selected elements into a broader framework of optimized resuscitation in a manner that accounts for resource availability. Certainly, the results of the ANDROMEDA-SHOCK and ANDROMEDA-SHOCK-2 trials together elevate the use of capillary refill time for assessing perfusion in septic shock, alongside (in fact, above) the use of Scvo₂ or serum lactate in a way that will inform practices and guidelines worldwide. Going forward, many clinicians caring for critically ill patients might choose to target capillary refill time alone or to incorporate it as part of a multimodal approach, in addition to Scvo₂, lactate, or the venous-to-arterial CO₂ gap. In a multimodal approach, treatment decisions should not rely on a single variable but rather on the synthesis of multiple variables using clinical judgment. Further resuscitation should not be undertaken solely on the basis of elevated lactate. However, in a patient with normal capillary refill time but with altered consciousness, low Svco₂, and an increased CO₂ gap, a thoughtful bedside physician might reasonably opt for additional resuscitation. Future research on capillary refill time should address interobserver variability, the influence of skin pigmentation, temperature, and local vasomotor tone, as well as assessment by clinicians outside of structured research environments. One intangible advantage of capillary refill time is that performing it brings the clinician back to the bedside to reassess the patient. This act may be itself the holy grail of early sepsis management.