Discordance Between Capillary Refill Time and Oxygen Extraction Ratio in Septic Shock

Background. In septic shock resuscitation, improvement of macrocirculatory targets such as blood pressure and capillary refill time (CRT) does not always reflect adequate oxygen utilization at the tissue level-a phenomenon described as loss of hemodynamic coherence. Although the ANDROMEDA-SHOCK trial highlighted CRT-targeted resuscitation, a subset of patients with a normalized CRT may still have impaired tissue oxygen balance. The oxygen extraction ratio (O₂ER) can deviate in two directions: a high O₂ER (>30%) suggests oxygen delivery that is insufficient for demand (e.g., low cardiac output, anemia), whereas a low O₂ER (<20%, with high central venous oxygen saturation) may reflect microcirculatory shunting and cytopathic hypoxia.

Primary aim. To determine the prevalence of the "discordant" phenotype-defined as a normalized CRT with an abnormal O₂ER at hour 6 of resuscitation-and to evaluate its association with 28-day mortality.

Secondary aims. (1) To examine the association between CRT/O₂ER-defined perfusion phenotypes and 28-day mortality, ICU and hospital length of stay, duration of mechanical ventilation, vasopressor-free days, and need for renal replacement therapy; (2) to validate the proposed O₂ER >30% threshold using ROC curve analysis and identify the optimal cut-off for mortality in this cohort; (3) to assess whether the discordant phenotype is an independent predictor of mortality after adjustment for age, APACHE II score, lactate, and noradrenaline dose using multivariable analysis; (4) to examine the relationship of discordance with complementary perfusion markers (lactate clearance, mottling score, venoarterial CO₂ difference).

Design and setting. Prospective, single-center, observational cohort study in an intensive care unit. Consecutive adult patients diagnosed with septic shock according to the Surviving Sepsis Campaign (SSC) 2026 guideline will be enrolled. Time zero (Hour 0) is defined as the time of vasopressor (noradrenaline) initiation.

Measurements. While patients continue standard care, serial clinical and laboratory measurements are recorded at hours 0, 6, 12, and 24, and organ function at hour 72. CRT is measured in a standardized manner on the ventral surface of the distal phalanx of the right index finger after 10 seconds of firm pressure. O₂ER is calculated as (SaO₂ - ScvO₂) / SaO₂. At hour 6, CRT and blood gas sampling are performed simultaneously (within 15 minutes). Inter-rater reliability of CRT is assessed in the first 20-30 patients by two independent, blinded observers (ICC for continuous values; Cohen's kappa for the ≤3 s normal/abnormal classification).

Phenotype grouping (Hour 6). Group 1 - Concordant normal (CRT ≤3 s + O₂ER 20-30%); Group 2a - Discordant, high O₂ER (CRT ≤3 s + O₂ER >30%); Group 2b - Discordant, low O₂ER (CRT ≤3 s + O₂ER <20%); Group 3 - Concordant abnormal (CRT >3 s + abnormal O₂ER). The primary group of interest is the combined discordant phenotype (Group 2a + Group 2b).

Outcomes. Primary outcome: 28-day mortality (counted from Hour 0). Secondary outcomes: ICU and hospital length of stay, duration of mechanical ventilation, time to vasopressor weaning, vasopressor-free days, hour-72 organ function (SOFA-2, ongoing vasopressor/ventilation, lactate), and need for renal replacement therapy.

Sample size and statistics. A target of approximately 100 consecutive patients was chosen so that an expected discordance prevalence of ~30% can be estimated with a 95% confidence interval half-width of ±9%. Descriptive statistics, normality testing (Shapiro-Wilk), and appropriate parametric/non-parametric group comparisons will be used. The prevalence of the discordant phenotype will be reported with 95% CI. ROC analysis (with AUC) will evaluate the O₂ER threshold for mortality. Multivariable binary logistic regression (parsimonious model: phenotype, APACHE II, lactate; events-per-variable ≈12) will identify independent predictors of mortality, reported as odds ratios with 95% CI. Survival across phenotype groups will be estimated by Kaplan-Meier analysis and compared with the log-rank test. Statistical significance is set at p < 0.05.