Timing of vasopressor initiation and mortality in septic shock: a cohort study

Data from a retrospective review of adult patients (≥18 years old) diagnosed with septic shock was used to create the Cooperative Antimicrobial Therapy of Septic Shock Database (member listing in Additional file 1). Consecutive adult septic shock patients from 28 medical institutions in Canada, the United States and Saudi Arabia for periods between 1996 and 2008 were retrospectively identified using either internal ICU registries/databases and/or International Classification of Diseases (ICD-9 or ICD-10) coding strategies. Patients from surgical, medical and mixed ICUs were included. Each potential case was screened to determine eligibility to meet the criteria for septic shock as described by the 1991 Society of Critical Care Medicine/American College of Chest Physicians consensus statement on sepsis definition [18]. All included cases were required to have no other obvious cause of shock. Each institution contributed a minimum of 50 cases. A waived consent protocol was approved by the Health Ethics Board of the University of Manitoba and at each individual participating center (listing in Additional file 2). The Ethics Boards waived the need for informed consent because of the retrospective, risk-free nature of the study in combination with the use of de-identified data.

Data including the time to vasopressor administration after documentation of persistent or recurrent hypotension refractory to fluid administration were retrospectively collected from clinical records using a uniform data extraction template by several trained research nurses or research assistants with medical training (medical students, residents, fellows). All data extractors reviewed >100 charts.

Hypotension was defined as a mean blood pressure <65 mmHg, a systolic blood pressure <90 mmHg, or a decrease in systolic pressure of 40 mmHg from the patient’s baseline consistent with the Society of Critical Care Medicine/American College of Chest Physicians criteria for septic shock [18]. An episode of hypotension was considered to represent the initial onset of septic shock when hypotension persisted from the onset despite fluid (>2 l saline or equivalent) administration (persistent hypotension), or when hypotension was only transiently improved (hypotension resolution for <1 hour) with fluid resuscitation (recurrent hypotension). Hypotension that resolved following fluid resuscitation alone (crystalloid or colloid) without subsequent clinical deterioration was not considered to represent the initial onset of septic shock-related hypotension. Similarly, patients exclusively treated with an inotropic agent without a vasopressor during the first 24 hours were excluded from the database. Organ failure was determined according to previously described criteria [3, 19].

Statistical analysis was performed using SAS version 9.1 (Cary, NC USA). Descriptive statistics were used to characterize the patient population, including mean and standard deviation for continuous variables (or median and inter-quartile range for skewed distributions) and frequency and proportion for categorical variables. Empirical logit plots were used to explore the functional form of the association between vasopressor delay fraction (analyzed continuously and also as categorized at decile cutpoints) and survival to hospital discharge. The shortest time delay decile (≤6 minutes) was excluded from the analysis as this usually represents cases where hypotension existed for an unknown period before arrival in the emergency department. In this circumstance, the true time from hypotension onset to vasopressor initiation is indeterminate.

The unadjusted association between survival to hospital discharge and vasopressor delay was estimated using simple logistic regression. A similar analysis was done with respect to the occurrence of individual and total number of organ failures after the day of shock (incremental organ failures from day 2 to day 10). A wide variety of epidemiologic factors (age, sex), comorbidities (AIDS, hematologic malignancy (lymphoma/leukemia/multiple myeloma), metastatic cancer, heart disease, organ transplant, hypertension, respiratory disease, renal disease, diabetes, autoimmune conditions, thromboembolism, neurological diseases), severity of illness (Acute Physiology and Chronic Health Evaluation (APACHE) score) [20], laboratory values (admission lactic acid and bicarbonate levels, white cell count) and therapeutic elements (time to initial appropriate antimicrobial therapy) were first assessed with respect to hospital survival and organ failure using univariate analysis. Those that were significant at P < 0.05 were retained for inclusion in the model. Multivariable logistic regression was then used to estimate the adjusted association and to identify independent correlates of mortality and organ failure. Mortality and individual organ failure results are expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Total incremental organ failure after the admission day (day 2 to day 10) was analyzed using Poisson regression with results expressed as rate ratios. Because hospital length of stay (LOS) and ICU LOS are count variables, these secondary outcomes were analyzed using generalized linear regression with a negative binomial distribution and logarithmic link function, adjusted for the same covariates as in the primary outcome analysis. Data are expressed as mean ± standard deviation or median with interquartile range as appropriate.

The baseline characteristics of the patients in the entire cohort are presented in Table 1. The average age was 62 ± 1 years with male predominance (57.0%). The most common existing comorbidities were diabetes inclusive of oral hypoglycemic and insulin-requiring (26.6%), chronic renal failure inclusive of dialysis (23.6%), and hypertension (19.1%). Illness severity is presented in Table 2 with the average APACHE II score being 26.1 ± 8.2. Baseline (day 1) laboratory results also presented in Table 2 showed elevated levels of serum creatinine (219 ± 181 μmol/l), leukocyte count (16.3 ± 16.1 × 10⁶ cells/l), International Normalized Ratio (1.5 ± 1.4) and serum lactate (4.8 ± 4.4 mmol/l). The heart rate was elevated at 115 ± 29 beats/minute. Approximately 40% of cases were due to nosocomially acquired infection (Table 2). Culture negative and bacteremic/fungemic patients each accounted for about one-third of the cohort. The lungs, abdomen and urinary tract were the most common infection sites and Escherichia coli, Staphylococcus aureus and Streptococcus pneumoniae were the most frequently isolated pathogens (Table 2).

The median time to vasopressor initiation was 3 hours (25 to 75% range: 1 to 7.1 hours). The distribution of vasopressor use is presented in Table 3. The most commonly used vasopressor was norepinephrine in about two-thirds of patients, with dopamine being the second most common used in approximately one-half. Use of a given vasopressor was not exclusive of use of others. Dobutamine, an inotropic agent, was used for at least 30 minutes during the first 24 hours after pressor initiation in 12.2% of cases. However, inotropes were never initiated before pressors and an intrope alone was never used (per inclusion criteria). Steroids were used in 32% of patients.

The overall unadjusted mortality rate was 53%. Unadjusted mortality among deciles ranged from 47.6% to 63.0% (Figure 1).

The significant independent correlates of mortality from the multivariable analysis are presented in Table 4 in order of descending influence on mortality based on Wald χ² values. Among these correlates, the APACHE II score was most significant with an OR of 1.11 per point (95% CI = 1.10 to 1.12). Antimicrobial delay was the next most important variable, each hour of delay was associated with a 7% increase in mortality (OR = 1.07, 95% CI = 1.06 to 1.08) and age was associated with a 2.6% increase in mortality per year of life (OR = 1.03, 95% CI = 1.02 to 1.03). Among categorical variables, liver failure had the strongest association with mortality (OR = 3.46, 95% CI = 2.67 to 4.48). A history of hypertension was found to convey a protective effect (OR = 0.62, 95% CI = 0.52 to 0.73).

After adjusting for independent correlates of mortality (AIDS, hypertension, liver failure, neutropenia, malignancy, metastatic disease, APACHE II score and delay in appropriate antimicrobials), there was a weak association of delay of vasopressors with in-hospital mortality (adjusted OR = 1.02, 95% CI = 1.01 to 1.03, P < 0.001). To examine the impact of delays in vasopressor initiation further, deciles of delay were examined in the model. The results are shown in Figure 2. At increasing delays of approximately 0.50 to 1.15 hours, 1.16 to 2.00 hours, 2.01 to 2.90 hours, 2.91 to 4.00 hours, 4.01 to 5.75 hours, 5.76 to 8.45 hours, 8.46 to 14.10 hours and >14.10 hours (reference second decile, 7 to 30 minutes as per the analysis protocol), the adjusted OR of survival was significantly increased only for the final, latest decile (OR = 1.34, 95% CI = 1.03 to 1.76, P = 0.048).

Secondary outcomes were adjusted for the same independent predictors of mortality as the primary outcome. In both unadjusted and adjusted analyses, a strong trend or actual significance was found between the delay to pressor initiation and the occurrence of organ failures. Adjusted P values were as follows: renal, P = 0.0182; respiratory, P < 0.0001; hematologic, P = 0.0788; central nervous system, P = 0.0208; coagulation, P = 0.0089; metabolic, P < 0.0001. Notably, in each case, the last decile (>14.1 hours) accounted for the impact of pressor delay on the occurrence of organ failure. In addition, the total incremental organ failures after the day of presentation (that is, day 2 to day 10) was associated with pressor delay. Again, this relationship was driven by the last decile of delay (Figure 3).

For the survivors, while controlling for significant variables, delay in vasopressor initiation was not predictive of hospital LOS (P = 0.19) or ICU LOS (P = 0.17). In addition, there was no significant impact on duration of vasopressor/inotropic therapy (P = 0.313) and only a trend towards a longer duration of ventilator support (P = 0.055) among survivors.