Despite advancements in understanding and treatment, septic shock remains a worldwide healthcare problem. With an increasing annual incidence in the developed world, mortality remains between 25 and 50% of those afflicted 1 2 3 . The pathophysiology of septic shock is complex and involves vasodilatation, relative and absolute hypovolemia, myocardial dysfunction, increased metabolic rate and altered regional and microvascular blood flow 4 5 6 7 8 9 10 11 . Septic shock appears to cause a loss of autoregulation, making the perfusion of many vital organs and tissues dependent on blood pressure 5 12 13 . Early and aggressive fluid resuscitation of sepsis has been suggested to have a critical role in optimization of organ perfusion, preservation of end organ function and improvement of survival 14 .

Hypotension despite adequate fluid resuscitation therapy is a defining criterion in the diagnosis of septic shock 15 . To maintain organ perfusion, current guidelines recommend maintaining a mean arterial pressure (MAP) of 65 mmHg with fluid therapy and vasopressors even when hypovolemia has not yet been resolved 15 . According to the Surviving Sepsis Campaign this recommendation is considered ‘strong’ although supporting evidence is considered ‘weak’ 15 .

Many studies have compared different vasopressor agents for the resuscitation of septic shock but very few have investigated the role that the timing of vasopressor initiation in relation to hypotension onset plays in outcome 16 17 .

There were a total of 8,670 patients that fit the diagnostic criteria for septic shock. Thirty patients did not have a time of vasopressor initiation available and were excluded. Another 2,126 patients were excluded due to inadequate data acquisition of other significant analytic variables, primarily time to appropriate antimicrobial therapy from documentation of hypotension. In total, 6,514 observations were included in this analysis.

Hypotension is a central feature in the pathophysiology of septic shock. The duration of hypotension before intervention in cardiogenic shock caused by massive myocardial infarction, obstructive shock due to pulmonary embolus and hypovolemic shock due to major trauma/hemorrhage is a key determinant of survival 21 22 23 24 25 . Outcome in these conditions is closely associated with earlier initiation of therapy 21 22 23 24 25 26 . Similarly, in septic shock, early initiation of fluid resuscitation and rapid administration of appropriate antimicrobials are critical determinants of outcome and central tenets of management 14 27 28 . Based on these factors, we hypothesized that longer duration of hypotension without hemodynamic support using vasopressor infusion may result in a higher mortality rate and an increased incidence of organ failure in septic shock patients.

Our study demonstrates that the interval between diagnosis of septic shock and the administration of vasopressor agents is a significant although modest independent correlate to in-hospital mortality and development of late organ failure. The entire increasing mortality effect with increased delays in vasopressor initiation is related to the increased mortality in the final decile group (>14 hours post hypotension documentation) relative to the reference group. Similarly, increasing probability of incremental aggregate organ failures after the day of shock (that is, day 2 to day 10) is only seen in the highest delay decile groups (>14 hours post hypotension documentation). New-onset renal, respiratory, central nervous system, coagulation and metabolic failures were also individually associated with pressor delays >14 hours. Perhaps because of the modest strength of the correlation between pressor delay and mortality/organ failure, there is no association in the survivor group with ICU or hospital length of stay, ventilator duration or total vasopressor administration time.

Studies have shown that septic shock as defined in part by persistent hypotension is an indicator of a marked increase in morality risk in septic states 29 30 . At least two retrospective human septic shock studies show an increasing mortality with increasing severity and duration of hypotension 31 32 . Varpula and colleagues showed in 111 septic shock patients that the time spent below a MAP of 65 mmHg in the first 48 hours was a strong predictor of mortality 31 . In another retrospective study, Dünser and colleagues similarly measured the area under the curve for MAP and effect on mortality in 274 sepsis patients 32 . This study demonstrated that the time spent with MAP <55 mmHg was associated with increased risk of death. However, a similar correlation did not exist with the duration when MAP was <60 mmHg, <65 mmHg, <70 mmHg and <75 mmHg.

While there has been much study into the comparison of vasopressors/inotropes individually and in combination 33 34 35 , there has been a relative paucity in the literature regarding the timing of their initiation in septic shock. The 2012 Surviving Sepsis Guidelines recommend that vasopressor support be started for fluid-refractory shock as part of the 6-hour bundle based solely on expert opinion 15 . A rat model of endotoxic shock has suggested potential benefit with a higher proportionate splanchnic blood flow, lower lactate levels and less overall fluid support requirement for early compared with delayed norepinephrine administration 36 . A porcine model of fecal peritonitis/shock has demonstrated that delayed resuscitation (inclusive of antibiotics, fluids and pressors) was associated with increased physiologic instability and higher pressor requirements 37 . Conversely, in a small (n = 95) retrospective human study, no difference in organ dysfunction or ICU LOS was noted with early (<1.37 hours) versus late (>1.37 hours) administration of vasopressors 16 . These studies have their limitations in that two were animal studies and none utilized survival as an endpoint.

In our study, the timing of initiation of vasopressors following documentation of hypotension is only weakly associated with mortality in septic shock, as indicated by the low Wald X ² values in Table 4. The Wald X ² value for delays in antimicrobial initiation, the other remediable treatment parameter in the multivariate analysis, is 16.7 times higher. Note that this does not suggest that duration of hypotension before resuscitation (inclusive of appropriate antimicrobials and fluid resuscitation) is only weakly correlated to outcome. On the contrary, appropriate antimicrobial delays relative to hypotension and early fluid resuscitation are well established to have critical roles in improving outcome of septic shock 14 28 . Only the delay of vasopressors appears to have a limited impact on outcome in this retrospective analysis.

Given the modest strength of the association, the statistical significance of time to vasopressor initiation relates primarily to the extraordinarily large number of cases in this dataset. The only decile group that appears to carry an increased mortality or specific organ failure risk relative to the reference group is the latest group (>14 hours post hypotension documentation). All included deciles to that point appear to carry no significant increased mortality or specific organ failure risk after adjustment for multiple morbid/epidemiologic factors. This finding is entirely congruent with the findings of Subramanian and colleagues, who showed no impact of vasopressor delays up to 12 hours on organ function in a smaller cohort of <100 patients 16 .

A history of hypertension conveying a protective effect was an unexpected result on multivariate analysis. It is possible that this finding may be explained by user bias, in that these patients may have activated the healthcare system more frequently to gain a diagnosis of an otherwise silent condition. Hypertension is normally a silent condition, which may suggest that these patients had more routine access to medical care. Alternatively, the study entry criteria (decrease in systolic pressure >40 mmHg) used for many of these patients may be overly sensitive with respect to diagnosing septic shock. The impact of antimicrobial delay on mortality is not surprising because an earlier version of this database demonstrated this same finding 28 and animal studies demonstrate parallel results 38 39 .

Overall, the results of this study are congruent with the limited available human data. The study contributes significantly by adding statistical power with a larger sample size while correcting for known confounders (antimicrobial delay, disease severity). There are still significant study limitations. The study did control for delays in antimicrobial administration. However, we were unable to adjust for early fluid administration using this dataset. Although fluid resuscitation is considered a vital part of the initial resuscitation by emergency room physicians and intensivists 15 , there are studies suggesting increased mortality associated with over-resuscitation of fluids 40 41 . Other studies conversely suggest increased mortality with under-resuscitation with fluids 14 42 . Significant interactions between the timing of vasopressor initiation and early fluid resuscitation that we are unable to capture in this dataset may exist. This is a significant limitation of this study and future analyses should also attempt to factor in fluid resuscitation.

There are other limitations to this study. This is a retrospective review with its inherent inability to account for all potential confounders. However, there has yet to be a randomized controlled trial of timing of vasopressor initiation in any critical illness. Given the ethical concerns of exposing moribund patients to potential harm, a prospective, randomized human study of timing of vasopressor initiation in septic shock would be challenging. Another limitation is that the use of hypotension as the defining criteria for septic shock in this patient group may be imperfect. MAP is at best a surrogate of inadequate microvascular perfusion in shock. It does not directly capture microcirculatory perfusion and cellular injury that lead to organ dysfunction and death 7 11 13 . Nonetheless, other metabolic markers such as serum lactate and bicarbonate levels as well as severity of illness scores (APACHE II scores) were incorporated into the model to help adjust for variations in shock severity. Despite these limitations of blood pressure monitoring, given its universal access and ease of use it is the most relied upon clinical parameter for guiding therapy and will remain a mainstay in the treatment of septic shock for the foreseeable future.

From this study, we conclude that markedly delayed initiation of vasopressor medications in patients with septic shock is modestly associated with increased organ failure risk and decreased survival. Substantial delays of vasopressor initiation (>14 hours after hypotension documentation) are required to see these effects. Given the almost universal use of vasopressors in septic shock and the critical need for precise titration, further study of this area is warranted.

•Delays in initiation of vasopressor therapy following the first documentation of hypotension in septic shock are modestly associated with increased specific organ failure and mortality risk.

•This increase in specific organ failure and mortality risk is entirely driven by the decile of patients with the greatest delays of >14 hours.

•Vasopressor initiation delays are not associated with increased time on vasopressors or on mechanical ventilation among survivors.

•Delay of initiation of appropriate antimicrobial, age and APACHE II score are also independent correlates of mortality.

APACHE: Acute Physiology and Chronic Health Evaluation; CI: confidence interval; LOS: length of stay; MAP: mean arterial pressure; OR: odds ratio.

AK received unrestricted funding for the initial development of this database from Lilly, Pfizer, Astellas, Merck and Bayer. Additional support was provided through grants from the Manitoba Health Research Council, the Health Sciences Foundation and the Deacon Foundation. The current analysis/paper was not funded by any sponsor. JEP consulted with Sangart, Artisan, Philips, and Immunetrics. All other authors have no other relevant competing interests.

AK had full access to all the data in the study and is responsible for the integrity of the database and the accuracy of the data analysis. This specific research concept, the septic shock database and manuscript were developed by AK. AK, DC, AP, GLB and VB were responsible for the methodological design issues and data analysis. AK and VB drafted the manuscript. AK, VB, DC, AP, GLB, SZ and JEP contributed to data interpretation and manuscript revisions. All authors read and approved the final manuscript.