We conducted a study of patients presenting to the ED of a teaching hospital in Melbourne, Australia with severe sepsis or septic shock and who met clinical criteria for EGDT [1
]. We measured the incidence of candidacy for EGDT and mortality rates in such patients. We found that approximately 1.0% of patients with a diagnosis of infection were candidates for EGDT. In the absence of EGDT, hospital mortality in these patients was 30.2%. We further made comprehensive attempts to identify any other patients who might have had an elevated lactate in isolation as the sole criterion for EGDT. After their inclusion, the incidence of the syndrome (reflected by candidacy for EGDT) was still only 1.6% and its non-NFR mortality lower. These findings are relevant to other institutions considering the application of EGDT and to the powering of future randomized controlled trials.
It must be borne in mind that this (like the EGDT trial [1
]) is a single centre study. Thus, our observations might not apply to other centres. However, this is precisely the point and major finding of our investigation, namely that in different health care systems, geographical settings and institutions the syndrome described in the EGDT trial may be relatively uncommon and carry a different mortality. However, we note that our findings are concordant with those reported by Shapiro and coworkers [5
]. Those American investigators identified ED patients at risk for infection, as indicated by blood culture order. Of the 2,070 patients so identified, only 52 (2.5%) had septic shock (defined as severe sepsis plus persistent hypotension after an initial fluid challenge of 20–30 ml/kg), with a 28-day hospital mortality rate of 26.9%. This mortality rate is similar to that in our study. Using two electronic reference libraries and broad search strategies, we could not find any other studies of the outcome of ED patients with severe sepsis or septic shock.
Our study is retrospective, with all the limitations inherent to such studies. However, we used a specific and reproducible strategy to identify patients for inclusion, making it possible for other investigators to confirm or refute our findings elsewhere. Furthermore, the data used were electronically and prospectively collected and stored in various hospital databases, and were not amenable to bias or manipulation. Some candidates for EGDT might have been admitted to the general ward and missed by our identification process. However, because of our hospital's policy, all patients requiring vasopressor therapy or who remain hypotensive despite fluid resuscitation are admitted to the ICU or HDU in our institution. If any such patients were ever admitted to the general wards, it would be for palliative care only. This would have excluded them from allocation to EGDT. Nonetheless, we studied all deaths to ensure that no possible EGDT candidates who died would be missed. Only two such patients were identified who had died in the ward and were not NFR. These patients did not fulfill the clinical EGDT trial criteria and their base deficit was not greater than -3 mEq/l. Nonetheless, according to study protocol, these two patients who died were added to the overall cohort.
We might have missed patients who did not fulfill clinical criteria for EGDT but who would have been randomized to EGDT on the basis of an isolated lactate concentration above 4 mmol/l. (We note that the EGDT trial did not report how many such patients were randomized on the basis of an elevated lactate only.) However, only one patient had a lactate above 4 mmol/l in isolation. That patient survived. Because lactate was not routinely measured in all septic patients admitted to hospital, we used BE and bicarbonate as surrogates to identify patients who might have had an isolated yet unmeasured high lactate level. We thus identified several such patients and included them in our analysis. Their inclusion further lowered the mortality rate and only slightly increased the number of possible candidates for EGDT. We note that our protocol was biased toward inflating mortality by identifying all deaths among the initial cohort and potentially missing patients who fulfilled the clinical or biochemical criteria for EGDT but went on to be treated in the ward and did not die.
Some patients might have had severe sepsis or septic shock but were not identified by the ED ICD coding, medical and nursing notes, and hourly observations. This is unlikely. However, even if this were true, these patients would not have been candidates for EGDT, because such treatment requires recognition of sepsis as the cause of the patient's illness. We might have incorrectly classified patients as having severe sepsis or septic shock when in fact they were less ill. This is essentially impossible because the search criteria were prospectively determined, numerical and objective in nature, and could only be verified with independently recorded observations or laboratory findings.
The low incidence of the syndrome targeted by EGDT in our hospital may reflect a particular health care system, geographical setting, or institution. However, this is the very point we wish to make – each hospital must assess its characteristics with regard to this syndrome and its outcome. We note, however, that the incidence found in our hospital is concordant with that reported by Shapiro and coworkers [5
]. We also note that our hospital is typical of a teaching hospital in any large Australian city. However, the Australian health care system is quite different from the American system and offers free access to medical care to all patients, irrespective of health insurance status. Some patients with sepsis might present much later in the US system because concern about lack of health insurance and associated cost of care.
The low mortality in our cohort might reflect a group that was less critically ill. This once again supports our point about the need for each institution to assess its own population. Nonetheless, the mean APACHE II score was 19.8 versus 20.4 in the EGDT study; the percentages of patients with severe sepsis and septic shock in our study were 44.0% and 56.0%, respectively, compared with 48.7% and 51.3% in the control group of the EGDT study. Furthermore, our mortality rate is concordant with the findings reported by Shapiro and coworkers [5
] and those of the recent PROWESS trial [6
], which reported 30.8% mortality in septic patients randomly assigned to placebo (pre-randomization APACHE II score 25). The mortality reported in the control arm of the EGDT trial, however, is discordant with these findings.
Many of our patients were assigned to palliative care only. Some of these patients might have been considered for active therapy elsewhere and received active intervention in the EGDT study. Such patients might have died more frequently and thus led to a high mortality. However, if this were the case then this difference would provide further proof of the need to assess more broadly the generalizability of the EGDT approach.
Finally, the reported advantages of EGDT might not reflect the style of fluid and haemodynamic management per se
] but rather the general consequences of early involvement of expert physicians in the management of severe sepsis and septic shock. Early experienced team intervention, rather than specific interventions targeted at high superior vena cava oxygen saturation, may be the variable within the EGDT intervention bundle that improves survival. If this is so, then in the model of care operative at our institution (formal training in emergency medicine, senior emergency medicine staff rapidly involved in patient care 24 hours/day, early co-management and 'closed' ICU system) such advantages might already be realized. The intervention rates of central line insertion in 78% of patients, arterial cannulation in 72% of patients, and antibiotic administration in the first 6 hours in 90% of cases all support this view. More importantly, the initial mean central venous pressure CVP of above 10 mmHg in our patients (compared with between 5 and 6 mmHg in the EGDT study [1
]) and a 70% rate of vasopressor drug administration (compared with approximately 30% in the EGDT study) suggest that more aggressive and prompt resuscitation was already taking place routinely in our hospital even before the publication of the EGDT protocol.