Another retrospective study proving early antibiotics in sepsis improves mortality

Christopher W. Seymour, Foster Gesten et el published a retrospective study in NEJM looking at impact of implementation of bundle treatment in sepsis.


Looked at the data(On patients with sepsis and septic shock, community acquired only) reported to New York Health Department from April 1, 2014, to June 30, 2016. More than 98% of the patients with data entered in the database were confirmed to have had severe sepsis or septic shock on manual audit


New York Department of health mandated following protocols on patients suspected of sepsis.

3-hour bundle

A. blood culture before the administration of antibiotics

B.  measurement of the serum lactate level

C. the administration of broad-spectrum antibiotics.

6-hour bundle
Bolus of 30 ml of intravenous fluids per kilogram of body weight in patients with hypotension or a serum lactate level of 4.0 mmol or more per liter, the initiation of vasopressors for refractory hypotension, and the remeasurement of the serum lactate level within 6 hours after the initiation of the protocol.


21,046 patients (18.8%) who were ineligible(no explanation why), 32,665 (29.2%) who had protocols initiated outside the emergency department, 3648 (3.3%) who had protocols initiated after 6 hours, and 5126 (4.6%) who did not have the 3-hour bundle completed within 12 hours.


Rapid completion of a 3-hour bundle of sepsis care and Rapid administration of antibiotics was associated with lower risk adjusted mortality.

This benefit was not seen with rapid administration of IV fluids.


  1. Retrospective study
  2. No analysis using excluded patients
  3. From one state, may not apply to other locations.
  4. Based on what was reported to health department
  5. Appropriateness of antibiotics was not analyzed.


In a scathing criticism of the “early antibiotics” and “each hour delay”, Mervyn Singer writes that so far all of the evidence pointing towards using early antibiotics in sepsis are based on retrospective review of data collected by regulatory authorities. Raw data are heavily adjusted statistically. There is no randomized trial to support this notion.

Problem with data collection from such sources misses the following : – 1. confirmation of infection 2. adequacy of antibiotic choice 3. antibiotic dosing 4. source control.

Furthermore, Non-infectious mimics accounted for 18% of patients initially diagnosed and treated as septic in an US emergency department (ED).

He cites several prospective studies which failed to show any difference in mortality with early antibiotics.

Biggest problem he cites is the point where we start counting time… For all studies, time is counted from presentation, which varies in the course of disease.

Shown in this picture


Following untoward outcomes may happen if we use antibiotics indiscriminately.

Increasing levels of antimicrobial resistance and harm from antibiotics themselves which include organ injury, mitochondrial dysfunction, the impact on the microbiome, and overgrowth by fungi and Clostridium difficile.


It will be very hard to have a study which would compare “early” vs “late” antibiotics.  The debate will go on. Furthermore, we are clumping all infections in one bundle of sepsis. Bacterial pneumonia definitely behaves different than meningitis.  It is time we separate infections based on system, and study the impact of therapy and its timing on the outcome.  In the interim, we have to follow the guidelines by various societies as well as government regulators.

Pantoprazole reduces GI bleeding in intubated patients without improving mortality.Read more