Faster antibiotics in sepsis save lives

Overview

Seymour CW et al.. N Engl J Med 2017 · PMID 28528569 · DOI 10.1056/NEJMoa1703058

Each hour of delay in antibiotics was linked to about 4% higher odds of death.

Does the study support the specific claim? No · Methodological quality: high

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Seymour CW et al.

Journal

N Engl J Med

Year

2017

Identifier

PMID 28528569 DOI 10.1056/NEJMoa1703058

Verification

verified – PubMed efetch 2026-10-25

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Supported by NIH grants R35GM119519 and K08 GM115859, VA HSR&D grant 11-109 13-079, and IPRO; author disclosure forms at NEJM.org (not in text).

2. Research question

Is faster sepsis treatment associated with lower mortality?

3. Study design and population

Study design

Cohort study – Retrospective observational cohort of state-mandated sepsis protocols (New York State Department of Health database), multilevel risk-adjusted logistic models

Population

Adults (>17 years) with severe sepsis or septic shock (Sepsis-2 criteria) and sepsis protocol initiated in the emergency department within 6 h of arrival, in 149 New York hospitals (median age 73)

Country

USA (New York)

Period

1 April 2014 to 30 June 2016

Sample size

49,331 patients at 149 hospitals (from 111,816 at 185 hospitals); 26,978 in fluid-bolus analysis

Inclusion and exclusion criteria

Included: protocol initiated in ED within 6 h; 3-hour bundle completed within 12 h. Excluded: ineligible (18.8%), protocol outside ED (29.2%), after 6 h (3.3%), bundle >12 h (4.6%), contraindication, treatment-limiting advance directives, refusal, trial enrolment; hospitals with <50 cases

4. Methods and endpoints

Exposure / intervention

Time (hours) from protocol initiation to completion of 3-hour bundle (blood culture, broad-spectrum antibiotics, lactate), to antibiotic administration, and to completion of 30 ml/kg fluid bolus

Comparator

Per-hour delay; completion within 3 h vs 3-12 h

Endpoints

Risk-adjusted in-hospital mortality

Statistical methods

Multilevel logistic regression with hospital random effects, a-priori covariates (demographics, comorbidities, infection site, admission source, shock, lactate, platelets, ventilation); internally validated risk model; sensitivity analyses; quantitative bias analysis for unmeasured confounding

5. Actual results

Results

82.5% had bundle completed within 3 h; median times: bundle 1.30 h (IQR 0.65-2.35), antibiotics 0.95 h (0.35-1.95), fluid bolus 2.56 h (1.33-4.20). In-hospital mortality 22.8% overall. Per hour to bundle completion OR 1.04 (95% CI 1.02-1.05); bundle in 3-12 h vs <=3 h OR 1.14 (1.07-1.21). Per hour to antibiotics OR 1.04 (1.03-1.06); antibiotics in 3-12 h vs <=3 h OR 1.14 (1.06-1.22). Associations stronger in patients receiving vasopressors. Fluid bolus per hour OR 1.01 (0.99-1.02; P=0.21). Bias analysis: results robust unless an unmeasured confounder were at least twice as prevalent with later treatment and raised odds of death by >1.35.

Absolute risks

Bundle completion at 6 h vs within the first hour associated with approximately 3 percentage points higher in-hospital mortality (predicted, typical patient); crude in-hospital mortality 22.6% (bundle <=3 h) vs 23.6% (3-12 h)

Authors' conclusion

Longer time to completion of the 3-hour bundle and to antibiotics was associated with higher risk-adjusted in-hospital mortality; time to fluid bolus was not; if causal, faster recognition and treatment could reduce avoidable deaths.

6. Strengths

• Large mandated statewide dataset with audits (>98% of audited cases confirmed sepsis)
• Pre-specified covariates and validated risk model
• Multiple sensitivity analyses and quantitative bias analysis
• Dose-response by hour

7. Weaknesses

• Not randomised; residual confounding (e.g. diagnostic difficulty delaying treatment)
• Appropriateness of antibiotic choice not known
• One US state; Sepsis-2 definitions
• Patients with bundle >12 h excluded; time-zero inaccuracies possible
• Per-hour effect small (4% odds increase)

8. Risk of bias

Orienting tool

ROBINS-I

Type of assessment

structured assessment based on the full text by a single reviewer (no formal double review)

Full formal assessment

no

Overall risk of bias

some concerns

Rationale

Single-reviewer structured ROBINS-I-oriented judgement based on the full text ('some_concerns' corresponds to ROBINS-I 'moderate'); not formal (needs two independent reviewers and a protocol).

Full-text review (25 October 2026): Added inclusion/exclusion flow (111,816 to 49,331), median treatment times, overall mortality 22.8%, 3-12 h vs <=3 h ORs, absolute difference (~3 percentage points), bias analysis, funding; ROBINS-I domains (overall unchanged).

9. Assessment of the statistical interpretation

Small per-hour relative effects (OR 1.04) on a high baseline mortality (about 23%) translate into an estimated ~3 percentage-point higher mortality when bundle completion is delayed from 1 to 6 hours; association, not proven causation.

10. Relevance to the specific claim

Contradicts C-ABX-12.

11. Independent contrary findings

12. Transferability to Germany and Europe

High.

13. Overall methodological judgement

Strong observational evidence that delayed antibiotics in sepsis are associated with higher mortality; directly contradicts a blanket refusal of antibiotics in serious infection.

14. Open questions

15. Linked fact checks

• Antibiotics should not be used as a treatment at all; nobody should take an antibiotic for any reason (not even after a tooth extraction); the body can cope on its own. – Studies contradicting or limiting the claim