Initiation of Inappropriate Antimicrobial Therapy Results in a Fivefold Reduction of Survival in Human Septic Shock

Kumar. A. Chest. 2009 Nov;136(5):1237-48. doi: 10.1378/chest.09-0087

Clinical Question

• In patients with septic shock, what is the relationship between appropriateness of initial empiric antimicrobial therapy and survival?

Design

• Retrospective observational study
• Pre-determined rules for assessing appropriateness of antimicrobial agents. For unanticipated scenarios not covered by these pre-determined rules, two infectious disease/critical physicians who were blinded to outcome, reviewed the data to determine appropriateness of antimicrobials
• Patients identified from ITU registries or ICD coding

Setting

• 22 hospitals in Canada, USA, Saudi Arabia
• 1996-2006

Population

• Inclusion:
  • Adult patients with septic shock as defined by:
    • Documented or suspected infection
    • Persistent hypotension requiring pressor therapy
    • ≥2 of the following
      • HR >90
      • RR >20 or PCO2 <32mmHg
      • Temp <36°C or >38°C
      • WCC <4 or >12 or >10% band forms
• Exclusion: none defined
• n=5715
  • Patient characteristics
    • Mean age: 62.6 years
    • 56.3% male
    • Mean APACHE II score: 25.2
    • Treatment with:
      • Drotrecogin-alfa: 2.2%
      • Low-dose steroid: 27.1%
    • Indication for source control: 43.4%
    • Community acquired infections 55%
    • Documented infection: 82.2%
    • Suspected infection without either plausible pathogen isolated, or definitive radiological, surgical, autopsy or biopsy evidence of infection: 17.8%
  • Comparing patients who received appropriate vs. inappropriate antimicrobials
    • Co-morbidities
      • Immunosuppressive chemotherapy or long term steroid therapy: 15% vs. 19.8%, p<0.05
      • COPD: 13.6% vs. 14.1%
      • Chronic renal failure: 14.6% vs. 21.6%, p<0.01
      • Long-term dialysis dependent” 7.3% vs. 10.7%
    • Clinical site infections
      • Pneumonia: 38.1% vs. 37.2%
      • Intra-abdominal 30.1% vs. 30%
      • Genitourinary: 10.9% vs. 7.3%, p<0.01
      • Skin and soft tissue: 8.1% vs. 4.6%, p<0.01
      • Intravascular catheter: 3.4% vs. 5.5%, p<0.01
      • Systemically disseminated infection (including yeast and TB): 1.7% vs. 6.8%, p<0.01
    • Primary microbiological pathogen
      • Gram negative: 53.9% vs 34.4%, p<0.01
      • Gram positive: 33.9% vs. 31.8%
        • S. aureus: 13.2% vs. 19.7%, p<0.05
      • Yeast/fungi: 6.5% vs. 28.8%, p<0.01
      • Anaerobes: 4.3% vs. 2.1%, p<0.05

Intervention

• Appropriate antimicrobial administered
  • Defined as positive if:
    • The first new antimicrobial agent given following recurrent/persistent hypotension or an antimicrobial initiated within 6 hours of the administration of the 1st new antimicrobial had:
      • in vitro activity appropriate for isolated pathogen
      • Or for culture negative shock, antimicrobial administered consistent with accepted norms modified for local flora

Control

• Inappropriate or no antimicrobial administered

Outcome

• Primary outcome: survival to hospital discharge (included discharge to chronic care facility)
  • 43.7%
• Secondary outcomes:
  • Appropriate antimicrobials initiated in 80.1%
  • Comparing appropriate vs. inappropriate antimicrobials
    • Survival to hospital discharge was significantly improved (52% vs. 10.3%, OR 9.45, 95% C.I. 7.74-11.54, p<0.0001)
    • Multivaribale logistic regression analysis performed to correct potential confounding variables: survival remained significantly improved (OR 8.99, 95% C.I. 6.60-12.23, p<0.0001)
• Subgroup analysis
  • Significantly improved mortality in patients treated with appropriate vs. inappropriate antimicrobials when analysis restricted to:
    • Culture positive cases (OR 6.95, 95% C.I. 3.11-15.52)
    • Gram positive cocci (n=1356), p<0.0001
    • Gram negative bacilli (n=2002), p<0.0001
    • Anaerobes (n=156), p=0.0048
    • Yeast (n=443), p<0.0001

• In patients with septic shock, ~20% are initially treated with inappropriate antimicrobial therapy. This is associated with a 5-fold decrease in survival.

Strengths

• Large sample size, multi-centre
• Pre-determined rules for determining appropriateness of antimicrobials
• Some known biases are adjusted for by the multivariate analyses, suggesting that the observed effect is likely to be real, even if exact magnitude cannot be determined from this study
• Simple statistical approach
• Large effect size (OR 6.95-9.45) so clinically relevant
• Similar results found in all subgroups thereby giving us confidence in results

Weaknesses

• Retrospective analysis, therefore confounded by multiple biases
• Irrespective of unknown biases, the cases and controls definitely differed significantly according to parameters which are very likely to have affected the outcome: specifically patients who didn’t receive appropriate antibiotics were more likely to have been immunosuppressed,  have had surgical complications, line infections, disseminated infections and hard to treat / highly pathogenic organisms (P aeruginosa, S aureus,  yeasts/fungi)
• The study give no data on patients lost to follow up; this could be an additional source of bias
• In the case of culture negative infections the gold standard (2 or 3 expert opinions / concordance with guidelines) is weak, although similar to that seen in clinical practice
• Duration of follow up is not stated
• Indication for source control in >40% but it was not stated if this was performed
• Data is historical: some as early as 1996. Overall mortality is very high based on current standards