The 7% and the liar, liar, pants on fire!


Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock, by Kumar et al [1].


When we had a discussion about the old Lancet’s article here in the blog (link), I told that someone might throw the Kumar’s 7% increase in mortality for each hour we delay the antibiotic crap on me, and I also told that the Kumar’s article would be discussed soon. Well, today is the day! Although most of us agree that patients with infections need antibiotic treatment, we still not sure how to do it. Specially in critically ill patients. We don’t know much about timing of initiation, choice of antibiotic, dose, and duration of treatment. Therefore, stories that began as myths, throughout time, are taken as unconditional true. Questionable evidences became part of daily living protocols. Worse than that, new evidences, although also weak, that contradicts previous beliefs are hardly adopted. For some situations, like crowded emergency rooms, or development countries ICUs without adequate human resources, I agree with the motto: better safe than sorry. But now, let’s put the pessimism aside, if that’s possible, and focus on what matters.

Once upon a time

In the middle 2000s, few studies had evaluated the effect of delay in antimicrobial therapy in critically ill patients, and most of them evaluated “operational delays”, like timing to admission in ICU or ER. So, Kumar and his team, based on animal models, decided to study the delay in antimicrobial therapy from initial onset of septic shock and its effects on mortality [1]. Of course every health care system, private or public, has interests for this kind of research. They can save money, increase de quality of care, and maybe decrease mortality. And, adding one good thing to another (Kumar’s will to design a study and health care system interests), Kumar and his team did what everyone is supposed to do: received “unrestricted grants from Eli-Lilly, Pfizer, Merck, and Astra-Zeneca”. A retrospective review of three cohorts of adult patients with septic shock (SS) was performed. First cohort: ICU patients of seven hospitals with SS (1999 to 2004); Second: All patients with SS at a tertiary institution (1989 to 1999); Third: Consecutive SS patients from three hospitals (approximately 150 each from 1999-2004). Why didn’t he specified the number of patients in each cohort? Can you imagine how different our practice was in 1989? We should take this questions into account when interpreting the trial’s results.

The criteria for septic shock of the 1991 sepsis definitions was used. They had definitions of clinical infections, how to qualify a potential pathogen of causing shock, time of initiation of effective antimicrobial therapy, and “questionable cases or data elements were reviewed by the principal investigator (PI) for adjudication”. Wait a second! Can they do that? The PI giving the final word? Very suspicious, dear Watson. For culture negative SS, empirical antibiotics based on the clinical syndrome were given. Documented infections were those in which a pathogen was identified from the infection site or blood in the context of a compatible clinical, radiological, surgical, or phatologic diagnosis. Otherwise, the infections were considered to be suspected. The primary outcome was survival to hospital discharge. They used a logistic regression to examine survival as function of time delay (interval data).

What they found

A total of 2,731 cases were identified. All cohorts were similar regarding APACHE, infections, time to antibiotics and outcomes. The demographic data was shown in the text (boy, I miss a Table 1). The average age was 62, average APACHE II score of 26, some patients used steroids (657) and an indication for source control existed in 39% of patients. The authors didn’t give any other informations regarding source control. How many patients had surgery or percutaneous intervention? Well, this is important, because as a friend of mine says: “You cannot treat clinically a surgical patient”. The most common infections were: lung infections 37%, intraabdominal 29%, and genitourinary 10%. Documented infections accounted for 77,9% of cases, while 22,1% represented suspected infections. We, doctors, have the nasty habit of blaming infections for everything. It was no different here. One fifth of the cases were labeled as suspicious infections. Maybe they couldn’t fin the source, but maybe it wasn’t infection at all, and until the last time I read about it, antibiotics usually don’t work under noninfectious situations!

The overall mortality was 56,2%. Around 20% of patients (558) were on effective antibiotics prior the onset of hypotension, and almost 45% of these patients required source control. Pay attention to the following numbers: the survival rate of those 558 patients who were on effective antibiotics prior the onset of hypotension was 52.2%! Of the remaining patients who received antibiotics after the onset of hypotension the survival was: 82,7% if antibiotics administered <30min of shock onset; 77,2% in the second half our, and 42% in the sixth hour. If we look at the graph published in the article, we can see that the survival rate became less than 50% only after the fourth hour. I’ve seen a lot of crazy things in my life, but how can you explain that a patient who receive appropriated antibiotics before the onset of hypotension has 52% survival rate, while patients receiving antibiotics after the onset of shock will only achieve a survival rate lower than 50% after 4 hours of shock. This contradicts the whole authors’ argument about the benefit of early antibiotics. The article gets to its climax when the authors say: over the first 6 hrs after the onset of hypotension, each hour of delay in initiation of effective antimicrobial therapy was associated with mean decrease in survival of 7.6%! For bullshit enthusiasts, that was it! Early antibiotics for everyone, but not too early. Multivariate and subgroup analysis were made, but don’t worth the discussion.

But wait! It ain’t over. Why didn’t the authors specify how many patients received antibiotics in each time-interval (hourly)? Also, only 19 patients (0,7%) did not receive effective antibiotics before death. They might be really good, but a 0.7% rate of inappropriate antibiotics is too good to be true. In addition, we don’t have any information about organ dysfunctions, vasoactive drugs and appropriated interventional (surgical or percutaneous) treatment when indicated. Do you believe in fairy tales? I don’t! I can’t recall any study in critical care that claimed such a huge benefit of an intervention. A retrospective study, with more holes than a Swiss cheese, and industry money involved became a trend in any ICU discussion in the middle 2000s. How come? I didn’t invent the numbers, they are there, you just need to pay attention. Of course I’m not saying that you shouldn’t give antibiotics to your patients. What I’m saying is that there is no free lunch, and when you hear a story too good to be true, it ain’t!


1.     Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med. 2006;34(6):1589-1596.


Photo credit

hier houd ik van 

What's Your Reaction?

Confused Confused
Damn Damn
Geeky Geeky
Like Like
Scary Scary
Win Win
Angry Angry
Happy Happy

The 7% and the liar, liar, pants on fire!

log in

Become a part of our community!

reset password

Back to
log in