When I received this invitation for a PRO-CON, I accepted within 1 minute. Only later to realize that it was on “Optimised dosing according to PK/PD principles in patients – does it improve the efficacy of antibiotics?” Luckily I was given the CON, but I was in a poor position upfront: In a twitter poll 93% of voters were PRO (bias not excluded) and my opponent was Jason Roberts. So, this was my line of reasoning:
There is a lot of theoretical and preclinical data on PK/PD that justify a more personalized dosing approach of antibiotics for patients, but only few studies attempted to demonstrate that real patients actually benefit. And that is what all reviews and guidelines say, including some recent opinion papers, even drafted by my opponent. So, this debate looked like a “done-deal” before it started. Yet, in my own hospital we monitor drug concentrations and modify dosing for aminoglycosides (when used >3 days, which is mainly in CF patient and the NICU), vancomycin and azoles (both mainly in paediatic haemato-oncology). Yet, we’re not measuring drug concentrations in ICU patients receiving beta-lactam antibiotics, but most are treated with continuous infusions. So, in our hospital continuous infusion (which is fully based on PK/PD principles) has been implemented, apparently without convincing evidence for patient benefit (or safety).
Just before starting to systematically review all the literature available on this topic, I came across a French guideline based on that exercise performed by about 20 French colleagues (published 2 weeks ago!). In recognition of the absence of definitive RCTs they decided – upfront – to only provide “optional recommendations” and the level of their own agreement. Nevertheless, they unanimously agreed to suggest “administering beta-lactam antibiotics by prolonged or continuous infusions in critical care patients with septic shock and/or a high severity score in order to improve the clinical cure rate.”
The justification for that conclusion was based on 7 meta-analyses! Which is kind of surprising if there are few good quality RCTs. So, that allowed me to give my reflections on meta-analyses (as I did before), and I highlighted 2 of these. One included 22 studies, which obviously differed enormously in patient populations, infections, antibiotics, bias risk and endpoints (huge clinical heterogeneity). The pooled effect estimate yielded a 30% mortality reduction and the authors concluded that “Prolonged infusion of antipseudomonal β-lactams for the treatment of patients with sepsis was associated with significantly lower mortality than short-term infusion. Further studies in specific subgroups of patients according to age, sepsis severity, degree of renal dysfunction, and immunocompetence are warranted.” I agree on the second part, but feel that the first part is too strong (based on this meta-analysis).
The other meta-analysis used the individual patient data from 3 studies, that were very much alike (all 3 performed by the same research group, with my opponent included) and had the lowest risk of bias in the before mentioned meta-analysis. So, this, I think, is – in contrast to a classical meta-analysis – a very solid approach of gathering evidence. And the outcome was similar: 30% mortality reduction, with a number needed to treat of 12. Still, they also concluded that a definitive RCT is needed. In fact, they are doing that trial – BLING III – with 1,000 of 7,000 subjects enrolled. And I questioned whether there still is sufficient equipoise to randomize patients to a treatment strategy with a more than 50% risk of inferiority …. That’s a strong PRO argument, I guess.
To return to my CON position I only saw one remaining opportunity: When do beta-lactam antibiotics work optimally? From old textbooks I recall that if the time >MIC is 35%, and for penicillins and carbapenems (which were tested) that still holds. Now, intermittent dosing achieves 50% and continuous dosing 100% of time above MIC. But can we really expect that that change reduces hospital mortality in critically ill patients with 30%? So better convince me with a trial.
In the (twitter-)discussion Jason explained that the BLING III trial is powered on an absolute risk reduction of 3.5% for 90-day mortality. Not sure what the baseline mortality is, but if that is 20%, than the relative risk reduction is 17.5%. Hopefully we will see the results soon.
At the stage Jason asked me what I would prefer when a family member would be critically ill in ICU and in need of IV beta-lactam antibiotics. I said continuous, but I regret that I did not immediately counter with the question whether he would have a family member being randomized. The exit poll after the debate was close to 100% to zero.