A couple of articles on fluid resuscitation worth mentioning. Not necessarily for their quality, but because they will be quoted and used, and critical appraisal of the content and conclusion is, without a doubt, necessary to us soldiers in the trenches.

The first one, Interaction between fluids and vasoactive agents on mortality in septic shock: a multi-center, observational study, from the october issue of the CCM Journal (2014) by Wechter et al, for the Cooperative Antimicrobial Therapy of Septic Shock Database Research Group, is a large scale effort do shed some light on one of the finer points of resuscitation, which is when to initiate vasopressors in relation to fluids in the face of ongoing shock/hypotension.

So they reviewed 2,849 patients in septic shock between 1989 and 2007, trying to note the patterns of fluid and vasopressor therapy which were associated with the best survival.  They found that survival was best when combining an early fluid loading, with pressors started somewhere in the 1-6 hour range.  I do invite you to read it for yourself, it is quite a complex analysis with a lot of permutations.

So…is it a good study?  Insofar as a retrospective study on a highly heterogeneous bunch of patients, I think so. But can I take the conclusion and generalize it to the patient I have in front of me with septic shock? I don’t think so. In all fairness, in the full text conclusion the authors concede that this study, rather than a clinical game-changer, is more of a hypothesis generator and should prompt further study. That, I think, is the fair conclusion.

In the abstract, however, the conclusion is that aggressive fluid therapy should be done, withholding vasopressors until after the first hour.  This is somewhat of a concern to me, since it isn’t uncommon for some to just read that part…

So why is this not generalizable?  First of all, I think that the very concept of generalizing is flawed.  We do not treat a hundred or a thousand patients at a time, and should not be seeking a therapeutic approach that works best for most, but for the one patient we are treating. Unfortunately, this is the inherent weakness of any large RCT and even more so in meta-analyses, unless the right subgroups have been drawn up in the study design.

Let me explain.

Patient A shows up with his septic peritonitis from his perforated cholecystitis. He’s a tough guy, been sick for days, obviously poor intake and finally crawls in. If you were to examine him properly, you’d have a hard time finding his tiny IVC, his heart would be hyperdynamic, his lungs would have clear A profiles, except maybe for a few B lines at the right base. You’d give him your version of EGDT, and he’d do pretty well. A lot better than if you loaded him with vasopressors early and worsened his perfusion. Score one for the guideline therapy.

Patient B shows up with his septic pneumonia, also a tough guy, but happens to be a diabetic with a past MI. He comes is pretty quick cuz he’s short of breath.  If you examine him properly, he has a big IVC, small pleural effusions, right basal consolidation and B lines in good quantity. He gets “EGDT” with an aggressive volume load and progressively goes into respiratory failure, which is ascribed to his severe pneumonia/ARDS, but more likely represents volume overload, as he was perhaps a little volume responsive, but not volume tolerant. An example of Paul Marik’s “salt water drowning.” (http://wp.me/p1avUV-aD) Additionally he goes into acute renal failure, ascribed to severe sepsis, but certainly not helped by the venous congestion (http://wp.me/p1avUV-2J). If he doesn’t make it, the thought process will likely be that he was just so sick, but that he got “gold standard” care. Or did he?

It may very well be that the studied group may include more Patient A types, and less B types, whose worse outcome will be hidden by the “saves” of the As. If you have a therapy that saves 15/100 but kills 5/100 you still come out 10/100 ahead… Great for those 15, not so much for the 5 outliers.

We, however, as physicians, need to apply the N=1 principle as we do not treat a hundred or a thousand patients at a time. I would not hesitate to be much more conservative in fluid resuscitating a B-type patient, regardless of the evidence.

Unfortunately, until trials include a huge number of important variables (an accurate measure of volume status, cardiac function, capillary leak, extravascular lung water, etc), it will be impossible to extrapolate results  to an individual patient.  These trials will, I suppose, eventually be done, but will be huge undertakings, and I do look forward to those results.

So, bottom line?

It’s as good a study of this type as could be done, but the inherent limitations make it of little clinical use, unless your current practice is really extreme on fluids or pressors. What it will hopefully be, however, is an onus to do the highly complex and integrative trials that need to be done to determine the right way to treat each patient we face.

 

thanks!

 

Philippe

 

COMMENTS:

Lawrence Lynn says:

Excellent post. This thoughtful quote should be read and understood by every sepsis trialists!!

“We do not treat a hundred or a thousand patients at a time, and should not be seeking a therapeutic approach that works best for most, but for the one patient we are treating.”

This single quote exposes the delay in progress caused by the ubiquitous oversimplification which defines present sepsis clinical trials. Bacteria (and viruses) generate “extended phenotypes” which are manifested in the host. These phenotypes combine with the phenotypic host response to produce the range of “dynamic relational hybrid phenotypes of bacterial and viral infection”. These hybrid phenotypes are also affected by the innoculum and/or the site of infection (vis-à-vis, your example of peritonitis).

Certainly Wechter et al and the Cooperative Antimicrobial Therapy of Septic Shock Database Research Group should be commended for beginning the process of moving toward the study of the dynamic relational patterns of complex rapidly evolving disease and treatment.

We are excited to see the beginning of the move of trialists toward the study of dynamic state of disease and treatment. However, before they can help us with meaningful results, trialists will need to study and define the range of “the dynamic relational phenotypes of severe infection” and then study the treatment actual phenotypes. This will not be easy as these organisms have had hundreds of thousands of years of evolution writing the complex genotypes which code for the extended of human infection. Sepsis trailists need to be encouraged by clinicians to rise to the task.

The clinicians must actively teach the trialists, (as you have in your post) that we expect trails which help to identity the therapeutic approach that works best in response to the dynamic hybrid phenotype “we are treating”.

The two linked articles below explain the present oversimplified state of the science of sepsis trails and why we clinicians must teach the trailists not to oversimplify and assure that they move quickly toward the study of the actual dynamic phenotypes of severe infection.

http://www.ncbi.nlm.nih.gov/pubmed/24834126

http://www.ncbi.nlm.nih.gov/pubmed/24383420

This is a paradigm shift so we, as clincians, must act to teach trailists this move is necessary. Otherwise we will continue to be left with hypotheses, which, while nice, are not useful at the bedside.

Lawrence Lynn