For many years, sepsis has been “the clinical syndrome defined by the presence of both infection and a systemic inflammatory response.” In other words, two SIRS criteria plus a source of infection. Throw in end-organ dysfunction (or, preferably, an elevated lactate per the criteria insurance companies use) and you have “severe sepsis.” If the patient requires vasopressors, then severe sepsis becomes septic shock.
Each step up in severity represents an increase in deleterious host inflammation – a gathering “cytokine storm.” At least, that’s how I learned it. Best I can tell, this teaching is still pretty pervasive. The problem is, the more we study sepsis, the more it looks like the traditional model is completely wrong.
The new definition of sepsis
In 2016, the Third International Consensus Definitions for Sepsis and Septic Shock updated the definition of sepsis that had been in place since 1992. Instead of focusing on the systemic inflammatory response, they described sepsis as “life-threatening organ dysfunction caused by a dysregulated host response to infection.” But why change this central tenet of medical practice after almost 25 years of broad agreement?
The short answer is that the old model didn’t pan out when examined objectively over time. For starters, those SIRS criteria can be present for a whole lot of reasons other than sepsis – drug reaction, hypovolemia, pain, physical exertion and post-operative state, to name a few. When they appear in other contexts, SIRS criteria are generally not the harbingers of life-threatening organ dysfunction that we expect from sepsis. Furthermore, even when the changes are related to sepsis, SIRS criteria are pretty poor predictors of infection-related mortality. Our old model implies that more SIRS = worse outcomes, but that’s just not what we see in practice.
The cytokine storm blows over
In fact, we have a growing body of literature portraying sepsis as a state of immunosuppression, rather than uncontrolled host inflammation. For example, one study from 2011 examined cytokine levels in spleen and lung tissue from a cohort of patients who died of sepsis and compared them to a group who died of non-infectious causes. At the time, everyone thought they already knew the outcome – severe sepsis means severe inflammation, so the sepsis patients will have higher cytokine levels, right?
Wrong. Cytokine levels in the sepsis group were less than 10% of the levels observed in those who died from other causes. There was some attempt to explain this away by saying that maybe the sepsis patients had such a robust inflammatory response prior to death that they actually exhausted their ability to make more cytokines. Nice try, but this explanation doesn’t hold up against mounting evidence.
Consider a couple of studies from the early 2000s that attempted to use TNF-alpha antagonists to improve outcomes in severe sepsis. These patients were obviously still alive at the time of enrollment, yet their serum levels of TNF-alpha averaged only 41 pg/mL in one study and 9 pg/mL in the other. By comparison, another study from the same era found that a group of healthy volunteers had an average TNF-alpha level of 73 pg/mL. So, patients in the midst of their “cytokine storm” had half as much TNF-alpha in their blood as healthy people? Maybe that’s just a one-off. Maybe we’re not studying the right cytokines?
Nope. Look back even further to 1996, and you’ll find a study examining serum levels of IL-1beta in post-operative patients. The investigators also examined levels of anti-inflammatory mediators IL-1ra and IL1-RII and compared all of these levels to patients who met sepsis criteria. The surgical patients had marked elevations in IL-1beta, while – to everyone’s surprise – levels in the sepsis patients were undetectable. At the same time, levels of the anti-inflammatory markers were significantly elevated in sepsis patients compared to the surgical cohort.
Still don’t buy it? Another clue that we’ve been off track all these years is the consistent finding that anti-inflammatory agents don’t improve sepsis outcomes. We’ve tried NSAIDs, IL-1 receptor antagonists, TNF-alpha receptor antagonists, monoclonal antibodies against TNF-alpha, monoclonal antibodies against endotoxin, N-acetylcysteine, bradykinin antagonists, activated protein C and corticosteroids, to name a few. All were intended to reduce inflammation in some fashion, and none of them budged the needle on sepsis mortality. (Corticosteroids may be the one minor exception for a subset of patients with septic shock, but you get the idea.)
So, what causes sepsis?
These findings beg the question – if inflammation isn’t the problem in sepsis, then what is? As it turns out, we already have data hinting at the culprit, and it’s not what most of us would expect.
This issue of inflammation versus immune suppression in the setting of sepsis has been examined in animal models. One study took a group of mice with experimentally induced bacterial peritonitis and gave them infusions of necrotic splenocytes, apoptotic splenocytes or saline. Necrotic cells should release cytokines, while apoptotic cells should have their cytokine expression suppressed.
Viewed through our traditional lens, you might expect the mice getting necrotic splenocytes to have the worst outcomes. However, investigators found that mice receiving the necrotic cells actually had a significant improvement in survival over saline controls, while those receiving apoptotic cells fared worse. Another mouse study of peritonitis supported this finding by showing that administration of nelfinavir or ritonavir (which both have anti-apoptotic effects) led to improved survival over controls.
Apoptosis may seem an unlikely offender in the pathogenesis of sepsis, but we have at least one precedent to show that apoptosis can be lethal in the absence of inflammation. Bacillus anthracis, the bacterium that causes anthrax, produces a protein called lethal factor – a metalloprotease that inhibits production of TNF-alpha while inducing apoptosis of macrophages. Even when separated from its bacterial source, lethal factor lives up to its name. In a rat model where lethal factor alone was continuously infused, 53% of rats in the experimental group died by day seven, compared to none in the control group. Despite their worse outcomes, the rats receiving lethal factor had no significant increases in TNF-alpha, IL-1beta, IL-6 or IL-10 compared to controls.
A better model
So, we have at least one example of a bacterial exotoxin that kills by suppressing the host’s inflammatory response – perhaps there could be others? If so, this might shed some light on situations where antibiotics that inhibit protein synthesis outperform those that disrupt the cell wall: linezolid versus vancomycin for MRSA pneumonia, clindamycin versus penicillin for streptococcal myositis and fidaxomicin versus oral vancomycin for Clostridioides difficile, among others.
The explanatory power of bacterial immunosuppression doesn’t stop there, however. It could also account for those patients who are clearly infected with resultant organ dysfunction, yet they don’t display expected signs like fever or leukocytosis. It might explain why the presence of SIRS is such a poor predictor of clinical outcome – if bacteria are trying to shut down the inflammatory host response, then signs of systemic inflammation may actually suggest that the host is “winning.” It’s consistent with our experience that anti-inflammatory therapies don’t work to treat sepsis, and it fits with the observation that sepsis mortality increases rapidly with even small delays in antibiotic therapy (just not for the reasons we previously thought).
That’s not to say that a “cytokine storm” is never to blame for adverse outcomes, which is probably why the new sepsis definition went with “dysregulated host response,” rather than “downregulated host response.” COVID-19 provides a timely example of how some pathogens can, in fact, provoke deleterious inflammatory reactions that require anti-inflammatory treatment. That’s just not the case for an overwhelming majority of bacterial infections for the reasons outlined above.
Why does it matter?
Let’s say we’re wrong about how sepsis works – does it really change current practice? After all, we should still be giving antibiotics as quickly as possible, volume resuscitating as needed and assessing for source control and evidence of end-organ dysfunction. While the practical points of sepsis treatment may not shift much for now, I think that reconsidering our position on pathophysiology is important for several reasons.
First, I believe it’s time to shift resources away from studying anti-inflammatory treatments for bacterial sepsis. You know that old definition of insanity? Well, we’ve effectively been trying the same thing to mitigate sepsis mortality for the last 30 years, and we have yet to see a significant overall improvement in outcomes.
On a related note, I think the new model presents some interesting new opportunities for study. Maybe we should try using more protein synthesis inhibitors, bucking the deep-seated (and unfounded) notion that bactericidal agents are inherently better. Maybe we need more researchers looking for novel bacterial exotoxins, rather than pursuing yet another study of steroids in sepsis and hoping for a breakthrough with more rigorous subgroup analysis. (My bias: if something actually works to treat a condition, you generally shouldn’t have to massage the data to find a very specific type of patient who benefits. Xigris, anyone?)
Lastly, I think this is a good reminder that we should never get too comfortable with the status quo in medicine. We tend to cling tightly to those core beliefs and practices we learned in training, but scientific progress has a way of upending incontrovertible truths.
Remember early goal-directed therapy? Through medical school and much of my post-graduate training, EGDT stood unquestioned as best practice in the management of sepsis. It took over 15 years of follow up and meta-analysis to show us that it didn’t work. Could it be that we haven’t really advanced our treatment of sepsis because our basic understanding of it is unsound?
Sepsis is just one example of how crucial it is to keep an open mind and try to stay abreast of new data. Establish your framework, but be willing to move some pieces around as new information emerges. The old saying that half of what you learn in medical school is wrong may very well be true.