Among the many things which have had a pendulum-like course in the ICU literature, probably none has seesawed quite as much as the subject of using steroids in septic shock. Public opinion first swung hard towards the use of steroids, then hard away, then back (in the era of "relative adrenal insufficiency") and now oscillates somewhere between "maybe useful" and "maybe pointless". However, when up against the wall, most reasonable people would agree that nobody should die of sepsis without a trial of corticosteroids.
Steroids in shock generally are asked about in Question 10 from the second paper of 2000. Evidence for steroids in sepsis and ARDS is explored in Question 22 from the first paper of 2008. Steroids in sepsis specifically are discussed in Question 25 from the second paper of 2019 and Question 16 from the first paper of 2013. Obviously, this is a fairly important issue, and because the controversy still rages, one might expect it to come up in future papers. Of specific concern are the written papers of late 2018 and early 2019, as this would be peak season to ask about the recent ADRENAL and APROCCHSS trials.
The "brief summary" below has become distended by severe bloat. A better, more concise review is available from LITFL (Corticosteroids in Refractory Shock and Steroids and Septic Shock Literature Summaries from the CCC). In point form, the key issues are as follows:
Rationale for steroids in sepsis:
- Reversal of relative adrenal insufficiency
- Reversal of inflammatory overactivity
- "Reprogramming" of the immune response (less destructive)
- Improved response to catecholamines
- Deactivation of nitric oxide synthase
- Improved cardiac tolerance of bacterial endotoxin
- Improved retention of resuscitation fluid
- Repair (or at least protection) of the endothelial glycocalyx
Classic studies and guidelines to quote:
- French study in 2002: improved mortality (severe sepsis, mortality ~ 70%)
- CORTICUS: no improvement in mortality ("mild" sepsis)
- Disagreement among meta-analysis authors (Annane vs Sligl)
- No mortality benefit in ADRENAL study (Venkatesh et al, 2018)
- Small survival benefit when used with fludrocortisone in very severe sepsis (APROCCHSS study, Annane et al 2018)
- Supported by the Surviving Sepsis Guidelines (GRADE 2B)
Rationale for the use of steroids in sepsis
Djillale Annane, the great champion of steroids in sepsis, has published a 2011 review of the topic which contains within it an excellent summary of what an optimist might expect from their "stress dose" steroid treatment.
Reversal of relative adrenal insufficiency
The concept of relative adrenal insufficiency, and precisely how one might identify "insufficient" patients, is discussed in detail elsewhere. Suffice to say these septic patients seem to require more vasopressors, and seem to die more frequently. Exogenous corticosteroid administration may be helpful in addressing their increased need for cortisol. There are probably several mechanisms responsible for this effect.
Reversal of inflammatory overactivity
At all times, within the human organism there seems to exist a proinflammatory-antinflammatory homeostasis, which is regulated by endogenous cortisol to a considerable extent. The proinflammatory system, mediated by nuclear factor kappa-B (NF-κB), promotes the secretion of inflammatory mediators. In opposition to is, the glucocorticoid receptor alpha (G-GRα) complex regulates antiinflammatory cytokines and dampens the inflammatory response. It would appear that an inadequate cortisol response to stress results in increased proinflammatory activity in critical illness, and thus to greater organ damage, increasing the mortality.
Reprogramming of the immune response
Data collected by Ehrchen et al (2007) suggests that instead of simply inhibiting the immune response, corticosteroids reorganise an redirect it into a less destructive pattern of behaviour. In their study, corticosteroids treatment "did not cause a global suppression of monocytic effector functions but results in differentiation of a specific anti-inflammatory phenotype which seems to be actively involved in resolution of inflammatory reactions." Thus redirected, monocytes go about their reconstructive tasks without laying waste to the organism with corrosive chemicals. The upshot of this could be an improvement in the vessel tone which reflects on vasopressor dose, and the improvement of microcirculatory failure which is observed in sepsis.
Improved responsiveness of α-1 receptors
When one is using α-1 agonists like noradrenaline, one tends to become concerned when the patient requires ever escalating doses. After the administration of corticosteroids, vascular reactivity to α-1 agonists tends to be restored withon several hours. The mechanism for this is poorly understood.
Correction of vasoplegia by deactivation of nitric oxide synthase
Corticosteroids seem to address the cardinal vascular dysfunction in sepsis - the overwhelming activation of nitric oxide synthase (NOS) by bacterial endotoxin and inflammatory cytokines. Mind you, this data comes from rat studies, but the message is clear - one's cardiovascular system can tolerate a whole lot more lipopolysaccharide with steroids than without them.
Improved cardiac tolerance of bacterial endotoxin
Endotoxin tends to depress cardiac function, and early isolated heart muscle studies have demonstrated abundantly that pre-treatment of the muscle with corticosteroids resulted in a marked preservation of contractility. It is difficult to assess exactly how much steroid treatment contributes to the reversal of sepsis-induced cardiomyopathy, but this laboratory evidence is encouraging for the steroid enthusiast.
Improved retention of resuscitation fluid
This may seem silly, but corticosteroids (by exerting a mild mineralocorticoid effect) produce a retention of sodium and water, thereby ensuring that all those litres of crystalloid remain inside the patient. Of course, the leakyness of the capillaries is probably going to distribute all that fluid into useless compartments, but one can hope that at least some of it will remain in circulation, and contribute to the correction of the relative hypovolemia of sepsis.
Repair (or at least protection) of the endothelial glycocalyx
Corticosteroids are among the few "well accepted" techniques which are expected to somehow defend the endothelial glycocalyx from getting denuded in septic shock. This"acceptance" is actually far from solid. The data comes from guinea pig studies. Not just guinea pig studies, but disembodied guinea pig heart studies;and they weren't even investigating a model of sepsis. The hearts were subjected to ischaemia-reperfusion injury; those which had received some hydrocortisone were found to be oozing a little less syndecan-1, a glycocalyx constituent used as a marker of its shedding. The authors concluded that " Hydrocortisone preserves the endothelial glycocalyx, sustaining the vascular barrier and reducing interstitial edema.". To generalise these findings to the population of septic humans under your care would be difficult.
Evidence for the use of steroids in sepsis
First, we loved the steroids
The idea of using steroids in sepsis is far from novel. This early age has given us much of the physiological justifications for the use of steroids in sepsis. These are fairly well summarised in the 1969 article by Schumer and Nyhus (Schumer was the Djillale Annane of the 1960s and 70s). The early interest in this strategy was manly sparked by animal study findings which suggested that corticosteroids promoted an improvement in haemodynamics and microcirculatory perfusion in septic shock. Not all of these were of the same level of clinical relevance. For instance, Jack Bloch et al (1963) were able to demonstrate an improvement of microcirculatory perfusion in septic dogs, and Tanz et al (1959) tortured some cat papillary muscle with electricity and remarked that they tolerated this better when soaked in cortisone acetate.
Not everybody were impressed by the quality of this evidence. "The use of steroids in the shocked patient is a topic which has lurked too long in the shadowy area of anecdotal reporting and conflicting laboratory results", wrote contemporary authors disappointed with the level of scientific rigor of their peers. To answer these concerns, Schumer et al (1976) ran an RCT comparison of saline (placebo), dexamethasone and methylprednisolone. The trial demonstrated a massive 20% mortality benefit among septic humans. This was sufficient for the author to promote this as a good idea in multiple subsequent papers. "30 mg/kg of methylprednisolone sodium succinate or 3 mg/kg of dexamethasone sodium phosphate is administered and repeated only once in four hours if no improvement is noted", he wrote (Schumer, 1979). Judging by the fact that nobody reacted to this (they were mainly concerned about where he put his punctuation) one might assume that at this point the use of high dose steroids in sepsis was a fairly well-accepted strategy.
Then, we cooled off about the steroids.
Obviously, the level of acceptance for the use of steroids in sepsis remained variable. The role of steroids was viewed as sufficiently controversial that people wrote "pros and cons" article duets (1981, Blainsdell con, Schumer pro.) To settle the issue, RCTs were called for. The ones from the early 1960s and 1970s were methodologically flawed, people said; and moreover they reported wildly different and unbelievable results (eg. 80% decreases or 30 increases in mortality). Sprung et al (1984) published an RCT to put this matter to rest. In summary, though the reversal of shock was much faster with steroids, the overall mortality was unchanged in these patients. The authors were forced to admit that mortality is not affected even though by reversing shock the steroids "buy time" for other therapies to work. Then, in 1987, the publication of a larger trial by Bone et al (NEJM) put a fly in the ointment by finding not only a lack of benefit in mortality, but an increased risk of infection as well. The reaction to these studies was fairly strong. A contemporary 1991 review article by Cohen and Glauser concluded that "with the possible exception of typhoid fever, there is no indication for high-dose steroids in the management of septic shock."
Then, we got excited about "stress dose" steroids
Initial studies of this technique were highly favourable. Briegel (1999) gave their patients a continuous hydrocortisone infusion and found that the reversal of septic shock (i.e. turning off the vasopressors) happened much more quickly. Then the French in 2002 reported a significant improvement in mortality among their 300-strong cohort of septic patients - from 70% to 58%. These rates seem high, but these patients were extremely ill, requiring 1.1μg/kg/min (75ml/hr) of noradrenaline.
However, subsequently results of the CORTICUS trial published in 2008 refuted the above findings. The CORTICUS investigators did not find any mortality difference associated with the use of steroids.
However, the CORTICUS patients were less "shocked", requiring on average only 0.5μg/kg/min (35ml/hr) of noradrenaline. In fact the CORTICUS investigators enrolled all-comers, regardless of the severity of sepsis, whereas the earlier French group only enrolled the doomed patients, which is reflected in their higher mortality (70% for the placebo group!).
Then, stress dose steroids became buried in negative data.
As trial data piled higher, this area of investigation turned into a battle in which meta-analysis publications were the main ordinance. A meta-analysis by Annane - a steroid enthusiast - had concluded (on the basis of 17 trials) that there is an overall mortality benefit. In contrast, Sligl et al performed the same analysis - excluding all but 6 well-designed trials - and came up with the conclusion that steroids did nothing for survival. Both reviewer groups still did agree on one thing - shock reversal was more rapid with steroids, which is the same thing people have been seeing since the 1960s.
Steroids in the era of mega-trials
With the combined might of such kaiju as ADRENAL and APROCCHSS, one might develop the impression that the matter has been finally put to rest. However the situation remains as ambiguous as ever. These studies merit closer inspection, as the trainees will likely be expected to be familiar with them (and, let's face it, it is not unlikely that one might find oneself seated across from one of the ADRENAL investigators during the viva stations).
ADRENAL (Venkatesh et al, 2018) was a towering ANZICS monstrosity (n = 3658) which enrolled enough patients to provide 90% power to detect a 5% absolute risk reduction, from a baseline mortality rate of 33%. They were mostly quite sick (about 17% needed vasopressin, at least half required 15mcg/min of a catecholamine). As is predictable in such things, there was no mortality difference at 90 days. Also as predictable, the patients spent less time on vasopressors (by 1 day), less time on the ventilator (by 1 day) and less time in the ICU (by 2 days). The investigators used a continuous infusion rather than intermittent doses, which may not matter in the least, but has been brought up by people as an oddity (i.e. "we don't do that round here").Another weird issue was that the use of steroids improved the mortality only in Australia and New Zealand (70% of the enrolment), but not in the other sites. In any case, steroid supporters will be heartened by the lack of serious adverse effects, whereas steroid-haters will have their prejudices reinforced by the steroids' failure to affect "body count", to borrow a turn of phrase from Chris Nickson.
APROCCHSS (Annane et al, 2018) in contrast was somewhat smaller (n= 1241) and compared routine care with the use of hydrocortisone plus fludrocortisone. The patients were sicker; the anticpiated mortality was 45%, and they all got shocking Euro-doses of vasopressors (on average, 1mcg/kg/min of noradrenaline, or around 70ml/hr of the standard 6mg/100ml dilution in most Australian units). The investigators found a 6% mortality reduction in the steroid group. Again, parameters which promote cessation of vasopressors, extubation and discharge all were improved by the steroids. Unfortunately, the fragility index for the primary outcome is in the single digits, i.e. a couple more patients dead in the steroid group and we are back in the "no benefit" territory. The trial was stopped several times by the steering committee, had ungeneraliseably high doses of vasopressors, and was intended to include recombinant activated protein C (drotrecogin alfa) but never did.
So, what have we learned? In general, on the basis of the available evidence one is forced to conclude that steroids will not benefit patients with "mild" septic shock, who are not requiring large doses of vasopressors and who are not experiencing multi-organ system failure. And at the terminal stages of their ICU training, an exam candidate should probably be mature enough to admit that clinical trials looking for mortality differences in management of something as complex as sepsis will probably be frustrated by the multitudes of confounding factors all of which contribute to survival. However, it also must be acknowledged that if you are a patient whose mortality from sepsis is likely to be over 60%, your condition may improve with 200mg/day of hydrocortisone, even if this manifests only in a reduction of the vasopressor doses (i.e. you get to keep your fingers and toes). This is the spirit in which the Surviving Sepsis Guidelines have recommended (2B) that steroids in sepsis should be reserved for those patients who are refractory to fluids and vasopressors. In short, this is a formal way of restating the old position of the pragmatic intensivist, that nobody should die of sepsis without having had a trial of steroids first.