Hello, my name is Natalie Cvjanovic. I'm a pediatric intensivist at the UCSF Benioff Children's Hospital in Oakland, California. Today I will discuss the topic of corticosteroids and sepsis, why, who, and how much. I have nothing to disclose. Before continuing, I would like to take a moment to honor and thank Hector Wong, whose vision, mentorship, and friendship is largely responsible for my having the opportunity to speak with you today. His passion for clinical care and the refinement of precision medicine to bring questions from the bedside to the bench and back to the bedside, has informed much of the work described in this presentation, and likely many other presentations at this conference. His untimely death has left an immeasurable void, but I'm confident that our critical care community can carry on his pioneering work and continue to improve the lives of children and adults worldwide. Thank you, Hector. This is obviously a big topic for 15 minutes, so I'll plan on briefly reviewing why steroids might be beneficial in sepsis, but we'll spend most of the time discussing who might benefit, and who might actually be harmed from steroids, and methods by which we hope to answer those questions. Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. It's a major cause of death worldwide, and sepsis-related care accounts for a majority of healthcare costs. Current guidelines reflect principles of prevention, early recognition, and supportive care. Within the guidelines for the management of patients with septic shock, there remains controversy surrounding the appropriate administration of steroids. There's little question for those of us at the bedside, that corticosteroids have direct effects on the cardiovascular system that can improve patient hemodynamics in the short term. However, there's also the concept of critical illness-related corticosteroid insufficiency, or SIRSI. SIRSI is defined as inadequate cellular corticosteroid activity relative to the severity of the patient's critical illness, and is manifest by insufficient glucocorticoid-glucocorticoid receptor-mediated down-regulation of pro-inflammatory transcription factors. The principle behind steroid supplementation in septic shock, therefore, is not fundamentally to suppress an overly exuberant response to infection, but rather to restore, balance, and calibrate the host response. The prevalence of SIRSI is thought to be as high as 60%. However, there are currently no reliable diagnostic criteria for the determination of SIRSI. We're thus left with balancing the risk-benefit ratio of treatment based on insufficient data. There is, of course, inherent risk in the administration of exogenous steroids. The best-known side effects of steroids have been well-described and are important to consider. Less well-understood, but arguably more importantly, there seems to be a signal for harm in some subpopulations of patients with septic shock, which I will describe to you in greater detail in a few minutes. So we are left with the question, to whom should we give steroids? I'm going to give you the easy ones first. Patients who are known to be adrenally insufficient, such as those with Addison's disease, panhypopituitarism, or long-term steroid use, definitely need stress-dose steroids in the setting of septic shock. Although some proportion of critically ill patients are presumed to have SIRSI and thus be functionally adrenally insufficient, as I described earlier, neither random cortisol levels nor the response to ACTH stimulation tests reliably predict this functional steroid deficiency. So how can we evaluate this question further, you might ask? The obvious answer is a randomized control trial. However, despite multiple carefully conducted randomized trials of steroids in sepsis, there is no consistent mortality benefit, as has been shown in this summary slide. So what do we do when the data don't show what we hope? Lots and lots and lots and lots of meta-analyses, of course. The overall conclusion seems to be that there's no consistent mortality benefit of steroid administration in sepsis, but there is likely earlier resolution of shock, shorter duration of mechanical ventilation, and shorter ICU length of stay. Based on data from three large randomized trials and an updated meta-analysis demonstrating faster shock resolution and more vasopressor-free days in patients treated with steroids, when weighed against relatively few long-term side effects and the widespread availability of an inexpensive drug, the Surviving Sepsis Campaign last year issued an updated recommendation for adults seen here. Although still described as a suggestion, this is a stronger recommendation in favor of using steroids compared to the 2016 guidelines. So what about data in children? Although this meta-analysis is titled Corticosteroids for Treating Sepsis in Children and Adults, it's notable that of the more than 12,000 patients who were included, as Dr. Agus pointed out after its publication, only 533 of these patients were children, and only 390 actually had sepsis. There was no difference in mortality in the pediatric subgroup, and furthermore, the pediatric population studied was significantly different at baseline than the adults included in the analysis. Here, therefore, is the Surviving Sepsis Campaign guideline for children, and consistent with the lack of evidence in children, the recommendation is far weaker than that for adults. In fact, the suggestion is that either IV hydrocortisone or no hydrocortisone may be used if adequate fluid resuscitation and vasopressor therapy are not able to restore hemodynamic stability. So how can we refine the application of a biologically plausible therapy with potentially divergent effects in complex populations? This is an illustration of prognostic and predictive enrichment. Prognostic enrichment is defined by selection of patients who are at higher risk for a particular disease outcome, for example, death, independent of treatment. In this figure, the baseline population is a heterogeneous cohort of patients with the same condition. Rather than treating the entire cohort similarly, the subjects are first stratified into low versus high risk based on their baseline prognosis. The low-risk patients, light blue and green in this figure, are assigned to standard care, whereas the high-risk patients, dark blue and pink, are further stratified into groups defined by the likelihood of a positive treatment response based on their biological subgroup. For example, use of a monoclonal antibody targeting the HER2 receptor in women with HER2 receptor-positive breast cancer. This second stratification is known as predictive enrichment. There are several advantages to this approach. If the entire group is treated similarly, not only is the likelihood of seeing a significant treatment effect of a novel therapy diluted by the low-risk patients who would have had a good outcome regardless of treatment, but importantly, the risk-benefit ratio of the novel treatment is higher for low versus high-risk patients. By differentiating between patients at high and low risk for an outcome, low-risk patients are spared exposure to potentially high-risk therapies. In addition, the likelihood of measuring an effect attributable to that treatment is optimized, increasing the power of a study and reducing the necessary sample size. So how do we employ these strategies in studying septic shock? Since sepsis and septic shock are really more a syndrome with heterogeneous ideologies and responses, the task is to define septic shock subclasses in a clinically feasible and time-efficient way. Of course, that strategy has been applied to clinical trials for decades in the selection of eligibility criteria for a given study. In fact, the difference between outcomes of an initial study of steroids in septic shock compared to the CORTICUS trial has been attributed by some to the difference in baseline mortality risk between the two study populations. Based on severity of illness scores, the patients in the ANOG trial were sicker, and therefore that study population can be considered prognostically enriched compared to the CORTICUS trial. That is, of course, an oversimplification. Otherwise, subsequent trials could just have selected sicker patients. Ongoing advances in technology and machine learning have enabled more refined approaches to prognostic enrichment. One example of prognostic enrichment has been the development of PERSEVERE, the Pediatric Sepsis Biomarker Risk Model. This model uses a panel of protein biomarkers obtained on day one of septic shock to estimate the baseline mortality risk in children with septic shock. The biomarkers were developed from discovery-oriented gene expression profiling that identified biologically plausible biomarkers for which assays already exist. Classification and regression tree analysis, or CART, is a machine learning algorithm that explains how a target variable's values can be predicted based on other values. Here's the classification and regression tree for the PERSEVERE patients. In this case, the target variable is death, and the predictive values are the biomarker levels. In this model, patients can be stratified into high versus low-risk nodes. The nodes outlined in orange are the low-risk nodes. In node seven, for example, only two of the 174 patients in that node died. The area under the curve for the recalibrated decision tree was 0.883. An updated model has a positive predictive value for mortality of 32% and a negative predictive value of 99%, thus generating two cohorts having a 30-fold difference in mortality. Again, keep in mind that these are biomarker levels obtained on day one of septic shock diagnosis. Excluding patients with a low risk of mortality based on PERSEVERE would therefore prognostically enrich a study population, which becomes especially important in pediatrics, where the baseline risk of mortality is much lower than in adults. Using an updated model which added the platelet count to PERSEVERE, known as PERSEVERE-2, patients could be further stratified into low, intermediate, and high-risk groups, again, using CART analysis. As you can see here, the low-risk group has a nearly 100% 28-day survival, whereas the high-risk group has a less than 60% 28-day survival. In this one example of prognostic enrichment by applying a therapeutic intervention only to patients at high risk of mortality, the low-risk group is spared undue exposure to potential harm, and the power of this study to show a benefit is maximized and sample size is decreased, increasing the likelihood of successful study completion. To further refine the trial design, in addition to prognostic enrichment, one can use techniques for predictive enrichment. Remember that predictive enrichment is based on shared biological features, not necessarily dependent on clinical phenotypic similarities. There are many techniques by which this can be achieved, but one of the most powerful is to use differential gene expression patterns among patients with a particular disease or syndrome. And since this is a talk on sepsis, I will give you two examples in this population. Here you see an example of predictive enrichment where we were able to identify pediatric septic shock subtypes A and B using transcriptomic data generated from whole blood-derived RNA obtained within 24 hours of a septic shock diagnosis. The 100 genes depicted in this expression mosaic reflect adaptive immunity and glucocorticoid receptor signaling. The composite mosaics represent the mean expression values of the 100 subclass-defining genes within each subclass. Red intensity correlates with increased gene expression and blue intensity correlates with decreased gene expression. Those assigned to endotype A are characterized by repression of the majority of these genes relative to patients assigned to endotype B. Examples one and two are patients with endotype A and three and four are patients with endotype B. Importantly, patients in these two subclasses or endotypes are clinically completely indistinguishable. Assignment to endotype A was independently associated with poor outcome. In this cohort of patients, clinical care, including the administration of steroids, was completely at the discretion of the care team. However, those who received corticosteroids were analyzed further and corticosteroid prescription was independently associated with poor outcome among endotype A subjects with a fourfold greater risk of death when compared to endotype B subjects. A similar strategy in adults with sepsis was used in the VANISH trial. This was a study designed to assess the effect of early vasopressin versus norepinephrine on kidney failure in patients with septic shock. Over 400 adult patients with septic shock who required vasopressors despite adequate fluid resuscitation were enrolled. It was a two by two factorial design with patients randomized to vasopressin or norepinephrine and hydrocortisone or placebo. The primary outcome was the development of kidney failure. In this study, there was no difference in kidney failure between presser groups and no interaction with hydrocortisone across the study population. In a secondary analysis of the subjects in the VANISH trial, the investigators stratified them by a transcriptomic sepsis response signature or SRS previously described based on genome-wide expression profiling in adult patients with sepsis, with community acquired pneumonia and fecal peritonitis. Similar to the pediatric data I just reviewed, although the candidate genes were different, there was a group of subjects who were relatively immunosuppressed or SRS1 and a group that was relatively immunocompromised or SRS2. Overall, the SRS1 group had increased mortality, but there was no difference in effect of vasopressor treatment on the outcome in either group. However, the patients in the SRS2 group had a significantly higher mortality if they were treated with hydrocortisone shown on the right side of this figure. This more than eightfold higher risk of mortality persisted after adjustment for age, sex, disease severity with APACHE2 scores, and comorbidities. Remember that the SRS2 group is the more immunocompetent with a better baseline mortality risk, but with steroid treatment, the mortality was similar to that of the already immunosuppressed SRS1 group. Importantly, in this study, steroid use was protocolized and uniform in contrast to the pediatric study I just described. So now we know there is signal for harm in both children and adults, yet the subgroups are not yet distinguishable at the bedside. If we use both prognostic and predictive enrichment, we can potentially optimize trial design. In a post hoc application of that strategy, where we looked at the association between endotypes, steroids, and complicated course defined as death or persistent organ failure, those patients with endotype B who were at highest risk of complicated course had a more than tenfold reduction in complicated course with the administration of corticosteroids. Remember that patients with endotype B have a general overexpression of adaptive immunity genes. This is a pediatric cohort, and an equivalent adult study has not been executed to my knowledge. So where does that leave us? The stress hydrocortisone in pediatric septic shock trial is underway, impeded, of course, by the impact of COVID, as so many studies were. It's a multicenter, double-blinded, placebo-controlled randomized trial of stress dose hydrocortisone in pediatric patients with septic shock, requiring two or more vasoactive agents, or high dose epinephrine or norepinephrine alone. The primary objective is to determine whether hydrocortisone decreases poor outcomes at 28 days. Secondary objectives include determining whether hydrocortisone decreased development of new organ dysfunction and or was associated with adverse events. Most relevant to this presentation, it will also determine whether biomarker-based prognostic and predictive enrichment strategies allow for identification of children who may be most likely to benefit from hydrocortisone. Our adult colleagues are also working on this question. The stated overall objective is to determine whether different signatures of immune status and or corticosteroids biological activity influence the responses to hydrocortisone plus flutocortisone of adults with sepsis. In the meantime, we must keep in mind that the adage no one should die without steroids is not necessarily true, and we must maintain equipoise as we continue to refine our treatment strategies. Thank you.

corticosteroids

sepsis

septic shock

steroid supplementation

diagnostic criteria

Pharmacology

Pharmacokinetics Pharmacodynamics

Presentation

Select

2022

Intermediate

Advanced

Sepsis Resources