2019 CYIR Finalized

4 events occurred in the hydrocortisone group vs. placebo, but these were not thought to impact patient outcomes. Comments 1. This is an incredibly well-conducted, placebo controlled trial, with one of the largest samples size in which to test the efficacy of corticosteroid therapy. 2. The patient population included septic shock with respiratory failure and was less ill with greater rate of surgical admissions compared the simultaneously published APROCCHSS trial (which found benefit for a combination of hydrocortisone and fludrocortisone vs. placebo among 1241 patients for 90-day mortality). 3. Despite the absence of benefit for 90-day mortality, the hydro- cortisone arm improved shock resolution consistent with prior studies. 4. Different from prior trials, these authors administered cortico- steroids as a continuous infusion and did not use a taper at the end of 7 days. 5. No treatment was guided by corticotropin testing or other sur- rogate measures of adrenal insufficiency. Hernández G, Ospina-Tascón GA, Damiani LP, Estenssoro E, Dubin A, Hurtado J, Friedman G, Castro R, Alegría L, Teboul J, Cecconi M, Ferri G, Jibaja M, Pairumani R, Fernández P, Barahona D, Granda-Luna V, Biasi Cavalcanti A, Bakker J; for the ANDROMEDA-SHOCK Investigators and the Latin America Intensive Care Network (LIVEN). Effect of a Resuscitation Strategy Targeting Peripheral Perfusion Status vs Serum Lactate Levels on 28-Day Mortality Among Patients With Septic Shock: The ANDROMEDA-SHOCK Randomized Clinical Trial. JAMA , 2019; 321(7):654-664. Summary International clinical practice guidelines recommend that the resuscitation of patients with septic shock focus on the measurement of serum lactate or lactate clearance, thought to be surrogates for tissue hypoperfusion and adequacy of treatment. However, serum lactate can be elevated in other conditions, may not be elevated in all shock states, and lactate clearance may not be pragmatic to guide clinician decision making. An alternative approach is to monitor the bedside clinical examination using capil- lary refill time, but there is little empiric evaluation of this technique. In this study, Hernandez and colleagues report the results of a 2-arm clinical trial among 424 intensive care unit patients with early septic shock, randomized to an 8-hour resuscitation strategy that was guided by every 30 minute capillary refill time (CRT) vs. every 2-hour serum lactate measurement. All administration of fluids and vaso- pressors were recommended using a standard study proto- col, and on average, the course of resuscitation was similar in both arms. The 28-day mortality rate was 43% in the lactate group compared to 35% in the CRT-guided group, corresponding to a hazard ratio of 0.75 (95%CI=0.55- 1.01, p=0.06). The authors also report lower rates of organ dysfunction measured with Sequential Organ Failure Assessment (SOFA) at 72 hours in the CRT arm (p=0.045). Comments 1. The trial just missed a traditional cutoff for statistical sig- nificance and may have been underpowered to detect the pre- planned 15% absolute reduction in 28-day mortality. 2. CRT-guided therapy may be superior to lactate guided resusci- tation but a larger confirmatory trial is required. 3. The study was not blinded and may be susceptible to bias. 4. Although the 2 arms tested may represent state-of-the-art care, the trial did not include a usual care arm or arm that combines the techniques. 5. A simple, standardized bedside clinical measurement of periph- eral perfusion may be useful in the resuscitation of septic shock, and is particularly worthy of studying in resource-limited settings. NEW STRATEGIES FOR CARE Komorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. The Artificial Intelligence Clinician learns optimal treat- ment strategies for sepsis in intensive care. Nat Med. 2018; 24(11): 1716-1720. Summary Life-threatening organ dysfunction from sepsis is typically treated with vasopressors and intravenous (IV) fluids, yet the optimal combination is unknown. As a result, wide variability in clinicians’ use of these treatments exists, perhaps contributing to unanticipated poor outcomes. Recently, clinical decision algorithms based upon rein- forcement learning (“artificial intelligence (AI)”) are pro- posed to improve patient outcomes in the intensive care unit. In this study, Komorowski et al. developed the “AI Clinician”, a computational model developed from the Medical Information Mart for Intensive Care III (MIMIC-III) dataset. Among 17,083 patients with Sepsis-3, they used 48 variables to define 4-hour patient states, and then mapped the total IV fluid volume and maximum pressor dose administered to patients in each state in 25 combina- tions. After using a Markov decision process, the authors could determine the treatment patterns that optimized patient mortality using a series of rewards and penalties (“AI Clinician policy”). The authors then studied the eICU Research Institute Database, defined another cohort of sepsis patients (N=79,073), and identified vasopressor and fluid treatments that were consistent or discordant with that recommended by the optimal AI Clinician policy. They found that the clinicians’ used a vasopressor close to that recommended by the AI policy in half of patients (56%) and only gave the recommended dose of IV fluids in one out of three patients (36%). The administration of more or less of either treatment was associated with worse mortality in a dose-dependent fashion. Comments 1. This retrospective evaluation is the first to model vasopressor and intravenous fluid treatment decisions for septic patients using reinforcement learning and artificial intelligence tools. 2. The validation of the AI policy against clinician actions in an external dataset is a significant strength. SEPSIS AND SEPTIC SHOCK

RkJQdWJsaXNoZXIy MTM1ODMw