Journal Club Fraser Morton

Journal Club De-escalation versus continuation of empirical antimicrobial treatment in severe sepsis: a multicenter non-blinded randomized non-inferiority trial Leone, M et al Intensive Care Med, Aug 2014

Conflicts of Interest None

Current Concepts De-escalation of antibiotics involves switching to a more appropriate narrow-spectrum antimicrobial regime once the causative pathogen is identified Aims to prevent super infection and drug resistance in the community, improve patient safety and reduce costs Excessive use of broad spectrum antibiotics promotes emergence of antibiotic- resistant bacterial pathogens in ICUs and poorer patient outcomes2 Surviving Sepsis Guidelines (2012)3 recommends daily re-assessment for potential de-escalation Infectious Disease Society of America systemic review of antimicrobial stewardship (2013)4 Recommends de-escalation, but recognises this is based on moderate quality evidence from clinical studies with a lack of prospective RCTs

Current Concepts Two relevant studies: Ganarcho-Montero et al., Intensive Care Med, 20135 Prospective observational study of 628 patients with severe sepsis or septic shock of which 35% had antimicrobial de-escalation De-escalation had improved mortality (OR 0.58) even with adjustment for baseline severity and risk factors Mokart et al., Intensive Care Med, 20136 Prospective observational study of 101 patients with neutropaenic sepsis of which 44% had de-escalation No change in mortality hazard ratio with de-escalation out to 1 year after discharge (with multivariate logistic regression)

Current Concepts Both studies recognised approximately only 50% of potential cases of sepsis with known microbiology had de-escalation Significant potential for improvement Cochrane Review of de-escalation of antimicrobial treatment for adults with sepsis, severe sepsis or septic shock (2013)7 Found no adequate direct evidence of RCTs assessing anti- microbial de-escalation efficacy or safety Significant clinical question with a lack of relevant evidence

Null Hypothesis De-escalation of empirical antimicrobial therapy in patients with severe sepsis or septic shock is inferior to continuation of empirical antimicrobial therapy Aim is to prove non-inferiority Primary outcome as length of ICU stay Secondary aims of comparing the two strategies on: Mortality Number of super-infections Organ failure C.difficile infection Vasopressor use Duration antimicrobial therapy Duration mechanical ventilation

Study Design Non-inferiority trial – aims to demonstrate intervention is not worse than comparator by a specified non-inferiority margin (delta) 95% Confidence Interval must be within margin to satisfy non-inferiority

Non-Inferiority Studies Common non-inferiority study design concerns8: determining non-inferiority margin – who and why not demonstrating any superiority, therefore not actually asking a significant clinical question ethical issues with assigning acceptable mortality or failure margins Susceptible to favourable bias with high discontinuation rates and intention-to-treat analysis Calculation of adequate statistical power

Study Design Multicenter non-blinded randomized non-inferiority trial Enrolled patients from 9 ICUs from Feb 2012 to April 2013 Centralised list by principal investigator with inclusion criteria assessed prior to enrollment and randomisation Ethics approved and consent from patients or relatives Patients enrolled once causative organism identified ?with sensitivities 1:1 blocked random assignment to de-escalation or continuation group No mention of stratified randomisation within centres At risk of selection bias (exacerbated by high non-inclusion) No attempts at blinding

Inclusion Criteria Presence of severe sepsis requiring empirical antimicrobial treatment Defined as SIRS and suspected infection with evidence of at least one organ failure (conventional definition) Appropriate use of empirical antimicrobial agents related to suspected source as per International Guidelines

Inclusion Flow Chart

Enrollment Issues Significant exclusion (504) – only enrolled 19% of those screened Majority lost due to logistic issues (poor documentation, consent, early discharge) 101 did not meet inclusion criteria or managed under another protocol – questions validity with current practice Only 2 excluded due to carbapenemase production (0.3%) – a very low incidence representing an abnormal and potentially more well population Incidence found between 3-70% depending on organism and geographical location9,10 Minimal loss to follow-up

Treatment Arms De-escalation strategy Continuation strategy Primary empirical antibiotic changed to narrowest spectrum possible once cultures known and based on guidelines and de-escalation hierarchy as specified Companion antimicrobial ceased at day 3, MRSA agents ceased if not identified Continuation strategy Primary antibiotic continued for duration of treatment (provided it is active for microbe) Potential for de-escalation after 8-15 days if continuing >15 days Companion drug ceased day 3-5, MRSA agents continued as per guidelines

Data collection Baseline characteristics and risk factors for multi- drug resistant bacteria collected at inclusion Duration of ICU stay as primary outcome Secondary outcomes as mentioned with follow-up to 90 days “Monitored” for organ failure, although frequency not designated Death during ICU stay reported as discharge – possible skewing of hospital stay if mortality increased

Data analysis Differences between groups assessed using multiple sampling distribution testing (i.e.. chi squared) Used linear regression modeling of ICU stay after adjustment for baseline SAPS II score Cox models performed for adjustment on SAPS II of secondary outcomes Appropriate forms of analysis for given data Post-hoc analysis to subgroup of patients with VAP Risk of Type I error

Non-inferiority margin Used data from two previous observational studies to estimate mean duration ICU stay of 10 days Using a non-inferiority margin of 2 days with 80% statistical power – calculated 51 patients would be required in each arm to establish non-inferiority (recruited 120 to allow for loss to follow-up) Duration of ICU stay expressed as median and interquartile range, mean difference and 95% CI Non-inferiority is present if upper confience bound of mean difference of ICU stay is less than 2 days

Baseline Characteristics Unbalanced groups – favours de-escalation Poor randomization (simple 1:1) Possible selection bias (not blinded) Risk of confounding bias - requires adjustment of covariates - other covariates likely to be unaccounted Longer inclusion time – potentially longer duration empirical antibiotics - weaken effect of de-escalation

Baseline Characteristics Infection source was homogenous, although pneumonia over-represented in de-escalation group - Potential selection bias – easier to identify organism and de-escalate Initial antibiotic use did not differ apart from carbapenems - Poor randomisation

Adherence to Protocol Duration of companion antibiotics significantly less in de-escalation group (2.0 vs 3.0 days), consistent with strategy goals De-escalation occurred at median day 3 [IQR 2-4], likely earliest possible availability of culture Latest de-escalation day was 9 – weakens effect od intervention

Outcomes

Outcomes Duration of ICU stay had a mean of 15.2±15.0 in the de-escalation group, and 11.8±12.6 in the empirical group (P = 0.71) Higher than estimated 10 days for sample size calculation The greater SD and IQR of de-escalation may recognise a subset in patients ineffectively treated Antibiotic duration for initial episode was unchanged, however were used significantly more in the de- escalation over the entire admission (14.1 vs 9.9 days) Potentially represents recurrent of incompletely treated infection, or super-infection as discovered (inconsistent with hypothesis)

Mortality Insignificant difference in mortality with 31% and 23% in de-escalation and empirical groups at 90 days Possible Type II error given small number and study not powered for mortality comparisons or superiority Higher than Australian figures - 18.4% absolute mortality in severe sepsis in Australian ICUs8, questioning validity although small number No significant increase in Hazard Ratio of Cox regression following adjustment Other secondary outcomes were not significantly different

Super Infection Super infection episodes requiring antibiotics occurred significantly more in the de-escalation group, 27% vs 11% Correlates with previous super-infection rates ie. 26-33% in CORTICUS study9 This accounts for the significantly increased number of days of antibiotic use in de-escalation (14.1 vs 9.9), given same antibiotic duration for initial episode No increase in mortality or organ failure despite super infection May represent colonisation rather than super-infection (given high incidence) Of the super infections, 56% were from new pathogens Possible synergistic effect of empirical therapy, multiple causative microbes or incorrectly identified initial pathogen

Results Discussion Mean difference between de-escalation group and continuation group was 3.4 (95% CI -1.7 to 8.5) Therefore the de-escalation group was not found to be non- inferior for ICU length of stay than continuation of empirical antibiotics with a non-inferiority margin of 2 days Significance of this result is debatable: indirect nature of length of stay as measure of non-effectiveness arbitrary selection of margin of 2 days Larger duration of stay than calculated for sample size Unbalance age and wellness of groups Lack of blinding could influence decision to discharge

Limitations Relatively small sample size Lack of blinding of treatment Potential weakening of effect due to heterogeneity of infective sources Attempted post hoc analysis on VAP patients with too small a sample size Not specified if de-escalation antibiotic choice made on organism alone, or culture sensitivities No interim results analysis or provision for cessation of the study if unexpected harm or benefit (although unlikely, not possible with non-inferiority model and short study)

Improvements Superiority trial Adaptive biased randomisation or at least stratified within centres Strict criteria of choice of antibiotic and de-escalation decision, with further clarification of adherence to treatment guidelines Repeat study with more narrow inclusion criteria of specific infective source Use more direct measures of infection resolution Confirm super-infection vs colonisation with same measures Blinding

Discussion Relevance New information Research question Clear relevance of topic forming major part of recommended antibiotic stewardship in sepsis management New information Paucity of evidence with no relevant RCTs identified by Cochrane Review Extrapolation of findings limited by study design Research question Clear clinical question – is de-escalation as a strategy non- inferior to empirical antimicrobial management in severe sepsis Weakened by ambiguity in definition of de-escalation and identification of suitable antibiotics, although attempted to rectify issue with development of antimicrobial hierarchy and use of international guidelines for management

Discussion Appropriateness Elimination of bias Protocol Adherence Study design compromised by desire to establish non- inferiority by an ambiguous margin on an indirect measure rather than compare strengths of treatment Elimination of bias Many potential sources of bias in unbalanced randomisation, small study population and lack of blinding Protocol Adherence Appeared to adhere to protocol although large variation in population and not explicitly confirmed Minimal loss to follow-up

Discussion Statistical analysis Supportive Data Conflicts of Interest Attempted to adjust for unbalanced populations Appropriate statistical analysis Supportive Data Data did justify conclusions although unable to extrapolate given design Conflicts of Interest Multiple authors had links to pharmaceutical companies, however would seem irrelevant given broad nature of trial

Summary The study could not demonstrate non-inferiority of de- escalation of antibiotics, despite a favourable demographic of the treatment group Primarily due to super infection of new pathogens for an unknown reason with no postulated cause, although mortality and other outcomes were unchanged – possibly represents colonisation Multiple issues with study design weakening strength of conclusion and validity Unclear whether this will alter current practice, as the large potential benefits of de-escalation are not patient specific and not assessed in this trial

References Leone, M., et al., De-escalation versus continuation of empirical antimicrobial treatment in severe sepsis: a multicenter non-blinded randomized noninferiority trial. Intensive Care Med, 2014. Eliopoulos, G.M., D.L. Paterson, and L.B. Rice, Empirical antibiotic choice for the seriously ill patient: are minimization of selection of resistant organisms and maximization of individual outcome mutually exclusive? Clinical infectious diseases, 2003. 36(8): p. 1006-1012. Dellinger, R.P., et al., Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive care medicine, 2013. 39(2): p. 165-228. Dellit, T.H., et al., Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America Guidelines for Developing an Institutional Program to Enhance Antimicrobial Stewardship. Clinical Infectious Diseases, 2007. 44(2): p. 159-177. Garnacho-Montero, J., et al., De-escalation of empirical therapy is associated with lower mortality in patients with severe sepsis and septic shock. Intensive Care Medicine, 2014. 40(1): p. 32-40.

References Mokart, D., et al., De-escalation of antimicrobial treatment in neutropenic patients with severe sepsis: results from an observational study. Intensive Care Medicine, 2014. 40(1): p. 41-49. Gomes Silva, B.N., et al., De-escalation of antimicrobial treatment for adults with sepsis, severe sepsis or septic shock. Cochrane Database Syst Rev, 2010. 12. Ricci, S., What Does ‘Non-Inferior to’ Really Mean? Cerebrovascular Diseases, 2010. 29(6): p. 607-608. Gupta, N., et al., Carbapenem-Resistant Enterobacteriaceae: Epidemiology and Prevention. Clinical Infectious Diseases, 2011. 53(1): p. 60-67. Kotsanas, D., et al., Down the drain”: carbapenem-resistant bacteria in intensive care unit patients and handwashing sinks. Med J Aust, 2013. 198(5): p. 267-9. Finfer, Simon, et al. "Adult-population incidence of severe sepsis in Australian and New Zealand intensive care units." Intensive care medicine 30.4 (2004): 589-596. Sprung, C.L., et al. "Hydrocortisone Therapy for Patients with Septic Shock", The New England journal of medicine, vol. 358, no. 2, pp. 111-24.