Sepsis: An update Dr David Harrison Senior Statistician Intensive Care National Audit & Research Centre (ICNARC)

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Landmark trial by Rivers et al, 2001 Patients (n=263) presenting to ED with emerging septic shock in single US hospital Compared six-hour, haemodynamic goal-directed, resuscitation protocol (EGDT) with usual resuscitation Significant 16% absolute risk reduction in hospital mortality (from 46.5% to 30.5%) Reduction in hospital length of stay from 18.4 days to 14.6 days

Sepsis: An update | David Harrison Hypothesis …that, in adults presenting to ED with signs of early septic shock, delivery of protocolised resuscitation (EGDT) is superior to delivery of usual resuscitation To estimate the clinical effectiveness on all-cause mortality at 90 days To compare the incremental cost-effectiveness at one year Primary objectives

Sepsis: An update | David Harrison Participating sites 56 NHS hospitals in England (approx. 25%)

Thank you to patients and staff at…

Sepsis: An update | David Harrison Inclusion criteria Eligibility (met in ANY order):    Infection Presumed or known Hyperlactataemia Lactate (≥4 mmol l -1 ) or Refractory hypotension SBP <90 mmHg or MAP <65 mmHg (after minimum one litre fluid challenge within 60 minutes) IV antibiotics commenced SIRS – two of Temperature (≥38°C or ≤36°C) Heart rate (≥90 beats min −1 ) WBC (≤4 ×10 9 l −1 or ≥12 ×10 9 l −1 or ≥10% immature neutrophils) Respiratory rate or hyperventilation (≥20 breaths min −1 or PaCO 2 ≤4.3 kPa or acute mechanical ventilation)

Sepsis: An update | David Harrison Timeline Presentation at ED Eligibility criteria met Informed consent/agreement Randomise Up to SIX hours Up to TWO hours Clock starts ticking as soon as physiological inclusion criteria met IV antibiotics commenced before randomisation Insert PreSep TM catheter (if randomised to EGDT) Up to ONE ‘golden’ hour

Sepsis: An update | David Harrison Sample size Aim to achieve 80% power to detect an 8% absolute risk reduction in 90-day mortality from 40% to 32% (P<0.05) Required 1260 patients (630 per group) including refusals/loss to follow-up

Sepsis: An update | David Harrison

Sepsis: An update | David Harrison Eligible but not recruited 970 no research staff present – Out of hours and weekends 343 delayed referral – Referral to research staff not in time frame 354 declined consent – Time pressured stressful environment

Sepsis: An update | David Harrison Randomisation by day of week

Sepsis: An update | David Harrison Randomisation by hour of day

Sepsis: An update | David Harrison Baseline characteristics EGDT (n=625) Usual resuscitation (n=626) Mean age6664 Male57%59% Refractory hypotension54%56% Hyperlactatemia65%64% Median IV fluids1950ml2000ml Mean APACHE II score1918 Mean SOFA score4.24.3

Sepsis: An update | David Harrison Catheter insertion EGDT (n=625) Usual resuscitation (n=626) CVC insertion92%*51% Median time from randomisation to insertion 1.1h1.4h Arterial catheter insertion74%62% Median time from randomisation to insertion 1.1h1.0h * capable of continuous ScvO 2 monitoring - CVC not mandated in usual resuscitation group - Arterial catheter not mandated in either group

EGDT physiological targets

EGDT interventions

Sepsis: An update | David Harrison Analysis principles Intention to treat Pre-specified statistical analysis plan – Power GS et al. The Protocolised Management in Sepsis (ProMISe) trial statistical analysis plan. Crit Care Resusc 2013; 15: P<0.05 (two-sided) No adjustment for multiple tests

Sepsis: An update | David Harrison Primary outcome All-cause mortality at 90 days

Sepsis: An update | David Harrison Primary outcome All-cause mortality at 90 days EGDT, deaths (%) Usual resuscitation, deaths (%) P-value 184 (29.5)181 (29.2)0.90 Effect estimate (95% CI) P-value Absolute risk reduction-0.3 (-5.4 to 4.7) Relative risk1.01 (0.85 to 1.20) Adjusted* odds ratio0.95 (0.74 to 1.24)0.73 *Adjusted for components of the MEDS score and site-level random effect

Sepsis: An update | David Harrison Secondary outcomes Secondary outcomes were: – SOFA score at six and 72 hours – Receipt and duration of organ support – Duration of stay – All-cause mortality at: 28 days Discharge from acute hospital one year – Duration of survival at one year

Sepsis: An update | David Harrison Secondary outcomes EGDT group had: Increased mean SOFA score at six hours  Mean 6.4 vs 5.6 Higher proportion receiving advanced cardiovascular support  37% vs 31% Longer duration of ICU stay  Median 2.6 vs 2.2 days No difference in other secondary outcomes

Sepsis: An update | David Harrison Subgroup analyses

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison Rationale for cost-effectiveness analysis NICE provides national guidance and advice to improve health and social care Determine which treatments are cost effective for use in the National Health Service

Sepsis: An update | David Harrison Rationale for cost-effectiveness analysis New York Times, October 2008 ‘Mr Hardy is British, and the British health authorities refused to buy the medicine.’ ‘Mr Hardy’s life is not worth prolonging, according to a British government agency, the National Institute for Health and Clinical Excellence.’

Sepsis: An update | David Harrison Rationale for cost-effectiveness analysis The Guardian, October 2008 ‘NICE has to decide what treatments represent best value for the NHS. If these [expensive cancer] treatments were provided on the NHS, other patients would lose out on treatments that are both clinically and cost effective.’

Sepsis: An update | David Harrison What is economic evaluation? ‘The comparative analysis of alternative courses of action in terms of their costs and consequences’ (Drummond et al, 2005)

Sepsis: An update | David Harrison Quality-adjusted life years (QALYs) Generic - compares interventions across the board Quality of life (QOL) and length of life QOL anchored at 0 (death) and 1 (perfect health) Year of unhealthy life worth less than 1 Calculate QALY gain as a result of treatment

Sepsis: An update | David Harrison Cost-effectiveness plane New treatment less costly but less effective New treatment more effective New treatment less effective New treatment less costly Old treatment dominates New treatment dominates New treatment more effective but more costly New treatment more costly NICE tends to recommend interventions that cost less than £20,000 to gain a QALY Cost per QALY

Sepsis: An update | David Harrison Methods – resource use up to one year CategoryItemSource InterventionCatheter/consumables Protocol duration Staff time CRF Expert Opinion Initial admissionED Critical care unit General medical CRF & CMP database ReadmissionCritical care unit General medical CRF & CMP database Questionnaire CommunityGP visits Residential care Questionnaire

Sepsis: An update | David Harrison Mean total costs per patient: one year £14,375 £15,139 Post-discharge

Sepsis: An update | David Harrison CEA results: one year EGDT (n=625) Usual resuscitation (n=626) Incremental effect Mean (95% CI) Costs (£)15,13914, (-1,402 to 2,930) EQ-5D (survivors) QALY ( to 0.040) Incremental net benefit (£) * -725 (-3,000 to 1,550) * The mean QALY gain (or loss) is valued at £20,000

Sepsis: An update | David Harrison Cost-effectiveness plane: one year

Sepsis: An update | David Harrison Cost-effectiveness acceptability curve Typical NICE threshold £20,000

Sepsis: An update | David Harrison Summary – cost-effectiveness analysis On average, EGDT led to higher costs with similar quality of life and QALYs Probability that EGDT is cost-effective compared with usual resuscitation is below 30%

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison Evaluating a treatment effect Treatment received Outcome

Sepsis: An update | David Harrison Evaluating a treatment effect Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect Treatment received Outcome Confounders Risk adjust

Sepsis: An update | David Harrison Evaluating a treatment effect Treatment received Outcome Confounders Propensity match

Sepsis: An update | David Harrison Evaluating a treatment effect Instrumental variable Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect An instrumental variable is… Associated with the treatment Independent of the confounders Not associated with the outcome except through its association with treatment Instrumental variable Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect An instrumental variable is… Associated with the treatment Independent of the confounders Not associated with the outcome except through its association with treatment Instrumental variable Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect An instrumental variable is… Associated with the treatment Independent of the confounders Not associated with the outcome except through its association with treatment Instrumental variable Treatment received Outcome Confounders 

Sepsis: An update | David Harrison Evaluating a treatment effect An instrumental variable is… Associated with the treatment Independent of the confounders Not associated with the outcome except through its association with treatment Instrumental variable Treatment received Outcome Confounders 

Sepsis: An update | David Harrison Evaluating a treatment effect If an instrumental variable can be found, we can use a two-stage least-squares (2SLS) regression to get an unbiased estimate of treatment effect Stage 1: Regress treatment on the instrument Stage 2: Regress outcome on predictions from stage 1 Instrumental variable Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Treatment received Outcome Confounders  Randomisation

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Intention to treat analysis Allocated treatment Outcome Confounders  Randomisation

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Intention to treat analysis Evaluates effectiveness not efficacy In the real world, the allocated treatment is not necessarily the treatment received Allocated treatment Outcome Confounders  Randomisation

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders Adherence

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders Allocated treatment is an instrumental variable

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders Allocated treatment is an instrumental variable Associated with the treatment

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders Allocated treatment is an instrumental variable Associated with the treatment Independent of the confounders

Sepsis: An update | David Harrison Evaluating a treatment effect - RCT Allocated treatment Treatment received Outcome Confounders Allocated treatment is an instrumental variable Associated with the treatment Independent of the confounders Not associated with the outcome except through its association with treatment

Sepsis: An update | David Harrison Adherence-adjusted analysis - ProMISe Use allocated group as instrumental variable Adherence defined as proportion of the six hours that the patient was adherent to the EGDT protocol Assume relationship between adherence and treatment effect is linear

Sepsis: An update | David Harrison Adherence-adjusted analysis - ProMISe Use allocated group as instrumental variable Adherence defined as proportion of the six hours that the patient was adherent to the EGDT protocol Assume relationship between adherence and treatment effect is linear Relative risk (95% CI)P-value Original1.01 (0.85 to 1.20)0.90 Adherence-adjusted1.02 (0.78 to 1.32)0.90

New England Journal of Medicine 2015; 372:

Health Technology Assessment, in press

The ProMISe trial was funded by the National Institute for Health Research Health Technology Assessment Programme (project number 07/52/03) The views and opinions expressed are those of the authors and do not necessarily reflect those of the HTA Programme, NIHR, NHS or the Department of Health

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison

Sepsis: An update | David Harrison Australia Based on Wunsch et al 2008 (with Australia added)

Sepsis: An update | David Harrison Harmonised approach… Intensive Care Medicine 2013; 39:

Sepsis: An update | David Harrison …with prospective IPD meta-analysis Intensive Care Medicine 2010; 36:11-21.

Sepsis: An update | David Harrison The three trials (as planned)… ProCESS – US, 25 sites – 1950 patients (3 arms) – First patient: March 2008 ARISE – Australia/New Zealand, 40 sites – 1600 patients – First patient: October 2008 ProMISe – UK, 48 sites – 1260 patients – First patient: February 2011

Sepsis: An update | David Harrison Patients left to recruit (April 2013)

Sepsis: An update | David Harrison ProCESS… Mortality lower than expected Re-sized to detect same absolute risk reduction (larger relative risk reduction) Approved by DSMB & funders New target: 1350

Sepsis: An update | David Harrison Patients left to recruit (April 2013)

March 2014 October 2014 March 2015

Sepsis: An update | David Harrison Recruitment

Sepsis: An update | David Harrison Results of ProCESS, ARISE and ProMISe No difference in mortality No difference in hospital LOS Increased use of vasoactive agents (particularly dobutamine) Increased use of critical care

Sepsis: An update | David Harrison 90-day mortality

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison Intensive Care Medicine 2015; 41:

Primary outcome for each trial All trials recruiting in ED

ProCESS, ARISE and ProMISe 90-day mortality No difference in 90-day mortality…!

Sepsis: An update | David Harrison Where are we now with EGDT?

Sepsis: An update | David Harrison Management of septic shock – before Early identification Early antibiotics Early fluid resuscitation Monitoring and treatment

Sepsis: An update | David Harrison Management of septic shock – before Early identification Early antibiotics Early fluid resuscitation Monitoring and treatment – ‘EGDT’?

Sepsis: An update | David Harrison Early identification Patients were identified early in both/all arms of all three trials

Sepsis: An update | David Harrison Early identification Patients were identified early in both/all arms of all three trials Time from ED presentation to randomisation (hours) EGDTUsual careProtocolised standard care ProCESS3.3 [1.9]3.0 [1.6]3.1 [1.9] ARISE2.8 (2.1, 3.9)2.7 (2.0, 3.9) ProMISe2.7 [1.3] 2.5 (1.8, 3.5) 2.8 [1.4] 2.5 (1.8, 3.5) mean [SD] or median (quartiles)

Sepsis: An update | David Harrison Early antibiotics Patients received early antibiotics in both/all arms of all three trials

Sepsis: An update | David Harrison Early antibiotics Patients received early antibiotics in both/all arms of all three trials In both ARISE and ProMISe the first dose of antibiotics had to be given before randomisation In ProCESS, 76% received antibiotics before randomisation and 97% by hour 6

Sepsis: An update | David Harrison Early fluid resuscitation Patients received early fluid resuscitation in both/all arms of all three trials

Sepsis: An update | David Harrison Early fluid resuscitation Patients received early fluid resuscitation in both/all arms of all three trials ProCESS required a minimum of 1 litre over 30 minutes to meet refractory hypotension (initially 20 ml/kg) ARISE and ProMISe required 1 litre over 60 minutes to meet refractory hypotension

Sepsis: An update | David Harrison Early fluid resuscitation Patients received early fluid resuscitation in both/all arms of all three trials Volume of IV fluid pre-randomisation (litres) EGDTUsual careProtocolised standard care ProCESS2.3 [1.5]2.1 [1.4]2.2 [1.4] ARISE2.5 [1.2]2.6 [1.3] ProMISe1.9 [1.1]2.0 [1.1] mean [SD]

Sepsis: An update | David Harrison Monitoring High proportions of patients in the usual care arms received arterial catheters/CVCs

Sepsis: An update | David Harrison Monitoring High proportions of patients in the usual care arms received arterial catheters/CVCs Percentage receiving arterial catheter by hour 6 EGDTUsual careProtocolised standard care ProCESSNR ARISE91%76% ProMISe74%62%

Sepsis: An update | David Harrison Monitoring High proportions of patients in the usual care arms received arterial catheters/CVCs Percentage receiving CVC by hour 6 EGDTUsual careProtocolised standard care ProCESS93%58%57% ARISE90%62% ProMISe92%51%

Sepsis: An update | David Harrison Treatment Differences in treatment were significant but not large (in absolute terms)

Sepsis: An update | David Harrison Treatment Differences in treatment were significant but not large (in absolute terms) Volume of IV fluid from randomisation to hour 6 (litres) EGDTUsual careProtocolised standard care ProCESS2.8 [2.0]2.3 [1.9]3.3 [1.7] ARISE2.0 [1.4]1.7 [1.4] ProMISe2.2 [1.4]2.0 [1.3] mean [SD]

Sepsis: An update | David Harrison Treatment Differences in treatment were significant but not large (in absolute terms) Percentage receiving vasopressors from randomisation to hour 6 EGDTUsual careProtocolised standard care ProCESS55%44%52% ARISE67%58% ProMISe53%47%

Sepsis: An update | David Harrison Treatment Differences in treatment were significant but not large (in absolute terms) Percentage receiving dobutamine from randomisation to hour 6 EGDTUsual careProtocolised standard care ProCESS8%1% ARISE15%3% ProMISe18%4%

Sepsis: An update | David Harrison Treatment Differences in treatment were significant but not large (in absolute terms) Percentage receiving red blood cells from randomisation to hour 6 EGDTUsual careProtocolised standard care ProCESS14%8% ARISE14%7% ProMISe9%4%

Sepsis: An update | David Harrison Management of septic shock – after Early identification Early antibiotics Early fluid resuscitation Monitoring and treatment – but not ScvO 2 – and not all patients need arterial catheters or CVCs (although many do)

Sepsis: An update | David Harrison Where are we now with EGDT? In adults, presenting to ED with signs of early septic shock – identified early and receiving IV antibiotics and adequate fluid resuscitation - haemodynamic management according to a strict EGDT protocol does not lead to an improvement in outcome

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison What is a prospective meta-analysis? A meta-analysis of studies (usually RCTs) that were identified, evaluated and determined to be eligible for the meta-analysis before the results of those studies became known Cochrane Handbook of Systematic Reviews

Sepsis: An update | David Harrison What is a prospective meta-analysis? A meta-analysis of studies (usually RCTs) that were identified, evaluated and determined to be eligible for the meta-analysis before the results of those studies became known Cochrane Handbook of Systematic Reviews

Sepsis: An update | David Harrison What is a prospective meta-analysis? A meta-analysis of studies (usually RCTs) that were identified, evaluated and determined to be eligible for the meta-analysis before the results of those studies became known Usually undertaken by a collaborative group Usually analyse individual patient data Cochrane Handbook of Systematic Reviews

Sepsis: An update | David Harrison Protocolised Resuscitation In Sepsis individual patient data Meta-analysis PRISM ProCESS, ARISE and ProMISe investigators Independent Chief Investigator (Michael Reade)

Sepsis: An update | David Harrison Prospective IPD meta-analysis IPDMA will allow us to explore questions of interest that cannot be answered in any one trial alone – Can we identify subgroups of patients, based on site factors, patient factors or care delivery factors, for whom EGDT is effective at reducing 90-day mortality compared with usual care? – Can we identify elements of EGDT/usual care that are associated with lower 90-day mortality?

Sepsis: An update | David Harrison Prospective IPD meta-analysis Dataset harmonisation ongoing Analysis/write-up planned for December

Sepsis: An update | David Harrison

Sepsis: An update | David Harrison Sepsis: An update The ProMISe trial – Background, methods and results – Cost-effectiveness analysis – Adherence-adjusted analysis Where are we now with EGDT? – The three harmonised EGDT trials – Trial level meta-analysis – Prospective individual patient data meta-analysis Trends in severe sepsis epidemiology

Sepsis: An update | David Harrison Severe sepsis epidemiology Considerable international variation in reported incidence and mortality of severe sepsis Characteristics of patients admitted to critical care influenced by: – Provision of critical care beds – Trajectories of critical illness – Timing of recognition and response Currently no international consensus for standardised reporting of severe sepsis case mix and outcomes

Sepsis: An update | David Harrison Severe sepsis epidemiology In 2014, Kaukonen et al reported a large decrease mortality for severe sepsis in Australia and New Zealand from

Sepsis: An update | David Harrison Severe sepsis epidemiology

Sepsis: An update | David Harrison Severe sepsis epidemiology In 2014, Kaukonen et al reported a large decrease mortality for severe sepsis in Australia and New Zealand from How do results for England compare? What factors influence mortality and have these remained constant?

Sepsis: An update | David Harrison Case Mix Programme National clinical audit for adult critical care 100% coverage of adult, general critical care units in England (70-90% during study period) Severe sepsis identified from clinical data recorded during the first 24h following admission: – Infection recorded as reason for admission – 2 or more SIRS criteria – 1 or more organ dysfunctions

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Characteristics and outcomes

Sepsis: An update | David Harrison Combination of SIRS criteria

Sepsis: An update | David Harrison Combination of organ dysfunctions

Sepsis: An update | David Harrison Illustrative example – risk categories Four risk categories defined based on combinations of organ dysfunctions Categories based on crude mortality observed in previous analysis of Case Mix Programme database (Padkin et al, 2003)  %  %  %  75%+

Sepsis: An update | David Harrison Risk categories

Sepsis: An update | David Harrison Risk categories

Sepsis: An update | David Harrison Comparison with Kaukonen et al Adjusted odds ratio (95% CI) per year: Kaukonen et al0.94 (0.94, 0.95) CMP0.94 (0.94, 0.95)

Sepsis: An update | David Harrison Comparison with Kaukonen et al

Sepsis: An update | David Harrison Summary – sepsis trends The characteristics of the severe sepsis population changed over time and so did their relationship with mortality Differences in case mix preclude direct international comparisons of incidence and outcomes We have illustrated a case for developing an international consensus on standardised reporting of severe sepsis epidemiology

Sepsis: An update | David Harrison