Protocolised Management In Sepsis (ProMISe): a multicentre, randomised controlled trial of the clinical and cost-effectiveness of early protocolised resuscitation for emerging septic shock

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 07/37/47. The contractual start date was in April 2010. The draft report began editorial review in March 2015 and was accepted for publication in July 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Dr Tiffany M Osborn declares grants from ImaCor Inc. and Cheetah Medical during the trial.

Copyright statement

© Queen’s Printer and Controller of HMSO 2015. This work was produced by Mouncey et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Background and rationale

The incidence of infections severe enough to cause systemic sepsis and septic shock in adults is estimated to range from 56 to 91 per 100,000 population per year. 1 Affected patients have high mortality, morbidity and resource utilisation. 2–5 Efforts to improve care for these patients have been hampered by multiple factors including limited evidence regarding the timing and delivery of therapies. It has been suggested that there are ‘golden hours’ in the initial management of emerging septic shock during which prompt, rigorous, protocolised care may reduce unwanted consequences and improve clinical outcomes.

In 2001, Rivers et al. 6 reported the results of a single-centre, randomised controlled trial, which took place in the USA. This trial investigated the delivery of 6 hours of early goal-directed therapy (EGDT), with pre-determined haemodynamic goals, to patients presenting at an emergency department (ED) with emerging septic shock. EGDT, compared with usual resuscitation, significantly reduced hospital mortality (from 46.5% to 30.5%) and shortened hospital length of stay for survivors. The rationale for EGDT is that many patients with emerging septic shock have global tissue hypoxia that is not adequately identified using traditional resuscitation end points (such as blood pressure) and that rapid correction of occult tissue hypoxia leads to improved survival. Accordingly, resuscitation incorporating EGDT incorporates the invasive measurement of central venous oxygen saturation (ScvO₂) to detect occult global tissue hypoxia. EGDT aims to optimise tissue oxygen transport by continuous monitoring of pre-specified physiological targets – central venous pressure, mean arterial pressure and ScvO₂ – to guide delivery of intravenous fluids, vasoactive drugs and packed red blood cell transfusions.

The plausible biological rationale for EGDT, combined with the results of the Rivers et al. 6 trial and some observational studies,7–12 led to its recommendation for the initial management of patients with septic shock by the Surviving Sepsis Campaign guidelines for the resuscitation and management of severe sepsis13–15 and incorporation into the associated ‘bundles’ of care. 16 However, adoption of, and compliance with, these resuscitation and management bundles has been limited. 2,17

The lack of adoption of EGDT has primarily been due to concerns about the external validity of results from a single-centre trial, generalisability into other health-care settings, the complexity of delivery of EGDT, potential risks of the components of EGDT and the resources required for implementation. 18,19

Reports of successful implementation of EGDT have identified important enablers, including leadership (local champion); communication, education and training; buy-in to the protocol; provision for protocol transition from ED to the intensive care unit (ICU); and locally determined delivery. 20,21

Resuscitation practice in the UK, though not standardised across hospitals, usually involves intravenous fluid and vasoactive drug administration, with the intensity of resuscitation typically being determined by clinical assessment. Therapeutic strategies to improve ScvO₂ are not routinely employed during resuscitation in UK hospitals.

Despite its promising results, the Rivers et al. 6 trial can be considered only as ‘proof of concept’, and it is necessary to establish whether or not these results are generalisable to the UK NHS. The sample size was small (n = 263 patients) and single-centre studies often reflect local, and sometimes unique, processes of care. It may not be possible to replicate the results of single-centre studies in larger, multicentre studies, and important examples of this have recently been reported in the critical care literature. 22

To address these concerns, three research teams collaborated to conduct multicentre trials of EGDT in the USA (Protocolized Care for Early Septic Shock: ProCESS),23 Australasia (Australasian Resuscitation In Sepsis Evaluation: ARISE)24 and England (Protocolised Management In Sepsis: ProMISe). The three trials employed harmonised methods25 and, following full reporting, data will be pooled into one individual patient data meta-analysis. 26 Both ProCESS23 and ARISE24 have published their results (in March 2014 and October 2014, respectively) and reported no benefit of EGDT. However, both trials reported mortality in the usual-resuscitation group that was lower than anticipated (ProCESS, 60-day in-hospital mortality, 18.9% observed, 30–46% anticipated; ARISE, 90-day mortality, 18.8% observed, 38% anticipated). Consequently, neither trial could exclude, with 95% confidence, the potential for a 20% relative reduction in 90-day mortality for EGDT compared with usual resuscitation [ProCESS, relative risk 0.94, 95% confidence interval (CI) 0.77 to 1.15; ARISE, relative risk 0.98, 95% CI 0.80 to 1.21].

Aim

The overall aim of the ProMISe trial was to test the hypothesis that EGDT is superior, in terms of both its clinical effectiveness and its cost-effectiveness, to usual resuscitation in patients presenting with early septic shock to NHS EDs in England.

Objectives

Trial design

ProMISe was a pragmatic, open, multicentre, parallel-group randomised controlled trial with an integrated economic evaluation.

Research governance

ProMISe was sponsored by the Intensive Care National Audit & Research Centre (ICNARC) and co-ordinated by the ICNARC Clinical Trials Unit (CTU). An ethics application was made to the North West London Research Ethics Committee 1 on 4 May 2010 and a favourable opinion was received on 2 August 2010 (reference number 10/H0722/42).

Global NHS permissions were obtained from London North West Comprehensive Local Research Network (CLRN) on 8 September 2010 and local NHS permissions were obtained from each participating NHS hospital trust. A clinical trial site agreement, based on the model agreement for non-commercial research in the NHS, was signed by each participating NHS hospital trust and the sponsor (ICNARC).

The National Institute for Health Research (NIHR) Clinical Research Network (CRN) Portfolio details high-quality clinical research studies that are eligible for support from the NIHR CRN in England. The trial was adopted onto the NIHR CRN Portfolio on 11 July 2011.

To ensure transparency, the trial was registered for an International Standard Randomised Controlled Trial Number (ISRCTN). Registration was confirmed on 19 November 2009 (ISRCTN36307479).

Following guidelines from the NIHR, a Trial Steering Committee (TSC), with a majority of independent members, was convened to oversee the trial on behalf of the funder (NIHR) and the sponsor (ICNARC). The TSC met at least annually during the trial and comprised an independent chair (an experienced triallist); independent lay members (representing patient perspectives); independent clinicians (specialising in critical care medicine and emergency care medicine); the chief investigator (KR); and a co-investigator (JB) representing the Trial Management Group (TMG).

Additionally, an independent Data Monitoring and Ethics Committee (DMEC) was convened to monitor trial data and ensure the safety of trial participants. The DMEC met at least annually during the trial and comprised two expert clinicians specialising in critical care medicine and emergency care medicine, and was chaired by an experienced statistician.

Management of the trial

The trial manager (PM) was responsible for the day-to-day management of the trial with support from the research assistant (RJ), data manager (JT) and trial statistician (SP). The TMG was responsible for overseeing the day-to-day management of the trial and comprised the chief investigator (KR), SH, TO and the co-investigators (DB, JB, TC, DH, MS and DY). The TMG met regularly throughout the trial to ensure adherence to the trial protocol and to monitor the conduct and progress of the trial.

Network support

To maintain the profile of the trial, regular updates on trial progress were provided at quarterly meetings of the NIHR CRN Critical Care Specialty Group and at local CLRN meetings. In addition, updates were provided at national meetings, such as the Annual Meeting of the Case Mix Programme and the UK Critical Care Research Forum.

Design and development of the protocol

As part of the international collaboration to evaluate the effectiveness of EGDT for managing patients with septic shock, the ProMISe TMG worked closely with the ProCESS and ARISE TMGs in developing the trial protocol to ensure common standards, design elements and the data variables collected across the three trials. This will enable a prospective individual patient data meta-analysis to be conducted on completion and publication of all three trials. 25

Individuals representing emergency medicine, acute medicine and critical care medicine from NHS hospitals across the UK were invited to attend a meeting to discuss the trial protocol and the proposed intervention, EGDT. The meeting took place on 16 March 2010 and was attended by 91 clinicians from 54 NHS hospitals. The chairperson of the ARISE TMG also attended the meeting to share experiences in the set-up and ongoing delivery of the ARISE trial in Australasia.

Following the meeting, minor changes were made to Rivers’ EGDT protocol6 as follows:

• arterial catheter – insertion of an arterial catheter was changed from being mandated to recommended

• physiological goals – rather than a range for physiological goals, clinicians agreed that they preferred a minimum physiological goal, with no upper limit, for both central venous pressure and blood pressure

• blood pressure – a minimum physiological goal was agreed for systolic blood pressure as well as for mean arterial pressure to allow for variation in practice across NHS hospital trusts.

The trial protocol was approved by the TSC and DMEC.

Amendments to the trial protocol

Following receipt of a favourable opinion of the trial protocol from the research ethics committee on 2 August 2010, five substantial amendments were submitted and received favourable opinion. In summary, these were as follows.

Amendment 1 (March 2011): the consultee consent form was amended to the consultee agreement form to clarify that, in accordance with the Mental Capacity Act 2005,27 personal/professional consultees were being asked for their agreement, rather than their consent, for the patient to participate in the trial. A telephone agreement form was added to document cases where personal/professional consultee agreement was obtained via telephone and an emergency consent form was added to document cases where emergency consent was obtained from an independent clinician.

Amendment 2 (September 2011): in consultation with the research ethics committee, guidance was added for situations where a patient did not regain the mental capacity to provide informed consent (retrospectively) to continue participating in the trial; where possible, agreement was to be sought from a personal consultee. The exclusion criterion – immunosuppressive agents for uncured cancer or immunosuppression for organ transplantation or from systemic disease – was removed following review by the trial clinicians, who felt that this was an important group of patients who potentially might benefit from an intervention for septic shock. In addition, minor semantic changes were made to the trial protocol and the patient follow-up letter.

Amendment 3 (January 2012): the letter to the patient’s general practitioner informing them of the patient’s participation in the trial was amended for use in cases where the patient was known to have died. The patient follow-up letters were amended to be specific to the follow-up time point, namely 90 days and 1 year post randomisation. Following feedback from patients, relatives and clinicians, a short version of the patient information sheet was produced which provided salient information about the trial.

Amendment 4 (November 2012): the exclusion criterion ‘known to be participating in an interventional study’ was removed following review by the TMG; it was agreed that patients could be co-enrolled into two interventional studies if, after careful consideration, there were no concerns about patient safety, risk of biological interaction or the scientific integrity of the trial. Local principal investigators (PIs) were advised to contact the trial on a case-by-case basis to discuss the co-enrolment of patients. In addition, minor semantic changes were made to the trial protocol and the consent/consultee agreement forms.

Amendment 5 (November 2013): a newsletter for patients participating in the trial was produced and sent with the follow-up questionnaires at 90 days and at 1 year post randomisation. Permission was also sought from the research ethics committee to e-mail follow-up questionnaires to patients, if requested.

NHS support costs

Trials in emergency and critical care are challenging and expensive to conduct. Unlike in other areas of health care, such as oncology, recruitment cannot take place solely within usual office hours. Resources are needed to enable screening and recruitment 24 hours per day, 7 days per week. Patients with severe sepsis and emerging septic shock are more likely to present at the ED in the afternoon through to late at night. Another challenge of emergency and critical care research is the informed consent process, which often has to be completed within a very short time frame, as treatments are often time limited. For ProMISe, consent and randomisation occurred within 2 hours of the patient meeting eligibility. Critically ill patients usually lack the mental capacity to be able to provide informed consent prior to randomisation, in which case it is necessary to involve a personal or professional consultee in accordance with the Mental Capacity Act 2005. 27 Senior, experienced staff are needed to be able to assess the patient’s mental capacity and to be able to effectively communicate information about the trial to the patient and/or their relatives in a stressful situation.

To this end, resources equivalent to 0.9 whole-time equivalent band 8 research nurse NHS support costs were successfully agreed with the London North West CLRN on 3 December 2010. Resources were based on an estimated 22 eligible admissions per site per year, of whom 14 would be recruited and 7 would be randomised to receive EGDT. Using these recommendations, participating sites, assisted by the TMG, negotiated resources required locally for the trial with their respective research and development departments and CLRNs.

Trial equipment

The central venous catheters with ScvO₂ monitoring capability (PreSep^TM central venous oximetry catheter), for use in patients allocated to the EGDT (intervention) group, were purchased from Edwards Lifesciences Ltd (Newbury, Berkshire) and distributed to participating sites by the ICNARC CTU. Edwards Lifesciences loaned the Vigileo^TM monitor required for continuous monitoring of ScvO₂ to each participating site for the duration of the trial. Both the PreSep^TM central venous oximetry catheter and the Vigileo^TM monitor are manufactured by Edwards Lifesciences and are commercially available and licensed for use in the UK. Each participating site received training in the use of the PreSep^TM central venous oximetry catheter and the Vigileo^TM monitor, provided free of charge by Edwards Lifesciences. In addition, Edwards Lifesciences provided 24-hour, 7-days-per-week telephone support for any technical queries. Edwards Lifesciences had no further role in the trial.

Patient and public involvement

Engagement with patients was vital to the successful conduct of the trial. Two former critical care patients were independent members of the TSC and provided input into the conduct of the trial, including reviewing literature to be given to patients and their families (e.g. patient information sheets and patient newsletters).

Participants: sites

The trial aimed to recruit a representative sample of 48 NHS hospitals in the UK. The criteria for inclusion were:

• EGDT, including continuous monitoring of ScvO₂, was not already part of usual resuscitation for patients presenting with severe sepsis/septic shock

• agreement from senior clinical staff in emergency care, acute care and critical care to recruit eligible patients and to adhere to the trial protocol – sites were asked to identify a ‘champion’ from each specialty to promote the trial locally

• identification of a local PI and a dedicated research nurse to take responsibility for the local conduct of the trial

• provision of timely data on recruited patients entered onto a secure, dedicated, electronic case report form.

Invitations for expressions of interest were sent to lead clinicians in acute medicine, emergency medicine and critical care medicine at NHS hospitals throughout the UK. Invitations were also circulated via the College of Emergency Medicine, the Society of Acute Medicine and the Intensive Care Society. The trial was promoted through presentations at national meetings of all three organisations.

Participants: patients

The trial procedures for recruitment and follow-up of patients are summarised in Figure 1.

FIGURE 1.

Summary of trial procedures for the recruitment and follow-up of patients. HSCIC, Health and Social Care Information Centre.

Screening and recruitment

Following attendance at a site initiation meeting, screening and recruitment was commenced at participating sites once the clinical trial site agreement had been signed and all necessary approvals were in place.

To promote awareness of the trial and facilitate recruitment, posters providing information about ProMISe were displayed in the ED and in family/visitor waiting rooms.

Potentially eligible patients were identified and approached by authorised members of staff about taking part in the trial. Information about the trial was provided to the patient; this included the purpose of the trial, the consequences of taking part or not, data security and funding of the trial. This information was also provided in a patient information sheet (see Appendix 1), along with the name and contact details of the local PI, which was given to the patient to read before making the decision whether or not to take part in the trial. A short version of the patient information sheet, summarising the salient information about the trial, was also provided (see Appendix 2).

If the patient lacked mental capacity (because of their acute illness) to understand the information about the trial, then, in accordance with the UK Mental Capacity Act 2005,27 a personal consultee, who could be a relative or close friend, was identified with whom the patient’s participation in the trial could be discussed. If there was no personal consultee available, the patient was provided with a professional consultee – an independent mental capacity advocate appointed by the NHS hospital trust – with whom the patient’s participation in the trial could be discussed. If there was neither a personal nor a professional consultee immediately available in person or via the telephone, an independent clinician (senior doctor or nurse) was consulted in person or via telephone for emergency consent. The personal/professional consultee or independent clinician was provided with the same information as patients (see Appendix 1) along with an explanation that they were being asked for their agreement for the patient taking part in the trial. Patients, personal/professional consultees and independent clinicians were provided with an opportunity to ask questions before being invited to sign the consent form, personal/professional consultee agreement form or emergency consent form, as appropriate.

Informed consent

Staff members who had received training on the background, rationale and purpose of ProMISe and on the principles of the International Conference on Harmonisation Good Clinical Practice guidelines were authorised to take informed consent from patients, informed agreement from a personal or professional consultee or emergency consent from an independent clinician.

Once the staff member taking informed consent, consultee agreement or emergency consent was satisfied that the patient, personal/professional consultee or independent clinician had read and understood the patient information sheet and all their questions about the trial had been answered, the patient, personal/professional consultee or independent clinician was invited to sign the consent form, personal/professional consultee agreement form or emergency consent form, as appropriate.

For patients who had lacked mental capacity prior to randomisation, informed consent to continue participating in the trial was sought as soon as possible after the patient had regained mental capacity. If a patient did not regain mental capacity, then, if possible, agreement from a personal consultee was obtained for the patient to continue participating in the trial.

Randomisation and allocation procedure

Following informed consent from the patient, agreement from a personal/professional consultee or emergency consent from an independent clinician, eligible patients were randomised within 2 hours of meeting eligibility via a central 24-hour, 7-days-per-week, telephone randomisation service hosted by Sealed Envelope Ltd. Patients were randomly allocated 1 : 1 to either the EGDT group or the usual-resuscitation group, by computer-generated randomised permuted blocks (with variable block lengths of 4, 6 and 8) stratified by recruiting site. A manual randomisation list was prepared a priori by the trial statistician in case the central telephone randomisation service was not available for any reason. Staff at participating sites were advised to call the 24-hours-per-day, 7-days-per-week telephone support service if they experienced any problems with the central telephone randomisation service. Manual randomisation was carried out, as required, by the on-call member of the TMG.

Screening log

To enable full and transparent reporting for the trial, brief details of all patients who met eligibility criteria or who met all inclusion criteria plus one or more of the exclusion criteria were recorded in the screening log. The reasons for eligible patients not being recruited were recorded, which included the patient declining the invitation to take part, the patient being excluded by the treating clinician, logistical reasons, etc. No patient identifiers were recorded in the screening log.

Treatment groups

Early, goal-directed therapy (intervention)

For patients randomised to the EGDT group, during the first hour (defined as the next whole hour, e.g. if randomised at 09.24, then by 11.00), a PreSep^TM central venous oximetry catheter was inserted into either a subclavian or an internal jugular vein using standard techniques for central venous access and calibrated against a sample aspirated from the catheter and analysed by co-oximetry. Central venous catheters were managed according to the guidelines of the Central Venous Catheter Care Bundle. 29 If not already initiated, supplemental oxygen was administered, with intubation and mechanical ventilation as needed, to maintain an arterial oxygen saturation (SpO₂) of ≥ 93%. An arterial catheter was recommended, but not mandated.

The EGDT resuscitation protocol (Figure 2) was followed for 6 hours (intervention period) with personnel involved and treatment location decided by each site. At least one trained member of staff was available throughout the 6-hour intervention period. All other treatment, during the intervention period and after, was at the discretion of the treating clinician(s).

FIGURE 2.

Early goal-directed therapy resuscitation protocol. CVC, central venous catheter; CVP, central venous pressure; Hb, haemoglobin; i.v., intravenous; MAP, mean arterial pressure; PRBC, packed red blood cells; SBP, systolic blood pressure.

Each element of the resuscitation protocol was administered in series or simultaneously, depending on the clinical assessment of the patient’s requirements. For example, the clinical team could choose to administer intravenous fluids in conjunction with vasopressors if a patient was in extremis.

Central venous pressure

Intravenous fluid boluses in half-litre or equivalent increments were given every 30 minutes until a minimum central venous pressure of 8 mmHg was achieved, unless the treating clinician discerned a risk to patient safety. The type of intravenous fluid and the rate of administration were at the discretion of the treating clinician(s).

Blood pressure

If the mean arterial pressure was < 65 mmHg or the systolic blood pressure was < 90 mmHg and the central venous pressure was at least 8 mmHg, vasopressors were administered and titrated to a achieve a minimum mean arterial pressure of 65 mmHg or a systolic blood pressure of 90 mmHg. The choice of vasopressor was at the discretion of the treating clinician(s) based on best evidence, the patient’s clinical needs and local policy. If the mean arterial pressure was > 90 mmHg, clinicians could consider administering a vasodilator agent to reduce afterload, if clinically indicated.

Central venous oxygen saturation

Once the central venous pressure was at least 8 mmHg and the mean arterial pressure was at least 65 mmHg or the systolic blood pressure at least 90 mmHg, treatment was initiated, if necessary, to achieve a minimum ScvO₂ of 70%. If the ScvO₂ was < 70% and the post-fluid resuscitation haemoglobin was < 10 g/dl, packed red blood cells were transfused. If the ScvO₂ was < 70% and the haemoglobin was at least 10 g/dl, an infusion of dobutamine was commenced, at an initial rate of 2.5 µg/kg/minute for 30 minutes, and then increased by 2.5 µg/kg/minute every 30 minutes, to a maximum dose of 20 µg/kg/minute, until a ScvO₂ of ≥ 70% was achieved. The dose of dobutamine was reduced or the infusion discontinued if there was concern about drug-induced tachycardia or arrhythmia. If the ScvO₂ remained < 70%, the clinician could consider mechanical ventilation (with sedation and paralysis) to decrease oxygen consumption.

Monitoring

Once all physiological goals for central venous pressure, blood pressure and ScvO₂ were met, the patient was monitored continuously for the remainder of the intervention period (a total of 6 hours). If the central venous pressure, blood pressure or ScvO₂ fell below its physiological goal during the 6-hour intervention period, the EGDT resuscitation protocol recommenced. At the end of 6 hours, continuous ScvO₂ monitoring was no longer mandated and the patient returned to standard care.

Outcome measures

The primary clinical effectiveness outcome was all-cause mortality at 90 days following randomisation and the primary cost-effectiveness outcome was incremental net monetary benefit (INB) gained at 1 year, at a willingness to pay of £20,000 per quality-adjusted life-year (QALY). Secondary outcomes were as follows:

• Sequential Organ Failure Assessment (SOFA) score30 at 6 and 72 hours

• receipt of and days alive and free (up to 28 days) from advanced cardiovascular, advanced respiratory or renal support31

• ED, critical care and acute hospital length of stay

• duration of survival

• all-cause mortality at 28 days, at acute hospital discharge and at 1 year

• health-related quality of life, resource use and costs at 90 days and at 1 year

• lifetime incremental cost-effectiveness.

Safety monitoring

Patients were monitored for adverse events that occurred between randomisation and 30 days following randomisation. Specified adverse events were defined as follows:

• pneumothorax – defined as any new pneumothorax requiring insertion of a chest drain (intercostal catheter)

• haemo-pneumothorax – defined as any new haemo-pneumothorax requiring insertion of a chest drain

• bleeding – defined as any new, overt blood loss requiring transfusion of one or more units of blood

• thrombosis – defined as any new clinical and radiographic evidence of a deep-vein thrombus

• pulmonary emboli – defined as any new evidence from computed tomography pulmonary angiogram with appropriate clinical history

• vascular catheter infection – defined as any new vascular catheter-related infection in which a vascular catheter, such as a central venous catheter, was identified as the primary source of infection and associated with signs and symptoms of infection requiring antimicrobials

• pulmonary oedema – defined as any new radiographic evidence consistent with pulmonary oedema

• blood transfusion reaction – defined as any allergic reaction to blood transfusion, haemolysis related to incompatible blood type or alteration of the immune system related to blood transfusion

• myocardial ischaemia – defined as any new acute electrocardiogram changes with appropriate clinical findings and changes in cardiac troponins or non-ST segment elevation myocardial infarction with appropriate increases in cardiac troponins but without electrocardiogram changes

• peripheral ischaemia – defined as any new sustained depression or loss of arterial pulse (as determined by palpation or Doppler ultrasonography) resulting in symptoms consistent with ischaemia or obvious gangrene.

Unspecified adverse events were defined as an unfavourable symptom or disease temporally associated with the use of the trial treatment, whether or not it was related to the trial treatment, that was not deemed to be a direct result of the patient’s medical condition and/or standard critical care treatment.

All adverse events were recorded in the electronic case report form and reported, as part of routine reporting throughout the trial, to the DMEC and the research ethics committee. Adverse events that were assessed to be serious (i.e. prolonging hospitalisation or resulting in persistent or significant disability/incapacity), life-threatening or fatal – collectively termed serious adverse events – were reported to the ICNARC CTU and reviewed by a clinical member of the TMG. Serious adverse events that were unspecified and considered to be possibly, probably or definitely related to the trial treatment were reported to the research ethics committee within 15 calendar days of the event being reported.

Data collection

A secure, dedicated electronic case report form, hosted by ICNARC, was set up to enable trial data to be entered by staff at participating sites. The electronic case report form was accessible only to authorised users, and access was approved centrally by the trial manager or the data manager (after cross-checking the site delegation of trial duties log). Each individual was provided with a unique username and password, and had access to data only for the patients recruited at their site.

The data set for ProMISe included the minimum data required to confirm patient eligibility, to describe the patient population, to monitor and describe delivery of the intervention, to assess primary and secondary outcomes and to enable linkage to the ICNARC Case Mix Programme, the national clinical audit of adult critical care32 (see Appendix 3).

Longer-term follow-up

Following randomisation, a letter was sent to the patient’s general practitioner informing them of the patient’s participation in the trial and issuing a request for assistance with follow-up, if required. All patients who survived to leave hospital were followed up at 90 days for the primary clinical effectiveness outcome (all-cause mortality) and secondary outcomes (health-related quality of life and resource use), and at 1 year for secondary outcomes (all-cause mortality, duration of survival, health-related quality of life and resource use) and to calculate the primary cost-effectiveness outcome (INB).

Data linkage with death registration

Follow-up of patients was carefully monitored to prevent any potential distress to those who care for the patient receiving a letter addressed to a deceased relative, partner or friend. The follow-up process started at 75 days for the 90-day follow-up and at 350 days for the 1-year follow-up to allow for the administrative processes. Each week a list of all patients who had been discharged alive from hospital and who were either 75 days or 350 days post randomisation was sent to the Health and Social Care Information Centre Data Linkage and Extract Service to confirm their mortality status. Patients indicated as having died were logged and the follow-up process ended.

Follow-up procedure

Patients identified by the Health and Social Care Information Centre Data Linkage and Extract Service as not having died started the follow-up process, as summarised in Figure 3. A questionnaire pack was sent from the ICNARC CTU, by post, to the patient. Following evidence-based practice for maximising responses to postal surveys,35 the questionnaire pack included a cover letter (see Appendix 6); the patient information sheet (see Appendix 1) or patient newsletter (which replaced the patient information sheet in November 2013); two questionnaires – the Health Questionnaire (see Appendix 7) and the Health Services Questionnaire (see Appendix 7); a stamped, addressed return envelope; and a pen. The Health Questionnaire (see Appendix 7) included the required questions from the European Quality of Life-5 Dimensions-5-Level (EQ-5D-5L) questionnaire to evaluate health-related quality of life and the Health Services Questionnaire (see Appendix 7) included questions about the patient’s use of health services following discharge from the acute hospital and was used to cost subsequent use of health services. The cover of the questionnaires included a ‘do not wish to participate’ tick box.

FIGURE 3.

Patient follow-up process at 90 days and at 1 year. HSCIC, Health and Social Care Information Centre.

If no response was received after 2 weeks, a reminder letter was sent with another questionnaire pack. If no response was received after a further 2 weeks, the patient was telephoned, if his or her contact details were available. Telephone calls were made at various times from Monday to Friday between 08.30 and 20.30 to maximise the chances of contacting the patient. Patients who were successfully contacted by telephone were asked if they had received the questionnaire pack and were invited to complete the questionnaires over the telephone, if this was convenient. In addition, patients were reminded about completing the questionnaire when they attended hospital follow-up appointments.

Follow-up ended on receipt of a completed (or blank) questionnaire; on receipt of a questionnaire with a ticked ‘do not wish to participate’ box; on notification to the ICNARC CTU by telephone or e-mail that the patient wished to withdraw from the trial; or if there was no response to the telephone follow-up. For questionnaire packs returned indicating that the recipient was not known at the address, the contact details for the patient were checked with the recruiting hospital and/or general practitioner.

For patients who were identified as being either a hospital inpatient or resident in a care home or rehabilitation centre, the relevant institution was contacted to establish the status of the patient and the most appropriate way to proceed with follow-up. If the patient had the mental capacity to consent but required assistance in reading and/or completing the questionnaire, health-care professionals usually assisted the patient. For patients who lacked the mental capacity to consent, institutions advised on the most appropriate person to contact to complete the questionnaires.

If patients were identified as having no fixed abode but were registered with a general practitioner or had regular contact with a homeless shelter, the questionnaire pack was sent to be given (when appropriate) to them at their next appointment or visit.

Data linkage with the Case Mix Programme

The linkage of patient identifiable trial data to the ICNARC Case Mix Programme database provided information on subsequent admission to adult, general, critical care following discharge from the acute hospital. 32

Data for the CMP are collected by trained data collectors to precise rules and definitions. The data then undergo extensive local and central validation for completeness, illogicalities and inconsistencies prior to pooling.

Data management

Data management was an ongoing process. Data entered by sites onto the electronic case report form were monitored and checked throughout the recruitment period to ensure that they were as complete and accurate as possible.

Two levels of data validation were incorporated into the electronic case report form. The first was to prevent obviously erroneous data from being entered, for example entering a date of birth that occurred after the date of randomisation. The second level involved checks for data completeness and any unusual data entered, for example a physiological variable, such as blood pressure, that was outside the pre-defined range. Site staff could generate data validation reports, listing all outstanding data queries, at any time via the electronic case report form. The site PI was responsible for ensuring that all data queries were resolved. Ongoing data entry and validation at sites were closely monitored by the data manager (JT) and any concerns were raised with the site PI.

The contact details for patients and their general practitioners (name and postal address) were checked weekly for completeness to avoid unnecessary delays in sending out questionnaire packs at 90 days and at 1 year.

Adherence to the trial protocol was closely monitored, including adherence to all elements of the EGDT resuscitation protocol. Any queries relating to adherence were generated in a separate report which was sent to the site PI.

Data received from completed European Quality of Life-5 Dimensions (EQ-5D) and Health Services Questionnaires were entered centrally into a secure database at the ICNARC CTU following a standard operating procedure. All identifiable information, such as names (e.g. of patients, family members or hospital staff members), was removed. All queries relating to data entry were reviewed by two members of the TMG (SH/PM) and any disagreement was reviewed and discussed with a third member (KR).

To ensure that data were entered accurately, all questionnaire data entered into the database were cross-checked by a second member of the CTU team. Any errors found were logged and corrected on the database.

Sample size

Estimates for baseline mortality in the usual-resuscitation group were based on the ICNARC Case Mix Programme database. 32 Between 1 January 2005 and 31 December 2006, there were 24,155 patients admitted to 156 participating adult general ICUs direct from the ED. Of these, 6671 (28%) met at least two SIRS criteria during the first 24 hours following ICU admission and had evidence of infection. Acute hospital mortality for these patients was 35%. To allow for additional deaths after discharge from hospital and before 90 days, sample size calculations were based on an anticipated mortality at 90 days of 40% in the usual-resuscitation group. To achieve 80% power to detect a 20% relative reduction in mortality at 90 days (corresponding to an 8% absolute reduction) from 40% to 32% associated with EGDT compared with usual resuscitation (p < 0.05, two-sided) required a sample size of 589 patients per treatment group (Stata/SE version 10.1, StataCorp LP, College Station, TX, USA). Allowing for 6% of patients refusing consent to follow-up (in the PAC-Man trial, 2% of patients refused consent after randomisation36) or being lost to follow-up before 90 days, our aim was to recruit 630 patients per group (1260 patients in total). This sample size provided > 99% power to detect an absolute risk reduction of the magnitude observed in the Rivers et al. trial (i.e. 16%). 6

Interim analysis

Unblinded comparative data on recruitment, withdrawal, adherence with the trial protocol and serious adverse events were regularly reviewed by the DMEC. Without specific analysis of the primary outcome, the DMEC reviewed data from the first 50 trial participants and continued to review data at least 6-monthly to assess potential safety issues and to review adherence with the trial protocol. A single planned formal interim analysis was performed once 90-day outcome data from the first 500 patients enrolled were available. A Haybittle–Peto stopping rule (p < 0.001) was used to guide recommendations for early termination owing to harm.

Analysis principles

All analyses were based on the intention-to-treat principle. Patients were analysed according to the treatment group they were randomised to, irrespective of whether or not the allocated treatment was received (i.e. regardless of whether or not they adhered to the EGDT algorithm). All tests were two-sided with significance levels set at p < 0.05 and with no adjustment for multiplicity. All a priori subgroup analyses were carried out irrespective of whether or not there was strong evidence of a treatment effect associated with the primary outcome. As missing data for the clinical effectiveness primary outcome were anticipated to be minimal, a sensitivity approach was taken when the primary outcome was missing (see Secondary analyses of the primary outcome). Missing data for the cost-effectiveness analysis, as well as missing baseline data for adjusted analysis of clinical outcomes, were handled by multiple imputation.

Multiple imputation

Missing data in baseline covariates, resource use and health-related quality of life variables at 90 days and 1 year were handled with multivariate imputation by chained equations. 37 Under this approach each variable was imputed conditional on fully observed baseline variables such as age, sex, past medical history, site of sepsis, SOFA score, MEDS score, admitted from nursing home, length of stay in critical care and general medical wards up to 90 days and 1 year, and all other imputed variables. Patients who were eligible for 90-day follow-up (i.e. alive at 90 days) but did not return or fully complete the EQ-5D questionnaire administered at 90 days, had their EQ-5D utility scores imputed from those survivors who did fully complete the questionnaire. Similarly, for those eligible patients who did not return the Health Services Questionnaire, information on the use of outpatient services up to 90 days following randomisation, was imputed from those patients who did complete this questionnaire. In the same way, patients who were eligible for 1-year follow-up but did not return or fully complete the EQ-5D questionnaire or the Health Services Questionnaire administered at 1 year also had their information imputed from those survivors who did fully complete the questionnaire. When addressing the missing data, multiple imputation assumes that the data are missing at random conditional on the observed data.

The same multiple imputation approach was used to address the administrative censoring, which applied to the total costs, vital status and quality of life at 1 year for patients randomised after 12 November 2013. In this case it was assumed that the data were censored completely at random, which was plausible as the censoring was administrative, that is it is unlikely that there would be systematic differences between those whose end points (cost, vital status and quality of life) were observed and those who were censored. One-year cost and quality-of-life end points were conditional on survival status; as such, the imputation was conducted in 2 stages. In the first stage, imputation models were specified for mortality at 1 year according to baseline covariates and auxiliary variables, including duration of the initial inpatient stay, and costs at 90 days. In the second stage, for each of the imputed data sets from stage 1, imputation models were specified for costs and quality of life at 1 year for those patients who were missing these but were known to be alive at 1 year, or were predicted to be alive by the first-stage imputation model. These imputation models included those variables in the first-stage imputation model but also information on costs and quality of life at 1 year for those individuals for whom this end point was observed. Each of the resultant estimates was combined with Rubin’s rules, which recognise uncertainty both within and between imputations. All multiple imputation models were implemented in the statistical package R (The R Foundation for Statistical Computing, Vienna, Austria).

Statistical analysis: clinical effectiveness

Statistical analyses were conducted according to a pre-specified, published statistical analysis plan. 38 The final analyses were conducted using Stata/SE version 13.0.

Cost-effectiveness analysis

A full cost-effectiveness analysis was undertaken to assess which treatment strategy, EGDT or usual resuscitation, was more cost-effective. This analysis assessed whether or not any intervention costs associated with EGDT were offset by any subsequent reduction in morbidity costs, for example from reduced use of critical care, and whether there were improvements in either mortality or health-related quality of life. The cost-effectiveness analysis was reported for three time periods: randomisation to 90 days, randomisation to 1 year and lifetime. For each time period the cost-effectiveness analysis took a health and personal health services perspective,45 using information on health-related quality of life collected at 90-day and 1-year follow-up, combined with information on vital status, to report QALYs. Each QALY was valued using the National Institute for Health and Care Excellence (NICE)-recommended threshold of willingness to pay for a QALY gain (£20,000), in conjunction with the costs of each treatment strategy to report the INBs of EGDT versus usual resuscitation.

The primary objective of the cost-effectiveness analysis was to compare incremental cost-effectiveness at 1 year between the treatment groups. There were also a number of secondary objectives:

• to compare health-related quality of life at 90 days and 1 year between the treatment groups

• to compare resource use and costs at 90 days and 1 year between the treatment groups

• to estimate the lifetime incremental cost-effectiveness between the treatment groups.

The main assumptions of the cost-effectiveness analysis were subjected to extensive sensitivity analyses.

Participants: sites

Expressions of interests were received from 83 NHS hospitals in the UK. A total of 57 hospitals in England obtained local NHS permissions and opened to recruitment between 15 February 2011 and 25 March 2013. Forty-four sites were opened within the first 9 months of the trial’s opening on 15 February 2011 (Figure 4).

FIGURE 4.

Sites open to recruitment during the trial recruitment period.

The rate at which sites were opened for the ProMISe trial was higher than for ProCESS and ARISE. Within 12 months of the trial opening, 47 sites had been opened for ProMISe, compared with 20 each for ProCESS and ARISE (Figure 5).

FIGURE 5.

Recruitment of sites for ProMISe compared with ProCESS and ARISE.

The median time from local NHS permission to the trial opening at sites (i.e. start of screening) was 83 (IQR 51–151) days (Figure 6). Reasons for delays in opening were issues related to the confirmation of NHS support costs from the CLRN and delays in the local set-up of the trial, for example training staff.

FIGURE 6.

Time (in days) from local NHS permission to start of screening.

Overall, sites participated in the ProMISe trial for a median of 30 (IQR 19–35) months. Of the 57 sites that opened, seven were closed early because of poor recruitment (one site recruited no patients), two were closed because of insufficient resources locally for screening and recruitment and two were closed for other local logistical reasons. As part of the staggered close-down of the trial, nine sites were closed in October 2013 and a further eight were closed in April 2014, with 29 sites remaining open until the end of recruitment in July 2014 (Figure 7).

FIGURE 7.

Duration of participation of sites.

There were eight sites that had at least one period when screening and recruitment was suspended either because of insufficient resources (n = 6) or to enable new staff to be trained in delivery of the trial protocol (n = 2) (see Figure 7).

Participants: patients

In total, 6192 patients were screened between 15 February 2011 and 24 July 2014. Of these, 2415 (39.0%) met one or more exclusion criteria. There were 2517 (40.6%) patients who, although eligible for inclusion in the trial, were not recruited. The most frequently reported reason for not recruiting eligible patients was logistical issues, mainly no research staff being available, for example if the patient presented at the ED outside usual office hours. Other reported reasons included refusal by the treating clinician to recruit the patient; the patient declined to take part; or the patient was identified as eligible for the trial outside the 2-hour window for obtaining consent and randomising (Figure 8).

FIGURE 8.

Screening, randomisation and follow-up. AIDS, acquired immune deficiency syndrome.

The 1260 (20.3%) patients were recruited between 16 February 2011 and 24 July 2014, with 630 randomised to the EGDT group and 630 randomised to the usual-resuscitation group (Figure 9). There was variation across the 56 sites in the rate of recruitment (Figure 10), the overall median recruitment rate being 0.15 (IQR 0.10–0.22) patients per site per week, with a highest recruitment rate of 0.48 patients per week. Patients were recruited over a relatively shorter time period than in ProCESS and ARISE (Figure 11). Manual randomisation was required for three patients.

FIGURE 9.

Patient recruitment.

FIGURE 10.

Patient recruitment rate.

FIGURE 11.

Recruitment of patients for ProMISe compared with ProCESS and ARISE.

Patients were generally recruited into ProMISe during weekdays (Monday to Friday) and during usual office hours (Figure 12); most of the recruiting sites reported having insufficient resources to enable screening and recruitment at weekends and outside usual office hours.

FIGURE 12.

Randomisation by (a) day of the week and (b) time of day.

Almost half of patients provided informed consent prior to randomisation (n = 624, 49.5%). For the remaining patients, agreement was obtained from a personal (34.8%) or professional (2.9%) consultee or from an independent clinician using emergency consent (12.8%) (Table 6). Four patients withdrew from the trial, requesting the removal of all of their data from the analysis, and five patients were ineligible and recruited in error, resulting in data on 1251 for initial analysis (n = 625 EGDT; n = 626 usual resuscitation). Eight patients subsequently withdrew before 90 days, resulting in 1243 patients for analysis of outcomes (n = 623, 99.7% EGDT; n = 620, 99.0% usual resuscitation). Owing to the truncation of follow-up, 127 patients were not followed up at 1 year (n = 65 EGDT; n = 62 usual resuscitation) (see Figure 8 and Table 6). Follow-up was completed on 30 October 2014.

Adherence to the protocol

Most patients randomised to EGDT (n = 545, 87.3%) had timely insertion of a PreSep^TM central venous oximetry catheter (Table 9). Two (0.3%) patients in the usual-resuscitation group had one inserted in error but these were not used for monitoring ScvO₂. The reasons for failure of insertion in the EGDT group were that patients were determined either to no longer meet inclusion criteria or to now meet exclusion criteria (n = 22); process of care (lack of equipment, staff, beds, communication, error; n = 20); technical or patient difficulties (n = 18); clinician decision (n = 9); refusal by the patient (without withdrawal from the trial; n = 5); and death before insertion (n = 2). No reason was provided for four patients. The mean first ScvO₂ value recorded after catheterisation (at hour 1) was 70% (SD 12%). Standard central venous catheters (not mandated) were inserted in 318 (50.9%) of the usual-resuscitation group and measurement of ScvO₂ from aspirated blood samples occurred in six patients. Arterial catheters (not mandated) were inserted in the majority of patients (n = 462, 74.2% EGDT; n = 389, 62.2% usual resuscitation). EGDT was stopped prematurely in 21 patients (median time to cessation, 3 hours) owing to withdrawal of active treatment (n = 9); patient no longer considered to be septic (n = 5); error (n = 3); transfer to operating theatre (n = 1); and refusal by the patient (n = 1). No reason was provided for two patients. Of the patients who died within 6 hours (n = 17 EGDT; n = 18 usual resuscitation), five in the EGDT group and six in the usual-resuscitation group had active treatment withdrawn prior to death.

Figure 13 shows the adherence with each element of the EGDT protocol across the 6-hour intervention period. The bars plotted on each hour report the percentage of patients (of those meeting the previous targets) who did not meet or met each physiological target at that hour, or for whom the relevant physiological value was not recorded (within 15 minutes either side of the hour). The bars plotted between each hour report the percentage of patients (of those who did not meet the physiological target at the previous hour) that received the associated action either during the intervention period or after the intervention period, or who no longer required the action as the target was subsequently met. Adherence with the protocol was generally good, particularly for delivery of fluid and vasopressors, but there were delays in obtaining packed red blood cells and dobutamine, such that the ScvO₂ value had often resolved spontaneously before these were delivered.

FIGURE 13.

Adherence to the EGDT protocol. (a) Target: CVP; action: fluids; (b) target: MAP/SBP; action: vasopressors; and (c) target: ScVO₂; action: PRBC transfusion (L) or dobutamine (R). Numbers at the foot of each bar represent the denominator for the percentages in that bar. Each target is considered sequentially (i.e. those meeting the target for CVP become the denominator for MAP/SBP and those meeting the target for MAP/SBP become the denominator for ScvO₂). Of the two smaller bars, (L) denotes the left-hand bar and (R) denotes the right-hand bar. Thirty-two patients who no longer met eligibility criteria or declined the intervention were excluded from the evaluation of adherence. CVP, central venous pressure; MAP, mean arterial pressure; PRBC, packed red blood cells; SBP, systolic blood pressure.

Figure 14 reports the protocol adherence compared with ProCESS and ARISE, according to the adherence algorithms reported in their respective publications. 23,24 For the comparison with ProCESS, adherence was assessed at the end of the 6-hour intervention period only. Patients who died, were discharged or were withdrawn from the intervention prior to 6 hours were excluded (47/625, 7.5%, for ProMISe; 35/439, 8.0%, for ProCESS). Overall adherence among the evaluable patients was similar for the two trials (85.6% for ProMISe vs. 88.1% for ProCESS). The greatest difference was on failure to insert a central venous catheter with ScvO₂ monitoring capability; on all other measures, non-adherence for ProMISe was lower than for ProCESS. For the comparison with ARISE, adherence was assessed hourly as the percentage of patients meeting each physiological target (of those for whom the relevant physiological value was recorded) and, for those with physiological values recorded at 2 consecutive hours, the percentage who either met the target at the start or the end of the hour or received the associated action during the hour. The percentage of patients meeting the physiological targets at each hour, for ProMISe compared with ARISE, was similar for central venous pressure, higher for mean arterial pressure/systolic blood pressure and lower for ScvO₂. When reported as the percentage either meeting the target or receiving the associated action, adherence was extremely high (and similar to ARISE) for both receipt of intravenous fluids and receipt of vasopressors, but somewhat lower for receipt of packed red blood cells or dobutamine (although this did reach a level of 89% by the final hour of the intervention period).

FIGURE 14.

Protocol adherence compared with ProCESS (a) and ARISE [(b), CVP; (c) i.v. fluids; (d) MAP or SBP; (e) vasopressors; (f) ScvO₂; and (g) PRBC and/or dobutamine]. For the comparison with ProCESS, the numbers of evaluable patients were 578 (ProMISe) and 404 (ProCESS). For the comparison with ARISE, numbers at the foot of each bar represent the denominator for the percentages in that bar. CVP, central venous pressure; i.v., intravenous; MAP, mean arterial pressure; PRBC, packed red blood cells; SBP, systolic blood pressure.

Delivery of care by treatment group

During the 6-hour intervention period, patients randomised to EGDT received more intravenous fluid than those randomised to usual resuscitation (see Table 9). Hourly fluid volume decreased over the 6 hours but patients in the usual-resuscitation group received a greater volume initially (Figure 15). In both groups, crystalloid was used more frequently than colloid. More patients in the EGDT group received vasopressors and dobutamine. Although more patients in the EGDT group received packed red blood cell transfusions, larger volumes were transfused in the usual-resuscitation group. Administration of platelets and fresh-frozen plasma was similar, although volumes of both were higher in the EGDT group. Physiological values normalised slightly over the 6-hour intervention period in both groups (Figure 16). After 6 hours, central venous pressure, mean arterial pressure, systolic blood pressure and haemoglobin, where measured (with greater frequency in the EGDT group), were similar (Table 10).

FIGURE 15.

Interventions delivered during the intervention period. The values reported are the mean and 95% CI (volume of i.v. fluid) or percentage of patients receiving the intervention (all other panels) during each hour of the intervention period. The numbers at the foot of each bar represent the denominators for the means/percentages in that bar. (a) Volume of i.v. fluid; (b) vasopressors; (c) PRBC transfusion; (d) dobutamine; (e) sedatives; (f) neuromuscular blocking agents; and (g) mechanical ventilation. i.v., intravenous; PRBC, packed red blood cells.

FIGURE 16.

Physiology during the intervention period. (a) CVP; (b) MAP; (c) SBP; and (d) ScvO₂. The values reported are the mean plus/minus SD for values recorded within 15 minutes either side of the specified time point. For the EGDT group, first recorded values after insertion of the PreSep^TM central venous oximetry catheter are at hour 1. ScvO₂ measurements are not reported for the usual-resuscitation group owing to very small numbers of patients for whom these were recorded. CVP, central venous pressure; MAP, mean arterial pressure; SBP, systolic blood pressure.

Between 6 and 72 hours, use of intravenous fluids was similar but usual-resuscitation patients received higher volumes (Table 11). Intravenous colloid use was higher in the EGDT group but volumes were similar in the two groups, intravenous crystalloid use was similar but volumes were higher in the usual-resuscitation group and use of packed red blood cell transfusions was higher in the EGDT group but the volumes delivered were higher in the usual-resuscitation group. Although the use of platelets and fresh-frozen plasma was similar, the volume of platelets transfused was higher in the EGDT group and the volume of fresh-frozen plasma was higher in the usual-resuscitation group. Vasopressor and dobutamine use remained higher in the EGDT group. At 72 hours, physiological, biochemistry and SOFA values were similar (Table 12).

Primary outcome: clinical effectiveness

At 90 days following randomisation, 184 (29.5%) patients randomised to EGDT had died, compared with 181 (29.2%) patients randomised to usual resuscitation, corresponding to an absolute risk reduction of −0.3 (95% CI −5.4 to 4.7, p = 0.90) and a relative risk of 1.01 (95% CI 0.85 to 1.20; Table 13). This difference remained non-significant after adjustment for baseline characteristics (adjusted odds ratio 0.95, 95% CI 0.74 to 1.24; p = 0.73; unadjusted odds ratio 1.02, 95% CI 0.80 to 1.30).

Secondary outcomes: clinical effectiveness

For patients in the EGDT group, both the mean SOFA score at 6 hours (p < 0.001) and the proportion receiving advanced cardiovascular support (p = 0.026) were significantly higher than for patients in the usual-resuscitation group. The median total length of stay in critical care was significantly longer for patients in the EGDT group than the usual-resuscitation group (p = 0.006). There were no significant differences between the groups in any of the other secondary outcomes, including duration of survival (log-rank test, p = 0.63; Cox proportional hazards model, adjusted hazard ratio 0.94, 95% CI 0.79 to 1.11; p = 0.46) (Figure 17 and see Table 13).

FIGURE 17.

Kaplan–Meier curves for survival to (a) 90 days and (b) 1 year following randomisation.

Safety monitoring

Thirty (4.8%) patients in the EGDT group and 26 (4.2%) patients in the usual-resuscitation group experienced one or more serious adverse events within 30 days following randomisation (relative risk 1.16, 95% CI 0.69 to 1.93; p = 0.58; Table 14). The most commonly reported serious adverse events were pulmonary oedema (four patients in the EGDT group and seven patients in the usual-resuscitation group) and myocardial ischaemia (seven patients in the EGDT group and four patients in the usual-resuscitation group). Three serious adverse events associated with EGDT (two pulmonary oedema and one arrhythmia) were reported as probably related to the intervention, compared with four (in three patients) serious adverse events (two pneumothoraces and one pulmonary oedema probably related, and one ventricular fibrillation definitely related) reported as associated with usual resuscitation.

Subgroup analyses of the primary outcome

There was no difference in the effect of EGDT on the primary outcome of all-cause mortality at 90 days following randomisation according to the pre-specified subgroups defined by the degree of protocolised care/usual resuscitation, age, MEDS score, SOFA score and time from ED presentation to randomisation (p-values for test of interaction 0.39 to 0.72; Figure 18).

FIGURE 18.

Subgroup analyses of the primary outcome. p-values are for tests of interaction. The x-axis is presented on a log scale. The solid line represents no difference between the groups and the dashed line represents the overall effect estimate (adjusted odds ratio).

Secondary analyses of the primary outcome

Eight patients were missing the primary outcome of all-cause mortality at 90 days following randomisation (two in the EGDT group and six in the usual-resuscitation group). The effect of missing data on the results was minimal. Sensitivity analyses making alternative extreme assumptions for the missing outcomes reported relative risks of 0.99 and 1.03 (Table 15). Although the learning curve analysis suggested increased odds of mortality for the first patient randomised to EGDT in each site, this effect was not significant (p = 0.56; Figure 19 and see Table 15). Adjusting for non-adherence to the EGDT protocol resulted in minimal change to the relative risk (see Table 15).

FIGURE 19.

Learning curve for delivery of EGDT. The y-axis is presented on a log scale.

Comparison with other early goal-directed therapy studies

Comparing the usual-resuscitation group for ProMISe with that from the original randomised controlled trial of EGDT by Rivers et al. ,6 patients in ProMISe were randomised, on average, slightly later than those in the Rivers et al. trial (Table 16). Demographics were similar. Patients in ProMISe had slightly lower blood pressure, blood lactate measurements and APACHE II scores. Patients in the usual-resuscitation group in the Rivers et al. 6 trial received considerably more fluid and packed red blood cell transfusions, and were more likely to be mechanically ventilated; however, those in ProMISe received more vasopressors and dobutamine. Hospital mortality was substantially higher in the Rivers et al. 6 trial than in ProMISe.

Comparing the usual-resuscitation group for ProMISe with those from ProCESS23 and ARISE,24 time from ED presentation to randomisation was similar but patients in ProMISe had a shorter length of stay in the ED than those in ARISE (Table 17). Demographics were similar. The mean volume of fluid received prior to randomisation was higher for ARISE than for ProCESS or ProMISe. Blood pressure and blood lactate measurements were similar for ProMISe and ProCESS, but mean arterial pressure was slightly higher and blood lactate was slightly lower for ARISE. This lower severity of illness for patients in ARISE was also reflected in APACHE II scores, which were lowest for ARISE and highest for ProCESS. A greater proportion of patients in ProMISe met both the refractory hypotension and the hyperlactataemia inclusion criteria. To explore the hypothesis that the combination of both refractory hypotension and hyperlactataemia was associated with higher mortality than either alone, we examined data from the Case Mix Programme, the national clinical audit for adult critical care. Among 12,004 patients admitted to 183 adult general ICUs in England between February 2011 and June 2014 (the recruitment period of ProMISe) direct from the ED with infection, meeting two or more SIRS criteria during the first 24 hours following admission, and with hypotension (lowest systolic blood pressure < 90 mmHg) and/or hyperlactataemia (highest blood lactate ≥ 4 mmol/l), acute hospital mortality was around 30% for patients meeting a single criterion but almost double for patients meeting both the refractory hypotension and the hyperlactataemia criteria (Table 18).

During the intervention period, the mean volume of fluid received by patients in the usual-resuscitation group was highest for ProCESS and lowest for ARISE (see Table 17); however, a greater proportion of patients in ARISE received vasopressors. Dobutamine use was highest in ProMISe and lowest in ProCESS and the proportions of usual-resuscitation group patients receiving packed red blood cell transfusions in ProCESS and ARISE were approximately double that in ProMISe. In all three trials, around 20% of patients received mechanical ventilation, and the central venous catheter insertion rates varied from 51% in ProMISe to 62% in ARISE. Reflecting the pattern seen in severity of illness scores, 90-day mortality for usual-resuscitation group patients in ProCESS was slightly higher than in ProMISe (34% vs. 29%), whereas for those in ARISE it was substantially lower (19%). Mortality in ARISE was similarly lower at other comparable time points. Of patients discharged alive from hospital in both ProMISe and ARISE, approximately 80% were discharged home, compared with only just over 50% of similar patients in ProCESS.

Cost-effectiveness at 90 days following randomisation

Cost-effectiveness at 90 days

The incremental QALY gain for EGDT versus usual resuscitation was negative, but with 95% CIs that included zero (see Table 23). The average costs were higher for the EGDT group, but this difference was not statistically significant. The INB for EGDT versus usual resuscitation was negative at −£1000 (95% CI −£2720 to £720; see Table 23).

When the uncertainty in the incremental costs and QALYs is represented on the cost-effectiveness plane, the majority of the points are in those quadrants that show EGDT has, on average, higher costs (Figure 20). The probability that EGDT is more cost-effective than usual resuscitation, given the data, is never greater than 20%, irrespective of how much society is willing to pay for a QALY gain (Figure 21).

FIGURE 20.

Uncertainty in the mean costs (£) and QALY differences at 90 days and their distribution for EGDT vs. usual resuscitation.

FIGURE 21.

Cost-effectiveness acceptability curve reporting the probability that EGDT is cost-effective at 90 days at alternative willingness to pay.

The estimated INB were similar for the scenarios considered in the sensitivity analyses (Figure 22). For example, the INB remains around −£1000 whether EGDT is provided in the ED or in critical care. Similarly, excluding readmissions that were reported from responses to the Health Services Questionnaire, to avoid any risk of double counting, had only a small impact on the mean INB (−£800 vs. −£1000).

FIGURE 22.

Sensitivity analyses for the incremental net benefit at 90 days following randomisation according to alternative assumptions, compared with the base case. The vertical dashed line indicates incremental net benefit in the base-case analysis. The solid vertical line indicates no difference in net monetary benefits between the treatment groups.

The results of the subgroup analyses are presented in Table 24, and show that the incremental QALYs were similar across all subgroups. Although there were some subgroups for which the incremental costs of EGDT were negative and hence the INBs were positive, the statistical uncertainty surrounding these findings was high. Hence for each subgroup, as for the overall results, the 95% CIs around the INB included zero. Adjusting for adherence to the EGDT protocol decreased the INB to −£1438, but the 95% CI still included zero (see Table 24).

TABLE 24 - Cost-effectiveness at 90 days: subgroup and secondary analyses

Subgroup	Incremental cost (£) (95% CI)	Incremental QALYs (95% CI)	Incremental net benefit (£) (95% CI)
Degree of protocolised resuscitation in usual-resuscitation group
Low	806 (−1213 to 2825)	0.002 (−0.004 to 0.009)	−765 (−2789 to 1259)
High	1655 (−1822 to 5131)	−0.005 (−0.017 to 0.006)	−1764 (−5247 to 1719)
Age (years)
18–56	3253 (−155 to 6662)	−0.001 (−0.012 to 0.011)	−3265 (−6684 to 154)
57–67	398 (−3021 to 3818)	0.003 (−0.008 to 0.014)	−329 (−3758 to 3100)
68–77	−1511 (−4884 to 1862)	−0.003 (−0.015 to 0.008)	1444 (−1943 to 4831)
78–95	2359 (−1112 to 5830)	0.003 (−0.008 to 0.015)	−2296 (−5777 to 1185)
MEDS score
0–4	2129 (−2644 to 6902)	0.002 (−0.014 to 0.018)	−2089 (−6864 to 2686)
5–6	2700 (−815 to 6215)	0.002 (−0.009 to 0.014)	−2652 (−6173 to 869)
7–9	−250 (−3308 to 2807)	0.005 (−0.005 to 0.015)	351 (−2715 to 3417)
10–20	196 (−2934 to 3326)	−0.009 (−0.019 to 0.001)	−377 (−3514 to 2760)
SOFA score
0–2	1947 (−1482 to 5375)	0.002 (−0.009 to 0.014)	−1898 (−5327 to 1531)
3–4	623 (−2351 to 3598)	0.001 (−0.009 to 0.011)	−603 (−3580 to 2374)
5	−1506 (−5705 to 2692)	−0.007 (−0.021 to 0.006)	1359 (−2848 to 5566)
6–14	2658 (−701 to 6016)	−0.004 (−0.014 to 0.007)	−2736 (−6099 to 627)
Time from ED presentation to randomisation (hours)
0.2–1.8	1291 (−2114 to 4697)	−0.002 (−0.013 to 0.009)	−1322 (−4734 to 2090)
1.8–2.5	2849 (−515 to 6214)	0.004 (−0.007 to 0.015)	−2776 (−6147 to 595)
2.5–3.5	1123 (−2344 to 4590)	−0.003 (−0.014 to 0.009)	−1179 (−4655 to 2297)
3.5 +	−1453 (−4882 to 1976)	−0.001 (−0.012 to 0.01)	1426 (−2008 to 4860)
Adherence adjusted analysis	1423 (−1042 to 3888)	−0.001 (−0.009 to 0.007)	−1438 (−3909 to 1033)

Cost-effectiveness at 1 year following randomisation (primary outcome)

Cost-effectiveness at 1 year

At 1 year following randomisation, a slightly higher proportion of patients in the EGDT group were alive than in the usual-resuscitation group (see Secondary outcomes: clinical effectiveness). The net effect of the EGDT group having higher survival but a lower average EQ-5D utility score resulted in similar 1-year QALYs between the treatment groups (see Table 29). The mean total cost was higher in the EGDT group, with an incremental cost of £764 (95% CI −£1402 to £2930). Hence the INB for EGDT versus usual resuscitation was negative at −£725 (95% CI −£3000 to £1550). The distribution of incremental costs and QALYs in the cost-effectiveness plane is shown in Figure 23.

FIGURE 23.

Uncertainty in the mean costs (£) and QALY differences at 1 year and their distribution for EGDT vs. usual resuscitation.

The cost-effectiveness acceptability curve (Figure 24) shows that at 1 year the probability that EGDT is more cost-effective than usual resuscitation, given the data, is below 30% at the £20,000 willingness-to-pay threshold stipulated by NICE.

FIGURE 24.

Cost-effectiveness acceptability curve reporting the probability that EGDT is cost-effective at 1 year at alternative levels of willingness to pay.

The estimated INBs were similar for the scenarios considered in the sensitivity analyses (Figure 25). This shows that the base-case results are robust to alternative assumptions.

FIGURE 25.

Sensitivity analyses for the incremental net benefit at 1 year following randomisation according to alternative assumptions, compared with the base case. The vertical dashed line indicates incremental net benefit in the base-case analysis. The solid vertical line indicates no difference in net monetary benefits between the treatment groups.

The estimated INBs were similar across all pre-specified subgroups (Table 30). Although there were some subgroups for whom EGDT was cost-saving and hence their INBs were positive, there was high statistical uncertainty around surrounding these findings. Hence for each subgroup, as for the overall results, there is high statistical uncertainty surrounding INBs.

TABLE 30 - Cost-effectiveness at 1 year: subgroup and secondary analyses

Subgroup	Incremental cost (£) (95% CI)	Incremental QALYs (95% CI)	Incremental net benefit (£) (95% CI)
Degree of protocolised resuscitation in usual-resuscitation group
Low	801 (−1718 to 3319)	0.014 (−0.03 to 0.058)	−525 (−3172 to 2122)
High	739 (−3562 to 5040)	−0.017 (−0.093 to 0.059)	−1084 (−5578 to 3410)
Age (years)
18–56	3643 (−626 to 7913)	0.011 (−0.064 to 0.086)	−3422 (−7932 to 1088)
57–67	257 (−3977 to 4490)	0.025 (−0.049 to 0.098)	238 (−4234 to 4710)
68–77	−1924 (−6101 to 2252)	−0.028 (−0.104 to 0.047)	1357 (−3042 to 5756)
78–95	2038 (−2227 to 6302)	0.041 (−0.035 to 0.116)	−1226 (−5720 to 3268)
MEDS score
0–4	1972 (−3966 to 7909)	0.015 (−0.088 to 0.118)	−1670 (−7866 to 4526)
5–6	2401 (−1931 to 6733)	0.008 (−0.068 to 0.084)	−2241 (−6789 to 2307)
7–9	−909 (−4684 to 2866)	0.045 (−0.022 to 0.112)	1806 (−2160 to 5772)
10–20	615 (−3305 to 4534)	−0.047 (−0.114 to 0.021)	−1551 (−5636 to 2534)
SOFA score
0–2	1579 (−2755 to 5913)	0.020 (−0.055 to 0.095)	−1183 (−5694 to 3328)
3–4	696 (−3033 to 4425)	0.007 (−0.059 to 0.072)	−566 (−4450 to 3318)
5	−2836 (−8057 to 2384)	−0.036 (−0.127 to 0.056)	2118 (−3332 to 7568)
6–14	2769 (−1401 to 6939)	−0.016 (−0.087 to 0.055)	−3097 (−7446 to 1252)
Time from ED presentation to randomisation (hours)
0.2–1.8	1739 (−2507 to 5985)	−0.010 (−0.085 to 0.065)	−1940 (−6387 to 2507)
1.8–2.5	3576 (−633 to 7786)	0.032 (−0.043 to 0.107)	−2934 (−7350 to 1482)
2.5–3.5	1111 (−3225 to 5446)	−0.005 (−0.082 to 0.071)	−1215 (−5785 to 3355)
3.5 +	−3535 (−7762 to 692)	−0.007 (−0.083 to 0.068)	3388 (−1037 to 7813)
Adherence adjusted analysis	1099 (−2013 to 4211)	0.003 (−0.051 to 0.057)	−1042 (−4312 to 2228)

Lifetime incremental cost-effectiveness

Long-term survival

The Kaplan–Meier survival curves show that when the time horizon was extended beyond 1 year, for those with survival data available, the probability of survival remained similar between treatment groups (Figure 26).

FIGURE 26.

Kaplan–Meier survival curves for long-term survival.

To calculate QALYs over 20 years, the long-term survival for each patient was estimated by combining the observed survival for each patient up to 1 year with their predicted survival from 1 year to 20 years. We compared alternative parametric extrapolation approaches to predict longer-term survival of patients recruited to ProMISe. Figure 27 considers alternative parametric extrapolations for survival, using the observed survival data after day 30. The survival data were pooled across the treatment groups, given that there was no evidence of an effect of treatment group on survival. Of the alternative survival functions, log-normal appeared to fit the observed data best in that it reported the lowest Akaike information criteria and Bayesian information criteria (Table 31). However, the Gompertz function offered the most plausible projections of future survival (see Figure 27 and Table 31), in that the levels of survival remained constant over time from 5 years following randomisation onwards. The parametric models predicted excess mortality in patients recruited to ProMISe compared with the age-/sex-matched general population. In the base case, we applied death rates according to the most plausible parametric model (i.e. Gompertz) between year 2 and year 15. At year 16, predicted survival overlaps with age-/sex-matched survival and, therefore, from year 16 onwards we applied age-/sex-matched general population death rates.

FIGURE 27.

Comparison of alternative parametric extrapolations of survival.

TABLE 31 - Fit of alternative parametric survival functions applied to the ProMISe data set after day 30

AIC, Akaike information criterion; BIC, Bayesian information criterion.

Distribution	AIC	BIC
Gompertz	1658.0	1687.1
Log-normal	1642.4	1671.5
Logistic	1644.5	1673.5
Weibull	1645.8	1674.8
Exponential	1691.0	1715.2

Lifetime incremental cost-effectiveness

Table 32 presents the resultant lifetime QALYs, lifetime costs and INB according to the base-case assumptions. Overall, at the NICE-stipulated threshold of £20,000 per QALY the INB was negative, but with a wide 95% CI that included zero. The distribution of incremental costs and QALYs in the cost-effectiveness plane is shown in Figure 28.

TABLE 32 - Lifetime cost-effectiveness

Incremental net benefit is calculated according to NICE methods guidance, by multiplying the mean QALY gain (or loss) by £20,000, and subtracting from this the incremental cost.

Values are mean (SD), unless stated otherwise. Results are reported after applying multiple imputation to handle missing data.

End point	EGDT (n = 625)	Usual resuscitation (n = 626)	Incremental effect (unadjusted), mean (95% CI)	p-value
QALYs	4.584 (3.546)	4.582 (3.720)	0.002 (−0.411 to 0.414)	0.99
Costs (£)	33,620 (25,012)	32,142 (25,798)	1478 (−1434 to 4390)	0.32
Incremental net benefit (£)a			−1446 (−8102 to 5210)	0.67

FIGURE 28.

Uncertainty in the mean costs (£) and QALY differences at 1 year and their distribution for EGDT vs. usual resuscitation.

The cost-effectiveness acceptability curve shows that that the probability of EGDT of being cost-effective compared with usual resuscitation is never > 50% irrespective of how much society is willing to pay for a QALY gain (Figure 29).

FIGURE 29.

Cost-effectiveness acceptability curve reporting the probability that the EGDT is cost-effective (at lifetime) at alternative levels of willingness to pay.

The sensitivity analyses on the long-term results suggest that these findings are robust to alternative assumptions including those applied to extrapolation of long-term survival, quality of life for survivors and costs (Figure 30). For example, a large versus smaller decrement in quality of life over 15 years had only marginal impact on the mean INB.

FIGURE 30.

Sensitivity analyses for lifetime incremental net benefit according to alternative assumptions, compared with the base case. The vertical dashed line indicates incremental net benefit in the base-case analysis. The solid vertical line indicates no difference in net monetary benefits between the treatment groups. QOL, quality of life.

The results of the subgroup analysis presented in Table 33 show that there were some differences in the direction of mean incremental effects but high statistical uncertainty surrounds these findings. Across subgroups, incremental QALYs were small. Although there were some subgroups of patients for whom the incremental costs of EGDT were negative, and hence the INBs were positive, 95% CIs around these INBs included zero. Hence for each subgroup, as for the overall result, cost-effectiveness estimates were surrounded by high statistical uncertainty.

TABLE 33 - Lifetime incremental cost-effectiveness: subgroup and secondary analyses

Subgroup	Incremental cost (£) (95% CI)	Incremental QALYs (95% CI)	Incremental net benefit (£) (95% CI)
Degree of protocolised resuscitation in usual-resuscitation group
Low	1905 (−1495 to 5305)	0.104 (−0.38 to 0.587)	168 (−7630 to 7966)
High	965 (−4874 to 6805)	−0.131 (−0.968 to 0.706)	−3589 (−17,056 to 9878)
Age (years)
18–56	4881 (−736 to 10,498)	0.096 (−0.715 to 0.907)	−2957 (−16,236 to 10,322)
57–67	1357 (−4289 to 7003)	0.114 (−0.692 to 0.919)	915 (−12,166 to 13,996)
68–77	−1945 (−7570 to 3681)	−0.201 (−1.007 to 0.606)	−2069 (−15,097 to 10,959)
78–95	4635 (−1113 to 10,383)	0.521 (−0.309 to 1.351)	5787 (−7640 to 19,214)
MEDS score
0–4	3344 (−4654 to 11,343)	0.139 (−1.006 to 1.284)	−560 (−19,050 to 17,930)
5–6	3759 (−2120 to 9638)	0.139 (−0.700 to 0.977)	−985 (−14,470 to 12,500)
7–9	819 (−4262 to 5900)	0.280 (−0.441 to 1.001)	4788 (−6840 to 16,416)
10–20	−140 (−5456 to 5176)	−0.361 (−1.104 to 0.383)	−7077 (−18,979 to 4825)
SOFA score
0–2	3008 (−2798 to 8814)	0.151 (−0.661 to 0.963)	4 (−13,077 to 13,085)
3–4	1886 (−3143 to 6915)	0.114 (−0.591 to 0.819)	386 (−10,938 to 11,710)
5	−3279 (−10,351 to 3793)	−0.312 (−1.300 to 0.677)	−2952 (−18,774 to 12,870)
6–14	1792 (−3854 to 7439)	−0.369 (−1.163 to 0.426)	−9164 (−21,917 to 3589)
Time from ED presentation to randomisation (hours)
0.2–1.8	2471 (−3280 to 8223)	0.005 (−0.819 to 0.829)	−2371 (−15,643 to 10,901)
1.8–2.5	6390 (707 to 12,073)	0.523 (−0.276 to 1.323)	4074 (−8777 to 16,925)
2.5–3.5	693 (−5140 to 6525)	−0.279 (−1.130 to 0.571)	−6275 (−20,079 to 7529)
3.5 +	–3759 (–9472 to 1953)	–0.230 (–1.052 to 0.591)	–847 (–14,099 to 12,405)
Adherence adjusted analysis	2126 (–2056 to 6307)	0.002 (–0.591 to 0.595)	–2079 (–11,654 to 7496)

Principal findings

Among adults identified with early signs of septic shock presenting to the ED of one of 56 NHS hospitals in England and receiving 6 hours of protocolised resuscitation, there was no significant difference in mortality at 90 days when compared with usual resuscitation (relative risk 1.01, 95% CI 0.85 to 1.20). Although mortality was lower than anticipated, our results rule out, with 95% confidence, a relative risk reduction with EGDT of > 15%. On average, EGDT increased costs and, given similar QALYs across groups, INB at 1 year was negative (−£725, 95% CI −£3000 to £1550). The probability that EGDT is cost-effective (at a willingness to pay of £20,000 per QALY) is below 30%. Sensitivity analyses found that this conclusion is robust to alternative assumptions to those made in the base-case analysis.

There was no significant interaction between treatment effect and mortality at 90 days across pre-specified subgroups on the basis of degree of protocolisation of usual resuscitation, age, MEDS score, SOFA score or time to randomisation. More patients receiving EGDT were admitted to ICU, resulting in significantly more days spent in critical care in this group. Treatment intensity was greater for the EGDT group, driven by adherence to the protocol, and indicated by the increased use of central venous catheters, intravenous fluids, vasoactive drugs and packed red blood cell transfusions. Increased treatment intensity was reflected by significantly higher SOFA scores and more advanced cardiovascular support days in critical care for the EGDT group. There were no significant differences in any other secondary outcomes including health-related quality of life, measured in health-state utility values, which was substantially poorer in this severely ill patient group at both 90 days (0.61) and 1 year (0.62–0.65) than for the age-/sex-matched general population (0.80). 58 At 12 months post randomisation, approximately 30% of responders reported ‘severe’ or ‘extreme’ problems with mobility or undertaking usual activities, indicating substantial ongoing morbidity for this patient group.

Strengths

ProMISe was set in a real-world context, in a large, representative, mixed sample of approximately one-quarter of NHS hospitals in England, and was pragmatic, with staff and locations for delivery of the protocol determined locally, as would be the case if the intervention were to be adopted into routine NHS care. Site set-up was rapid, resulting in our trial recruiting its full sample of 1260 patients over a relatively shorter time period than the two similar studies already reported from the USA and Australasia; minimising the potential for other changes in treatments to impact on the trial.

Loss to follow-up was minimal and all analyses were conducted according to a pre-specified, published statistical analysis plan including, given the complex intervention, adjusted analyses to address the degree of adherence to EGDT and the possibility of the existence of a learning curve for its delivery.

Unlike the previous studies, our trial reported on quality of life at 90 days and 1 year post randomisation and our results include an integrated analysis of the cost-effectiveness of EGDT. This prospectively designed economic evaluation ensured that resource use data were collected on both primary admissions and readmissions for each patient randomised. The resource use measurement harnessed information from three sources: the case report forms, linked data from the Case Mix Programme database and responses to follow-up Health Service Questionnaires. This approach enabled detailed resource use measurement for those events that were anticipated to be the key drivers of the incremental costs of EGDT versus usual resuscitation. The cost-effectiveness analysis also measured quality of life with the EQ-5D-5L;59 this version of the EQ-5D instrument was anticipated to be sensitive to differences in health status between the treatment groups. To address missing data, we undertook the recommended approach of multiple imputation and imputed missing values, conditional on all the information observed.

Limitations

As with all studies enrolling patients presenting as emergencies, recruitment was more challenging at weekends and out-of-office hours. As a result of this, together with other logistical issues, only around one-third of eligible patients were recruited. However, there were no important differences in baseline characteristics between patients recruited during usual working hours or at weekends and out-of-office hours. In addition, exclusion by a clinician was comparatively rare.

With recruitment rates much lower – and eligible patients missed – at nights and at weekends, alongside a number of sites undergoing a period of suspension or closing early to recruitment owing to lack of available resources, the trial recruited behind the planned schedule. Owing to the short time-windows to recruit and commence treatment in randomised controlled trials in emergency and critical care settings, it is vital that research infrastructure within the NHS is delivered 24 hours per day, 7 days per week.

The intervention could not be blinded to those caring for patients but the risk of bias was minimised through central randomisation to ensure allocation concealment and the use of a primary outcome not subject to observer bias. The mortality observed in the usual-resuscitation group was lower than anticipated as the basis for the sample size calculation (29.2% vs. 40%). This was also true for both ProCESS (60-day in-hospital mortality, 18.9% observed, 30–46% anticipated23) and ARISE (90-day mortality, 18.8% observed, 38% anticipated24). However, unlike both ProCESS and ARISE, based on the 95% CI, our results were able to rule out with 95% confidence, in our setting of NHS EDs, the relative risk reduction in 90-day mortality of 20% that was the basis for our sample size calculation. Although able to provide, relatively, the most precise overall estimate of effect, we have limited power to address many important subgroups for either the clinical effectiveness or the cost-effectiveness.

The long-term cost-effectiveness analysis was inevitably required to make assumptions, in particular about the mortality, quality of life and cost in the time period beyond the observed data. The study made maximum use of the available trial data to inform these assumptions. For example, the analysis of the mortality data found that mortality was similar between the treatment groups at each time point, and that there was excess mortality versus the age-/sex-matched general population for up to 15 years post randomisation. These findings informed the assumptions made in the base-case analysis concerning the long-term survival extrapolation. The study made these and other requisite structural assumptions transparent and subjected them to extensive sensitivity analyses.

Our findings in context

Implications for health care

Our results suggest that usual resuscitation has evolved over the fifteen years since the Rivers et al. 6 trial; NHS hospitals now achieve similar levels of in-hospital survival to those achieved with EGDT in the Rivers et al. 6 trial for patients with septic shock identified early and receiving intravenous antibiotics and adequate fluid resuscitation. The addition of continuous ScvO₂ monitoring and strict protocolisation of care was, on average, more costly and did not improve outcomes.

Although adherence to the EGDT protocol in ProMISe was similar in most respects with both ProCESS and ARISE, it is of note that the adherence with administration of packed red blood cells was generally low and occurred considerably slower than for other interventions. The lower adherence may reflect concerns over the relatively high haemoglobin threshold for blood transfusion in the EGDT protocol, as more recent research evidence suggests that lower transfusion thresholds may be preferable,71 but the long time lag from reaching the transfusion threshold to administration of blood may warrant local investigation to ensure adequate processes are in place for rapid provision of blood products when required.

Recommendations for research

Research staff at participating sites

We acknowledged that there have been many other individuals who made a contribution within the participating units. It is impossible to thank everyone personally; however, we would like to thank the following research staff:

Addenbrooke’s Hospital (Vazeer Ahmed, Adrian Boyle and Andy Scott-Donkin); Arrowe Park Hospital (Heather Black, Christopher Smalley, Reni Jacob and Andrea Wooten); Barnsley Hospital (Julian Humphrey, Sally Anne Pearson and James Griffiths); Bedford Hospital (Devasena Subramanyam, David Niblett and Sunil Krishnanankutty); Birmingham Heartlands Hospital (Fang Gao-Smith, Teresa Melody and Keith Couper); Blackpool Victoria Hospital (Raj Nichani, Emma Brennan and Simon Tucker); Bristol Royal Infirmary (Jonathan Benger, Judith Edwards and Kathryn Pollock); Broomfield Hospital (Dilshan Arawwawala, Alex Hieatt and Fiona McNeela); Chelsea and Westminster Hospital (Derek Bell, Theresa Weldring and Jaime Carungcong); Derriford Hospital (Peter MacNaughton, Helen McMillan and Kate Tantam); Dorset County Hospital (Tony Doyle, Sarah Moreton and Stephanie Jones); Frenchay Hospital (Jason Kendall, Ruth Worner and Anna Gilbertson); Hinchingbrooke Hospital (Colin Borland, Suzanne Boys and Shashank Ranjan); Hull Royal Infirmary (Ian Smith, Neil Smith, Victoria Mendham and Paul Smith); John Radcliffe Hospital (Duncan Young, Roser Farras-Araya and David Vallance); Kettering General Hospital (Phil Watt, Parizade Raymode and Laszlo Hollos); King’s College Hospital (Phil Hopkins, Paul Riozzi, Harriet Couper and Sinead Helyar); Leicester Royal Infirmary (Jonathan Thompson, Dawn Hales, Zubeir Essat and Prem Andreou); Leighton Hospital (Susan Gilby, Phil Chilton and Richard Miller); Manchester Royal Infirmary (John Butler, Alison Jefferies and Richard Clark); Medway Maritime Hospital (Graeme Sanders, Nuno Pinto and Catherine Plowright); Musgrove Park Hospital (Richard Innes, Dawn Bayford and Pippa Richards); New Cross Hospital (Shameer Gopal, Jagtar Singh Pooni and Hazel Spencer); Newham University Hospital (James Napier and Ena Warrington); North Devon District Hospital (Liam Kevern, Jane Hunt and Colin Barrett); North Tyneside General Hospital (Eliot Sykes, Karen Connelly and Bryan Yates); Peterborough City Hospital (Coralie Carle, Critical Care Outreach Team and Theresa Croft); Poole Hospital (Nick Jenkins, Henrik Reschreiter, Julie Camsooksai and Helena Barcraft-Barnes); Queen Elizabeth Hospital Birmingham (Catherine Snelson, Colin Bergin and Frank Keats); Queen Elizabeth Hospital, Gateshead (Vanessa Linnett, Jenny Ritzema and Steve Christian); Queen’s Medical Centre (Daniel Harvey, Philip Miller, Claudia Woodford and Anna Bolland); Royal Berkshire Hospital (Liza Keating, David Mossop and Carys Jones); Royal Bournemouth Hospital (David Martin, Emma Willett and Peter Swallow); Royal Lancaster Infirmary (Sam McBride, Asim Ijaz, Jay Datta and Jayne Craig); Royal Preston Hospital (Thomas Owen, Alex Williams, Sean McMullan and Jackie Baldwin); Royal Surrey County Hospital (Mehrun Zuleika and Peter Carvalho); Royal Sussex County Hospital (Dan Agranoff, Fiona Ingoldby, Laura Ortiz-Ruiz De Gordoa and Carrie Ridley); Royal Victoria Infirmary (Ian Clement and Charley Higham); Salford Royal Hospital (Bruce Martin, Kate Clayton and Julie Chadwick); South Tyneside District Hospital (Christian Frey, Diane Miller and Philippa Laverick); Southend University Hospital (Khurram Iftikhar, David Higgins and Victoria Katsande); Stafford Hospital (Moses Chikungwa and Clare Jackson); The Great Western Hospital (Malcolm Watters, Paul Liddiard and Kate Gannon); The Ipswich Hospital (Richard Howard-Griffin, Stephanie Bell and Heather Blaylock); The James Cook University Hospital (Isabel Gonzalez, Emmanuel Cirstea and Stephen Bonner); The Queen Elizabeth Hospital, King’s Lynn (Parvez Moondi, Kate Wong and Joseph Carter); The Royal Blackburn Hospital (Stephen Hartley, Iain Crossingham, Joanne Hinchcliffe and Leanne Phoenix); The Royal London Hospital (Tim Harris, Jason Pott and Geoffrey Bellhouse); Torbay Hospital (Michael Mercer, Pauline Mercer and Hazel Robinson); University College Hospital (David Brealey, Jung Ryu, Georgia Becardes and the Critical Care Trials Team); University Hospital of North Staffordshire (Anne-Marie Morris, Mark Poulson, Loretta Barnett and Ian Massey); Wansbeck General Hospital (Eliot Sykes, Karen Connelly and Bryan Yates); Whipps Cross University Hospital (Tim Harris and Imogen Skene); Whiston Hospital (Patrick Nee, Susan Dowling and Amanda McCairn); Worthing Hospital (Roger Duckitt, Richard Venn and Jordi Margalef); and York Hospital (Jonathan Redman, Helen Milner and Sara Ma).

Contributions of authors

Paul R Mouncey (senior trial manager) managed the trial, contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Dr Tiffany M Osborn (associate professor) contributed to the design of the trial, the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

G Sarah Power (statistician) contributed to the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Dr David A Harrison (senior statistician and honorary senior lecturer in medical statistics) contributed to the design of the trial, the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Dr M Zia Sadique (lecturer in health economics) contributed to the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Richard D Grieve (professor of health economics methodology) contributed to the design of the trial, the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Rahi Jahan (research assistant) contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Jermaine CK Tan (trials data manager) contributed to the analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Dr Sheila E Harvey (CTU manager and senior research fellow) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Derek Bell (professor of acute medicine) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Julian F Bion (professor of intensive care medicine) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Timothy J Coats (professor of emergency medicine) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Mervyn Singer (professor of intensive care medicine) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor J Duncan Young (professor of intensive care medicine) contributed to the design of the trial, contributed to the acquisition, analysis and interpretation of the data, and drafted and critically reviewed the manuscript.

Professor Kathryn M Rowan (Director of Scientific & Strategic Development at ICNARC and honorary professor) conceived and designed the trial, contributed to acquisition, analysis and interpretation of the data, and drafted and critically revised the manuscript.

Trial Management Team

Paul Mouncey (senior trial manager); Sarah Corlett (previous trial administrator); Dr David Harrison (senior statistician); Dr Sheila Harvey (CTU manager/senior research fellow); Rahi Jahan (research assistant); Hannah Muskett (previous assistant trials manager); Sarah Power (trial statistician); Professor Kathryn Rowan (chief investigator); Dr Rachael Scott (previous senior trial manager); and Jermaine Tan (trials data manager).

Trial Management Group

Professor Derek Bell (co-investigator); Professor Julian Bion (co-investigator); Professor Timothy Coats (co-investigator); Dr David Harrison (senior statistician); Dr Sheila Harvey (CTU manager/senior research fellow); Dr Tiffany Osborn (co-investigator); Professor Kathryn Rowan (chief investigator); Professor Mervyn Singer (co-investigator); and Professor Duncan Young (co-investigator).

Trial Steering Committee

Professor Steve Goodacre (independent chairperson); Professor David Bennett (previous independent); Professor Julian Bion; Caroline Bowers (independent); Phillip Crow (independent); Professor Rupert Pearse (independent); Professor Kathryn Rowan; and Professor David Yates (independent).

Independent Data Monitoring and Ethics Committee

Professor Jon Nicholl (chairperson); Professor Alasdair Gray; and Professor Timothy Walsh.

Publications

Power GS, Harrison DA, Mouncey PR, Osborn TM, Harvey SE, Rowan KM. The Protocolised Management in Sepsis (ProMISe) trial statistical analysis plan. Crit Care Resusc 2013;15:311–17.

Huang DT, Angus DC, Barnato A, Gunn SR, Kellum JA, Stapleton DK, et al. Harmonizing international trials of early goal directed resuscitation for severe sepsis and septic shock: methodology of ProCESS, ARISE, and ProMISe. Intensive Care Med 2013;39:1760–75.

Mouncey PR, Osborn RM, Power GS, Harrison DA, Sadique MZ, Grieve RD, et al. Trial of early, goal-directed resuscitation for septic shock. N Engl J Med 2015;372:1301–11.

Angus DC, Barnato AE, Bell D, Bellomo R, Chong CR, Coats TJ, et al. A systematic review and meta-analysis of early goal-directed therapy for septic shock: the ARISE, ProCESS and ProMISe Investigators. Intensive Care Med 2015;41:1549–60.

Data sharing statement

Data can be obtained from the corresponding author.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Acute Physiology And Chronic Health Evaluation version II

The APACHE II Acute Physiology Score consists of weightings for 12 physiological parameters to give a total score ranging from 0 to 60, with higher scores indicating greater severity of illness. 33 The 12 physiological parameters are as follows:

• temperature

• mean arterial pressure

• heart rate

• respiratory rate

• alveolar–arterial gradient (if FiO₂ ≥ 0.5) or PaO₂ (if FiO₂ < 0.5)

• arterial pH (or serum bicarbonate if no arterial blood gas recorded)

• serum sodium

• serum potassium

• serum creatinine (with double weighting for acute renal failure)

• haematocrit (estimated from haemoglobin)

• white blood cell count

• GCS score (assumed to be normal for patients sedated or paralysed).

The APACHE II score comprises the Acute Physiology score plus additional weightings for age and severe comorbidities in the past medical history to give a total score ranging from 0 to 71. Severe comorbidities must have been present and documented in the past medical history within the 6 months prior to presentation at hospital and are defined as follows:

• severe liver condition – presence of portal hypertension, biopsy proven cirrhosis or hepatic encephalopathy

• severe cardiovascular condition – presence of fatigue, claudication, dyspnoea or angina at rest (New York Heart Association Functional Class IV)

• severe respiratory condition – presence of permanent shortness of breath with light activity due to pulmonary disease, or on home ventilation

• renal condition – receiving chronic renal replacement therapy (haemodialysis, haemofiltration and peritoneal dialysis)

• immunological condition – receiving chemotherapy, radiotherapy or daily high-dose steroid treatment (≥ 0.3 mg/kg, prednisolone or equivalent) for 6 months, or diagnosis of human immunodeficiency virus (HIV)/acquired immunodefiency syndrome (AIDS), lymphoma, acute or chronic myelogenous/lymphocytic leukaemia, multiple myeloma and active metastatic disease.

Sequential Organ Failure Assessment

The SOFA score consists of weightings for six organ systems to give a total score ranging from 0 to 24, with higher scores indicating a greater degree of organ failure. 30 Organ dysfunction is defined as follows:

• respiratory dysfunction, defined as PaO₂/FiO₂ < 400 mmHg

• cardiovascular dysfunction, defined as mean arterial pressure < 70 mmHg (irrespective of vasopressor use)

• renal dysfunction, defined as creatinine of ≥ 1.2 mg/dl (110 µmol/l)

• neurological dysfunction, defined as GCS score of ≤ 14

• hepatic dysfunction, defined as bilirubin of ≥ 1.2 mg/dl (20 µmol/l)

• coagulation dysfunction, defined as platelets < 150 × 10⁹/l.

Mortality in Emergency Department Sepsis

The MEDS score is derived from nine variables to give a total score ranging from 0 to 33, with higher scores indicating a greater risk of death. 34 The nine variables are as follows:

• terminal illness, defined as presence of metastatic disease [distant (not regional lymph node) metastases documented by surgery, imaging or biopsy]

• respiratory difficulties, defined as tachypnea (respiratory rate of > 20 breaths per minute) or hypoxia (SpO₂ < 90% or FiO₂ of ≥ 0.4)

• septic shock, defined as severe sepsis plus hypotension (systolic blood pressure < 90 mmHg) that persists after initial fluid challenge of 20–30 ml/kg body weight of intravenous crystalloid

• low platelet count, defined as < 150 × 10⁹/l

• bandemia, defined as baseline immature neutrophils (band forms) < 5%

• age > 65 years

• suspected lower respiratory tract infection

• nursing home residence

• altered mental status, defined as a recent change in sensorium (confusion, disorientation, drowsiness, obtundation, stupor or coma) by history or physical examination or GCS score of ≤ 14.

Definitions

Duration of organ support in the critical care unit was defined as the number of days alive and free from support of each of the following organ systems, as defined by the UK Department of Health Critical Care Minimum Data Set,³¹ during the first 28 days following randomisation. Patients that died within the first 28 days were assigned 0 days alive and free from organ support. Organ support definitions were as follows:

• advanced respiratory – indicated by one or more of invasive mechanical ventilatory support through a translaryngeal tube or tracheostomy; bilevel positive airway pressure through a trans-laryngeal tube or tracheostomy; continuous positive airway pressure through a trans-laryngeal tube; or extracorporeal respiratory support

• advanced cardiovascular – indicated by one or more of receipt of multiple intravenous and/or rhythm controlling drugs (of which at least one must be vasoactive) when used simultaneously to support or control arterial pressure, cardiac output or organ/tissue perfusion; continuous observation of cardiac output and derived indices; an intra-aortic balloon pump or other assist device; or a temporary cardiac pacemaker

• renal – indicated by receipt of acute renal replacement therapy (e.g. haemodialysis, haemofiltration, etc.) or receipt of renal replacement therapy for chronic renal failure where other acute organ support is received.