Protocol for fever control using external cooling in mechanically ventilated patients with septic shock: SEPSISCOOL II randomised controlled trial


Background and rationale

Sepsis is a common syndrome responsible for multiple organ failure and septic shock, which lead to a high mortality rate.1 Patients with septic shock typically experience a disproportionate host response to infection so that fever may contribute to excessive inflammation and worsen organ failure. Fever has a high metabolic cost, so cooling febrile critically ill patients reduces their heart rate, cardiac output and oxygen consumption.2 The deleterious effects of fever on inflammation have mostly been shown in experimental models of acute lung injury.3–6 Because experiments have also reported protective effects of heat stress on host defences and lung inflammation, the controversy persists.7 It remains unclear whether fever or the physiological response to fever causes morbidity, and whether management of fever with pharmacological agents or physical cooling confers benefit. The benefit-to-risk ratio of fever treatment may be favourable in the presence of severe infection.8 9 Fever is associated with worse disease evolution in patients with acute respiratory distress syndrome (ARDS).10 Reducing oxygen demand, cardiac output, right-to-left lung shunt and carbon dioxide CO2 production can increase partial pressure of oxygen PaO2 and allow more protective ventilation strategies.11 12 As trials have reported harmful effect of induced hypothermia in sepsis, avoiding hypothermia represents an important goal when controlling fever.13 14

The SEPSISCOOL-I (SC-I) pilot study focused on febrile patients with septic shock and compared a strategy of fever control using external cooling to no fever treatment.15 Cooling significantly improved organ function and decreased mortality on day 14 with an adjusted OR of 0.36 (95% CI 0.17 to 0.76). There were no significant adverse events (AEs) observed in the cooling group. Of note, the need for sedation had to be included to limit the risk of shivering and to produce a metabolic benefit.2 There was a trend towards a higher incidence of early acquired nosocomial infection in the cooling group. Discrepant results regarding the risk of acquired infection when using fever control have been reported.16 17

Only six randomised controlled trials (RCTs) assessing fever treatment in intensive care unit (ICU) patients with infection have reported mortality. Four of these RCTs used pharmacological intervention and two used external cooling.15 18–22 The efficacy in controlling fever was higher when using cooling.23 Two meta-analyses in adult ICU patients, with or without sepsis, concluded that active temperature management neither increases nor decreases the mortality risk.24 25 Drewry et al24 reported a relative risk (RR) of 0.93 (95% CI 0.77 to 1.13) and Dallimore et al25 reported an RR of 1.01 (95% CI 0.81 to 1.28). The wide CIs indicate uncertainty in the estimate of the treatment effect. The authors of these meta-analyses did not take into account the severity of the treated patients. An individual personal data meta-analysis assessed fever control in adult ICU patients according to their baseline characteristics.23 The RR of reducing death using fever control was the largest (22%) among patients requiring invasive mechanical ventilation and vasopressor (OR 0.72, 95% CI 0.50 to 1.04); the CI includes a potential clinically important benefit of fever control in this subgroup of the most severe patients.

Objectives and outcomes

Primary objective and outcome

The study hypothesis is that a strategy of fever control at normothermia using external cooling will significantly reduce mortality at day 60 after randomisation compared with no fever treatment.

Secondary objectives and outcomes

The following secondary outcomes will be examined:

  1. Evolution of organ failure assessed by organ-support-free days on day 28 and evolution of the Sequential Organ Failure Assessment (SOFA) score.

  2. Acute kidney injury (AKI, based on the creatinine Kidney Disease Improving Global Outcomes criteria)26 up to day 7, criteria to initiate renal replacement therapy (RRT)27 and to start new RRT up to day 28.

  3. ARDS acquired after randomisation up to day 3.28

  4. Mortality at day 28 after randomisation.

  5. Incidence of secondary ICU-acquired nosocomial infections through day 28.

  6. Tolerance of the two strategies:

    • The number of patients with shivering through 48 hours.29

    • The number of patients with seizures through 72 hours, clinically documented and/or proven by electroencephalography.

    • The number of patients with hypothermia through 72 hours defined by a core body temperature below 36°C.

    • The number of patients with >1 new episode of supraventricular or ventricular cardiac arrhythmia through 72 hours.

Study design

SEPSISCOOL-II is an investigator-initiated, pragmatic, multicentre, open-label, randomised controlled, superiority trial. Primary outcome across two parallel groups in a 1:1 ratio is assessed on statistical difference. After stratification based on the presence of ARDS, adaptive randomisation and an adaptive multistage population-enrichment design will be used.30 31 Inclusions started on October 2022. We complied with the SPIRIT 2013 checklist ( recommendations for this protocol draft (see online supplemental material 1) and provided all WHO Trial Registration Data Set (see online supplemental material 2).

Supplemental material

Rationale for adaptive study design

To test the hypothesis that patients with ARDS will benefit the most from fever control

Determining groups of patients who are more likely to benefit from treatments represents an important goal for individualised medicine.32 33 Heterogeneity decreases the likelihood that studies will show treatment benefits. The same intervention can benefit one group but worsen another.34 35 A multistage population-enrichment design has been proposed to adapt the study design in the presence of different treatment effects among groups of patients.30 Two prespecified strata of patients according to the presence or the absence of ARDS will be identified at randomisation.30 31 According to data from the pilot study, half of the included patients are expected to present with ARDS.15

To ensure balance at interim and final analyses

The trial intervention is not masked for the investigators, healthcare workers and patients, as blinding of cooling is not feasible. This increases the risk of a selection bias of the included patients, resulting in non-comparable groups. Increasing the number of patients reduces the risk of imbalance. Notably, interim analyses are performed on necessarily small populations, but balance in important predictors is required to permit confident decisions. Adaptive randomisation based on predictors of mortality would ensure balance at interim and final analyses, helping to make adequate decisions on study continuation and interpretation of the final results.36

To warrant adequate power

Available outcome data on febrile patients with the Sepsis-3 criteria are limited.37 An adaptive design with sample size reassessment at interim analysis based on the conditional power (CP) observed between the cooling and no cooling group will ensure the study has sufficient power.

Methods and analysis

Eligibility criteria

Patients with septic shock defined by the Sepsis-3 definition will be screened for enrolment.37 Only patients requiring invasive mechanical ventilation and sedation will be included. Sedative treatments are mandatory to avoid induction of shivering. There is no consensus regarding the definition of fever; the same threshold applied in the pilot study will be used—that is, 38.3°C.

There is no maximal delay for inclusion after the start of vasopressor. Fever can occur later when a patient comes from the operating room or is hypothermic at the initial phase of shock. However, at the time of inclusion and randomisation patient must still need vasopressor.

Patients with other indications for temperature control will not be included (table 1).

Table 1

Eligibility criteria


Fever management

Interventions for fever management is applied by the ICU team similarly to the sepsiscool-I trial.15 In the experimental group, fever is controlled by external cooling during the first 48 hours following randomisation. The objective of cooling is to maintain normothermia at 36.5°C–37°C over 48 hours. Cooling is applied by automated surface methods according to the technique available in each centre. The temperature target is set at 36.7°C.

In the control arm, no cooling must be applied. No intervention switch is allowed within the trial. Antipyretic methods are not allowed except when fever surpasses 41°C. A subject may be discontinued from allocated intervention at any time if the investigator or the sponsor feels that it is not in the subject’s best interest to continue. If a subject is withdrawn from allocated intervention due to an AE, the subject will be followed and treated by the investigator until the abnormal parameter or symptom has resolved or stabilised.

Other interventions in the two groups

Septic shock management is let at the discretion of clinical staffs according to usual centre practices. Weaning of vasopressors is performed according to the algorithm used and published in the sepsiscool-I pilot trial.15 Management of ARDS is performed according to French expert guidelines.38

Sample size calculation

Justification of estimations

The Sepsis-3 definition of septic shock will select patients with a higher mortality risk than those reported in older studies.35 39 The expected mortality has been extracted from the results of the pilot study and from the databases of two RCTs performed in France (ie, ‘HYPER2S’ and ‘SEPSISPAM’) among patients meeting the Sepsis-3 criteria (n=698 patients).15 40 41 Half of the patients had ARDS and the observed mortality was 54%; mortality was similar among patient with and without ARDS.

In a meta-analysis of personal data, the RR of death using fever control among patients requiring mechanical ventilation and vasopressors was 22% (OR 0.72, 95% CI 0.50 to 1.04).23 The RR difference estimation in large RCTs comparing two treatments for sepsis was 22% in the ‘HYEPR2S’, ‘SEPSISPAM’, ‘CASS’ and ‘APROACCHSS’ studies. The RR difference estimations in large RCTs comparing treatments for ARDS were 19% in the ‘PETAL’ study and 33% in ‘EPVent-2’ study.13 40–44


An expected mortality of 52% with an expected RR of death of 19% is assumed. The hypothesis is to reduce mortality at day 60 from 52% to 42%. For a two-sided alpha of 5%, a power of 80% and a cooling-to-no cooling ratio of 1:1, 780 evaluable patients are needed. Due to lost to follow-up or other reasons, the final sample size of 820 patients is planned. The calculation was performed using the nQuery V.8 software (PTT0).


ICU medical investigators complete the process of informed consent and randomise the patients during their ICU stay. The treatment arm allocation will be provided by an interactive web response system (IWRS). After randomisation, cooling will be initiated immediately in patients allocated to the experimental group (figure 1). Allocation will be performed by stratification based on ARDS and by covariate-adaptive procedures. The ARDS criteria and the following patient characteristics—known to be strong predictors of mortality—will be recorded and added to the IWRS at randomisation to power the algorithm and ensure to the groups are balanced:

  • Immunosuppression (ie, cancer chemotherapy, HIV, corticosteroids >1 mg/kg and/or >1 month, solid organ transplant, bone marrow transplant, immunosuppressant drugs for autoimmune disease).

  • Arterial oxygen partial pressure to fractional inspired oxygen PaO2/FiO2 ratio (≤4 hours before randomisation).

  • Lactate level (≤4 hours before randomisation).

  • Age.

  • Vasopressors (epinephrine+norepinephrine).

  • Clinical centre of randomisation.

Figure 1
Figure 1

SEPSISCOOL-II trial design. ARDS, acute respiratory distress syndrome.

The aforementioned characteristics are routinely monitored: their real-time availability guarantees the feasibility and the robustness of the real-time successful randomisation. These variables will be monitored and recorded in the study database. Hence, the distribution of covariates will be continually updated as the trial progresses.

Dynamic hierarchical randomisation will be used to tackle the challenges of balancing a large number of stratification variables.45 The minimal sufficient balance randomisation method developed by Zhao et al46 will be implemented. As recommended by regulators, a random component will be incorporated to reduce the predictability of the allocation.47 48 The imbalance control limit will be set to p≥0.3 for all covariates, and the covariate-adaptive allocation procedure will start after 20 patients have been randomised in each ARDS stratum.46 For each covariate (either continuous or categorical), a vote is conducted to decide whether this covariate should be selected for the randomisation and how to reach balance. The algorithm then allocate the definitive arm, with a predefined probability, according to the results of the votes for the selected covariates. This process is performed for each new participant.

Development and test of the algorithm

The proposed algorithm for randomisation has been tested internally and widely on simulated data (>10 000 datasets simulated). Nineteen scenarios were computed and included worst-case and best-case schemes. Once we will have fully implemented the randomisation in this study, a separate research paper will provide a comprehensive overview of all relevant developments (ie, simulations and real-world data) used for assessing and validating the performances of the proposed algorithm.


As blinding of cooling is not feasible, the trial intervention is not masked for investigators, healthcare workers and patients. Several methodological aspects limit the risk of bias. First, adaptive randomisation precludes the risk of allocation expectation. Second, interim and final statistical analyses will be performed by independent statistician. At interim analysis, the need for a blind break is let at the discretion of the data monitoring committee (DMC). Third, the primary endpoint of the study, that is, mortality, is unlikely to be influenced by the lack of blinding. Fourth, the majority of secondary endpoints are specifically recorded for the study according to strict definitions. However, they are not validated by an endpoint adjudication committee.

Data collection

All information on data collection methods and management can be found in the protocol version V.3.0 dated 8 June 2023.

Method and timing

Data will be collected on an electronic case report form (CRF) by a trained investigator or research technician among 27 French hospitals (see exhaustive list in online supplemental material 3). Research assistants will regularly monitor all centres on site to attest to protocol adherence and the accuracy of the recorded data. The details of the recorded data are indicated in online supplemental material 4.

Quality control

A plan for quality control will be established prior to the beginning of the study. This plan of monitoring, tailored to the specific human subject protection and data integrity risks of the study, will describe the strategy, methods, responsibilities of all parties involved, and requirements for monitoring the study.

Statistical methods

Except where mentioned otherwise, for all analyses (primary, secondary and interim), two-sided p<0.05 will be considered statistically significant. The strategy for design, analysis and report will be made in compliance with the Consolidated Standards of reporting Trials CONSORT statement ( Any modification of the presented statistical analysis strategy will be indicated in the final publication.

Compliance with the protocol and study treatment

Protocol deviations will be described prior to database lock in a blinded fashion regarding either the use of cooling or antipyretic drugs in the control group or no cooling performed in the experimental group. This step will lead to establishing the populations (randomised and treated per protocol).

Demography and clinical characteristics at baseline

Demographic and clinical characteristics of patients at baseline (H0) will be compared between the two groups according to standard statistical tests: Student’s t-test or the non-parametric Wilcoxon test for quantitative parameters (based on the variable’s distribution) and the χ2 test or Fisher’s exact test for qualitative parameters. The results will be presented as mean±SD if the parameter follows a normal distribution and median (IQR) if the parameter follows another distribution.

Primary analysis

Principal analysis on primary endpoint

The difference in the proportion of deceased patients through day 60 after randomisation will be compared using logistic regression in the randomised population according to the intention-to-treat principle.49 The following hypotheses will be tested:

The results will be presented overall (pooled ARDS strata) as the absolute difference and 95% CI (unbiased estimates). After programming, closed testing together with the p value combination function will be performed to control for type I errors.

Sensitivity analysis on primary endpoint

The model used for the primary analysis will be challenged at the end of the study with the following (but not limited to) sensitivity analyses:

  • The analysis will be repeated on the per-protocol population, that is, all randomised patients without any major protocol deviation.

  • An analysis based on the ‘worst-case scenario’ may be considered for patients with missing vital status information at day 60.

  • A survival analysis will be conducted using Kaplan-Meier curves and the associated log-rank test.

Secondary analyses

Evolution of organ failure

Evolution of the SOFA score between day 0 and day 7 will be analysed using a mixed model for repeated measures, with a random patient effect (subject’s intercept) and an interaction term time×patient (subject-specific evolution).

Organ-support-free days—that is, ventilator, vasopressors and RRT from randomisation to day 28—will be assessed as proposed by Yehya et al.50 The number of free days between the two treatment groups will be compared by Student’s t-test or the non-parametric Wilcoxon test for quantitative parameters (based on the variable’s distribution).

The proportions of patients with AKI through day 7 and the proportion of patients who develop ARDS after randomisation through H72 will be compared using a logistic regression as described for the primary analysis.

Day 28 mortality

Day 28 mortality will be compared between the two treatment groups using logistic regression as described for the primary analysis.

Expected AEs

The proportion of patient with shivering hypothermia, new cardiac arrhythmia through 48 hours and seizure through 72 hours will be compared between the two treatment groups using a logistic regression as described for the primary analysis.

The incidence of secondary ICU-acquired nosocomial infections through day 28 will be reported for 1000 ICU days and compared by Poisson or negative-binomial regression according to overdispersion and zero-inflation of the data.15 The definitions used for nosocomial infections are indicated in online supplemental material 5.

Preplanned subgroup analysis

The effect of cooling on all-cause mortality at day 60 (primary endpoint) will be assessed among the subgroups of:

  • ARDS status.

  • Immunosuppressed patients.

  • Patients not receiving treatment potentially influencing temperature evolution, that is, steroids and RRT.

  • Patients randomised less than 12 hours after vasopressor start.

Interim analyses

A first interim analysis will be performed after enrolment of 200 patients for safety assessment only. Exploratory outcomes, mortality at day 28, the Simplified Acute Physiology III score and variables used in the algorithm for adaptive randomisation will be provided to the DMC. Data will be presented for four groups according to the ARDS strata and study arms (ARDS YES with cooling YES, ARDS YES with cooling NO, ARDS NO with cooling YES, ARDS NO with cooling NO) in a blinded fashion. No statistical test will be performed between groups; the decision to stop the study for safety reason will be let at the discretion of the DMC.

A second interim analysis is planned after enrolment of 50% of patients. In addition to the data listed above, the number of patients receiving antipyretics in the no cooling group and mortality at day 60 will be provided. Interim analysis will be performed by a blinded independent statistician. The sponsor and all investigators will be kept blinded to the data and the interim analysis results. The decision for breaking the randomisation code will be let to the discretion of the DMC.

Statistical criteria for termination of the study at the second interim analysis

Because of the limited number of subjects at interim analysis, early stop for efficacy may strongly increase the type I error.51 The stopping rules are planned for futility only.

For each ARDS stratum, the CP (ie, the probability of rejecting the null hypothesis at the final stage of the trial, given the current data) calculated by the method proposed by Metha and Pocock, applying stochastic curtailment. The calculation will be made according to (a) the observed difference of the primary endpoint between the experimental and control group at the interim step and the interim information level computed as a function of (b) the observed number of patients in each treatment group at the interim step and (c) the average of primary endpoint expected in the experimental and in the control groups in the original trial planning .30 52 53 In the absence of an ARDS stratum effect (CP similar in the two strata), then we will compute a unique CP by pooling the strata (figure 2A); else if the CP is different between the two strata, then we will make two separate decisions depending on the CP of each stratum (figure 2B,C). Three zones for CP-based decision-making are defined as follows:

  • Unfavourable: CP<CPmin, the CPmin cut-off value will be set at 10%.

  • Promising: CPmin<CP < 80% (1–β).

  • Favourable: CP>80%.

Figure 2
Figure 2

Interim analyses and rules for decision-making. See the text for the definition of conditional power. (A) Absence of the acute respiratory distress syndrome (ARDS) stratum effect; (B) effect of cooling in favour of ARDS; (C) effect of cooling in favour of no ARDS. CP, conditional power.

If CP falls in the unfavourable zone, the stratum or study will be stopped. If the CP falls in the promising zone, the sample size will be re-estimated based on the considered ARDS strata or whole population. If the CP falls in the favourable zone, the stratum or study will continue with the initial sample size estimation. The decision-making process is summarised in the figure 2. Depending on the inferential statistical results at this interim analysis, the DMC will make one of the following recommendations:

  • To continue the trial to the final step with both ARDS strata (initial sample size prevails).

  • To continue the trial with enrichment for both ARDS strata after sample size reassessment.

  • To continue the trial to the final step for one ARDS strata (initial sample size prevails) with enrichment for the other strata after sample size reassessment.

  • To stop one ARDS strata and to continue the trial without enrichment for the other strata (initial sample size prevails).

  • To stop one of the two ARDS strata and to continue the trial with enrichment for the other strata after sample size reassessment.

  • To completely stop the study.

For strong control of the familywise error rate at the set alpha level, the use of the closed testing principle is planned at the very end of the study (any elementary hypothesis must be rejected alone, but all intersections that contain this elementary hypothesis must also be rejected) together with a p value combination function to combine the p values observed at each stage of the sequential design.30 54 55

Missing data management

During data management, invalid data will be considered missing data. Except when stated otherwise, no imputation is planned for either the primary or secondary analyses.

Data monitoring

Data monitoring committee

The DMC comprises three clinicians and one statistician (see composition in online supplemental material 6). Collectively, they have experience in the conduct, monitoring and analysis of RCTs. None of the members are directly involved in the study. DMC membership is restricted to individuals free of conflict of interest, and each member has signed a conflict of interest form. The DMC has approved the study protocol and the operating charter (further details in version v1–0 dated 23 December 2020).

At interim analysis, the committee will monitor the rate of inclusion and recorded expected adverse effects and will make the final decision to terminate the study or not. Results of interim analysis on day 60 mortality will be provided as well as the incidence of expected adverse effects. Patients’ characteristics recorded at randomisation for treatment allocation will be also provided. Because the study is at minimal risks, no rules will be defined to stop the study for safety reasons but early stop for safety reasons will be left at the discretion of the DMC.

Monitoring of AEs

This study will not test drugs, devices or human products, which has been classified by the ethics committee as a trial exposing patients to minimal risks and constraints. Patients with septic shock are at high risk for developing complications no matter the strategy of fever management. Therefore, serious AEs will not be recorded as an entity. The most important AEs will be captured in the exploratory outcome measures: death, seizure, shivering, nosocomial infections, cardiac arrhythmia and skin lesions related to surface cooling. The list was validated by the DMC.

ICU patients are closely monitored 24 h a day by expert nurses and physicians allowing rapid detection of AEs.

According to the French law, in study at minimal risks and constrains, vigilance and safety reporting will correspond to those performed in usual care. AEs will be reported to the corresponding agency ‘pharmacovigilance’, ‘materiovigilance’ or ‘biovigilance’. A monitoring of expected adverse effects will be specifically performed for the study. The following known and expected adverse effects will be notified in the CRF but not reported to the sponsor (Article R1123-49): death, seizure, shivering, nosocomial infections, cardiac arrhythmia, skin lesions related to surface cooling.

Data availability statement

According to the Best Clinical Practice, the coordinating investigator permits the direct access to the trial documentation to any authorised party to examine, analyse, verify and reproduce any records and reports for evaluation of a clinical trial.

The sponsor and the clinical research unit of Assistance Publique Hôpitaux de Paris APHP–Hôpital Européen Georges Pompidou, in charge of the statistical analyses, will have access to the final trial dataset. Datasets and statistical code are not publicly available but request on access may be submitted to the corresponding author or to the sponsor (clinical research unit, Centre Hospitalier Intercommunal de Créteil, 40 avenue de Verdun, 94 000 Créteil, France).

No additional data are available at this date.

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.


The SEPSISCOOL-II trial was developed to provide evidence on the beneficial effect of fever control at normothermia by external cooling in a selected population of severe patients expected to benefit the most from this treatment.


The hypothesis tested in the SEPSISCOOL-II trial comes from current evidence reported prior to study development in systematic reviews. The feasibility has been tested in a pilot RCT.15 The design is an adaptive multicentre, randomised controlled, superiority trial allowing adaptations for collected evidence. Interim analyses are key to refine the effect of cooling in patients with ARDS. We will conduct an interim analysis after 50% of the participants have been followed for 60 days to allow early identification of the treatment response according to whether ARDS is present and to adapt the study continuation according to strict rules. The confidence in making the decision will be increased assuming between-group comparability. When validating the algorithm, imbalance across clinical centres and heterogeneity based on randomisation variables have typically been studied and did not lead to considerable impact on randomisation process.

Several methods of adaptation in randomisation have been described, but few have been tested in the reality of a clinical trial. This trial will use the method of Zhao et al,46 which has been applied in a large RCT (NCT04494074) to control the between-group balance in a population at risk for heterogeneity, as regularly observed in studies on septic shock.


To construct the randomisation algorithm, we selected a priori variables previously found to be associated with mortality in patients with septic shock. The drawback of this approach is the need to include variables that are trustworthy and easily available at the time of randomisation and at each clinical site. This is the reason why we did not select calculated predictive scores. Other factors influencing the primary endpoint could lead to imbalances between the groups. Sensitivity analysis with adjustment is planned to limit that risk.

The trial intervention is not masked for investigators, healthcare workers and patients, as blinding of cooling is not feasible. Hence, there is an increased risk of bias. However, our primary outcome on mortality is unlikely to be affected by the lack of blinding.

Participants may be subjected to protocol violations, expected to occur more frequently in the no cooling group. To maximise protocol compliance, the inclusion criteria include the investigator’s agreement with equipoise for fever treatment.

Ethics and dissemination


The clinical research unit of the Centre Hospitalier Intercommunal de Créteil is the sponsor of the trial. The patients will be not able to provide informed consent; therefore, a process of deferred consent will be applied. Their next of kin or other designated persons available at screening will be requested to express their wishes to include the patient and to sign an informed consent form. As soon as possible, a definitive post hoc consent form will be obtained from each patient to continue to participate in the study.

The protocol, informed consent forms and subject information sheet have been reviewed and approved (no: by a French ethics committee (Comité de Protection des Personnes Nord Ouest II (Amiens, France)). For any major change in the protocol, the sponsor will request the approval of the ethic committee (article L. 1123-9), inform the French health authority (ANSM), the investigators and the DMC, and update trial registry

The data will be recorded and managed in compliance with the French Data Protection Act and with the European Data Protection Regulation before, during and after the study. All original records will be archived at the trial sites for 15 years.

This manuscript corresponds to V.3.0 dated 8 June 2023.


Communications and scientific reports corresponding to this study will be carried out under the responsibility of the coordinating investigator (Frédérique Schortgen) with the agreement of all investigators. Publications will be approved by the statistician of the clinical research unit of Assistance Publique Hopitaux de Paris APHP–Hopital Européen Georges Pompidou (Armelle Guénégou-Arnoux), the coordinating investigator and all the investigators. The coordinating investigator, the statistician and the investigators of trial sites with at least one inclusion will be granted one authorship. Publication rules will follow international recommendations.56 Participants will be informed of the overall results of the study, if wished (Article L1122-1). We do not intend to use professional writers at this date.


The authors would like to thank Vidhya Raghavan, Cécile Hoffart and Marwa Touaiti from the clinical research unit of the Centre Hospitalier Intercommunal de Créteil for their assistance in the conduct of the SEPSISCOOL-II trial.

This post was originally published on