Introduction: Intravenous (IV) fluid is a key intervention in the management of septic shock. The benefits and harms of lower versus higher fluid volumes are unknown and thus clinical equipoise exists. We describe the protocol and detailed statistical analysis plan for the conservative versus liberal approach to fluid therapy of septic shock in the Intensive Care (CLASSIC) trial. The aim of the CLASSIC trial is to assess benefits and harms of IV fluid restriction versus standard care in adult intensive care unit (ICU) patients with septic shock. Methods: CLASSIC trial is an investigator‐initiated, international, randomised, stratified, and analyst‐blinded trial. We will allocate 1554 adult patients with septic shock, who are planned to be or are admitted to an ICU, to IV fluid restriction versus standard care. The primary outcome is mortality at day 90. Secondary outcomes are serious adverse events (SAEs), serious adverse reactions (SARs), days alive at day 90 without life support, days alive and out of the hospital at day 90 and mortality, health‐related quality of life (HRQoL), and cognitive function at 1 year. We will conduct the statistical analyses according to a pre‐defined statistical analysis plan, including three interim analyses. For the primary analysis, we will use logistic regression adjusted for the stratification variables comparing the two interventions in the intention‐to‐treat (ITT) population. Discussion: The CLASSIC trial results will provide important evidence to guide clinicians' choice regarding the IV fluid therapy in adults with septic shock.
Intravenous (IV) fluid therapy is a mainstay intervention in the treatment of sepsis. However, international guidelines lack firm evidence for the optimal volume.[
In a landmark trial in African children with infection and circulatory impairment, increased mortality was observed in children who received fluid boluses as compared to those who did not.[
In the CLASSIC pilot trial,[
In a systematic review of RCTs aiming at separation in IV fluid volumes in adult patients with sepsis, we have currently identified 9 trials including a total of 637 patients; all trials were judged to have overall high risk of bias. We found a very low quantity and certainty of the evidence for lower versus higher fluid volumes across all patient‐important outcomes (Meyhoff 2019, in preparation).[
Altogether, there is a need for high‐quality data from RCTs to inform the decision on fluid volume strategies in septic shock.
The aim of the CLASSIC trial is to assess the effects of IV fluid restriction versus standard care on patient‐important outcome measures in adult ICU patients with septic shock. We hypothesise that, fluid restriction will improve patient‐important outcomes. In this manuscript, we outline the rationale, methods, and the detailed statistical analysis plan.
The CLASSIC trial is an investigator‐initiated, pragmatic, international, parallel‐grouped, centrally randomised, stratified, analyst‐blinded trial with adequate allocation sequence generation, and allocation concealment.
The protocol has been prepared according to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines.[
Eligible patients fulfilling all inclusion criteria and no exclusion criteria are randomised 1:1 using a centralised web‐based system according to a computer‐generated allocation sequence list, the stratification variables, and varying block sizes. The allocation sequence list and block sizes are only known by the data manager at the Copenhagen Trial Unit and remain concealed from the investigators until the last patient has completed the follow‐up.
The trial intervention is not blinded for investigators, clinical staff, and patients, as blinding of IV fluid restriction versus standard care is not feasible. The statistician assessing all outcomes will be masked for the allocation. The data for 90‐day and 1‐year mortality will be collected from electronic patient records or central national registries in most participating countries. The remaining outcomes will be provided by local investigators, who are not blinded for the intervention, by entering the data from patients' files. The management committee will write the preliminary abstract with the group allocation masked; this abstract will be published as a supplement to the primary trial report.[[
Patients must fulfil all the inclusion criteria;
- Age ≥18 years.
- Admitted or planned admitted to the ICU regardless of trial participation.
- Septic shock defined according to the Sepsis‐3 criteria:[
22 ] - o Suspected or confirmed site of infection or positive blood culture.
- o On‐going infusion of vasopressor/inotrope agent to maintain a mean arterial blood pressure of 65 mm Hg or above.
- o Lactate of 2 mmol/L or above in any plasma sample performed within the last three hours.
- At least 1 L of IV fluid was administered (all crystalloids, colloids or blood products including that given with medication and nutrition) in the last 24 hours prior to screening.
A detailed description of the inclusion criteria is available in Supplementary Appendix 1.
Patients who fulfil all inclusion criteria and one or more of the exclusion criteria below will be excluded;
- Septic shock for more than 12 hours at the time of screening
- Life‐threatening bleeding
- Acute burn injury of more than 10% of the body surface area
- Known pregnancy
- Consent not obtainable as per the model approved for the specific site
A detailed description of the exclusion criteria is available in Supplementary Appendix 2.
The CLASSIC intervention period is the time spent in ICU from randomisation to a maximum of 90 days. If a patient is discharged and readmitted to a CLASSIC ICU within the 90‐day period, the allocated fluid intervention continues.
No IV fluids should be given unless one of the below extenuating circumstances occurs; in these cases, IV fluid may be given in measured amounts:
- In case of severe hypoperfusion or severe circulatory impairment defined by either:
- Lactate 4 mmol/L or above _B_OR
- Mean arterial blood pressure below 50 mm Hg (with or without vasopressor/inotrope) _B_OR
- Mottling beyond the kneecap (mottling score >2, Supplementary Appendix 3) _B_OR
- Urinary output less than 0.1 mL/kg bodyweight/h, but only in the first 2 hours after randomisation A bolus of 250‐500 mL of IV crystalloid solution may be given followed by reevaluation. These criteria identify patients at increased risk of death,[[
23 ]] and were found feasible and not associated with harm in the CLASSIC pilot trial.[11 ] - In case of overt fluid losses (eg, vomiting, large aspirates, diarrhoea, drain losses, bleeding or ascites tap) IV fluid may be given to correct for the loss, but not above the volume lost
- In case the oral/enteral route for water or electrolyte solutions is contraindicated or has failed as judged by the clinical team, IV fluids may be given to:
- Correct dehydration or electrolyte imbalances
- Ensure a total fluid input of 1 L per 24 hours (fluids with medication and nutrition count as input)
IV fluids may be given as a carrier for medication, but the volume should be reduced to the lowest possible volume for the given medication. Oral/enteral fluids are not restricted in the IV fluid restriction group.
There will be no upper limit for the use of IV or oral/enteral fluids. In particular:
- IV fluids should be given in the case of hypoperfusion or circulatory impairment and should be continued as long as hemodynamic variables improve including static or dynamic variable(s) as chosen by the clinicians. These criteria are based on the Surviving Sepsis Campaign guidelines[
1 ] - IV fluids should be given as maintenance if the ICU has a protocol recommending maintenance fluid
- IV fluids should be given to substitute expected or observed loss, dehydration or electrolyte imbalances
The primary outcome measure is all‐cause mortality at day 90 after randomisation.
We have 7 secondary outcome measures and are as follows:
- Number of participants with one or more SAEs in the ICU defined as ischaemic events (cerebral, cardiac, intestinal, or limb ischaemia) or a new episode of severe AKI (modified KDIGO3[
26 ]) - Number of participants with one or more SARs to IV crystalloids in the ICU (Supplementary Appendix 5)
- Days alive at day 90 without life support (vasopressor/inotropic support, invasive mechanical ventilation, or renal replacement therapy [RRT])
- Days alive and out of the hospital until day 90
- All‐cause mortality at 1 year after randomisation
- HRQoL at 1 year after randomisation measured using the EuroQoL EQ‐5D‐5L scores[
27 ] - Cognitive function at 1 year after randomisation as assessed by the telephone MINI‐Montreal Cognitive Assessment (MINI‐MoCA) score[
28 ]
Detailed definitions of the outcome measures are available in Supplementary Appendix 4.
• Sex
• Age
- Date of admission to the hospital
- Date and time of the first ICU admission during the current hospitalisation
- From where the participant was admitted to ICU
- The focus of infection (pulmonary, gastrointestinal, urinary tract, skin or soft tissue or other infectious focus documented or suspected including meningitis, endocarditis, osteomyelitis, arthritis, and bacteraemia)
- Comorbidities prior to ICU admission (active hematological malignancy, metastatic cancer, ischemic heart disease or heart failure, chronic hypertension, or chronic RRT)
- Blood values, interventions, and vital parameters including participant weight, highest plasma lactate within 3 hours prior to randomisation, highest dose of noradrenaline within 3 hours prior to randomisation, volume of IV fluid 24 hours prior to randomisation, use of systemic corticosteroids 24 hours prior to randomisation, highest plasma creatinine value within 24 hours prior to randomisation, and the use of acute RRT 3 days prior to randomisation and habitual plasma creatinine value prior to current hospitalisation
- Values for the Simplified Mortality Score in ICU (SMS‐ICU),[
29 ] which is based on 7 variables obtained within the 24 hours prior to randomisation (age, lowest systolic blood pressure, acute surgical admission, hematologic malignancy or metastatic cancer, and the use of vasopressors/inotropes, respiratory support (invasive or noninvasive mechanical ventilation) and RRT)
- Fluid input and output cumulated from the 24‐hour ICU charts
- The total volume and specific type of IV isotonic crystalloids
- The total volume of other IV fluids: eg, <10% glucose
- The total volume of albumin
- The total volume of fluids with medications: Both parenteral and enteral
- The total volume of fluids with enteral and parenteral nutrition
- The total volume of nonnutritional enteral/oral fluids
- The total volume of blood products
- Urinary output on this day
- Any fluid volume removed during RRT
- The total volume of other losses on this day including drainage, aspirates, stools, and bleeding
- Major protocol violation on this day:
- Restrictive group: IV fluids were given without one of the extenuating circumstances on this day
- Standard care group: The violations (no IV fluid given) will be assessed from all day form registrations regarding the fluid input at the end of the trial
- Infusion of vasopressor or inotrope on this day
- The use of systemic corticosteroids on this day
- The use of invasive mechanical ventilation on this day
- The use of RRT on this day
- Plasma concentration of creatinine on this day
- SAEs (at least one episode of either the following):
- Ischemic events defined as either
- Cerebral ischemia
- Acute myocardial ischemia
- Intestinal ischemia
- Limb ischemia
- A new episode of severe AKI (defined as modified KDIGO3,[
26 ] Supplementary Appendix 4)
• SARs:
- General tonic‐clonic seizures
- Anaphylactic reactions
- Central pontine myelinolysis
- Severe hypernatremia defined as p‐Na >159 mmol/L
- Severe hyperchloraemic acidosis defined as pH <7.15 AND p‐chloride >115 mmol/L
- Severe metabolic alkalosis defined as pH >7.59 AND SBE >9 mmol/L
- Death (y/n, if yes, date of death)
- Date of discharge from the hospital
- Additional hospital admissions
- Death (y/n, if yes, date of death)
- EQ‐5D‐5L and EQ‐VAS score
- Telephone MoCA score
Detailed definitions of the registered variables are available in Supplementary Appendix 6.
If less than 5% of the data required for any specific analysis on primary or secondary outcomes are missing, a complete case analysis will be performed. If more than 5% are missing, and it is concluded that data are not "missing completely at random" (MCAR criterion),[[
In any multiple imputation, we will use all relevant outcomes and the stratification variables (site and metastatic or hematologic cancer), SMS‐ICU[
The analyses will be done in the ITT population defined as all randomised participants for whom there is consent for the use of data. The conclusion of the trial will be based on the ITT analysis.
The per protocol population is defined as the ITT population except those having one or more major protocol violations.
Four analyses will be performed for the primary outcome:
- Logistic regression analysis adjusted for the stratification variables 34 (site and hematologic malignancy/metastatic cancer) in the ITT population. P values will be two‐tailed
- 2. A two‐tailed logistic regression analysis adjusted for the stratification variables, the SMS‐ICU[
29 ] focus of infection (other foci vs. urinary tract infection)[[29 ], [35 ]] and use of corticosteroids[36 ] - 3. A two‐tailed logistic regression analysis in the per protocol population adjusted for the stratification variables
- 4. Two‐tailed logistic regression analyses in the pre‐planned subgroups adjusted for the stratification variables
We will report absolute and relative risk ratios with 95% CIs for the primary analysis of the primary outcome (analysis 1) (P value <.05), computed using GLM models with appropriate link functions and binomial error distribution. Further, we will report the crude event rates in each group and a Kaplan‐Meier survival curve for the crude data of the primary outcome.
Dichotomous secondary outcomes will be analysed in the same way as the primary analysis for the primary outcome, ie,:
- SAEs Two‐tailed logistic regression adjusted for the stratification variables in the ITT population
- SARs Two‐tailed logistic regression adjusted for the stratification variables in the ITT population
- 1‐year mortality Two‐tailed logistic regression adjusted for the stratification variables in the ITT population
A Kaplan‐Meier survival curve will be reported for the crude data for secondary outcome 1‐year mortality.
The remaining secondary outcomes are continuous measures which we expect to be skewed (nonnormally distributed), because of inflation of specific values such as zero for "days alive outside hospital" for all patients who die while in the ICU. The outcome measures will be analysed as follows:
- Days alive at day 90 without life support Generalised linear model or nonparametric test stratified for the site in the ITT population
- Days alive and out of the hospital at day 90 Generalised linear model or nonparametric test stratified for the site in the ITT population
- HRQoL 1‐year after randomisation[
27 ] Generalised linear model or nonparametric test stratified for the site in the ITT population - Cognitive function 1‐year after randomisation[
28 ] Generalised linear model or nonparametric test stratified for the site in the ITT population
For the generalised linear model we will initially use Poisson distribution, alternatively negative binomial.[
The following secondary outcome measures are composite; SAEs, SARs, and days alive at day 90 without life support. We will report each component of these outcomes in an appendix to the primary publication without P‐values due to the lack of adjustment for multiple comparisons.
We will report absolute and relative risk ratios with 99% CIs for dichotomous secondary outcomes. For continuous secondary outcomes we will report mean differences with 99% CIs if they are normally distributed and medians with 99% percentile‐based bootstrapped CIs for nonnormally distributed continuous secondary outcomes (P value.01) due to the multiplicity of these. Definition of the level of significance for the secondary outcomes is available in Supplementary Appendix 7.
We plan to enrol 1554 (2 × 777) patients to be able to show a 15% relative risk reduction (RRR) (7% absolute) in the restrictive group on the primary outcome from an estimated 45% 90‐day mortality in the standard care group at type 1 and 2 error levels of 5% and 20%, respectively, corresponding to a number needed to treat (NNT) of 14 or less. The anticipated RRR of mortality and the estimated mortality in the control group are based on the data from previous RCTs and systematic reviews.[[
Trial sequential analysis of existing trials (n = 9) have shown that 15% of the required information size of 4115 patients to detect or reject a 15% RRR in all‐cause mortality with lower versus higher fluid volumes has been reached. (Meyhoff 2019, in preparation)[
We expect to have the following statistical power for the secondary outcomes based on 2 × 777 participants, a type 1 error level of 1% and a RRR of 15% in the fluid restriction group versus standard care group:
- 50% power for the number of participants with one or more SAEs (control event rate 25%[[
39 ]]) - 10% power for the number of participants with one or more SARs (control event rate 5%[[
39 ]]) - 80% power for the mortality at 1 year (control event rate 55%[[
39 ]])
The estimates of the control event rates originate from the data of previous septic shock trials.
We expect the following outcomes to be highly skewed (nonnormally distribution): Days alive without life support and out of the hospital at day 90 and HRQoL and cognitive function at 1 year. As we lack sufficient knowledge on the details of the nonnormal distribution, no realistic power analysis can be provided. We therefore refrain from this in order to avoid creating a false impression of precision.
We plan to assess the heterogeneity of intervention effects of the primary outcome in the following 5 subgroup analyses based on patient characteristics at baseline:
- Respiratory support at randomisation (hypothesised increased effect of fluid restriction in patients receiving respiratory support)
- Severe AKI defined as modified KDIGO2 or above[
26 ] at randomisation (hypothesised increased effect of fluid restriction in patients with severe AKI)[26 ] - Severe metabolic failure at randomisation defined as the plasma lactate level above 4 mmol/L (hypothesised increased the effect of fluid restriction in patients with severe metabolic failure)
- Participant weight at randomisation with higher weight defined as bodyweight (measured or estimated) ≥76 kg[[
11 ], [39 ]] versus lower weight as <76 kg (hypothesised increased effect of fluid restriction in patients with lower weight) - Patients who received ≥30 mL/kg body weight IV fluids in the 24 hours prior to randomisation versus patients who received a lower volume (hypothesised increased the effect of fluid restriction in patients with less fluids given 24 hours prior to randomisation)
For all subgroups a P value <.01 in the test of interaction will be considered statistically significant. Detailed definitions of the subgroups analysis are provided in Supplementary Appendix 8.
At the trial completion, the flow of trial participants will be reported according to the Consolidated Standards of Reporting Trials (CONSORT) statement.[
A Data Monitoring and Safety Committee (DMSC) has been formed, consisting of independent ICU trialists/clinicians and a biostatistician who collectively have experience in the management of ICU patients and in the conduct, monitoring, and analysis of RCTs. The charter for the DMSC is available from Supplementary Appendix 9.
We will perform 3 interim analyses:
- Interim analysis when 10% of patients have completed 30‐days follow‐up
- Interim analysis when 30% of patients have completed 30‐days follow‐up
- Interim analysis when 50% of patients have completed 90‐days follow‐up
For the first two interim analyses, the DMSC will evaluate data on:
- Fluid volumes and protocol violations
For the third interim analysis, the DMSC will evaluate data on:
- Fluid volumes, protocol violations, 90‐day mortality, and rates of SAEs and SARs in the ICU
The DMSC will be provided with the following masked (as group 0 and 1) data from the coordinating centre:
- Number of patients randomised
- Number of patients randomised per intervention group
- Number of patients stratified per stratification variable per intervention group
- Number of events, according to the outcomes, in the two groups
Based on the evaluation of these outcomes, the DMSC will decide if they want further data from the coordinating centre. The DMSC can, at any time during the trial, request the distribution of events, including outcome measures and SARs according to intervention groups. Further, the DMSC can request unblinding of the interventions. Additionally, the DMSC will be informed yearly about SARs in the two groups of the trial.
The interim analyses will be performed by an independent statistician. The DMSC may recommend pausing or stopping the trial if a group difference in the primary outcome measure, SARs or SUSARs are found in the interim analyses with statistical significance levels adjusted according to the LanDeMets group sequential monitoring boundaries based on the O'Brien Fleming alpha‐spending function.[
Furthermore, the DMSC can recommend pausing or stopping the trial if continued conduct of the trial clearly compromises participant safety. The management committee will make the final decision regarding the continuing, pausing or stopping of the trial. However, stopping for futility to show an intervention effect of 15% RRR or RRI will not occur, as intervention effects <15% RRR or RRI in all‐cause mortality may be clinically relevant.
The optimal strategy for IV fluid therapy in adults with septic shock is largely unknown and the existing data on lower versus higher fluid volumes are of low/very low quality.
Together with existing data and the data from other on‐going RCTs,[[
Prior to the initiation of the CLASSIC trial, we have assessed current evidence in a systematic review[
The CLASSIC trial is a large international trial designed to provide data on the effect of fluid volumes on patient‐important outcomes with the lowest possible risk of bias. The trial is monitored according to the standards of GCP.[
The CLASSIC trial intervention is not masked for investigators, clinicians, and patients, as blinding of different fluid strategies is not feasible. This increases the risk of bias,[[
Our trial participants may be subjected to protocol violations, expected to occur more frequently in the restrictive group. In general, protocol violations may be more frequent in trials of complex interventions in ICU.[[
In conclusion, the CLASSIC trial is a large European trial assessing the effects of IV fluid restriction versus standard care on patient‐important outcome measures in adult ICU patients with septic shock. The trial results will provide important evidence to guide IV fluid therapy in this critically ill population.
The CLASSIC trial is registered at the European Clinical Trials Database (2018‐000404‐42) and
The final de‐identified dataset used for analysis will be made publicly available 9 months after the publication of the outcome data according to the recent ICMJE recommendations.[
After the completion of the CLASSIC trial, all trial results will be submitted to a peer‐reviewed medical journal irrespective of the direction of the results. Further, the trial results will be published at
The Trial was initiated on November 27th 2018 and is expected to be completed within 2 years.
We would like to thank the clinical staff at all ICUs who participate or will participate in the CLASSIC trial and all patients and relatives who have provided consent for the trial.
The Department of Intensive Care, Rigshospitalet receives support for research from Ferring Pharmaceuticals.
GRAPH
By Tine Sylvest Meyhoff; Peter Buhl Hjortrup; Morten Hylander Møller; Jørn Wetterslev; Theis Lange; Maj‐Brit Nørregaard Kjær; Andreas Bender Jonsson; Carl Johan Steensen Hjortsø; Maria Cronhjort; Jon Henrik Laake; Stephan M. Jakob; Marek Nalos; Ville Pettilä; Iwan Horst; Marlies Ostermann; Paul Mouncey; Kathy Rowan; Maurizio Cecconi; Ricard Ferrer; Manu L. N. G. Malbrain; Christian Ahlstedt; Søren Hoffmann; Morten Heiberg Bestle; Lars Nebrich; Lene Russell; Marianne Vang; Michael Lindhardt Rasmussen; Christoffer Sølling; Bodil Steen Rasmussen; Anne Craveiro Brøchner and Anders Perner
Reported by Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author