Efficacy and Safety of Colloids Versus Crystalloids for Fluid Resuscitation in Critically Ill Patients

BACKGROUND

Fluid resuscitation is a very common treatment in the ICU, and every day a thousands of critically ill patients are treated around the world with crystalloids or colloids to correct hypovolaemia (1, 2). A wide diversity of fluids is available, and new products are coming in the very near future. In 1989, a French consensus conference recommended to abandon the use of fresh frozen plasma and to limit the use of albumin to very specific situations (e.g. contraindication to other colloids, serum protein levels below 35 g/l) (1). Crystalloids and gelatins were considered as the best solutions for fluid resuscitation in the critically ill. In 1997, as starch was increasingly used, this guidelines were updated (2). It was concluded that isotonic crystalloids are as efficacious as starch pending the amount of fluid to be administered.

Three systematic reviews of the literature provided an accurate summary of data available from randomised controlled trials evaluating human albumin (3), or comparing crystalloids to colloids in fluid resuscitation in critically ill patients (4,5). The first systematic review (3) has analysed 30 randomised trials and concluded that there is no evidence that albumin administration reduces the risk of death in critically ill patients. By contrast, this review suggested that albumin administration might increase by 6% the risk of death. In the Cochrane Injury Review Group systematic review (4), 40 randomised trials were analysed. The authors concluded that the administration of colloids might be associated with an absolute increase of 3.8% in the risk of death. The last review has analysed only 17 randomised trials as they decided to exclude studies of hypertonic saline. This review showed no significant difference in the risk of death between colloids and crystalloids or in other outcomes like prevalence of pulmonary edema and length of stay at the ICU and at the hospital (5). However, when the authors considered only the studies with a high methodological quality score, they observed a nice trend toward an increased survival rate in favour of crystalloids. They reached the same conclusions for the subset of trauma patients.

JUSTIFICATION FOR THE STUDY

It seems reasonable to abandon the use of fresh frozen plasma in fluid resuscitation in critically ill patients. There are no definite criteria to decide which of crystalloids and colloids should be preferred. It is unquestionable that, to achieve a given haemodynamic effect, the amount of crystalloids needed is almost twice the amount of colloids (1, 2). Colloids and crystalloids have different effects on a range of important physiological parameters. As most of the critically ill patients underwent one or more volume replacements, even a small increase in the risk of death (around 5%) has considerable clinical implications. The meta analyses suggested a 5% absolute risk reduction in mortality in favour of crystalloids (4, 5). The number needed to treat, an estimate which is more clinically meaningful (6), is of 20, suggesting about 1 additional death (with colloids) for every 20 patients resuscitated. Given the number of patients exposed to fluid resuscitation, about 60 additional deaths might be observed per year in a ICU,receiving 600 patients a year.

A large trial is needed to compare the safety and efficacy of colloids and crystalloids (4,5).

OBJECTIVES

Primary objective:

To compare the effects on 28-day mortality of colloids versus crystalloids in ICU patients who need fluid resuscitation.

Secondary objectives:

ICU and hospital mortality and morbidity, and safety.

STUDY DESIGN

A multinational, randomised, controlled trial, on two parallel groups.

STUDY TREATMENTS

• Arm A: crystalloids
• Arm B: colloids Allocated treatment must be started immediately after randomisation (Day-0)
• The amount and speed of fluid loading will be at the physicians' discretion. The amount of starch should not exceed 30 ml/kg/24 hours. In case additional volume replacement is necessary, gelatins or albumin may be used. During all ICU stay, the patients will receive only crystalloids or only colloids for fluid resuscitation, according to randomisation.

Double blind seems unfeasible as the time window for inclusion is extremely short (treatment should be available promptly at bedside) and the amounts of volume replacement for all ICU stay could not be predicted a priori.

Allowed co-interventions:

• Any treatment required for a pre-existing condition
• Any type of inotropes or vasopressors
• Red cells, platelets, fresh frozen plasma, which are used should follow general guidelines (2).

Co-interventions not allowed:

Any other volume replacement solution than those above mentioned. Albumin is not allowed in patients allocated to crystalloids unless plasma albumin levels are below 20 g/dl (2).

ENDPOINTS

Main endpoint:

• 28-day mortality rate.

Secondary endpoints:

• ICU and Hospital mortality rates
• Number of ICU days the patient is alive and free of
• mechanical ventilation,
• vasopressors and inotropes,
• renal replacement therapy,
• organ system failure according to the sequential organ failure assessment (SOFA) scoring system[7]
• Total amount of volume replacement
• Difference in
• AUC of mean arterial pressure between HO and H24 at Day-0
• weight between Day-0 and Day-1 and Day-2
• PaO2/FiO2 ratio between H0 and H12 and H24, at Day-0
• chest X-ray score between day-0 and Day-1 and day-2(8)
• Frequency of adverse events
• Anaphylaxis reaction: skin or mucous rash, nausea, vomiting, shock, bronchial spasm, respiratory or cardiac arrest
• Coagulation disorders: fall in prothrombin time (PT) rate, in factor VIII
• Impairment in renal function: increase in creatinine > 200 µmol/l
• Length of stay at ICU
• Length of hospital stay

RANDOMISATION

The list of randomisation will be generated by computer. We will use block randomisation stratified by site and diagnosis,

1. trauma or haemorrhage,
2. sepsis,
3. other diagnoses.

Modalities:

• Day of randomisation = Day-0
• The delay between the decision to resuscitate the patients with fluids and randomisation should be as short as possible (15 minutes or less). Thus, the use of sealed envelopes seems to be the best method for allocation concealment in this case. There will be at each centre, 3 sets of sealed envelopes, one for each strata (i.e., trauma or hemorrhage, sepsis, and other diagnoses). To randomise a patient, the investigator must use the first available sealed envelope (according to allocation number) in the corresponding (i.e., trauma or hemorrhage, sepsis, other diagnoses).
• Choice of strata: when a patient has more than one diagnosis, the choice of the strata must be as follows: trauma or haemorrhage > sepsis > others. For example, a patient with trauma and sepsis will be randomised in the " trauma " strata.
• An envelope must be used only once.
• Investigators must declare by fax (to the coordinating centre) each inclusion within 2 working days, and provide the following information:

FOLLOW UP

From H0 (time of randomisation) to H24:

• Total amount of each type of fluid infused
• Number of red cell units
• Mean of mean arterial pressure at hourly intervals from H0 to H24
• PaO2/FiO2 ratio at H0, H12 and H24
• SOFA score
• PT time, (worse values)
• Plasma total proteins and albumin levels
• Adverse events (cf. supra)

Daily from Day-1 to ICU discharge:

• Patient's status: dead or alive
• Total amount of each type of fluid infused
• SOFA score
• Number of red cell units
• Surgical procedures
• Specialised radiographic procedures
• Adverse events (cf. supra)
• Chest X-ray score (annexe 6) and weight at Day-1 and day-2

At ICU discharge:

• Alive or date of death
• Length of ICU stay
• Time on mechanical ventilation
• Omega scores 1,2,3 and total (annexe 8 [14])
• Number and type of adverse events (cf. supra)
• Where the patient is discharged to: home, another ward in the same hospital, another hospital, rehabilitation centre, home for disabled people

At hospital discharge (if not the same as that of ICU discharge):

• Alive or date of death
• Length of hospital stay
• Where the patient is discharged to: home, another ward in the same hospital, another hospital, rehabilitation centre, home for disabled people

SAMPLE SIZE

This study is designed to show an absolute difference of 5% in 28-day mortality between colloids and crystalloids. Assuming a mortality rate of 20% in the crystalloids group, with alpha = 5% and beta = 10%, 1504 patients per treatment arm are needed. A total of 3010 patients will be enrolled. All randomised patients must be followed up (at least) till ICU discharge. One hundred active centres will be recruited in Europe. The participation of each centre will be around 3 to 6 months.

STATISTICAL ANALYSIS

Interim analyses and stopping rules

The boundaries of the sequential plan are drawn to demonstrate an absolute difference of 5% in 28-day mortality rate between the two treatment arms, assuming a 20% mortality rate in the crystalloids group, and with alpha and beta of 5% and 10% respectively. The analyses will be performed every 100 deaths. The figure displays the boundaries of the sequential plan. Briefly, Z represents the difference between the two groups and V the number of patients that have been included. When a boundary is crossed, the enrollments in the study must be stopped, and the conclusion depends on which boundary has been crossed (see figure). Simulations allow to estimate how many inclusions are saved: when difference in mortality rates is nil, 1109 patients have to be included to reach the conclusion. When difference is 5%, 1477 patients have to be included to reach the conclusion.

Final analysis

The final analysis will be performed according to the intention to treat principle, after inclusions in the study will be stopped.

Baseline characteristics of patients will be compared between the two treatment arms: categorical variables will be expressed as number and percentage and compared by Chi-2 tests, means, standard deviation, and range will be given for continuous variables, which will be compared by Student t tests.

Methods for analyses of efficacy and safety will depend on the type of outcome. Survival curves will be constructed according to the Kaplan-Meyer method, and compared by log-rank tests. The comparison will be adjusted the main prognosis variables with Cox models. Categorical variables will be compared by Chi-2 tests, and continuous variables Student t tests or analysis of variance for repeated measures.

STUDY ORGANISATION

Steering committee

PI: Djillali Annane (Garches), Statistician: Sylvie Chevret (Paris) Yves Cohen (Avicenne), Samir Jaber (Montpellier), Gilles Troché (Versailles) Fékri Abroug (Monastir, Tunisie) Olivier Lesur (Sherbrook, Canada),

Advisory Board:

Jean François Baron, François Feihl (Lausanne, Suisse), Jean Louis Vincent (Bruxelles, Belgique)

Safety and Efficacy Monitoring Board:

Edward Abraham (Birmingham, USA), Déborah Cook (Hamilton, Canada), Mervyn Singer (London, UK), Charles Sprung (Jérusalem, Israel)