Wednesday, 12 November 2014

FACTT - The Fluid and Catheter Treatment Trial in ARDS

Paper Link

New England Journal of Medicine (2006); 354:2564-75

At a glance

  • Intervention - conservative versus liberal approach to fluid administration protocols in acute lung injury
  • Randomised - yes via an automated system
  • Controlled - yes, head to head with a liberal "usual care" arm
  • Placebo used - no
  • Blinded - no
  • Population / Sample size calculation - 1000 patients in 20 academic north american intensive care units from 2000-2005
  • Groups same at randomisation - yes
  • Groups treated the same after randomisation - almost.  Higher transfusion requirements and possibly steroid use in liberal arm
  • Intention to treat analysis - yes
  • Follow-up - 99% to hospital discharge.  Assumed 60 day survival if discharged (see text)
  • Significant primary endpoint - no significant difference (25% v 28%) in 60 day mortality
  • Harms analysed - yes, no clinically significant differences found


The Background
Acute respiratory distress syndrome is a diffuse disease of the lungs that results in disordered pulmonary epithelial function.  Leaky pulmonary capillaries allow increased fluid and cell migration into the alveoli, resulting in impaired gas exchange.
If you buy into Starling’s forces theory, then increasing hydrostatic and falling oncotic pressure in the pulmonary capillaries is likely to exacerbate alveolar fluid leak, worsening this effect.
Prior to FACTT, there was speculation that “running the patient dry” might improve lung function.  This in turn may help shorten ventilation and reduce exposure to harm.  Animal studies had demonstrated that this effect was plausible.  Observational studies in humans with ARDS had also indicated lower vascular pressures and fluid balances were associated with improved outcomes.
However, there was equally concern that this strategy might lead to haemodynamic instability and potentially increased renal dysfunction.  Adding to this concern was the knowledge that most patients who die with ARDS actually die from non-pulmonary organ failure.  It was unclear where the balance of benefit and harm lay, and little consensus on the optimum fluid strategy existed.
FACTT sought to determine whether an aggressive approach towards a drier fluid balance after the resuscitation phase of ARDS would improve outcomes.  

The Trial

1000 patients with early ARDS were recruited between 2000 and 2005 in 20 academic North American ICUs.  They were randomised to one of 4 arms - 2 arms used a pulmonary artery catheter to guide therapy, and 2 used a CVC alone.  In this factorial trial design, the PAC group and the CVC group were subdivided into a dry and a wet group.  

ARDS was defined in a standard fashion, using the pre-Berlin definition of the day.

The results were presented in 2 separate papers.  The first paper compared the use of a CVC and a PAC in the fluid management of ARDS, and found that the use of a PAC did not improve outcomes, consistent with the findings of a number of other studies into PACs.  Consequently, the second paper analysed patients in just 2 groups - dry versus wet - and ignored the catheter method used.

The study was powered for a 90% chance to detect a 10% drop in mortality, assuming a 31% mortality in the "standard" therapy arm.  As it turned out, the actual mortality was only 25%, resulting in a somewhat underpowered trial.  The patients were similar at baseline and received similar care after randomisation, with the exception that blood transfusion was required significantly more often in the liberal group (39% vrs 29%).  There was also a trend towards higher steroid use in the liberal group (37% vrs 32% p=0.09), though the indication for this is unclear.  Ventilation was performed in an evidence based, lung protective fashion, and weaning from the ventilator was protocolised.

The trial used a highly protocolised approach to fluid management in these patients.  In short, after the patients became haemodynamically stable, they were fluid restricted and diuresed with frusemide (furosemide).  This strategy was continued so long as they were stable, until they achieved either a central venous pressure of 4cmH2O in the CVC group or a wedge pressure of 8.  The control group was treated similarly, but aimed for targets thought to reflect usual care, aiming for a CVP of 10-14 or a wedge pressure of 14-18cmH2O.  This protocol was commenced within 48 hours of ICU admission and continued for up to 7 days.

Of course, there is plenty of conjecture as to whether ANY of these markers (CVP, PCWP, urine output, skin perfusion, MAP etc) are reliable markers of tissue perfusion, which is what we're supposedly titrating our therapy to.

Compliance with the protocol was checked regularly and at least twice daily, and 90% compliance is reported in both arms of the study - not a bad effort.

Strangely enough, strictly following the policy could result in a haemodynamically stable patient with no evidence of defective circulation and a normal urine output receiving fluid boluses in the liberal group, based purely on their CVP or PCWP.  This doesn't appear consistent with common practice, but it is unclear how many patients were exposed to this.

The primary outcome was mortality at 60 days.  Secondary outcomes included ventilator free days, days free from ICU and a number of physiological variables associated with improved lung function.

The results
Overall, the patients were reasonably unwell, as suggested by an average APACHE 3 score of 94 and PaO2/FiO2 ratio of 150.  1/3 were shocked and receiving vasoconstrictor support at randomisation.  50% of the patients had ARDS due to pneumonia, a further 25% due to sepsis.  Importantly, each group had already achieved a positive fluid balance of 2.5L by randomisation.  

A major criticism of the paper was the exclusion criteria.  Of 11,500 patients screened, only 9% were included in the trial.  The most common reason for exclusion was that they already had a PAC in place (nearly 20%), and the second most was because they were excluded by their clinician (16%).  9% were rejected because they were already on renal replacement therapies.  Taken together, they raise the possibility of significant selection bias against sicker patients.

The paper achieved excellent separation of care between the groups.  At the end of the study, the dry group had an average change in fluid balance of -136ml, while the wet group were nearly 7L positive.  The fluid management of the control arm resulted in very similar fluid balances to other ARDS trials such as ARMA and ALVEOLI, confirming the authors' claim that this reflects "usual practice" at the time.  This separation was present whether the patient was shocked at the start of the trial or not.

The primary endpoint of death at 60 days was reported to be the same in both groups - 25% vrs 28%.  Sadly, this finding is undermined by the fact that the trial authors did NOT follow the patients out to this time point.  Strangely, the authors assumed that if a patient left hospital, they had survived to 60 days.  You will need to judge for yourself how to interpret this.

However, supporting the view that this strategy was, in fact, good for patients with ARDS, was that there was a significant reduction in ventilator free days (14.6 vrs 12.1) and days out of ICU.  Similarly, many respiratory physiology variables seemed to be improved.

The concerns that a restrictive fluid policy would be harmful appear unfounded, as there was no change in the rate of haemodynamic instability between groups (despite some minor changes in soft markers such as stroke volume), nor an increase in renal replacement.  In fact, there appears an increased renal replacement requirement in the wet group (14% vrs 10%), a very interesting finding indeed.  Perhaps waterlogged kidneys work less well than desiccated ones.

Another interesting note from the study was that african american and hispanic patients were significantly more likely to die after correction for baseline risk factors and illness severity.  Whether or not this is related to a genetic predisposition remains to be determined.

What does it all mean?

The authors stress the point that this is not an early goal directed therapy trial.  While 30% of patients still required resuscitation at randomisation, the protocol was not implemented until they had achieved stability.  The average time to first implementation of the protocol was nearly 24 hours after randomisation, over 43 hours after admission to ICU.

The study appears to confirm the belief that a positive fluid balance worsens pulmonary oedema, and that reducing this burden can improve ventilation parameters, allowing patients to be liberated from the ventilator sooner.  It also suggests that this can be done without exacerbating non-pulmonary organ failure.  However, these results were not sufficient to generate a mortality benefit in this study (though the issue of under powering may be partly to blame).

Several studies have also suggested this type of approach - its worth looking and Greg Martin's work examining albumin and frusemide to assist the removal of fluid in ARDS too.  It also compliments the work of other groups that demonstrate a markedly positive fluid balance is bad for patient outcomes, even after correcting for baseline risk factors (e.g. the SOAP study).

The major limitations of the study are that it is unblinded, there is a significant potential for selection bias against the sickest patients, followup was not technically complete and that it is somewhat underpowered.

Some commentators have questioned whether a more aggressive approach, leading to an overall negative balance might have made a bigger impact - further evaluation is required to confirm this.

There is one very interesting criticism of the paper, found in the letters to the editor link below.  The authors endorse this approach because it showed a reduction in ventilation time and ICU stay, despite not showing a mortality benefit.  One reader pointed out that the same group (the ARDS-Net group) had rejected the use of steroids in ARDS, despite remarkably similar results - read the interesting exchange of views below.

At the end of the day, it appears we should be drying our ARDS patients out a bit.  However, the endpoints for resuscitation (or de-resuscitation in this case) still remain incredibly nebulous and without validation - a fact that will plague these types of trials until a reliable indicator of tissue perfusion is developed.


Editorials and reviews