A Black Water Day ?

Mortality after Fluid Bolus in African Children with Severe Infection. Maitland, K et al. New England Journal of Medicine, May 26th 2011 (epub ahead of print) (DOI: 10.1056/NEJMe1105490)
An interesting paper published in this weeks NEJM will cause substantial comment and concern after it’s headline result showed increased mortality with rapid fluid resuscitation in paediatric sepsis. This surprising result is potentially extremely important as, if verified, undermines much of peadiatric (and adult)emergency care.

Before examining the paper in detail it’s worth making a few comments on what we think we know already….

  1. Septic shock is multi-factorial comprising relative hypovoleamia, vasodilatation and cardiogenic shock. These aspects are of different importance in different populations, aetiologies and even within the same patient at different points of their disease.
  2. Cardiac performance in hypovolemaic patients is supply dependent (volume responsive).
  3. Traditional measures of volume responsiveness have been shown to be inaccurate (HR, MAP, CVP), instead dynamic measures are required (fluid challenge, passive leg raising, or predictive measures of fluid responsiveness – SVV, PPV, FTc etc)
  4. Tissue oedema (secondary to capillary leak & reduced oncotic pressure) may occur whilst the cardiovascular system remains fluid responsive (amazing how often this key fact, which is after all enshrined in the very definition of ARDS, seems to be forgotten).
  5. Microvasular flow is grossly abnormal in septic states, there is wide variation amongst organ beds, pathologies and patients. Hence perfusion may be poor despite systemic “normal heamodynamics”.
  6. Even if perfusion is good, cellular (mitochondrial) function may be impaired.
  7. Timing of resuscitation is important, perfusion re-established late will be futile, as Shoemaker said “even oxygen can’t resuscitate dead cells”. Early “goal directed” therapy has been shown to be effective in adult septic shock (although see later) and peri-operatively.
  8. Late resuscitation may even be harmful as it potentially risks tissue oedema and reactive oxygen species with no additional benefit (see 2 recent trials – hyperoxia following cardiac arrest and the FACTT trial in ARDS).
  9. Targeting supra-normal perfusion has been shown to be harmful in adults with septic shock.

OK, with that in mind what did they do in this study?

This was a multi-centre, randomised trial conducted entirely in Africa (Uganda, Kenya, Tanzania) and was not blinded. children with febrile illness and impaired perfusion were randomised to receive either a bolus of 20/kg of saline or albumin, or no bolus, on admission to hospital. The huge majority of those enrolled were not hypotensive (hypotensive patients all received bolus, but were randomised to fluid type and received 40mls/kg). Impaired perfusion was defined as a capillary return > 3 seconds, limb temp gradient, weak volume pulse or tachycardia. Malnourished children and those with gastroenteritis were excluded (clear evidence exists in these groups). 57% of those enrolled had a diagnosis of malaria. There were no intensive care facilities available. Primary end point was 48 hour mortality with raised ICP, pulmonary oedema and 4 week mortality secondary outcomes. The study was stopped early (at 3141 patents, planned 3600) with an excess mortality in both bolus arms (10% mortality vs 7.3%, RR 1.45 (1.13-1.86)). Follow up was remarkably good for a trial in the 3rd world. The patients were young (mean age 24 months), severely anaemic (mean Hb 7),acidotic and in respiratory distress (83%) on enrolment.

A few questions spring to mind….

  1. Is there reason to suspect that malaria may behave in a significantly different way to other forms of septic shock, ie is the result not generalisable ? Yes. Malaria is known to carry a significant risk of cerebral oedema, hence fluid administration may risk increased intercranial pressure, with predictably worse outcomes. This effect would be magnified if either there were insufficient safe guards in place to identify this complication, or a lack of resource to deal with it. There was no increase in intercranial pressure reported in any of the cohorts, but critically it’s not clear how this was monitored or diagnosed (access to CT and ICP monitoring clearly non existent). The reported rates of raised intercranial pressure was approx 2% and yet 40% of patients had a seizure during treatment. Clearly a majority of these are likely to represent febrile convulsion, but I would argue that the actual incidence f cerebral oedema is unknown.
  2. Likewise, malaria (black water fever) is also associated with heamolysis and renal failure, and patients resuscitated with clear volumes with either complication could be expected to potentially do worse. The lack of intensive care back up means that patients who develop pulmonary oedema , however mild, temporary or reversible, could be expected to be at significant extra risk which may well be greater than any benefit. Respiratory distress was the norm on enrolment, yet the reported rates of pulmonary oedema was only approx 15%. Obvious haemolysis was present in 15% and jaundice in 30%. Again rates of anaemia, acidosis and lactate were reported as being equal across all groups which would tend to refute this as a significant factor, but would again be dependent on the accuracy of monitoring and diagnosis. Patients with HB < 5 were transfused, but this was after the clear volume loading (resuscitation in the first world would very likely be with blood at an earlier stage). Deaths were reviewed but only post-hoc and it’s not clear how deaths at least partly attributable to pulmonary odema or intercerebral oedema would be have been picked up.
  3. Should we be concerned that no sub group showed an improvement? Yes. I’m not going to get into the sub-group fishing argument here, but a massive study of peadiatric sepsis should be relatively safe from most of them. The fact that even in sub-groups with pneumonia and elevated lactate showed excess mortality is very concerning. The subgroup with normal lactate (<5) showed compounded RR mortality of 2.17, so there is a “dose response” signal here.
  4. Is the use of saline or albumin problematic ? Maybe. Saline is known to cause metabolic acidosis but the physiological importance is disputed, and in any case seems unlikely to be significant here. Albumin has previously been suspected to have an adverse effect on mortality in shock (BMJ meta-analysis) but the ANZICS SAFE study (7000 adult patients) was reassuring. Interesting sub-group analysis of this study showed patients with traumatic brain injury (another condition with disordered cerebral blood flow and capillary function) did have an increased mortality. Therefore potentially albumin in the setting of cerebral oedema may be harmful. Even if this were important it wouldn’t explain the excess mortality in the saline cohort.
  5. Can this study be applied to children with hypotension & septic shock ? No. There were only 30 patients in the entire trial who were hypotensive, all of whom received bolus fluids.
  6. Does the result inform adult practise ? Unknown. Young children have relatively fixed stroke volume, instead regulating cardiac output via HR, therefore their response to fluid loading will be different, I doubt this is significant. The aetiology, pathogenesis and additional co-morbidities present in adult populations are are far more importance and limit the applicability. We await the results of the upcoming major GDT trails with interest (PROMISE and OPTIMISE recruiting at NUH now or in the near future).
  7. Should practise in the 3rd world change as a result of this study ? I’m not qualified to say, but probably.
  8. Should we change practise on the basis of this study ? No. The major concern is that all of the “excuses” for a poor result above are of marginal importance which leaves us with 2 possibilities a) it’s a fluke b) it’s real and early volume expansion in this setting is associated with a mortality increase. The latter may be true but that should not lead to a restriction of fluids in our practise in the UK. The failure of the vast majority of therapies and trials in intensive care medicine over that last decade (with a few notable exceptions) can be partly explained by the misguided and naive application of a single interventions to a massively disparate cohort of patients. We know that the involvement of an experienced intensivist reduces mortality (Pronovost’s work) and this is probably because of the ability to apply the evidence base in an individualised manner. The factors listed above explain why an individual patient may not respond to a therapy that seems grounded in physiology and common sense. Rivers work itself was the application of an algorithm approach to maximising physiological function, but as been pointed out elsewhere the main intervention may well have been Rivers himself (and his team). Patients matching enrolment criteria for this study would certainly have been admitted to intensive care in this country, very likely intubated for a period of time and certainly had their fluid administration individualised. I still think that would be correct, although the global quality, type and timing of administration in our practise may be suboptimal. This trial doesn’t help us there.
  9. Is there now equipoise for a study in the first world ? Maybe, any of my peadiatric critical care colleagues wish to comment ?
  10. Is it ethical, and how does this alter the ethics of future trials ? (I’ve some reservations about the ethics here, but having had insufficient time to consider then I’ll leave that topic for another post!)

The future of trials in the critically ill is the identification of features which predict the response of a patient to a given therapy, followed by its logical application. These predictors may be pathological (steroids in pneumocoocal meniginits, immunoglobulin in MRSA PVL pneumonia), physiological (as in SVV, PPV), genetic (as in oncology), chronological (fluid in ARDS), biochemical (BNP, SB100, aPC levels etc) or related to co-morbidity. It seems very unlikely that the application of a magic bullet to “sepsis” will be successful, recent advances in mortality reduction have instead come from a package (bundle) of interventions applied across populations but importantly after consideration of their individual applicability (sepsis, VAP bundles). Having said all that, if someone had asked me to pick a single magic bullet to reduce the mortality of paediatric sepsis in the 3rd world I would have chosen early fluid bolus. Now I’m not so sure.


2 comments so far

  1. danharvey on

    I know it’s been a while since the last post, it’s been a very busy few months! Hopefully will be able to post more frequently on Critical Insight over the next few months.

  2. Andy Norris on

    Dan, very thoughtful analysis. Death within 48hrs is pretty heavyweight endpoint. Clearly this was a very ill cohort. I do think is worrying that an intervention that would “normally” be expected to improve a physiological derangement doesnt necessarily help. The use of norepinephrine appears to have lasted longer than most, but many rapid interventions do seem to cause trouble…i hastened the demise of a few people with mechanical ventilation, correction of electrolyte and pH

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: