Maitland et al published an impressive trial (1) on early fluid resuscitation. Now (2) they present an in depth analysis on excess mortality in the intervention group.
We do think that an explanation for stratum A is not quite as complicated and was partly suggested by Ford and Vishram (3) without being fully discussed. In spite of a robust looking randomization and the notion from Maitland et al (4) that anaemia cannot be held responsible for excess mortality, this is not really the case. Severe anaemia challenges the very randomization.
Looking into the published appendix (5) we see that in the first hour children with an Hb lower than 5 g/dl received in the albumin bolus group their blood only in 26 cases - compared to 207 children in the non-bolus group. Nevertheless the children without transfusion and an Hb lower than 5 g/dl received 20ml/kg of albumin (or even 40 ml/kg after the amendment of the protocol).
Assuming a blood volume of 80 ml/kg and accepting different opinions on pharmacokinetics we will end up with different amounts of plasma volume expansion after one hour - when the blood transfusion was finally initiated in the bolus group for children with an Hb lower than 5 g/dl. The plasma expansion and the consecutive reduction of Hb and Hk will depend on the respectively used pharmacokinetic model and its prediction on the percentage of non-blood fluids staying in the intravascular compartment. This will be between 100% (leading to an Hb of 3.9 g/dl or 3.3 g/dl in the amended group) and 0 % (leading to an Hb of max. 4.99 g/dl).
In this moment than FEAST effectively randomized a group of children with an Hb of 3.9 - 4.99 g/dl (or even 3.3 - 4.99 g/dl in the amended group) against children with an Hb of 5 or more (the children of the non-bolus group with an initial Hb of lower than 5 g/dl, who already received a blood transfusion).
The argument of Maitland et al that the blood volume per kilogramm given in the first hour was only minimally different between the two groups cannot really be followed (4) when we see that almost 8 times more children received blood in the non-bolus group than e.g. in the albumin bolus group (5) and that the argument does not relate to the fact that the non-blood receiving children were haemodiluted.
A child in the albumin group (323 children (1) with a Hb between 3.9/3.3 g/dl and 4.99 g/dl ) has not much reserves left.
10 additional deaths in the albumin group would render the study to statistical insignificance. This is even more important considering the fact that 5 children of the albumin bolus group died before they got the bolus (and nobody in the non-bolus group (1)). This is statistically correct (intention to treat) but should not be forgotten in the search for an explanation at least as long as a no real death analysis of these 5 children could be shown.
The beauty of the FEAST trial is a low overall mortality reached through the dedicated work of the collaborators - a result which should be emphasized much more.
References:
1. Maitland K et al.: Mortality after Fluid Bolus in African Children with Severe Infection, N Eng J Med; 2011;364;2483-95
2. Maitland K et al.: Exploring mechanisms of excess mortality with early fluid resuscitation: insights from the FEAST trial, BMC Medicine, open access, 14 March 2013
3. Ford S, Vishram A: Correspondence to the editor , N Eng J Med, 2011;365;14;1348
4. Maitland et al. Correspondence to the editor - The Authors Reply, N Eng J Med,2011;365;14;1348
5. Maitland K et al.: Supplementary Appendix to: Mortality after Fluid Bolus in African Children with Severe Infection, N Eng J Med; 2011;364;2483-95
Competing interests
A competing interest exists when your professional judgment about a paper could possibly be influenced by considerations other than the paper's validity or importance. Detail possible competing interests here...
FEAST - is the explanation really so complicated?
7 June 2013
Maitland et al published an impressive trial (1) on early fluid resuscitation. Now (2) they present an in depth analysis on excess mortality in the intervention group.
We do think that an explanation for stratum A is not quite as complicated and was partly suggested by Ford and Vishram (3) without being fully discussed. In spite of a robust looking randomization and the notion from Maitland et al (4) that anaemia cannot be held responsible for excess mortality, this is not really the case. Severe anaemia challenges the very randomization.
Looking into the published appendix (5) we see that in the first hour children with an Hb lower than 5 g/dl received in the albumin bolus group their blood only in 26 cases - compared to 207 children in the non-bolus group. Nevertheless the children without transfusion and an Hb lower than 5 g/dl received 20ml/kg of albumin (or even 40 ml/kg after the amendment of the protocol).
Assuming a blood volume of 80 ml/kg and accepting different opinions on pharmacokinetics we will end up with different amounts of plasma volume expansion after one hour - when the blood transfusion was finally initiated in the bolus group for children with an Hb lower than 5 g/dl. The plasma expansion and the consecutive reduction of Hb and Hk will depend on the respectively used pharmacokinetic model and its prediction on the percentage of non-blood fluids staying in the intravascular compartment. This will be between 100% (leading to an Hb of 3.9 g/dl or 3.3 g/dl in the amended group) and 0 % (leading to an Hb of max. 4.99 g/dl).
In this moment than FEAST effectively randomized a group of children with an Hb of 3.9 - 4.99 g/dl (or even 3.3 - 4.99 g/dl in the amended group) against children with an Hb of 5 or more (the children of the non-bolus group with an initial Hb of lower than 5 g/dl, who already received a blood transfusion).
The argument of Maitland et al that the blood volume per kilogramm given in the first hour was only minimally different between the two groups cannot really be followed (4) when we see that almost 8 times more children received blood in the non-bolus group than e.g. in the albumin bolus group (5) and that the argument does not relate to the fact that the non-blood receiving children were haemodiluted.
A child in the albumin group (323 children (1) with a Hb between 3.9/3.3 g/dl and 4.99 g/dl ) has not much reserves left.
10 additional deaths in the albumin group would render the study to statistical insignificance. This is even more important considering the fact that 5 children of the albumin bolus group died before they got the bolus (and nobody in the non-bolus group (1)). This is statistically correct (intention to treat) but should not be forgotten in the search for an explanation at least as long as a no real death analysis of these 5 children could be shown.
The beauty of the FEAST trial is a low overall mortality reached through the dedicated work of the collaborators - a result which should be emphasized much more.
References:
1. Maitland K et al.: Mortality after Fluid Bolus in African Children with Severe Infection, N Eng J Med; 2011;364;2483-95
2. Maitland K et al.: Exploring mechanisms of excess mortality with early fluid resuscitation: insights from the FEAST trial, BMC Medicine, open access, 14 March 2013
3. Ford S, Vishram A: Correspondence to the editor , N Eng J Med, 2011;365;14;1348
4. Maitland et al. Correspondence to the editor - The Authors Reply, N Eng J Med,2011;365;14;1348
5. Maitland K et al.: Supplementary Appendix to: Mortality after Fluid Bolus in African Children with Severe Infection, N Eng J Med; 2011;364;2483-95
Competing interests
A competing interest exists when your professional judgment about a paper could possibly be influenced by considerations other than the paper's validity or importance. Detail possible competing interests here...
No competing interests