Another interesting finding was that transcapillary filling appeared to be delayed by exogenous noradrenaline (by increasing the venous pressure therefore increasing hydrostatic pressure). The use of pressors in the case of traumatic bleeding is contentious, and in the days of aggressive resuscitation I’m not sure this effect is still relevant – it's still interesting though.
The clinical consequence of transcapillary filling is the usefulness of [Hb] to detect bleeding and to guide transfusion. If bleeding is recognised and i.v. fluid is given, a drop could be expected through haemodilution if nothing else. But if not immediately recognised (as can be the case in intensive care), a drop in [Hb] can only be expected if transcapillary filling has occurred – which the above studies tell us requires an uncertain and sometimes lengthy time period (I also wonder whether the low protein levels we see in critical care reduces transcapillary filling further). I can’t find any data for the critically ill, although I suppose by definition, studying unrecognised blood loss can be problematic! There is however data in the setting of trauma.
This study looked at the Hb concentration for all patients presenting to a level 1 trauma centre over a two-month period, and found that a [Hb] in the first 30 mins of <10 g/dl was associated with a 95% specificity for intervention, but that it was only 15% sensitive. This means that if the [Hb] is low, the patient is very likely to be bleeding BUT that relying on a positive test (low [Hb]) would lead to missing many bleeding patients. There are many weaknesses of this study and I don’t agree with the conclusions, but the fact that there were so many patients without a reduced [Hb] only reinforces that transcapillary filling occurs at a varying rate if at all.
The same authors conducted a second study looking at whether a low Hct predicted blood transfusion (to quote the authors verbatim “At first glance, it might appear that the practical impact of this study is minimal”). Even for this outcome, the sensitivity of a Hct <30 was only 29% - only 29% of patients requiring blood had a Hct <30 (a [Hb] of approx 10 g/dL). Giving a patient blood is a human behaviour rather than a physiological endpoint, but I can see how it is used as a surrogate for bleeding. Unfortunately, despite claims made throughout the article, the authors can only conclude that “… use of the initial Hct might not help clinicians identify all patients who will require a transfusion…”.
This study took a different approach. They looked at a cohort of approx. 200 trauma patients who required surgery within 4hrs. of presentation. 43% of patients with a Hct within the normal reference range (>40 for males) had an estimated blood loss of >1000ml (which I think might be an underestimate unless their theatres are ultra-efficient – something made the surgeons take them to theatre quickly and usually that’s only visceral injury or bleeding). Once again, the data in this paper doesn’t lead me to the same conclusion as the authors that Hct is useful to diagnose significant blood loss.
If we can make sense of this by saying that transcapillary filling might not have had enough time to occur when [Hb]/Hct is measured, one strategy might be to make serial measurements. Unfortunately, this doesn’t seem to help as this refreshingly honest paper shows (when reading it you can almost see the author banging their head off the desk!). The investigators found that repeating normal [Hb] measurements in 2 hours resulted in another normal result in 70% of patients. In those whose repeat [Hb] was low (14%) it was in keeping with a known diagnosis, and there were no clinical consequences in the 16% that did not have a repeat test.
This data shows that we cannot rely on a normal [Hb] as a rule out test (but we frequently do), and in the critically ill patient the sensitivity may be even lower than in trauma (this is only my suspicion, don’t take it as fact). If the [Hb] doesn’t drop but we suspect bleeding, what should we do when our mental models are based on the combination of a bleeding patient and a low [Hb]?
By definition the intravascular volume will be reduced – if we give volume (crystalloid) that will temporise matters but at the cost of subsequent tissue oedema (incidentally the ‘weakness’ of this strategy was shown in this paper in 1932!), so is giving blood the answer? We know that a high [Hb] target is harmful in critical illness in general, sepsis, and in upper GI bleeding, but we also know that aggressive resuscitation of the bleeding patient is required. If we give blood, it is conceivable that the [Hb] would be expected to rise further. If protein is key what about Albumin? Or hypertonic saline to increase osmolality? What about giving crystalloid to haemodilute thereby justifying subsequent blood? - I think that’s what often ends up happening even if not by intention.
All of this perhaps raises more questions than it answers, particularly for the critically ill patient on the unit with occult bleeding. Do we really see a fall in [Hb], and is transcapillary filling the cause? The more I read the less sure I am...
Finally, if you haven’t lost the will yet here’s a review of the pathophysiology of haemorrhagic shock and an here’s an article about the methods used to measure plasma volume.