The Revised Starling Principle: Implications for Rational Fluid Therapy
“Doctrine once sown strikes deep its root, and respect for antiquity influences all men.”
-William Harvey 
The use of hyperoncotic albumin to draw fluid from the interstitial space permeates dark corners of the critical care community. The ‘pull and push’ of 25% albumin followed by furosemide remains somewhat of a cryptic lore – its use often spoken of in hushed tones, as if this special physiology may be called upon only in the most dire of situations and only by the most venerable of clinicians. It is a physiology that I have invoked whilst managing patients with cirrhosis – or others in whom the mystical creature of ‘hypervolemic, yet volume deplete’ is raised.
Yet the data for this practice is mixed and contemporary – and brilliant – reappraisals of the original Starling principle of capillary filtration have seriously challenged the reasoning behind this practice.
In the late 19th century, Starling noted that isotonic saline injected into a dog’s hind limb was reabsorbed whereas, serum was not. From this, he deduced the capillaries and post-capillary venules are semipermeable membranes. Movement of fluid then became a competition between transendothelial hydrostatic pressure [that is, the pressure within the capillary [Pc] less the hydrostatic pressure within the interstitial space [Pi]] and the colloid osmotic pressure difference between the capillary and interstitial space [πc – πi]. The colloid osmotic pressure is determined largely by albumin and the degree to which albumin permeates the endothelium is reflected in the Staverman’s osmotic reflection coefficient [σ] which ranges from 0 [fully permeable] to 1 [impermeable]. We are left with the following – simplified – equation determining net fluid flux [Jv]:
Jv = [Pc – Pi] – σ[πc – πi] [note: I have knowingly omitted the hydraulic conductivity Lp]
If a ‘sum-of-forces’ approach is adopted, the following pictorial analysis may be employed [Figure 1A]. Note that the force favoring filtration is Pc while the summative force opposing filtration [Pco] may be expressed by the following equation
Pco = σ[πc – πi] + Pi
figure 1A: the hydrostatic pressure within the capillary [Pc] is represented by the sloped brown line. The sum-of-pressures opposing the Pc is the Pco, represented by the dotted red line in the middle. When the Pc is above the Pco, filtration occurs, when the Pc is below the Pco, absorption occurs. This is the traditional Starling Model.The capillary filtration opposition pressure [Pco] should be intuitive for if the capillary colloid osmotic pressure [πc] rises or if the interstitial osmotic pressure [πi] falls, fluid should be retained within the capillary. Similarly, if the pressure surrounding the capillary [Pi] rises, filtration is opposed. The Pco is illustrated by a dotted, red line on figures 1 & 2; if its value rises, filtration is opposed [reabsorption is enhanced] while if its value falls filtration is augmented. In the early 20th century, Pc was first successfully measured and found to be roughly 35-45 mmHg at the arterial end and 12-15 mmHg at the venular end. At that time, it was not possible to simultaneously measure πi and it was assumed to be quite low. Similarly, σ was assumed to be 1.0. Based on said assumptions, it was concluded that the Pc falls below the Pco in the middle of the capillary and therefore filtration predominates at the arterial end while absorption emerges at the venular end.
Figure 1B & 1C: Hypothetical changes in the opposition pressure. Note that the Pco may rise [B] in response to an increase in πc or Pi or a fall in πi. This favours absorption. Conversely, Pco will fall [C] secondary to a fall in πc or Pi, or a rise in πi. This favours filtration.Revised Model
However, when techniques became available to simultaneously measure all of the Starling forces, the Pco was found to be surprisingly low – due to the relatively high πi [i.e. 16 mmHg, note that increasing πi lowers the Pco] and low Pi [actually subatmospheric at heart level] such that the Pc remains above Pco throughout the entirety of the capillary; importantly this is also true for tissues with the lowest Pc [e.g. lung]. In other words, there is no absorption. This has been found true for most tissues. There are notable exceptions to the steady-state no-absorption rule, and these tissues include the intestinal mucosa [but not the mesentery], the renal cortex and medulla as well. These tissues manage to keep the πi quite low [which raises the Pco] such that absorption is observed.
Figure 2: The no-absorption rule [in the steady state]. Note that this occurs in the vast majority of capillaries. The high πi and low Pi both diminish the Pco such that Pc is > Pco throughout the capillary and filtration dominates [downward blue arrow].Transient versus Steady State
Capillary absorption can be seen in tissues which normally do not absorb along their length when there is a transient fall in Pc; however, within a period of minutes, the sum-of-forces returns to net filtration. This fact highlights the important linkage between Jv [i.e. fluid filtration], πi and Pi. When Jv falls in response to a drop in Pc, the colloid oncotic pressure of the interstitium πi, rises with time and the Pi falls. Consequently, the Pco falls and net filtration across the capillary is regained; this effect tends to occur within 30 minutes before net filtration is, again, achieved. In theory, the converse is also true, that a transient rise in Pc will momentarily augment filtration, but over a period of minutes the Pco will also rise – an effect which will buffer the initial increase in Jv.
Importantly, even when the revised model with simultaneously measured ‘sum-of-forces’ is utilized, there is still an order of magnitude difference between the predicted lymphatic flow and the observed lymphatic flow. Per the above model, the predicted filtration, and therefore afferent lymph drainage should be higher than what is observed. If the venular side of the capillary does not reabsorb in the steady-state, where is the excess filtrate going? It appears now that the colloid oncotic pressure difference which determines Jv, is no longer a trans-endothelial force per se, but rather an intra-endothelial force. This realization has come about in response to the presence of the endothelial glycocalyx [EG]. The EG is a mesh of mucopolysaccharides associated with proteoglycans and glycosaminoglycans; the EG acts as a brushy border within capillaries separating red blood cells and other large proteins from the sub-endothelial surface. In health, the EG may have a volume of 1700 mL. It is likely that the Staverman osmotic reflection coefficient represents the ability of this border to reflect albumin from the subendothelial space. Thus the modified Starling equation becomes:
Jv = [Pc – Pi] – σ[πc – πsg].
Normally, the colloid oncotic pressure of the subglycocalyx [πsg] is quite low, but this force is entirely within the capillary such that Jv across the endothelium is a function of Pc and Pi while the colloid osmotic difference across the EG simply retards filtration. The aforementioned principles still hold in terms of transient and steady-state effects, however, this raises the possibility that the hyperoncotic effect of albumin is simply to dehydrate the subendothelial space and EG rather than drawing any significant amount of fluid from the interstitium.
Figure 3: The glycocalyx model showing filtration throughout the capillary, but at a lower value due to the difference between the colloid osmotic pressure within the capillary [πc], less the low colloid osmotic pressure in the sub-glycocalyx space [πsg].Implications for practice
The revised Starling-Glycocalyx model explains why there is little difference in hemodynamic outcome and volume infused between colloid and isotonic crystalloid in a many number of trials. Because the colloid oncotic pressure differential is an ‘intra-endothelial’ force rather than ‘trans-endothelial’ the volume expansion effects of colloids are diminished as predicted by the traditional model. It is argued that the greater the reduction in Pc, the stronger the argument for isotonic crystalloid – which will ‘rehydrate’ the EG. The revised model thereby turns our focus to the pressure differential [Pc – Pi] as the key determinant of capillary filtration. Many patients in the intensive care unit are inflamed – for a variety of reasons. Inflammation dilates pre-capillary arterioles which increases Pc. Simultaneously, inflammation changes the characteristics of the interstitium – the extracellular matrix changes its characteristics, consequently increasing its compliance; thus, Pi is diminished and the trans-endothelial pressure differential rises. Ostensibly, the treatment of edema should focus on the underlying cause of inflammation. It also suggests a protective mechanism of alpha agonists [e.g. norephinephrine] which constricts arterioles, subsequently attenuating Pc. As well, keeping intra-thoracic pressure low should promote lymphatic drainage to the great veins [e.g. low central venous pressure].
The aforementioned physiology also calls into question the use of hyperoncotic albumin to draw fluid from the interstitial space, especially in the inflamed ICU patient [i.e. with a reduced Staverman coefficient]. An albumin bolus will raise Pc favoring filtration, however, the hyperoncotic effect of 25% albumin [which has an oncotic pressure ~ 4 times that of human plasma] is argued to oppose filtration and even cause resorption. In septic patients, 200 mL of 20% albumin resulted in an increase in plasma volume of 430 mL, with a maximum effect occurring in the first 30 minutes. There was an equally transient improvement in oxygenation during this time. However, it is entirely possible that the increase in plasma volume was due to dehydration of the EG layer rather than imbibing interstitial fluid. Additionally, the transient improvement in oxygenation may reflect improved oxygen delivery to the tissues with consequent increase in mixed venous oxygen saturation, as well as, diminished dead space perfusion. Importantly, the FADE Trial is set to better expand our knowledge here, but should albumin-furosemide not prove fruitful, it may well confirm that many of us, myself included, have been suffering from a ‘colloid delusion.’