Ultrafiltration as a mechanism of fluid removal

This chapter is related to the aims of Section H3(i) from the 2017 CICM Primary Syllabus, which expects the exam candidate to "describe the principles of dialysis and renal replacement fluid". The process of fluid removal by ultrafiltration is certainly one of these foundational concepts which should probably be explored early in the course of ICU training, and it integrates nicely into the renal physiology section because this is also something the glomerulus does. Having said this, no question in the primary exam has ever asked anything about dialysis, let alone ultrafiltration or techniques of fluid removal. In contrast the fellowship exam has brought this up in one instance so far, that being Question 30 from the second paper of 2014. By some completely arbitrary flight of whimsy, the author has decided to put a short summary of dialysis terminology in the Part II revision section, and this detailed exploration of dialysis mechanisms into Part I. This is less reflective of what content needs to be known at the primary exam level, and more reflective of what the author wishes he knew when he had first started this job.

Of the latter, the salient points can be summarised as follows: 

• Ultrafiltration is filtration through a semipermeable membrane, where small particles and macromolecules are separated from the body fluid water
• Ultrafiltration is driven by a pressure gradient between the blood compartment and effluent compartment
• The pressure gradient is generated by positive pre-filter blood pressure (generated by the blood pump) and negative post-filter effluent pressure (generated by the effluent pump)
• This pressure gradient is the Transmembrane Pressure (TMP) 
• The equation which describes TMP is  (Pf+Pr)/2 - Pe,  where
  • Pf = pre-filter pressure
  • Pr = return pressure
  • Pe = effluent pressure
• The TMP is usually around 100-150 mmHg
• The ultrafiltration coefficient (KUF) is  the permeability of a membrane to water per unit of pressure and surface area
• A typical filter will have a KUF between 10-25 ml/h/mmHg/m² 

For supporting literature, there are several articles of varying levels of detail. It goes without saying that the time-poor primary exam candidate can safely omit these altogether. For somebody in need of a superficial treatment whcih covers the important stuff, an excellent article outlining some of the practical matters is "Technical aspects of hemodialysis" by Mitra and Mitsides (2016). For pure theory (and a little bit of history) one should read the 2013 review by Ficheux et al. The latter has sufficient depth to bury even the most detail-hungry reader up to the neck in equations. 

Definition of ultrafiltration

For both biological systems and dialysis membranes, ultrafiltration is when a hydrostatic pressure forces a liquid through a semipermeable membrane. This is pretty much what happens at the membrane of the glomerulus.  Ultrafiltration rate depends upon transmembrane pressure and ultrafiltration coefficient (KUF, discussed later in this chapter). There are, of course, official definitions. Medical dictionaries typically define it as

"filtration through a semipermeable membrane or any filter that separates colloid solutions from crystalloids or separates particles of different size in a colloid mixture"

The "ultra" in ultrafiltration seems to signify that it separates really tiny particles from the solvent, whereas microfiltration seems to be something to do with larger particles. For even larger particles, the term media filtration is used. When particles get small enough (eg. around the size off sugars) their mass separation process seems to earn the name nanofiltration, and for something as miniscule as dissolved ions the correct terminology appears to be reverse osmosis,  Where the size barrier lies to separate these two terms is somewhat arbitrary. Ann-Sofi Jönsson (2013) wrote an excellent chapter on this topic for Separation and purification technologies in biorefineries (p.205-231), and there she used the following definitions:

• Microfiltration: particles, 0.1–10 μm; classified according to membrane pore size
• UF membranes: macromolecules,  1–20 nm; classified according to molecular weight
• Nanofiltration: small molecules, 1 nm
• Reverse osmosis or "hyperfiltration": 0.1 nm

A handy diagram illustrates this scale (misappropriated for medical education purposes from the Safe Drinking Water Foundation)

Mechanism of ultrafiltration

Ultrafiltration takes place as the result of pressure. The blood is forced crudely against a fine mesh sieve, and water is forced through it under pressure while large molecules and cells remain behind. The filtration mechanism in dialysis is only slightly more sophisticated than the filtration mechanism in the coffee plunger. Both compartments might be under pressure (and in fact in CVVHDF the dialysate pump produces probably 50-100mmHg of pressure on the effluent side of the filter) but so long as the pressure in the 

This is something which took a while to accomplish, in terms of dialyser technologies. The earliest experimental machines used membranes so flimsy that it was impossible to apply any significant pressure across them. Haemodialysis was purely a technique of removing uraemic toxins by diffusion until Alwall's cellophane membrane dialyser, which permitted ultrafiltration because the membrane was trapped between two layers of the filter casing and was therefore not going to expand when pressure was applied. In this manner Alwall was able to perform the first ultrafiltration session in 1947. On one side of the membrane,the blood compartment pressure was supplied either by the patient's own arterial circulation (i.e CAVH, continuous arterio-venous haemodialysis) or by a roller pump. The pressure could be increased further by applying a clamp to the post-filter blood return line, causing pressure to back up in the blood compartment. On the other side of the filter, Alwall created a negative dialysate pressure by the highly scientific method of hanging the effluent hose out of the window.

These days, the pressure gradient in a CVVHDF circuit is accomplished by a series of roller pumps. The blood pump is pushing blood through the circuit with a pressure of around 100-120 mmHg, or at least that's the pressure generated by the resistance of ten thousand tiny hollow fibres in response to the blood flow rate of 200ml/minute. On the effluent side, a pressure drop is generated by the difference between the dialysate pump and the effluent pump. The  dialysate pump might be pushing the dialysate through at 2000ml/hour, and the effluent pump is sucking the effluent out at 2200 ml/hour. The difference between these pump flow rates (200ml/hr) is the ultrafiltration rate, i.e. how much fluid removal you will accomplish. The diagram below illustrates the basic principle and some of the pressures generated are "realistic", i.e measured in a real-life CVVH circuit.

There is an excellent article on the precise pressures which can be expected in a CVVH circuit. Ejaz et al (2004) went through all the data collected by the dialyser pressure sensors over the course of 91 treatments and summarised the findings. From these values, the poorly labelled graph above and the better labelled graph below are generated. 

After some rounding to improve memorability, pressure values from Ejaz et al were:

• Pre-filter blood pressure: 120-100 mmHg
• Post-filter blood pressure: 50-40 mmHg
• Effluent pressure: -20 to -70 mmHg
• Pressure drop (blood side): 60-70 mmHg

As you can see, throughout the circuit there is a pressure difference between the blood side of the membrane and the effluent side of the membrane, which is generated by the "push" of the blood pump and the "suck" of the effluent pump. This pressure difference across the membrane is unimaginatively known as the "Transmembrane Pressure", and is one of the the major determinants of ultrafiltration rate.

Transmembrane pressure

Transmembrane pressure, usually abbreviated TMP, has a precise ADQI definition:

The college had asked about this in the Fellowship exam, where Question 30 from the second paper of 2014 asked for a series of basic definitions related to dialysis. Instead of offering a wordy definition, the college defined it in terms of the following equation:

In short, transmembrane pressure is the average of all the blood pressures in the circuit, from which the effluent pressure is subtracted. In a perfect would you'd actually use the area under the pressure curve to calculate your average (a sort of "Mean Filter Pressure" if you will) but because we cannot accurately measure the pressure at every point in every single hollow fibre the difference between the pre-filter and post-filter blood pressures is the next least stupid surrogate. As a helpful point of reference, it is probably worth knowing that a normal range of TMP values one should expect during a run of CVVHDF is around 100-150 mmHg, depending on the rate of fluid removal prescribed. 

The influence of colloid oncotic pressure on ultrafiltration

It isworth remembering that pre and post filter blood pressures are not the only forces influencing the water as it decides whether to move from one side of the membrane to the other. Another important force to consider is the plasma protein oncotic pressure. This opposes the TMP, i.e. it is a force which attracts water back into the blood compartment of the haemofilter. Whats worse is that as haemoconcentration takes place along the filter, the plasma proteins become more concentrated and their oncotic pressure increases. In normal blood this pressure is usually about 25mmHg. Generally speaking, in dialysis which is managed with pre-dilution the protein oncotic pressure is decreased by the addition of fluid before the filter, which mitigates this effect to a significant degree.

From the above, it would stand to reason that ultrafiltration rate (in ml/min) would vary along a haemofiltration filter as the concentration of protein increases. This is indeed seen in real life. On top of that, if there is excess concentration of protein (or if transmembrane pressure is insufficient) there could even be backfiltration, i.e. the migration of water back into the concentrated blood compartment. This was seen by Ronco et al (1992) who used radiolabelled albumin to determine the concentration of albumin (and hence water) across a haemofilter.

Transmembrane pressure values as indicators of "filter health"

The TMP value can help troubleshoot circuit alarms. 

• High TMP with normal return pressure suggests there is a filter problem.  If the TMP is high with normal return pressure, it means that either the effluent pressure is extremely negative (impossible unless you attached it to wall suction!) or - more likely - that the pre-filter blood pressure is increasing due to high filter resistance. This scenario means that the filter is probably clotting. 
• High TMP with high return pressure suggests either the filter or the return line are the problem. Given that the effluent pressure is fairly fixed, the presence of a high return pressure and high TMP suggests that the venous return is somehow occluded, i.e. a clot has formed in the return line. This scenario does not exclude the possibility that the filter is also clotting. Both could be happening simultaneously.

But, these discussions are a digression into matters all too practical for the primary exam candidate who is supposed to be submerged in pure theory. Let us return to it now, and specifically to the primary question of the chapter, which is "how does water get removed by dialysis?". The specific variable which determines this (and into which TMP is factored) is KUF, the membrane ultrafiltration coefficient. 

Ultrafiltration coefficient

This single value is designed to represent the ultrafiltration efficiency of the haemofilter, and is a very important performance characteristic. It answers the question, "how easy will it be to suck water through this membrane?" ANSI defines KUF as "the permeability of a membrane to water" per unit of pressure and surface area, and it is described by the following equation:

The unit of measurement is ml/h/mmHg/m²  which is sufficiently cumbersome to warrant the use of a famous name for it (one supposes they could have called it the Bellomo, such that 1ml/h/mmHg/m² = 1 Bellomo). For any given filter the ultrafiltration coefficient is basically the KUF multiplied by the membrane surface area of that filter. KUF defines the filter's "flux", i.e. generally it is thought that anything below 10 Bellomos is a "low-flux" membrane and anything above 25 Bellomos  is "high-flux".