Furosemide

This chapter is relevant to the aims of Section H2(i) from the 2017 CICM Primary Syllabus, which expects the exam candidate to have "an understanding of the pharmacology of diuretics". Frusemide (furosemide) is such a ubiquitous ICU instrument and so strongly represented in the exam, that one could safely say that frusemide is diuretics ("La classe ç'est moi") and limit one's study of this whole tranche of agents to the detailed understanding of this one drug. That would be a ridiculous shortcut from an epistemological standpoint, but totally fair for the time-poor exam candidate, particularly looking at this list which represents the entire repertoire of diuretic questions from past papers:

• Question 14 from the first paper of 2020 (frusemide solo)
• Question 18 from the first paper of 2019 (frusemide solo)
• Question 16 from the first paper of 2018 (frusemide vs acetazolamide)
• Question 20 from the first paper of 2013 (frusemide vs acetazolamide)
• Question 16 from the first paper of 2012 (frusemide solo) 
• Question 2 from the second paper of 2007 (classification)

The success rates for these questions was surprisingly poor, rising to its highest in 2020 (51%). "Several candidates gave confused answers as to the mechanism(s) or drew pictures of a tubule with directional arrows for electrolytes with inadequate explanation", the examiners chuckled. Feeling highly exposed by this trenchant criticism, the author has expended considerable effort to offer adequate explanations for his confused electrolyte arrow diagrams. 

In summary:

In general, for diuretics,  Lang & Hropot (1995) is the best resource for general chemical and pharmacological properties, and Breyer & Jacobson (1990) fill in the blanks with regards to mechanisms. For furosemide specifically, nothing beats the two articles by Ponto & Schoenwald (from 1990, Part I and  Part II).  Unfortunately none of these articles are available for free. One must resort to excellent free articles such as Huang et al (2016), Oh & Han (2015) or literally a thousand other such resources, all of which generally give the same information abut furosemide pharmacokinetics. Unfortunately, pharmacodynamic mechanisms are more difficult to find. Particularly, good explanations of the acid-base effects and electrolyte consequences are difficult to find; they are either maddeningly incomplete, or poorly referenced ("see this out-of-print textbook from the 1960s"), or scattered across dozens of papers. To soothe the angry nerves of the baffled reader, every effort was made to dredge this material out of the depths, laying it out into a calm of orderly paragraphs.

Chemical structure and chemical properties of frusemide

Frusemide, known in Australia mainly as frusemide and elsewhere in the world as furosemide, is an anthranilic acid derivative more properly referred to as 4-chloro-N-(2-furyl-methyl)-5-sulfamoyl-anthranate. In the classic tradition of pharmacology writers, a chemical structure is ritualistically paraded before the reader, as if it has some kind of significance for their study or practice:

Furosemide was released on to the market in 1964.  Karl Sturm's team discovered it in 1959, giving it the name Salu 58 (Strum et al, 1966). The "furo" in the name comes from the methylated five-carbon furan ring hanging off the side of it, substituted for the NH₂ group of sulfamoylanthranilic acid. The name difference ("fruse" vs "furose") is a completely pointless artifact of international drug marketing, similar to the underlying reasons for why adrenaline and epinephrine ended up divided by the Atlantic ocean. Furosemide is the recommended International Non-Proprietary Name (rINN), and frusemide had, at one stage, been the British Approved Name of furosemide, as listed in the British Pharmacopoeia. It appears England is in the process of gradually converting to rINNs and double-labelling of this drug is routine there (even though "some people in the United Kingdom will deplore these changes, partly because they will regard them as a wholesale abandonment of British approved names in favour of American ones"). Similarly, Australia is gradually integrating into the world by transitioning from stubborn British names, and double-labelling of furosemide (frusemide) will persist here until 2023. In this piece, the two names will be used randomly and interchangeably.

Chemical relatives

Class relatives (comrades? classmates?) of frusemide are loop diuretics such as torsemide,  bumetanide and ethacrynic acid. This class rapidly became the most popular class of diuretic agents in the 1960s (prior to this, thiazides and organic mercury compounds were the dominant diuretics in routine use). Frusemide was by far the most popular, because unlike ethacrynic acid it did not have crippling gastrointestinal symptoms.

They are all chemically different. Frusemide and bumetanide are anthranilic acid derivatives, torsemide is a sulfonylurea,  ethyacrynic acid is a phenoxyacetic acid derivative and bumetanide is a 4-phenoxybenzoic acid (Wile et al, 2012). Anthranilic acid on which frusemide is based is an amino acid that acts as a precursor for tryptophan, and is the backbone of numerous molecules, none of which have very much pharmacological interest (though there are some NSAIDs among them). Frusemide is also technically a "sulfa drug", as it contains a sulfonamide group and is capable of causing allergic reactions in susceptible individuals.

Structure and function relationship

The main structural feature which makes these drugs into loop diuretics is an acid group attached to the meta position of the sulfonamide group.  As you can see, most of these drugs have a carboxyl group here, but a tetrazole will do just as well. When Sturm et al (1983) added SO₃^- groups there, the resulting molecules also had potent loop diuretic properties.

It appears that this acid group facilitates the binding of the diuretic molecule to the chloride binding site of the NKCC2 symporter protein. The mechanism of action of these drugs rests in their ability to compete with chloride for this pore, as demonstrated by experiments where high chloride concentrations were used to displace bumetanide off its binding site (O'Grady et al, 1987). Apart from that, it is difficult to relate the structure of these molecules to their performance characteristics. For example, substituting the tetrazole and changing the furan over to a thienyl makes azosemide, but all that does is make the drug less bioavailable for some reason. Keeping the carboxyl group and replacing the furan ring with a long butylamino substituent makes bumetanide (that butyl puts the "bu" in "bumetanide"), and somehow this makes it about 60 times more potent than frusemide, even though the supposed chloride pore binding site is totally unchanged. Clearly, the relationship between the molecular structure and loop diuretic properties is not completely understood.  Fortunately, nobody obedient to the unwritten laws of decency would ever ask about this in the CICM exams, or any other exams for that matter. 

Pharmacokinetics of frusemide

According to legend, the name "Lasix" was chosen because the duration of action of furosemide is about six hours. Because it lasts six hours. This spectacular choice, if true, occurred during the golden age of commercial pharmacology, when inventing a name for the agent occupied barely one microscopic sub-mote of the patent clerk's attention. These days, when drug names are generated through focus groups and linguistic experts load them with plosive letters to suggest power (P, T, D, K, Q) or fricative letters (X, F, S, or Z) to imply speed, furosemide would have ended up been called Xaqnatra, or something similar. 

Routes of administration 

Furosemide is usually administered orally or intravenously (in which case it can be given as boluses or as an infusion). A sublingual route is also apparently possible, to occupy that narrow window in illness acuity where you're sick enough to want immediate effect onset but not sick enough to merit a peripheral cannula (Haegeli et al, 2007).

It can also be nebulised, which has been used in the treatment of asthma - unsuccessfully, it was felt - though some authors reported improvement in the mild and short-lasting category of asthma attacks (Pendino et al, 1998). The supposed benefit of this was completely unrelated to the loop diuretic effect: Anderson et al (1991) found that it has some sort of anti-inflammatory effect, decreasing the release of leukotrienes and histamine from some isolated fragments of lung tissue. The fact that they have probably never heard of this should immediately suggest to the reader that this therapy has never made it into the arsenal of standard asthma management. However, it does get absorbed systemically when given as a neb, which means it could be administered to palliative patients who should not have invasive lines. In case you are wondering, 80mg seems to be the dose. 

Absorption of frusemide

Absorption of frusemide from the GI tract is erratic and individual. Studies comparing CCF patients with varying degrees of illness severity seem to suggest that it varies more between individuals than it does between groups with different stages of heart failure. According to (undoubtedly, messy) rat experiments by Chungi et al (1979),  the site of absorption is the stomach, where pH is 3.0. Absorption in the rather more alkaline rat jejunum (pH = 5.0) was markedly worse. The pKa of this drug is 3.6, which means it is highly ionised (thus highly water-soluble and poorly lipid soluble) at normal intestinal pH (8.0-8.5).  It is otherwise rather poorly water-soluble, and actually requires a substantial amount of sodium hydroxide as an alkalinising excipient to make it available as an IV formulation.

Solubility, protein binding and distribution

This acidic pKa means furosemide is highly water-soluble at normal physiological pH and therefore distributes to mainly the circulating and extracellular water. Cutler et al (1974) found that its volume of distribution was something like 11-18% of total body weight, or 0.1-0.2 L/kg.

It is highly protein-bound, mainly to albumin (95-97%), which produces an interesting effect. In the absence of sufficient albumin to transport it, frusemide ends up distributing god knows where. Pichette et al (1996) created a population of cursed mice, born without any albumin at all,  and demonstrated that the volume of distribution of frusemide in this group was massively increased. Moreover, they were barely secreting any of it via the proximal tubule, which is how it has its effect (it has to bind to its target at the luminal surface of the tubule). The authors were forced to conclude that frusemide behaves in a completely opposite way to the way other protein-bound drugs behave, where the decrease in albumin increases the free fraction of the drug and makes it more potent. Instead, in the case of frusemide, a decrease in albumin produces a decreased delivery of the drug to its site of active secretion in the kidney, which decreases its pharmacological effect.

This pharmacokinetic oddity has given rise to the impression, historically held by many physicians, that the concomitant infusion of albumin together with frusemide into patients with hypoalbuminaemia will somehow result in an increased diuretic effect. Give them albumin, so they cand deliver more frusemide to the proximal tubule and therefore secrete more frusemide, increasing their urine output - that was the rationale. Unfortunately, though physiologically plausible, this has never been demonstrated to improve the clinical response to frusemide. For example, Chalasani et al (2002) found no difference, no matter whether they gave the albumin as an infusion or as boluses.  "The coadministration of albumin and furosemide ... should not be used clinically", the authors concluded somberly.

Metabolism and clearance

As hinted at in the spoiler above, frusemide is cleared by three main mechanisms:

• Glomerular filtration (minor route, because of the extremely high protein binding)
• Proximal tubular secretion (usually, 50%)
• Metabolism in the kidney (~ 50%)

A large proportion of frusemide is cleared renally, and most of this happens by active secretion, via OAT-1 organic anion transport proteins in the proximal tubule. Beermann et al (1977) determined that approximately 50% of the administered dose is cleared in this way. Basolateral transporters have such a high affinity for organic anions that they are able to strip the drug molecules off the albumin as it passes them (Shankar & Brater, 2003). 

The rest of an administered dose appears to be metabolised in the kidney itself, where it is converted into a glucuronide which has approximately 25% of the parent drug's activity. Theoretically, this can (and probably does) happen in the liver as well, but this does not account for a lot, and first-pass effect is not thought to play an important role. The half-life is shorter than the duration of effect, about 30-120 minutes (Ponto & Schoenwald, 2012)

Pharmacodynamics of furosemide

Furosemide specifically, and loop diuretics in general, are so called because they act on the thick ascending limb of the loop of Henle. There, furosemide competes with chloride for the chloride pore of the main transport protein (NKCC2). The result is a failure to transport anything (as the co-transported only works when all the ions are available). 

The mechanisms of action of furosemide are:

Or, in case these concepts somehow remain unclear, let's try it in point form:

• Blockade of the NKCC2 transporter decreases the reabsorption of sodium potassium and chloride in the thick ascending limb
• This increases the delivery of sodium potassium and chloride to the distal nephron.
  • There, sodium is reabsorbed by aldosterone-regulated ENaC channels.
    • As body water volume decreases with frusemide therapy, so the reabsorption of sodium here escalates, as aldosterone is released in response to hypovolemia
    • Potassium removal via the urine is also increased because of sodium and potassium being exchanged with the tubular fluid
  • Chloride and ammonium elimination is also increased, as ammonium is an alternative substrate for NKCC2
  • The consequences of this are:
    • Metabolic alkalosis (hypochloraemia)
    • Hypernatremia (as sodium is retained)
    • Hypokalemia (as potassium excretion is increased by the increased sodium delivery to the distal nephron)
  • The increased delivery of chloride to the distal nephron results in urinary acidification.
• The decreased reabsorption of solutes into the inner medulla and the increased delivery of solutes to the distal nephron produces the diuretic effect:
  • The osmolality of the collecting duct fluid increases because of decreased tubular sodium and chloride reabsorption
  • The osmolality of the medullary intestitium decreases for the same reason
  • Ergo, the gradient for osmotic reabsorption of water via the aquaporins in the collecting duct is impaired
  • Thus, more water is excreted, and this is the basis of the diuretic effect 

Ceiling dose and "threshold dose"

How much sodium/chloride/potassium is actually lost, the engineering hindbrain of the intensivist might ask? According to Shankar & Brater (2003), an IV dose of 40mg of furosemide ends up producing a diuresis of about 3-4 litres of urine, and approximately 200-250 mmol of sodium, over the course of about 3-4 hours. This amounts to a fractional excretion of about 20% of all filtered sodium, and represents a "ceiling dose" for frusemide, i.e. a dose beyond which no further sodium excretion increases would be possible.  Interestingly, for patients with CCF, this dose is close to 80mg, and for patients with chronic renal failure it may be as high as 200mg  (Wilcox, 2002).

The mechanism of diuretic effect is also often said to have a "threshold dose", below which one does not detect any clinically relevant increase in urine output. For furosemide, for example, in healthy kidneys a dose less than 10mg will not produce any diuresis, so it is said. Oh & Han (2015) also give the Wilcox article as the reference for this, but it does not contain any mention of a 10mg threshold dose. It is however mentioned in numerous other publications. But does it actually exist? Plotting the dose-response curve for furosemide in the oral range of 20-80mg produces a curve which is close to linear (Marciniak, 2019). This suggests that the threshold dose is some sort of myth, arising from a misinterpretation of a logarithmic dose-response curve (which can visually give the impression of a threshold effect). Or at least within the clinically relevant dose range the threshold dose is probably even lower then 10mg in healthy kidneys. 

On the other hand, it would be reasonable to expect some sort of minimum concentration below which the furosemide molecules simply lose their competition with chloride.  And it would be reasonable to expect patients with chronic renal failure to respond so minimally to small doses that the clinical diuretic effect could be described as a non-response, even though technically speaking some NKCC2 channels do probably have furosemide molecules attached to them. Often, when this comes up in the literature, it tends to be accompanied by a diagram which looks like this:

This graph itself looks plausible and is probably good enough to illustrate the concepts and get an exam candidate out of trouble. It is reproduced in numerous official publications, textbooks and review articles. However the reader must be made aware that the actual shape of the curves is purely speculative, an invention on the part of the author. It does not help that most publications do not reference the original source for these data, which was probably Brater et al (1980). The investigators measured fractional sodium excretion as a function of urinary furosemide concentration, producing a messy series of curves from multiple individuals with different degrees of renal impairment. Together, these demonstrated a markedly altered pharmacodynamic response in damaged kidneys. 

Infusion versus bolus dosing

The use of continuous furosemide infusion is a staple of the cardiologist and intensivist who wants to produce a sustained diuresis. It is not clear whether this has any benefit in terms of total fluid removal or patient-centred outcomes. It is certainly difficult to draw conclusions from the conflicting data in the literature, where authors are forced to say things like "continuous intravenous infusion is more effective than, or in some cases, similar to, intravenous bolus therapy". The more important difference between these two strategies is probably best illustrated by a graph which plots toxic and desirable effects over duration of therapy. With an infusion, it is possible to achieve the desired effect without risking otoxotic peak concentrations. 

Mechanism of furosemide-induced hypokalemia

“Pseudo-Bartter-Syndrome”, where the patient loses potassium during furosemide therapy, is so well known that other diuretic classes which do not do this are occasionally referred to as "potassium-sparing" diuretics. It can lead to quite extreme hypokalemia (a potassium concentration below 2 mmol/L), with horrific complications such as arrhythmias and hypokalemic rhabdomyolysis (Ruisz et al, 2013).  To summarise, the mechanism of this effect is as follows:

• The ENaC channel reabsorbs sodium in the collecting duct
• This generates an apical transmembrane potential, as positively charged ions are being removed from the lumen of the duct
• This negative apical membrane charge produces a movement of potassium out of the cells via the ROMK channel (Welling, 2016)
• In this fashion, sodium is exchanged for potassium in the collecting duct
• Thus, increasing the delivery of sodium to the distal nephron creates an increased potassium excretion.

Or, in the form of pictograms:

Therein lies the origin of the potassium-sparing effect of spironolactone, which downregulates the expression of the ENaC channel. By means of decreasing sodium reabsorption here, aldosterone and ENaC antagonists can prevent potassium excretion and therefore ameliorate the hypokalemia and hypernatremia associated with furosemide use. Intensivists love to exploit this pharmacodynamic interaction, as it results in them having to prescribe fewer doses of electrolyte replacement.

Mechanism of furosemide-induced metabolic alkalosis

This was demonstrated by Zazzeron et al (2016) whose term were able to measure the urinary electrolyte concentration of surgical patients in real time following a 20mg furosemide dose. The investigators managed to get a hold of a urinary electrolyte analyser for this purpose. Their data  really does speak for itself, which probably makes the following reinterpreted diagram completely redundant, but the author could not resist his primitive instincts:

As you can see, chloride excretion and sodium excretion are immediately increased, but then sodium reabsorption begins to develop, and the urinary sodium starts to decline. Not so for the chloride anion, as there is no ENaC for that- the chloride gets dumped in the urine. To the practitioner of physicochemical Stewart voodoo, this will immediately make sense- the reabsorption of a strong cation and the loss of a strong anion will increase the strong ion difference of the extracellular fluid and should therefore become an alkalinising influence. However, one should be fair and try to explain this in classical terms. This will require the following contortions:to explain this in classical acid-base terms, we will need to contort ourselves into the following line of reasoning:

• Furosemide decreases the extracellular fluid volume, and produces hypokalemia
• The extracellular volume contraction stimulates the release of aldosterone
• Both aldosterone and hypokalemia stimulate H⁺ ATPase activity in the collecting duct, increasing the elimination of H⁺ into the lumen of the collecting duct
• This loss of acid results in a systemic metabolic alkalosis

Eiam-Ong (1993) discussed this in their article, relying mainly on references to nephrology textbook chapters. This makes their explanation perfect for our purpose, as it represents a distillation of classical textbook explanations, and is therefore likely to match what examiners might expect in some kind of written answer.

Mechanism of the diuretic effect

How does the blockade of solute reabsorption in the thick ascending limb result in litres of urine? Well. Those solutes being reabsorbed in the medulla have a tendency of increasing the medullary interstitial osmolality, and at the same time decreasing the osmolality of the tubular fluid. The resulting osmotic gradient is then used to reabsorb water from the collecting duct. When the gradient for reabsorption is decreased, less water is reabsorbed, and more diuresis results. Consider a thought experiment where the urinary osmolality and the inner medullary osmolality are in perfect equilibrium. No water absorption though collecting duct aquaporins would be possible, because there is no concentration gradient to drive it. It would be as if one had developed diabetes insipidus (i.e. those aquaporins may as well not be there). 

From this, it follows that after furosemide you should have a period where your inner medulla is is relatively hypoosmolar. This exact effect has been demonstrated by experimental data. Spitalewitz et al (1982) measured the tissue osmolality and tissue sodium concentration in the renal papilla of the dog, pre and post furosemide. The osmolality dropped from 1005 to 381, and the papillary interstitial sodium concentration dropped from 152 mmol/L to 87 mmol/L. Recall that the urinary sodium (in the graph literally one paragraph above) increases significantly following furosemide. With so much solute in the tubular fluid, and so little solute in the inner medullary interstitium, there is nothing to osmotically attract water back out of the urine. If one needed to express this in the form of a diagram, one would do something like this (though one would probably want to label it a little better for any sort of exam purposes):

 

Effects of furosemide on magnesium and calcium handling by the kidney

A reader had pointed out that this is a common and frightening detour in CICM viva voce examinations, where the examiner, after calmly drifting around furosemide pharmacodynamics, might suddenly swerve dangerously into oncoming calcium or magnesium metabolism. "Ah yes, and how does frusemide affect this?" they might ask, laughing at your discomfort. This electrolyte derangement is commonplace in critical care, but seems to be one of those things nobody ever really bothers to explore.  Fortunately, at least one good article (Alexander & Dimke, 2017) is available. 

In summary, the main source of the magnesium and calcium loss is an interference with the normal mechanisms of magnesium and calcium handling in the thick ascending limb, where much of aforementioned handling takes place. 

Recall that the NKCC2 channels in the thick ascending limb permit the movement of two cations and two anions, which should theoretically maintain electroneutrality. However, the thick ascending limb also features ROMK potassium channels, which allow the potassium to escape again. The increased potassium concentration and decreased chloride concentration in the tubular lumen is clearly a recipe for a positive change in the tubule lumen, and that is exactly what happens. 

Thus, a transepithelial potential difference develops, with the tubular lumen more positive than the extraluminal fluid (+10mV is the figure measured by Greger & Schlatter, 1983 ) The result is a repulsion of cations. Potassium, magnesium and calcium are therefore pushed out of the tubular lumen. In this fashion,  25% of the calcium and 60% of the total filtered magnesium are reclaimed here. That this is a purely electrical phenomenon was demonstrated elegantly by Di Stefano et al (1993), who made the tubular fluid negative, and observed calcium and magnesium flow back into the lumen out of the medullary interstitium. Obviously, anything which might affect the reabsorption of chloride (like frusemide, or Bartter and Gitelman syndrome) will decrease the positive luminal charge and discourage the reabsorption of calcium and magnesium, resulting in the loss of these cations.

Adverse effects of furosemide

As one is occasionally called upon to list the potential adverse effects of furosemide, they are listed here in a convenient pointform table

• Hypotension (esp. orthostatic)
• RAAS activation
• Ototoxicity, especially in combination with aminoglycosides
• Electrolyte disturbances:
  • Hypokalemia
  • Metabolic alkalosis (hypochloraemia)
  • Hypernatremia (as sodium is retained)
  • Hypomagnesemia
  • Hypophosphatemia
• Pharmacological interactions:
  • Displacement of warfarin from albumin binding sites

Furosemide toxicity

As one is occasionally called upon to list the potential adverse effects of furosemide, they are listed here in a convenient pointform table

• Hypotension (esp. orthostatic)
• RAAS activation
• Ototoxicity, especially in combination with aminoglycosides
• Electrolyte disturbances:
  • Hypokalemia
  • Metabolic alkalosis (hypochloraemia)
  • Hypernatremia (as sodium is retained)
  • Hypomagnesemia
  • Hypophosphatemia
• Pharmacological interactions:
  • Displacement of warfarin from albumin binding sites

Ototoxicity due to the use of furosemide is a well documented but rare side effect. It does not appear to be unique to massive doses - Gallagher & Jones (1979) reported deafness even with doses as low as 40mg. The rate of administration appears to be a critical determinant - it appears that one can decrease the risk of this complication if one limits oneself to a dose rate of less than 4mg/min.

How high a dose is too high? The authors colleagues bring harrowing tales of cardiologists using infusions in excess of 100mg/hr, for many hours. In the community, frusemide overdoses are limited to case reports of military personnel attempting rapid weight loss to comply with airforce weight standards. Incidentally, no hearing and only a small amount of decorum was lost (one unlucky staff sergeant experienced comical toilet syncope). Theoretically, a truly massive overdose would probably lead to severe dehydration and some sort of electrolyte-related cardiotoxicity, but because of the ceiling effect, one cannot expect a fatal scenario due to the diuretic effect alone. 

Now, apart from just mentioning these abnormalities, occasionally we are called upon to also explain them all in the space of ten minutes. For the prepared candidate, that should not be a major imposition. For example, Question 16 from the first paper of 2012, by the examiner's own description, was "a relatively straightforward question with marks available for listing the abnormality and then discussing its origin", which really calls for a tabulated answer, such as this one: