Pharmacology of insulin preparations and glucagon

This chapter is related to Section U2(ii) and U2(vi) from the 2017 CICM Primary Syllabus, which expect the exam candidate to "understand the pharmacology of insulin preparations" and "outline the pharmacology of glucagon". The difference between "understand" and "outline", in the mind of this author, is the permission to completely lose his thread among the smallest pointless details of insulin pharmacology; whereas for glucagon only a sober summary is permitted. The reader is reminded that only Question 6 from the second paper of 2013 has ever asked about the pharmacology of insulin (actrapid was the specific subject), which means this chapter is superfluous to the needs of the fatigued exam candidate and can be ignored completely.

For the casual reader and the trainee preparing for CICM exams, insulin pharmacology is abundantly available in many different forms. Good professional peer reviewed resources include  "Donner & Sarkar (2015), Bushra (2004), Joshi et al (2007) or Heinemann & Richter (1993). What these may lack, you do not need; i.e. there are minute details (eg. the pKa of hexameric insulin) that might be difficult to track down without dumpster-diving out the back of the medical library, but nobody could possibly need that level of detail for exams, and outside of exams you could simply Google it. 

Chemical relatives and chemical classification

All insulins are synthetic polypeptide hormones, of 51 amino acids in length, which means their "chemical relatives" would have to include all the other peptide hormones (eg. those secreted by the hypothalamus and pituitary) as well as peptide cytokines and paracrine mediators. Within the family of insulins all members are roughly the same molecular weight and the variation is rather minor, usually differing by only one or two amino acids. For example, porcine insulin has an alanine instead of threonine at the end of the B-chain, and bovine insulin has two different amino acids on the A-chain (instead of threonine and isoleucine at positions A8 and A10, cows get alanine and valine). These days we have phased out animal extracts, and the insulin available to human patients is human insulin, also known as  "regular insulin" or "neutral insulin". It is generally marketed as Actrapid (a registered trade name), and it is identical to the molecule which is made in your pancreas, except these days it is mass-produced by an enslaved yeast species. 

It would be doubly pointless to report all the minor molecular differences between the various insulin formulations, firstly because this could never be included in any sort of exam question, and secondly because the structural variations do not follow any kind of predictable pattern and are therefore without value from an educational perspective, as there is no high overarching concept to be explained by discussing them. Fortunately, pointlessness is no barrier for Deranged Physiology.

An Unnecessary Summary of Chemical Differences between Different Formulations of Insulin

Insulin preparation	Structural variation
Neutral human insulin (Actrapid, Humulin)	Two chains, B (30 amino acids) and A (21 amino acids), with amino acids labelled by chain letter and number.
Bovine insulin	Two amino acid substitutions at A8 and A10: alanine and valine instead of threonine and isoleucine
Porcine insulin	Alanine instead of threonine at B30
Insulin aspart (NovoRapid)	Aspartic acid substituted for proline at B28
Insulin lispro (Humalog)	Lysine replaces proline at B28 and proline replaces lysine at B29 Koivisto (2009)
Insulin glulisine (Apidra)	Asparagine is replaced by lysine at position B3, and lysine is replaced by glutamic acid at B29.
Isophane insulin, or NPH* (Protaphane)	A complex of regular human insulin and protamine
Insulin glargine (Optisulin, Lantus, Toujeo)	Asparagine is replaced with glycine in position A21, and the B-chain is extended by 2 arginine residues, which makes this formulation less soluble at physiological pH.
Insulin detemir (Levemir)	At position B29, myristic acid (a fatty acid) is attached to the lysine, which makes this insulin bind to albumin with greater affinity
Insulin degludec (Tresiba)	At position B29, hexadecanedioic acid is added via an amide linkage to the lysine. With this, the insulin can form multi-hexamers, greatly slowing its release ("ultra-long-acting" is sometimes used to describe this formulation)

Structure and function relationship of insulins

Contradicting some of the comments above, there are a few general statements to be made about some molecular rules important for the structure-activity relationship of insulin. What there is to say on this subject is said in the most intelligent way by Mayer et al (2007), and what follows is a series of dot-point oversimplifications. 

• Insulin molecules have some essential sequences (about a third of all the amino acids), and these must not vary in order for the molecule to remain active. We have discovered them by looking very broadly at the insulin-like peptides synthesised by different animal species. These sequences 
• There are also sequences that are not essential, and we can fool around with these as much as we like. This is the most important fact to absorb about synthetic insulin analogs: we can mess with the insulin molecule to alter its solubility and other properties, and we can do this without interfering with the function of the hormone, so long as the essential sequences remain intact.
• Messing with the end of the B chain for example (eg. with amino acids in the B28 and B29 positions) tends to produce a molecule with a weakened self-association behaviour, which means it will be very reluctant to form hexamers. The unassociated peptide molecules can then easily diffuse away from the subcutaneous injection site. Analogs with this sort of modification (eg. Novorapid and Humalog) are therefore rapid-onset and very short acting. 
• On the other hand, adding some amino acids to the end of the B chain seems to change the solubility of the insulin, making it rather insoluble at normal body pH. An acidic insulin solution is injected as a subcutaneous depot, where it immediately reverts to neutral pH and precipitates, 
• Changing the insulin molecule is actually a rather recent development, in the sense that we only got our first modified analog in 1996 (it was Humalog). Apart from changing the molecule, previous methods of influencing the duration and onset of effect required people to add things to regular insulin, for example protamine and various phenols. This was available since NPH was developed in the late 1930s.  "NPH" stands for Neutral Protamine Hagedorn, after Hans Christian Hagedorn who designed it in an effort to produce a longer lasting formulation. The name also recalls neutral pH, which was the main point of performing this molecular manipulation (to adjust the pKa of the molecule and make insulin less soluble at physiological pH). The same can be said for the term "isophane", which sounds like "iso-pH" but was in fact coined by Hagedorn because it means " same appearance" and refers to the quantitative titration of protamine into insulin which ultimately resulted in the proportion of protamine to insulin being the same in solution and in precipitate.

What exactly is a "unit" of insulin

In the modern era, one international unit of insulin is defined as 38.5 µg (0.03846 mg) of dry insulin crystals. This was decided by the WHO Committee on Biological Standardisation in 1987, which was the last such revision, in a long string of progressively improving efforts to define and standardise the dosing of insulin worldwide.

Like units of vasopressin and units of heparin, this measurement standard is bioefficacy based, i.e. it corresponds to the minimum amount of insulin required to do some physiological thing. In this case the physiological thing is "make a 2kg rabbit so hypoglycaemic that they develop seizures". Specifically, this was the definition used by Banting et al in 1922, when they were injecting experimental animals with pancreatic extract, and trying to somehow rate the potency of their concoction. They settled on using rabbits (mice being the other option) and the measurement they used was "the number of cubic centimetres which lowers the percentage of blood sugar in normal rabbits to 0.045 in from two to four hours", that percentage being a level of hypoglycaemia necessary to cause convulsions.

Unfortunately, after generating whole piles of comatose rabbits, the investigators were forced to acknowledge that there was some considerable variability in the seizure threshold of individual animals. Moreover, reasonable people would agree that convulsions are not the correct clinical endpoint for therapeutic insulin administration, and therefore most doses would have to be less than one "rabbit seizure unit" in magnitude. This consideration had led to a confusing proliferation of units and standards, lovingly detailed by Lacey (1967). In short, when dry pancreatic extract finally became available (around 1923),  several independent laboratories tested its potency (again, by killing rabbits) and finally declared 0.125mg as the standard unit. This value has decreased over the years mainly because the purity and therefore the potency of dry insulin has improved, with modern insulin capable of taking out a healthy bunny with only 0.03846 mg.

Administration and absorption of insulin

Being a rather large peptide hormone, insulin would be regarded by the digestive system in much the same way as a pork chop and would have zero oral bioavailability as the result. This means one could consume any amount of fried pancreas without the risk of insulin overdose. Valiant efforts to help it withstand enzymatic degradation have so far failed to yield a commercially available product. As the result, the regular insulin user needs to become comfortable and familiar with needles, as their only means of accessing the benefits of this drug would be to inject it into themselves. 

From this, it follows that any discussion of insulin bioavailability is a discussion of how it performs when it is absorbed from a subcutaneous depot. True, some insulins can be given intravenously, but for the majority of people regular intravenous injections would be too much of an imposition, and compliance with regular IV drugs would be rather poor (unless they possess heroin-like addictiveness). Moreover, the half-life of insulin in the circulation is rather short (4-6 minutes), which is why the pancreas secretes it at a constant rate, and why a constant infusion of insulin would be required instead of intermittent doses. Thus, subcutaneous administration is the most common and convenient mode, and the most important modifications made to insulin molecules in the course of their commercial development have focused on delaying its absorption from the depot at the injection site, to make it more infusion-like.

Thus, the "absorption" of insulin is really its diffusion away from the site where it was injected. Hildebrandt (1991) and Søeborg et al (2009) offer great overviews of these kinetics and the factors that influence them. Fortunately, aside from the quality of the microcirculation at the site of injection, there are only two main influences:

• Concentration of the insulin: the more insulin there is in the depot, the more likely the insulin molecules are to associate into oligomers. The association state ranges from 50% at 0.2 mmol (33 units per ml) to 100% at 1.5 mmol (250 units per ml). At a concentration of 5 units per ml, the insulin is mainly monomeric. Monomeric insulin is a much smaller molecule and diffuses into the systemic circulation more readily, whereas the insulin in large concentrated subcutaneous depots will mainly be present in the form of hexamers, and diffusion would be slow. This is important because it is the absolute opposite of what is normally expected to happen; large concentrations create gradients that tend to favour diffusion, but here it is inhibited.
• Oligomerisation of insulin is therefore an important determinant of its absorption, with short acting rapid-onset formulations (eg. insulin aspart, Novorapid) having their oligomerization behaviour intentionally weakened by the modifications made to the insulin molecule. Oligomerisation does not need to be happening exclusively between insulin molecules; for example isophane insulins are a precipitate of protamine and insulin (which you'd have to call crystallisation, as oligomers are complexes made of several copies of the same molecule) . Older animal-derived insulin formulations had absorption kinetics similar to those of modern intermediate-acting formulations mainly because of various impurities forming complexes with the insulin molecules and slowing their diffusion.

Given that the range of available insulin concentrations is limited (most present as 100 IU/ml), and considering we never dilute it before subcutaneous injection, the onset of clinical activity is mostly dependent on the species of insulin you have injected. A quick Google Image search immediately yields numerous similar-looking concentration/time curves where different insulins are compared to one another. On closer inspection, the vast majority of these are totally made up - or, to insult them more precisely, they are either rough sketches made to illustrate the concept (without much attention to pharmacological accuracy), or they are created using pharmacokinetic modelling software (with all the caveats inherent therein). Valuable though these are, a regular reader of this site will probably expect real pharmacokinetic data from some kind of cruel human experiments, such this one taken directly from Heinemann & Anderson (2004). 

Those are not insulin concentration measurements, of course, as that assay is difficult  and expensive - instead, the authors used a euglycaemic glucose clamp technique to indirectly assess the effect of the drug over time. The jaggedness of their data is a reflection of this fact, as the automated glucose controlled struggled to maintain a stable BSL, but you don't need smooth curves here - in fact, the reader will have realised by this point that the actual shape of the curves is rather meaningless, and what is probably more important is the time of onset, timing of the peak, and total duration of the effect, which would work better as a table. Here's a representative one, compiled mainly from Arshag et al (2006) and Freeman (2009):

Insulin species	Onset	Peak	Duration of action
Neutral human insulin (Actrapid, Humulin)	30-60 min	2-3 hrs	6-10 hrs
Insulin aspart (NovoRapid)	10-20 min	1-3 hrs	3-5 hrs
Insulin lispro (Humalog)	15-30 min	0.5-2.5 hrs	3-6.5 hrs
Insulin glulisine (Apidra)	10-15 min	1.0-1.5 hrs	3-5 hrs
Isophane insulin, or NPH* (Protaphane)	1.5-4 hrs	6-14 hrs	16-24 hrs
Insulin glargine (Optisulin, Lantus, Toujeo)	1-3 hrs	N/A	~24 hrs
Insulin detemir (Levemir)	50-120 min	N/A	~24 hrs
Insulin degludec (Tresiba)	30-90 min	N/A	24-36 hrs

Distribution,  solubility and protein binding of insulin

The volume of distribution for normal regular insulin is approximately the same as the extracellular fluid space. According to Turnheim & Waldhäusl (1988), there is a rapid early elimination (2-5min) from the intravascular space by distribution to the extracellular fluid, and then a slower elimination which is accounted for by metabolism. Insulin does not enter the intracellular waters, as its molecular size concentration and lipid solubility would not permit this, which means its VOD should be something like 0.2 L/kg. To tell whether this approximation is true or not is remarkably difficult, as hardly anybody seems to have ever published a straight answer for the different species; but fortunately Potocka et al (2011) at least has the VOD for regular human insulin, which they reported as 30.7L, or about 0.44 L.kg.

For something that is supposed to play nice with the bloodstream into which is is naturally secreted,  on paper insulin is remarkably insoluble in water at normal physiological pH. Regular insulin preparations are said to have a pKa of 5.4, and are rendered soluble only by titrating the solution to a facemelting pH of 2-3, or by adding various solubility-enhancing excipients like glycerol and phosphate. At least this is the series of steps taken to create a concentrated commercial product, or to homecook an aqueous solution of lyophilized dry insulin crystals which have been delivered to your lab. As Brange & Langkjœr (1993) patiently point out, this is because at high concentrations insulin molecules naturally associate into hexamers - this is why they develop these weird solubility problems. On their own, insulin monomers are easily soluble in water, which is how insulin tends to traffic in the bloodstream, where the concentration is nanomolar.

One other important feature needs to be mentioned, which does not really belong in any other section. Endogenously secreted insulin goes directly to the liver via the portal circulation, whereas IV insulin is delivered to the systemic bloodstream. This has all sorts of implications, discussed by Arbit (2004). In summary, the result is a delay in the magnitude, the onset and the offset of effect:

• Skeletal muscle is not capable of disposing of very large amounts of glucose
• The onset of action is delayed, as skeletal muscle takes longer to start absorbing glucose and depositing glycogen (whereas the liver can do this very rapidly)
• Skeletal muscle is also slow with stopping this function, which means it keeps absorbing glucose even after the insulin stimulus is gone, which could lead to hypoglycaemia

Protein binding is only really a feature of insulin detemir, which has a fatty acid hanging off it. This trailing acid causes it to catch on passing albumin molecules, enhancing its slow release characteristics and limiting its bioavailability in the bloodstream. There is little published material out there to describe the protein binding of other insulin species, but we can generally assume that it is probably low.

 Metabolism and elimination

Insulin is a rather large molecule, and meets a fate similar to other peptides and protein fragments, where it is ultimately captured by endocytosis and degraded. About 50-60% is degraded by the liver, around 30% by the kidneys, and the rest by skeletal muscle, adipocytes, CNS and the myocardium. 

• Metabolism by insulin-sensitive tissues: any tissues that bind insulin via the insulin receptor can also endocytose the receptor-ligand complex, drop the pH in the vesicle to something like 6 to force them to dissociate, return the receptor to the cell surface, and then digest the insulin (Duckworth, 1988). This seems to happen in skeletal muscle, cardiac muscle, adipose tissue, and the liver - with the liver probably accounting for most of the degradation of endogenously secreted insulin.
•  Metabolism by the kidney: renal metabolism can account for as much as 30% of the total elimination of insulin (Lohr et al, 1998). Insulin is filtered at the glomerulus, ends up being reabsorbed by the proximal tubule cells, and undergoes degradation in lysosomes. Only about 1% is eliminated as unchanged drug. 

Pharmacodynamic effects

Insulin is a pleiotropic hormone that exerts its effect by binding to a large transmembrane receptor with an intracellular tyrosine kinase domain. These receptors are present in all mammalian cells (though they are expressed with different density). The intracellular signalling pathway is mainly the PI3K secondary messenger pathway. Detailed discussions on the physiology of insulin are offered elsewhere, and here only the main effects are: 

• Intracellular effects of insulin receptor binding:
  • Target tissue cells contain vesicles which, on their inner membrane surface, express GLUT4 glucose transport proteins. 
  • In response to the binding of insulin to its receptors, these cells perform a sort of exocytosis, where these vesicles migrate to the surface membrane and fuse with it
  • The end result is the appearance of more GLUT4 transport proteins on the cell surface
  • This increases glucose uptake by this cell
• Physiological effects of insulin
  • Cabohydrate metabolism
    • Increased glucose uptake by skeletal muscle (80%), myocardium, adipose tissue, and liver
    • Decreased glycogenolysis and decreased gluconeogenesis
    • Increased deposition of muscle and hepatocyte glycogen
  • Lipid metabolism
    • Decreased free fatty acid mobilisation by adipose tissue (decreased activity of hormone-sensitive lipase)
    • Increased triglyceride synthesis in liver and adipose tissue
    • Increased synthesis of VLDLs and increased activity of lipoprotein lipase
  • Protein metabolism
    • Decreased protein catabolism, increased protein synthesis
    • Decreased gluconeogenesis from amino acids, and thus decreased urea production
  • Intracellular electrolyte shift
    • Intracellular shift of potassium and phosphate 
  • Haemodynamic effects
    • Increased cardiac contractility
    • Increased coronary blood flow
    • Decreased afterload due to decreased peripheral vascular resistance
    • Increased sympathetic nervous system activity
  • Paracrine effects
    • Decreased release of glucagon

Prescription of regular subcutaneous insulin

The pancreas secretes insulin continuously, and then increases production when stimulated by a meal. If it stops doing this because of diabetes, the diabetic needs some constant baseline insulin source as well as pre-meal doses. The vast majority of the time this is achieved by prescribing some of the total daily insulin as a long-acting formulation, and then splitting the short-acting doses for the three meals of the day. An example approach would be to split the total daily dose into halves, with 50% of the total dose given as basal insulin (usually 0.1-0.5 units/kg/day, i.e. 7-30 units/day), and the rest split evenly into three prandial bolus doses. There are all manner of algorithms to guide this more specifically and to protocolise the titration of insulin; interested readers are redirected to the diabetes section of the RACGP website and the ADEA guidelines. 

For the intensivist, the more relevant topic would be a discussion of how to commence regular subcutaneous insulin after an insulin infusion is no longer required.  Intravenous insulin has a relatively short half life due to its distribution, and will disappear from the circulation very quickly. For this reason, an insulin infusion needs to be overlapped with a subcutaneous dose by at least a couple of hours. Moreover, for a variety of reasons the total 24-hr dose of insulin is not really reflective of what the subcutaneous requirements are going to be, and some suggest only using 80% of the 24-hour dose.

Pharmacology of glucagon

As promised, here is a perfunctory and oversimplified detour into glucagon pharmacology. The use of this substance has diminished, and its significance for the critical care trainee along with it. These days the CICM exam candidate needs to be aware of this drug only in the context of its historical use as an antidote for β-blocker toxicity, and that awareness need only be dim and limited, as in this application glucagon is rapidly being surpassed by high dose insulin euglycaemic therapy. For the reader with a deeper need, Müller et al (2017) do glucagon over 46 exhausting pages.

Name	Glucagon
Class	Pancreatic hormone analog
Chemistry	Peptide hormone (29 amino acids)
Routes of administration	IV, subcutaneous, IM
Absorption	Essenially zero oral bioavailability. Absorbed well from subcutaneous depot, with onset of effect within about 20 minutes
Solubility	pKa of around 7.1; poor water solubility at high concentrations (because it self-associates into a trimer)
Distribution	VOD=0.25 L/kg; minimally protein bound
Target receptor	Glucagon receptors - G-protein (Gs and Gq) coupled receptors which activate adenylyl cyclase and therefore produce increased cAMP. Mainly found in the liver
Metabolism	30% metabolised in the liver, 30% metabolised in the kidney, the rest degraded by (probably) the reticuloendothelial system
Elimination	Minimal free drug is eliminated in the urine
Time course of action	Half life is about 20-30 minutes
Mechanism of action	By increasing intracellular cAMP, glycogen stimulates cAMP-dependent protein kinases and therefore activates pathways of glycogen breakdown and glucose release (among many metabolic pathways)
Clinical effects	Mainly hepatic effects: Increased glycogenolysis, decreased glycogen synthesis, increased gluconeogenesis, decreased synthesis of VLDLs and increased β-oxidation of fatty acids, leading to ketosis. Also decreased release of insulin, decreased appetite, increased basal energy expenditure. In high doses, increased cardiac contractility and heart rate
Single best reference for further information	Müller et al, 2017