Claims surrounding insulin poisoning cannot be reached from one blood test.

The prosecution alleges that Lucy Letby added exogenous insulin (ActRapid) to ternary parenteral nutrition bags (TPN) and dextrose solutions. TPN and/or dextrose boluses are given to neonates when they cannot take food by mouth, so the nutrients are injected directly into their veins in a hypotonic solution. The prosecution relies on two blood tests from two infants who were experiencing transient hypoglycaemia (low blood glucose) in the period immediately after birth. It appears that the expert witness, Professor Hindmarsh, failed to acknowledge that ~33.7% of all preterm neonates experience hypoglycaemia during this period (Sharma et al., 2017).  The same witness also failed to explain that premature twins have a higher incidence of hypoglycaemia in the neonatal period (Zanardo et al., 1997).  In a study comparing 216 premature twins and 1284 premature singletons, born in the same two-year period, researchers showed that there was a significantly higher risk of hypoglycaemia in twins vs singletons (54% vs 32%).  The same group also found that the risk of twin deliveries was increased at 30-33 weeks gestational age (25% vs 15%), and that twins had higher rates of hospitalisation (50% vs 40%) and an increased risk for requiring cardiorespiratory resuscitation (Zanardo et al., 1997). The mechanism causing hypoglycaemic events in preterm infants is poorly understood. However, both infants presented in a unique set of circumstances which have additionally been overlooked by the expert witnesses. Child F was administered insulin in the immediate days prior, and there is evidence that when insulin administration coincides with infectious disease it can result in the production of autoantibodies to insulin.  Child L was born to a mother suffering from gestational diabetes, a condition that can result in maternal antibodies to insulin crossing through the placenta and causing changes to the kinetics of insulin. The prosecution allege, and this is supported by Prof. Hindmarsh, that findings from blood plasma of the infants show very high levels of insulin (Child F: 4657 pmol/L (or 4657 mU/L) | Child L: 1099 pmol/L (or 1099 mU/L)*), relative to low c-peptide concentrations (Child F: 169 pmol/L ) | Child L: 264 pmol/L ), and that this is shows that the infants had been improperly treated with insulin. There is no other testing performed relating to the insulin levels at the time of hypoglycaemia. Prof Hindmarsh provides no information or scientific explanations which might provide an innocent reason for the discordance in insulin and c-peptide concentration. He fails to mention any factors that might raise doubt over the readings, despite the fact that many exist (Marks, 2005). Additionally, he tolerates the claim that the blood test must mean that insulin was added to the TPN bag. Prof. Hindmarsh has no evidence or expertise which would permit him to claim that a blood test such as that performed in this case is evidence of insulin poisoning, nor proof that insulin was added to TPN. Such a claim is outside the bounds of his expertise, denies a body of evidence that provides an alternative explanation, and is simply speculative. Finally, the calculations that he provides do not appear to be sufficient to result in the insulin concentrations observed.   *it is not clear which units are being used. 

Professor Vincent Marks, MA, DM, FRCP, FRCPath, MAE

The evidence in most cases rests on the results of simultaneous insulin and C-peptide measurement showing inappropriately high insulin in the presence of undetectable, or extremely low, C-peptide levels. This combination has, however, been accepted – too uncritically – as conclusive proof of exogenous insulin administration even when there was no supporting evidence, such as an insulin injection site, and has led to several miscarriages of justice.

Insulin: A short background into blood glucose control

The control of blood glucose levels involves two key hormones: glucagon and insulin. When blood glucose is high, the pancreas releases insulin from β cells.  When insulin binds to its receptor, this triggers the transport of glucose into cells, stimulates glycogen synthesis (glycogenesis) in the liver and muscles and promotes glucose conversion into fatty acids for storage.  When blood glucose is low, glucagon is released from the α cells of the pancreas and this breaks down glycogen to release glucose to make energy for cells. This intricate system of hormonal regulation and feedback mechanisms ensures that blood glucose levels are maintained within a narrow range, providing cells with a constant supply of energy while avoiding harmful fluctuations. Disruptions in this balance can lead to changes in blood glucose levels which results in hyperglycaemia and hypoglycaemia.

Glucagon and Insulin regulate blood glucose

In preterm neonates, hypoglycaemia (low blood glucose) is common in the immediate period after birth.  A key factor in the development of hypoglycaemia relates to the limited storage of glycogen in the unborn foetus, prior to 27 weeks gestation.  The amount of glycogen stored in the liver slowly increases between 17 - 26 weeks gestation, which means infants born prior to 27 weeks are at a particularly high risk for developing hypoglycaemia. After 27 weeks gestation, the glycogen stores rapidly increase, which reduces the risk of incidence of hypoglycaemia in infants born after 27 weeks gestation.  Glycogen is a large molecule which is made up of repeating units of glucose.  Glycogen is made in the liver, muscle and fat cells, and requires the release of insulin.  Insulin binds to a receptor on the outside of cells and this initiates a molecular cascade, which converts glucose molecules to glycogen. Insulin binding to its receptor causes the specialised channels in the cell to increase the transport of glucose form the blood stream to inside the cell.  Preterm infants often have lower glycogen stores, decreased glucose production and reduced insulin sensitivity compared to full-term infants. As a result, they are at a higher risk of developing hypoglycaemia shortly after birth and during the early days of life.

Proinsulin the precursor to Insulin

Proinsulin is an inactive precursor molecule of insulin that is synthesized in the beta cells of the pancreas. It serves as the immediate precursor to insulin, the hormone responsible for regulating blood glucose levels. Proinsulin is a single-chain polypeptide consisting of three regions: the N-terminal B chain, the C-peptide (connecting peptide), and the C-terminal A chain. In neonates, the concentration of proinsulin in the blood is very high and can account for up to 70% of all insulin detected by immunoassay (Hawdon et al., 1995). It is important to note that proinsulin cannot function in the same manner as insulin, and therefore plays no role in the control of blood glucose.  Proinsulin has a half-life ~ 60 minutes, which exceeds that of both insulin and C-peptide (Xu et al., 2022).

Blood testing to detect insulin 

ELISA, or Enzyme-Linked Immunosorbent Assay, is a commonly used laboratory technique for detecting and quantifying the presence of a specific substance, such as insulin, in a sample.  This is the method used for determining the concentration of insulin and c-peptide.  The sandwich ELISA, also known as a capture ELISA, is a variant of the Enzyme-Linked Immunosorbent Assay (ELISA) technique. It is commonly used for the detection and quantification of specific antigens in biological samples. In a sandwich ELISA, the target antigen is "sandwiched" between two antibodies—a capture antibody and a detection antibody. The capture antibody is immobilized on a solid surface, such as a microplate, through non-covalent binding or covalent attachment. This capture antibody is specific to the target antigen and is selected to bind to a different epitope on the antigen than the detection antibody.   A substrate solution specific to the enzyme conjugated to the detection antibody is added. The enzyme catalyzes a reaction with the substrate, producing a detectable signal (such as a colour change or fluorescence) that is directly proportional to the amount of antigen present.  The signal generated is measured using a spectrophotometer or a specialist ELISA plate reader, which quantifies the optical density or fluorescence intensity. This signal is used to determine the concentration of the target antigen in the sample by comparing it to the standard curve generated using known antigen concentrations.

There are significant issues with relying on one single blood test as proof of insulin poisoning - this approach would not be permitted in clinical practice

The insulin concentrations obtained in the blood tests from Child F and Child L are so high that they are only seen in cases of fatal insulin overdose (Marks, 2005; Garg et al., 2012). However, neither infant was reported to experience any adverse consequences, despite the claim that the insulin was exogenous. It is not clear whether the concentration of insulin that has been provided is in pmol/L or mU/L. This matters, as it will provide further indication as to whether the test results are an artefact (meaning whether or not they are valid and usable).  In either instance, the concentrations are incredibly high, and given that the blood glucose level of both infants never reached a concentration which would cause brain injury (<<0.8 mmol/L), there is no real evidence that the insulin that was measured actually constitutes a valid test result.    One particular issue, which was dismissed by witness Dr Anna Milan, is the loss of glucose in the sample.  For Child F, the blood glucose sample that was sent to the lab was tested as having a concentration of 1.3 mmol/L, whereas the in-hospital sample tested at 1.9 mmol/L.  This finding is strongly indicative of the samples not being immediately frozen.  It is widely understood that glucose continues to be converted into biological metabolites when extracted in blood. In absolute terms, a loss in glucose of about 0.67 mmol/L (12 mg/dL) occurs at a concentration of 5.55 mmol/L (100 mg/dL) after 2 h at room temperature (Bruns and Knowler, 2009).  In the case of Child F, the mere fact that the concentration of glucose decreased by ~0.6 mmol/L should lead one to question whether the samples were adequately handled after collection.  Additionally, the tests used to measure the blood concentration of insulin may cross-react with proinsulin, which will result in the insulin levels being reported as artificially high.  Both infants were at additional risk of producing antibodies to insulin. Child F had been given insulin five days prior, and individuals with compromised immune responses, such as babies, experience a greater likelihood of their immune system creating antibodies to insulin if given exogenously (Shen et al., 2019; Liu et al., 2022).  In the case of Child L, the mother had been diagnosed with gestational diabetes, which is associated with the production of antibodies to insulin.  Given that maternal antibodies cross the placenta, it may be that maternal antibodies to insulin extended the half life of insulin there by resulting in a discordance between the concentration recorded in the test and the concentration one would expect to see (Naserke et al., 2001).  In order to demonstrate that the expert witnesses fail to properly assess the relevance of the insulin blood tests, one should assume that the test is valid and then determine what the consequence of the test result is for other variables that can be predicted and modelled. One particular issue is discussed below where, given the high concentration of insulin, the C-peptide test would need to have the capacity to measure C-peptide at the concentration that is predicted in order to be valid. However, the blood tests used to test C-peptide do not give reliable measures of the C-peptide at the concentration we would predict based on the insulin concentration. This means that even if the insulin level were produced endogenously, we would never see the predicted concentration of c-peptide because it is outside the boundaries of the test.  This phenomenon is known as the Hook Effect. 

The Hook Effect and blood testing

The Hook Effect, also known as the prozone phenomenon, is a phenomenon that can occur in certain immunoassays, including enzyme-linked immunosorbent assays (ELISAs) - the test used to detect C-peptide hormone for Child E and L. The Hook Effect describes a phenomenon in which the concentration of C-peptide in the blood is so high that it perturbs the functional abilities of the blood test and results in an artificially low reading, and in some instances no C-peptide will be detected at all. Given that the half life of C-peptide is longer than that of insulin (~30 min v. >10. min), we would expect that a concentration of insulin measured at 4657 pmol/L (or 4657 mU/L depending on the units used), would result in a correspondingly high concentration of c-peptide. Using data on the insulin and c-peptide concentration in preterm infants, we can roughly estimate the concentration of c-peptide we would expect to see given the concentration of insulin. In a study assessing the concentration of insulin and c-peptide in preterm infants (<30 w gestation) it was identified that the insulin concentration was 8.6 mU/L, and the concentration of c-peptide was 1100 pmol/L (Mitanchez-Mokhtari et al., 2004). (*1 unit of insulin = 6 nmol = 6000 pmol, then 8.6 mU/L = 51.6 pmol/L).

Summary of the Hook Effect

The website for the testing facility that performed the blood test confirms that their blood test for insulin cannot determine that the insulin in the blood sample is synthetic.  Instead, Dr Milan relies on the fact that there are low levels of c-peptide detected in the blood tests.  The central claim is that owing to the fact that insulin and c-peptide are produced from proinsulin, it then follows that the concentration of C-peptide should be similar to that of insulin. However, given the high concentration of insulin they observed in their tests, it appears that the c-peptide would not be measured properly because the concentration would be out of range. The concentration of insulin they provided to the court appears to be outside the threshold range for the test parameters. This means that their sample is probably too concentrated and should have been diluted. This appears to be supported by the medical records.  It was reported that Child F’s blood sugar level returned to 9.9 mmol/L seven hours after the blood draw was made and the TPN removed.  However, an insulin concentration as high as that measured for the infant induces hypoglycaemia for days and is nearly always lethal (Garg et al., 2012) .  The test results are totally unreliable and the expert witnesses failed to inform the court about the limitations of the testing performed and even the reliability of the results obtained given the clinical status of the patients.

Identifying causes for hypoglycaemia

There is limited awareness and knowledge of conditions that may cause elevated insulin and low c-peptide.  Even where the Hook Effect appears to be a reliable cause of the low C-peptide concentration, this conclusion should not be the end of the investigation.  In a typical scenario where an infant presents with Whipple's triad (spontaneous hypoglycaemia, blood glucose <2.6 mmol/L and recovery after glucose administration), it is necessary to conduct further investigations surrounding the cause of hypoglycaemia.  A comparison of the insulin concentration recorded in both infants reveals that the test results measured resemble concentrations of those observed in fatal insulin toxicity.  However, such findings also occur in insulin autoimmune syndrome and are not associated with fatality.  Given that neither of the infants experienced any sequelae from the episodes, it seems more likely that the high insulin levels were unreliable and should not have been used to suppose infants had been poisoned.

Blood glucose levels can be altered by viral infection

In neonates, the immune response is very immature, and in preterm neonates this immaturity is worsened by the fact that the immune system develops during the third trimester. Every human derives their initial immunity from infectious disease from their mother during pregnancy.  The antibody response to infection is called the adaptive immune response and the mother passes her antibodies to infectious diseases to the foetus through the placenta. Infants born to term overcome the lack of adaptive immunity through supplemental protection afforded by maternal antibodies. However, this transfer of maternal antibodies occurs throughout the third trimester and babies born prior to 35 weeks gestation are at increased risk from infectious diseases owing to the incomplete transfer of maternal antibodies.  Picarnovirus family members represent a group of viruses which disproportionality infect neonates under the age of 3 months old.  It is believed that a primary reason for this skewed incidence of infection is a combination of the lack of maternal antibodies and the unique mechanisms these viruses employ to evade detection by the immune system.  A key feature of members of the picarnovirus family is their ability to perturb blood glucose levels in neonates under 3 months old and incorrectly develop antibodies to hormones in the blood glucose regulatory pathway.

Confounding variables surrounding claims of insulin poisoning

It was alleged that in both cases of insulin poisoning, insulin was added to the total parenteral nutrition (TPN) solution. In the case of the first infant, the recorded concentration of insulin in the blood was 4657 pmol/L. This level of insulin is incredibly high and it exceeds the levels typically observed with exogenous insulin administration. In the second case, the concentration of insulin was 1099 pmol/L. This is still very high, given the amount of exogenous insulin required to cause such a significant increase in serum insulin.

The differences between intravenous and subcutaneous insulin 

Owing to the short half-life of insulin, the exact amount required to produce the recorded blood concentration would be dependent on the final volume in which the insulin was diluted. Further, the concentration of insulin provided via the IV would require dosing which ensured that serum insulin was maintained in steady state conditions. The actual amount of insulin added to the TPN would have to be sufficient to maintain insulin levels at a similar concentration throughout the period in which the infant is hypoglycaemic in order to support the claims associated with contamination of the TPN bag with insulin.

Insulin adsorbs to the infusion sets and this would need to be factored into any calculated amount of insulin

There is direct evidence that insulin adsorbs to polyethylene (PE) and polyvinal chloride (PVC) lines, which are used to deliver TPN. Insulin adsorption capacity decreased hyperbolically with the flow rate for both PE and PVC, where low flow scenarios result in greater insulin adherence to infusion lines. When the infusion flow rate was halved from 1 to 0.5 mL/h, twice as much insulin adsorbed to the line. Insulin loss to adsorption resulted in up to ~50% of intended insulin not delivered over 24 hours in a low flow and low concentration context. Adsorptive insulin loss is greatest within the first one to two hours of new insulin infusion line use, where typically only 20%-80% of the desired insulin dose is delivered. Insulin adsorptive loss is much more significant where low concentrations and flow rates are used, as is common in neonatal and paediatric intensive care units (NICU and PICU) (Knopp et al., 2021).

Insulin added to TPN is not stable

There is evidence to suggest that the addition of insulin to TPN is not stable and that within the first 8 hours, the insulin concentration decreases by 30%-50%. The stability of insulin in TPN is dependent on the other micronutrients added and the pH of the solution. Given that TPN contains high amounts of dextrose, glycation of insulin via a Maillard reaction might be involved in the loss of insulin, as has been described by Fry et al. for amino acids. This phenomenon has been extensively described in vitro and in vivo for proteins in general and for insulin in particular (Henry et al., 2021). This finding is particularly notable given that Child F and Child L were prescribed TPN with 10% Dextrose, or a simple dextrose solution.  This means that based upon the findings in the described study, the amount of insulin available for infusion would have decreased by 50% by the time the blood was taken. This would require the starting concentration of insulin to be even higher than that assumed having factored in the insulin lost from adsorption to the TPN bag and the IV tubing.

Differential Diagnosis

Several studies have identified a link between Coxsackievirus B (CVB) infections and the development of autoimmune diseases. It has been observed that individuals with certain autoimmune conditions, such as type 1 diabetes, dilated cardiomyopathy and myocarditis often have a history of prior CVB infection (Root-Bernstein et al., 2023; Sioofy-Khojine et al., 2018). CVB infections can trigger autoimmunity through various mechanisms. One proposed mechanism is molecular mimicry, where viral proteins resemble self-antigens, leading to cross-reactive immune responses. In a study into the production of autoantibodies to insulin (IAA), it was observed that children who had signs of a CVB infection either alone or prior to infections by other CVBs were at the highest risk for developing IAA.  The authors identified 91 children who expressed IAAs in the immediate period of enterovirus infection (Sioofy-Khojine et al., 2018).  The cause of the hypoglycaemic episodes observed in the neonates may be due to a viral infection, such as Coxsackievirus B, which caused aberrant production of antibodies to insulin, thereby extending its half life and giving the appearance that the insulin in the blood has increased..  Without any blood testing, such a hypothesis cannot be ruled out.  Importantly, Coxsackievirus B was detected in the UK during the time of both infections and is linked to symptoms that were observed in the twin pair of each case (Kadambari et al., 2016).

Coxsackievirus B (CVB) induces autoantibodies to insulin which perturbs blood glucose levels

A recent studied identified that autoantibodies to insulin were produced in response to Coxsackievirus B (CVB) infection in infants. The Finnish study demonstrated that 91 children who had experienced CVB infections also produced autoantibodies to insulin in the period of infection (Sioofy-Khojine et al., 2018). These antibodies were not found in unaffected controls.  In a separate study, it was demonstrated that children treated with Actrapid who experienced viral infection also showed an increased production of autoantibodies to insulin in the period during infection.  Autoantibodies to insulin result in very high levels of insulin >1000 pmol/L. This is accompanied by a shift in the ratio of insulin: C-peptide. This is because the C-peptide is not affected by the autoantibodies, which means the C-peptide is excreted from the body at the usual rate (~30 minutes). Conversely, the half-life of insulin increases dramatically when autoantibodies are bound and the insulin is released slowly.  The result is discordance between the typical ratio between insulin and C-peptide.

Autoantibodies to insulin cause hyperglycaemia first

Molecular mimicry may link enterovirus infections with the development of autoantibodies to insulin. It involves a process whereby viral proteins may share similarities to insulin or other pancreatic antigens, leading to cross-reactive immune responses. In this scenario, the immune system, activated by the enterovirus, may mistakenly recognise insulin as foreign and mount an immune response against it. This cross-reactivity could result in the production of autoantibodies targeting both the viral proteins and insulin. When the body creates autoantibodies to insulin, the effect is a sudden hyperglycaemic event, which occurs due to the reduced availability of insulin and a concomitant increase in blood glucose levels. The hyperglycaemia is transient and is followed by hypoglycaemia. This is because the antibody-bound insulin results in increased levels of insulin in the blood.  This finding is particularly relevant given that Child F presented with hyperglycaemia which became hypoglycaemia.

The insulin blood test evidence was insufficient and lacked relevant information needed to properly assess the findings

It is notable that the infant with the insulin level of 4757 pmol/L was stated to have had high creatinine and urea. These molecules are associated with the rate of filtration via the kidney. The infant also had signs of jaundice, which indicates impaired liver function. If the plasma concentration of insulin in the infant had really been given exogenously, the concentration is so high (4657 pmol/L) that the infant would have been comatose for a period of days. Instead, the infant recovered from hypoglycaemia, and 24 hours after the first reading of 0.8 mmol/L glucose he had a blood glucose level of 9.9 mmol/L (hyperglycaemia). Individuals who overdose on insulin do not have blood levels as high as 4657 pmol/L. In one case, a woman took one vial of insulin and had a blood insulin level of ~2600 pmol/L. She was in a coma for three days (Gundgurthi et al., 2012). It seems more plausible that the initial hyperglycaemic event represented antibodies to insulin or the receptor being made and this caused insulin to be bound by antibodies and unable to control blood glucose levels. Given that the hyperglycaemia was observed in both infants soon after birth, it is plausible to assume that the infants were experiencing a viral infection such as Coxsackievirus B, which resulted in their immune systems producing antibodies to insulin. When antibodies are first produced, they will bind insulin and stop it from functioning. This would require that for the first twin, the insulin antibodies had high affinity for insulin and bound tightly, thereby sequestering the insulin and causing hyperglycaemia. For the second twin, the insulin antibodies would have a reduced affinity for insulin which would permit an initial hyperglycaemic event.  The antibodies would initially sequester insulin and then slowly release the insulin resulting in an increased insulin half life, which is discordant with c-peptide levels and causes prolonged, yet transient, hypoglycaemia.


The blood testing performed would not have been able to detect the presence of antibodies to insulin or proinsulin, and importantly the test cannot be used to claim that insulin was added to TPN or dextrose solutions. Such a claim is mere speculation.  In the Beverly Allitt case, blood testing was extensive and they tested for antibodies to insulin and proinsulin (Marks and Richmond, 2008). In the CoCH cases, at the time of testing the hospital was asked if they wanted to do further testing and the hospital declined. Had the hospital required the deceased twin to have an autopsy and had they collected blood samples from the deceased twin, it would have been possible to test the blood for antibodies to insulin (Shen et al., 2019). The doctors at CoCH did not collect any blood samples from the infants who died in these cases. As a result, it is impossible to know whether they had undiagnosed conditions or viral infections that could explain the presentation of illness in so many twin pairs.