Effect of Hypoglycemia on Measures of Myocardial Blood Flow and Myocardial Injury in Adults With and Without Type 1 Diabetes: A Prospective, Randomized, Open-Label, Blinded Endpoint, Cross-Over Study

Aims: Hypoglycemia provokes a profound autonomic response in humans including cardiovascular effects. This study examined the effect of experimentally-induced hypoglycemia on measures of myocardial blood �ow and myocardial injury in adults with, and without, type 1 diabetes. Methods and Results: In a prospective, randomized, open-label, blinded, endpoint cross-over study, 17 young adults with type 1 diabetes with no cardiovascular risk factors, and 10 healthy non-diabetic volunteers, underwent hyperinsulinemic euglycemic (blood glucose 4.5-5.5 mmol/L) and hypoglycemic (2.2-2.5 mmol/L) clamps. Myocardial blood �ow was assessed using transthoracic echocardiography Doppler coronary �ow reserve (CFR) and myocardial injury using plasma high-sensitivity cardiac troponin I (hs-cTnI) concentration. During euglycemia, a non-signicant trend for lower CFR was observed in participants with type 1 diabetes than in those without type 1 diabetes (3.66±0.47 versus 3.92±0.85), and a non-signicant lower trend also occurred during hypoglycemia (type 1 diabetes: 3.54±0.58 versus non-diabetes: 3.80±0.84). A generalized linear mixed-model analysis was performed, with diabetes status and euglycemia or hypoglycemia as factors affecting CFR. No signi�cant statistical difference in CFR was observed for diabetes status (p=0.23) or between euglycemia and hypoglycemia (p=0.31). No changes in hs-cTnI occurred during hypoglycemia or in the recovery period (p=0.86). Conclusions: While the observed reduction in coronary �ow reserve did not achieve signi�cance during exposure to insulin-induced hypoglycemia in healthy young men with type 1 diabetes, with no evidence of myocardial injury, adverse cardiovascular effects of hypoglycemia cannot be excluded in older people who have coronary disease. Further studies are required to investigate this putative problem.


Introduction
Diabetes is now a global pandemic (1) in which the principal cause of death is cardiovascular disease (2).While strict glycemic control improves some outcomes, clinical trials in people with type 2 diabetes who have cardiovascular disease have shown that intensive glucose-lowering therapy is associated with a higher mortality (3,4).It has been suggested that exposure to hypoglycemia may be contributing to this greater morbidity and mortality (5,6).
Hypoglycemia is a common occurrence in people with insulin-treated diabetes.It causes profound autonomic stimulation leading to activation of the sympatho-adrenal system with profuse catecholamine release.This has pronounced systemic and regional hemodynamic effects (7), including large increments in stroke volume and cardiac output (8), caused by increased left ventricular ejection fraction and myocardial contractility (9).This augmented cardiac workload may provoke myocardial ischemia or cardiac failure in people who have established cardiac disease.
Mechanisms that could promote adverse effects of hypoglycemia on cardiovascular function have been proposed (10).Endothelial dysfunction is a key pathophysiological pathway underlying macrovascular complications in type 1 and type 2 diabetes.A detrimental effect of hypoglycemia on peripheral endothelial function has been reported (11).However, several differences exist between peripheral and coronary arterial endothelium, such as the presence of shunt vessels and the microvascular architecture (12).Direct measurement of coronary vasomotor function may therefore provide a more valid measure of any cardiovascular impairment resulting from limited vascular responsiveness.Measurement of coronary microvascular dysfunction by calculation of coronary ow reserve (CFR) is the preferred investigative technique (13).Coronary ow reserve measures the capacity of the coronary circulation to increase ow during maximal resistance vessel vasodilatation.Maximal hyperemia is achieved by intravenous infusion of adenosine (14) and coronary ow velocity can be measured non-invasively by transthoracic Doppler echocardiography (15,16).CFR is reduced in people with type 1 diabetes during euglycemia (17) and in individuals with type 1 diabetes with evidence of microvascular disease in the form of retinopathy (18).
CFR has also been shown to correlate well with long term hard outcomes in coronary heart disease (19,20).It is unclear how hypoglycemia affects CFR.
Cardiac troponins are a biomarker of myocardial injury.In recent years, the introduction of a highly sensitive cardiac troponin I (hs-cTnI) assay allows detection of small degrees of myocardial damage.Lowering the diagnostic threshold for myocardial injury has improved outcomes for people with type 1 myocardial infarction (21).Hs-cTNI also provides prognostic information regarding long term outcomes in cohorts of people with underlying coronary heart disease (22).
People with type 2 diabetes often have multiple cardiovascular risk factors that can confound attempts to investigate hypoglycemia as a causative mechanism for cardiovascular morbidity and mortality.Younger people with type 1 diabetes of relatively short duration are less likely to have acquired these additional risk factors.To examine whether hypoglycemia might adversely affect myocardial blood ow and cause myocardial injury, coronary ow reserve and cardiac troponins were therefore measured during experimentally-induced hypoglycemia in people with type 1 diabetes who had no confounding factors.

Study Participants
Young healthy male adults with type 1 diabetes with no microvascular complications or cardiovascular risk factors were recruited from diabetes out-patient clinics in Lothian, Scotland.Men without diabetes, matched for age, were recruited by poster advertisements and from a database of volunteers.Only male subjects were studied to avoid the confounding effect of the variability of coronary ow reserve that occurs during the menstrual cycle (23).In addition, because the counterregulatory hormonal responses to hypoglycemia differ between men and women (24), the study was con ned to men to avoid a potential effect of gender on the magnitude of the sympatho-adrenal stimulus during hypoglycemia.
Exclusion criteria included co-existent systemic disease, malignancy, chronic alcoholism, psychiatric disorder, any history of cardiac conduction abnormality, impaired awareness of hypoglycemia (as assessed by the method of Gold et al (25)), past history of severe hypoglycemia, and any evidence of overt microvascular complications including retinopathy and neuropathy or the presence of microalbuminuria.The participants did not have any overt evidence of autonomic neuropathy.The use of any medications apart from insulin therapy, including beta-blockers and angiotensin converting enzyme inhibitors, was an exclusion criterion.
A total of 17 male adults with type 1 diabetes and 10 age-matched individuals without diabetes were studied (Table 1).Participants with type 1 diabetes had reasonable glycemic control, (average glycated hemoglobin (HbA1c): 8.0 ± 1.1%; 64 ± 11 mmol/mol), with a median duration of diabetes of 15 years (range 2-35 years), which is consistent with average quality of glycemic control recorded in the adult population with type 1 diabetes in Scotland.The two groups of participants did not differ in age or bodymass index (Table 1).The participants performed more frequent glucose testing 24 hours before each of the glucose clamp studies to ensure that they had not been exposed to antecedent hypoglycemia before the study.The study was conducted with informed written consent of all subjects, the approval of the Lothian Medical Research Ethics committee, and in accordance with the Declaration of Helsinki.

Study Design
Participants attended for two study visits, performed on separate days at least two weeks apart to avoid any potential carry-over effects.Two experimental conditions, hypoglycemia (blood glucose 2.5 mmol/L; 99 mg/dL) and euglycemia (4.5 mmol/L; 81 mg/dL), were studied in a prospective, randomized, open-label, blinded endpoint, (PROBE) cross-over study.Both the group with type 1 diabetes and matched individuals underwent the experimental conditions, and then crossed-over.The order of the experimental method was randomized using alternate randomization by the operator of the glucose clamp, while the endpoint was blinded to the echocardiographer (Supplemental Fig. 1).This study design has been employed previously by our group (26), with modi cations for the cardiovascular investigations.
Participants attended in a fasting state, having abstained from consumption of caffeine-containing food and beverages for 24 hours.Venous cannulae were inserted for intravenous infusion of dextrose and insulin, and blood sampling.A modi ed version of the hyperinsulinemic glucose clamp was employed (27).To arterialize blood samples, the nondominant arm was wrapped in a heated blanket with a retrograde intravenous cannula inserted into the forearm.An additional cannula was inserted into the non-dominant antecubital fossa to infuse insulin (human Actrapid; Novo Nordisk, Crawley, U.K.) and 20% dextrose.Insulin was infused at a constant rate of 1.5 mU/kg/min with a Gemini PCI pump (Alaris Medical Systems, San Diego, CA).Blood samples were taken at 5-min intervals and analyzed by a glucose oxidase method (2300 STAT; YSI, Yellow Springs, OH).The dextrose infusion rate was adjusted to maintain the appropriate arterialized blood glucose concentration.During a run-in period, arterialized blood glucose was maintained at 4.5 mmol/L for 20 min.Blood glucose was then either maintained at 4.5 mmol/L throughout (the euglycemia condition), or lowered over 20 min to 2.5 mmol/L (the hypoglycemia condition), and maintained at this level for 30 min before restoration of euglycemia.During the glucose clamp, the participants underwent an ultrasound examination by a trained ultrasound operator, using a well-described technique (16, 23).The timepoints were labelled as baseline, experimental (either euglycemia or hypoglycemia-blinded to the sonographer), and recovery (Supplemental Fig. 1).Continuous electrocardiographic monitoring and regular blood pressure monitoring were performed throughout the study.

Coronary Flow Velocity Measurements
During each study condition, the left anterior descending coronary artery was visualized by trans-thoracic echocardiography.Transthoracic Doppler echocardiography was used for a non-invasive estimation of coronary ow velocity (CFV), and maximal coronary vasodilatation was induced with an adenosine infusion to allow calculation of CFR.Imaging of the left anterior descending (LAD) artery and measurement of coronary blood ow velocity was performed using a 7.0 MHz transducer (Acuson Sequoia 512, Siemens Medical Solutions, Berkshire, UK).Baseline spectral Doppler signals were recorded initially in the distal portion of the LAD coronary artery over ve cardiac cycles at end-expiration.To measure coronary ow reserve (CFR), intravenous adenosine was administered (0.14 mg/kg/min; Adenocor, Sano ) for up to 4 min to record spectral Doppler signals during hyperemic conditions (14).Coronary velocities were measured at baseline and at peak hyperemic conditions from the Doppler signal recordings.Measurements were averaged over three cardiac cycles.CFR was de ned as the ratio of hyperemic to basal velocities, using maximum velocity (Vmax) parameters (Supplemental Fig. 2).Blood pressure was recorded at baseline, during adenosine infusion and at recovery.CFR was calculated at baseline, during the experimental phase (0-20 min) and in the recovery phase.

High-sensitivity Cardiac Troponin I Concentration
Blood samples were taken prior to assessment of CFR, during the experimental hyperinsulinemic clamp, and during the recovery period (Supplemental Fig. 1).High-sensitivity cardiac troponin I concentrations were determined using the ARCHITECT STAT high-sensitive troponin I assay (Abbott Laboratories, Abbott Park, IL).This is the rst clinically approved high-sensitivity troponin I assay, which has excellent precision at very low concentrations.The limit of detection is 1.2 ng/L and precision pro ling in our laboratory has demonstrated an inter-assay coe cient of variation (CV) of < 10% at 5 ng/L.The upper reference limit or 99th centile is 16 ng/L for women and 34 ng/L for men (28-30).

Statistical Methods
A power calculation was performed using results from a previous study using a similar technique (31), a sample size of 12 allows an 80% chance of detecting a 0.57 difference in CFR, which is considered clinically relevant.The effects of hypoglycemia on coronary ow reserve were assessed statistically by generalized linear mixed-effects modeling, with the experimental condition (hypoglycemia and euglycemia) and diabetes status as variables affecting CFR.We have speci ed per subject intercepts in our mixed effects model by tting individual subjects as a random effect to account for variation in baseline values across our study population.We also added an interaction term for diabetes status and experimental condition (hypoglycemia or euglycemia) in our mixed effects model Statistical signi cance was taken as a two-sided p < 0.05.Unless speci cally stated, results are mean ± standard deviation.The hsTnI data were log-transformed due to the skewed distribution.Heart rate, blood pressure, glucose and troponin data were analyzed using paired t-tests.Analysis of the results was performed using R stats (R Foundation for Statistical Computing, Vienna Austria, version 3.6.1).

Results
Study participants were all healthy young men with normal body-mass index (BMI).Baseline characteristics of both groups are provided in Table 1.During the hypoglycemia session (Table 2), the mean glucose nadir was 2.34 ± 0.2 mmol/L; 42.12 ± 3.8 mg/dL (p < 0.001 compared to baseline), which is su cient to stimulate a brisk sympatho-adrenal counterregulatory response.Heart rate, systolic and mean blood pressure did not change during euglycemia in either group.During hypoglycemia, a trend was apparent towards an increase in heart rate and systolic blood pressure (BP), and a decrease in diastolic BP in both groups (Table 2).In the type 1 diabetes group, the increments in heart rate and systolic blood pressure did not reach statistical signi cance.In the non-diabetic group, the heart rate increased from 71 ± 9 beats per minute (bpm) to 78 ± 8 bpm (p = 0.02) while the systolic BP increased from 116 ± 11 to 124 ± 12 mmHg (p = 0.001).

Coronary Flow Velocities and Reserve
No differences were observed between the coronary ow velocities of the group with type 1 diabetes and the non-diabetic group at baseline (Supplemental Fig. 3).During hyperemia after adenosine infusion, a trend for lower coronary ow velocities was observed in people with type 1 diabetes, but this did not reach statistical signi cance (Supplemental Fig. 3).
During euglycemia, a trend for a lower coronary ow reserve was observed in young adults with type 1 diabetes, compared to people without diabetes (3.66 ± 0.47 versus 3.92 ± 0.85), but this did not achieve statistical signi cance.During hypoglycemia, coronary ow reserve trended non-signi cantly lower in those with type 1 diabetes than in the non-diabetic participants (3.54 ± 0.47 versus 3.89 ± 0.89) (Fig. 1).
From the generalized linear mixed model analysis, no statistical signi cance was reached for euglycemia or hypoglycemia affecting CFR (p = 0.31).No statistical signi cance was noted for the effect of diabetes status on CFR (p = 0.23).This exploratory model did not demonstrate any signi cant interaction between diabetes status and experimental condition (P value = 0.938)

High-sensitivity Cardiac Troponin I
The hs-cTnI values were in a skewed distribution and were therefore log transformed for statistical analysis.No changes were observed in plasma high-sensitivity cardiac troponin I concentrations during euglycemia or hypoglycemia during the recovery phase in either group (Fig. 2).

Discussion
The present study used a well validated, non-invasive method to examine the effect of acute hypoglycemia on real-time coronary arterial ow in adults with and without type 1 diabetes, with the coronary ow ratio (CFR) being measured using transthoracic Doppler echocardiography.During acute hypoglycemia, young adult males with type 1 diabetes had a trend of lower coronary ow reserve compared with an age-matched group of non-diabetic males.The modest decline in CFR was well tolerated in young men with type 1 diabetes who were otherwise healthy and had no evidence either of microvascular complications or of coronary heart disease; normal coronary reserve was maintained.
The direct effects of hypoglycemia on cardiac function have proved di cult to elucidate as insulin per se exerts direct effects on the heart.Fisher et al (9) showed that administration of insulin caused an immediate increase in left ventricular ejection fraction and provoked sympathetic activation, both of which occurred before any fall in blood glucose.As blood glucose declined progressively, these responses became more pronounced, with the maximal changes coinciding with the glucose nadir (9).A strength of the present study was the ability to use a non-invasive real-time assessment of coronary ow reserve during acute hypoglycemia.By using a hyperinsulinemic glucose clamp it was possible to compare CFR during euglycemia and hypoglycemia, and between type 1 diabetes and the non-diabetic state.The changes observed are therefore related to the low blood glucose and to counterregulatory mechanisms, and not to the intravenous infusion of insulin per se.Furthermore, this study provides a direct assessment of coronary vasomotor function which excludes other potential confounders such as hypertension or microvascular disease.
The main limitation of the study is the small sample size.While this was in part a consequence of the robust exclusion criteria, the demanding study design also limited recruitment of potential participants, speci cally because of the use of adenosine to induce maximal hyperemia of the coronary blood vessels.The rationale for using adenosine was its short half-life, which meant that the CFR values from one measurement to the next would not be confounded by residual effects of adenosine.The disadvantage of this approach is that adenosine is often poorly tolerated because it induces unpleasant side-effects of chest tightness and facial ushing, which diminishes the willingness of volunteers to participate.In addition, the glucose clamp procedure is onerous and had to be repeated at least two weeks apart, which was a further limitation to recruitment.
In retrospect, the power calculation may have bene ted from an a priori calculation of possible attrition of participants, or non-concordance with the protocol.Additionally, the magnitude of the primary outcome (difference of 0.57 in CFR) may have been over-ambitious, given that our participants with diabetes were t and in good health.In view of the relatively small sample size of our study, it is possible that we did not have su cient power to detect a signi cant interaction between diabetes status and experimental condition in the generalised linear mixed model analysis.Therefore, while this study models the direct effect of hypoglycemia per se on coronary vasomotor function, the generalisability of the results must take into account the sample size and the exclusion of female participants.
A previous investigation by Rana and colleagues (32), which to our knowledge is the only other study to have explored the effect of acute hypoglycemia on the myocardial circulation, used sequential hyperinsulinemic glucose clamps and dipyridamole-induced stress echocardiography to measure myocardial blood ow reserve during euglycemia and acute hypoglycemia in adults of both sexes, 28 with, and 19 without type 1 diabetes (32).The age range was wider than in the present study and included people with microvascular disease.Hypoglycemia induced a signi cant fall in myocardial blood ow reserve in both groups, with lower values being observed in the group with type 1 diabetes at all times of measurement.A statistically signi cant association was observed with the presence of microvascular complications.In contrast to the present study design, no time interval was allowed between the initial euglycemia and the subsequent induction of hypoglycemia, so that myocardial blood ow reserve rose during the period of protracted euglycemia, which may have in uenced the effect on the subsequent hypoglycemia (32).In addition, the order of the euglycemic and hypoglycemic clamps was not randomized, which may have introduced observer bias and an order effect.Furthermore, the long halflife of dipyridamole might have in uenced the results.The present study may therefore have provided a more direct model of coronary ow reserve with fewer confounding factors such as the presence of microvascular disease and a possible residual effect of dipyridamole.
In the present study no signi cant change in CFR was observed during hypoglycemia.While a nonsigni cant trend towards a lower CFR was observed in the participants with type 1 diabetes during euglycemia, a trend towards a decline in CFR was also observed during hypoglycemia, consistent with previous observations (17).The lowest CFR values during hypoglycemia were observed in the participants with type 1 diabetes.The increments in heart rate and systolic blood pressure in the group with type 1 diabetes did not achieve statistical signi cance (Table 2), which was unexpected with this degree of hypoglycemia.The small sample size may not have allowed su cient sensitivity to detect small variations in pulse and blood pressure.An alternative interpretation is that some participants with type 1 diabetes may have had some degree of autonomic dysfunction or a diminished catecholamine response to hypoglycemia, which contributed to the modest hemodynamic changes and the lower CFR values observed in the group with type 1 diabetes.As formal assessment of autonomic function in the participants was not made nor were plasma catecholamines measured, this possibility cannot be excluded.
A Danish study using non-invasive cardiac magnetic resonance imaging has reported that myocardial blood ow reserve was higher at rest and lower during vasodilatory stress in people with type 2 diabetes compared with non-diabetic controls (33).Impaired myocardial blood ow reserve was associated with microvascular complications (albuminuria and retinopathy) of diabetes.The present study explicitly excluded people with overt microvascular disease; it is possible that the development of cardiac microangiopathy may underlie an abnormal response to hypoglycemia in type 1 diabetes (33).
While no change in the highly sensitive troponin values was observed during acute hypoglycemia, this was measured in close temporal proximity to the blood glucose nadir.It is possible that measurement in the immediate 'recovery' stage was made too early to detect a rise in plasma troponin and exclude evidence of myocardial insult.
The results of the present study imply that any putative cardiac harm of hypoglycemia is unlikely to be mediated solely through coronary vasomotor dysfunction.Other potentially harmful factors associated with hypoglycemia may be required, such as the promotion of pro-thrombotic mechanisms (34,35), endothelial abnormalities (11), or altered cardiac electrical conduction (36, 37) , .It should also be noted that this study speci cally studied acute hypoglycemia, and the cumulative effects of recurrent hypoglycemia have not been examined.
Recent randomized controlled trials (RCTs) that did not target strict glycemic control but used drugs with a low risk of hypoglycemia have shown bene cial cardiovascular outcomes (38,39).This is in direct contrast to previous RCTs, in which very strict glycemic control was pursued and the incidence of severe hypoglycemia was high (10).These ndings suggest that avoidance of hypoglycemia is important to achieve cardiovascular bene t.

Conclusion
Although in the present study hypoglycemia had no effect on markers of ischemia, a small reduction in CFR was apparent during hypoglycemia, with the lowest numerical value occurring in young adults with type 1 diabetes during hypoglycemia.Further larger studies that include female participants are required to con rm or refute this observation.If the present observation of a lower CFR is con rmed, this would raise concern that hypoglycemia may promote myocardial ischemia in older people with diabetes who have established coronary heart disease.While coronary blood ow should be examined during hypoglycemia in people with type 2 diabetes who have a greater risk of cardiovascular disease, this was not undertaken because of the potential cardiac risk to such participants.The present study should be repeated in a larger number of people with type 1 diabetes, using a more tolerable form of investigation and a real time biomarker of cardiac injury.As population studies have shown an association between severe hypoglycemia and cardiovascular risk (5,40) more mechanistic studies are required to elucidate potential reasons for this association.

Statement of Contributions:
RMN, ACW, NNL and NLM contributed to acquisition, analysis and interpretation of the data and to preparation and critical revision of the manuscript.KKL contributed to analysis of the data and preparation of the manuscript.DEN, CCL, AJG and BMF contributed to the study and design and to the preparation and critical revision of the manuscript.RMN is the guarantor of this work, had full access to the data, and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Figure 1 CFR (scatter and violin plot) in participants with type 1 diabetes and people without diabetes under euglycemic and hypoglycemic conditions.Blue represents people with diabetes, and red for people with diabetes.During euglycemia the mean CFR for people with diabetes versus without was 3.66±0.47versus 3.92±0.85respectively.During hypoglycemia the mean CFR was 3.54±0.47versus 3.89±0.89respectively.

Table 2
Plasma glucose, heart rate, systolic, and diastolic blood pressure values during hypoglycemia and euglycemia in the groups of participants, with and without type 1 diabetes