Severe COVID-19 associated hyperglycaemia is caused by beta cell dysfunction: a prospective cohort study

doi:10.21203/rs.3.rs-2203673/v1

Background

Hyperglycaemia was shown to be among the features of severe acute COVID-19 infection both in the acute and convalescence period. Mechanisms contributing to its development and/or maintenance in a post-COVID phase are still unclear.

Materials and Methods

Survivors of severe COVID-19 but without a known history of diabetes were examined at baseline (T0) and after 3 (T3) and 6 (T6) months: indirect calorimetry and OGTT. Insulin response and sensitivity (IS) were expressed as insulinogenic (IGI), disposition (DI), and Matsuda insulin sensitivity index (ISI), respectively. Resting energy expenditure (REE) was calculated using the Weir and Harris-Benedict equation and substrate preference using the respiratory quotient (RQ) and nitrogen losses.

Results

Thirty-two patients (12 women) were available for the analyses at T0 and 26 at T6. Baseline examination was within 21±6.5 days after COVID-19 diagnosis. Patients were hypermetabolic at baseline (30.7 ± 4.3 kcal/kg lean mass/day, ~120% predicted values) but REE declined over the 6 months (ΔT6-T0 mean difference (95% CI): -5.4 (-6.8, -4.1) kcal/kg lean mass/day, p<0.0001). 20 patients at T0 and 13 patients at T6 had hyperglycemia. None of the patients had positive islet autoantibodies. Insulin sensitivity at T0 was comparable between hyperglycaemic (H) vs. normoglycaemic (N) patients (T0 ISI_H=3.07 ± 1.18, ISI_N=3.23 ± 1.72, p=0.66), whereas insulin response was lower in the H group only (DI_H=3.19 ± 2.16 vs DI_N=8.78 ± 6.37, p=0.005). Over the 6 months, IS improved in the H group (ΔT6-T0 mean difference for ISI (95% CI): 1.84 (0.45, 3.24), p=0.01)), whereas IGI and DI did not improve in either group.

Conclusions

Severe COVID-19 infection was associated with hypermetabolism that did not persist over the follow-up. Patients were insulin resistant in the acute phase, but only those with insufficient insulin response developed hyperglycaemia. Insufficient beta cell response was a possible mechanism leading to hyperglycaemia.

Health sciences/Endocrinology/Endocrine system and metabolic diseases/Diabetes/Type 2 diabetes

Health sciences/Diseases/Endocrine system and metabolic diseases/Pre-diabetes

COVID-19

hyperglycaemia

diabetes

insulin resistance

beta cell dysfunction

hypermetabolism

COVID-19, a disease caused by a novel severe acute respiratory syndrome coronavirus 2 (SARS CoV-2), emerged in 2019 and has since then spread around the world, causing an unprecedented pandemic ¹. Although leading symptoms of COVID-19 relate to the respiratory system, metabolic disturbances, especially in glucose homeostasis, have been reported soon after the pandemic outbreak ^2,3. Hyperglycaemia was the most commonly described phenomenon that had been shown to be associated with an increased risk of adverse outcomes (e.g., the severity of illness, physiologic stress, and need for intubation) ^4,5, especially in patients with known diabetes ⁶. Although elevated blood glucose levels are common during a critical illness, the prevalence of hyperglycaemia in patients with COVID-19 infection appears to be unexpectedly high ⁷.

Since the first reports, intensive research has been conducted on possible causes for this phenomenon with many suggested mechanisms involved ⁴. Abnormities related to both insulin resistance (IR) and beta cell dysfunction were suggested ^8,9. IR is associated with hypermetabolism during the acute state of COVID-19 as a stress response and increased drive of counterregulatory hormones (such as catecholamines and cortisol) resulting in a decline in both skeletal muscle and hepatic insulin function ¹⁰. Increased lipolysis and elevated circulating non-esterified fatty acids (NEFA) were also described ¹¹. Damaged mitochondria in the peripheral tissues are unable to meet the increased energetic demands caused by the SARS CoV-2 triggered inflammation, with a subsequent increase in reactive oxygen species (ROS) production ¹². Beta cells, on the other hand, can be affected by many suggested mechanisms: systemic and islet renin-angiotensin-aldosterone system (RAAS) activation, islet redox stress, systemic and islet inflammation, islet amyloid deposition, islet fibrosis and/or failure due to apoptosis and capillary rarefaction ¹³. Direct damage by replicating the virus in beta cells was also suggested ¹⁴. The virus may also damage the endothelium, which further allows the impairment of microvasculature and pericytes in the islets of Langerhans ¹⁵.

Retrospective studies focusing on the incidence of new-onset diabetes or hyperglycaemia during the acute stage of the disease showed wide distribution between 2.8 and 49.2% in the general population ^16–19 and 35% to 85% in the critically ill ^9,20,21. The studies are mainly of retrospective design and methodologically very heterogeneous with different populations, end-point, and glycaemic cut-offs. Moreover, it is not clear how long the observed changes persisted over time and whether the insulin-glycaemic indices tend to improve, limited evidence suggests they persist for up to two months ⁸.

The aim of this prospective cohort study was to investigate whether hyperglycaemia persists in patients surviving severe COVID-19 over a follow-up period of six months and to explore possible causes of hyperglycaemia, including insulin sensitivity/secretion, energy expenditure, substrate preference, and metabolic flexibility.

Subjects and Design

The monocentric prospective cohort study COMETA (COVID-19 metabolic and nutritional consequences: prospective observational study) was conducted from March to November 2021 at University Hospital Královské Vinohrady, Prague, Czech Republic. All consecutive patients aged over 18 years who were short-term survivors (i.e. weaned from artificial ventilation/ECMO/oxygen) were screened (March-April 2021) at specialized COVID-19 wards. Inclusion criteria were severe COVID-19 (PCR/antigen nasopharyngeal test positive, bilateral COVID-19-associated pneumonia verified by CT/CXR with a respiratory failure defined as any need for oxygen support). Exclusion criteria were known diabetes in medical history, chronic lung disease, active cancer, neurological disease with impaired mobility, acutely decompensated endocrine disease (thyroid, adrenal, etc.), and pregnancy in women. Patients were to be examined once weaned from oxygen support, not later than 4 weeks from the onset of the disease to capture the acute phase. In total, 37 patients met the eligibility criteria and were enrolled. Of these, five participants were excluded from the analysis due to extreme values in the outcome variables of interest at the baseline visit (T0) (Figure S2). Patients were followed for 3 months (T3) and six months (T6) respectively. Six participants were lost to follow-up (not willing to participate, n=6). Therefore, data from 32 participants at baseline and 26 at follow-up were available for analysis. The STROBE flow chart is depicted in Figure S1.

All participants signed informed consent prior to the enrolment. The research protocol was approved by the Ethics Committee of University Hospital Kralovske Vinohrady (EK-VP-14-0-2021) and the study was conducted under GCP following the Declaration of Helsinki.

Clinical Examination

Anthropometry and medical examination

All examinations were performed after overnight fasting. Each subject underwent a basic medical check-up with an anthropometric examination (height, weight, body mass index (BMI), waist circumference, and waist-to-hip ratio). Body composition was determined by bioimpedance analysis (BIA, Quadscan, Bodystat, UK) and expressed as skeletal muscle mass, active tissue mass (i.e. fat-free mass), and fat mass in kg and percentages respectively. Each participant filled in a questionnaire on COVID-19-related symptoms under the guidance of a medical person. Data on in-patient course (i.e. date of admission, ventilatory support, corticosteroids, etc.) were derived from the available medical records.

Blood sampling and laboratory analysis

A peripheral venous blood sample was drawn from an indwelling cannula in each subject after 12-hour fasting. All collected blood samples were centrifuged at the CRU and sera were stored at −80°C until transported to a certified institutional laboratory. Parameters of glucose homeostasis (fasting plasma glucose, glycated haemoglobin (HbA1c), C-peptide, and insulin), lipid profile (total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and triglycerides), and further routine laboratory parameters (urea, creatinine, albumin, CRP, lactate, cortisol, TSH, fT4, anti-TPO, blood count, D-dimers) were assessed in a certified hospital laboratory. Fasting plasma glucose was assessed using the hexokinase reaction (KONELAB, Dreieich, Germany); C-peptide by using solid-phase competitive chemiluminescent enzyme immunoassay (Immulite 2000, Los Angeles, CA, USA); HbA1c by using high-pressure liquid boronate affinity chromatography (Primus Corporation, Kansas City, MO, USA); insulin using solid-phase competitive chemiluminescent enzyme immunoassay (Immulite 2000, Los Angeles, CA, USA); total cholesterol and triglycerides using an enzymatic method kit (KONELAB, Dreieich, Germany); high-density lipoprotein-cholesterol (HDL-c) measured using a polyethylene glycol-modified enzymatic assay kit (ROCHE, Basel, Switzerland); and low-density lipoprotein–cholesterol (LDL-c) calculated using the standard Friedewald equation. Beta cell-specific autoAb (antiIA2, GADA) were analyzed using ELISA (Medipan GmbH, Germany). Serum NEFA and glycerol were measured using an enzymatic colorimetric kit (Randox Laboratories Ltd., UK) and serum branched-chain amino acids (BCAA) were measured using counter-current ELFO ²².

Insulin sensitivity, secretion indices, and hyperglycaemia

An oral glucose tolerance test (OGTT, 75g glucose) was performed after an overnight (12 hrs) fasting and following standard WHO recommendations. First, baseline blood samples were obtained from an indwelling cannula, then in 15, 30, 60, 90, and 120 min post-ingestion yielding 6 values for each subject. Insulin sensitivity and secretion were evaluated using basal values of serum glucose, insulin, and C-peptide (HOMA indices ²³) and data from the oral glucose tolerance test (OGGT). Incremental AUCs for glucose and insulin were calculated using the trapezoid rule. Insulin sensitivity alone was expressed as HOMA indices and Matsuda insulin sensitivity index (ISI) as published ²⁴. Insulin secretion was expressed as the insulinogenic index (IGI). IGI was computed using change in insulinaemia over glycaemia in 0 to 30 minutes: ΔINS 30-0/ΔGLU 30-0 (μU/mL*mg/dL). To specifically assess beta cell function, the oral disposition index (DI) was calculated to adjust for actual insulin sensitivity as IGI*ISI in each participant ²⁵.

Hyperglycaemia was defined as fasting glycaemia ≥5.6mM and/or 2 hours OGTT glycaemia ≥7.8mM. Diabetes was defined as fasting glycaemia ≥7.6mM and/or 2 hours OGTT glycaemia ≥11.1mM ²⁶.

Resting energy expenditure and substrate preference

Indirect calorimetry was measured at each subject: 1/ after an overnight (12 hrs) fasting and 30 min bed rest and 2/ in 100-120 min of OGTT, using a canopy ventilated hood system (QuarkRMR, Cosmed, Italy). Measurements were performed for 20 min after the initial ventilation stabilization, data were averaged per 30 sec and percent variance was recorded to confirm that subjects were in steady-state. Gas sensors were calibrated using a mixture of known concentrations of gases and ambient air, flowmeter was calibrated using a semiautomated pump. All calibrations were done before each respective measurement. Measured ambient air temperature, pressure, and humidity were recorded. All subjects were asked to collect urine for 24 hours before the measurement and urea concentration in the mixed sample was used to calculate nitrogen output. Measured VO2 and VCO2 and fat-free mass were used to calculate daily non-protein resting energy expenditure using the Weir formula (REE) and respiratory quotient (RQ). VO2, VCO2, and nitrogen loss per 24 hours were used to calculate basal substrate utilization of carbohydrates, fat, and protein ^27,28. Harris-Benedict equation with fat-free mass weight was used to estimate predicted REE ²⁹. Change in RQ from baseline to 120 min OGTT (ΔRQ 120-0) was calculated as a parameter of metabolic flexibility and change in REE (ΔREE 120-0) as a parameter of the diet-induced thermogenesis.

Statistics.

Normally distributed data are presented as mean and standard deviation (SD), skewed distributed data as the median and interquartile range (IQR), and categorical variables as numbers and percentages. Mean differences with 95% confidence intervals (CI) between the normal glycaemia and hyperglycaemia groups were calculated by using unpaired T-tests for normally distributed variables and median differences with 95% CI by using Wilcoxon signed rank test for skewed distributed variables. Group differences were adjusted for age and sex.

Time differences between T6 and T0 were investigated with the paired T-test for continuous variables, and time differences across all three visits as well as between groups were determined using generalised linear models (repeated measures ANOVA).

To investigate whether the exclusion of the extreme outliers from our main analysis might have influenced our results, we conducted a sensitivity analysis by including the entire study sample (n=37 at T0, n=31 at T6; compare: Figure S1). Therefore, all analyses were repeated on the entire sample of the study. Differences between groups or between baseline and follow-up visits were determined according to the precision of the 95% CI (null value not included) and the corresponding p-value (p<0.05). All analyses were performed in SAS (version 9.4; SAS Institute, Cary, USA).

Clinical characteristics

At baseline, 32 patients, survivors of severe COVID-19, were examined on average 21±6.5 days after COVID-19 diagnosis. Patients had respiratory failure demanding high-flow (n=19) or low-flow (n=18) nasal oxygen therapy. Patients were administered high-dose corticosteroids in majority of cases (dexamethasone: n=18; prednisone: n=9; methyl-prednisolone: n=1; no corticosteroids: n=4). Clinical characteristics are summarized in Table 1. Twenty-six patients finished 6 months of follow-up, and six patients were not willing to participate in the prospective follow-up from the time constraints (Figure S1). The patients dropping out were comparable in terms of baseline characteristics.

The presence and persistence of COVID-19-related symptoms are depicted in Table S5. Nearly all patients reported the presence of some post-COVID-19 related symptom, the most common were fatigue and muscle weakness. At T6 about one-third of the patients had at least one remaining symptom (mainly fatigue, joint pain, and hair loss). The prevalence of hyperglycaemia at baseline was 63% (20 patients). Nine patients had baseline HbA1c above 48mmol/mol. Hyperglycaemia remained as high as 50% (13 patients) after 6 months. Of these patients, 10 were classified as having prediabetes (treated by lifestyle intervention only) and 3 had been diagnosed with Type 2 diabetes and were on pharmacological treatment (metformin n=3, gliflozin n=1).

We found no patient with a positive titer of antibodies, neither in GADA (>5.5±0.5 IU/L) nor antiIA2 (>3.7±0.4 IU/L).

Insulin function measures

To get insight into glycaemic outcomes we compared patients with normal glycaemia vs hyperglycaemia at T0 and T6. The clinical characteristics of the groups are summarized in Table S2. Groups were comparable in a wide range of parameters, they differed in fasting and postprandial glycemia and HbA1c.

Considerably worse glycaemic response (Figure 1A) combined with the lower insulinogenic response, dominantly in the first 30min (Figure 1B) was observed in the hyperglycaemia group. While insulinogenic response declined and was comparable between groups at T6 (Figure 1B), glycaemic response remained worse for the hyperglycaemia group (Figure 1A). This translated to the calculated glycaemic indices as summarized in Table 3. The hyperglycaemic group had worse insulinogenic and disposition indices at baseline (IGI_H=1.04 ± 0.55, IGI_N=2.94 ± 1.82, p=0.001; DI_H=3.19 ± 2.16 vs DI_N=8.78 ± 6.37, p=0.005), the difference that was not apparent at T6. When comparing differences in changes in insulin sensitivity (both ISI Matsuda and HOMA indices) improved in the hyperglycaemic group only [ΔT6-T0 mean difference for ISI (95% CI): 1.84 (0.45, 3.24), p=0.01] (Table 3).

In the same settings, we analysed circulating NEFA and glycerol as whole-body indices of basal and insulin-stimulated lipolysis (Figure 2A and 2B). Both basal values and kinetics of the response upon OGTT were comparable among groups both at T0 and T6. On the other hand, both groups had comparably lower basal and stimulated values at T6.

Neither BCAA basal values nor the suppression of its levels in OGTT showed any difference among groups and T0-T6 (Figure 2C).

Weight balance, resting energy expenditure, and substrate oxidation

Patients were examined at baseline in the weight-losing phase. Weight loss over the acute disease was 6.5±4 kg (~7% of the pre-hospitalization weight) which was associated with reduced self-reported oral intake of 80±32% (n=32). Total body weight increased over the 6 months by 6.5±4.3 kg (p<0.0001, n=32).

For the REE analyses, the whole group was compared in terms of measured over predicted REE per kg fat-free mass at baseline, 3 and 6 months follow-up. Data are summarized in Table 2 and Figures 3 and 4. At baseline the REE was 30.7 ± 4.3 kcal per kg of active tissue mass, which was ~19% above the predicted values from the Harris-Benedict equation; REE declined to predicted values over 3 months (Figure 3). Mean basal RQ was 0.70± 0.07 at baseline and slightly increased to 0.74 ± 0.05 (p=0.003) at 6 months (Table 2). We observed no differences over time in either ΔRQ_120-0or ΔREE_120-0 (Table 2). The oxidation of the individual substrates is shown in Figure 4. The predominant substrate oxidized under fasting conditions was fat. However, fat oxidation slightly decreased from T0 to T3 from ~76% to ~66% (p=0.04), and protein oxidation increased from ~17% to ~22% over the 6 months (ANOVA: p=0.02).

Sensitivity analyses

Sensitivity analyses were conducted on the entire study population including the five participants with extreme outliers (n=37). There were no noticeable differences in the characteristics when comparing the total study population with the study population excluding extreme outliers (Table S1). The findings on the clinical metabolic variables (Table S3) and macronutrient oxidation (Figure S3) were also comparable. When comparing the findings on insulin secretion indices, only slight differences were observed between the main and sensitivity analysis. While in the main analysis a clear reduction in IGI from baseline to six months follow-up was observed [-1.34 (95% CI: -2.48, -0.19); p=0.03] in the normal glycaemic group (Table 3), the reduction was not precisely estimated in the sensitivity analysis [-0.85 (95% CI: -2.59, 4.75); p=0.41] (Table S4).

The main findings of the current study are that 1/ highly prevalent hyperglycaemia is caused by an insufficient increase in insulin secretion to compensate for prevailing IR, 2/ hyperglycaemia persists as classified as prediabetes or diabetes in a large portion of patients despite improvement in their IS and 3/ REE is high during the acute phase and reverts to predicted levels already after 3 months follow-up.

Hypermetabolism is present in the acute phase but does not persist over time

We observed approximately 20% higher than predicted energy expenditure across the whole sample in the acute phase. These findings are in line with the results of other studies showing hypermetabolism in the acute COVID-19 phase ^30–33 For instance, Yu et al. found that COVID-19 patients have energy expenditure as high as 200% of the predicted one ³³. This increased energy expenditure in COVID-19 patients was even higher than the energy expenditure observed in patients with burn wounds or acute respiratory distress syndrome, which are both conditions associated with high REE and increased nutritional needs ³⁴. It was suggested that hypermetabolism persists in post-COVID-19 patients³⁵. Contrary, we found that already after 3 months of follow-up, the REE returned to predicted values. Unexpectedly, we observed slightly increased protein oxidation after 6 months. As we do not have dietary record data, we cannot conclude whether this reflects endogenous protein catabolism or anabolic resistance to a relatively high protein refeeding in these patients.

Mechanisms of increased REE in severe COVID-19 patients are not well understood, but in general, immune cell activation, medication, and increased high body temperature can increase the energy needs in critically ill patients ³⁶. High-dose corticosteroids are the main treatment for COVID-19 pneumonia, and acute administration of cortisol has been shown to increase energy expenditure ³⁷. Of note, measurements of REE in oxygen-dependent patients with highly transmissible diseases are challenging and make it difficult to obtain comparable prospective data because breath-by-breath measurements during artificial ventilation may yield different results than when the canopy is used. We were able to measure REE using ventilated canopy system as soon as patients were weaned from oxygen supply, so we did not capture the most critical phase of the disease; moreover, we focused only on survivors, which may explain the relatively smaller difference between measured and predicted REE than in previous reports.

Hyperglycaemia is a common feature in acute severe COVID-19

The prevalence of hyperglycaemia in the current study was as high as 63% (n=20) in a group of patients without previously diagnosed diabetes. In a meta-analysis, the prevalence of hyperglycaemia on acute COVID-19 was 14.4% in unselected patients population ³⁸ and as high as 85% in patients with critical illness ⁹. However, it is difficult to conclude from these epidemiological data whether the newly diagnosed diabetes is truly new-onset diabetes or a manifestation of diabetes that has been undiagnosed in the past. We found six patients in the hyperglycaemic group with HbA1c levels above 48 mmol/mol at baseline, and we cannot rule out that these patients have had diabetes for a long time that had not been diagnosed in the past. Moreover, we did not find a single patient with islet autoantibodies, making it unlikely that we enrolled a patient with new-onset type 1 diabetes in the study.

Hyperglycaemia relates to both insulin resistance and beta cell failure

We compared two groups of patients with normal glycemia and hyperglycaemia to explore the contribution of insulin resistance and beta cell failure to the manifestation of hyperglycaemia. The main difference we observed in the acute phase of COVID-19 was relatively lower insulin secretion in 30 and 60 minutes after glucose loading, suggesting a primary defect in the first phase of the insulin response. This also resulted in lower insulinogenic and disposition indices.

These findings are in contradiction with the current literature ^8,9. The findings from Reiterer et al. ⁹ showed that insulin resistance is the main cause of hyperglycaemia in critically ill patients with COVID-19. Here, the discrepancies could be explained by the fact that in the study only the ratio of C-peptide to glucose levels was used to classify the differences between insulin sensitivity and secretion, but the dynamics of insulin secretion under stimulated conditions were not measured. Moreover, the authors admit they had to rely on non-fasted samples which cannot preclude that the values may reflect an insulin-stimulated state. Another study from Montefusco et al. ⁸ pointed toward maintaining insulin secretion when assessed by an arginine stimulation test. Here, the discrepancy could be caused by different source populations (severe vs. non-severe COVID-19) and different designs, as we compared hyperglycaemic vs normoglycemic subjects, whereas the study ⁸ patients with acute and post-acute COVID19 that were examined were not stratified. Actually, in their sample of 10 vs. 10 subjects, there is a wide distribution of the response.

Concerning other potential mechanisms, we investigated treatment with corticosteroids during acute COVID-19. High-dose corticosteroids were used in all but four patients in the hyperglycaemic group. The effects of administered corticosteroids were measured by endogenous cortisol suppression which was comparable between groups. Corticosteroids were shown to affect IR ³⁹, even in a dose-response manner and the administration was shown to precipitate diabetes development ⁴⁰. On the other hand, short-term corticosteroid treatment probably does not influence insulin secretion ⁴¹ which determined hyperglycaemia in the current study. Therefore, we conclude that corticosteroids alone would not explain incident hyperglycaemia but they may have contributed to an overall decline in IS at T0. Of note, persistent suppression of endogenous cortisol was still present after 6 months, and therefore, it is likely that some of the post-covid symptoms overlap with adrenal insufficiency.

Adipose tissue insulin resistance does not contribute to the systemic IR

Above, we focused on glycaemic regulation that primarily reflects IR and insulin response of the liver and skeletal muscle, but the adipose tissue has also been shown to be an important organ contributing to systemic IR in COVID-19 patients ⁹. Among putative mechanisms, SARS-CoV-2 virus replication in adipocytes, leading to AT inflammation, and lower adiponectin and adiponectin/leptin ratios were suggested ^7,9. To address the contribution of AT we compared peripheral NEFA and glycerol suppressibility during OGTT. We found that both NEFA and glycerol were increased at T0 vs. T6 across all time points in OGTT but similarly in both groups. To conclude, indices of increased basal lipolysis do not relate to the presence/absence of hyperglycaemia. We also analysed changes in BCAA levels. BCAA were shown to be among important precursors for de novo lipogenesis and their circulating levels may reflect resistance to insulin function on protein turnover ⁴². We found no difference among groups and no change over the 6-month follow-up indicating that protein turnover remains unchanged.

Hyperglycaemia persists over the follow-up and there is a high prevalence of pre/diabetes in the post-COVID phase

After 6 months, insulin sensitivity improved in both groups, but more markedly in the hyperglycaemic group, whereas beta cell function (measured as disposition index) remained unchanged. Therefore, at 6 months follow-up, 13 hyperglycaemic patients remained in the group (50%). Three of these patients were classified as persistent Type 2 diabetes and pharmacological treatment was prescribed and 10 patients were classified as prediabetes and received lifestyle intervention only.

Strengths and limitations

Despite the strengths of the current study, namely the prospective 6-months follow-up and detailed metabolic phenotyping in survivors of severe COVID-19, further important limitations need to be discussed. We have encountered an important drop-out to follow-up (n=6; 18.8%) that limited the statistical analysis such as stratified analysis by sex or age groups. We included only COVID-19 survivors, within 4 weeks after COVID-19 diagnosis, so we may have missed the period of critical illness when signs of deterioration of insulin function and hypermetabolism may be more pronounced and prevalent. This limitation stemmed from feasibility reasons as performing OGTT and indirect calorimetry in unstable patients is both challenging and less ethical, and we also wanted to eliminate the mortality drop-out rate at baseline. The design of the study, i.e. without an adequate comparator group (i.e. other severely ill for the T0 and healthy for T6), makes it difficult to infer conclusions on causality between COVID-19 and hyperglycaemia and is prone to residual confounding. During the recruitment period, there were only a few non-COVID-19 critically ill patients and therefore the comparator was not feasible. As mentioned above, we cannot conclude the separation between new-onset and newly diagnosed diabetes. And lastly, six months follow-up period is relatively short to see further clinical evolution of pre/diabetes.

To conclude, hyperglycaemia is prevalent in patients with severe COVID-19 and an inadequate beta cell response to match for insulin resistance in the acute disease may be an explanation. Improvement in insulin resistance leads to improvement in glucose tolerance over the six months. Hypermetabolism is likely associated with the critical state of severe COVID-19 and its treatment by corticosteroids and normalizes relatively quickly during recovery. Despite all the uncertainties that remain about COVID-19 and diabetes, it is clear that these two pandemics are intertwined and closely interact. Diabetes is one of the major risk factors for severe COVID-19 disease, and on the other hand, COVID-19 induces diabetes in patients with impaired beta cell function. Therefore, identification of the underlying pathophysiology of COVID-19-related diabetes would enable the implementation of effective strategies to manage both diseases.

Conflict of interest

The authors have no conflict of interest to disclose.

Acknowledgments

Funding. Charles University Cooperatio METD, SVV 260531/SVV/2020, GAČR 22-22398S, Programme EXCELES, ID Project No. LX22NPO5104 - Funded by the European Union – Next Generation EU and EFSD mentorship programme supported by AstraZeneca

Contribution statement. JG, MA, SS were involved in the study conceptualization and design. KK, AO, AG, VŠ, JP and JG participated in the practical performance of data acquisition. AL, SS, MA, FK, LS, PT, JG participated in data analysis and interpretation. AL, SS, LR, and JG were involved in statistical analysis. KK and JG prepared the manuscript. All authors participated in critical revision of the manuscript and approved the final version of the manuscript. The corresponding author (JG) takes full responsibility for the work and/or the conduct of the study, he had access to the data and controlled the decision to publish.

Data availability. Pseudonymized datasets generated and analysed during the current study are available from the corresponding author upon reasonable request.

Huang C, Wang Y, Li X, Ren L, Zhao J, Hu Y et al. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. The Lancet 2020; 395: 497–506.
Guo W, Li M, Dong Y, Zhou H, Zhang Z, Tian C et al. Diabetes is a risk factor for the progression and prognosis of COVID-19. Diabetes Metab Res Rev 2020; 36. doi:10.1002/dmrr.3319.
Chen M, Zhu B, Chen D, Hu X, Xu X, Shen WJ et al. COVID-19 May Increase the Risk of Insulin Resistance in Adult Patients Without Diabetes: A 6-Month Prospective Study. Endocrine Practice 2021; 27: 834–841.
Khunti K, Prato S del, Mathieu C, Kahn SE, Gabbay RA, Buse JB. Covid-19, hyperglycemia, and new-onset diabetes. Diabetes Care 2021; 44: 2645–2655.
Rubino F, Amiel SA, Zimmet P, Alberti G, Bornstein S, Eckel RH et al. New-Onset Diabetes in Covid-19. New England Journal of Medicine 2020; 383: 789–790.
Schlesinger S, Neuenschwander M, Lang A, Pafili K, Kuss O, Herder C et al. Risk phenotypes of diabetes and association with COVID-19 severity and death: a living systematic review and meta-analysis. Diabetologia. 2021; 64: 1480–1491.
Langouche L, van den Berghe G, Gunst J. Hyperglycemia and insulin resistance in COVID-19 versus non-COVID critical illness: Are they really different? Crit Care. 2021; 25: 437.
Montefusco L, ben Nasr M, D’Addio F, Loretelli C, Rossi A, Pastore I et al. Acute and long-term disruption of glycometabolic control after SARS-CoV-2 infection. Nat Metab 2021; 3: 774–785.
Reiterer M, Rajan M, Gómez-Banoy N, Lau JD, Gomez-Escobar LG, Ma L et al. Hyperglycemia in acute COVID-19 is characterized by insulin resistance and adipose tissue infectivity by SARS-CoV-2. Cell Metab 2021; 33: 2174-2188.e5.
Wan L, Gao Q, Deng Y, Ke Y, Ma E, Yang H et al. GP73 is a glucogenic hormone contributing to SARS-CoV-2-induced hyperglycemia. Nat Metab 2022; 4: 29–43.
Cartin-Ceba R, Khatua B, El-Kurdi B, Trivedi S, Kostenko S, Imam Z et al. Evidence showing lipotoxicity worsens outcomes in covid-19 patients and insights about the underlying mechanisms. iScience 2022; 25: 104322.
Soto ME, Guarner-Lans V, Díaz-Díaz E, Manzano-Pech L, Palacios-Chavarría A, Valdez-Vázquez RR et al. Hyperglycemia and Loss of Redox Homeostasis in COVID-19 Patients. Cells 2022; 11. doi:10.3390/cells11060932.
Hayden MR. An Immediate and Long-Term Complication of COVID-19 May Be Type 2 Diabetes Mellitus: The Central Role of β-Cell Dysfunction, Apoptosis and Exploration of Possible Mechanisms. Cells. 2020; 9. doi:10.3390/cells9112475.
Müller JA, Groß R, Conzelmann C, Krüger J, Merle U, Steinhart J et al. SARS-CoV-2 infects and replicates in cells of the human endocrine and exocrine pancreas. Nat Metab 2021; 3: 149–165.
Kazakou P, Lambadiari V, Ikonomidis I, Kountouri A, Panagopoulos G, Athanasopoulos S et al. Diabetes and COVID-19; A Bidirectional Interplay. Front Endocrinol (Lausanne). 2022; 13. doi:10.3389/fendo.2022.780663.
Laurenzi A, Caretto A, Molinari C, Mercalli A, Melzi R, Nano R et al. No Evidence of Long-Term Disruption of Glycometabolic Control After SARS-CoV-2 Infection. J Clin Endocrinol Metab 2022; 107: e1009–e1019.
Liu Y, Lu R, Wang J, Cheng Q, Zhang R, Zhang S et al. Diabetes, even newly defined by HbA1c testing, is associated with an increased risk of in-hospital death in adults with COVID-19. BMC Endocr Disord 2021; 21. doi:10.1186/S12902-021-00717-6.
Lampasona V, Secchi M, Scavini M, Bazzigaluppi E, Brigatti C, Marzinotto I et al. Antibody response to multiple antigens of SARS-CoV-2 in patients with diabetes: an observational cohort study. Diabetologia 2020; 63: 2548–2558.
Cariou B, Pichelin M, Goronflot T, Gonfroy C, Marre M, Raffaitin-Cardin C et al. Phenotypic characteristics and prognosis of newly diagnosed diabetes in hospitalized patients with COVID-19: Results from the CORONADO study. Diabetes Res Clin Pract 2021; 175. doi:10.1016/J.DIABRES.2021.108695.
Ryrsø CK, Dungu AM, Hegelund MH, Jensen AV, Sejdic A, Faurholt-Jepsen D et al. Comparison of body composition, physical capacity, and immuno-metabolic profile of hospitalized patients with community-acquired pneumonia caused COVID-19, influenza, and bacteria. In: European Respiratory Journal. European Respiratory Society (ERS), 2021, p PA3266.
Plummer MP, Rait L, Finnis ME, French CJ, Bates CCRN S, Douglas J et al. Diabetes mellitus, glycaemic control and severe COVID-19 in the Australian Critical Care Setting – a nested cohort study. Australian Critical Care 2022. doi:10.1016/j.aucc.2022.05.002.
Tůma P, Gojda J. Rapid determination of branched chain amino acids in human blood plasma by pressure-assisted capillary electrophoresis with contactless conductivity detection. Electrophoresis 2015; 36: 1969–1975.
Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985; 28: 412–419.
Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care 1999; 22: 1462–1470.
Kile AJ, Hanna C, Hannon TS, Kirkman MS, Considine R v., Patel Y et al. The linearized disposition index augments understanding of treatment effects in diabetes. Am J Physiol Endocrinol Metab 2021; 320: E169–E177.
Association AD. Classification and diagnosis of diabetes: Standards of Medical Care in Diabetes-2020. Diabetes Care 2020; 43: S14–S31.
Frayn KN. Calculation of substrate oxidation rates in vivo from gaseous exchange. J Appl Physiol Respir Environ Exerc Physiol 1983; 55: 628–634.
Ferrannini E. The theoretical bases of indirect calorimetry: a review. Metabolism 1988; 37: 287–301.
Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO. A new predictive equation for resting energy expenditure in healthy individuals. Am J Clin Nutr 1990; 51: 241–247.
Niederer LE, Miller H, Haines KL, Molinger J, Whittle J, MacLeod DB et al. Prolonged progressive hypermetabolism during COVID-19 hospitalization undetected by common predictive energy equations. Clin Nutr ESPEN 2021; 45: 341–350.
Lakenman PLM, van der Hoven B, Schuijs JM, Eveleens RD, van Bommel J, Olieman JF et al. Energy expenditure and feeding practices and tolerance during the acute and late phase of critically ill COVID-19 patients. Clin Nutr ESPEN 2021; 43: 383–389.
von Renesse J, von Bonin S, Held HC, Schneider R, Seifert AM, Seifert L et al. Energy requirements of long-term ventilated COVID-19 patients with resolved SARS-CoV-2 infection. Clin Nutr ESPEN 2021; 44: 211–217.
Yu PJ, Cassiere H, DeRosa S, Bocchieri K, Yar S, Hartman A. Hypermetabolism and Coronavirus Disease 2019. Journal of Parenteral and Enteral Nutrition 2020; 44: 1234–1236.
Martindale R, Patel JJ, Taylor B, Arabi YM, Warren M, McClave SA. Nutrition Therapy in Critically Ill Patients With Coronavirus Disease 2019. JPEN J Parenter Enteral Nutr 2020; 44: 1174–1184.
Whittle J, Molinger J, MacLeod D, Haines K, Wischmeyer PE. Persistent hypermetabolism and longitudinal energy expenditure in critically ill patients with COVID-19. Crit Care 2020; 24: 581.
McClave SA, Taylor BE, Martindale RG, Warren MM, Johnson DR, Braunschweig C et al. Guidelines for the Provision and Assessment of Nutrition Support Therapy in the Adult Critically Ill Patient. Journal of Parenteral and Enteral Nutrition 2016; 40: 159–211.
Brillon DJ, Zheng B, Campbell RG, Matthews DE. Effect of cortisol on energy expenditure and amino acid metabolism in humans. Am J Physiol Endocrinol Metab 1995; 268. doi:10.1152/ajpendo.1995.268.3.e501.
Sathish T, Kapoor N, Cao Y, Tapp RJ, Zimmet P. Proportion of newly diagnosed diabetes in COVID-19 patients: A systematic review and meta-analysis. Diabetes Obes Metab. 2021; 23: 870–874.
Qi D, Rodrigues B. Glucocorticoids produce whole body insulin resistance with changes in cardiac metabolism. Am J Physiol Endocrinol Metab. 2007; 292. doi:10.1152/ajpendo.00453.2006.
Wu J, MacKie SL, Pujades-Rodriguez M. Glucocorticoid dose-dependent risk of type 2 diabetes in six immune-mediated inflammatory diseases: a population-based cohort analysis. BMJ Open Diabetes Res Care 2020; 8. doi:10.1136/BMJDRC-2020-001220.
Fine NHF, Doig CL, Elhassan YS, Vierra NC, Marchetti P, Bugliani M et al. Glucocorticoids reprogram b-cell signaling to preserve insulin secretion. In: Diabetes. American Diabetes Association Inc., 2018, pp 278–290.
Vanweert F, Schrauwen P, Phielix E. Role of branched-chain amino acid metabolism in the pathogenesis of obesity and type 2 diabetes-related metabolic disturbances BCAA metabolism in type 2 diabetes. Nutr Diabetes 2022; 12: 35.

Table 1: Clinical characteristics of the study population (n=32)

	T0 Baseline	T3 Visit after 3 months	T6 Visit after 6 months	p
n	32	28	26	-
Age [years]	57.2 ± 10.6	57.6 ± 10.8	59.0 ± 9.9	-
Sex [n female, %]	12 (38%)	11 (39%)	9 (35%)	-
Prevalence hyperglycemia [n, %]	20 (63%)	n.a.	13 (50%)	0.03
BMI [kg/m²]	30.8 ± 5.2	31.8 ± 4.7	31.6 ± 4.6	<0.0001
Body weight [kg]	90.4 ± 17.6	93.8 ± 15.7	94.7 ± 15.3	<0.0001
WHR	0.94 ± 0.08	0.93 ± 0.08	0.93 ± 0.08	0.11
Fat mass [%]	25 ± 10	24 ± 10	26 ± 13	0.30
Systolic blood pressure [mmHg]	131 ± 16	132 ± 16	135 ± 18	0.37
Diastolic blood pressure [mmHg]	84 ± 6	84 ± 9	84 ± 9	0.98
First random glucose [mmol/L]	6.5 (6.0, 7.5)	-	-	-
Fasting glucose [mmol/L]	5.3 ± 1.1	5.7 ± 0.5	5.4 ± 0.6	0.83
Glucose after 2 hours [mmol/L]	8.3 ± 2.5	n.a.	7.6 ± 1.8	<0.001
Fasting insulin [mIU/L]	11.8 (9.1, 16.7)	14.4 (11.4, 17.4)	7.7 (6.2, 13.4)	0.01
Insulin after 2 hours [mIU/L]	86.4 (55.0, 111.1)	n.a.	48.2 (36.7, 121.9)	0.06
Fasting C-peptide [pmol/L]	923 ± 324	713 ± 257	558 ± 215	<0.0001
C-peptide after 2 hours [pmol/L]	3523 ± 932	n.a.	2312 ± 716	<0.0001
Fasting NEFA [mmol/L]	0.97 ± 0.24	n.a.	0.77 ± 0.27	0.001
NEFA after 2 hours [mmol/L]	0.56 ± 0.20	n.a.	0.34 ± 0.14	<0.0001
Fasting glycerol [µmol/L]	254 ± 72	n.a.	156 ± 47	<0.0001
Glycerol after 2 hours [µmol/L]	212 ± 58	n.a.	121 ± 25	<0.0001
Insulinogenic index	1.23 (0.84, 2.19)	n.a.	1.49 (0.81, 1.53)	0.29
Insulin sensitivity index	3.12 ± 1.36	n.a.	4.53 ± 2.30	0.003
Disposition Index	5.12 ± 4.84	n.a.	4.89 ± 2.58	0.86
HOMA-beta	1.49 (1.13, 2.25)	n.a.	1.01 (0.81, 1.53)	0.01
HOMA-IR	2.49 (1.77, 4.79)	n.a.	1.77 (1.48, 3.42)	0.01
HbA1c [mmol/mol]	46.6 ± 12.4	35.5 ± 4.1	37.8 ± 4.2	<0.001
Triglycerides [mmol/L]	1.73 ± 0.84	1.70 ± 1.25	1.50 ± 0.83	0.08
Total cholesterol [mmol/L]	4.27 ± 0.88	5.11 ± 1.14	4.77 ± 0.92	0.11
HDL cholesterol [mmol/L]	1.28 ± 0.32	1.19 ± 0.25	1.19 ± 0.33	0.16
LDL cholesterol [mmol/L]	2.19 ± 0.77	3.24 ± 1.03	2.91 ± 0.90	0.01
ALT [μkat/L]	1.07 ± 0.55	0.49 ± 0.18	0.46 ± 0.19	<0.0001
AST [μkat/L]	0.46 ± 0.19	0.38 ± 0.10	0.36 ± 0.11	0.001
Urea [mmol/L]	5.62 ± 2.71	5.46 ± 1.46	5.76 ± 1.42	0.73
Creatinine [μmol/L]	67.1 ± 15.1	73.6 ± 15.6	75.8 ± 15.6	<0.001
Albumine [g/L]	40.5 ± 1.9	47.1 ± 5.7	43.0 ± 2.6	<0.0001
Cortisol [nmol/L]	238 ± 181	n.a.	249 ± 99	0.81
TSH [µmol/L]	1.85 (1.42, 2.78)	n.a.	1.51 (1.23, 1.85)	<0.001
Triiodothyronine (T3) [pmol/L]	5.23 ± 0.76	n.a.	5.74 ± 0.52	0.01
Thyroxine (T4) [pmol/L]	15.8 ± 3.6	n.a.	14.5 ± 1.9	0.03
White blood cells [x10⁹/L]	9.83 ± 3.95	6.67 ± 1.60	6.08 ± 1.67	<0.001
Platelets [x10⁹/L]	239 ± 104	252 ± 62	211 ± 57	0.26
D-dimer [μg/L]	550 (295, 1585)	520 (300, 860)	250 (210, 530)	<0.001
Normally distributed variables were expressed as mean ± standard deviation and skewed distributed variables as median and interquartile range P-values for differences between T0 and T6 were calculated with paired T-test for normal distributed variables, with Wilcoxon signed rank test for skewed distributed variables and via Chi-square test for categorical variables

Table 2: Clinical metabolic variables at baseline (T0) and follow-up visit after 3 (T3) and 6 months (T6) and the mean difference plus 95% CI between T0 and T6 in the study population

	T0		T3		T6		ΔT6-T0
	n	mean ± SD	n	mean ± SD	n	mean ± SD	n	mean difference (95% CI)	p
REE_{Harris Benedict}[kcal/d]	32	1712 ± 299	28	1751 ± 279	26	1767 ± 276	26	49 (31, 67)	<0.0001
REE₀[kcal/d]	32	2033 ± 352	28	1800 ± 296	26	1759 ± 285	26	-293 (-378, -207)	<0.0001
REE₁₂₀[kcal/d]	31	2128 ± 306	n.a.		26	1924 ± 277	25	-236 (-338, -134)	<0.0001
ΔREE_120-0[kcal/d]	31	106 ± 198	n.a.		26	165 ± 116	25	54 (-30, 139)	0.20
REE₀/REE_{Harris Benedict}[%]	32	119 ± 14	28	103 ± 9	26	100 ± 9	26	-20 (-25, -15)	<0.0001
REE₀/ATM ratio [kcal/kg]	32	30.7 ± 4.3	28	25.6 ± 2.4	26	24.8 ± 2.6	26	-5.4 (-6.8, -4.1)	<0.0001
RQ₀	32	0.70 ± 0.07	28	0.75 ± 0.06	26	0.74 ± 0.05	26	0.04 (0.02, 0.07)	0.003
RQ₁₂₀	31	0.80 ± 0.08	n.a.		26	0.82 ± 0.05	25	0.02 (-0.02, 0.06)	0.35
ΔRQ_120-0	31	0.10 ± 0.08	n.a.		26	0.07 ± 0.05	25	-0.03 (-0.06, 0.01)	0.12

Data are shown as mean ± SD. Differences between T6 and T0 are shown as mean difference ± 95% CI. P-values were derived from paired T-test.

ATM, active tissue mass; CI, confidence interval; n.a., not applicable; REE, resting energy expenditure; RQ, respiratory quotient; SD, standard deviation; T0, baseline visit; T3, visit after three months; T6, visit after six month

Table 3: Outcome indices at baseline (T0) and follow-up visit after 6 months (T6) and the mean differences plus 95% CI between these visits in the normal glycemic and hyperglycemic groups of the study population

	Normal glycemia	Hyperglycemia		Adjusted group differences
	mean ± SD	mean ± SD		mean difference (95% CI)	p
	T0 (n=32)
Insulinogenic index	2.94 ± 1.82	1.04 ± 0.55		-1.74 (-2.67, -0.81)	0.001
Insulin sensitivity index	3.23 ± 1.72	3.07 ± 1.18		-0.26 (-1.46, 0.94)	0.66
Disposition Index	8.78 ± 6.37	3.19 ± 2.16		-5.32 (-8.88, -1.76)	0.005
HOMA-beta	1.71 ± 1.09	1.91 ± 0.88		0.32 (-0.45, 1.08)	0.40
HOMA-IR	2.84 ± 1.98	3.82 ± 2.32		1.18 (-0.58, 2.94)	0.18
	T6 (n=26)
Insulinogenic index	1.35 ± 0.64	1.57 ± 2.28		0.52 (-1.25, 2.30)	0.55
Insulin sensitivity index	4.56 ± 1.27	4.52 ± 2.75		-0.37 (-2.59, 1.85)	0.73
Disposition Index	5.67 ± 1.98	4.45 ± 2.83		-0.91 (-3.32, 1.50)	0.44
HOMA-beta	1.20 ± 0.41	1.60 ± 1.24		0.52 (-0.49, 1.53)	0.30
HOMA-IR	2.04 ± 0.62	3.20 ± 2.83		1.38 (-0.89, 3.65)	0.22
	ΔT6-T0 (n=26)
	Normal glycemia			Hyperglycemia
	mean difference (95% CI)		p	mean difference (95% CI)	p
Insulinogenic index	-1.34 (-2.48, -0.19)		0.03	0.54 (-0.72, 1.79)	0.38
Insulin sensitivity index	1.02 (-0.17, 2.22)		0.08	1.84 (0.45, 3.24)	0.01
Disposition Index	-2.39 (-6.73, 1.96)		0.23	1.05 (-0.33, 2.44)	0.12
HOMA-beta	-0.43 (-1.12, 0.27)		0.20	-0.48 (-0.90, -0.05)	0.03
HOMA-IR	-0.65 (-1.88, 0.59)		0.26	-0.99 (-1.90, -0.09)	0.03

Data are shown as mean ± SD. Differences between the normal glycaemic and hyperglycaemic group are shown as age- and sex-adjusted mean difference ± 95% CI. P-values were derived from unpaired T-test.

Differences over time between T6 and T0 were derived from paired T-test

CI, confidence interval; HOMA-beta, homeostasis model assessment of beta cell function; HOMA-IR, homeostasis model assessment of insulin resistance; SD, standard deviation; T0, baseline visit; T6, visit after six month; ΔT6-T0, change between T6 and T0;

There is NO conflict of interest to disclose

Severe COVID-19 associated hyperglycaemia is caused by beta cell dysfunction: a prospective cohort study

Status:

Journal Publication

Version 1

Abstract

Figures

Background

Methods

Results

Discussion

Conclusion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1