Preserved C-peptide in survivors of COVID-19: Post hoc analysis

methods).


| INTRODUCTION
Hyperglycaemia 1 and new diagnoses of diabetes (including autoantibody-negative type 1 diabetes 2 ) have been reported in patients with coronavirus disease 2019 (COVID- 19). 1 Moreover, 77% of patients with COVID-19 and ketoacidosis (a condition typically associated with type 1 diabetes or pancreatic destruction) had type 2 diabetes. 3 The receptor necessary for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cellular entry, ACE2, and its permissive protein, TMPRSS2, are present in the pancreatic microvasculature 4 and, at postmortem, SARS-CoV-2 viral RNA has been detected in pancreatic beta cells. 5 In islets infected with SARS-CoV-2, in vitro insulin secretion, particularly glucose-stimulated insulin secretion, is attenuated. 6 Given this, it has been speculated that insulin secretory capacity may be curtailed by direct cytopathic action of SARS-CoV-2 on beta cells. However, it is unknown whether such changes are of clinical relevance in survivors of COVID-19.
Random non-fasting C-peptide (rCP) levels correlate with stimulated C-peptide levels derived during mixed meal tests, 7 and have high sensitivity and specificity compared with the gold standard threshold of 600 pmol/L considered to indicate insulin deficiency. 8 It remains to be determined whether there is evidence for deficient insulin secretion, consistent with beta cell destruction, in survivors of COVID-19.
Therefore, we assessed rCP in patients at least 3 months after COVID-19 infection to assess longer term beta cell secretory capacity. The study protocol has previously been described. 9 Participants were survivors of COVID-19 aged 18 years or older diagnosed with COVID-19 who attended Imperial College London NHS Healthcare Trust during March-November 2020, or who had responded to adverts placed on social media asking for patients who had tested positive for COVID-19 and then attended at least 3 months following diagnosis with COVID-19. Diagnosis of COVID-19 was confirmed using either real-time RT-PCR testing of a nasopharyngeal swab, confirmatory imaging (chest radiograph or CT scan), or a positive serum SARS-CoV-2 IgG antibody test after symptom onset. Exclusion criteria included those prescribed steroids following recovery from COVID-19, and those taking other medications known to affect cortisolbinding globulin (including the combined oral contraceptive pill, and hormone replacement therapy). Similarly, patients with underlying health conditions or states known to influence cortisol-binding globulin (including pregnancy, end-stage renal failure, or underlying malignancy) were also excluded.

Severity of COVID-19 was determined according to World
Health Organization (WHO) classification. 10 Serum glucose and insulin were measured using an Abbott Alinity ci-series analyser. The lower limits of detection were as follows: insulin 2.9 pmol/L, glucose 0.03 mmol/L. The inter-assay coefficients of variation were: insulin 2.1%, glucose 1%. The intra-assay coefficients of variation were: insulin 2.2%, glucose 1.2%. Serum C-peptide was measured using an Abbott Architect analyser. The lower limit of detection was 3.3 pmol/L, the inter-assay coefficient of variation was 4.0%, and the intra-assay coefficient of variation was 2.8%. Serum fructosamine was measured using the Roche Cobas kit on the Abbott Architect platform (colorimetric method); the intra-assay coefficient of variation was less than 0.5% at 299 and 256 μmol/L. The reference range was 205-500 μmol/L. Serum antibodies to SARS-CoV-2 N protein (IgG) were measured using Abbott Architect assay. For those with an indeterminate result, additional testing for antibodies to SARS-CoV-2 spike protein (RBD) (IgG) was performed using Imperial Hybrid DABA. 11 Study visits commenced 8:00 AM-9:30 AM and were non-fasted.
Serum samples were taken for measurement of insulin, glucose, and C-peptide levels. Fifty-four of 55 patients (98%) had complete insulin, C-peptide, and glucose measurements (one participant had missing data for C-peptide levels and was excluded from further analysis).
Statistical analysis was performed using GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA). Distribution of data was assessed by D'Agostino and Pearson normality tests. Data were presented as mean ± standard deviation (SD) if parametrically distributed, or median with interquartile range (IQR) if not. Student's t-test, or one-way analysis of variation (ANOVA), was used to compare two or more groups respectively, of parametrically distributed data. Mann-Whitney U test, or Kruskal-Wallis test with post hoc Dunn's test, was used to compare two or more groups respectively, of nonparametrically distributed data. Multivariable linear regression was used to assess the ability of selected variables to predict C-peptide, including age, sex, ethnicity, pre-existing diabetes, body mass index (BMI) and severity of COVID-19 disease, according to the WHO severity index.

| Baseline characteristics
In total, 55 participants attended for assessment at a median of 233 days post-COVID-19 infection. Of these, 15 (27.8%) had pre-existing type 2 diabetes; none had type 1 diabetes. No participants received a new diagnosis of type 2 diabetes following COVID-19 infection (Table 1). All participants with pre-existing diabetes were on metformin but none were treated with insulin.
3.2 | C-peptide is preserved at follow-up at ≥3 months following presentation with COVID-19 At the study visit appointment, which took place at least 3 months after initial presentation with COVID-19, median (IQR) C-peptide was 1319 (849, 1905) pmol/L and no patients had undetectable C-peptide levels (<3.3 pmol/L). Of the study cohort, 14.8% (n = 8) had C-peptide levels of less than 600 pmol/L with normoglycaemic contemporaneous glucose concentrations of 4.7-5.7 mmol/L, none of whom had known diabetes ( Figure 1A).

| DISCUSSION
Hyperglycaemia in patients with COVID-19 is associated with increased mortality rates, 12 and results from both impaired insulin secretion and increased peripheral insulin resistance. 13 However, the extent to which these causes of dysglycaemia persist beyond acute infection is unclear.
Both critical 14 and viral illnesses increase peripheral insulin resistance, particularly in people with pre-existing metabolic disorders. 6 Furthermore, the viruses themselves may contribute to beta cell destruction and dysfunction, 15 with cytomegalovirus, 16 Epstein-Barr virus, 17 rotavirus, 18 rubella virus, and enterioviruses (including coxsackievirus B) all being associated with pancreatic beta cell apoptosis and subsequent type 1 diabetes. 15 Similarly, SARS-CoV-2 induces pancreatic islet cell apoptosis, 19 and thus has the potential to result in beta cell destruction and resultant insulinopaenia. These data assess the clinical relevance of such findings.
In this analysis of 54 survivors of COVID-19, all participants (including those with pre-existing type 2 diabetes) had detectable Cpeptide levels. Although C-peptide levels were not associated with peak procalcitonin or CRP levels, C-peptide values were greater in participants with severe or critical disease compared with those with mild disease.
Similar to data from national cohorts, 20 BMI was also greater in those who had had critical disease compared with those with mild disease ( Figure S1), which may explain this observation.
Although not all the participants received steroid treatment, Cpeptide levels at follow-up were not different in those who had received dexamethasone. While the use of steroids probably contributes to insulin resistance during acute COVID-19, it does not appear to have a deleterious legacy effect on insulin secretion at follow-up.
Finally, while critical and viral illness-induced insulin resistance 14 probably explains a large proportion of the dysglycaemia observed, given that C-peptide levels were preserved at follow-up, these data suggest that any insulinopaenia during acute COVID-19 is not apparent at 3 months or more after presentation. These data are consistent with findings from other studies where it has been shown that, while a minority of patients with acute COVID-19 display characteristics of beta cell failure, in contrast to other causes of acute respiratory distress syndrome, the predominant cause of hyperglycaemia observed in patients with acute COVID-19 results from insulin resistance. 21 Furthermore, patients with COVID-19 have been observed to have reduced adiponectin levels, consistent with adipose tissue dysfunction, providing a potential mechanism for the insulin resistance observed. 21 Taken together, given the deleterious impact of hyperglycaemia on outcomes in patients with COVID-19, 22 agents that improve insulin resistance, such as metformin or thiazolidinediones, could have clinical utility, and hold potential as a focus for future research.
This post hoc analysis of data prospectively collected from survivors of COVID-19 is limited in that participants were not fasted.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/dom.14608.

DATA AVAILABILITY STATEMENT
The datasets generated during and/or analysed during the current study are not publicly available due to patient confidentiality but are available from the corresponding author on reasonable request.