Association of glycosylated haemoglobin HbA1c levels with outcome in patients with COVID‐19: A Retrospective Study

Abstract Patients with hyperglycemia tend to be susceptible to Coronavirus disease 2019 (COVID‐19). However, the association of HbA1c level with outcome of COVID‐19 patients was unclear. We performed a retrospective study of 2880 cases of COVID‐19 admitted in Tongji Hospital, Wuhan, China, among which 922 had detected the HbA1c levels. We found that COVID‐19 patients with either lower levels of HbAlc (3%‐4.9%) or higher levels of HbAlc (≥6%) were associated with elevated all‐cause mortality. Meanwhile, we observed that HbAlc levels were highly correlated with haemoglobin (Hb) and total cholesterol (TC) (P < .0001), moderately correlated with albumin (ALB) and high‐sensitive C reaction protein (hs‐CRP) (0.0001 < P<.001), and relatively low correlated with low‐density lipoprotein cholesterol (LDL‐C) (.001 < P<.01). These associated cofactors might together contribute to the clinical outcome of COVID‐19 patients. Furthermore, the mortality was higher in COVID‐19 patients with newly diagnosed diabetes mellitus (DM) compared with COVID‐19 patients with history of DM. Moreover, in patients with history of DM, the mortality was decreased in patients treated with anti‐hyperglycaemic drugs. In summary, our data showed that the in‐hospital mortality was increased in COVID‐19 patients with lower or higher levels of HbAlc. Meanwhile, initiation of appropriate anti‐hyperglycaemic treatment might improve the clinical outcome in COVID‐19 patients.

the concept that improved glycaemic control with better outcome in patients with COVID-19 and pre-existing type 2 diabetes.
While the close association between diabetes and increased mortality becomes clear, glycaemic-control achievements evaluation using glucose levels (FBG fasting glucose) are sometimes problematic.
Because blood glucose levels are highly variable and easily effected by short-term stress such as infection, glucocorticoid therapy and Somogyi phenomenon. Notably, the Somogyi effect is the tendency of the body to react to extremely low blood sugar (hypoglycaemia) by overcompensating, resulting in high blood glucose levels. 6,7 Therefore, FBG level alone might be unable to efficiently reflect the true glycaemic metabolism in COVID-19 patients. On the other hand, strict blood glucose surveillance or OGTT test might accelerate the shortage of medical resources during the COVID-19 pandemic.
HbA1c is produced by a non-enzymatic reaction that occurs between glucose and haemoglobin. 8 As plasma glucose increases, the fraction of HbA1c increases in a predictable way. This serves as a surrogate marker for average blood glucose levels over the previous 3 months prior to the measurement. 9 In 2009, the American Diabetes Association included HbA1c ≥ 6.5% (48 mmol/mol) as a diagnostic criterion for diabetes. [10][11][12] Fasting is not needed for HbA1c assessment and no acute perturbations (eg stress, diet and exercise) affect HbA1c. 13 Moreover, HbA1c captures chronic hyperglycaemia better than two assessments of fasting or 2-h oral glucose tolerance test plasma glucose. 13 However, to date, limited information is available regarding HbAlc levels and clinical outcome of COVID-19, which might be due to problems in standardization and variations in styles of HbA1c test among multiple-centred studies.
In this report, we performed a retrospective longitudinal study from a cohort of 992 confirmed COVID-19 cases enrolled in singlecentred Tongji hospital in Wuhan, China focusing on the association between plasma HbAlc level and clinic outcome in COVID-19 patients.
In addition, by using the cut-off value of HbA1c ≥ 6.5%, we sought to investigate the mortality of COVID-19 patients with in-hospital newly identified DM in comparison with previously diagnosed DM. Moreover, we further assessed the mortality of COVID-19 DM patients treated with different anti-hyperglycaemic drugs.

| Ethics
The study, approved by the Ethics Review Board of Tongji Hospital and Tongji Medical College, conforms to the principles outlined in the Declaration of Helsinki. Written and informed consent forms were waived by the ethics boards of the hospitals.

| Study design
The study included patients with COVID-19 diagnosed between 10 January 2020 and 30 March 2020. COVID-19 was diagnosed based on chest computed tomography (CT) manifestations and/or reverse transcription-polymerase chain reaction (RT-PCR) following the cri-

teria of the New Coronavirus Pneumonia Prevention and Control
Program (5th edition) published by the National Health Commission of China and WHO interim guidance. A total of 2880 patients with COVID-19 were initially screened for the study, and 922 of them had HbA1c level detected.

| Data collection
All clinical data (including basic information, clinical manifestations, laboratory findings, treatments and outcome during hospitalization) were obtained from patients' electronic medical records. The laboratory findings included routine blood test, fasting blood glucose (FBG) and HbA1c, C-reactive protein (CRP), D-dimer for liver function, kidney function, coagulation function and inflammation analysis.

| Statistical analysis
All statistical analysis was performed using SPSS 21.0 for Windows (SPSS Inc, Chicago, IL, USA). Categorical variables were presented as number (percentage), and continuous variables were presented as median (interquartile range). Categorical variables were compared using the chi-squared test or Fisher exact test. Normally and abnormally distributed continuous variables were compared using the Student's t test and the Mann-Whitney U test, respectively. Kruskal-Wallis ANOVA tests (nonparametric unpaired) was used among multiple groups. To assess the significance of the correlations, Spearman rank correlation coefficient was calculated. Unless otherwise stated, a value of P (or corrected P in case of multiple groups) <.05 was considered statistically significant.

| Association of HbAlc levels with mortality of COVID-19 patients
A total of 2880 confirmed COVID-19 patients were admitted to Tongji hospital, Wuhan, China from 10 January 2020 to 30 March 2020 during the pandemic, among which 922 patients had HbAlc level examined (Figure 1). We first analysed mortality by dividing these 922 COVID-19 patients into 6 subgroups based on FBG levels (Table 1). We found that higher levels of FBG expression were associated with increased COVID-19 mortality (Figure 2A) while lower FBG (3-4.9 mmol/L) was associated with the best clinical outcome (with mortality of 1.2%). We then analysed mortality by dividing these 922 COVID-19 patients into 6 subgroups based on HbAlc levels (Table 2). Unexpected, though HbAlc levels and FBG levels were highly correlated ( Figure 2B), lower levels (3%-4.9%) of HbAlc expression was associated with increased mortality (21.4%).
HbAlc levels between 5%-5.9% were associated with relatively improved outcome of patients with COVID-19 while higher HbAlc levels (≥6%) were associated with increased all-cause mortality ( Interestingly, we noted that in the subgroup when HbAlc was the lowest (3%-4.9%), FBG was unexpectedly higher (6.7 mmol/L) than the group with HbAlc levels among 5%-5.9% (FBG 5.3 mmol/L). Therefore, the high FBG levels in patients with lower HbAlc levels might be partly explained by overcompensated effects for unnoticed hypoglycaemia, similar as the Somogyi effect (See further in discussion).
These data suggested that lower levels of HbAlc (3%-4.9%) and higher levels of HbAlc (≥6%) were both associated with increased all-cause mortality of COVID-19 patients, which required attention in clinical practice.

| Association of HbAlc levels with other clinical parameters in COVID-19 patients
We then performed correlational analyses between HbAlc levels and    .048

| Mortality in DM patients treated with different anti-hyperglycaemic drugs
To evaluate the influence of long-term anti-hyperglycaemic drugs (not the short-term in-hospital treatment), we then assessed the in-hospital mortality of COVID-19 patients with DM history. Most of these patients were treated with anti-hyperglycaemic drugs regularly before the COVID-19 pandemic. We noted that the mortality in pre-diagnosed diabetic COVID-19 patients without any anti-hyperglycaemic drugs was 32.4% while the mortality was 7.9% in insulin-, 4.1% in metformin-, 0% in SU/GLN-and 2.3% in AGI-treated patients, respectively ( Table 4).
Because of the limited number of patients, we were unable to further divide patients into single anti-diabetic drug groups. Nevertheless, these data still suggested that early identification of diabetes and initiation of appropriate treatment might prevent worse outcome of COVID-19. Furthermore, we compared the difference between insulin therapy alone, insulin plus oral hypoglycaemic drug therapy and oral hypoglycaemic drug therapy alone in the Table S1. However, there was no difference of mortality between insulin along treated patients and patients without hypoglycaemic drug. Besides, the mortality in the oral hypoglycaemic group was much lower than that in the insulin group

| D ISCUSS I ON
In this study, we analysed the association of HbAlc levels with clinical outcome of COVID-19 patients, finding that all-cause mortality was increased in patients with lower levels of HbAlc (3%-4.9%) and higher levels of HbAlc (≥6%) compared with HbAlc levels between 5% and 5.9%. These data suggested that HbAlc might be a potential prognostic marker for assessing the risk of death in COVID-19 patients. Furthermore, by using the criterion of HbAlc ≥ 6.5%, we ob- Interleukin-1β (pg/ml) 8.5 (6.5-12.6) 7.7 (6.0-9.6) 8.9 (6.3-15.8) .244 High-sensitive C reaction protein (pg/ml) 10.8 (1.7-50.5) 13.5 (3.2-67.5)* 37.4 (5.5-99.9)* ,# <.001 (Continues) more robust indicator of glycaemia than a parameter describing it in the short term or in a given moment only. 13,14 The measurement of HbAlc equals the assessment of hundreds of fasting glucose levels as well as postprandial glucose peaks; therefore, it is a more robust and reliable measurement than FPG plasma glucose. 15  The Somogyi effect, also known as the 'chronic Somogyi rebound' or 'post hypoglycemic hyperglycemia', was described that when blood glucose levels dropped too low during the late evening, activation of counter-regulatory hormones such as adrenaline, corticosteroids, growth hormone and glucagon might be observed, leading to activation of gluconeogenesis and resultant hyperglycaemia in the early morning. 16 To further explore the potential relationship, we also divided the group with HbA1c between 3% and 4.9% into two subgroups according to whether the FBG level was higher than 7 mmol/L, one of the diagnostic criteria for diabetes. The mortality in group with FBG higher than 7 mmol/L was much higher than in group with FBG lower than 7 mmol/L (37.5% vs 0%, P =.209). Because of the small number of patients (6 vs 8)  .718 Temperature ( 19 Other study also showed that patients with newly diagnosed T2DM exhibited a marked chronic inflammatory state characterized by increased IL-6, TNFα, IL-1β, IL-2 and ferritin levels. After 1 year of treatment with acarbose or metformin, IL-6, TNFα, IL-1β and ferritin levels were significantly decreased compared with the baseline. The anti-inflammatory effects of acarbose and metformin were comparable and required a long-term treatment (1 year). 20 While recent study has suggested that host inflammatory response becomes a major cause of lung damage and subsequent mortality during COVID-19, 21 its entirely possible that anti-hyperglycaemic drugs such as acarbose and metformin protected against inflammatory damage caused by COVID-19 through their potential beyond glucose-lowering effects. These possibilities are intriguing subjects for future studies.
In terms of the insulin treatment, it should be noted that in this study, only DM patients treated with regularly basal levels of long-acting or intermediate-acting insulin analogues were considered as insulin users.
Short-acting human regular insulin was not included. Therefore, the effects of in-hospital short-term insulin treatment (or other oral antidiabetic drugs) are still unclear and remain to be determined. patients.

ACK N OWLED G EM ENTS
We thank our colleagues from the Division of Cardiology, Tongji Hospital for their tremendous efforts during this investigation.