At-admission HbA1c levels in hospitalized COVID-19 participants with and without known diabetes

Background To examine glycaemic status, and the impact of at-admission HbA1c levels on outcome, in a large group of participants hospitalized for COVID-19. Methods We inclued 515 participants with confirmed COVID-19 infection, with or without known diabetes, who met the following additional criteria: 1) age > 18 years, 2) HbA1c was determined at admission; 3) fasting plasma glucose was determined in the week of admission, and 4) discharge or death was reached before the end of the study. We examined attributes of participants at admission and 3–6 months post-discharge. To assess the associations of pre-admission attributes with in-hospital mortality, logistic regression analyses were performed. Results Mean age was 70 years, 98.8% were of white race, 49% were female, 31% had known diabetes (KD), an additional 7% met the HbA1c criterion for diabetes, and 13.6% died. In participants with KD, FPG and HbA1c levels were not associated with mortality in adjusted analyses; however, in participants without KD, whereas FPG showed direct association with mortality, HbA1c showed slight inverse association. Conclusions There was a very high prevalence of people without KD with HbA1c levels above normal at-admission. This alteration does not seem to have been related to blood glucose levels.


Introduction
There is now considerable evidence that diabetes increases the likelihood of poor outcome among people with coronavirus disease 2019 (COVID- 19), clinical studies in several countries having found that diabetes can increase the probability of a person hospitalized with COVID-19 progressing to intensive care and death [1][2][3]. In general, it is assumed that hyperglycaemia at the time of admission likewise increases the risk of poor outcome, regardless of prior diabetes status; indeed, tight glycaemic control significantly improves the prognosis of such peoples [4][5][6]. However, although this background suggests the advisability of checking glycosylated haemoglobin (HbA 1c ) upon hospital admission for COVID-19, since HbA 1c reflects average glycaemia over the preceding 2 to 3 months, the significance of admission HbA 1c for the management of COVID-19 peoples remains unclear [7]: some studies have found significant association between admission HbA 1c and disease progression or mortality among COVID-19 people, whereas others have not [8][9][10][11]. Furthermore, there have in fact been rather few studies that have measured admission HbA 1c in all people (rather than only in those known to have diabetes), and most of them have been quite small, in which HbA 1c data taken from medical records were often collected long before admission for COVID-19, generally only from participants with diabetes or suspected diabetes [12][13][14][15], or for<10% of participants without known diabetes [15].
The initial aim of the present study was therefore to determine the impact of admission HbA 1c on outcome in a large sample of participants hospitalized consecutively for COVID-19.

Participants and consent
The University Hospital Complex, Santiago de Compostela, Spain, is a tertiary centre serving an almost exclusively white population of approximately 450 000 people. This study concerned adult participants severely infected with acute respiratory syndrome coronavirus 2 (SARS-Cov-2, confirmed by polymerase chain reaction testing of nasopharyngeal specimens) who were admitted to the hospital between March 2020 and January 2021 and who met the following additional criteria: 1) age > 18 years, 2) HbA 1c was determined at admission, after<3 days of corticosteroid treatment, 3) fasting plasma glucose (FPG) was determined in the week of admission, and 4) the endpoint (discharge or death) was reached before the end of the study (Fig. 1).
Anonymised information was entered on database hosted on secure server. Individual consent was not required for anonymous data. The study was reviewed and approved by the Clinical Research Ethics Committee of Consellería de Sanidade. Xunta de Galicia, Spain and in accordance with the current Declaration of Helsinki.

Study design
We retrospectively assessed the glycaemic status at admission and COVID-19-related mortality in participants with and without known diabetes (KD). Association of mortality with age, sex, comorbidities was also evaluated.

Data collection and laboratory procedures
We retrospectively collected data from medical records concerning HbA 1c , FPG, corticosteroid administration at admission, demographic characteristics, comorbidities, complications and clinical outcome (discharge or death while hospitalized). Comorbidities included hypertension, coronary artery disease, past or active cancer, chronic kidney disease (CKD) associated with stage 3-5 kidney failure (estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m 2 as calculated by the Chronic Kidney Disease Epidemiology Collaboration equation), obesity (body mass index (BMI) ≥ 30 kg/m 2 ) and diabetes status. Following the American Diabetes Association (ADA) criteria, we defined prediabetes as an FPG level of 100-125 mg/dL (5.6-6.9 mmol/L) and/or HbA 1c of 39-47 mmol/mol (5.7-6.4%), and diabetes as HbA 1c ≥ 48 mmol/mol (6.5%), FPG level ≥ 126 mg/dL (≥7 mmol/L), or current use of an antidiabetic drug [16]; normoglycaemia was indicated by FPG and HbA 1c values both being below the cutoff points for prediabetes. "Known diabetes" (KD) was attributed when the participant's medical record showed a diagnosis of diabetes, or the participant reported having diabetes and/or being treated with antidiabetic drugs.
The presence of SARS-CoV-2 in respiratory specimens was detected by real-time RT-PCR methods. HbA 1C was determined by highperformance liquid chromatography in a Menarini Diagnostics Arkray Adams HA-8180 T analyser; all HbA1c values were converted to Diabetes Control and Complications Trial (DCCT)-aligned units in accordance with the US National Glycohemoglobin Standardization Program (NGSP) [17]. Fasting plasma glucose was determined by the glucose hexokinase method in a Siemens Healthcare Diagnostics Advia 2400 autoanalyser; to limit distortion due to insufficient fasting (at least 8 h at-admission), participants with abnormal values were attributed the lowest morning blood glucose level during the first 3 days of hospitalization. All analyses were performed on the day of collection in the Clinical Biochemistry Laboratory of the University Hospital Complex, Santiago de Compostela.

Statistical analysis
Descriptive statistics were calculated for all variables, with mean ± SD or median and interquartile range reported for continuous variables, and number and percentage of total for categorical and integer variables. The baseline characteristics of the subjects with KD and without KD were compared using the Student t test for continuous variables (a two-tailed P value ≤ 0.05 was deemed statistically significant) and the Pearson χ 2 test for categorical variables. To investigate an observed discrepancy between the distributions of FPG and HbA 1c among non-KD participants, we used χ 2 tests to compare, for each HbA 1c -assigned diabetes status range and each FPG-assigned range, the proportions of non-KD participants in that range at admission and 3-6 months postdischarge; and we performed univariable analyses to compare HbA 1c range subgroups of the non-KD group in regard to demographic characteristics, non-KD comorbidities, FPG, and outcome. To assess the associations of preadmission attributes with in-hospital mortality, we performed logistic regression analyses with death as the outcome variable. In multivariable logistic regression analyses we used all variables as predictors simultaneously (age, sex, hypertension, dyslipidemia, current or former smoking, obesity, chronic kidney disease, coronary artery disease, cancer, fasting plasma glucose, and HbA 1c ). The odds ratios from multivariable analyses are reported together with their 95% CIs. All calculations were performed using SPSS version 23 statistical software.

Results
During the study period, 3145 participants tested positive for SARS-CoV-2 in the hospital laboratory ( Fig. 1); of these, 618 (19.7%) required hospitalization. Of these 618 participants, 103 were excluded from the study because HbA 1c data were lacking or obtained after>3 days of steroid treatment; of the 515 participants included in the analysis, 413 (80.2%) had HbA 1c determined no later than the day on which steroid treatment started. The vast majority of participants were white (98.8%). Most participants (440 (85.4%)) had at least 1 comorbidity, KD in 159 cases (30.9%). The mean (SD) number of comorbidities other than KD was 4.1 (1.3) in the KD group as against 1.8 (1.4) in participants without KD (difference, 2.2; 95% CI, 2.0-2.5; P < 0 0.001). More than half of all participants (57.4%) had known exposure to someone with COVID-19. Table 1 summarizes baseline demographic and clinical characteristics. Although KD participants, as well as having more non-KD comorbidities than participants without KD, were also significantly older, the KD and non-KD groups had similar proportions of participants older than 60 years. Suprisingly, although the third quartile of the FPG distribution of non-KD participants, 5.5 mmol/L, was in the normoglycaemic range, the whole of their HbA 1c interquartile range, 39-44 mmol/mol (5.7-6.2%), was in the prediabetic range. Table 2 shows the glycaemic status of 215 non-KD participants (66.8% of survivors) for whom there were FPG and HbA 1c data not only at hospital admission but also 3-6 months after discharge. At admission, only 20.9% of these participants had HbA 1c < 39 mmol/mol (5.7%), but 3-6 months after discharge, 47.4% did (P < 0.0001). There was no such difference in regard to the proportion of participants with normoglycaemic FPG (75.8% at admission as against 68.4% 3-6 months after discharge; P = 0.09). Accordingly, only 39.1% of these participants were afforded the same glycaemic status by both ADA threshold criteria at admission, whereas 62.9% were 3-6 months after discharge. Similar results are observed if the at-admission data of all (3 5 6) non-KD participants are compared with the post-discharge data of the 215 (presented in Supplementary Table 1 ). The results with the WHO threshold criteria (presented in Supplementary Table 2 ) are numerically different, but qualitatively the sameor even more marked -as those obtained with the ADA criteria (Table 2).
At admission, approximately two-thirds of non-KD participants, 66.3%, had prediabetes (PD) on the basis of their HbA 1c levels. Univariable anlyses indicated that those with PD were more likely than those with normoglycaemia (NG) to be older than 60 years, and to suffer dyslipidemia (42.8 vs 31.0%, P = 0.03) and obesity (40.0 vs 28.8%, P = 0.03), but less likely to suffer chronic kidney disease (6.8 vs 14.3%, P = Data are n (%) or means ± SD or median (IQR). P value for comparison between groups with and without known diabetes. Data are n (%). 0.02) ( Table 3). Participants in the diabetic range (D) were more likely than NG participants to be older than 60 years (75 vs 56%, P = 0.03) and to have hypertension (66.7 vs 40.5%, P = 0.004), dyslipidemia (58.3 vs 31.0%, P = 0.003) and obesity (63.6 vs 28.8%, P < 0.001). During the study period, 70 deaths occurred in the whole study group (13.6%). In univariable analysis the risk of death was 2.8 times greater among KD participants than among non-KD participants (95% CI, 1.7-4.6). However, in the non-KD group, the death rate among participants with HbA 1c in the PD range was 2.6 times smaller than that of the NG subgroup, 14.3%, as well as being 4.5 times lower than that of the D subgroup, 25% (Table 3). There were no significant differences among the NG, PD and D subgroups of the non-KD group as regards the duration of hospitalization (median 10 days (range 1-62) in the whole non-KD group) or the proportion requiring ICU care (10.7% in the whole non-KD group) ( Table 3). Table 4 lists the results of multivariable analyses for factors associated with COVID-19 mortality in the KD and non-KD groups. In the KD group, age (OR 1.05, 95% CI 1.00-1.11; P = 0.04), cancer (OR 9.18, 95% CI 2.82-29.93; P < 0.001) and chronic kidney disease (OR 5.02, 95% CI 1.99-12.66; P = 0.001) were significant independent predictors of mortality, but there was no significant association with FPG or HbA 1c . In the non-KD group, age (OR 1.12, 95% CI 1.07-1.18; P < 0.001) and FPG (OR 2.06, 95% CI 1.45-2.92; P < 0.001) were significantly associated with higher mortality, but HbA 1c was independently associated with significantly lower mortality (OR 0.27, 95% 0.12-0.63; P = 0.002).

Discussion
This study focused on SARS-CoV-2-positive participants admitted to a single hospital since March 2020 and before the start of general vaccination. The prevalence of known diabetes in this cohort was 30.9% (greater than previously reported for Europe but similar to U.S. data [12,[18][19][20], and an additional 7% met ADA criteria for type 2 diabetes.
In keeping with the findings of previous studies [18,[21][22][23], participants with known diabetes were 2.4 times more likely to die than participants without diabetes, the significant independent risk factors for KD participants, as identified by multivariate analysis, being age, cancer, and chronic kidney disease; no significant association with FPG, HbA 1c , sex, hypertension, coronary artery disease, smoking, or obesity was observed.
However, results of this study concern participants without KD, who had not been receiving any kind of hypoglycaemic treatment, only age and FPG were identified as positive risk factors for mortality, and surprisingly, admission HbA 1c emerged as a negative risk factor. In particular, non-KD participants with prediabetic admission HbA 1c (66.3% of non-KD participants) had a significantly lower death rate than those with normal HbA 1c levels, as well as with respect to those with diabetic HbA 1c . Furthermore, 3-6 months after discharge, this "prediabetic" subgroup made up only 45.6% of the 215 participants with follow-up data (66.8% of survivors), while the proportion of non-KD participants with normoglycaemic HbA 1c had risen from 23.6% at admission to 47.4% (P < 0.001 in both cases). By contrast, the proportions of non-KD participants with FPG in the normoglycaemic and prediabetic ranges did not change significantly (and such change as there was, was in the opposite direction to that observed for the HbA 1cdefined subgroups). These results question the notions that elevated HbA 1c at admission in COVID-19 participants is 1) necessarily due to an increase in blood glucose levels, and 2) necessarily a predictor of poor outcome. Among people with diabetes, the relevance of HbA 1c to poor outcome has been questioned previously by studies that failed to find any association between them [10,13,24,25]. In this context (diabetes), in which it can indeed be assumed that high HbA 1c is probably due at least in great part to hyperglycaemia, questioning its relevance to outcome amounts to questioning whether the well-documented statistical association of diabetes with poor outcome is directly due to hyperglycaemia or reflects metabolic concomitants or clinical complications of diabetes; but it is also possible that the hyperglycaemia of these participants is in part due to metabolic stress due to COVID-19 [24]. In the present study, FPG was a predictor of mortality among non-KD participants, though whether as simply a marker of metabolic stress remains an open question.
In regard to the relationship between high HbA 1c and glycaemia (point 1 above), our findings suggest that the elevated HbA 1c levels observed in a large proportion of participants with COVID-19 may be due to a direct or indirect effect of SARS-CoV-2 infection on haemoglobin, rather than to de novo hyperglycaemia or, in participants with diabetes, a worsening of glycaemic control. Although interaction between the virus and haemoglobin through ACE2, CD147, CD26 and other receptors located on erythrocytes and/or blood cell precursors has been put forward as a possible cause of hypoxia [26], it is conceivable that conformational changes in hemoglobin due to binding of the virus [27] might promote glycosylation without leading to poor outcome. It should also be borne in mind that it is normal for a proportion of the population to have higher HbA 1c levels than those generally regarded as corresponding to their glycaemia [28].
Three-quarters of the non-KD participants in this study had normal FPG levels on admission. That considerably higher levels have been reported in some studies [5,18] may have been due to their not ensuring an 8-hour fast, possibly because of the difficulty of obtaining these samples. Studies in which the 8-hour fasting period was guaranteed have reported results similar to ours [29,30], not only in regard to the percentages of people with different glycaemic statuses, but also in that admission FPG was an independent predictor for mortality in non-KD participants.
The main strengths of this study are that, unlike most studies of the impact of HbA 1c on COVID-19 outcome, it only admitted participants whose HbA 1c was determined on admission to hospital before the start of general vaccination. Further, HbA 1c was determined in all cases by the same method in the same laboratory on the day the sample was obtained. There are also limitations to our work. Since HbA 1c was determined post-discharge only in 67% of the non-KD participants (and naturally not in those who died), bias may have occurred due to these participants having been prioritized for testing for reasons unknown to us. Also, the data are predominantly from Spanish whites, which may limit generalizability.
In conclusion, in a large sample of participants with COVID-19 at admission and without known diabetes, we found a large proportion of subjects, 76.4%, with HbA 1c levels above the ADA criterion for normoglycaemia. This alteration does not seem to be related to actual glycaemia, because only 24.4% of non-KD participants had nonnormoglycaemic FPG levels at admission; and because 3-6 months after discharge the percentage of non-KD participants with HbA 1c in the normal range was more than double that observed at admission. These findings may explain, at least in part, the absence of an association between HbA 1c levels and in-hospital mortality observed in several studies. More research seems necessary to determine whether these findings generalize to other populations and to study the causes of this intriguing behavior. Determining the HbA 1c of COVID-19 participants at admission to hospital would allow better stratification of their risk, and guide the care of both participants with diabetes and those without.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.