In this study we found potential predictors of multiple and complex clinical outcomes, considering the role of hyperglycemia, inflammation markers and clinical history. This information allowed us to add evidence to what has been debated or hypothesized in the literature.

High level of glucose at admission, an increase in inflammatory parameters and impaired renal function at admission to the hospital were found associated with an increased mortality during hospitalization, independently from BMI, age and sex. The mean age of our population was 72 years, one would expect that age alone would be responsible for the increased mortality as reported in literature [12]; however, hyperglycemia, inflammation and impaired renal function seems to be important, independently of age. Moreover, in this population even a slightly higher glucose level as 165 mg/dL, that may be tolerate in daily management, was associated with an increased mortality. This is an intriguing point; we can speculate that in older people the cut off level of hyperglycemia leading to an increased risk of mortality during COVID-19 might be lower than in younger; therefore, an early check and control of glycemia is important.

The full inflammatory profile in DM patients with COVID-19 is not characterized yet, however, hyperinflammation could be a possible response to the infection. Indeed, cytokine storm, an activation cascade of auto-amplifying cytokines production due to unregulated host immune response to COVID-19 infection, has been proposed as a pathological mechanism [22].

Diabetes is also characterized by chronic, low-grade inflammation, which is a prominent feature of its complications, and its pathophysiology shares several proinflammatory molecules from the COVID-19 cytokine storm cascade [23, 24]. The underlying chronic inflammatory state in diabetes may be “locked and loaded” for virus-induced damage, promoting a vicious cycle of cytokine release, leading to more widespread multiorgan damage, including injury to tissues already weakened by pre-existing diabetes complications [25]. Different authors have hypothesized that, as a chronic inflammatory condition, DM may predispose individuals to an increased inflammatory response since hyperglycaemia has traditionally been thought to be a major driver of inflammation [26]. However, in our study, the indirect effect of glucose on mortality through inflammatory markers was not significant for the majority of inflammatory biomarkers evaluated, except for absolute neutrophil count, CRP and PCT. Finally, hyperglycemia at admission to the hospital had a direct effect, not mediated by inflammation, on mortality.

Our data confirm what was recently described by Kho et al. which hypothesized that CRP is a partial mediator of the association between DM and severe COVID-19 [27], while, to our knowledge, we firstly described a similar role for PCT. It is known that PCT is associated with insulin resistance and association of plasma PCT in the general population [23], and now it has also a similar feature in diabetic COVID-19 patient.

This is an intriguing point because CRP and PCT could have a double effect on mortality, both direct and indirect mediated by hyperglycemia, being key markers and predictors of this outcome. Inflammation alone is also responsible of an increased mortality, probably mediated by the COVID-19 cytokine storm cascade. Therefore, both inflammation and hyperglycemia are associated to an increase mortality, but in parallel ways.

The reduction of insulin resistance and consequent inflammation, could explicate our finding that the patients treated with metformin before hospitalization had a reduced risk of mortality during hospitalization.

Our data also demonstrated that an increase in basophil count was associated with a reduction in mortality. Only few works in the literature described the role of basophils in non-diabetic patients with COVID-19, despite their vital role in the pulmonary pathologies and regulation of immune responses, probably because their low number in the circulation is an important limitation. These studies suggested that basophils may exert a protective role; also, they showed that their absolute count seems to be reduced in COVID-19 patients as compared to controls, as well as in severe COVID-19 disease compared to mild/moderate disease [28]. Basophil cytokine responses to COVID-19 might help reducing the inflammation and also promoting antibody responses to the virus. Furthermore, basophils store the secretory granules of heparin that is only released into the vasculature at sites of injury [29], therefore helping maintaining a proper blood flow by balancing the active anticoagulant and procoagulant processes in pato-fisological condition. This property could be hypothesized also during the COVID-19 infection of diabetic patients. To our knowledge our study is the first that described a protective role of basophils in diabetic patients hospitalized with COVID-19, as they could act both reducing inflammation and improving an anticoagulant process.

An increased neutrophil count was found associated with an increased mortality in our cohort; we can theorize regarding the possible mechanism, since it is not clear and there are only few recent data on humans. Hyperglycemia affects the hematopoietic stem and progenitor cells in the bone marrow leading to enhanced myelopoiesis and elevated number of neutrophils and monocytes in the blood [30]. Moreover, neutrophils have been implicated in the induction of adipose tissue inflammation and insulin resistance; in fact, deletion of neutrophil elastase results in decreased adipose tissue inflammation, reduced myeloid cell content, and improved glucose tolerance and increased insulin sensitivity in obese mice [30]. Finally, we found an elevated neutrophil-to-lymphocyte (N/L) ratio, that can be due to an exaggerated myelopoiesis that typically elevates it [31].

Severe COVID-2019 disease is characterized by microthrombosis, increased coagulation and profound inflammation; platelets mediate thrombosis and, in these patients could increase thrombotic or inflammatory profile [32]. Consumption of platelets into a growing thrombus or platelet apoptosis might explain the thrombocytopenia present in patients with COVID-19.

An important and pragmatic aspect of our work is that the biomarkers employed can be obtained by the emergency laboratory in less than an hour and they are in-expensive and frequently used also in developing countries. Standardized protocols, well defined range limits, quality internal and external controls and low costs compared to research assay make them more interesting for characterize diabetic patients hospitalized for COVID-19 [33, 34].

Finally, we evaluated the pre-hospital status of our study population. Patients with microvascular complications and chronic kidney disease have higher mortality during hospitalization. Interestingly, no association of the other variables (macrovascular complications or obesity) before hospitalization with clinical outcomes has been evidenced, despite diabetes and cardiovascular disease are frequent comorbidities in patients with COVID-19 and play a role in adverse outcomes. Moreover, considering complications during hospitalization only respiratory failure increased mortality but singularly cardiovascular, thromboembolic or neurological complications did not.

The impact of sitagliptin treatment on mortality in diabetic patients and COVID-19 has been debated in literature [35, 36]; however we have only seven patients on sitagliptin in our cohort and therefore no analysis has been done.

One question of interest is the role of past glycemic control and not only acute hyperglycemia on the outcomes. In our study HbA1c before hospitalization were not associated with the clinical outcomes.

Our study has several limitations. Firstly during the pandemic COVID 19, hospital admissions increased sharply and the hospital rapidly became overloaded with patients affected by pneumonia and respiratory failure, of whom a relevant proportion in need of ICU admission and artificial ventilation. Intensive care unit (ICU) beds and ventilators were not available for all of these patients, therefore data relative to these outcomes could be underestimated.

Secondly, our data contain a high number of missing values of HBA1c, due to the fact that patients were not regularly followed by our outpatient service but by general practitioners.

Thirdly the current study was retrospective, with all the inherent limitations of such studies.

In conclusion hyperglycemia at admission, renal function and inflammatory parameters were found to be predictors of in-hospital mortality, while an increased basophil count was protective. Hyperglycemia had a direct effect on mortality, the indirect effect was only through absolute neutrophil count, CRP and PCT and markedly lower than the direct one.