The present study suggests that, at least in elderly patients with COVID-19, adequate controlled diabetes at the time of admission and during hospitalization might reduce the risk of in-hospital mortality associated with T2D. This interpretation may explain why T2D per se was not demonstrated to be an independent predictor of death in our previous analysis in the same patients [16], in apparent contrast with other studies [17]. Zhu et al. [18] have recently described, in a retrospective longitudinal study, that well-controlled blood glucose (blood glucose values within 70 to 180 mg/dl) was associated with markedly lower mortality compared to individuals with poorly controlled blood glucose (blood glucose values exceeding 180 mg/dl).
Glucose control during hospitalization appeared to be appropriate both for the mean pre-meal blood glucose value and for CV, which were acceptable for hospitalized patients in our study. The CV in patients with diabetes was lower than 30%, which is the threshold value suggested by the international consensus on new glucometric parameters [19]. Furthermore, none of the patients managed with the titration algorithm had severe hypoglycaemia during hospitalization. The absence of hypoglycaemic events might have also influenced the result of the study. Indeed, as known from the literature, hypoglycaemia associates with increased mortality in hospitalized patients [20]. Therefore, the titration algorithm used during hospitalization seemed to be effective to obtain glucose control in our study. Unfortunately, in the studies conducted so far, the possible impact of glycemic control during hospitalization was not appropriately assessed as a time-dependent variable. Although, due to a small number of patients, we were not able to perform a survival analysis with glycaemia as a time-dependent variable, our study suggests that maintaining glycemic control may dilute the impact of T2D per se in the analyses conducted so far. Among variables which may impact on glucose levels, further analysis should also consider the effect of concomitant drugs. More than 2/3 patients in our group received hydroxychloroquine and approximately half of them received corticosteroids. Hydroxychloroquine may lower blood glucose and corticosteroids may raise it, so further studies in COVID-19 patients should consider their effects. Also, in clinical practice, the possible effects of these drugs on glucose control should be considered, with pre-prandial glycaemia measured even in non-diabetic patients taking these drugs. The benefit of a strict monitoring, should be balanced, however, against the risk of infections for the healthcare workers. For this reason, we prioritized to diabetic patients or those symptomatic for hypo or hyperglycaemia, while the other patients were not monitored.
An alternative interpretation for the lack of an association between T2D and mortality could be that possible pre-existing complications of T2D are more important. In apparent support to this interpretation, in our previous analysis [16], CV diseases (and not T2D) emerged as an independent predictor of death over T2D per se. More in general, cardiovascular disease is a predictor of mortality in infected patients regardless of diabetes and even in non COVID-19 patients [21]. Along the same line, Apicella et al [22] showed that, in COVID-19 patients, poorer prognosis of people with diabetes is likely to be the consequence of a syndromic nature of diabetes, in which hyperglycaemia, hypertension, obesity, and cardiovascular diseases all contribute to increase the risk of death. Therefore, albeit TD2 per se may not emerge as a risk factor, attention should be given to these patients who may be more fragile for comorbidities, especially the cardiovascular one.
We found that patients with diabetes who died were older and had lower lymphocyte count than patients with diabetes who survived. The recent meta-analysis by Huang et al [23] has demonstrated the strong relationship between diabetes and poor COVID-19 outcomes but not with ICU admission. The authors have also demonstrated that the relationship between diabetes and worst infection outcomes is complex and affected by the prevalence of other comorbidities as hypertension. In our sample the prevalence of hypertension was comparable between patients with and without diabetes as well as between deceased and survivors with diabetes.
Our study is affected by several limitations. First, the small number of patients may have precluded to get a definitive answer as to whether or not T2D per se is a variable independently correlated or predictive of death in COVID-19 patients. Second, unfortunately, we did not have a recent glycated haemoglobin value and could not measure it during hospitalization. Without HbA1c it is impossible to separate undiagnosed T2D (for example, HbA1c 11% and casual plasma glucose 250 mg/dl) and new onset diabetes after COVID-19 infection (for example, HbA1c 5.6% and casual plasma glucose 260 mg/dl). Notwithstanding these limitations, our study has strengths. First, as far as age is concerned, population was quite homogeneous. Second, almost all (50 patients) came from LTHCF where a COVID-19 outbreak occurred, so they were infected almost at the same time, making the analysis of intra hospital mortality more accurate and not biased by different length of infection prior to enrollment. Lastly, our operational study, evaluated for the first time, titration algorithm which appeared to be beneficial in allowing a good glucose control. At the same time, our protocol was designed with the aim of reducing the number of contacts between healthcare workers and infecting patients, thus allowing to contain the intra hospital risk of infection.