Association between prior cardiometabolic therapy and in-hospital mortality in very old patients with type 2 diabetes mellitus hospitalized due to COVID-19. A nationwide observational study in Spain

Advanced age and diabetes are both associated with poor prognosis in COVID-19. However, the effects of cardiometabolic drugs on the prognosis of diabetic patients with COVID-19, especially very old patients, are not well-known. This work aims to analyze the association between preadmission cardiometabolic therapy (antidiabetic, antiaggregant, antihypertensive, and lipid-lowering drugs) and in-hospital mortality among patients ≥ 80 years with type 2 diabetes mellitus hospitalized for COVID-19. We conducted a nationwide, multicenter, retrospective, observational study in patients ≥ 80 years with type 2 diabetes mellitus and COVID-19 hospitalized in 160 Spanish hospitals between 1 and May 2020 who were included in the SEMI-COVID-19 Registry. The primary outcome measure was in-hospital mortality. A multivariate logistic regression analysis were performed to assess the association between preadmission cardiometabolic therapy and in-hospital mortality. The regression analysis values were expressed as adjusted odds ratios (AOR) with a 95% condence interval (CI). In order to select the variables, the forward selection Wald statistic was used. Discrimination of the tted logistic model was assessed via a receiver operating characteristic (ROC) curve. The Hosmer-Lemeshow test for logistic regression was used to determine the model’s goodness of t.


Conclusions
We found important differences between cardiometabolic drugs and in-hospital mortality in older patients with type 2 diabetes mellitus hospitalized for COVID-19. Preadmission treatment with dipeptidyl peptidase-4 inhibitors and angiotensin receptor blockers may reduce in-hospital mortality; other antidiabetic drugs, angiotensin-converting enzyme inhibitors and statins seem to have a neutral effect; and acetylsalicylic acid may be associated with excess mortality.

Background
Aging is one of the most signi cant factors associated with a poor prognosis in coronavirus disease-2019 (COVID-19) [1][2][3][4]. In addition, diabetes is a common comorbidity among patients with severe COVID-19 [5,6] and while it does not increase susceptibility to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, it has also been correlated with a more severe course of COVID-19 [7,8].
Our aim was to analyze the association between preadmission cardiometabolic therapy (antidiabetic, antiaggregant, antihypertensive, and lipid-lowering drugs) with in-hospital mortality in patients ≥ 80 years with type 2 diabetes mellitus (T2DM) hospitalized due to COVID-19.

Study Design and Population
We conducted an observational, multicenter, nationwide study of patients ≥80 years of age with T2DM hospitalized with COVID-19 in Spain from March 1 to May 29, 2020. All patient data was obtained from the Spanish Society of Internal Medicine's SEMI-COVID-19 Registry, in which 160 hospitals in Spain participate. The SEMI-COVID-19 Registry retrospectively compiles data on the rst admission of patients ≥18 years of age with COVID-19 con rmed microbiologically by a reverse transcription polymerase chain reaction (RT-PCR) test. More in-depth information on the justi cation, objectives, methodology, and preliminary results of the SEMI-COVID- 19 Registry have recently been published [6].

De nition of variables
Patients were considered to have T2DM if this diagnosis was recorded on their electronic medical record and they were treated with antidiabetic drugs. We analyzed the use of antidiabetic drugs (metformin, DPP-4i, insulin, sodium-glucose cotransporter 2 inhibitors (SGLT-2i) and glucagon-like peptide-1 receptor agonist (GLP-1ra), antiaggregant drugs (ASA), antihypertensive drugs (ACEI, ARB), and statins. All preadmission comorbidities were collected from patients' electronic medical records, which were obtained from each hospital. In-hospital mortality was the primary outcome variable. More in-depth information about the de nition of other variables has recently been reported in manuscripts published by the SEMI-COVID-19 Network [4,6].

Statistical analysis
Patients were divided into two groups: survivors and non-survivors. The characteristics of each group were analyzed using descriptive statistics. Continuous and categorical variables were expressed as medians and interquartile ranges (IQR) and as absolute values and percentages, respectively. The differences between groups were calculated using the Mann-Whitney U test for continuous variables and Pearson's chi-square test for categorical variables. Values were considered to be statistically signi cant when p<0.05.
A multivariate analysis was performed to control for confounding variables. The regression analysis values were expressed as adjusted odds ratios (AOR) with a 95% con dence interval (CI). A multiple logistic regression analysis was used to identify independent variables of in-hospital mortality. In order to select the variables, the forward selection Wald statistic was used. Variables analyzed in the model were: demographics (age, sex, acquisition), comorbidities and dependence (degree of dependence, Charlson Comorbidity Index, hypertension, dyslipidemia, coronary disease, cerebrovascular disease, peripheral vascular disease, atrial brillation, heart failure, dementia, chronic lung disease, obesity, malignancy, moderate-to-severe renal disease), symptoms (dyspnea), physical examination (oxygen saturation <90%, temperature 37.8ºC, tachycardia, quick sequential organ failure assessment score ≥2), laboratory ndings (neutrophils, lymphocytes, hemoglobin, platelet count, glucose, estimated glomerular ltration rate, lactate dehydrogenase, c-reactive protein, alanine aminotransferase), and treatment (metformin, DPP-4i, insulin, SGLT-2i, GLP-1ra, ASA, ACEI, ARB and statin). Discrimination of the tted logistic model was assessed via a receiver operating characteristic (ROC) curve. The Hosmer-Lemeshow test for logistic regression was used to determine the model's goodness of t. Due to the fact that there were some missing values, variables which were not recorded for >25% of patients were excluded from the analysis. These included serum ferritin, D-dimer, interleukin-6, procalcitonin, venous lactate, and aspartate aminotransferase. Statistical data analysis was performed using IBM SPSS Statistics for Windows, Version 22.0 (Armonk, NY: IBM Corp).

Results
Baseline clinical characteristics, presentation, and laboratory data Of the 2,763 patients ³80 years old analyzed, 790 (28.6%) had T2DM. Figure 1 shows the patient inclusion owchart for this study. A total of 385 patients (48.7%) died during admission. Baseline clinical characteristics, clinical presentation, and laboratory data of patients grouped by non-survivors and survivors are shown in Table 1. The percentages of females in the non-survivor and survivor groups were 43.2% and 51.0%, respectively (p=0.03). The percentages of patients with moderate and severe dependency in the non-survivor group were higher than in the survivor group (52.6% vs. 44.8%, p<0.001).
The median Charlson Comorbidity Index value and prevalence of comorbidities were similar in the survivor and non-survivor groups, with the exception of dementia (37.9% vs 30.2%, p=0.03). The presence of dyspnea, pulse oximetry <90%, fever, tachycardia, and qSOFA ³ 2 were more common in the non-survivor group. In terms of laboratory values, a higher proportion of patients in the non-survivor group had high admission values of leukocytes, neutrophils, plasma glucose, lactate dehydrogenase, aminotransferases, C-reactive protein, lactate, procalcitonin, ferritin, and D-dimer (p<0.05) and low estimated glomerular ltration rate (eGFR) and lymphocyte values.

Preadmission cardiometabolic therapy
Preadmission cardiometabolic therapy of patients grouped by non-survivors and survivors is shown in Table 2. The antidiabetic drugs used before admission were similar between the non-survivor and survivor groups except for DDP-4i, which were used in 28.9% of non-survivors and 36.2% of survivors (p=0.03). Statin and ARB use were less frequent in non-survivors than in survivors (49.2% and 7.1% versus 56.6% and 19.8%, respectively, p<0.05) while the use of antiaggregants was similar in both groups.

Associations between preadmission cardiometabolic drugs and in-hospital mortality
On the multivariate analysis, the preadmission cardiometabolic medications found to be independent protective factors against in-hospital mortality were the use of DPP-4i (AOR 0.502, 95% CI 0. with in-hospital mortality. In this model, the goodness-of-t showed a p-value of 0.142 (Hosmer-Lemeshow test) and the area under the ROC curve was 0.788 (discrimination). Associations between preadmission cardiometabolic drugs and in-hospital mortality are shown in Table 3.

Discussion
The COVID-19 pandemic has affected older, frail individuals with diabetes particularly severely and there is still no effective treatment for COVID-19 available. Therefore, we performed a multicenter, nationwide, retrospective, observational study to analyze the impact of preadmission cardiometabolic medication on mortality in elderly diabetic patients hospitalized for COVID-19.
Our study suggests that preadmission treatment with DPP-4i and ARB could be associated with reduced in-hospital mortality in elderly patients with T2DM hospitalized for COVID-19 whereas treatment with ASA could be associated with excess mortality. Other antidiabetic drugs, ACEI, and statins showed neutral association with mortality.
So far, there has been no conclusive evidence in regard to the potential implications of cardiometabolic therapies on COVID-19 outcomes, especially among elderly patients. Additionally, whether certain antidiabetic drugs can improve the prognosis of diabetic patients with COVID-19 remains unknown. Studies on glucose-lowering drugs in patients hospitalized with COVID-19 have shown con icting results.
In some reports, the use of oral antidiabetics had a neutral effect on in-hospital mortality and the composite outcome of poor prognosis, de ned as progression to severe or critical illness and in-hospital death [19,20], whereas in other studies, metformin had a bene cial effect on clinical outcomes [16,18]. Insulin use was associated with a greater risk of poor prognosis compared to not using it [20,21].
In our study, which was conducted solely in very old patients with T2DM, all antidiabetic therapies showed neutral effects on the clinical outcomes of COVID-19 with the notable exception of DPP-4i which, after exhaustive adjustment for potential confounding factors, were associated with a signi cant reduction in in-hospital deaths. Overall, data from human studies on the effects of DPP-4i in COVID-19 are scarce. Better clinical outcomes in patients with T2DM and COVID-19 have also been reported in a population-based study of 832 patients from the National Health Review and Assessment Service database in Korea [22].
In a recent multicenter, case-control, retrospective, observational study of 338 patients with T2DM admitted to hospitals in northern Italy for COVID-19, sitagliptin treatment in conjunction with insulin administration upon admission was determined to be associated with reduced mortality and improved clinical outcomes when compared to standard of care [15]. These potential bene ts of DPP-4i in patients with T2DM and COVID-19 must be con rmed in further research and placebo-controlled trials. Currently, there are two ongoing trials analyzing the safety and e cacy of linagliptin in patients with T2DM hospitalized for COVID-19 (NCT04542213, NCT04371978).
Several possible mechanisms have been proposed to explain the potential bene ts of DPP-4i in individuals with T2DM and COVID-19 [23][24][25]. First, T2DM is characterized by an overexpression of DPP-4 receptors, thus their inhibition may have immunoregulatory and anti-in ammatory effects [26,27]. Second, aging is associated with changes in cellular and humoral immunity that could favor worse outcomes in COVID-19 [28]. Third, DPP-4 has been identi ed as a receptor for MERS-CoV [22] and, additionally, the structure of SARS-CoV-2 spike glycoprotein S1, which mediates virus entry into the host cell, has high degree of homology with DPP-4 and angiotensin-converting enzyme 2 (ACE2) [29]. This may indicate that DPP-4 can facilitate SARS-CoV-2 entry into respiratory tract cells [29,30] and as such, DPP-4 inhibition could contribute to reducing the viral load and improving in ammatory and immune responses so as to prevent a cytokine storm, which can entail lung injury and multiple organ failure in COVID-19 [31,32].
A previous work from the SEMI-COVID-19 Network that analyzed 2,666 patients with T2DM admitted for COVID-19 did not nd any signi cant associations between at-home glucose-lowering drugs and mortality or other adverse outcomes [19]. However, the mean age of patients in that study was much younger than those in our cohort (75 vs 86 years), a fact that could explain the difference in results. Aging may be associated with DPP-4 overexpression; a signi cant correlation between membrane DPP-4 activity and animal age has been found in murine models [33]. It has also been reported that DPP-4 receptor levels rise on the senescent cell surface, suggesting that DPP-4 could play an important role in the aging process [34]. Aging is also characterized by a state of chronic low-grade in ammation (termed "in ammaging") that could predispose patients to experiencing a cytokine storm in COVID-19. Moreover, some preclinical evidence suggests that the anti-in ammatory effects of DPP-4 may be more intense in elderly patients [35]. Taken as a whole, these data could explain why more bene cial effects of DPP-4i are observed in elderly patients with T2DM and COVID-19 than in younger populations.
We found that previous treatment with ARB, but not with ACEI, in elderly patients with T2DM hospitalized for COVID-19 was associated with a lower risk of all-cause mortality, a nding not previously reported in this population. The role of renin-angiotensin-aldosterone system (RAAS) inhibitors in the COVID-19 has not been fully characterized. Given that both ACEI and ARB induce up-regulation of ACE2 [36], it has been hypothesized that these drugs could augment susceptibility to and severity of SARS-CoV-2 infection [37]. However, data from observational studies indicate that use of RAAS inhibitors in patients with COVID-19 is safe [17,38]. A large meta-analysis which analyzed 28,872 patients and examined critical events and mortality data on patients prescribed ACEI and ARB found that their chronic use, especially among hypertensive patients with COVID-19, had bene cial effects [39]. In view of the foregoing, continuing ACEI and ARB treatment in COVID-19 patients has been recommended [40].
In spite of its upregulation of ACE2, the potential bene ts of RAAS inhibitor treatment in COVID-19 could be explained through its enhancement of the ACE2/Ang1 − 7/Mas axis. It converts angiotensin II into Ang1 − 7, which has anti-in ammatory properties that preclude lung injury due to COVID-19 [41,42]. Aging is associated with an upregulation of the angiotensin II proin ammatory pathway as well as a decrease in ACE2 levels and this likely predisposes elderly diabetic individuals to more severe COVID-19 disease [43].
Since ARB act on the nal step of the RAAS system, blocking the AT1 receptor of angiotensin II, it has been postulated that they might be superior to ACEI in terms of improving COVID-19 prognosis [44]. In fact, better outcomes have been described in patients with COVID-19 and hypertension who receive ARB versus ACEI [11]. Until more evidence is available, it would be wise to prioritize the use of ARB over ACEI in this population.
This work also found that preadmission therapy with ASA was associated with increased in-hospital mortality in very old patients with diabetes hospitalized for COVID-19. One retrospective study concluded that ASA use may have protective effects on the lungs and reduce the need for mechanical ventilation, ICU admission, and in-hospital mortality in hospitalized COVID-19 patients [14]. This is likely related to the antithrombotic and anti-in ammatory properties of ASA. However, in that work, the mean age of ASAtreated patients was 62 years and only 55% had diabetes, so these results cannot be extrapolated to older diabetic patients.
The impact of ASA use on all-cause mortality in elderly patients is uncertain, as both favorable [45] and unfavorable [46] effects have been reported. In the ASPREE trial [46], which analyzed healthy patients ≥ 70 years of age, low-dose aspirin signi cantly increased the risk of major bleeding events and mortality. In light of this nding, the potential bene ts of ASA use must be weighed against the risk of hemorrhage and other adverse effects in elderly diabetic individuals, especially those without prior cardiovascular disease [47].
Finally, preadmission therapy with statins showed a neutral effect on mortality in our population. It has been postulated that statins could have a potential role as an adjunct therapy in COVID-19 to mitigate endothelial dysfunction and dysregulated in ammation in patients with COVID-19 [48]. A possible antiviral effect of statins has also been postulated [49].
A meta-analysis [12] and two observational studies [13,50] -one of them in diabetic population [13]-have reported a signi cant reduction in mortality in patients with COVID-19 who received statins before admission. In another study, in-hospital statin use was linked to a reduced risk of mortality in individuals with COVID- 19 [51]. Once again, the mean age of patients included in these studies was signi cantly younger than in our population, so this potential bene t may not exist in older patients.
Our study has several limitations. First, its observational design does not allow us to determine causal relationships. Additional randomized controlled trials are needed to evaluate confounding factors that were potentially overlooked in our study. Second, our study was conducted in very old patients with diabetes hospitalized with severe SARS-CoV-2 infection and thus its conclusions cannot be extrapolated to other populations with COVID-19. Third, we did not have data on the characteristics of patients' T2DM, such as glycemic control before hospitalization, duration of diabetes, blood glucose levels during hospitalization, or in-hospital anti-hyperglycemic management. Lastly, the data provided about at-home glucose-lowering drugs did not include information on treatment adherence or treatment duration.

Conclusions
We found important differences between cardiometabolic drugs and outcomes in older patients with T2DM hospitalized for COVID-19. Preadmission treatment with DPP-4i and ARB may reduce in-hospital mortality; other antidiabetic drugs, ACEI and statins seem to have a neutral effect; and ASA may be associated with excess mortality. These ndings, which could have important clinical implications, must be con rmed in further controlled trials. All patients gave their informed consent. When there were biosafety concerns and/or when the patient had already been discharged, verbal informed consent was requested and noted on the medical record.
Data con dentiality and patient anonymity were maintained at all times, in accordance with Spanish regulations on observational studies. This study was carried out in accordance with the Declaration of Helsinki and was approved by the Institutional Research Ethics Committee of Málaga on March 27, 2020 (Ethics Committe code: SEMI-COVID- 19 27-03-20), as per the guidelines of the Spanish Agency of Medicines and Medical Products.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analysed during this study are included in this published article.

Competing interest
Page 11 /19 The authors declare that they have no competing interests.   A multivariate logistic regression analysis were performed to assess the association between preadmission cardiometabolic therapy and in-hospital mortality. The regression analysis values were expressed as adjusted odds ratios with a 95% con dence interval. In order to select the variables, the forward selection Wald statistic was used. Discrimination of the tted logistic model was assessed via a receiver operating characteristic (ROC) curve. The Hosmer-Lemeshow test for logistic regression was used to determine the model's goodness of t. Values were considered to be statistically signi cant when p<0.05.