Prognostic Factors for COVID-19 Hospitalized Patients with Preexisting Type 2 Diabetes

Background Type 2 diabetes (T2D) as a worldwide chronic disease combined with the COVID-19 pandemic prompts the need for improving the management of hospitalized COVID-19 patients with preexisting T2D to reduce complications and the risk of death. This study aimed to identify clinical factors associated with COVID-19 outcomes specifically targeted at T2D patients and build an individualized risk prediction nomogram for risk stratification and early clinical intervention to reduce mortality. Methods In this retrospective study, the clinical characteristics of 382 confirmed COVID-19 patients, consisting of 108 with and 274 without preexisting T2D, from January 8 to March 7, 2020, in Tianyou Hospital in Wuhan, China, were collected and analyzed. Univariate and multivariate Cox regression models were performed to identify specific clinical factors associated with mortality of COVID-19 patients with T2D. An individualized risk prediction nomogram was developed and evaluated by discrimination and calibration. Results Nearly 15% (16/108) of hospitalized COVID-19 patients with T2D died. Twelve risk factors predictive of mortality were identified. Older age (HR = 1.076, 95% CI = 1.014–1.143, p=0.016), elevated glucose level (HR = 1.153, 95% CI = 1.038–1.28, p=0.0079), increased serum amyloid A (SAA) (HR = 1.007, 95% CI = 1.001–1.014, p=0.022), diabetes treatment with only oral diabetes medication (HR = 0.152, 95%CI = 0.032–0.73, p=0.0036), and oral medication plus insulin (HR = 0.095, 95%CI = 0.019–0.462, p=0.019) were independent prognostic factors. A nomogram based on these prognostic factors was built for early prediction of 7-day, 14-day, and 21-day survival of diabetes patients. High concordance index (C-index) was achieved, and the calibration curves showed the model had good prediction ability within three weeks of COVID-19 onset. Conclusions By incorporating specific prognostic factors, this study provided a user-friendly graphical risk prediction tool for clinicians to quickly identify high-risk T2D patients hospitalized for COVID-19.


Introduction
Coronavirus disease 2019 (COVID- 19), which is caused by a novel betacoronavirus (named severe acute respiratory syndrome, coronavirus 2, or SARS-CoV-2), has been declared a global pandemic by the World Health Organization [1,2]. Studies report that 20-50% of COVID-19 patients have diabetes depending on different areas [3], and COVID-19 patients with diabetes have worse outcomes than nondiabetic patients [4][5][6][7]. Diabetes as a chronic disease affects more than 463 million people worldwide [8,9]. Its high prevalence combined with the poor COVID-19 clinical outcomes prompts the need for improving the management of diabetic patients to reduce complications and the risk of death.
Type 2 diabetes (T2D) accounts for 90-95% of all diabetes globally [10]. T2D patients may deteriorate more rapidly once infected due to the weakened immune system [1,11]. e COVID-19 pandemic represents a serious threat to this large vulnerable population. In this study, we identified clinical factors associated with COVID-19 outcomes specifically targeted at T2D patients and built a nomogram for clinicians to quickly identify high-risk T2D patients hospitalized for COVID-19 and for patients to easily do self-monitoring through risk assessment.

Study Design and Participants.
A retrospective cohort study was conducted of sequentially hospitalized COVID-19 patients from January 8 to March 7, 2020 in Tianyou Hospital, affiliated to the Wuhan University of Science and Technology in China. e final follow-up date was March 18, 2020. Patients were included based on the following criteria: (1) age ≥18 years, (2) chest computer tomography (CT) scans were performed at the time of admission and discharge, (3) blood glucose and other routine laboratory tests were performed on admission, and (4) treatments, complications, and definitive clinical outcomes (discharged or died) were clearly recorded between admission and the final follow-up date. is study was approved by the institutional ethics board of Tianyou Hospital, affiliated to the Wuhan University of Science and Technology (Ethical application no.: WKDTY-2020039).

Date Collection.
Demographics, symptoms, comorbidities, laboratory tests, treatments, disease progression, and outcomes of COVID-19 patients with and without T2D were collected from electronic medical records (EMR). Information about age of diabetes diagnosis, duration of diabetes, and diabetes treatment was collected from diabetic patients. All raw data were independently verified by two experienced physician teams to guarantee accuracy. e endpoint was inhospital death of COVID-19 patients.

Case Definition.
All patients were diagnosed with COVID-19 according to the WHO interim guidance [12]. Patients with SARS-CoV-2 infection can experience a range of clinical manifestations, and a patient's clinical status may change over time. Based on the National Institutes of Health (US) Coronavirus Disease 2019 (COVID-19) Treatment Guidelines [13], the patients were grouped into the following severity of illness categories according to their clinical presentation on admission: nonsevere (patients who showed no symptoms or fever, respiratory symptoms, and CT evidence of pneumonia); severe (patients who showed features of nonsevere patients as well as respiratory distress with respiratory rate ≥30 breaths/min, oxygen saturation less than 94% on room air at sea level, arterial partial pressure of oxygen/oxygen concentration less than 300 mmHg, or lung infiltrates >50%); and critical (patients who showed respiratory failure requiring ventilatory support as well as shock and organ dysfunction requiring intensive care and/or multiple organ dysfunction). T2D status was designated based on the patient's medical history and defined according to the guidelines of Chinese Diabetes Society [14]. All T2D patients took diabetes medication regularly in the past year and did not show obvious symptoms of acute cardiac injury on admission. Acute cardiac injury was defined as electrocardiographic and echocardiographic abnormalities or by cardiac biomarker (hypersensitive troponin I or creatine kinase) elevation above 99% of the upper reference limit [15]. Disease favorable outcome was defined as full recovery and discharge, progression from critical/severe to nonsevere disease status, and/or maintenance of nonsevere status; conversely, unfavorable outcomes included death, progression from nonsevere to severe/critical or severe to critical, and/or maintenance of severe or critical status.

Statistical Analysis.
Categorical variables were summarized as numbers with percentages, and continuous variables were presented as medians with interquartile range (IQR). Student's t-test or Wilcoxon rank-sum test was used for continuous variables. e chi-square test or Fisher exact test was used for categorical variables as appropriate. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated by univariate and multivariate Cox proportional hazard regression models. A final model selection was performed via a backward stepwise selection process. Timedependent receiver operating characteristic curve (ROC) and the calculated time-dependent area under the curve (AUC) were used to characterize the discrimination potential of the Cox regression model [16,17]. A predictive nomogram was developed to generate a combined indicator for estimating the mortality risk, and it was validated by bootstrap resampling 1,000 times. Discrimination was used to evaluate the model's ability to separate patients with different outcomes and quantified using Harrell's concordance index (C-index) [18,19]. Calibration was used to test how close the predictions were to the actual outcomes by demonstrating calibration curves.
If the number of events was too small to calculate the HR, the variables were excluded. Cumulative rates of in-hospital deaths were determined using the Kaplan-Meier method and Log-rank test.
e survival time was defined as the interval from the date of admission to the date of death. e analyses were performed by using R software (version 3.6.1, R Foundation, Vienna, Austria) or Statistical Analysis System (SAS) software (version 9.4, SAS Institute Inc., Cary, NC), with a statistical significance set at two-sided p < 0.05.

Participants and Clinical Outcomes.
e initial 539 confirmed COVID-19 patients were identified from January 8 to March 7, 2020 in Tianyou Hospital in Wuhan, China. A total of 157 patients were excluded due to incomplete or duplicate medical records (72 patients missed key clinical information, 11 patients were duplicated, 65 patients were transferred to other hospitals, and 9 patients were discharged within 24 hours), leaving 382 patients with a definitive clinical outcome (discharged or died) in this study, and 108 of them had T2D. e median age was 63 (range: 23- Figure 1.

Comparison of Clinical Features between COVID-19
Patients with and without T2D. e death rate among the 108 patients with T2D (14.8%) was two times higher than that among the 274 patients without T2D (7.7%). In Table S1, the complete clinical and biochemical panel provides the comparison of both COVID-19 patients with and without T2D, delineating the possible indicators predisposing patients with T2D to have worse clinical outcomes. e significant comparison results of COVID-19 patients with and without T2D were summarized in Table 1.
e median ages of the diabetes and nondiabetes groups were 68 and 60 years, respectively. Moreover, 41.7% of diabetic patients were unemployed and only 12.0% were employed. Hypnotics were more commonly administered to nondiabetic patients (28.6%) than to diabetic patients (16.2%). Fever and cough were the most common symptoms, with the former being more common in the nondiabetes group (85.4% vs. 66.7%) and the latter more common in the diabetes group (62.0% vs. 49.3%). e diabetes group had a significantly higher rate of preexisting comorbidities (63.9%, p < 0.0001), with hypertension in 60.2% and cardiovascular disease in 19.4%. As for disease severity, more than half of the diabetes patients showed severe disease, and 17.6% were critically ill on admission, compared to nondiabetes patients (5.1%).
During hospitalization, all T2D patients received diabetes treatment, either only oral medication, oral medication plus insulin, or only insulin. e clinical outcomes were significantly different between the three diabetes treatment groups (p < 0.0001). In the nonsurvivor group, most (75.0%) were treated with only insulin, and only two patients died with oral medication only and oral medication plus insulin treatment. e majority of the survivor group (68.5%) were treated by oral medication alone, with 17.4% of them using insulin only.

Development and Validation of Individualized Prediction
Nomogram.
e nomogram was constructed to predict the mortality risk of COVID-19 patients with T2D according to the final multivariate analysis. Each independent factor was assigned a score based on the point scale at the top of the nomogram, and the total points were calculated. Patient 7-, 14-, and 21-day mortality probabilities were obtained from the bottom point scale of the nomogram (Figure 3(a)). e C-index of the nomogram was 0.90 (95% CI, 0.84-0.96) and a high C-index value of 0.87 was reached in the internal validation assessed by bootstrapping (1,000 replicates), suggesting that the prediction of the nomogram was consistent with the actual observation for COVID - (Figure 3(b)).

Potential Benefits and Detriments of Diabetes Medication.
e diabetes treatments for COVID-19 patients with T2D during hospitalization were counted: 28 (25.9%) only insulin, 15 (13.9%) one or more oral medications, and 65 (60.2%) oral medications plus insulin. Based on the multivariate Cox model, COVID-19 patients with diabetes treated with oral medication only or with oral medication plus insulin had a significant lower mortality risk than those treated with insulin only. e survival curves for different diabetes treatments are shown in Figure S2. Considering the different disease severity of COVID-19 on admission may have a different reaction to diabetes treatment, COVID-19 patients with T2D were analyzed according to clinical disease severity to further validate the effect of diabetes medication on clinical outcomes. In the nonsevere and severe disease groups, favorable outcomes were more prevalent among COVID-19 patients receiving oral medication only or oral medication plus insulin than among the insulin-only group (100%, 100% vs. 60% for nonsevere, p � 0.020; 97.4%, 100% vs. 73.3% for severe, p � 0.034) (Figure 4(a)). In the critical disease group, favorable outcomes were still more prevalent among COVID-19 patients with oral medication only and oral medication plus insulin but was not statistically significant (83.3%, 60% vs. 25% for critical, p � 0.090) (Figure 4(a)). We also compared the blood glucose level of different diabetes treatment groups. As show in Figure S3, the blood glucose level of patients treated with oral medication plus insulin was higher than that of patients treated with insulin only. e insulin-only group had significantly increased acute cardiac injury (p � 0.0004) (Figure 4(b)).

Discussion
In this study, we identified prognostic factors in patients with T2D hospitalized for COVID-19 and developed a nomogram for predicting the individual mortality risk for this special patient group.
In our study, we found the death rate among the patients with T2D was twice as high as that among the patients without T2D which might be related to the higher prevalence of coexisting comorbidities and older age in T2D patients. ese findings are consistent with previous reports that older age and comorbidities are contributors to increased risk of COVID-19 susceptibility, disease severity, and clinical outcomes [4,[20][21][22][23][24]. Comorbidities could be considered a consequence of diabetes and have a synergistic effect with age on the mortality risk of patients with T2D. e blood glucose level has been reported to be an important prognostic factor for COVID-19 progression and fatality from a large retroactive study in China [25]. High glucose level contributes to acute respiratory distress syndrome risk in COVID-19 patients and other comorbidities. One study showed 1 mmol/L increase in plasma glucose level could increase 6% risk of hospitalization for pneumonia [26].
e elevated blood glucose could also facilitate the entire process of the COVID-19 infection [27]. COVID-19 patients with T2D have a higher blood glucose level than those without T2D leading to the severe adverse clinical outcome.
As identified in our model, oral diabetes medication may offer a protective impact on COVID-19.
is may be achieved through the medication's action on the receptor of SARS-CoV-2. It is worthwhile to explore the interactions of SARS-CoV-2 with commonly used diabetes medications, with the potential to repurpose them to prevent COVID-19 disease progression. Angiotensin-converting enzyme 2 (ACE2) is the principal cellular receptor for the novel SARS-CoV-2, mainly in cell membranes of lung alveolar epithelial cells and enterocytes of the small intestine, which bind with the receptor-binding domain (RBD) of the spike protein (S) on the virus surface [28][29][30][31]. In diabetes, ACE2 has been suggested as a potential therapeutic target for the management of diabetes and its complications [32,33] and reduction of the risk of type 2 diabetes development [34]. Metformin may have protective effects against SARS-CoV-2 by working with ACE2 in two ways [35,36]. First, metformin can activate the AMP-activated protein kinase (AMPK) signaling pathway, leading to phosphorylation of ACE2. eoretically, conformational and structural changes in ACE2 may lead to decreased binding to SARS-CoV-2 RBD due to steric hindrance by the addition of a phosphate group [37]. Second, AMPK phosphorylation of ACE2 enhances ACE2 stability, which may prevent SARS-CoV-2mediated ACE2 downregulation [38,39]. us, metformin may decrease the ACE2 binding ability of SARS-CoV-2 as well as increase the expression and stability of ACE2 through activating AMPK. Metformin can also be involved in the development of COVID-19 by inhibiting the mTOR signaling pathway [40,41] and altering the composition of the gut microbiota [42]. A recent study showed users of metformin had few adverse outcomes compared to nonusers [43]. Continued use of metformin reduced the risk of death or ICU admission during hospitalization [44].
Previous studies indicated that insulin therapy may be associated with increased cardiovascular events and mortality [45,46].Treatment with regular insulin may result in postmeal hyperglycemia and an increased risk of late postprandial hypoglycemia. Hypoglycemia is associated with an increased risk of cardiac arrhythmia due to a mismatch between insulin and carbohydrate intake, alcohol, or exercise [47]. Additionally, hypoglycemia can stimulate the catecholamine release and prolong the QT interval, which can lead to adverse cardiovascular events [48]. e use of insulin, therefore, should be closely monitored in COVID-19 patients with T2D.
SARS-CoV-2 infection can deregulate the immune response, which may trigger viral hyperinflammation in patients with severe COVID-19 [49] and increase the risk for SAA amyloid formation and subsequent pathologies [50]. SAA is a nonspecific acute phase protein mainly produced by the cytokines IL-1β, IL-6, and TNF-α in liver cells, which increase in chronic inflammatory processes, as in diabetes and obesity. Previous studies showed that patients with severe acute respiratory syndrome had significantly increased levels of SAA, suggesting SAA could be used as a biomarker to monitor the progression of respiratory diseases [51]. Recent studies indicated that COVID-19 patients with significantly increased SAA levels had poor prognosis [7,52], and SAA could be an excellent biomarker in discrimination between moderate and severe COVID-19 infection [53]. Meta-analysis also showed the severe COVID-19 patients had markedly higher SAA levels [54,55]. High mortality risk of COVID-19 patients with diabetes may be due to the acute increase of SAA after virus infection. In summary, our study provided a user-friendly graphical nomogram to help risk assessment for COVID-19 patients with T2D. By incorporating the prognostic factors of age, SAA, blood glucose, and diabetes treatment, this scoring tool could predict the probability of death at 7, 14, and 21 days. e potential protective effects of oral diabetes medication in COVID-19 highlighted a potential focus of future research.
Data Availability e data can be provided after the article is published. e corresponding author (dengy@hawaii.edu) has the right to decide whether to share the data or not based on the research objectives and the plan provided.

Conflicts of Interest
e authors declare no conflicts of interest.

Authors' Contributions
YD and LH conceived and supervised the study. LH, QSZ, CS, YM, LW, JJH, and CZW collected the epidemiological and clinical data. LH, YZ, SC, and HWR contributed to radiological figure interpretation. YF and ZW processed and conducted statistical data analyses. YF, ZW, and RY drafted the manuscript. All the authors reviewed and approved the final version for publication. YD and YF are responsible for the integrity of the data and the accuracy of the analyzed data.