Phenotypic characteristics and prognosis of inpatients with COVID-19 and diabetes: the CORONADO study

Study oversight

The CORONADO study was launched in all French hospitals volunteering to share data on hospitalised COVID-19 patients with diabetes. The study was sponsored by CHU (centre hospitalier universitaire) Nantes, designed in accordance with the declaration of Helsinki and conducted in accordance with French legislation with approval obtained from the local ethics committee (Institutional Review Board/Institutional Ethics Committee – GNEDS [groupe nantais d'éthique dans le domaine de la santé]; Ref. CORONADOV2), the CEREES (comité d'expertise pour les recherches, les études et les évaluations dans le domaine de la santé; n° INDS [institut national des données de santé]:1544730) and the CNIL (commission nationale de l'informatique et des libertés; DR-2020-155/920129). In light of the purely non-interventional design of this observational study and the emergency situation related to the COVID-19 pandemic, the CNIL and the GNEDS have repealed the systematic collection of written informed consent. They recommended that we collect an ‘oral non-opposition to participate’ as far as possible (in particular by publishing study information via posters in the hospitals). Living patients who were unable to give consent on admission all received information about their inclusion in the CORONADO study before discharge, and therefore had a clear and free choice to object to the use of their clinical data. Any patient declining to participate in the study or expressing his or her opposition to data collection from hospital information systems, even after hospital discharge, was excluded from the study.

Study design and participants

The aim of the CORONADO study was to describe the phenotypic characteristics and prognosis of individuals admitted to hospital with COVID-19 between 10 March and 10 April 2020. Inclusion criteria were (1) hospitalisation in a dedicated COVID-19 unit with COVID-19 diagnosis confirmed biologically (by SARS-CoV-2 PCR test) and/or clinically/radiologically (i.e. as ground-glass opacity and/or crazy paving on chest computed tomography [CT] scan); (2) personal history of diabetes or newly diagnosed diabetes on admission (i.e. HbA_1c ≥48 mmol/mol [6.5%] during hospitalisation).

Owing to the rapid recruitment rate and in order to make clinically relevant findings available as quickly as possible, the scientific committee, on 5 April 2020, suggested a premature database lock on 18 April 2020 for participants admitted in the period 10–31 March 2020, and continuation of recruitment with no further modification. A first set of analyses was performed in 1317 participants selected according to the following criteria: (1) meeting the eligibility criteria; (2) available information on the main outcome, recorded on day 7 following admission; (3) available data on age and sex (see flowchart in Fig. 1).

Patient follow-up and clinical outcomes

The composite primary endpoint combined tracheal intubation for mechanical ventilation and death within 7 days of admission. Secondary outcomes included death on day 7, tracheal intubation on day 7, admission to ICUs and discharge on day 7. Participants discharged before day 7 were systematically contacted to check for the non-occurrence of these events on day 7.

Data collection

Data collection was performed by clinical research associates and physicians in participating centres. They were instructed to systematically review the medical files of all COVID-19 inpatients, select those with diabetes, extract data from their medical files and, if needed, contact the patient’s general and/or specialist practitioners, regular pharmacist or biomedical laboratory. Collected data included clinical data (age, sex, ethnicity, BMI), classification of diabetes as noted in the medical file by the physician in charge of the patient, duration of diabetes, recent glycaemic control (i.e. two most recent HbA_1c levels determined before admission), microvascular and macrovascular complications and comorbidities. HbA_1c considered in the analysis was determined locally in the 7 days following admission or, if not available, was the result of a routine determination in the previous 6 months. Microvascular complications were defined as severe diabetic retinopathy (proliferative retinopathy and/or laser photocoagulation and/or clinically significant macular oedema requiring laser and/or intra-vitreal injections) and/or diabetic kidney disease (proteinuria [AER ≥300 mg/24 h; urinary albumin/creatinine ratio ≥300 mg/g; urinary albumin/creatinine ratio ≥30 mg/mmol creatinine; proteinuria ≥500 mg/24 h] and/or eGFR equal to or lower than 60 ml min⁻¹ [1.73 m]⁻², using the Chronic Kidney Disease Epidemiology Collaboration [CKD-EPI] formula) and/or history of diabetic foot ulcer. Macrovascular complications were defined as ischaemic heart disease (acute coronary syndrome and/or coronary artery revascularisation) and/or cerebrovascular disease (stroke and/or transient ischaemic attack) and/or peripheral artery disease (amputation owing to ischaemic disease and/or lower limb artery revascularisation). In addition, COVID-19-related clinical, radiological and biological characteristics were collected at admission as well as their clinical evolution during the hospital stay.

Statistical analysis

Quantitative data were expressed as mean ± SD or median [25th –75th percentile]. Categorical variables were given as number (percentage) of participants. As prespecified in the protocol, the main objective of the study was descriptive. We calculated that a population size of 300 participants with an attrition of 20% for missing data, and a percentage of 16% of our main outcome, would give us a 95% CI equal to 11.7–21.1% using the Clopper–Pearson estimate.

Univariate logistic regression models were used to calculate OR associated with primary outcome or death on day 7. Natural log transformation was consistently considered in cases of skewed distribution, which applied to BMI and biological features.

Age- and sex-adjusted ORs for the primary outcome were plotted for BMI, HbA_1c and admission plasma glucose using degree 2 fractional polynomial approaches [16].

Multivariable logistic regression models were used to separately assess the association of the primary outcome and death on day 7 with clinical and biological features. A standardisation process was also applied using z scores for the purpose of direct comparison. In our initial statistical analysis plan, four covariates were systematically forced in the models: age, sex, BMI and HbA_1c. However, since HbA_1c did not contribute to the risk of either the primary outcome or death on day 7, and owing to a significant number of missing data for HbA_1c and BMI, our multivariable models ultimately took only age and sex into account. Other variables were considered only if associated with the main outcome in univariate analysis (threshold: two-sided p value ≤0.10) and selected in the final model after a stepwise backward/forward selection process. In the event of obvious collinearity (such as alanine aminotransferase [ALT] with aspartate aminotransferase [AST], or white cell count with lymphocyte count), only the variable associated with the smaller p value was considered for multivariable analysis. In the final model, interactions were checked between all pairs of covariates.

We built two distinct multivariable models both separately for the main outcome and for the risk of death: (1) the first included covariates related to patient history prior to admission (chronic diabetes complications and other comorbidities) and routine medications; (2) the second included covariates related to medical presentation on admission, such as COVID-19 symptoms and biological determinations. This corresponds to the situation of a physician in an emergency room or department, assessing the prognosis of his/her patient.

All statistical tests were two-sided with a type 1 error set at 5%. All analyses were performed on available data, without imputation, and using statistical software R version 3.6.2 (https://cran.r-project.org/bin/windows/base/old/3.6.2/).

Population and clinical outcomes

The present analysis focused on 1317 participants with diabetes and confirmed COVID-19 admitted to 53 French hospitals during the period 10–31 March 2020.

A total of 382 patients (29.0%; 95% CI 26.6, 31.5) met the primary outcome. Overall, 410 patients (31.1%; 95% CI 28.6, 33.7) were admitted to ICUs within 7 days of hospital admission, including 267 individuals who required tracheal intubation for mechanical ventilation (20.3%; 95% CI 18.1, 22.5). One hundred and forty deaths (10.6%; 95% CI 9.0, 12.4) were recorded on day 7. In contrast, 237 participants (18.0%; 95% CI 16.0, 20.2) were discharged on day 7 (see flowchart in Fig. 1).

Demographic and diabetes-related characteristics

The clinical characteristics of the whole population are shown in Table 1. Mean (± SD) age was 69.8 ± 13.0 years and 64.9% were men. The classification of diabetes cases mainly included type 2 diabetes (88.5%), and less frequently type 1 diabetes (3.0%) or other aetiologies (5.4%). In addition, 3.1% of the participants were newly diagnosed with diabetes on admission (HbA_1c ≥48 mmol/mol [6.5%]). The median BMI was 28.4 (25th–75th percentile 25.0–32.7) kg/m². The mean HbA_1c value was 65 ± 21 mmol/mol (8.1 ± 1.9%). A medical history of hypertension and dyslipidaemia were found in 77.2% and 51.0% of the participants, respectively. Microvascular and macrovascular complications were reported in 46.8% and 40.8% of individuals, respectively. Regarding routine glucose-lowering medications, 38.3% of the participants were on insulin therapy while 56.6% received metformin and 21.6% dipeptidyl peptidase 4 (DPP-4) inhibitors. Moreover, treatment with renin–angiotensin–aldosterone system (RAAS) blockers (ACE inhibitors and/or angiotensin II receptor blockers [ARBs] and/or mineralocorticoid-receptor antagonists [MRAs]) and statins was used by 57.1% and 47.6% of the participants, respectively.

Characteristics of COVID-19 on admission

Characteristics of COVID-19 on admission are provided in Table 2. The median duration of symptoms before admission was 5 days (25th–75th percentile, 2–8 days). As expected, the most common signs were fever (77.9%), cough (68.7%), fatigue (62.4%), dyspnoea (61.8%) and digestive disorders (34.5%). SARS-CoV-2 PCR testing was performed in 1268 participants, with a positive result in 96.8%. Thoracic CT imaging demonstrated typical ground-glass opacity and/or crazy paving in 818 individuals (90.0%). Biological findings were consistent with obvious infection as illustrated by a median C-reactive protein (CRP) at 77.8 (38.4–132.7) mg/l. Median plasma glucose at admission was 9.20 (6.80–12.62) mmol/l.

Of interest, diabetes-related disorders were reported in 11.1% of the participants on admission with 132 episodes of severe hyperglycaemia, including 40 of ketosis, of which 19 were ketoacidosis, as well as 14 hypoglycaemic events, while severe anorexia was reported in 83 participants (6.3%).

Factors prior to admission associated with study outcomes

In univariate analysis considering the primary outcome, male sex was more frequent (69.1% vs 63.2%, p = 0.0420) and BMI was significantly higher (median 29.1 [25.9–33.6] vs 28.1 [24.8–32.0] kg/m², p = 0.0009) in patients who met the primary outcome compared with the others, as was the use of RAAS blockers (61.5% vs 55.3%, p = 0.0386) (Table 1 and electronic supplementary material [ESM] Table 1).

Furthermore, several characteristics prior to admission were associated with the risk of death on day 7 including age, hypertension, micro- and macrovascular diabetic complications and comorbidities such as heart failure or treated obstructive sleep apnoea (OSA). Among prior medications, metformin use was lower in people who died. In contrast, insulin therapy, RAAS blockers, β-blockers, loop diuretics and MRAs were found to be associated with death on day 7 (Table 1 and ESM Table 1).

When using age- and sex-adjusted nonlinear models, BMI was significantly and positively associated with the primary outcome (p = 0.0001) but not with death on day 7 (p = 0.1488) (Fig. 2). In contrast, HbA_1c level was neither associated with the primary outcome nor with death on day 7.

Multivariable analyses were then conducted with characteristics prior to admission. BMI remained associated with the primary outcome in a model where sex and age were forced into models. When comorbidities and routine treatment were entered in an adjusted model with stepwise selection, BMI was the only independent factor associated with the primary outcome, with an adjusted OR of 1.28 (95% CI 1.10, 1.47) (Table 3). Finally, age, history of microvascular or macrovascular complications, and treated OSA were found to be independently associated with the risk of death on day 7 (Table 4). A sensitivity analysis conducted only in patients with a positive SARS-CoV-2 PCR test found similar results for both the primary outcome and death (data not shown).

Factors on admission associated with study outcomes

Regarding COVID-19 symptoms on admission, dyspnoea was positively associated with the primary outcome and with death on day 7, whereas cephalalgia, upper respiratory tract symptoms (rhinitis and/or pharyngeal symptoms), and ageusia/anosmia, as well as time between symptom onset and admission, were negatively associated with death on day 7 (Table 2 and ESM Table 2). Several biological parameters reflecting the severity of the infection were also associated with both primary outcome and death on day 7, such as CRP, creatine phosphokinase (CPK) and lymphopaenia. Reduced kidney function, assessed by admission eGFR, and increased plasma AST level, was associated with both outcomes (Table 2 and ESM Table 2). In age- and sex-adjusted nonlinear models, admission plasma glucose was significantly and positively associated with the primary outcome (p = 0.0001) and with death on day 7 (p = 0.0059) (Fig. 2).

The prognostic value of characteristics on admission was finally investigated in multivariable models (Table 5). Among clinical symptoms, dyspnoea was the only predictor of the primary outcome. Regarding biological parameters, lymphopaenia on admission was independently associated with the primary outcome, as were increased AST and CRP concentrations. Dyspnoea and plasma levels of both AST and CRP were also independently associated with the risk of death on day 7, as well as decrease in platelet count and eGFR (Table 6). Once again, similar results were obtained for both the primary outcome and death on day 7 when considering only the patients with a positive PCR test (data not shown).