key: cord-0946211-1768lp0a
authors: Lazarus, G.; Audrey, J.; Wangsaputra, V. K.; Tamara, A.; Tahapary, D. L.
title: High admission blood glucose is an independent risk factor of poor prognosis in COVID-19: A systematic review and dose-response meta-analysis
date: 2020-09-27
journal: nan
DOI: 10.1101/2020.09.25.20200774
sha: 81431252977127eef50cb9f1516e8a395364b96b
doc_id: 946211
cord_uid: 1768lp0a

Aims To investigate the prognostic value of admission blood glucose (BG) in predicting COVID-19 outcomes, including poor composite outcomes (mortality/severity), mortality, and severity. Materials and methods Eligible studies evaluating the prognostic value of fasting BG (FBG) and random BG (RBG) levels in predicting COVID-19 outcomes were included and assessed for risk of bias with the Quality in Prognosis Studies tool. Random-effects high-vs-low meta-analysis followed by dose-response analysis using generalized least squares model in a two-stage random-effects meta-analysis were conducted. Potential non-linear association was explored using restricted cubic splines and pooled using restricted maximum likelihood model in a multivariate meta-analysis. Results The search yielded 35 studies involving a total of 14,502 patients. We discovered independent association between admission FBG and poor prognosis in COVID-19 patients. Furthermore, we demonstrated non-linear relationship between admission FBG and severity (Pnon-linearity<0.001), where each 1 mmol/L increase augmented the risk of COVID-19 severity by 33% (risk ratio 1.33 [95% CI: 1.26-1.40]). Albeit exhibiting similar trends, study scarcity limited the strength of evidence on the independent prognostic value of admission RBG. GRADE assessment yielded high-quality evidence for the association between admission FBG and COVID-19 severity, and moderate-quality evidence for its association with mortality and poor outcomes, while the other assessments yielded very low-to-low quality. Conclusion High level of FBG at admission was independently associated with poor prognosis in COVID-19 patients. Further researches to confirm the observed prognostic value of admission RBG and to ascertain the estimated dose-response risk between admission FBG and on COVID-19 severity are required.

The rapidly spreading coronavirus disease 2019 has placed significant burdens on healthcare systems worldwide, with millions of cases and hundred-thousands of deaths. 1 Despite significant efforts in comprehending the disease, its diagnosis and prognostication remain challenging, attributing to the prevalent non-specific and atypical symptoms. 2 In light of this, recent reports have indicated that admission blood glucose (BG) may yield prognostic values in predicting COVID-19 outcomes. 3, 4 Nevertheless, to the best of our knowledge, no meta-analysis has evaluated the prognostic value of admission BG level in predicting the outcomes of COVID-19 patients. Hence, this review intends to summarize the current knowledge regarding the role of admission BG level in determining the prognosis of COVID-19 patients.

This systematic review was conducted in accordance with the guideline recommended by the Cochrane Prognosis Methods Group 5 and reported based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 6 . A detailed protocol has been previously registered in PROSPERO (CRD42020154772 7 ).

Two independent investigators performed thorough literature searches, with discrepancies resolved by a third investigator in a blinded fashion. Searches were conducted through PubMed, Ovid EMBASE, CENTRAL, EBSCO MEDLINE, CINAHL, Scopus, and WHO COVID-19 database for studies published up to 8 September 2020. Grey literature (e.g., preprints) search and literature snowballing of references were also performed. No language restrictions were applied. Details on databases and keywords are further elaborated in Appendix Table S1 .

Inclusion criteria were set to filter primary studies investigating the association between admission BG level and poor outcomes among COVID-19 patients (see Appendix Table S2 ).

Admission BG level was determined from the first BG measurement following patients' . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09. 25.20200774 doi: medRxiv preprint admission to hospital prior to any intervention, while poor outcomes were further dichotomized into mortality and severity. 8 Conversely, studies were excluded if any of the following criteria were met: (1) case reports, case series, or letter to editors; (2) irretrievable full-text articles; or (3) non-English articles.

Data extraction and risk of bias assessment were performed by two independent reviewers using a pre-specified form, with discrepancies resolved by the consensus with an independent third investigator. Details on the data items and handling are further discussed in Appendix pg. 7 . The main outcome of interest in this review was the risk of poor composite outcomes (i.e. mortality/severity), mortality, and severity among COVID-19 patients. Whenever possible, outcomes on severity were further investigated per following criterion, including invasive ventilation, intensive care unit (ICU) admission, acute respiratory distress syndrome (ARDS), and shock. 8 Included studies were further assessed for methodological quality by using the Quality in Prognosis Studies tool 9 and subsequently judged to be yielding low, moderate, or high risk of bias (Appendix Table S3 ). Lastly, the certainty of the evidence was evaluated using the modified Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework, where the quality of evidence was regarded as high, moderate, low, or very low. 10

Analyses were performed for both adjusted and unadjusted estimates; however, adjusted estimates were primarily utilized for reporting and interpretation of results. 11 Pooled effects were converted to and presented in risk ratios (RRs; see Appendix pg. 11). Quantitative synthesis was first conducted by comparing the highest vs lowest categories of exposures by using the generic inverse variance method with the DerSimonian-Laird random-effects model 12 . Whenever appropriate (n10), potential publication bias was evaluated visually by contour-enhanced funnel plot 13 and quantitatively by Egger's 14 and Begg's 15 tests. Statistical heterogeneity was investigated with Cochran's Q test and I 2 statistics. Dose-response meta-analysis (DRMA) was conducted only for adjusted outcomes. Study-specific linear trend was estimated using the generalized least squares method and pooled using the two-stage random effects meta-analysis.

Meanwhile, potential non-linear dose-response trend was evaluated using restricted cubic splines with three-knots model at the 10 th , 50 th , and 90 th percentiles, and subsequently pooled using the . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint restricted maximum likelihood method in a multivariate random-effects meta-analysis. The Wald test was used to assess for non-linearity.

A priori, we determined subgroup and sensitivity analyses only for adjusted results.

Whenever available, subgroup analyses were carried out based on study design, location, sample size, risk of bias, number of categories, effect size type, and diabetic status. On the other hand, sensitivity analyses were conducted by leave-one-out analysis and the exclusion of studies with high-risk of bias. For DRMA, subset analysis was performed according to diabetic subgroups, while sensitivity analyses were conducted by assigning alternative approaches for open-ended categories (for linear trends) and alternative knots locations (for non-linear trends). Metaanalysis was conducted with R ver. 4.0.0 (R Foundation for Statistical Computing, Vienna, Austria) 16 , and additional analyses with MetaXL software ver.5.3. (EpiGear International, Queensland, Australia) 17 . The significance level was set at 5% for all analyses. Further details on DRMA and additional analyses are discussed in Appendix pg. 12-13.

The details on the literature search process are summarized on Figure 1 . The initial search yielded 1177 articles, of which 636 were deduplicated and 482 were excluded following title and abstracts screening, resulting in the retrieval of 59 records for full-text assessments-among which 13 inappropriate design, five inappropriate settings, four incompatible language, two irretrievable full-text articles, and one unidentifiable setting (see Appendix pg. 12 Table S4 ). In quantitative analysis, 10 studies were excluded as seven [18] [19] [20] [21] [22] [23] [24] only reported P-value and three [25] [26] [27] reported different effect measures.

From 35 included studies, more than half were conducted in China (23 studies), while the others were five each in America 18, 21, [27] [28] [29] and Europe 20, 25, [30] [31] [32] , and one each in Hong Kong 33 and South Korea 34 . FBG was utilized in 23 studies, while RBG in 13 studies. Bias assessment . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint revealed a predominant low-to-moderate risk of bias (16 and 12 studies, respectively). Most of the studies yielded unclear risk of bias in study participation and confounding domains (Figure 2 and Appendix Figure S1 ), which may partly be explained by the fact that all but two studies 21, 29 were conducted retrospectively.

The summary of adjusted and unadjusted pooled effects of high vs low meta-analysis are consecutively listed in Table 1 and Appendix Table S5 ; and the certainty of evidence as assessed with the GRADE approach is summarized in Appendix Table S6 . GRADE assessments of the prognostic value of FBG resulted in high-quality evidence for severity and moderate-quality evidence for mortality and poor outcome, whereas the remaining domains yielded very low-to-low-quality evidence. Overlapping populations were observed in four studies [35] [36] [37] [38] (Appendix Table S7 ), and analyses were prioritized to Wang et al. 35 due to larger sample size.

We demonstrated that FBG was independently associated with poor prognosis in COVID-19 patients (Table 1 and Figure 3A -C), although all models yielded substantial heterogeneity (I 2 =84% for poor outcome; I 2 =87% for mortality, I 2 =78% for severity; all with P heterogeneity <0.001). Subgroup analyses based on study design and location were not possible as all studies were retrospective and conducted in China. Furthermore, publication bias assessment was only eligible for mortality outcome.

We discovered that high admission FBG increased the risk of poor outcomes by 20% (RR 1.20 [95% CI: 1.04-1.39]; Figure 3A ). However, we were unable to establish firm evidence as the observed heterogeneity remained unexplained (see footnote in Table 1 ) and the observed effects were diminished following the exclusion of Fadini et al. 32 or Li et al. 38 (Appendix Figure   S2A ). In DRMA comprising of two studies 32,39 , we failed to observe exposure-response gradient (RR 1.23 [95% CI: 0.90-1.68]; P heterogeneity =0.008; Figure 4A ), although study-specific slopes indicated that such trend exists. Considering this, we deemed the quality of evidence to be moderate for qualitative assessment and low for quantitative assessment.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. Figure 3B ); although the observed heterogeneity remained unexplained following subgroup analysis, except when the analysis was stratified according to its category (i.e. I 2 =2%

for studies reporting two categories). Furthermore, we discovered that the smaller-sized studies yielded excessively wide CIs as compared to larger studies (Table 1) . Nonetheless, sensitivity analysis suggested that these small studies did not contribute much to the pooled estimate as our findings remained consistent, except when Fadini et al. 32 or Chang et al. 34 were excluded, which exaggerated the pooled estimates (Appendix Figure S2B ). We detected asymmetry in the funnel plot which was further ascertained by Egger's test (P<0.001; Appendix Figure S3 ), although Begg's test indicated otherwise (P=0.245), suggesting the presence of publication bias.

Nevertheless, we did not downgrade the outcome for publication bias upon qualitative assessment as most studies adequately adjusted for potential confounders (Appendix Table S8 ).

DRMA for mortality outcome included three studies 32, 35, 39 as overlapping populations were detected in two studies. 35, 38 We were also unable to observe any exposure-response gradients although most studies reported otherwise 35, 38, 39 . However, when analysis was conducted only for non-diabetic patients, we observed an elevated mortality risk of about 38% per 1 mmol/L increase (RR 1.38 [95% CI: 1.21-1.57]; Appendix Figure S4 ). Despite this, we decided to upgrade the certainty of evidence for exposure-response gradient only for qualitative summary, resulting in the judgment of evidence quality for qualitative and quantitative analysis to be moderate and very low, respectively. This implied that we were confident that admission FBG was independently associated with mortality in COVID-19 patients, but the interpretation of the pooled estimate should be made with caution. Non-linear DRMA for poor outcome and mortality were not conducted since only one study reported three or more categories for each outcome. 35, 39 With regards to outcomes on severity, we revealed that high admission FBG level increased the patient's risk of developing severe COVID-19 infection by more than three-folds (RR 3.65 [95% CI: 2.31-5.75]). Although considerable heterogeneity was observed (I 2 =78%, P heterogeneity <0.001), subgroup analysis according to risk of bias explained all heterogeneity (Table 1) , suggesting that studies with moderate-to-high risk of bias exaggerated the estimates.

Nevertheless, sensitivity analysis revealed similar estimates (Appendix Figure S2C) , suggesting that the pooled effects were mainly derived from studies with low bias risk.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. Figure 4B ).

Moreover, we observed a non-linear relationship (P non-linearity <0.001), where the risks of progression to severe COVID-19 cases decelerated starting from 6.6 mmol/L and reexponentiated starting from admission FBG level of 8.1 mmol/L ( Figure 4C ). Both linear and non-linear trends remained consistent following sensitivity analyses (Appendix Table S8 and Figure S5 ). When dose-response analysis was conducted only for non-diabetic patients, we observed a slightly higher risk (RR 1.36 [95% CI: 1.28-1.44]; Appendix Table S9 ).

The observed effects were more accentuated in patients without history of diabetes, where high admission FBG level increased the risk of poor outcome for diabetic patients by 10% and non-diabetic patients by a whopping 75% (diabetic vs non-diabetic; RR 1.10 [95% CI: 1.03-1.17] vs 1.75 [1.02-2.99]). Furthermore, diabetes-stratified analyses for mortality and severity outcomes revealed that the observed trends were non-significant for diabetic subgroup, whereas the models for non-diabetic subgroups remained robust (Table 1) .

Due to the paucity of studies reporting specific criteria of COVID-19 severity (i.e. ICU admission, invasive ventilation, ARDS, shock), we were unable to ascertain the independent effects to the respective outcomes. Nonetheless, the results were coherent to the main outcomes as the pooled unadjusted effects showed similar trends in predicting ICU admission, invasive ventilation, ARDS, and shock (Appendix Table S5 ). However, as our findings were limited, we judged the certainty of evidence to be low for ICU admission and invasive ventilation, and very low for ARDS and shock (Appendix Table S6 ).

Similar to FBG, we discovered that COVID-19 patients with high RBG level at admission were more susceptible to poor prognosis (Appendix Table S5 ). Nevertheless, we were unable to establish a strong evidence on the independent prognostic value of admission RBG due to paucity of studies and equivocal trends. High vs low meta-analysis for adjusted effects was only eligible for outcomes on ICU admission, which resulted in non-significant estimate (RR 1.25 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint Figure 4D ), although study-specific estimates suggested otherwise.

Considering this, we judged the quality of evidence to be low for mortality, and very low for the remaining outcomes (Appendix Table S6 ).

With regards to mortality, we found that the prognostic value of admission RBG was independent of age and sex. 25 Appendix Table S10 ). 31 However, we were unable to perform a formal dose-response analysis due to insufficient information. In addition, two studies also reported that the risk of mortality was more accentuated in patients without history of diabetes 30 31 Subgroup and sensitivity analyses, as well as publication bias assessment and DRMA was not conducted due to insufficient number of studies.

This meta-analysis showed that high admission BG level was associated with poor prognosis in COVID-19 patients. Although we were unable to establish a firm evidence on the independent prognostic value of admission RBG, our results on admission FBG level was consistent and robust. Furthermore, we also demonstrated dose-response trend between admission FBG level and COVID-19 severity. Although the potential non-linear association between admission FBG level with poor composite and mortality outcomes remained unexplored due to paucity of studies, we were able to establish non-linear relationship between admission FBG level and severity. However, in contrast with Zhu et al. 37 , we did not observe a J-shaped association between admission FBG and COVID-19 severity, which may be explained by the fact that all but one study 37 reported only three categories.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint Further analysis indicated that the observed effects were more accentuated in patients without prior history of diabetes; which was intriguing, considering the fact that there was as increasing proportion of COVID-19 patients presenting with hyperglycemia despite having no prior diabetic history. 32, 35, 43 Furthermore, our findings indicated that at-admission hyperglycemia was associated with poorer outcomes regardless of prior diabetes status, suggesting the existence of a more direct link between glycemic status and poor COVID-19 outcomes.

The relationship between COVID-19 severity and hyperglycemia is possibly bidirectional,

wherein infection might bring about state of stress and trigger an enhanced release of proinflammatory cytokines which may lead to insulin resistance. 44 Stress may also induce the release of stress hormones which trigger liver glycogenolysis, aggravating the effects. 45 Furthermore, SARS-CoV-2 is known to bind to angiotensin-converting enzyme 2 (ACE2) receptors, which are found to be expressed in pancreatic beta-cells, thus rendering it a target for the viral attack. Such binding provides a route for the virus to enter and damage the pancreatic islets, resulting in a defect of insulin production, as indicated in previous study with its SARS virus counterpart. 46 Together, these factors may contribute to the development of acute hyperglycemia in COVID-19 patients.

The mechanism by which acute hyperglycemia drives the progression of COVID-19 remains largely unexplored. A study by Fadini et al. found that a decline in respiratory parameters was most responsible for mediating the effects of hyperglycemia on the outcome. 32 Diabetes and hyperglycemia were previously known to induce structural changes in the lungs, giving rise to pulmonary remodelling and the subsequent restrictive respiratory pattern. 47 Moreover, hyperglycemia is also known to generate reactive oxygen species and induces oxidative stress 44 , leading to endothelial dysfunction which may cause further hyperglycemic pulmonary microangiopathy. 48 This is in line with the findings of a study by Lampasona et al. which demonstrated that inflammation and coagulopathy, rather than impaired antibody response as such present in individuals with diabetes, were more responsible in aggravating the outcomes. 49 Therefore, this explains the poorer prognosis found in hyperglycemic patients without prior diabetic history, and again supporting the direct link between glucose level and disease progression.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09. 25.20200774 doi: medRxiv preprint Altogether, these findings illustrate the potential utility of admission BG as a predictor for poor prognosis in COVID-19 patients. Considering that BG measurement is relatively practical and instant, its quantification upon admission would be beneficial in predicting the likelihood of progression to severe COVID-19 cases. Therefore, we encourage clinicians to routinely obtain FBG values of each COVID-19 suspected case at admission, thus providing a simple method of risk stratification for management of patients in clinical settings, which would be particularly helpful in streamlining the limited number of medical resources during the current pandemic.

Although our findings also favored over the use of admission RBG, future researches are required as we were unable to comprehensively explore the independent prognostic value of admission RBG due to paucity of studies. Furthermore, the current evidence indicated that the cut-off values to predict poor prognosis in COVID-19 patients are still equivocal 31,34,50,51 , suggesting that future large multicenter studies are required to obtain the most optimum cut-off value.

Despite the fact that our findings favored over the prognostic value of high admission BG, the observed unexplained heterogeneity and the fact that all studies were retrospective and conducted in China may limit the generalizability of our findings. Furthermore, most of the studies included in the DRMA on severity outcome yielded moderate risk of bias, thus indicating potential overestimation of the observed effects due to imprecision. These indicated that the observed effects should be interpreted cautiously, and future studies with higher quality of evidence are required to confirm the estimated risks. Nonetheless, our results were consistent with the independent prognostic value of admission FBG, thus we judged the certainty of evidence for severity to be high, and for mortality and poor outcome to be moderate. To the best of our knowledge, this is the first meta-analysis conducted to show the potential use of admission BG as a predictor of poor prognosis in COVID-19 patients. Although our eligibility criteria may introduce language bias, our study included a relatively large number of cohorts and only four non-English articles were excluded [52] [53] [54] [55] , suggesting that any potential bias was negligible. We hope that our findings may enhance the current knowledge on the management of COVID-19, thus contributing to the alleviation of the devastating disease burden.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09. 25.20200774 doi: medRxiv preprint In conclusion, this meta-analysis adds to the growing body of evidence corroborating the potential utility of admission BG as a predictor of poor prognosis in COVID-19 patients. There is a high-quality evidence on the prognostic value of admission FBG towards severity, and moderate-quality evidence on its prognostic value towards mortality and poor outcome, while the other outcomes yielded very low-to-low quality evidence. In addition, we demonstrated nonlinear dose-response relationship between admission FBG and COVID-19 severity. Further studies to ascertain the estimated risk in the DRMA and to confirm the observed prognostic value of admission RBG are required. 

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (6):1450-1454.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 27, 2020. . https://doi.org/10.1101/2020.09.25.20200774 doi: medRxiv preprint

Pathophysiology, Transmission, Diagnosis, and Treatment of Coronavirus Disease 2019 (COVID-19): A Review

Review on COVID19 disease so far

Effects of Hyperglycaemia on Complications of COVID-19: A Meta-Analysis of Observational Studies

A guide to systematic review and meta-analysis of prognostic factor studies

Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement

Admission blood glucose level as a prognostic factor for COVID-19 severity and mortality: a systematic review

Clinical management of COVID-19: interim guidance -27

Evaluation of the quality of prognosis studies in systematic reviews

Judging the quality of evidence in reviews of prognostic factor research: adapting the GRADE framework

Methodological issues and recommendations for systematic reviews of prognostic studies: an example from cardiovascular disease

Meta-analysis in clinical trials

Contour-enhanced meta-analysis funnel plots help distinguish publication bias from other causes of asymmetry

Bias in meta-analysis detected by a simple, graphical test

R: A language and environment for statistical computing

Higher Comorbidities and Early Death is Characteristic of Hospitalized African-American Patients with COVID-19. medRxiv

Clinical findings of patients with coronavirus disease 2019 in Jiangsu province, China: A retrospective, multi-center study

Development of a severity of disease score and classification model by machine learning for hospitalized COVID-19 patients. medRxiv

Impaired glucose metabolism in patients with diabetes, prediabetes, and obesity is associated with severe COVID-19

CLINICAL CHARACTERISTICS OF 28 PATIENTS WITH DIABETES AND COVID-19 IN WUHAN, CHINA

Clinical characteristics and outcomes of patients with severe covid-19 with diabetes

Clinical analysis of risk factors for severe COVID-19 patients with type 2 diabetes

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

Factors Associated with Early Invasive Ventilation in Critically Ill Patients with COVID-19: A Single-Centered, Retrospective, Observational Study

Admission hyperglycemia and radiological findings of SARS-CoV2 in patients with and without diabetes

Glycemic Characteristics and Clinical Outcomes of COVID-19 Patients Hospitalized in the United States

A Risk Score to Predict Admission to Intensive Care Unit in Patients With COVID-19: The ABC-GOALS Score. medRxiv

Outcomes in Patients With Hyperglycemia Affected by COVID-19: Can We Do More on Glycemic Control? Diabetes Care

Hyperglycemia at Hospital Admission Is Associated With Severity of the Prognosis in Patients Hospitalized for COVID-19: The Pisa COVID-19 Study. Diabetes Care

Newly-diagnosed diabetes and admission hyperglycemia predict COVID-19 severity by aggravating respiratory deterioration

Identifying main and interaction effects of risk factors to predict intensive care admission in patients hospitalized with COVID-19: a retrospective cohort study in Hong Kong. medRxiv

Fasting Plasma Glucose Level Independently Predicts the Mortality of Patients with Coronavirus Disease 2019 Infection: A Multicenter, Retrospective Cohort Study

Fasting blood glucose at admission is an independent predictor for 28-day mortality in patients with COVID-19 without previous diagnosis of diabetes: a multi-centre retrospective study

Association of diabetes mellitus with disease severity and prognosis in COVID-19: A retrospective cohort study

Correlation between Fasting Blood Glucose Level at Admission and Mortality in COVID-19 Patients: A Retrospective Study

Newly diagnosed diabetes is associated with a higher risk of mortality than known diabetes in hospitalized patients with COVID-19

Impaired Fasting Glucose and Diabetes Are Related to Higher Risks of Complications and Mortality Among Patients With Coronavirus Disease

Hyperglycemia is a strong predictor of poor prognosis in COVID-19

J-shaped Association Between Fasting Blood Glucose Levels and COVID-19 Severity in Patients without Diabetes

Higher fasting plasma glucose reduced the survival rate of 306 hospitalized patients with COVID-19 in Wuhan

Elevation of blood glucose level predicts worse outcomes in hospitalized patients with COVID-19: a retrospective cohort study

Stress-hyperglycemia, insulin and immunomodulation in sepsis

COVID-19 and diabetes : a review literature

Binding of SARS coronavirus to its receptor damages islets and causes acute diabetes

The diabetic lung: an easy target for SARS-CoV-2?

Diabetic microvascular disease and pulmonary fibrosis: The contribution of platelets and systemic inflammation

Antibody response to multiple antigens of SARS-CoV-2 in patients with diabetes: an observational cohort study

Risk factors influencing the prognosis of elderly patients infected with COVID-19: a clinical retrospective study in Wuhan

Admission biochemical test associated with the prognosis of COVID-19: a multi-centered retrospective cohort study

Logistic regression analysis of death risk factors of patients with severe and critical coronavirus disease