In our living systematic review and meta-analysis, we summarised the current knowledge on associations between phenotypes of individuals with diabetes and confirmed SARS-CoV-2 infection regarding COVID-19-related death and COVID-19 severity and evaluated their certainty of evidence. Moderate to high certainty of evidence for higher risk of COVID-19-related death was observed for male sex, older age, CVD, CKD, COPD, high plasma blood glucose at admission and chronic insulin use. Metformin use was inversely associated with death. For COVID-19 severity, similar associations were observed in general, but estimates were lower and less precise.
Older age, male sex, obesity, hypertension, chronic pulmonary diseases, CVD, active cancer [3,4,5,6, 21], laboratory parameters (e.g. low lymphocyte count, and elevations in CRP, ALT and AST) [48] have been linked to a poor prognosis of COVID-19 in the general population infected with SARS-CoV-2. These risk factors among the general population are in line with the risk factors we identified in the diabetes populations, with some exceptions. Interestingly, we did not observe a positive association for obesity or hypertension with COVID-19 severity or death in people with diabetes and COVID-19. In addition, higher white blood cell (leucocyte) and neutrophil counts and lower lymphocyte counts also increased the relative risk for COVID-19-related death and COVID-19 severity in our meta-analyses. Nevertheless, we observed no clear associations for CRP (the most frequently measured biomarker of inflammation) or for liver enzymes (ALT, AST). However, only two to four primary studies in our meta-analyses included these biomarkers, and the certainty of evidence was low or very low, meaning that it is likely or very likely that further studies might change the results. Interestingly, findings from a large representative study in England indicated that higher HbA1c was associated with poor prognosis of COVID-19 in individuals with diabetes: the RR (95% CI) for HbA1c < 7.5% (<58.5 mmol/mol) and death was 1.31 (1.24, 1.37), and for HbA1c ≥ 7.5% (≥58.5 mmol/mol), it was 1.95 (1.83, 2.08) compared with individuals without diabetes [20]. In another population-based study of participants with diabetes (but not all with SARS-CoV-2 diagnosis), associations between HbA1c levels and death related to COVID-19 was less clear [23], which is comparable with our findings. Only at HbA1c values of ≥10% (≥85.8 mmol/mol) was a clear association observed (RR 2.23 [95% CI 1.50, 3.30]) compared with individuals with HbA1c values between 6.5% (47.5 mmol/mol) and 7% (53 mmol/mol).
Furthermore, high blood glucose at admission has been shown to be a marker for poor prognosis of COVID-19, even in individuals without pre-existing diabetes [49]. One study reported that individuals with well-controlled diabetes had a better prognosis of COVID-19 compared with individuals with poorly controlled diabetes [47]. In our meta-analysis, higher blood glucose at admission was also associated with worse prognosis of COVID-19. The certainty of evidence was very low to moderate because of the limited number of original studies.
Regarding glucose-lowering medication, chronic insulin use was associated with higher risk of COVID-19-related death, while metformin use was associated with lower risk. We speculate that it is not the treatment, per se, that is associated with prognosis of COVID-19, but rather that it represents an indicator of severity of diabetes. Unfortunately, our meta-analyses did not allow for stratification by diabetes severity or duration. We could also not stratify our meta-analyses by diabetes type.
Nevertheless, we observed a higher relative risk for COVID-19-related death when comparing type 2 with type 1 diabetes, but the findings were not statistically significant and only based on two studies and, thus, certainty of evidence was very low. On the contrary, a large population-based study from England indicated that, when compared with individuals without diabetes, individuals with type 1 diabetes had a worse prognosis than individuals with type 2 diabetes [22]. Holman et al. showed in their study (which included participants with diabetes but not all with SARS-CoV-2 infection) that age, sex, hypertension, obesity and comorbidities were associated with COVID-19-related death for both type 1 and type 2 diabetes [23]. Moreover, in these population-based studies, socioeconomic deprivation was associated with COVID-19-related death [21, 23]; we could not investigate associations with socioeconomic deprivation in our meta-analysis because these data were not available from the primary studies included.
After our last update, eligible studies providing important data have been published on this topic, such as the report from McGurnaghan et al., which covered the whole Scottish population, including individuals with diabetes [50]. In general, the data support our findings but provide new insights on further risk factors not included in our study. For example, a higher level of deprivation, any admission due to diabetic ketoacidosis or hypoglycaemia in the previous 5 years, pre-existing immune diseases, use of immunosuppressants and evidence for retinopathy were all associated with severity of COVID-19.
Taken together, the risk group we identified for the population with diabetes and COVID-19, i.e. older individuals with comorbid conditions and using insulin, might simply reflect severity of diabetes or poor health conditions per se. Nevertheless, considering these phenotypes can be helpful for identifying people with diabetes and COVID-19 at high risk for poor outcomes and, therefore, those most likely to require early intensified treatment.
The strengths of our report include the comprehensive literature search that used four different databases, the assessment of risk of bias of the primary studies using a validated tool, the consideration of risk of bias due to confounding in our analysis and the assessment of the certainty of evidence by following the GRADE approach. In addition, our living review will be updated continuously and provide information regarding the best evidence on prognosis of COVID-19 among individuals with diabetes.
Our study has a number of limitations. First, there was a low number of primary studies for some of the associations assessed, including diabetes-specific factors (e.g. diabetes duration, HbA1c, use of specific glucose-lowering medications), certain comorbidities (e.g. cancer, liver disease, and dementia) and laboratory parameters at admission (e.g. CRP, differential blood cell count, liver enzymes). For these associations, the certainty of evidence was mainly graded as low or very low, reflecting that future research might change the findings. Second, the risk of bias was high for eight studies (36% of the studies included), mainly due to insufficient adjustment of important confounders. However, we stratified our meta-analyses by risk of bias due to confounding and the overall findings were robust, with two exceptions; older age and CKD were more strongly associated with COVID-19-related death in studies with high risk of bias compared with studies with low or moderate risk of bias due to confounding. Third, it was not possible to conduct stratified analyses by study design, data collection and diabetes type. Fourth, the primary studies did not account for possible specific treatment of COVID-19. Fifth, a general question remains as to whether deceased participants died with or due to COVID-19. Finally, these findings cannot be generalised to all individuals with diabetes infected with SARS-CoV-2 because only individuals with the more severe form of COVID-19 are included in the primary studies and the majority of the studies were conducted in the hospital setting.
In conclusion, our living systemic review and meta-analysis provides the best current evidence on associations between phenotypes of individuals with diabetes and confirmed SARS-CoV-2 and COVID-19-related death and severity of COVID-19. Male sex, older age and pre-existing comorbidities (CVD, CKD and COPD), as well as the use of insulin, most of which are potential indicators for a more progressive course of diabetes, were associated with increased risk of COVID-19-related death and severity in individuals with diabetes and SARS-CoV-2, whereas metformin use had associations in the opposite direction. To strengthen the evidence, more primary studies investigating diabetes-specific risk factors, e.g. type and duration of diabetes or additional comorbidities (such as liver disease and neuropathy), and accounting for important confounders, are needed. We will continuously update this report to strengthen the evidence of already examined associations and to investigate further outcomes, such as long-term complications due to COVID-19 for individuals with diabetes.