This primary care study found an increased incidence of type 2 diabetes in individuals with Covid-19 after recovery. The strengths of the study are its use of a nationwide primary care database that is representative of diagnoses and drug prescriptions . Second, the sample size was larger than in previous studies, which were mostly hospital-based . Third, recall bias was unlikely because of the use of original data collected in primary care.
The characteristics of individuals with Covid-19 in the present study were largely as expected from national statistics . The median age of all 2.4 million people with Covid-19 in Germany from January 2020 to February 2021 was 44 years, with a slightly larger proportion of women (52%). Of these 2.4 million, 10% were hospitalised, a higher proportion than in the DA, most likely reflecting missing data due to emergency hospitalisations not recorded in the practices. Furthermore, because of the exclusion of individuals with a diabetes history, a healthier sample with milder Covid-19 infections was selected, requiring fewer hospitalisations. Finally, 3.1% of people with Covid-19 in Germany died . Thus, although death is not recorded in the DA, this bias is negligible.
A main limitation of the study is that there was no detailed information on hospitalisations available because the database only includes data obtained by primary care physicians in their clinical practice. Diagnostic data from external specialists and hospitals are only recorded in the database if the primary care physician adds this information. Because of the potential for incomplete medical records, information bias may have been introduced. However, it is most likely that misclassification of exposure and outcome was non-differential, meaning that the errors were the same in the two groups being compared. Furthermore, individuals with undetected diabetes were not included. Transient acute hyperglycaemia, which may have been mistaken as manifest diabetes, could not be ascertained. Individuals with a Covid-19 diagnosis outside of primary care practice (e.g. in hospitals or at Covid-19 test centres) were also not necessarily completely included. Negative SARS-CoV-2 tests were recorded in the practices in only 12,981 (36%) of the AURI group during 365 days after the index date. BMI was not controlled for due to missing data. Furthermore, the number of censored patients due to change of practice or death could not be determined. In the German healthcare system, patients are free to choose a primary care practice. Nevertheless, the vast majority of patients are consistently treated by a single primary care practice. Usually, death of patients is not recorded in German practices. However, the likelihood of short-term mortality is low in patients with mild Covid-19. The incidence of type 1 diabetes was not investigated due to the small number of cases. Finally, the generalisability (external validity) of the data, i.e. the extent to which our results may be applied to other circumstances, requires analysis using separate databases.
The present results from primary care are in line with a retrospective cohort study of hospitalised Covid-19 patients in UK, comprising individuals with more severe disease, higher virus load and greater immune activation . Over a mean follow-up of 140 days, the incidence of new-onset diabetes in 47,780 Covid-19 patients (mean age 65 years; 55% men) was 29 (95% CI 26, 32) per 1000 person-years, yielding a rate ratio of 1.5 (95% CI 1.4, 1.6) compared with matched controls from the general population . Furthermore, merging three databases in the USA (administrative claims, outpatient laboratory testing, inpatient hospital admissions), the hazard ratio for newly diagnosed diabetes in the 6 months after the acute phase of Covid-19 was 2.47 (95% CI 1.14, 5.38) in 266,586 patients infected with SARS-CoV-2 (age 18–65 years) compared with matched controls .
Insulin resistance and impaired insulin secretion have been described in individuals without diabetes history who recovered from SARS-CoV-2 infections . Cytokines and TNF-α remain upregulated after remission of Covid-19, which may induce beta cell dysfunction and insulin resistance . Thus, there are plausible mechanisms for a causal relationship between Covid-19 and newly diagnosed type 2 diabetes.
It has been suggested that diagnosis and treatment of post-Covid syndrome require integrated rather than disease-specific approaches . If confirmed, the results of the present study indicate that diabetes screening in individuals who have recovered from even mild Covid-19 should be recommended.
However, there are still a number of unanswered questions for future research. First, it is unclear whether pre-existing diabetes becomes apparent during Covid-19 as a consequence of immunological activation or stress hyperglycaemia . Second, it should be investigated if post-Covid diabetes may be reversed after full recovery . Third, the management of new-onset diabetes after Covid-19 should be evaluated. Diabetes ketoacidosis has been observed in some individuals without known diabetes even months after Covid-19 . Thus, serological testing for diabetes-associated autoantibodies and C-peptide may be indicated in individuals without known risk factors for diabetes after Covid-19. Finally, the risk of hyperglycaemia in individuals with Covid-19 is most likely a continuum, depending on risk factors such as injury of beta cells, an exaggerated proinflammatory response and changes in health behaviour during the pandemic. In particular, by using fixed cut-off points for the definition of incident diabetes, our study may have missed changes in hyperglycaemia risk below these cut-off points. Future studies should investigate the effects of SARS-CoV-2 infections on glucose and HbA1c measurements on a continuous scale.
In conclusion, the present primary care study indicates a temporal relationship between mostly mild Covid-19 and newly diagnosed type 2 diabetes. If confirmed, this study supports the potential relevance of active monitoring of glucose dysregulation after recovery from mild forms of SARS-CoV-2 infection.