This population-based study of patients with type 2 diabetes suggests that average glycemic control, and in particular current glycemic control—assessed as the latest updated HbA1c level—is important for the risk of infection in type 2 diabetes, particularly for hospital-treated infections. In contrast, there seemed to be no strong association between baseline HbA1c levels obtained soon after the start of therapy and later infections.
Our findings underscore the importance of the present guidelines for HbA1c targets. Our results indicate that for infectious complications, current hyperglycemia as measured by the single latest HbA1c level is important, supporting the hypothesis of an acute and reversible impact of hyperglycemia on infections. There may be differences in the mechanisms at play for infection and micro- and macrovascular complications. For vascular diabetes complications, Lind et al. suggested that mean or updated mean HbA1c values in general are more important than single HbA1c measurements.
Evidence from similar cohort studies on the association between glycemic control over time and risk of infection in type 2 diabetes is limited.[15–18,30] Our study corroborates findings from a smaller Dutch study of general practices. In that study, Bartelink et al. reported no overall difference in mean HbA1c levels in type 2 diabetes patients with and without infection, whereas patients who presented with an infection at some point during follow-up showed higher HbA1c levels in that period compared with periods without any infection. Other studies have assessed single-point HbA1c values, focusing on specific selected infections. Those investigators reported increased risks associated with poor glycemic control for bloodstream infections,[16,17] pneumonia requiring hospitalization, tuberculosis, vaginitis and balanitis, and urinary tract infection. The Copenhagen City Heart Study, a study of the Copenhagen general population, assessed plasma glucose at baseline and found a particularly increased risk of urinary tract infections and skin infections with increased glucose levels. This is in line with our results. Among patients undergoing surgical cardiac procedures, acute hyperglycemia is a known predictor of wound infections,[3,31] and randomized trials have shown that intensive insulin treatment may reduce the risk of subsequent sepsis or wound infection.
In our study, we found increased risk of infection at HbA1c levels below 5.50%, in addition to high levels. A similar J-shaped association has been observed between HbA1c and mortality and cardiovascular disease. We observed that the patient group with HbA1c values less than 5.50% was younger than other patients, more likely to have high comorbidity (notable given their younger age), and more likely to have alcoholism-related conditions (Web Table 1). We speculate that some of these patients had mild or borderline diabetes detected during clinical workup and treatment for other severe conditions, and that the apparently higher infection risk associated with very low HbA1c might be explained by unmeasured comorbidity and other risk factors in these patients. Other researchers[35,36] observing similar J- or U-shaped outcome curves have suggested that patients with very low HbA1c levels may have more comorbidity and another "phenotype" of type 2 diabetes. Alternatively, causal and detrimental effects of hypoglycemia may be at play. Hypoglycemia thus might lead to adverse outcomes through sympathoadrenal activation, thrombogenesis, vasoconstriction, and the release of inflammatory mediators and cytokines. Carson et al. reported that an HbA1c concentration as low as <4% is a marker of unfavorable red blood cell factors and is suggestive of underlying inflammation and liver disease. Additional research with more detailed clinical and biomarker data than our study possessed is needed to investigate the exact mechanisms.
In our study, the setting of the Danish health-care system permitted the use of a population-based design with inclusion of all patients who had hospital- or drug-treated type 2 diabetes within a well-defined region and a homogenous population, as well as complete follow-up and availability of laboratory data for assessment of glycemic control. These features largely eliminated the selection problems prevalent in smaller follow-up studies based on limited numbers of participants. By using both prescription and hospital-based data, we were able to identify all infections requiring medical attention, unlike previous studies, which often focused exclusively on infections treated in the hospital.
Our study also had limitations. First, we defined incident diabetes on the basis of prescriptions and hospital diagnoses in Danish registries, not exact measurements of hyperglycemia. For example, we missed type 2 diabetes patients treated with lifestyle interventions, and we relied on HbA1c measurements as ordered by physicians in the course of routine clinical care. Second, patients with poor glycemic control versus good glycemic control may have a lower threshold of antiinfective or hospital treatment when infection is suspected (surveillance bias), leading to overestimation of the association. Third, we cannot exclude the possibility of reverse causality in some patients in clinical practice—that is, latent infection's leading to increasing HbA1c levels. Fourth, most of our confounders were measured at the index date, and values for some of them may have changed during follow-up. However, follow-up was short due to early outcome events in many patients, and factors that may be affected by exposure to high HbA1c levels during follow-up should not be adjusted for. Fifth, we did not have information on certain infection risk factors that may have altered HbA1c values, such as blood transfusions or enteral or parenteral nutrition, which could have led to HbA1c misclassification in some patients. Sixth, as in any observational study, other imperfectly measured, unmeasured, or unknown factors may have affected the observed associations, including high body mass index, smoking, low physical activity, and other adverse lifestyle and socioeconomic measures. Nonetheless, we were able to adjust for a wide range of medical conditions closely associated with these adverse factors, likely reducing their confounding effect. Finally, we had no genetic data available with which to examine any genetic predisposition for insulin resistance, diabetes, and infections.
It has been hypothesized that increased risk of infection may be mediated primarily by long-term chronic hyperglycemia via chronic tissue inflammation or development of other complications, which in turn increase risk of infection.[3,6,7] As has been reviewed elsewhere, numerous in vitro studies have demonstrated that hyperglycemia may impair the innate immune system, inhibit adaptive immunity, and interfere with complement cascade through glycosylation of immune proteins.[6,7] Such processes may underlie our finding of increased risk of infection associated with current hyperglycemia. Alternatively, unmeasured factors associated with high HbA1c levels, such as high body mass index and lower socioeconomic status—both of which are documented risk factors for infection[38,39]—may explain our findings in part. A large proportion of patients with very high HbA1c levels (≥10.50%) used neither glucose-lowering nor statin treatment. Such poor glucose control may be a marker of decreased compliance with preventive therapies in general, including use of other cardiovascular drugs and possibly vaccinations.
In summary, our population-based cohort study provides evidence that among patients with type 2 diabetes, current hyperglycemia is associated with increased risk of community-treated and hospital-treated infections. The findings from this study suggest that infections in persons with type 2 diabetes may be prevented with appropriate and consistent glycemic control.
This work was supported by the Danish Centre for Strategic Research in Type 2 Diabetes (DD2) and the Program for Clinical Research Infrastructure, established by the Lundbeck Foundation and the Novo Nordisk Foundation. The DD2 is supported by the Danish Agency for Science (grants 09–067009 and 09–075724), the Danish Health and Medicines Authority, the Danish Diabetes Association, and an unrestricted donation from Novo Nordisk A/S. Project partners are listed on the DD2 website (www.DD2.nu).
The Department of Clinical Epidemiology, Aarhus University Hospital, receives funding for other studies from companies in the form of research grants to (and administered by) Aarhus University.
CCI, Charlson Comorbidity Index; CI, confidence interval; DNPR, Danish National Patient Registry; HbA1c, hemoglobin A1c; HR, hazard ratio.
Am J Epidemiol. 2017;186(2):227-236. © 2017 Oxford University Press