Article Text

Download PDFPDF

Original article
Inflammatory markers are associated with cardiac autonomic dysfunction in recent-onset type 2 diabetes
  1. Christian Herder1,2,
  2. Imke Schamarek1,2,
  3. Bettina Nowotny1,2,
  4. Maren Carstensen-Kirberg1,2,
  5. Klaus Straßburger2,3,
  6. Peter Nowotny1,2,
  7. Julia M Kannenberg1,2,
  8. Alexander Strom1,2,
  9. Sonja Püttgen1,2,
  10. Karsten Müssig1,2,4,
  11. Julia Szendroedi1,2,4,
  12. Michael Roden1,2,4,
  13. Dan Ziegler1,2,4
  14. for the German Diabetes Study Group
    1. 1Institute for Clinical Diabetology, German Diabetes Center, Leibniz Center for Diabetes Research at Heinrich Heine University Düsseldorf, Düsseldorf, Germany
    2. 2German Center for Diabetes Research (DZD), München-Neuherberg, Germany
    3. 3Institute for Biometrics and Epidemiology, German Diabetes Center, Leibniz Center for Diabetes Research at Heinrich Heine University Düsseldorf, Düsseldorf, Germany
    4. 4Department of Endocrinology and Diabetology, Medical Faculty, Heinrich Heine University Düsseldorf, Düsseldorf, Germany
    1. Correspondence to Dr Christian Herder MSc, Institute for Clinical Diabetology, German Diabetes Center, Auf'm Hennekamp 65, Düsseldorf 40225, Germany; christian.herder{at}ddz.uni-duesseldorf.de

    Abstract

    Objective Cardiovascular autonomic neuropathy is a common but underestimated diabetes-related disorder. Associations between cardiovascular autonomic dysfunction and subclinical inflammation, both risk factors of diabetic comorbidities and mortality, have been proposed in non-diabetic populations, while data for type 1 and type 2 diabetes are conflicting. Our aim was to investigate associations between inflammation-related biomarkers and cardiac autonomic dysfunction in patients with diabetes.

    Methods We characterised the associations between seven biomarkers of subclinical inflammation and cardiac autonomic dysfunction based on heart rate variability and cardiovascular autonomic reflex tests (CARTs) in 161 individuals with type 1 and 352 individuals with type 2 diabetes (time since diagnosis of diabetes <1 year). Analyses were adjusted for age, sex, anthropometric, metabolic and lifestyle factors, medication and cardiovascular comorbidities.

    Results In individuals with type 2 diabetes, higher serum interleukin (IL)-18 was associated with lower vagal activity (p≤0.015 for association with CARTs), whereas higher levels of total and high-molecular-weight adiponectin showed associations with very low frequency power, an indicator of reduced sympathetic activity (p≤0.014). Higher levels of soluble intercellular adhesion molecule-1 were associated with indicators of both lower vagal (p=0.025) and sympathetic (p=0.008) tone, soluble E-selectin with one indicator of lower vagal activity (p=0.047). Serum C-reactive protein and IL-6 were also related to cardiac autonomic dysfunction, but these associations were explained by confounding factors. No consistent associations were found in individuals with type 1 diabetes.

    Conclusions Biomarkers of inflammation were differentially associated with diminished cardiac autonomic dysfunction in recent-onset type 2 diabetes.

    Statistics from Altmetric.com

    Introduction

    Cardiovascular autonomic neuropathy (CAN) is a common and serious complication of both type 1 and type 2 diabetes with prevalence rates of approximately 20% in unselected patient groups, but reaching up to 65% with increasing age and duration of diabetes.1 Symptomatic CAN is mainly a consequence of long-standing diabetes, whereas alterations in resting heart rate variability (HRV) and cardiovascular autonomic reflex tests (CARTs) usually precede the onset of autonomic symptoms and allow an early detection of CAN at asymptomatic stages, which is especially relevant in prediabetes or recent-onset diabetes.1 ,2 Even in the absence of symptomatic CAN, impaired HRV and CARTs represent independent predictors of cardiovascular problems and increased mortality.2

    Risk factors for CAN include age and poor glycaemic control in both diabetes types and obesity, dyslipidaemia and hypertension in type 2 diabetes.1 Inflammatory processes may contribute to distal sensorimotor polyneuropathy (DSPN) in individuals with diabetes,3 but their relevance for the development of CAN is not yet fully understood. The analysis of inflammation as a potential risk factor for CAN is complicated by the fact that the autonomic nervous system (ANS) itself also exerts immunomodulatory functions.4 ,5 Biomarkers of subclinical inflammation were associated with reduced HRV in study populations with and without diabetes, but most studies were limited to C-reactive protein (CRP) and/or interleukin (IL)-6.6 Moreover, the issue of confounding due to obesity-related factors and comorbidities has not been thoroughly addressed, and data for individuals with recent-onset diabetes are very limited.7 Thus, it is currently not clear if inflammation-related biomarkers are associated with CAN in patients with diabetes.

    Therefore, we aimed to test the hypothesis that serum levels of biomarkers representing different aspects of subclinical and vascular inflammation (acute-phase protein, proinflammatory cytokines, adiponectin, soluble adhesion molecules) are associated with measures of resting HRV and CARTs reflecting different degrees of sympathetic and parasympathetic activity in patients with recently diagnosed type 1 and type 2 diabetes.

    Methods

    Study population

    The German Diabetes Study (GDS) is an ongoing prospective cohort study investigating the natural history of recent-onset diabetes and the development of its complications.8 The study is being conducted in accordance with the Declaration of Helsinki and was approved by the ethics committee of Heinrich Heine University, Düsseldorf, Germany (ClinicalTrials.gov Identifier: NCT01055093 still recruiting). All participants gave written informed consent.

    Inclusion criteria for the GDS are a diagnosis of type 1 diabetes or type 2 diabetes, known diabetes duration of <1 year and age between 18 and 69 years at the baseline examination. Exclusion criteria, the experimental design including a structured interview, anthropometry and measurement of haemoglobin A1c (HbA1c) and lipids have been reported previously.8 ,9

    This cross-sectional analysis is based on consecutive participants who entered the study between its start in September 2005 and December 2011 (n=513). Due to missing data for CAN variables, the analyses were limited to two largely overlapping samples of n=413 and n=482 for analyses of HRV variables and CARTs, respectively (see online supplementary table S1 and table 1).

    Table 1

    Patient characteristics stratified by diabetes type (study sample for the CART analyses)

    supplementary table

    Patient characteristics stratified by diabetes type (study sample for the HRV analyses)

    Assessment of heart rate variability

    R-R intervals were recorded in the supine position during a hyperinsulinaemic-euglycaemic clamp over 3 hours using a digital Spider View Holter recorder with seven electrodes to record three-channel ECGs (Sorin Group, Munich, Germany). HRV was analysed from the Holter monitor recordings with the SyneScope V.3.00 analysis system (Sorin Group). The sampling rate of the ECG signal was 200 Hz (5 ms resolution). The system automatically edits all artefacts and ectopic beats and obtains a regular signal by linear interpolation of the heart rate tachogram. Time-domain and frequency-domain measures were computed in accordance with the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Time-domain measures included normal-to-normal mean R-R intervals, the standard deviation of N-N averages (SDANN) over 5 min and the root mean square of successive differences (RMSSD). Frequency-domain indices included the very low frequency band (VLF; 0.003–0.04 Hz), the low frequency band (LF; 0.04–0.15 Hz), the high frequency band (HF; 0.15–0.4 Hz) and the LF/HF ratio.

    Cardiovascular autonomic reflex tests

    CARTs including heart rate changes in response to a Valsalva manoeuvre (Valsalva ratio) and standing up (maximum-to-minimum (max/min) 30:15 ratio) were performed using a VariaCardio TF5 system (MIE Medical Research, Leeds, UK).

    Measurement of biomarkers of subclinical and vascular inflammation

    Serum concentrations of high-sensitivity C-reactive protein (hsCRP), IL-6, IL-18, total and high-molecular-weight (HMW) adiponectin, soluble intercellular adhesion molecule-1 (sICAM-1) and soluble E-selectin (sE-selectin) were measured as described.9 ,10

    Statistical analysis

    Data are presented as mean±SD, median (25th; 75th percentiles) or percentages (%). We tested for differences between diabetes types with Student's t test, Mann-Whitney U test and Fisher's exact test. Variables without Gaussian distribution (triglycerides, all biomarkers of subclinical inflammation and all HRV variables) entered regression models after log-transformation.

    Associations between z-standardised inflammation-related biomarkers and continuous variables of cardiac autonomic dysfunction were estimated using multivariable linear regression models (separate models for the associations of each of the eight biomarkers with each of the eight variables of cardiac autonomic dysfunction) with increasing complexity stratified by diabetes type. Associations between inflammation-related biomarkers and covariables in these regression models were estimated using Spearman's rank correlation coefficients and corresponding p values to provide a comprehensive overview for associations between the biomarkers and variables used for adjustment.

    All statistical analyses were carried out using SPSS V.22 (IBM, Ehningen, Germany) and R V.3.2.4 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria). p Values <0.05 were considered to indicate statistically significant differences or associations. Analyses were not adjusted for multiple testing and should therefore be considered exploratory.

    Results

    Study population

    The two overlapping study samples for analyses regarding HRV variables and CARTs consisted of 285 individuals with type 2 diabetes and 128 individuals with type 1 diabetes (see online supplementary table S1) and 329 individuals with type 2 diabetes and 153 individuals with type 1 diabetes, respectively (table 1). In both study samples, individuals with type 2 diabetes were older and had a less favourable cardiometabolic risk factor profile than those with type 1 diabetes, with the exception of lower HbA1c in type 2 diabetes. Moreover, persons with type 2 diabetes had higher levels of most inflammation-related biomarkers, but lower adiponectin levels than persons with type 1 diabetes. Individuals with type 2 diabetes showed lower HF, LF and VLF power, lower RMSSD as well as lower Valsalva and 30:15 ratios than individuals with type 1 diabetes, but there was no difference in the LF/HF ratio between both diabetes types (table 1). Correlations between inflammation-related biomarkers and covariables used in subsequent regression analyses are summarised in online supplementary tables S2 and S3 for type 2 and type 1 diabetes, respectively.

    supplementary table

    Correlations between biomarkers of subclinical inflammation and phenotypes used as covariables in multiple regression analyses for patients with type 2 diabetes

    supplementary table

    Correlations between biomarkers of subclinical inflammation and phenotypes used as covariables in multiple regression analyses for patients with type 1 diabetes

    Biomarkers of subclinical inflammation and cardiac autonomic dysfunction in type 2 diabetes

    Initial age-adjusted and sex-adjusted analyses (model 1) indicated 23 significant associations (based on a total of 64 tests) between biomarkers of inflammation and parameters of cardiac autonomic dysfunction. In detail, serum hsCRP and IL-6 were inversely associated with HF, LF and VLF power in model 1, but most of these associations were explained by the covariables in models 2 and 3 (table 2). IL-18 was inversely associated with HF power and RMSSD and positively with the LF/HF ratio in model 1, but these associations were also attenuated to non-significance by further adjustment. Total and HMW adiponectin levels displayed stable inverse associations with VLF power after adjustment. Serum sE-selectin and sICAM-1 showed comparable associations with at least four out of the six HRV variables in model 1, but only the inverse associations between sICAM-1 and SDANN and between sE-selectin and RMSSD remained robust after adjustment (table 2).

    Table 2

    Association between biomarkers of subclinical inflammation and measures of HRV in patients with type 2 diabetes

    When cardiac autonomic dysfunction was assessed using CARTs, serum IL-18 showed consistent inverse associations with the Valsalva and max/min 30:15 ratios, while sICAM-1 was inversely related to max/min 30:15 ratio in all models (table 3).

    Table 3

    Association between biomarkers of subclinical inflammation and CARTs in patients with type 2 diabetes

    Biomarkers of subclinical inflammation and cardiac autonomic dysfunction in type 1 diabetes

    The same analyses were also conducted in the subgroup of individuals with type 1 diabetes, but no consistent associations between biomarkers of subclinical inflammation and HRV were observed (see online supplementary tables S4 and S5). Only serum IL-6 was inversely associated with Valsalva ratio (see online supplementary table S5).

    supplementary table

    Association between biomarkers of subclinical inflammation and measures of heart rate variability (HRV) in patients with type 1 diabetes

    supplementary table

    Association between biomarkers of subclinical inflammation and cardiovascular autonomic reflex tests (CARTs) in patients with type 1 diabetes

    Comparison of effect sizes for associations between biomarkers of subclinical inflammation and cardiac autonomic dysfunction in both diabetes types

    Figure 1 presents regression coefficients and 95% CIs for all associations that were statistically significant in model 1 (adjusted for age and sex) for either diabetes type. Overall, associations observed in patients with type 2 diabetes showed larger effect sizes compared with type 1 diabetes, and in about half of the cases associations even appeared in opposite directions. Further adjustment (model 2) led to attenuated effect sizes, but had no substantial impact on the differences between both diabetes types (figure 2).

    Figure 1

    Overview of regression coefficients and corresponding 95% CIs for significant associations between biomarkers of subclinical inflammation and markers of cardiac autonomic dysfunction in age-adjusted and sex-adjusted analyses (model 1). Dotted lines, type 1 diabetes; black lines, type 2 diabetes. HF, high frequency; hsCRP, high-sensitivity C-reactive protein; IL, interleukin; LF, low frequency; max/min, maximum-to-minimum; RMSSD, root mean square of successive differences; SDANN, standard deviation of N-N averages over 5 min; sE-selectin, soluble E-selectin; sICAM-1, soluble intercellular adhesion molecule-1; VLF, very low frequency.

    Figure 2

    Overview of regression coefficients and corresponding 95% CIs for significant associations between biomarkers of subclinical inflammation and markers of cardiac autonomic dysfunction in analyses based on model 2 (adjusted for age, sex, time since diagnosis of diabetes, haemoglobin A1c, waist circumference, height, total cholesterol, hypertension). Dotted lines, type 1 diabetes; black lines, type 2 diabetes. HF, high frequency; hsCRP, high-sensitivity C-reactive protein; IL, interleukin; LF, low frequency; max/min, maximum-to-minimum; RMSSD, root mean square of successive differences; SDANN, standard deviation of N-N averages over 5 min; sE-selectin, soluble E-selectin; sICAM-1, soluble intercellular adhesion molecule-1; VLF, very low frequency.

    Discussion

    In the subgroup with recent-onset type 2 diabetes, all five immune mediators tested here, reflecting increased cardiovascular risk, were associated with at least three parameters of cardiac autonomic dysfunction in age-adjusted and sex-adjusted analyses. Independent associations with cardiac autonomic dysfunction were observed for IL-18, sICAM-1 and sE-selectin, whereas associations of hsCRP and IL-6 were partly explained by confounding factors. Moreover, increased levels of total and HMW adiponectin were positively associated with cardiac autonomic dysfunction in type 2 diabetes. For type 1 diabetes, no consistent associations between biomarkers of subclinical inflammation and cardiac autonomic dysfunction were observed.

    The first novel finding is the association between increased serum IL-18 and cardiac autonomic dysfunction, which was reflected by inverse associations between IL-18 levels and Valsalva ratio and max/min 30:15 ratio. This pattern suggests a link between elevated IL-18 and impaired efferent vagal activity. Spectral analysis suggests that the parasympathetic nervous system (PSNS) is a major contributor to both parameters related to IL-18.2 Furthermore, our data suggest a positive association of IL-18 with LF/HF ratio, which points to a shift of sympathovagal balance towards sympathetic predominance at higher IL-18 levels, although the concept of sympathovagal balance remains a matter of debate.

    IL-18 belongs to the IL-1 family of cytokines with mainly proinflammatory properties. With respect to diabetes-related complications, a recent meta-analysis showed that high IL-18 levels are associated with an increased risk for non-fatal myocardial infarction or death from coronary heart disease independently of conventional cardiovascular risk factors.11 IL-18 has also been implicated in neuroinflammation and pain.12 ,13 Underlining the biological plausibility of our finding, IL-18 release can be downregulated by the PSNS under certain experimental condition.12 Thus, IL-18 may be directly linked to vagal suppression and sympathetic predominance.

    The second novel findings are the inverse associations of total and HMW adiponectin levels with VLF power, suggesting a link to reduced sympathetic activity as opposed to IL-18.2 Interestingly, the direction of association, that is, increased serum adiponectin being related to diminished sympathetic tone, is the same for the adiponectins as for hsCRP, IL-6 and the soluble cell adhesion molecules. Our data are in line with one report of a positive association between adiponectin and CAN,14 but differ from other studies that found direct associations between adiponectin and sympathetic activation.7

    The inverse association between adiponectin and HRV is remarkable because adiponectin is often considered as a protective adipokine against diabetes-related complications in rodents.15 However, positive relationships between adiponectin and microvascular and macrovascular risk have previously been demonstrated in patients with type 2 diabetes.16 ,17 Moreover, DSPN was related to increased adiponectin in two studies.14 ,17 Also in line with our finding is one report of an inverse association between sympathetic activity and release of adiponectin in patients with essential hypertension.18 Although the mechanistic basis for these observations remains uncertain, it has been hypothesised that increased adiponectin levels may result from counterregulating proinflammatory and/or metabolic insults and should therefore be interpreted as an indirect risk indicator rather than genuine risk factor of diabetic complications.19 Taken together, the aforementioned findings suggest that hyperadiponectinaemia and diminished HRV may be pathophysiologically linked in individuals with type 2 diabetes.

    CRP and IL-6 were found to be inversely related to HRV with possibly more evidence for links with parasympathetic rather than sympathetic activity.6 ,20 However, the interpretation of previous reports may be difficult due to insufficient adjustment for confounding factors such as obesity,6 and given that observations from non-diabetic study samples may not be generalisable to individuals with diabetes.

    We report here that although increased levels of CRP and IL-6 were inversely associated with multiple HRV indices of both parasympathetic and sympathetic tone, these associations were confounded by anthropometric, metabolic and lifestyle variables. Three previous studies addressed the potential link between CRP and/or IL-6 and cardiac autonomic dysfunction in individuals with type 2 diabetes and came to conflicting results.14 ,21 ,22 However, these studies were conducted in smaller samples with longer diabetes duration. Moreover, these three studies were performed in East Asia. This latter point is relevant because East Asians have on average much lower circulating CRP levels than Europeans.

    This study also found multiple associations of sICAM-1 and E-selectin with both lower parasympathetic and sympathetic tone. However, the two soluble adhesion molecules differed in their sensitivity to confounding. Serum sICAM-1 showed independent inverse associations with both measures of sympathetic and vagal tone, whereas sE-selection was inversely associated with a measure of vagal activity after full adjustment. Noteworthy, circulating levels of both soluble adhesion molecules did not correlate with diminished HF and/or LF power after adjustment in a previous population-based study including only 13% of individuals with diabetes.23 To our knowledge, up to this point comparable studies in patients with diabetes have not been performed so far.

    In contrast to type 2 diabetes, we did not observe any consistent associations between biomarkers of subclinical inflammation and cardiac autonomic function in individuals with recent-onset type 1 diabetes. Our study was not designed and adequately powered to thoroughly investigate the aspect of differences in these associations between diabetes types. However, our data as visualised in figures 1 and 2 provide first preliminary evidence (to be corroborated in further studies) that the virtual absence of significant findings for type 1 diabetes cannot solely be explained by statistical power limitations and may have pathophysiological causes. Both subgroups differed in age, obesity and a range of associated factors, but these differences reflect the differences in pathogenesis of both diabetes types in adulthood. Associations between IL-6, but not hsCRP, and diminished HRV were observed in one study,24 whereas a different study found that CRP was related to reduced HRV in type 1 diabetes.25 However, the mean diabetes duration in these studies was 14 and 22 years, respectively. Taken together, we can speculate that an association between subclinical inflammation and cardiac autonomic dysfunction may not be very pronounced in younger adult individuals with recent-onset type 1 diabetes, but may develop with increasing age and duration of diabetes. The difference between type 2 diabetes and type 1 diabetes in our study supports the hypothesis that risk factors and pathomechanisms for DSPN may differ between both diabetes types.26

    Most studies that investigated the relationship between inflammation-related biomarkers and HRV hypothesised that the ANS modulates the immune system. This hypothesis is supported by anti-inflammatory effects mediated by vagal activity in conditions of pathogen invasion and tissue injury.4 However, the relationship between the ANS and chronic, subclinical inflammation, which is characteristic of obesity and type 2 diabetes, is less clear. An association between low baseline HF power and subsequent increases in CRP was observed in a small sample of healthy adults.27 If the ANS represented the dominant driver in this relationship, we would also expect associations in the subgroup with type 1 diabetes. The notion of a bidirectional link between inflammation and the ANS is supported by studies showing that the induction of proinflammatory activities in humans reduces HRV.28 We are aware of only one prospective study which addressed this issue and found a positive association between CRP and changes in HF power.29 But even if the precise mechanistic link remains to be explored, both biomarkers of subclinical inflammation and cardiac autonomic dysfunction represent risk factors for the development of cardiovascular outcomes and mortality in type 2 diabetes. Therefore, further studies should address the directionality of these associations and whether both risk factors act additively or even synergistically in the development of macrovascular complications. These studies should not be limited to individuals with diabetes, because the prevalence of CAN may be increased in those with prediabetes.30 Given that both alterations of HRV and subclinical inflammation are modifiable risk factors responding to lifestyle-based and pharmacological interventions,2 a better understanding of the associations reported here might help to better understand and reduce cardiovascular risk in individuals with type 2 diabetes.

    Strengths and limitations

    Strengths of the present work are the comprehensive assessment of cardiac autonomic dysfunction and subclinical inflammation and the unique study sample consisting of a relatively homogeneous group of individuals with recent-onset type 1 and type 2 diabetes. This study population allows the analysis of associations between subclinical inflammation and cardiac autonomic dysfunction before the onset of potentially confounding comorbidities.

    The main limitation is the cross-sectional rather than prospective study design of all analyses presented here, so that the temporal relationship between subclinical inflammation, covariables in regression analyses and development of cardiac autonomic dysfunction cannot be resolved at present. As a consequence of the present design, subgroups with type 1 and type 2 diabetes could not be equally sized, entailing differences in statistical power. We did not adjust for multiple testing, but the substantially higher than expected number of significant findings for type 2 diabetes points towards real associations between biomarkers of inflammation and cardiac autonomic dysfunction in this subgroup. Due to the selection of our study sample, our results are not representative of the general population with diabetes.

    Conclusion

    In summary, biomarkers of subclinical inflammation were associated with cardiac autonomic dysfunction in recent-onset type 2 diabetes. Higher serum IL-18 and E-selectin levels were associated with impaired vagal activity, whereas higher adiponectin showed associations with reduced sympathetic activity, while higher sICAM-1 levels were associated with both. Inverse associations between hsCRP and IL-6 were explained by confounding factors. These findings may improve our understanding of the pathophysiology leading to CAN and may have implications for screening or assessment of prognosis.

    Key messages

    What is already known on this subject?

    • Studies have indicated that biomarkers of subclinical inflammation and cardiovascular autonomic dysfunction are associated in non-diabetic individuals, but data on inflammation and cardiac autonomic dysfunction in individuals with diabetes are very scarce.

    What might this study add?

    • This study showed that in type 2 diabetes, higher serum interleukin-18 and E-selectin levels were associated with impaired vagal activity, whereas higher adiponectin showed associations with reduced sympathetic activity, while higher levels of soluble intercellular adhesion molecule-1 were associated with both. No consistent associations were observed in type 1 diabetes.

    How might this impact on clinical practice?

    • These findings implicate a diabetes type-specific role for inflammatory processes in the development of cardiac autonomic dysfunction, which may have implications for prevention and treatment.

    Acknowledgments

    The authors appreciate the voluntary contribution of all study participants. They also thank the staff of the Clinical Research Center of the German Diabetes Center (DDZ, Düsseldorf) for excellent technical assistance and taking care of the patients. The authors further thank Ulrike Partke, Gabi Gornitzka, Margret Behler and Maria Schroers-Teuber, DDZ, for excellent technical assistance.

    References

    Footnotes

    • MR and DZ contributed equally.

    • Collaborators The German Diabetes Study (GDS) Group consists of M. Roden (speaker), A.E. Buyken, J. Eckel, G. Geerling, H. Al-Hasani, C. Herder, A. Icks, J. Kotzka, O. Kuss, E. Lammert, J. Lundbom, K. Müssig, P. Nowotny, W. Rathmann, J. Szendroedi, D. Ziegler and their coworkers who are responsible for the design and performance of the GDS.

    • Contributors CH designed the study, researched data, contributed to data analysis, interpreted the data and wrote the manuscript. IS designed the study, performed the statistical analysis, interpreted the data, contributed to discussion and critically reviewed the manuscript. BN researched data and critically reviewed the manuscript. MC-K researched data, contributed to discussion and critically reviewed the manuscript. KS contributed to data analysis, contributed to discussion and critically reviewed the manuscript. PN, AS and SP researched data and critically reviewed the manuscript. JMK contributed to the statistical analysis, interpreted the data, contributed to discussion and critically reviewed the manuscript. KM and JS researched data, contributed to discussion and critically reviewed the manuscript. MR designed the study, contributed to discussion, critically reviewed and edited the manuscript. DZ designed the study, researched data, interpreted the data, contributed to discussion, critically reviewed and edited the manuscript. All authors approved of the final version of the manuscript.

    • Funding The German Diabetes Study was initiated and is performed by the DDZ, which is funded by the German Federal Ministry of Health (BMG) and the Ministry of Innovation, Science, Research and Technology (MIWF) of the State North Rhine-Westphalia. This study was supported in part by a grant from the German Federal Ministry of Education and Research (BMBF) to the German Center for Diabetes Research (DZD e.V.) and by a grant of the Helmholtz Alliance Imaging and Curing Environmental Metabolic Diseases.

    • Competing interests None declared.

    • Ethics approval Ethics committee of Heinrich Heine University, Düsseldorf, Germany.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.