Post hoc analysis of a cohort study in type 2 diabetes and healthy controls studying temporal changes in myocardial function in patients with type 2 diabetes and non-diabetic controls (13). The study was performed in compliance with the Declaration of Helsinki. All participants gave informed written consent, and the study protocol was approved by a regional research ethics committee (H-19024534).

In 2013-2014, 60 persons with type 2 diabetes and 30 age- and sex-matched healthy controls were recruited at Steno Diabetes Center Copenhagen (7). In the type 2 diabetes group, 30 had normoalbuminuria (<30mg/24h) and 30 had albuminuria (≥30mg/24h). Participants were between 35 and 80 years of age and free of overt cardiovascular disease. In 2019-2020, all living participants (n=82) were invited for a follow up visit. A total of 48 participants from the original cohort participated. Reasons for not participating were severe illness (n=5), no response (n=8) and not interested (n=21). Cardiovascular autonomic function data was missing for six participants (5 with type 2 diabetes and 1 control) due to lack of compliance or malfunction of the equipment at the time of measurement. Thus, the present post hoc analysis includes 42 participants (24 with type 2 diabetes and 18 controls) who had a complete dataset on cardiovascular autonomic function at baseline and on the study outcomes at baseline and follow up ( Figure 1 ).

Laboratory variables included HbA_1c, lipid profile and plasma creatinine measured by standard methods. Upper arm blood pressure was measured in sitting position on the right arm after 10-minutes rest with a validated blood pressure monitor and the mean of three consecutive measurements was calculated.

Participants were advised to abstain from hard physical activity 24 hours before the examination. All tests of cardiovascular autonomic function were performed between 8:00 A.M. and 2:00 P.M. in a quiet examination room. Three standard cardiovascular autonomic reflex tests (CARTs) and 5-min resting heart rate recordings for calculating time- and frequency-domain heart rate variability indices were obtained once (3, 14). Parameters of cardiovascular autonomic function were obtained using the Vagus Device (Medicus Engineering, Aarhus, Denmark) (15). The CARTs included the heart rate response to: 1) standing (30:15); 2) deep breathing (E:I); and 3) the Valsalva test. The 30:15 ratio, E:I ratio and Valsalva tests are measures mainly of parasympathetic function (1). The CARTs were evaluated using age-related reference intervals (16). CAN was defined using the American Diabetes Association criteria and participants with two or three pathological CARTs were classified as having CAN (17). From the 5-min resting heart rate recordings, time-domain heart rate variability indices were derived: the standard deviation of normal-to-normal (SDNN) intervals and the root mean square of successive differences (RMSSD). Frequency-domain heart rate variability indices were calculated using fast Fourier transformation and included low frequency (LF) power (0.04–0.15 Hz), high frequency (HF) power (0.15–0.4 Hz), and the total (Total) power (0.4 Hz). The ratio of LF-to-HF (LF/HF Ratio) was also calculated. Where RMSSD and HF power reflects parasympathetic activity, the SDNN, LF power and Total power are measures of combined sympathetic and parasympathetic activity.

The estimated glomerular filtration rate (eGFR) was calculated by the CKD-EPI equation (18). The urine albumin excretion was measured at the baseline visit in two 24 hours urine collections and estimated as the urine albumin excretion rate and measured at the follow up visit in three consecutive morning urine samples and estimated as the urine albumin creatinine ratio. Cardiac ⁸²Rb Positron emission tomography computed tomography (PET/CT) imaging provides a quantitative measurement of the myocardial perfusion (ml/min/g of tissue) during stress and rest. This enables calculation of the myocardial flow reserve (19). The cardiac ⁸²Rb PET/CT scans were performed using a hybrid PET/CT scanner in 3D mode (Siemens Biograph mCT 128, Siemens, Munich, Germany) after administration of 1,100 MBq ⁸²Rb (CardioGen-82, Bracco Diagnostics, Monroe Township, NJ, USA). The scans were performed at rest and at stress (adenosine infusion to induce maximum myocardial hyperemia). The scans were performed with similar equipment, protocol and analyzed using identical software (syngo.via VB20A) by the same single observer at the baseline and follow up visits. The coronary artery calcium score was calculated by the method described by Agatston et al. as the sum of coronary artery calcium content in the three main coronary arteries (20) using semi-automated commercially available software (Syngovia 4.0).

Clinical characteristics of the participants are presented as n (%), mean ± standard deviation (SD) or if skewed distributed, as medians with interquartile range [quartile 1 to quartile 3]. The variables with skewed distributions were log transformed (natural logarithm) before analyses and normal distribution was obtained. A value of 1 was added to the coronary artery calcium score before log transformation, since the skewed distribution included values of zero. We applied the Chi-squared test or Fisher’s exact test for categorical data and unpaired Student’s t-tests for continuous variables to analyze differences between the participants with type 2 diabetes and the healthy controls. Cross-sectional associations in the total population were assessed in the baseline study as already publiched (7). We repeated these analyses in the present sub-population, using unadjusted linear regression models to evaluate the association between baseline cardiovascular autonomic function parameters and baseline levels of outcomes. Next, to control for possible confounders, we adjusted the significant associations for sex, baseline age, smoking, systolic blood pressure, heart rate, HbA_1c and body mass index. Change from baseline to follow up was analyzed using paired Student’s t-tests. Changes in the outcomes were expressed per year as the change from the baseline to the follow up visit, divided by the individual follow up time in years. Unadjusted and adjusted linear regression models were applied to assess associations between the baseline cardiovascular autonomic function parameters (30:15, E:I, Valsalva, log (SDNN), log (RMSSD), log (LF), log (HF), log (Total) and log (LF/HF Ratio)) and yearly changes in the outcomes. Adjusment included sex, baseline age, smoking, systolic blood pressure, heart rate, HbA_1c, body mass index and the baseline value of the given outcome. All participants were pooled in the linear regression analyses and standardized regression coefficients are reported. We used SAS Enterprise Guide, version 7.15 for analyses. Two-sided p values < 0.05 were considered statistically significant. Due to the exploratory nature of this post hoc analysis, we did not perform a sample size calculation for this study.

Age and sex distributions were equal in the 24 persons with type 2 diabetes and the 18 healthy controls. The persons with type 2 diabetes had a higher body mass index and heart rate and lower LDL and total cholesterol compared with controls. The participants with type 2 diabetes had a median [Q1 to Q3] diabetes duration of 11 [5 to 19] years, mean HbA_1c of 53 ± 10 mmol/mol, eGFR of 79 ± 22 ml min⁻¹ 1.73 m⁻² and 11 (46%) had albuminuria ( Table 1 ).

All measures of cardiovascular autonomic function at baseline, except the 30:15, the Valsalva and the LF/HF ratio, were lower in the participants with type 2 diabetes compared with the healthy controls. Only one participant (with type 2 diabetes) had CAN ( Table 1 ). Baseline associations between cardiovascular autonomic function and measures of heart and kidney function are shown in Table S1 . A higher (more beneficial) 30:15 ratio was associated with a higher eGFR. Higher values of all measures of cardiovascular autonomic function, except the E:I ratio, RMSSD, high frequency power and the LF/HF ratio, were associated with lower urine albumin excretion rate. Higher values of all measures of cardiovascular autonomic function, except the Valsalva ratio, High Frequency power and LF/HF ratio, correlated with higher baseline myocardial flow reserve. Finally, higher values of all cardiovascular autonomic function parameters, except the LF/HF ratio, were associated with lower baseline coronary artery calcium score. None of the associations at baseline remained significant after confounder adjustment ( Table S1 ).

Baseline and follow-up values of outcomes are shown in Table 2 . The mean ± SD follow up time was 6.2 ± 0.2 years for participants with type 2 diabetes and 6.3 ± 0.2 years for the healthy controls. The mean change per year in eGFR was -1.2 ± 1.3 ml min⁻¹ 1.73 m⁻² for participants with type 2 diabetes and -0.7 ± 0.8 ml min⁻¹ 1.73 m⁻² for healthy controls, with no significant difference between groups (p=0.16). The median change per year in urine albumin was -0.9 [-7.3 to 0.3] mg g^-1 for the type 2 diabetes group and -0.2 [-0.5 to 0.1] mg g^-1 for the healthy controls, with no significant difference between groups (p=0.34). There was no difference between groups in the yearly change in myocardial flow reserve. The median change per year in coronary artery calcium score was significantly larger in the type 2 diabetes group (p=0.003) ( Table 2 ).

Associations between cardiovascular autonomic function at baseline and changes in heart and kidney function are shown in Table 3 . A lower response in heart rate to the Valsalva test, reflecting impaired cardiovascular autonomic function, was associated with a larger decline in eGFR in the unadjusted model (β per 1 SD = 0.37 ml min⁻¹ 1.73 m⁻² increase per year; p=0.04), but not after adjustment (p=0.12). No other associations between cardiovascular autonomic function at baseline and changes in eGFR were found. No associations with changes in urine albumin were found. A higher response in heart rate to standing (30:15), reflecting better cardiovascular autonomic function, was associated with a larger decline in myocardial flow reserve in the unadjusted model (β per 1 SD = 0.04 decrease per year; p=0.02) and after adjustment (p=0.02). No other associations between cardiovascular autonomic function at baseline and changes in myocardial flow reserve were demonstrated. A higher response in heart rate to the Valsalva maneuver was associated with a larger increase in coronary artery calcium score (β per 1 SD = 0.34 increase per year; p = 0.02), but the association became insignificant after adjustment (p = 0.16). There were no other associations between cardiovascular autonomic function at baseline and changes in coronary artery calcium score ( Table 3 ).

Changes in medical therapy from baseline to follow up are shown in Table S2 . In the type 2 diabetes group, a total of 21 (88%) had changes in their diabetes therapy from baseline to follow up (insulin; sodium-glucose cotransporter 2 inhibitor (SGLT2i); glucagon-like peptide-1 receptor agonist (GLP1-RA); metformin) which might have influenced our results. To take this into account, we adjusted our analyses further for change in diabetes therapy from baseline to follow up (yes vs. no). Results were confirmatory. To account for the influence on beta-blockers on the cardiovascular autonomic function tests, we performed a sensitivity analysis, excluding the two participants treated with a beta-blocker at baseline. Results were confirmatory.

The investigated population was small, and this introduces a risk of false negative findings due to lack of power. Only 42 out of the original 90 persons participated in the follow up. The participants in the follow-up examination were younger, had a higher myocardial flow reserve, better kidney function and a lower coronary artery calcium score at the baseline examination compared with non-participants (13). The baseline characteristics of the full original cohort of 90 subjects and the follow up cohort of 42 subjects are presented in Table S3 . This introduces a considerable risk of selection- and survival bias which weakens our chance to find associations. Thus, our results should be interpreted as exploratory and with this limitation in mind. When conducting a large number of tests as we have done there is a risk of type 1 errors. The myocardial flow reserve did not change in neither of the groups during follow up which makes it difficult to study associations between autonomic function and change in myocardial flow reserve. It is a limitation that participants were not abstinent from medication before measurements of cardiovascular autonomic function at baseline. Moreover, we did not evaluate the cardiovascular autonomic function at follow up with the Vagus device as in the baseline examination and are therefore not able to relate changes in cardiovascular autonomic function to the outcomes.