Determinants and Prognostic Value of Cardiovascular Autonomic Function in Coronary Artery Disease Patients With and Without Type 2 Diabetes

Cardiovascular autonomic function, which can be assessed, for example, by measuring postexercise heart rate recovery (HRR), and by analyzing heart rate (HR) variability and HR turbulence (HRT) from ambulatory electrocardiogram (ECG) recordings, has been shown to be altered among patients with ischemic heart disease (1–3). These three methods describe in part different aspects of autonomic regulation. The decline of HR within the first minute after cessation of exercise is caused mainly by parasympathetic reactivation, with sympathetic withdrawal being more prominent 2 min after exercise (4). Therefore, HRR immediately after exercise can be considered to reflect short-term vagal reactivity. The SD of normal R-R interval (SDNN) reflects global 24-h HR variability, including all the cyclic components responsible for fluctuations in sinus rate, and can be considered a marker of long-term vagal tone (5,6). Long-term HR variability is also partly influenced by physical activity at the time of recordings (7,8). HRT describes short-term fluctuations in sinus rhythm after spontaneous ventricular premature complexes (9,10). Overall, the HRT measurement provides an indirect assessment of baroreflex sensitivity (9).

Previous studies have suggested that autonomic dysfunction among cardiac patients may be explained by abnormalities in cardiac structure and/or function, for example, angiographic severity (11,12), extension of myocardial infarction (13), and the severity of left ventricular dysfunction (14), as well as by markers of inflammation (15,16). However, in the population without cardiac disease, autonomic function has been observed to be associated with several cardiac risk factors, such as obesity (17,18), hypercholesterolemia (19), reduced physical activity (20–22), and low physical fitness (22–24). Type 2 diabetes, which is a common comorbidity among coronary artery disease (CAD) patients, is also associated with abnormal cardiovascular regulation (25). The role of these features as determinants of the autonomic function of patients with CAD is not well known. Therefore, we hypothesized that autonomic dysfunction in CAD patients may be related not only to abnormal cardiac function and/or structure, but also to other cardiovascular risk factors. Moreover, we assumed that type 2 diabetes has some effect on autonomic dysfunction among CAD patients. In addition, we wanted to study whether autonomic markers can provide short-term prognostic information independent of other risk variables in the stable CAD population without a recent acute coronary event.

This study is part of the larger ARTEMIS (Innovation to Reduce Cardiovascular Complications of Diabetes at the Intersection) Study, which is being conducted in the Division of Cardiology of Oulu University Hospital and is registered at ClinicalTrials.gov (record 1539/31/06; identifier number NCT01426685). The goals of the ARTEMIS Study are to compare cardiovascular autonomic markers, ECG markers, and biomarkers between CAD patients with and without type 2 diabetes, and to assess the prognostic value of these markers in predicting cardiovascular events among CAD patients with and without type 2 diabetes during 2- and 5-year follow-ups. The study population of this substudy consisted of 530 patients with stable CAD and 530 CAD patients with type 2 diabetes who were recruited from the consecutive series of patients who had undergone coronary angiography in the Division of Cardiology of Oulu University Hospital 3–6 months before inclusion in the assessment of their risk profile. More than 80% of the patients had undergone coronary intervention 3–6 months before entering the study, and almost half of the patients had experienced an acute coronary event >3 months before enrollment into the risk profile study (Table 1). The patients with and without type 2 diabetes were matched in terms of age, sex, left ventricular ejection fraction, history of myocardial infarction, and angiographic severity of CAD. Patients with age <18 years or >85 years, impaired glucose tolerance or impaired fasting glucose, New York Heart Association class IV, a permanent pacemaker or implantable cardioverter defibrillator, planned implantable cardioverter defibrillator implantation, or end-stage renal failure needing dialysis were excluded from this study, as well as patients who had a life expectancy of <1 year or who were psychologically or physically (due to any other illness) unfit for participation in the study. The study was performed according to the Declaration of Helsinki, the local committee of research ethics of the Northern Ostrobothnia Hospital District approved the protocol, and all the subjects gave written informed consent.

Type 2 diabetes was verified according to the World Health Organization criteria (26). Patients who had a fasting capillary plasma glucose level of ≥7.0 mmol/L and/or a 2-h postload value of ≥12.2 mmol/L in the oral glucose tolerance test, or were receiving hypoglycemic medication based on a prior diagnosis of diabetes were classified as having type 2 diabetes. Left ventricular systolic function was assessed by two-dimensional echocardiography (Simpson method), and diastolic function was assessed by tissue Doppler echocardiography. Blood samples and urine samples were obtained for the analysis of renal function, microalbuminuria, and levels of inflammation markers, blood lipids, plasma glucose, and glycated hemoglobin A_1c (HbA_1c). Blood pressure was measured with the patient in a supine position after a 10-min resting period. All laboratory measurements and exercise tests were performed after a 12-h overnight fast, and the patients were instructed to avoid smoking before the tests.

The patients performed an incremental symptom-limited maximal exercise test on a bicycle ergometer for assessment of maximal exercise capacity. The test was started at 30 W, and the work rate was increased by 15 W in men and 10 W in women every minute until voluntary exhaustion. The maximal workload was calculated as the workload at the end of the test, and maximal exercise capacity was calculated from the workloads during the last minute of the test. The patients moved to a supine position within 30 s after the cessation of exercise, and no cooldown period was allowed. HRR was calculated as the reduction in HR from the rate at maximal exercise to the rate 1 min after the cessation of exercise. A 15-lead ECG (CAM-14; GE Healthcare, Freiburg, Germany) was collected during the exercise and 10 min after the exercise.

A 24-h ambulatory ECG was recorded with a digital Holter recorder (Medilog AR12; Huntleigh Healthcare, Newtownabbey, U.K.) with an accuracy of 4 ms and was saved on a computer for further analysis with custom-made software (HEARTS software; Heart Signal Oy, Oulu, Finland). From the total of 1,060 patients, 27 patients did not undergo the 24-h ECG recording and 66 patients were excluded from the analysis as a result of a large amount of technical and biological disturbances (e.g., periods of atrial fibrillation). In total, 496 CAD patients and 471 CAD patients with type 2 diabetes were included in the analyses of 24-h ECG recordings. A HRT was measured from 423 CAD patients and 414 CAD patients with type 2 diabetes.

To exclude all undesirable beats, the R-R intervals were edited by visual inspection using the interpolation method. This method replaces edited, removed R-R intervals with a local average of the previous accepted normal R-R intervals (27). SDNN was used as a time-domain measure of HR variability. Postectopic HRT was analyzed according to a previously published method (10). The HRT slope was defined as the maximum slope of the regression line assessed over any sequence within five subsequent sinus R-R intervals (RR) during the first 15 sinus beats after an ectopic beat, as described earlier (28). During the measure of cardiovascular autonomic function by 24-h ambulatory ECG recording, patients were advised to carry on with their normal daily routine without any limitation.

The follow-up was 2 years after the patient’s first measurement in the ARTEMIS Study in Oulu University Hospital. The end point was defined as a combination of cardiovascular complications, including cardiovascular death, acute coronary event, stroke, and hospitalization for heart failure. The follow-up data were collected from patient records of Oulu University Hospital and the mortality statistics of Statistics Finland and the Causes of Death Register. The patients who were alive were also contacted by telephone after 2 years. The follow-up was completed for 427 CAD patients and 525 patients with type 2 diabetes until January 2013, at the time the query was run.

A Kolmogorov-Smirnov z test was performed for all variables to assess the assumption of Gaussian distribution. Because some of the data were skewed, a Spearman nonparametric test was used to assess the correlations among HRR, SDNN, HRT, and potential predictors, including demographic characteristics, laboratory measurements, echocardiographic parameters, markers of inflammation, and exercise capacity. Linear regression analyses with stepwise regression analyses were used to construct the predictive regression models for HRR, SDNN, and HRT. The group variable (type 2 diabetes) and all the significant demographic, laboratory, and echocardiographic variables, which correlated with autonomic markers in the Spearman correlation analyses (P < 0.05), and medication were included as covariates in the linear regression analyses. SDNN and HRT slope had non-Gaussian distribution and were transformed into natural logarithms before the parametric statistical tests.

Because the cutoff values of autonomic markers have not been previously defined for this type of population with stable CAD, the optimal cutoff value for each of the measures of autonomic function was defined from receiver operating characteristic curves as the maximum sum of sensitivity and specificity below the median, with a sensitivity of at least 20% using a composite of cardiovascular complications as the end point. Univariate Cox regression analysis was used to obtain values for hazard ratio with 95% CIs for categorized autonomic markers. Thereafter, multivariate Cox regression analysis was performed, including covariates that were associated with a composite cardiovascular end point: sex, age, presence of type 2 diabetes or HbA_1c, high-sensitivity C-reactive protein (hs-CRP), triglycerides, microalbuminuria, left ventricular systolic and diastolic function, and exercise capacity. Cox regression analyses were also performed for the subgroup of CAD patients with (n = 525) and without (n = 427) type 2 diabetes. The statistical analyses were performed using SPSS software, versions 19.0 and 21.0 (SPSS Inc., Chicago, IL). A P value <0.05 was considered statistically significant.

The demographic characteristics and medication of the study population are presented in Table 1.

All significant correlations between HRR and other potential variables are presented in Table 2. In CAD patients with and without type 2 diabetes, HRR had the strongest univariate correlation with exercise capacity (r = 0.55), age (r = −0.41), fasting plasma glucose (r = −0.29), and left ventricular diastolic function (r = −0.29, P < 0.001 for all). CAD patients with type 2 diabetes had reduced HRR compared with their counterparts without diabetes (23 ± 11 vs. 29 ± 11 bpm, P < 0.001). Similarly, patients who used calcium antagonists had reduced HRR compared with those who did not take these drugs (23 ± 11 vs. 27 ± 12 bpm, P < 0.001). When all factors that correlated with HRR were entered into a stepwise linear regression model (R = 0.60, P < 0.001), exercise capacity (R = 0.34, P < 0.001), age (R = −0.24, P < 0.001), type 2 diabetes (R = 0.09, P = 0.002), hs-CRP (R = −0.10, P < 0.001), left ventricular diastolic function (R = −0.07, P = 0.007), calcium antagonists (R = −0.06, P = 0.013), and triglycerides (R = −0.06, P = 0.024) emerged as the significant predictors of HRR for the CAD patients with and without type 2 diabetes.

Significant correlations between SDNN and other potential variables are presented in Table 2. In CAD patients with and without type 2 diabetes, SDNN had the strongest univariate correlation with exercise capacity (r = 0.40), fasting plasma glucose (r = −0.32), and HbA_1c level (r = −0.26, P < 0.001 for all). CAD patients with type 2 diabetes had reduced SDNN compared with their counterparts without diabetes (124 ± 41 vs. 149 ± 39 ms, P < 0.001). When all factors that correlated with SDNN were entered into a stepwise linear regression model (R = 0.48, P < 0.001), exercise capacity (R = 0.33, P < 0.001), type 2 diabetes (R = 0.16, P = 0.001), triglycerides (R = −0.11, P < 0.001), left ventricular mass index (R = 0.09, P = 0.003), hs-CRP (R = −0.09, P = 0.004), BMI (R = 0.09, P = 0.008), depression score (R = −0.08, P = 0.013), and albuminuria (R = −0.07, P = 0.028) were the significant predictors of SDNN in CAD patients with and without type 2 diabetes.

Significant correlations between HRT slope and other potential variables are presented in Table 2. In the CAD patients with and without type 2 diabetes, there was a clear positive correlation between HRT slope and exercise capacity (r = 0.23, P < 0.001), and negative correlations among HRT slope and age (r = −0.18), fasting plasma glucose (r = −0.18), and HbA_1c (r = −0.17, P < 0.001 for all). CAD patients with type 2 diabetes had reduced HRT slope compared with their counterparts without diabetes (7.30 ± 7.85 vs. 8.94 ± 8.37 ms/RR, P < 0.001). Similarly, patients who used calcium antagonists had reduced HRT slope compared with those who did not use these drugs (6.74 ± 6.90 vs. 8.62 ± 8.51 ms/RR, P = 0.001). When all factors that correlated with HRT slope were entered into a stepwise linear regression model (R = 0.38, P < 0.001), exercise capacity (R = 0.13, P = 0.001), left ventricular systolic (R = 0.18, P < 0.001) and diastolic function (R = −0.11, P = 0.002), calcium antagonists (R = −0.07, P = 0.026), depression score (R = −0.09, P = 0.008), age (R = −0.12, P = 0.002), and HDL cholesterol level (R = 0.10, P = 0.005) were significant predictors of HRT slope in CAD patients with and without type 2 diabetes.

During the follow-up period, 127 patients (13%) reached a composite end point, including 16 patients (2%) with cardiovascular death, 78 patients (8%) with acute coronary events, 29 patients (3%) with strokes, and 16 patients (2%) hospitalized for heart failure. In the receiver operating characteristic analysis, areas under the curve were 0.606 (P = 0.001), 0.552 (P = 0.072), and 0.564 (P = 0.020), respectively, for HRT, SDNN, and HRR. In univariate analyses with optimal cutoff points, both blunted HRT slope (<3.4 ms/RR), SDNN (<110 ms), and HRR (<21 bpm) were powerful predictors of a composite end point (hazard ratio 2.1 [95% CI 1.4–3.2], P < 0.001; 1.9 [1.3–2.7], P = 0.001; 1.6 [1.1–2.2], P = 0.012, respectively) in the total population. The sensitivity, specificity, and positive and negative predictive accuracy for the optimal cutoff points were 42, 76, 21, and 90% for HRT; 39, 76, 19, and 89% for SDNN; and 44, 67, 17, and 89% for HRR. However, after multivariate adjustment, none of the autonomic markers remained predictive of a composite end point (Table 3).

Among CAD patients with type 2 diabetes, cardiovascular end points occurred in 80 patients (15%), including 10 cardiovascular deaths (2%), 43 acute coronary events (8%), 23 strokes (4%), and 14 hospitalizations for heart failure (3%). A composite end point was predicted by blunted HRT slope (n = 133), SDNN (n = 172), and HRR (n = 239) in univariate analyses (hazard ratio 2.1 [95% CI 1.3–3.4], P = 0.003; 2.0 [1.2–3.1], P = 0.005; and 1.7 [1.1–2.6], P = 0.020, respectively). The sensitivity, specificity, and positive and negative predictive accuracy for the optimal cutoff points were 50, 71, 24, and 89% for HRT; 51, 66, 21, and 89% for SDNN; and 58, 57, 19, and 88% for HRR. After multivariate adjustment, hs-CRP remained as the only independent predictor of a composite end point (Table 3, Fig. 1).

Among patients with CAD alone, cardiovascular complications occurred in 47 patients (11%) (P = 0.05 compared with CAD patients with type 2 diabetes), including 6 cardiovascular deaths (1%), 35 acute coronary events (6%), 6 strokes (1%), and 2 hospitalizations for heart failure (1%). None of the autonomic markers predicted cardiovascular end points in univariate or multivariate analyses in CAD patients without type 2 diabetes (Table 3, Fig. 1).

The current study shows that low exercise capacity, age, type 2 diabetes, and hs-CRP levels are the most important determinants of cardiovascular autonomic function in patients with stable CAD with and without type 2 diabetes. HRT was also influenced by left ventricular systolic function. Autonomic function assessed by HRR, HR variability, or HRT predicted short-term cardiovascular end points in univariate analyses in CAD patients with type 2 diabetes, but not in their counterparts without type 2 diabetes. However, autonomic measures did not provide independent short-term prognostic information, even in the patients with type 2 diabetes, after adjusting for clinical, demographic, and echocardiographic risk variables, including a marker of low-grade inflammation, which remained as the independent prognostic marker.

Many studies have found an association between autonomic function and exercise capacity or physical fitness in the general population and in various patient groups, including cardiac patients (22,23,29–31). Both impaired HRR and HR variability are related to reduced exercise capacity. The results of our study are consistent with previous studies, and indicate that abnormalities in HRR and HR variability strongly correlate with low exercise capacity in CAD patients. Moreover, we also found a similar association between HRT and exercise capacity, which has not been reported previously.

Autonomic dysfunction, assessed by blunted HRR, has been shown to be associated with an increased level of fasting plasma glucose in CAD patients without type 2 diabetes (32). In a previous study, we found that blunted HRR was more common among CAD patients with type 2 diabetes than in those without (31), and, similar to our results, Nonaka et al. (33) demonstrated that abnormal HRR is associated with type 2 diabetes in patients with suspected CAD. We also found in the current study, as expected, a significant association among autonomic function, and fasting plasma glucose and HbA_1c levels among CAD patients with and without type 2 diabetes. Moreover, type 2 diabetes was one of the strongest predictors of HRR and SDNN, but it did not predict HRT. Therefore, our present study confirmed the relationship between autonomic dysfunction and type 2 diabetes also among the patients with stable CAD.

Previous studies have suggested that autonomic dysfunction in CAD patients is associated with the angiographic severity and degree of coronary occlusion, but not with a history of myocardial infarction (11,12). The results regarding the association between left ventricular function and autonomic markers have been variable. Szydlo et al. (34) found an association between lower HR variability and left ventricular systolic dysfunction, which also seems to be the key factor influencing HRT parameters in CAD patients (35). However, some recent studies did not find significant correlations between decreased HRR and left ventricular systolic dysfunction (12,33). In the current study, we observed a weak correlation between a history of myocardial infarction and all three measurements of autonomic function, whereas left ventricular systolic function correlated only with the HRT.

In patients with CAD, depressive symptoms are associated with cardiac autonomic dysfunction (36). Also, inflammatory markers are inversely correlated with autonomic function among CAD patients (15,16). In our study, both depression score and hs-CRP level were negatively associated with all three autonomic markers, but depression score did not independently predict autonomic function in the multiple linear regression analyses.

Most previous association studies have usually used only one measure of the autonomic nervous system—either HRR or HR variability—whereas the determinants of HRT are less well studied. In the current study, cardiovascular autonomic function was measured by three different methods, which describe in part different aspects of autonomic regulation. Postexercise HRR can be considered to reflect short-term vagal reactivity, SDNN is a marker of long-term vagal tone, and HRT provides an indirect assessment of baroreflex sensitivity. Moreover, previous studies have mainly examined the association between autonomic function and only a few associated factors, whereas in the current study we have taken account of many potential determinants of autonomic function at the same time. Furthermore, we concentrated here on studying the short-term prognostic value of autonomic markers in patients with stable CAD, whereas previous studies have mainly focused on long-term prognostic value of postinfarction patients. It can be speculated that the risk profiles and their predictive value may change over time, especially after an acute coronary event. Recent studies have shown that autonomic function measured late after an acute myocardial infarction (AMI) predicts mortality better than measurements performed in the early postinfarction phase (28,37,38). Therefore, the analysis of factors predicting cardiovascular complications after stabilization of an acute coronary event may in fact provide more relevant clinical information.

The univariate analysis showed that all three autonomic markers predicted short-term cardiovascular complications in CAD patients with type 2 diabetes, but not in their counterparts without type 2 diabetes. This emphasizes the role of autonomic markers as predictors of short-term cardiovascular events, especially in patients with type 2 diabetes and stable CAD. In agreement with many similar studies in other populations (5,9), the positive predictive accuracy of HRT and HR variability remained at a relative low level also in patients with type 2 diabetes. However, the negative predictive values were high, suggesting that well-preserved autonomic function prevents the early occurrence of cardiovascular events in this group of patients. HRT had the highest positive predictive value and the highest hazard ratio for cardiovascular events. The results are in agreement with a previous substudy of a German postinfarction population, which showed that HRT was a strong predictor of mortality in patients with type 2 diabetes (39).

The prognostic value of autonomic function in CAD patients with type 2 diabetes did not remain statistically significant after multivariate adjustment, including hs-CRP, a marker of low-grade inflammation. Biasucci et al. (40) have investigated a prognostic role for hs-CRP in patients with acute coronary syndrome during a 1-year follow-up. Interestingly, their findings were opposite to the results of our study, which suggested that hs-CRP is strongly associated with death and AMI, especially in patients without type 2 diabetes, but not in patients with type 2 diabetes. However, Biasucci et al. (40) had enrolled a relatively small number of patients with unstable angina (n = 251), and hard events were observed in only seven patients with type 2 diabetes, whereas we report the prognostic power of hs-CRP for cardiovascular events in a larger population with type 2 diabetes and stable CAD.

The prognostic part of the current study was limited by a relatively short follow-up and a mixture of various end points. The low number of end points in the CAD patients without type 2 diabetes limits the generalizability of the results in terms of the negative predictive value of the autonomic markers. Similarly, the negative predictive value of these markers in multivariate analysis in patients with type 2 diabetes may be partly due to the small sample size and the short follow-up. However, we wanted specifically to study the short-term prognostic values of risk markers, which could provide insight into the therapeutic targets of the well-defined population of CAD patients who had recently undergone coronary angiography, most of whom were treated with a recent coronary intervention. In this respect, the independent prognostic power of hs-CRP is of potential interest in patients with type 2 diabetes, suggesting that therapeutic interventions to reduce inflammation might be of pivotal importance in attempting to reduce early cardiovascular complications of patients with type 2 diabetes and CAD. A larger sample size of the ongoing study will reveal the definite value of hs-CRP in this context.

It is notable that the present measurements of autonomic function were performed under continued prescribed medication, and that the majority of the patients (88%) were treated with β-blockers. However, the use of β-blockers did not independently predict autonomic function in the multiple linear regression analyses probably because of the low number of patients in the nonuser group. During the follow-up, cardiovascular end points occurred in 119 patients (14%) who were treated with β-blockers and in 8 patients (8%) who were not treated with β-blocker medication. Although the difference in the incidence of cardiovascular end points was not statistically significant, it is possible that the users of β-blockers may be at increased risk of short-term cardiovascular events. A large survey in patients with stable CAD without a recent myocardial infarction also showed that the composite end point, including cardiovascular death, nonfatal myocardial infarction, stroke, and hospitalization for atherothrombotic events or a revascularization procedure, was higher among the users versus nonusers of β-blocking drugs in the current era (41).

In conclusion, cardiovascular autonomic function in CAD patients is associated with several variables, including exercise capacity, age, inflammation, and left ventricular systolic function. Although autonomic dysfunction predicts short-term cardiovascular events among CAD patients with type 2 diabetes, it did not provide independent prognostic information after multivariate adjustment with other risk variables. The definite role of hs-CRP in risk stratification of CAD patients with type 2 diabetes needs further investigation.

The authors thank Polar Electro (Kempele, Finland) and Hur Oy (Kokkola, Finland) for technical and financial support. The authors also thank the registered nurses Pirkko Huikuri, Päivi Koski, Päivi Kastell, and Sari Kaarlenkaski, Oulu University Hospital, for help with the data collection.

Funding. This study was supported by grants from the Finnish Technology Development Centre (Helsinki, Finland).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.J.K. analyzed and interpreted data and wrote and reviewed the drafts of the article. A.M.K. contributed to the statistical analyses and reviewed the final draft of the article. A.J.H., O.-P.P., E.S.L., M.A.P., O.H.U., and P.S.M.H. contributed to the data collection and reviewed the final draft of the article. H.V.H. and M.P.T. contributed to the design of the ARTEMIS Study, guided the process of data analysis and interpretation, and read and reviewed the drafts of the article. M.P.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.