Impaired Autonomic Function Is Associated With Increased Mortality, Especially in Subjects With Diabetes, Hypertension, or a History of Cardiovascular Disease: The Hoorn Study

The present study is part of the Hoorn Study, a prospective study of glucose tolerance in a general population of Caucasian subjects (50–75 years of age) (11). The baseline examination was conducted in 1989–1992, as previously described in detail (15). Briefly, a random sample was drawn from the municipal population of Hoorn, the Netherlands. All 2,484 participants (71%), except those who were taking glucose-lowering medication, underwent 75-g oral glucose tolerance testing. A sample of 708 subjects, stratified by 2-h glucose values, age, and gender, was invited to undergo extensive examinations of diabetes-related complications, and 631 subjects (89%) participated. For the present analyses, we excluded subjects who had pacemakers (n = 2), who were taking antiarrhythmic medication (n = 17), or who had self-reported neurologic disease (n = 8), resulting in a study population of 605 subjects.

Fasting and 2-h plasma glucose levels were determined with a glucose dehydrogenase method (Merck, Darmstadt, Germany). Subjects were classified into glucose tolerance groups, based on the mean values of two oral glucose tolerance tests (16). A total of 282 individuals had normal glucose tolerance (NGT), 164 had IGT, 85 had newly diagnosed diabetes, and 74 had known type 2 diabetes, as defined by the use of a diet or glucose-lowering medication. HbA_1c was determined by ion-exchange high-performance liquid chromatography (Bio-Rad, Veenendaal, the Netherlands).

Blood pressure was measured in duplicate on two separate occasions on the right arm of seated subjects after at least 5 min of rest, by means of a random-zero sphygmomanometer (Hawksley-Gelman, Lancing, Sussex, U.K.). The averages of the four blood pressure readings were calculated. Hypertension was defined as a systolic blood pressure ≥160 mmHg, a diastolic blood pressure ≥95 mmHg, and/or the use of antihypertensives (17). These cutoff points were used because of the definition of hypertension at the time of the data collection. Use of antihypertensives was defined as current treatment with alpha blockers, beta blockers, calcium antagonists, angiotensin-converting enzyme inhibitors, diuretics, and/or a rest group including centrally acting antihypertensives.

Cardiovascular disease was defined as coronary artery, cerebrovascular, and/or peripheral arterial disease. Coronary artery disease was defined as self-reported history of myocardial infarction, coronary artery bypass grafting, or Minnesota codes 1-1 or 1-2 on electrocardiography. Peripheral arterial disease was defined as self-reported peripheral arterial reconstruction, nontraumatic limb amputation, and/or an ankle brachial pressure index <0.50 (18). Cerebrovascular disease was defined as self-reported history of stroke or carotid stenosis >80% (19,20).

For assessment of cardiac autonomic function, participants were asked to refrain from smoking and drinking coffee for 2 h before the tests. Tests took place between 8:30 a.m. and 4:00 p.m., at least 1 h after participants ate a light meal, in a quiet ambience with room temperature of 19–22°C. Cardiac cycle duration (R-R interval) and continuous finger arterial pressure were continuously recorded on a PC-based data-acquisition system, in supine position 1) during spontaneous breathing for 3 min and 2) during metronome breathing at six breaths/min for 1 min. When off-line spectral analysis showed that breathing was not performed at the appropriate frequency, the record was discarded. The test session started with a resting period of at least 10 min. R-R intervals were obtained by electrocardiography using a QRS detector with an accuracy of 1 ms. Blood pressure was recorded using the Finapres method (finger arterial blood pressure, model BP2000; Ohmeda, Englewood, CO) (20). Systolic blood pressure values were obtained from the sampled blood pressure signal by means of an automatic procedure, which was verified by visual inspection.

Seven measures of cardiovascular autonomic function were computed, as previously described in detail (14,18,21,22) (Table 1). Individual data were missing (57 for spectral measures, 51 for Ewing tests, 87 for BRS) because the test schedule was not completed, because the test was not performed correctly, or because the quality of the data were insufficient for processing.

Information on vital status was obtained from the population register of the city of Hoorn. Causes of death were obtained from the medical records at the general practices, the local hospital, and the local nursing home and coded according to the International Classification of Diseases, Injuries and Causes of Death, 9th Revision (ICD-9) (23). Cardiovascular disease mortality, including sudden death, was defined by the ICD-9 codes 390-459 and 798.

The study protocol was approved by the Ethics Committee of the University Hospital of the Vrije Universiteit. All study participants gave informed consent.

All analyses were performed using SPSS statistical software (SPSS, Chicago, IL). For survival analyses, the lowest 25th percentile values of the autonomic function measures in the NGT group were considered the cutoff to define autonomic dysfunction. There are no generally accepted cutoff points to define impaired autonomic function, and furthermore, there are indications that the association with mortality risk may not be continuous. The lower quartile was selected to have substantial numbers in each group for the statistical analysis. Hazard ratios (relative risk) and 95% CIs for determinants of cardiovascular and all-cause mortality were determined by Cox’s proportional hazards analysis. All analyses were adjusted for age, gender, and glucose tolerance because of the stratification procedure. We primarily adjusted for other risk factors that were statistically significantly associated with mortality and secondarily for all other risk factors. As possible confounders, we considered BMI, HbA_1c, fasting insulin, total cholesterol, HDL cholesterol, current smoking (yes/no), and alcohol consumption (yes/no). To test for possible effect modification, the risk factor of interest, the autonomic function measure under consideration, and the product term were included in the model. A significant product term was interpreted as effect modification by that risk factor. Whenever effect modification was likely, a stratified analysis was performed. For all analyses, two-sided probability values <0.05 were considered statistically significant.

Follow-up continued until January 1999. Median follow-up was 7.9 years (range 0.5–9.2). During the follow-up period, 101 of the 605 persons died. Information on cause of death could not be obtained for 14 of the subjects who had died, because the individuals had moved and medical records could not be retrieved. The baseline characteristics for survivors (n = 504) and nonsurvivors are shown in Table 2 . Subjects who died had significantly lower scores on all parameters of autonomic function, except for the HRV power ratio, than those who survived (Table 2).

Cutoff points for impaired autonomic function, taken from the lowest 25th percentile in the NGT group, were 878 ms for the mean of all normal-to-normal (NN) (sinus rhythm) R-R intervals, 25.7 ms for the standard deviation of all NN (sinus rhythm) R-R intervals (SDNN), 125 ms² for low-frequency (LF) power in the R-R–interval spectrum between 0.04 and 0.12 Hz, 93 ms² for high-frequency (HF) power in the R-R interval spectrum between 0.12 and 0.40 Hz, 0.41 for LF/(LF + HF), 107 ms for expiration-inspiration (EI) difference in R-R intervals during breathing at six breaths/min, and 6.1 ms/mmHg for BRS.

The survival analyses, shown in Table 3, yielded systematically different results for subjects with and without diabetes. In diabetic subjects, but not in nondiabetic subjects, impaired autonomic function was consistently associated with an approximately doubled risk of mortality but was not statistically significant for all parameters. Of these seven parameters of autonomic function, EI difference was significantly associated with all-cause mortality, after adjustment for age, gender, and glucose tolerance, and an additional four parameters showed a tendency (P < 0.10): mean NN, LF power, HF power, and BRS. A survival curve of categories of the EI difference is shown in Fig. 1. For six of the seven measures of autonomic function, the relative risks were slightly higher in diabetic subjects than in nondiabetic subjects; for SDNN, the product term was significant. The associations with cardiovascular disease mortality were similar.

Further stratified analyses showed an elevated risk of mortality associated with low autonomic function in subjects with hypertension, subjects taking antihypertensive medication, or subjects with a history of cardiovascular disease (data not shown). Therefore, a stratified analysis was performed for subjects at low or high risk, with high risk defined as the presence of diabetes and/or hypertension and/or a history of cardiovascular disease. In subjects at low risk, impaired autonomic function did not seem to be associated with all-cause or cardiovascular mortality. In the high-risk group, however, autonomic dysfunction showed a consistent association with mortality but was not statistically significant for all parameters (Table 3). For the EI difference, the product term, indicative of effect-modification by baseline risk, was significant.

Potentially confounding variables that were statistically significantly associated with all-cause mortality were BMI, smoking, and fasting insulin (data not shown). Including these potentially confounding variables into the models did not materially change the estimated relative risks for impaired autonomic function.

This study shows that autonomic function parameters are associated with risk of all-cause and cardiovascular death, even after adjustment for age, gender, glucose tolerance, and use of antihypertensive medication. The strength of the relation between survival and autonomic function was generally stronger in subjects with diabetes, hypertension, and/or a history of cardiovascular disease.

Relatively short recordings of heart rate variability (15 s to 15 min) have been shown to be predictive for both cardiovascular and all-cause mortality in postmyocardial infarction patients and also in the general population (3,5). Autonomic function assessment is used mainly in two clinical fields: in cardiology for risk-stratification after myocardial infarction and in internal medicine to assess (diabetic) autonomic neuropathy. Cardiologists tend to use HRV measures (SDNN, LF power, and HF power) as obtained from 24-h Holter monitor recordings (14), whereas internists tend to use controlled breathing and the lying-to-standing maneuver (12,24), both known as Ewing tests, although HRV analysis has been introduced as well. Recently, the BRS has gained more interest (4,25). The current study is the first to address all these measures of autonomic function: Ewing test, HRV, and BRS. In general, the measures obtained from the deep-breathing test (EI difference and BRS) showed the highest relative risks, especially in individuals with hypertension or diabetes. However, the measures obtained during spontaneous breathing also showed a clear association with mortality. Use of certain medication, such as beta blockers and sympaticomimetics, may be confounding because they affect autonomic nervous system function (14,18). Therefore, additional stratified analyses were performed, and indeed, hypertension and use of antihypertensives were effect modifiers. This finding has not been reported before and has not been taken into account in previous studies (3,4,5). In the present study, a large proportion of the diabetic subjects were taking antihypertensive medication; the initial tendency of effect modification by diabetes was partly attributable to the differences in use of antihypertensives in the diabetic subjects (40 vs. 23% in nondiabetic subjects).

In the present study, the lowest quartiles of the distribution of autonomic function parameters were taken as indications for impaired autonomic function. In this population-based study, few subjects had levels lower than previously reported cutoff points. However, when we did apply these cutoff points to our data, the estimated relative risks were not consistently higher, and the results with respect to the observed interaction with high-risk status were similar.

Studies on the relation between autonomic function and survival in diabetes are scarce (2,26). This is the first study to address the influence of diabetes on the association between autonomic function parameters and mortality in a large cohort, stratified by glucose intolerance. The Framingham Heart Study (6), the Zutphen study (5), and the Autonomic Tone and Reflexes After Myocardial Infarction (ATRAMI) study (4) have shown that impaired autonomic function is predictive for (cardiovascular) death, both in the general population and after myocardial infarction. In the Zutphen Study (5), these possible effect modifiers (4,5) were not taken into account, whereas in the Framingham Heart Study, data have been adjusted for diabetes and use of diuretics. However, as we observed in our study, diabetes and hypertension may be effect modifiers, and in that case, a stratified analysis is indicated to present the results. In the ATRAMI study, subjects with type 1 diabetes were excluded, but hypertension or use of antihypertensives was not accounted for. These differences of accounting for diabetes and use of antihypertensives, together with the different ways of assessing autonomic function, the different cutoff points used, and the different populations, limit detailed comparison with previous results. However, the association of autonomic dysfunction with mortality risk is consistent. Our study confirms this for a Ewing test measure, HRV measures, and BRS. We showed that these associations were especially strong in a high-risk population, defined by the presence of diabetes, hypertension, and/or a history of cardiovascular disease.

The increased risk of mortality in myocardial patients with impaired autonomic function can be attributed to the elevated risk of life-threatening arrhythmia (3,14). In the general population, impaired autonomic function by low HRV or low BRS is also associated with nonfatal cardiovascular events and even with future risk of hypertension (27,28,29); therefore, other mechanisms must be involved. Autonomic dysfunction could be a marker or even a risk factor for endothelial dysfunction and atherosclerotic disease. The present finding that impaired autonomic function is especially harmful in subjects at high risk may add strength to the hypothesis that impaired autonomic function may complicate underlying cardiovascular disease rather than being a risk factor. However, further studies are needed to elucidate the involved mechanisms.

The strength of the relation between survival and autonomic function was dependent on the presence of diabetes, hypertension, or a history of cardiovascular disease. This suggests that in subjects already at high risk, cardiovascular autonomic dysfunction may be especially hazardous.

This study was supported by a research grant from the Dutch Diabetes Research Foundation and by a travel grant from the Netherlands Organization for Scientific Research (NWO) to J.G.

We thank Femmie de Vegt for her accurate collection and ICD-9 coding of the causes of mortality for the Hoorn Study population.