Study population

One hundred and two women, who survived hospitalisation for AMI or underwent a revascularisation procedure, percutaneous transluminal coronary angioplasty (PTCA), or coronary artery bypass grafting (CABG) at Huddinge University Hospital or St Göran’s Hospital, Stockholm, Sweden, were enrolled. We based the diagnosis of AMI on World Health Organization criteria of typical enzyme patterns and chest pain or diagnostic ECG changes. Independently of the present report patients participated in a randomised controlled intervention trial. Consecutively, all eligible women below 75 years were asked to participate in the trial and those who agreed to do so were randomly assigned to either the control or the intervention group. The intervention comprised a stress management program specifically designed for women with CHD and targeted at improving social functioning, self esteem, and coping and decreasing depression and type A behaviour.³⁰

Ambulatory ECG monitoring and analyses of HRV

One year and five months (±2.5 months) after randomisation a two channel ECG (Spacelab 90205; Spacelab Inc, Redmond, Washington, USA) was recorded over 24 hours during normal daily activities. The ECG electrodes were attached at the position of CM-V5 (left anterior axillary line sixth rib) and CS-V1 (fourth rib at the sternal border). Patients were asked to continue their usual medications.

The 24 hour Holter tape recordings were digitised and QRS labelled by a commercially available PC based system (Aspect Holter System, Daltek, Borlänge, Sweden). Arrhythmias were automatically analysed and the QRS complexes classified. Consecutive RR intervals were expressed in centiseconds and analysed in units of five minute epochs by custom software. To be accepted for additional analysis, we required at least 96% of the QRS complexes to be classified as normal by the Aspect system.³¹ The time series of RR intervals were resampled at a frequency of two samples a second. Gaps in the time series due to non-normal RR intervals (QRS labelled by the Aspect System classification as noise or extrasystoles) were filled with values calculated by linear interpolation between adjacent normal RR intervals. Misclassified dropped beats deviating more than 3.0 SD from the normal RR interval of each epoch were also automatically checked.

Both time and frequency domain measures were analysed. The mean of the standard deviations of all normal to normal intervals for all five minute segments of the entire recording (SDNNI,³² in ms) was obtained from the time series of normal RR intervals. Frequency domain parameters were calculated by an autoregressive method as described previously³³: high frequency power, 0.15–0.40 Hz; low frequency power, 0.04–0.15 Hz; very low frequency power, 0.0033–0.04 Hz; and total power (in ms²).

Patients were not enrolled if they had non-sinus rhythm or if less than 50% of the original ECG recording was available for analysis.

Evaluation of alcohol consumption

Consumption of alcoholic beverages was assessed by the corresponding items of the Willett food frequency questionnaire,³⁴ which has shown excellent correlation with alcohol consumption as measured by four diet records of one week each (obtained 3–4 months apart) in a Swedish female cohort.³⁵ The questionnaire asks about the usual frequency and quantity of intake of five beverage types: regular beer, strong beer, wine, light spirits (for example, liqueur, vermouth, and port), and spirits. The estimated alcohol content of these beverages was 3.5%, 4.9%, 11%, 19%, and 39%, respectively. Average daily alcohol intake was calculated in grams. Daily consumption of alcohol was categorised as follows: 0 (abstainers), 0.1–5.0 g/day (light drinkers: up to half a standard drink a day), and > 5.0 g/day (moderate drinkers: over half a standard drink a day). Daily consumption of alcohol from regular beer, strong beer, wine, light spirits, and spirits was also calculated. Because overall consumption of strong beer and light spirits was low (mean intake 0.14 and 0.22 g/day, respectively), we grouped these beverages with regular beer and regular spirits, respectively, in beverage specific analyses. Binomial categorical variables were created to indicate whether patients consumed beer, wine, or spirits at all.

Other covariates

Educational attainment was classified into three levels: mandatory schooling only, completion of high school, and some college. Menopausal status was categorised as premenopausal, postmenopausal, taking hormone replacement therapy, and postmenopausal without hormone replacement therapy. Smoking status was categorised as never, current, or former smoking. Clinical history of diabetes mellitus was assessed. Height and weight were measured and body mass index was calculated.

Statistical analysis

HRV parameters were logarithmically transformed, as their distribution was skewed. Student’s t test was used to detect significance of differences between continuous variables for two groups; one way analysis of variance was applied in the case of three groups. Categorical data were compared by χ² test.

After adjustment for potentially confounding factors, least squares mean (SE) was calculated according to the general linear model for the parameters of HRV in each category of alcoholic beverage. Potential confounders were age, menopausal status, body mass index, smoking habits, educational status, history of diabetes mellitus, and treatment status (intervention versus control). In addition we adjusted for medications. Stratified analyses assessed possible effect modification. SAS version 8.2 (SAS Institute, Cary, North Carolina, USA) and SPSS version 11.0 for Windows (SPSS Inc, Chicago, Illinois, USA) were used for statistical analyses.

The study was approved by the ethics committee of Karolinska Institute at Huddinge University Hospital. Informed consent was obtained from all patients.

HRV, heart rate, and alcohol consumption

Table 2⇓ presents the HRV parameters and mean RR interval in the group of abstainers and light (0.1–5.0 g/day) and moderate (> 5.0 g/day) drinkers before and after adjustment. Although HRV parameters appeared to be higher with increasing alcohol intake, we found no significant differences. However, mean RR interval was significantly differently distributed between the alcohol consumption categories; the highest values were registered among light drinkers.

HRV and consumption of specific alcoholic beverages

Table 3⇓ presents the HRV parameters and mean RR interval according to the consumption of the three beverage types. Most of the HRV measures were significantly higher among women drinking wine in the unadjusted as well as in the adjusted models. HRV parameters also tended to be higher among beer drinkers and spirit drinkers than among women who did not drink beer or spirits, but these differences did not reach the level of significance. These findings remained essentially unchanged after further adjustment for β blockers and calcium channel blockers (ln SDNNI among wine drinkers and wine non-drinkers was 3.89 v 3.59, respectively, in the model with further adjustment for medication, p  =  0.019). Although mean RR interval was longer among wine drinkers (p  =  0.026 and p  =  0.064 for the unadjusted and adjusted models, respectively), adjusting for mean RR interval when testing the relation between wine drinking and HRV parameters gave similar results (for example, ln SDNNI was 3.92 v 3.68, p  =  0.043 in the model with further adjustment for mean RR). Furthermore, we examined maximum heart rate as an indicator for physical activity during the 24 hour recording period and found no relation with wine intake and virtually no effect of adjustment.

To evaluate in more detail the relation of HRV to specific beverage intake, we also categorised consumption of beverage types as non-drinking, drinking below the median, and drinking above the median. ln SDNNI was 3.58 versus 3.96 versus 3.80 (p  =  0.025, adjusted model) in these three groups of wine drinking, respectively. The other HRV parameters also differed significantly between the three groups, except for high frequency power, which was only marginally significant. However, there were no significant differences in HRV measures between the corresponding categories of consumption of spirits or beer.

Furthermore, we performed stratified analyses in selected clinical subgroups to ensure that our results were consistent. The results were not materially different when analyses were restricted to the control or to the intervention group, to patients ⩾ 65 years or to patients < 65 years old, or to patients with AMI or patients who underwent CABG or PTCA.

We also tested the hypothesis that patients drinking beer or spirits may have a trend for higher HRV only due to their higher prevalence of wine drinking. When we simultaneously controlled for intake of other beverage types, wine intake remained significantly associated with HRV (ln SDNNI among wine drinkers was 3.61 and among non-drinkers, 3.89; p  =  0.041, adjusted model). However, the tendency towards higher HRV among the consumers of beer and spirits was considerably attenuated in the simultaneous analyses (ln SDNNI 3.76 v 3.74, p  =  0.85, and 3.76 v 3.73, p  =  0.86, for drinkers versus non-drinkers of beer and spirits, respectively).

Comparison with previous studies

According to previous observations acute alcohol intake decreases HRV, especially the indexes of vagal activity.^40–,⁴³ It has also been shown that chronic excessive intake of alcohol is associated with decreased HRV.^44,45

However, studies on usual daily alcohol consumption and HRV found varying and apparently opposite effects. Ryan and Howes²⁷ reported a negative relation between HRV and usual alcohol intake in apparently healthy men. In the Framingham study, Tsuji and colleagues²⁵ found a positive association between HRV and alcohol intake, but the association was no longer significant after controlling for potential determinants of HRV. In line with this finding alcohol was not a significant independent predictor in the studies of Virtanen and colleagues²⁸ and Stolarz and associates.²⁹ Nevertheless, according to Kupari and colleagues,²⁴ alcohol consumption is positively related to HRV in multivariate models in women, but not in men. All these studies examined apparently healthy people or population samples, except for the investigation of Virtanen and colleagues,²⁸ where patients with newly diagnosed hypertension were studied in addition to healthy volunteers.

Christensen and colleagues,²⁶ in a study population similar to ours, found that consumption of wine, but not beer, was associated with higher HRV among patients referred for coronary angiography. However, this effect was not independent of consumption of polyunsaturated fatty acids derived from fish, which was strongly related to HRV.

Study limitations

Several potential limitations of our study need to be considered, such as the use of self reported assessment of alcohol intake, which may lead to an underestimation of the real intake. However, there is no reason to believe that patients with lower HRV would underestimate their alcohol consumption to a greater extent.

The validity of our findings may be limited by the design of our study. As with any observational study, unevenly distributed characteristics associated with the amount of HRV may lead to overestimation or underestimation of the true effect of alcohol consumption. Multivariate adjustment here did not attenuate the association between wine intake and HRV. However, we cannot exclude the possibility of residual confounding, for example, by the effect of polyunsaturated fatty acids, which was a serious confounder in the study by Christensen and colleagues.²⁶ At the same time, a remaining confounder would need to be associated with both alcohol consumption and HRV and generally unrelated to the factors considered in our multivariate analyses.

The sample size may not have been sufficient in this study to detect an association between HRV and consumption of beer or spirits or total amount of alcohol consumed. Nevertheless, the association with wine intake was already significant at this level of statistical power, and the simultaneous analyses of the beverages types showed virtually no tendency for higher HRV among beer or spirit consumers.

Another potential limitation of the present study is the possibility that some women ceased drinking in response to the severity of their illness. Since we have no information on drinking habits in the past, the “sick quitter” hypothesis⁴⁶ may provide an alternative explanation for our findings.

We investigated women with established CHD and our findings are most readily generalised to similar populations. Women differ from men in drinking habits and alcohol metabolism, and recent studies suggest that wine drinking may have stronger effects on atherosclerosis in women.^47–,⁴⁹ In addition, as alcohol consumption was rather modest among our patients we cannot extrapolate our results to heavier alcohol consumption.

Conclusion

Intake of wine, but not spirits or beer, was positively and independently associated with HRV in women with CHD. Our results, suggesting that HRV may be an important linking factor, may contribute to the understanding of the complex relation of alcohol consumption with CHD.