This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Mental Health, is properly cited. The complete bibliographic information, a link to the original publication on http://mental.jmir.org/, as well as this copyright and license information must be included.
Sleep is fundamental for good health, and poor sleep has been associated with negative health outcomes. Alcohol consumption is a universal health behavior associated with poor sleep. In controlled laboratory studies, alcohol intake has been shown to alter physiology and disturb sleep homeostasis and architecture. The association between acute alcohol intake and physiological changes has not yet been studied in noncontrolled real-world settings.
The aim of this study was to assess the effects of alcohol intake on the autonomic nervous system (ANS) during sleep in a large noncontrolled sample of Finnish employees.
From a larger cohort, this study included 4098 subjects (55.81%, 2287/4098 females; mean age 45.1 years) who had continuous beat-to-beat R-R interval recordings of good quality for at least 1 day with and for at least 1 day without alcohol intake. The participants underwent continuous beat-to-beat R-R interval recording during their normal everyday life and self-reported their alcohol intake as doses for each day. Heart rate (HR), HR variability (HRV), and HRV-derived indices of physiological state from the first 3 hours of sleep were used as outcomes. Within-subject analyses were conducted in a repeated measures manner by studying the differences in the outcomes between each participant’s days with and without alcohol intake. For repeated measures two-way analysis of variance, the participants were divided into three groups: low (≤0.25 g/kg), moderate (>0.25-0.75 g/kg), and high (>0.75 g/kg) intake of pure alcohol. Moreover, linear models studied the differences in outcomes with respect to the amount of alcohol intake and the participant’s background parameters (age; gender; body mass index, BMI; physical activity, PA; and baseline sleep HR).
Alcohol intake was dose-dependently associated with increased sympathetic regulation, decreased parasympathetic regulation, and insufficient recovery. In addition to moderate and high alcohol doses, the intraindividual effects of alcohol intake on the ANS regulation were observed also with low alcohol intake (all
Alcohol intake disturbs cardiovascular relaxation during sleep in a dose-dependent manner in both genders. Regular PA or young age do not protect from these effects of alcohol. In health promotion, wearable HR monitoring and HRV-based analysis of recovery might be used to demonstrate the effects of alcohol on sleep on an individual level.
Sleep is a crucial period of physiological restoration, and it is the optimal state to assess the tonic component or the most relaxed state of the autonomic nervous system (ANS) in real-life conditions [
Heart rate variability (HRV) is a widely used marker of cardiac autonomic regulation reflecting fluctuations in R-R intervals in short or extended time recordings [
In addition to the cardiac autonomic regulation, HRV analysis may also provide useful information on sleep [
The effect of acute alcohol intake on the ANS using heart rate (HR) and HRV parameters has been shown in the previous studies. In laboratory settings, high acute alcohol consumption (0.7 g/kg-1.0 g/kg) was associated with decreased HRV and increased HR in awake subjects [
However, the effect of acute alcohol intake on the ANS during sleep has not been studied in noncontrolled free-living conditions or with large samples. Most published studies considering the effects of acute alcohol intake on HRV have involved only males, been rather small in number of participants, and included no comparison between genders or objective measurements of physical activity (PA) and recordings during sleep [
The widespread use of wearable and connected consumer devices enables unobtrusive collection of massive amounts of data from large number of individuals during their daily life. These health-related datasets gathered under normal day-to-day circumstances outside of traditional clinical trials represent so called real-world data [
Alcohol consumption is a universal health behavior associated with poor sleep [
The original data sample contained 111,025 measurement days from 42,086 Finnish employees representing a wide range of blue- and white-collar workers in varying size companies. Employees had voluntarily participated in a preventive occupational health care program with the aim of improving their health habits and stress management. The program included a continuous beat-to-beat R-R interval recording for a few days during the participant’s normal life. The R-R interval recordings were performed using Bodyguard (Firstbeat Technologies Ltd, Jyväskylä, Finland) wearable device that was attached on the chest with two electrodes. HRV indices, stress, recovery, and PA were computed with Firstbeat Analysis Server (Firstbeat Technologies Ltd) from the recorded R-R interval data, and together with other physiological measurements, they were used as health promotion tools at employees’ workplaces. Employees were instructed not to participate in recordings if they had any disease stages or medications possibly affecting R-R intervals, for example, chronic heart rhythm disturbance, very high blood pressure (≥180/100 mm Hg), type 1 or 2 diabetes with autonomic neuropathy, severe neurological disease (eg, advanced multiple sclerosis or Parkinson disease), fever or other acute disease, or BMI >40 kg/m2[
All the R-R interval recordings performed on the employees were analyzed and stored anonymously to a registry administered by Firstbeat Technologies Ltd. Each service provider conducting recordings for participants signed an agreement allowing Firstbeat Technologies Ltd to store the anonymized data and to use it for development and research purposes. The employers were responsible to inform their employees about the data usage. Following the agreements, a dataset was extracted from the registry for this study. The use of the dataset for research purposes was approved by the ethics committee of Tampere University Hospital (Reference No R13160).
The dataset extracted from the registry to this study included the R-R interval recordings performed with the Bodyguard device (Firstbeat Ltd, Jyväskylä, Finland). The sampling frequency of the device is 1000 Hz for the R-R interval recording [
From the artifact-corrected beat-to-beat R-R intervals, the average of HR in 10-min nonoverlapping windows and RMSSD with 5-min windows were calculated [
For this study, the exclusion criteria were unknown or very high reported alcohol consumption (>12 portions of alcohol) during the recording day, unknown or very short self-reported sleeping time, and poor quality of HRV recordings (
As background information, age, gender, and self-reported weight, height, and PA class modified from Ross and Jackson [
The selection of study population for the analyses.
HR, RMSSD, LF/HF ratio, time considered as recovery (recovery percentage), and average of the momentary absolute levels of recovery reactions (recovery index) from a 3-hour period starting 30 min after the self-reported bedtime onset were considered as the outcome variables. Only the first 3 hours of sleep were analyzed, as most of the slow wave sleep typically occurs during the first hours of sleep [
All analyses were conducted in a within-subject repeated-measures manner by comparing the participants’ outcome variables between days with and without alcohol intake. The within-subject design was used, as it allows studying the intraindividual effects of acute alcohol intake and controls for possible unknown confounders.
For the within-subject repeated-measures two-way analysis of variance (ANOVA), the participants’ hourly averages of outcome variables were calculated for days with and without alcohol intake, and the participants were categorized into low (≤0.25 g/kg), moderate (>0.25-0.75 g/kg), and high (>0.75 g/kg) dose groups according to their alcohol intake during the day. Note that the groups also include the participant’s reference with no alcohol, and the participants may have data in one, two, or all three dose groups. If a participant had more than 1 day with low, moderate, or high or with no alcohol intake, the outcome variables were averaged over those days. A repeated-measures two-way ANOVA was performed separately for each dose group to evaluate the difference and the shape of the hour-by-hour pattern in the outcome variables between the days with and without alcohol intake.
In the second analysis, the linear regression model was fitted for the difference in the average of the 3-hour HR and HRV parameters between the participant’s days with and without alcohol intake. First, the 3-hour averages of the outcome variables were calculated for each measurement day. Thereafter, the difference in the participant’s averages between the days with and without alcohol intake was calculated. If the participant had more than one measurement day without alcohol intake, the average of the measurement days’ 3-hour outcome variable averages was employed. A dataset including the measurement day with the highest amount of reported alcohol intake from each participant was extracted, and a linear regression was fitted to the data using the difference in the outcome variables between the days with and without alcohol intake as a dependent variable. In addition to alcohol intake, all information available about the subjects was employed as independent variables in the regression models. The independent variables were continuous variables of alcohol dose (g/kg), age, PA class, BMI, and the 3-hour average of HR (bpm) during a night after a day without alcohol intake (baseline sleep HR) and gender as a categorical variable. Age, BMI, and the baseline sleep HR were subtracted to baseline levels of 18 years, 18.5 kg/m2, and 38 bpm, respectively. In addition, a linear regression with interactions between alcohol doses and other predictors was fitted.
All statistical analyses were conducted using R (The R Foundation for Statistical Computing) version 3.2.2. The level of significance in all analyses was set at <0.05. However, with data of this size, it is more important to focus on effect sizes than
From a larger cohort, this study included 4098 subjects who had continuous beat-to-beat R-R interval recordings of good quality with for at least 1 day with and for at least 1 day without alcohol intake. There was a significant proportion of female subjects in this study (
Neither PA class nor BMI differed significantly between the dose groups (
The means and 99% CIs for HR, the LF/HF ratio, RMSSD, the recovery percentage, and recovery index calculated from intraindividual HRV recordings during the first 3 hours of sleep in low, moderate, and high dose groups were calculated (
High alcohol intake had the greatest effect on the outcome variables. On average, HR was increased by 1.4 bpm with low, 4.0 bpm with moderate, and 8.7 bpm with high alcohol intake. The LF/HF ratio was increased by 0.1 with low, 0.3 with moderate, and 0.5 with high alcohol intake. RMSSD was decreased by 2.0 ms with low, 5.7 ms with moderate, and 12.9 ms with high alcohol intake. The recovery percentage was decreased by 9.3 percentage units with low, 24.0 percentage units with moderate, and 39.2 percentage units with high alcohol intake. The recovery index was decreased by 7.1 with low, 20.8 with moderate, and 40.2 with high alcohol intake.
For each dose group, the within-subject repeated-measures two-way ANOVA showed significant differences in all outcome variables (all
Characteristics of the study population.
Demographic characteristic | All (N=4098), mean (SD; range) | Males (N=1811), mean (SD; range) | Females (N=2287), mean (SD; range) |
Age (years) | 45.1 (9.6; 19-65) | 45.2 (9.4; 19-65) | 44.9 (9.8; 19-65) |
Physical activity classa | 4.8 (1.8; 0-10) | 4.9 (1.7; 0-10) | 4.8 (1.8; 0-10) |
Body mass index (kg/m2) | 26.0 (4.0; 18.5-39.9) | 26.7 (3.5; 18.9-39.5) | 25.4 (4.3; 18.5-39.9) |
aPhysical activity class range: 0 (physically inactive) to 10 (physically very active).
Characteristics of low, moderate, and high dose groups during the heart rate variability (HRV) recordings.
Demographic characteristic | Low ≤0.25 g/kg (n=1752) | Moderate >0.25-0.75 g/kg (n=2194) | High >0.75 g/kg (n=716) | |
Number of male subjects, n (%) | 671 (38.29) | 1010 (46.03) | 380 (53.1) | <.001a |
Age in years, mean (SD) | 45.6 (9.0) | 46.3 (9.3) | 42.3 (10.7) | <.001b |
Physical activity class, mean (SD) | 4.9 (1.6) | 4.9 (1.7) | 4.6 (1.9) | .59b |
Body-mass index in kg/m2, mean (SD) | 25.9 (4.2) | 26.0 (3.8) | 25.8 (3.7) | .10b |
Weight in kg, mean (SD) | 77.5 (16.2) | 77.9 (14.3) | 78.1 (14.1) | .31b |
aChi-square test.
bOne-way analysis of variance (ANOVA) adjusted for gender.
The effect of alcohol intake during the three first hour of sleep on a) heart rate, b) low frequency/high frequency (LF/HF) ratio, and c) root mean square of the successive differences (RMSSD). The marks green ●=low dose (≤0.25 g/kg), blue ▲=medium dose (>0.25-0.75 g/kg), and red ■=high dose (>0.75 g/kg) denote the averages, and corresponding white symbols denote the measures for the same persons without alcohol. Due to the size of the data and clarity of the figure, 99% CIs are shown, and the lines between hours are only shown for alcohol dose groups.
This shows that the hour-by-hour pattern was different between the days with moderate and no alcohol intake only for the LF/HF ratio and recovery percentage. In low dose comparisons, the hour-by-hour pattern in the LF/HF ratio (
The effect of alcohol intake during the three first hour of sleep on a) recovery percentage, and b) recovery index. The marks green ●=low dose (≤0.25 g/kg), blue ▲=medium dose (>0.25-0.75 g/kg), and red ■=high dose (>0.75 g/kg) denote the averages, and corresponding white symbols denote the measures for the same persons without alcohol. Due to the size of the data and clarity of the figure, 99% CIs are shown, and the lines between hours are only shown for alcohol dose groups.
In linear model analysis, alcohol intake significantly affected the outcome variables (
Alcohol intake increased the LF/HF ratio, and this effect was slightly stronger among males and subjects with higher PA level (
Recovery percentage decreased significantly by increased alcohol intake (
When the effects for alcohol and background characteristics were controlled, the difference in recovery percentage was strongly correlated with the difference in HR (Pearson partial correlation coefficient and the coefficient of determination:
The linear regression models without interaction components for the average of heart rate (HR), low frequency/high frequency (LF/HF) ratio, root mean square of the successive differences (RMSSD), recovery percentage, and recovery index during the first 3 hours of sleep. BMI: body mass index.
Outcome | HR | LF/HF ratio | RMSSD | Recovery percentage | Recovery index |
Intercept, estimate (SE) | 10.87 (0.66)a | 0.715 (0.115)a | −15.32 (0.98)a | −56.09 (3.25)a | −28.27 (3.70)a |
Alcohol (g/kg), estimate (SE) | 8.49 (0.29)a | 0.425 (0.051)a | −12.24 (0.44)a | −33.67 (1.45)a | −36.63 (1.65)a |
Physical activity class, estimate (SE) | −0.48 (0.06)a | −0.019 (0.011) | 0.37 (0.09)a | 1.62 (0.31)a | 1.81 (0.35)a |
Age (0=18 years), estimate (SE) | −0.03 (0.01)b | −0.001 (0.002)a | 0.14 (0.02)a | −0.06 (0.05) | −0.08 (0.06) |
BMI (0=18.5 kg/m2), estimate (SE) | 0.22 (0.03)a | 0.002 (0.005) | −0.26 (0.04)a | −0.81 (0.14)a | −0.89 (0.15)a |
Gender (0=female, 1=male), estimate (SE) | −1.70 (0.21)a | 0.014 (0.037) | 2.09 (0.32)a | 7.22 (1.06)a | 5.24 (1.21)a |
Baseline sleep HR (0=38 bpm), estimate (SE) | −0.33 (0.01)a | −0.021 (0.002)a | 0.41 (0.02)a | 1.67 (0.06)a | 0.86 (0.07)a |
Adjusted coefficient of determination for the model | 0.267 | 0.039 | 0.245 | 0.230 | 0.135 |
a
b
The linear regression models with interaction components for the average of heart rate (HR), low frequency/high frequency (LF/HF) ratio, root mean square of the successive differences (RMSSD), recovery percentage, and recovery index during the first 3 hours of sleep.
Outcome | HR | LF/HF ratio | RMSSD | Recovery percentage | Recovery index |
Intercept, estimate (SE) | 7.61 (1.04)a | 0.724 (0.182)a | −10.55 (1.55)a | −49.34 (5.13)a | −26.73 (5.86)a |
Alcohol (g/kg), estimate (SE) | 14.47 (1.56)a | 0.386 (0.272) | −20.68 (2.32)a | −46.05 (7.69)a | −38.48 (8.78)a |
Physical activity class, estimate (SE) | −0.33 (0.10)b | −0.051 (0.018)b | 0.51 (0.15)a | 2.29 (0.51)a | 2.57 (0.59)a |
Age (0=18 years), estimate (SE) | 0.03 (0.02) | 0.004 (0.003) | −0.008 (0.03) | 0.07 (0.09) | −0.05 (0.10) |
BMI (0=18.5 kg/m2), estimate (SE) | 0.19 (0.04)a | 0.005 (0.008) | −0.18 (0.07)b | −0.60 (0.22)b | −0.42 (0.25) |
Gender (0=female, 1=male), estimate (SE) | −1.38 (0.35)a | −0.010 (0.062) | 2.04 (0.53)a | 6.47 (1.75)a | 2.95 (2.00) |
Baseline sleep HR (0=38 bpm), estimate (SE) | −0.28 (0.02)a | −0.019 (0.004)a | 0.31 (0.03)a | 1.25 (0.10)a | 0.71 (0.11)a |
Alcohol x physical activity class, estimate (SE) | −0.27 (0.17) | 0.065 (0.029)c | −0.32 (0.25) | −1.47 (0.83) | −1.59 (0.95) |
Alcohol x age | −0.12 (0.03)a | −0.009 (0.005) | 0.26 (0.04)a | −0.04 (0.14) | 0.27 (0.16) |
Alcohol x BMI, estimate (SE) | 0.08 (0.08) | −0.009 (0.014) | −0.15 (0.12) | −0.39 (0.39) | −1.05 (0.45)c |
Alcohol x gender, estimate (SE) | −0.56 (0.60) | 0.251 (0.104)c | −0.13 (0.89) | 1.35 (2.95) | 4.87 (3.37) |
Alcohol x baseline sleep HR, estimate (SE) | −0.08 (0.03)c | −0.005 (0.005) | 0.18 (0.05)a | 0.83 (0.15)a | 0.30 (0.17) |
Adjusted coefficient of determination for the model | 0.271 | 0.042 | 0.255 | 0.236 | 0.137 |
a
b
c
Impact of alcohol on autonomic nervous system control during sleep has been earlier demonstrated in controlled conditions with relatively small samples. This study demonstrated that this effect is also clearly seen in noncontrolled conditions with wearable HR monitoring and HRV analysis. In the large heterogeneous, noncontrolled, and free-living study population, alcohol intake caused a dose-dependent effect in cardiac autonomic regulation during the first 3 hours of self-reported sleep time. Intraindividually, HR remained elevated, parasympathetic recovery was delayed, and sympathetic dominance was prolonged after alcohol intake compared with recordings with no alcohol. The effects in cardiac autonomic regulation were observed already with low doses of alcohol.
These findings accord with previous studies reporting dose-related effects of alcohol on parasympathetic indices of HRV during sleep in laboratory conditions [
The strength of this study was the large study population representing a sample of Finnish employees with the whole span of working age, different BMI categories and PA levels, and both genders. With the large free-living sample, this study provided real-world evidence and enabled further studying the effects of personal background parameters on the effects of alcohol intake on the ANS. The main limitations of the study were not knowing the exact alcohol doses and the exact times of alcohol consumption and sleep. The higher alcohol intakes may have been underestimated. In addition, the alcohol drinking habits of the participants were not known.
Most previous studies considering the effects of alcohol on the ANS have used male subjects only, and differences between the sexes have not been examined [
Our findings on the modifiable disease risk factors are in agreement with previous data on that physical inactivity and high BMI reduce HRV [
Poor sleep associates with negative health behaviors, ill health, and decreased work ability [
The study demonstrates, with big uncontrolled data from unobtrusive wearable monitoring, that alcohol intake results in suppression of parasympathetic regulation of the ANS in a dose-response manner. Being physically active and young appears to provide no protection from alcohol-induced suppression of parasympathetic regulation, a finding that needs to be considered given the literature evidence that increased PA associates with higher alcohol usage among nonalcoholics. Personalized HRV measures such as recovery percentage may be more practical in occupational health settings to demonstrate the effect of alcohol on sleep than, eg, RMSSD, which is strongly age-dependent.
Autonomic nervous system
Analysis of variance
Body mass index
high frequency
low frequency
heart rate
heart rate variability
physical activity
root mean square of the successive differences
This study was supported by the Finnish Funding Agency for Technology and Innovation (TEKES).
HL is currently employed in the Digital Health Lab of Nokia Technologies, and TM and I are currently employed by Firstbeat Technologies Ltd, Jyväskylä, Finland.