Effect of different vibration frequencies on heart rate variability and driving fatigue in healthy drivers

  • Kun Jiao
  • Zengyong Li
  • Ming Chen
  • Chengtao Wang
  • Shaohua Qi
Original Article

DOI: 10.1007/s00420-003-0493-y

Cite this article as:
Jiao, K., Li, Z., Chen, M. et al. Int Arch Occup Environ Health (2004) 77: 205. doi:10.1007/s00420-003-0493-y

Abstract

Objective

This investigation was to assess the effect of different vibration frequencies on heart rate variability (HRV) and driving fatigue in healthy subjects during simulated driving, by the use of power spectrum analysis and subjective evaluation.

Materials and methods

Sixty healthy subjects (29.6±3.3 years) were randomly divided into three groups, A, B and C, and the subjects of each group participated in the simulated driving for 90 min with vertical sinusoidal vibration (acceleration 0.05 g) of 1.8 Hz (group A), 6 Hz (group B) and no vibration (group C), respectively. Low-frequency (LF) and high-frequency (HF) components of HRV, reflecting sympathetic and parasympathetic activities, and the LF:HF ratio, indicating sympathovagal balance, were measured throughout all periods. All indices of HRV were calculated in the pre-experiment period, mid-experiment period and end-experiment period, and were analyzed by repeated measures analysis of variance. Subjective responses to a questionnaire were obtained after the simulated task for the three groups.

Results

Significant differences in all indices of HRV were observed between different experiment periods and between any two groups. The ratings of subjective fatigue exhibited significant differences between any two groups.

Conclusion

The drivers’ fatigue ratings were associated with vibration frequencies in simulated driving. The study quantitatively demonstrated that different effects on autonomic nerve activities were induced by different vibration frequencies.

Keywords

Vibration Heart rate variability Fatigue Sympathetic Parasympathetic Sympathovagal balance 

Introduction

The development of modern life involves ever-increasing exposure of the human body to the action of unfavourable environmental factors, especially vibration and noise. Many people are exposed to vibration in vehicles: cars, buses, trains, ships and airplanes, on a daily basis. It was confirmed that vibration caused discomfort, fatigue and physical pain (Liu et al. 1995). There have been several reports that describe how vibration interferes with people’s working efficiency, safety and health (Mcleod and Griffin 1995; Qassem et al. 1996). Prolonged exposure to vibration from power hand tools, transmitted to the human hands, has been related to a series of disorders in the vascular, sensorineural and musculoskeletal hand–arm system (Gemme and Taylor 1983).

Vehicle vibration is a substantial problem and a serious threat in the city; extensive analytical and experimental studies performed on vehicle and driver vibration have established a relationship between the magnitude and frequencies of vehicle vibration and driver fatigue (Griffin 1990). It was reported that vibration influenced human physiological reactions such as heart rate, blood pressure and respiration (Kubo et al. 2001). Some researchers measured and analyzed autonomic responses of young passengers under different speeds and driving modes of a vehicle (Byung et al. 2002), and, therefore, some researchers have concentrated their efforts on vehicle vibration and its influences.

Heart rate variability (HRV) allows analysis of the interaction between activity in the sympathetic and parasympathetic nervous systems by modulation of the heart beat-to-beat interval. Power spectrum analysis of HRV is a non-invasive, reliable technique able to evaluate the autonomic modulation to the heart (Malliani et al. 1991; Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology 1996; Pagani et al, 1997). HRV has previously been used to evaluate the influence on the autonomic nervous system of different body positions and various anaesthetics, and, qualitatively, different levels of stress (Widmark et al. 1998; Mohr et al. 2002). Frequency fluctuations in the range of 0.04–0.15 Hz [low frequency (LF)] are, however, regarded with some controversy and are considered to be a marker of sympathetic activity, and high-frequency (HF) fluctuations in the range of 0.15–0.40 Hz are considered to be a marker of parasympathetic or vagal activity (Task Force of the European Society of Cardiology 1996). A reciprocal relationship exists between these two frequency domains and is similar to that characterizing the sympathovagal balance (Malliani et al. 1991).

To our knowledge, few studies based on power spectrum analysis of HRV have explored the effect of vibration on autonomic nervous activities and driving fatigue in healthy drivers. In particular, there have been no studies on different effects of different vibration frequencies within the 1–10 Hz range on HRV in healthy drivers. The purpose of this study was, by the use of power spectrum analysis of HRV, to assess the effect of different vibration frequencies on HRV and fatigue in healthy subjects during simulated driving.

Materials and methods

Subjects

To avoid the influence of gender and age on HRV (Duanping et al. 1995), we recruited 60 healthy, male, volunteers aged 29.6±3.3 years (mean ± SD) from staff and students within Shanghai Jiao Tong University of China. Prior to investigation their informed consent was obtained. All subjects were healthy and none of them was under medication. The university ethics committee approved the experimental procedures.

Driving simulator

The vibration testing system HVT V/H-5 (Saginomiya, Japan) was the vibration generator for the experiments, and could generate random or fixed vibration frequencies. Subjects controlled the steering wheel and stared at the 32-in. monitor at a distance of 2.5 m. A 3-h videotape showed a view of the road ahead, and included other vehicles and roadside objects such as trees and traffic signs along a scenic highway having few intersecting roads and scarcely any buildings. A speedometer was provided. To enable an indicated speed of 60 km/h to be maintained we monitored the accelerator pedal position. A small red light mounted in the middle of the dashboard stayed on when pedal pressure was maintained within prescribed limits.

Experimental procedure

All subjects were randomly divided into three groups: A, B or C, and each group contained 20 subjects. All subjects were required to be well rested before the experiment and were instructed not to drink or smoke 2 h prior to the task so that the influence of circadian fluctuation on HRV could be avoided (Fauchier et al. 1998). Subjects of groups A and B, under vertical sinusoidal vibration of 1.8 Hz and 6 Hz (vibration acceleration of 0.05 g), and group C, without vibration, were required to perform simulated driving at the same time on the day of the study (between 8.00 a.m. and 12.00 a.m.). To avoid respiratory events that might influence LF the subjects were made to breathe at a rate of at least 9 breaths per min (0.15 Hz; Piccirillo 1998).

Simulated driving for 90 min followed the pre-experiment period of 10 min. Both the end-experiment period and mid-experiment period were 10 min (Haker et al. 2000). HRV was recorded throughout the whole time (Fig. 1).
Fig. 1

Experiment design of the simulated driving. Pre-experiment period (Pre-expe), mid-experiment period (Mid-expe) and end-experiment period (End-expe) each continued for 10 min. HRV was recorded throughout the whole time. Period 1 the first 50-min, Period II remaining 40 min

HRV recording and analysis

The subjects were connected to an ECG measuring system (FDP-1, version 3.01, Shanghai Medical University) by three electrodes attached to the chest for recording ECG signals and subsequent calculation of HRV and LF and HF components while they were in the driving position. Surface ECG signals were sampled at 250 Hz and recorded directly onto the hard disk of the computer through an AD converter with a resolution of 12 bits. All data were collected in a data logger and then transferred to a computer for subsequent blind analysis. The autoregressive (AR) power spectrum from the RR interval trendgram was calculated by means of the Marple algorithm (Marple 1980; Hayano et al. 1991), and the power of every spectral component was computed according to the method of Zetterberg (1969). We determined the optimal AR model order by minimizing the value of the final prediction error. Using the AR model, we estimated the power spectrum of HRV from 256 RR intervals of the heart rate. For frequency domain analysis, the LF component was calculated as the power within the frequency range of 0.04–0.15 Hz, and the HF component as the power within the frequency range of 0.15–0.4 Hz (Malliani et al. 1991). The ratio between LF and HF power was also calculated. The LF and HF components were expressed in absolute values (milliseconds squared) as well as in normalized units (NU) [LF(NU)=LF×100/(LF+HF) and HF(NU)=HF×100/(LF+HF)] (Hans et al. 1997). Subjective responses to a questionnaire were obtained after the simulated task for the three groups (Tables 1 and 2).
Table 1

Survey table of subjective fatigue symptoms

Symbol

Symptoms

Symbol

Symptoms

a

Physically tired

h

Stiffness of shoulder

b

Lazy

i

Lumbago

c

Want to lie down

j

Easily absent-minded

d

Irritable

k

Eyestrain

e

No energy

l

Feel asleep

f

Mentally sluggish

m

Nausea

g

Headache

n

Trembling of hands and legs

Table 2

Subjective fatigue rating and descriptions

Rating

Fatigue description

1

Not tired at all

2

Minimally tired

3

Somewhat tired

4

Tired

5

Considerably tired

6

Very tired

7

Extremely tired

Statistical analyses

We used repeated measures analysis of variance (ANOVA) to test both the effects of the three experimental conditions and the variability of HRV within each experimental condition. The Bonferroni test was used, after adjustment for multiple comparisons based on observed means when repeated measures ANOVA were significant. We used the chi-square test to compare the different occurrences of fatigue symptoms in the section “Subjective evaluation”, below. Results are given as mean values ± SD. P<0.05 was considered to be the level of statistical significance.

Results

The LF and HF components and the LF:HF ratios of HRV, analyzed with repeated measures ANOVA, are shown in Tables 3, 4 and 5, respectively. Adjustment for multiple comparisons, based on observed means with the Bonferroni test, are shown in Table 6.
Table 3

Effects of different vibration frequencies on LF of HRV

Group

Pre-experiment

Mid-experiment

End-experiment

Total

F

P

A (1.8 Hz)

33.11±14.12

37.42±9.21

42.18±11.34

37.57±12.12

4.436

0.019*

B (6 Hz)

35.92±14.81

43.17±13.65

47.77±12.82

42.29±14.41

3.829

0.031*

C (no vibration)

32.44±12.85

34.13±11.48

38.56±10.22

35.04±11.66

1.401

0.259

Total

33.82±13.80

38.24±12.00

42.84±11.94

38.30±13.07

8.667a

0.000a*

A–B

t

0.614

1.562

1.461

F=4.501a

P=0.015a

0.379b

0.824b

P

0.543

0.128

0.152

A–C

t

0.157

0.999

1.060

P

0.876

0.324

0.296

B–C

t

0.794

2.267

2.512

P

0.432

0.029*

0.016*

*P<0.05

aF statistic and P value of main effect

bF statistic and P value of crossover effect

Table 4

Effects of different vibration frequencies on HF of HRV

Group

Pre-experiment

Mid-experiment

End-experiment

Total

F

P

A (1.8 Hz)

16.72±6.74

13.24±4.13

10.65±2.44

13.54±5.32

9.083

0.001*

B (6 Hz)

20.63±7.08

12.19±3.42

8.64±2.86

13.86±6.90

52.434

0.000*

C (no vibration)

16.97±7.62

14.61±5.34

12.15±2.46

14.58±5.81

3.459

0.042*

Total

18.11±7.26

13.35±4.41

10.52±2.87

13.99±6.03

38.783a

0.000a*

A–B

t

1.789

0.876

1.935

F =0.544a

P =0.584a

3.652b

0.008b

P

0.082

0.387

0.060

A–C

t

0.110

0.908

2.317

P

0.913

0.370

0.026*

B–C

t

1.573

1.706

4.134

P

0.124

0.096*

0.000*

*P<0.05

aF statistic and P value of main effect

bF statistic and P value of crossover effect

Table 5

Effects of different vibration frequencies on LF:HF of HRV

Group

Pre-experiment

Mid-experiment

End-experiment

Total

F

P

A (1.8 Hz)

2.27±1.22

2.87±1.35

4.09±1.88

3.08±1.67

13.634

0.000*

B (6 Hz)

2.08±1.57

3.76±1.63

6.12±1.91

3.99±2.37

35.593

0.000*

C (no vibration)

2.13±1.39

2.46±1.31

3.27±1.28

2.62±1.39

4.429

0.019*

Total

2.16±1.38

3.03±1.52

4.49±2.07

3.23±1.93

48.794a

0.000a*

A–B

t

0.427

1.881

3.385

F=8.168a

P=0.001a

6.740b

0.000b

P

0.672

0.068

0.002*

A–C

t

0.337

0.975

1.612

P

0.738

0.336

0.115

B–C

t

0.108

2.779

5.539

P

0.915

0.008*

0.000*

*P<0.05

aF statistic and P value of main effect

bF statistic and P value of crossover effect

Table 6

Adjustment for multiple comparisons based on observed means with Bonferroni test

Index

Between groups

Mean difference

P

95% CI for difference

Lower bound

Upper bound

LF

A–B

−4.717

0.178

−10.762

1.328

A–C

2.526

0.921

−3.519

8.572

B–C

7.243

0.014*

1.198

13.288

HF

A–B

−0.321

1.000

−2.840

2.198

A–C

−1.040

0.938

−3.559

1.479

B–C

−0.719

1.000

−3.238

1.800

LF:HF

A–B

−0.910

0.032*

−1.759

−0.061

A–C

0.457

0.570

−0.392

1.306

B–C

1.367

0.001*

0.517

2.216

*P<0.05

The low-frequency components

No statistically significant differences in LF during the pre-experiment period between the three groups were observed. A statistically significant difference in LF between groups B and C was observed in the mid-experiment and end-experiment periods (P<0.05, P<0.05). According to the analysis of main effect of time, significant differences were exhibited between different experiment periods (F=8.667, P=0.000). Significant differences between different experiment periods were observed in both group A and group B, while no difference was observed in group C. Significant differences were also observed between different groups (F=4.501, P=0.015). No significant crossover effects of time and vibration factor were observed (F=0.379, P=0.824). (Table 3)

The high-frequency components

No statistically significant differences in HF during the pre-experiment period between the three groups were observed. A statistically significant difference in HF between groups B and C was observed in the mid-experiment and end-experiment periods (P<0.05, P<0.05). Moreover, a significant difference in HF between group A and group C was observed in the end-experiment period. According to the analysis of main effect of time, significant differences were exhibited between different experiment periods (F=38.783, P=0.000). Significant differences between different experiment periods were observed in all three groups. No significant differences were observed between different groups (F=0.544, P=0.584). Significant crossover effects of time and vibration factor were observed (F=3.652, P=0.008). (Table 4)

The LF:HF ratio

No statistically significant differences in HF during the pre-experiment period between the three groups were observed. A statistically significant difference in HF between group B and group C was observed in the mid-experiment and end-experiment periods (P<0.05, P<0.05). Moreover, a significant difference in HF between groups A and B was observed in the end-experiment period. According to the analysis of main effect of time, significant differences were exhibited between different experiment periods (F=48.794, P=0.000). Significant differences between different experiment periods were observed in all three groups. Significant differences were observed between different groups (F=8.168, P =0.001). Significant crossover effects of time and vibration factor were observed (F=6.740, P=0.000). (Table 5)

Adjustment for multiple comparisons

The results of adjustment for multiple comparisons, based on observed means with the Bonferroni test, are shown in Table 6. A significant difference was obtained in LF components between groups B and C; a significant difference was obtained in LF components between groups A and B and between groups B and C.

Subjective evaluation

Self-reporting is the most frequently used method for assessing fatigue. The subjective responses to the questionnaire given after the simulated task for both A and B are shown in Table 1. The occurrence frequencies of subjective fatigue symptoms are shown in Table 7 and Fig. 2, and were compared among the three groups by use of the 3×2 table chi-square test for each fatigue symptom.
Table 7

3×2 Table chi-square test for each fatigue symptom during subjective evaluation

Symbol of fatigue symptoms

Minimum expected count

Chi-square test value

df

P

a

9.67

8.943

2

0.011*

b

9.67

3.737

2

0.154

c

9.00

11.313

2

0.003*

d

9.33

2.143

2

0.343

e

7.33

4.450

2

0.108

f

7.00

3.956

2

0.138

g

3.67

1.656

2

0.437

h

8.00

10.417

2

0.005*

i

8.33

0.960

2

0.619

j

10.00

10.000

2

0.007*

k

6.67

2.850

2

0.241

l

8.67

0.950

2

0.622

m

1.33

4.543

2

0.103

n

0.67

4.534

2

0.104

*P<0.05

Fig. 2

Occurrence frequencies of subjective fatigue symptoms were compared among the three groups using 3×2 table chi-square test for each fatigue symptom during subjective evaluation. *P<0.05

The subjective fatigue rating and descriptions are shown in Table 2.The rating of subjective fatigue in the three groups were 3.65 (group A), 5.1 (group B), and 2.3 (group C), Statistically significant differences were observed between any two groups (P<0.05, P<0.05, P<0.05).

Examples from the three groups

In Figs. 3, 4 and 5, three examples of the power spectrum analysis from, respectively, groups A, B and C, are shown.
Fig. 3

Power spectrum density (PSD) calculated from HRV signals by AR modelling and spectral decomposition in LF and HF components. One example was selected from group A with vibration of 1.8 Hz. Pre-expe pre-experiment, Mid-expe mid-experiment, End-expe end-experiment, Freq. frequency

Fig. 4

Power spectrum density (PSD) calculated from HRV signals by AR modelling and spectral decomposition in LF and HF components. One example was selected from group B with vibration of 6 Hz. Pre-expe pre-experiment, Mid-expe mid-experiment, End-expe end-experiment, Freq. frequency

Fig. 5

Power spectrum density (PSD) calculated from HRV signals by AR modelling and spectral decomposition in LF and HF components. One example was selected from group C without vibration. Pre-expe pre-experiment, Mid-expe mid-experiment, End-expe end-experiment, Freq. frequency

Discussion

Vibration, fatigue and autonomic nervous activities

Drivers and passengers perceive the vibration of vehicles. Their subjective evaluation or judgments of discomfort may influence their opinion of the vehicle. The influence of whole-body vertical vibration of 1–10 Hz on the dynamic human–seat interface was investigated, and it was concluded that vibration could induce fatigue (Wu et al. 1998). Both vibration–load test and road test suggest that vertical vibration at a low frequency of 1 to 2 Hz induces a driver to accumulate fatigue gradually and steadily during long-distance running (Taguchi and Inagaki 1999). In real-life situations, at workplaces, in vehicles, etc., vibrations are usually complex. They consist of multi-frequency, translational and rotational motions, they occur simultaneously in different directions and they are interspersed with shocks of various strengths (Hansson and Wikström 1981).

According to Grandjean (1997), fatigue has been defined as a functional state, which graduates in one direction into sleep and in the other direction into a relaxed, restful condition, both of which are likely to reduce attention and alertness. In other studies, it was defined as a disinclination to continue performing a task and to involve an impairment of human efficiency when work continued after the subjects had become aware of their fatigued state (Brown 1994). Bunnell and Horvath (1989) suggested that reactive changes in various organs and tissues due to the effects of external influences such as vibration are regarded as a complex of energy-dependent structural and functional rearrangements.

Although fatigue may be difficult to be assessed quantitatively, a behavioural scale from no fatigue to unbearable (Appenzeller and Oribe 1997) has been used to consider its relationship with sympathetic nerve activity, as measured by direct intraneural recording from sympathetic nerves in muscle (MSNA; Saito et al. 1989). In many studies of circulatory physiology and clinical medicine as well as of occupational and environmental health, analysis of HRV has been used to estimate cardiac autonomic function (Cook et al. 1991; Harada et al. 1990; Murata and Araki 1996). Some studies have suggested that vibration input has a significant effect on sympathetic and parasympathetic activities as well as autonomic functions in patients (Bovenzi 1990; Griffin 1990; Harada et al. 1990; Harada 1994). Using autoregressive spectral (frequency-domain) analysis of HRV, Murata et al. (1991) found depressed parasympathetic activity during quiet breathing in subjects exposed to vibration.

Experimental period

In the current investigation, we applied power spectrum analysis of HRV to evaluate the effects of different vibration frequencies on autonomic nervous activity in healthy subjects during simulated driving, and we used a subjective questionnaire to evaluate the subjective fatigue rating of subjects. As recommended by the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (1996), LF and HF components were expressed in normalized units. The normalized power of LF and HF components of HRV has been shown to provide a quantitative index of sympathovagal interaction modulating cardiovascular function, LF being the marker of sympathetic activity and HF the marker of parasympathetic activity.

The results showed that the LF component did not exhibit a significant difference during simulated driving without vibration (Table 3) while HF and the LF:HF ratio did exhibit a significant difference (Tables 4 and 5). However, all indices of HRV were changed significantly with vibrations of 1.8 or 6 Hz throughout the simulated driving period. The results indicated that the sympathovagal balance of healthy drivers under vibration conditions had changed significantly, while no similar significant change was obtained in the subjects under no vibration conditions in the same time session. The results have been proved for our previous study (Li, 2003) and supported by Gohara et al. (1996). Considering crossover effects of time and vibration factor, we observed significant crossover effects in HF and LF:HF and no significant crossover effect in LF during simulated driving. This indicated that crossover effects of time and vibration influenced parasympathetic nervous activities significantly but did not influence sympathetic nervous activities. For subjective fatigue rating evaluation, the fatigue-rating value of group C (2.3) is much lower than that of group A (3.65) and group B (5.1). These results were in line with Taguchi and Inagaki’s conclusion (1999) that vibrations of 1–10 Hz would induce fatigue gradually.

No significant differences in LF and HF components and LF:HF ratio were observed between any two groups in the pre-experiment period, and significant differences were observed between groups B and C in the mid-experiment period and post-experiment (P<0.05, P<0.05). That means vibration frequency of 6 Hz induced autonomic nerve activity quickly when compared with the group without vibration (P<0.05). It indicated that different vibration at 6 Hz could significantly change parasympathetic nerve activities at different levels. In the post-experiment period, a significant difference in HF component was obtained between groups A and C, while no significant difference in LF component was obtained between these two groups. It indicated that vibration of 1.8 Hz mainly changed parasympathetic nerve activities. The result most likely relates to the whole body resonance frequency within 5–7 Hz (Wu et al. 1998) and to a head resonance frequency of 1.8 Hz or so (Taguchi and Inagaki 1999).

Although the real mechanism for why different vibration frequencies cause different levels of autonomic nervous activity during simulated driving is not clear, we suggest that it might be related to the resonance frequencies of different parts of the human body. Furthermore, it is reported that whole-body vibration and local vibration could induce tractor drivers’ discomfort and change their physiological and psychological condition. Both vibrations cause the nerve centre to excite and make the autonomic nervous system unbalanced (Li et al. 1995). In addition, it seems likely that long-term exposure to the vibration could develop hyperactivity in the sympathetic nervous system (Harada 1994). Although rhythms should not simply be equated to neural structures, functional states likely to be accompanied by an increase in sympathetic nerve activity are characterized by a shift of the LF:HF balance in favour of the LF component; the opposite occurs during presumed increases of parasympathetic activities (Malliani et al. 1991). The findings indicate that the HRV properties are greatly affected by different frequencies of vibration. HRV parameters are shown to become sensitive to vibration according to the accumulation of mental and physical fatigue, showing considerable change in autonomic modulation. Thus, we have to quantify the subjects’ background mental and physical condition and take it into consideration when deriving a meaningful stress index from the HRV.

Methodological considerations

In the present study we tried to avoid the influence of gender and age on HRV (Duanping et al. 1995), so 60 male, healthy, volunteers aged 29.6±3.3 years were selected to participate in our study. To avoid respiratory events that might influence LF the subjects were required to breathe at rates of at least 9 breaths per min (0.15 Hz; Piccirillo et al. 1998).

Conclusion

The drivers’ fatigue ratings were associated with vibration frequencies in simulated driving. By using power spectrum analysis of HRV and subjective fatigue evaluation, our study quantitatively demonstrated that different effects on the autonomic nerve activities were induced by different vibration frequencies. It was concluded that vibrations of 6 Hz influenced both sympathetic and parasympathetic nerve activities, while vibrations of 1.8 Hz mainly influenced parasympathetic nerve activities. These findings represent physiological support for the proposal that different vibration frequencies cause different levels of mental stress and fatigue in healthy drivers.

Acknowledgements

The authors wish to thank all the participants from Shanghai Jiaotong University and Takashimaya Nippatsu Kogyo Co., Ltd, who funded this research.

Copyright information

© Springer-Verlag 2004

Authors and Affiliations

  • Kun Jiao
    • 1
  • Zengyong Li
    • 1
  • Ming Chen
    • 1
  • Chengtao Wang
    • 1
  • Shaohua Qi
    • 2
  1. 1.Institute of Life Quality via Mechanical EngineeringShanghai Jiaotong UniversityShanghaiChina
  2. 2.Department of Rehabilitation, Zhong Shan HospitalFu Dan University Medical CollegeShanghaiChina

Personalised recommendations