Journal of Psychosomatic Printed in Great Britam.

Research,

Vol. 34. No. I, pp. 21-27.

1990. 0

THE RELATIONSHIP BETWEEN HEART MOOD IN REAL LIFE

0022-3999/w $3.00 + .oo 1990 Pergamon Press plc

RATE AND

DEREK W. JOHNSTONand PAVLOS ANASTASIADES* (Received

21 April 1989; accepted in original form

13 June 1989)

Abstract-Verylittle is known about the relationship

between stress and cardiovascular responses in everyday settings. The three subjective states of Stress, Arousal and Time Pressure were measured every 30 min during a normal day in 32 healthy male volunteers and related to heart rate, which was measured continuously using standard ambulatory techniques. An index of the subjects physical activity was derived from the muscle activity of the thigh. Heart rate related to emotional state in very few subjects when time-series statistical methods, which take into account the autocorrelated nature of the data, were used. The relationship was further reduced when allowance was made for concurrent physical activity. The minority of subjects who exhibited a significant association between heart rate and mood variations were significantly more anxious, reported more anger, and had higher systolic blood pressures at rest than subjects who did not show a relationship between mood and heart rate.

INTRODUCTION

Two APPROACHESto the study of cardiovascular responsiveness in real life can be discerned; the study of highly selected individuals undergoing specific stressors and, less commonly, the study of individuals facing the normal repetitive and largely minor vicissitudes and joys of everyday life. Examples of the former are studies of racing drivers competing [l], cardiac surgeons and anaesthetists in theatre [2] and agoraphobic patients entering frightening situations [3]. Such studies are valuable demonstrations of the occurrence of cardiovascular arousal under real life circumstances but suffer from some of the constraints of laboratory investigations, i.e., they are only possible with some psople, some of the time and give little idea of the extent to which stress contributes to cardiovascular reactivity in normal life. This is important since theories which suggest that stress has an important role in the aetiology of conditions such as hypertension [4], imply that there must be frequent, although possibly minor, stress related elevations in cardiovascular activity throughout the day, in some individuals at least. Furthermore the evaluation of stress reduction techniques in such conditions would be greatly advanced if such effects could be first documented and then included as outcome variables in treatment trials. Studies of the cardiovascular effects of unselected everyday stressors are rare. Sokolow er al. [5] examined the relationship between various subjective mood states, presumably correlates of environmental and other stressors, and heart rate and blood pressure when all were measured repeatedly during the waking day. In fifty hypertensive patients, mood and the cardiovascular measures were taken every 30 min and significant correlations found between mood and heart rate and blood pressure in a minority of patients. This pioneering work, which has been curiously *University of Oxford. Address for correspondence: School, London. SW17 ORE.

D. W. Johnston,

Psychology

21

Department,

St George’s

Hospital

Medical

22

DEREK W. JOHNSTONand PAVL~S ANASTASIADKS

ignored until recently, suffers from three major defects: (I) the cardiovascular measurement required that the subject inflate a blood pressure cuff; this obviously disturbs ongoing events and may mean that measurement is omitted or delayed during particularly stressful periods. (2) Sokolow et al. did not measure physical activity which can have a powerful effect on heart rate and blood pressure which can be confounded with the effects of stress. (3) They examined the relationship between mood and cardiovascular activity using simple correlations. It is very likely that the measures of mood and cardiovascular activity are autocorrelated in such a way that current observations relate positively to previous ones. This is likely to lead to inflated estimates of the significance of the correlations [6]. In the present study we examine the relationship between heart rate and mood using methodology that eliminates some of the difficulties with the earlier work. Heart rate was determined from a continuous 24-hr ECG record using standard ambulatory recording techniques that do not interfere with ongoing activities. An estimate of physical activity was obtained from a measure of the subject’s thigh EMG level [7] and time series methods, which take into account the autocorrelated nature of the data, used to examine the relationship between heart rate, activity and mood. Mood was assessed from questionnaires that were completed every 30 min and which measured Arousal, Stress [S] and Time Pressure [S]. METHOD

The subjects were 32 healthy male volunteers from Oxford Polytechnic aged between 18 and 24. They were paid f8 for participating in a laboratory session [described m 91 and the 24 hr monitoring.

All signals were recorded on a 4.channel Medilog 4.24 Cassette Recorder (Oxford Medical Systems) usmEAD amplifiers. The ECG was recorded using disposable AgIAgCI electrodes rn the modified Lead II position. The’EMG was recorded with elcctrodesplaced one-thiyd and two-thirds above the knee along the muscle of the left thigh. The third channel received signals from an event marker. This device also generated a tone every 30 min. The fourth channel was used to monitor variations in recording speed and thus allow resetting of the replay speed for a constant replay to record ratio of x25.

Cassettes were replayed on an Oxford Instruments Replay Unit with adjustable speed control and demodulating replay amplifiers. Signals were replayed to a PDP 1 I /34 minicomputer (DEC) via a multiplexed 12 bit AjD converter and interface. Signals were extensively reviewed prior to analysis to identify any missing or distorted srgnals, adjust amplification levels and ensure a correct replay speed. The ECG was sampled every 3 msec with respect to record time (every 120msec replay time). and analysed for the occurrence of an R-wave using a threshold detection method. Intervals between successive R-waves were stored as IBI in msec. Failure to detect an R-wave for a period of more than 2 set was considered an artefact and the time counter reset. The EMG signal was sampled every 75 msec, rectified and integrated over a period of 3 sec. and further averaged every 30 sec. Signals were continuously displayed on an oscilloscope and monitored for artefacts. An interactive facility allowed the indication of such instances for later editing. After such editing 30 set means of heart rate and activity were computed and stored along with the time of day and corresponding event marks.

Subjects attended the laboratory early in the morning (approximately 9 am) and were prepared for 24 hr ambulatory recording. The ECG. EMG and Event leads were connected to the Medilog recorder, which was worn on a belt round the waist. The quality of the signals was checked using a Medilog monitoring device and on a polygraph. A brief recording was taken using the cassette and this reviewed to ensure that the entire system was functioning properly.

Heart

rate and mood

in real life

23

Subjects were required to complete diaries every 30 min. when the event marker buzzer sounded. The diary consisted of 30 copies of a brief mood questionnaire. This contained 6 adjectives relating to Arousal, 6 to Stress (3 positive and 3 negative in each case) and 4 measuring Time Pressure. The Arousal and Stress adjectives were taken from MacKay’s Stress Arousal check list [S] and Time Pressure from Sokolow [5]. Instructions and scoring followed the published procedures. The adjectives were present in 6 random orders. Subjects briefly noted their activities during the preceding 30min, recorded the exact time of completing the questionnaire and pressed the event marker button. Subjects were advised that they should proceed with their day in a normal fashion. They were shown how to adjust the belt with the recorder when they wished to go to the lavatory, undress or go to sleep and how to disconnect the buzzer prior to sleep. Finally’the cassette was inserted in the recorder, the Medilog switched on and the exact time noted. Subjects returned on the next day for the laboratory test session when they completed the Spielberger Anxiety Inventory [lo], the Profile of Mood States [I l] with respect to the previous 24 hr and, for the final 17 subjects, a resting blood pressure was taken using a random zero sphygmomanometer. Statistical analysis The relationship between mood and heart rate was determined individually for each subject. Mean heart rate and activity every 30 set were calculated from the Medilog recordings and averaged to provide 6-min means that form the basic datum for the analyses. This measurement frequency was chosen since it provides enough data to readily determine the most appropriate times series model using the methods at our disposal, and was not at markedly higher frequency than our measurement of mood. Following expert advice* autoregressive models were fitted using Ordinary Least Squares regression. The analysis proceeded in number of stages using programmes from the BMDP suite of programmes implemented on the PDP 1 l/34. Using P2T the autocorrelogram and partial autocorrelogram was examined for each heart rate series and a provisional model identified. In most of the subjects a first order autoregressive model appeared appropriate, i.e., the autocorrelations died out exponentially and the partial autocorrelation was only significant at lag 1 [see 12 for an accessible discussion of this technique]. Models were fitted using PlR and P2R and the residuals examined with P2T. An adequate model should result in a residual series that is not autocorrelated, i.e., a white noise series. Mood can then be incorporated into the model and the resulting estimates of the relationship between mood and heart rate should have credible probability values. The criteria for determining that the residuals were white noise was that the autocorrelations be non-significant at lags 1 and 2 and the Ljung-Box Q test, a measure of randomness in a time series, be non-significant over the first 20 lags. The effect of activity was examined and removed using a similar procedure in which multivariate models of the heart rate series were developed in which both the heart rate and ACT were used as lagged variables. Further details of the autoregressive models have been presented elsewhere [9]. When satisfactory autoregressive models of the heart rate series, with and without activity, were developed then the three mood measures were incorporated separately. Heart rate was computed five times more frequently than mood but the analysis requires that all variables be measured with the same frequency. We approached this problem in three ways: (1) we computed the simple correlations between mood every 30 min and the average heart rate of the 6-min period during which the mood was measured. (2) A full analysis of all the heart rate data was conducted using estimates of the mood for the four heart rate observations that were obtained without a mood measure. Estimates of the subjects’ mood every 6 min were obtained by assuming that mood was a first order autoregressive process and using the data obtained every 30 min to estimate the intervening values. It should be emphasised that the values estimated in this way are unlikely to lead to spuriously significant relationships between mood and heart rate but may underestimate the relationship, since the inaccuracies in the estimates may be substantial. As we have indicated the alternative procedure runs the opposite danger and can suggest that there are more significant relationships than is in fact the case. (3) Finally the residuals (i.e., the difference between the observed value and that predicted for the time series analyses) from the autoregressive modelling for the 6-min period during which the diaries were completed, were correlated with the mood ratings. These correlations are unaffected by the autocorrelations in the heart rate and activity series and do not involve the rather restrictive assumptions necessary to calculate the intervening mood values. RESULTS

One subject did not complete his mood ratings reliably and his results were not analysed and one reported too little variation in Time Pressure for analysis. There *We wish to thank David Hendry, Professor assistance with the time series analysis,

of Economics,

The University

of Oxford,

for his generous

DEREKW. JOHNSTONand PAVLOSANASTASIADES

24

are therefore either 30 or 31 subjects available for analysis dependent on the mood measure. Adequate univariate models of the heart rate series were obtained for all bar two subjects and satisfactory multivariate models, incorporating activity, were obtained for all bar four. The univariate models were mostly of the simplest kind, first order autoregressive, i.e., an uncorrelated series of observations were obtained when the correlation between the current observation and the immediately preceding one was removed. The multivariate models mostly consisted of a first order autoregressive heart rate effect along with the current level of activity and the activity in the immediately preceding time period. The few series that were not adequately modelled had only slight defects that are most unlikely to have affected the relationship between mood and heart rate. A more complete report of the results of the time series analysis can be found in Johnston et al. [9]. The number of subjects showing significant relationships between mood and heart rate in the autoregressive analyses are shown in Table 1. It can be seen that only a minority of subjects showed any relationship between a particular mood and heart rate but that when it was significant it was usually in the predicted positive direction. Stress had marginally more significant relationships than the other mood measures. Eleven subjects showed a significant positive relationship with at least one mood measure and four negative. When activity is allowed for this diminishes with 6 subjects showing a positive relationship and two negative. The partial correlations between mood and heart rate of the significant positive relationships vary between 0.19 and 0.34. The significant negative relationships were all marginal, r = -0.18 to - 0.22. The number of significant correlations between a specific mood and heart rate measured every 30min are shown in Table 2. These refer to the simple product moment correlation between heart rate and mood and the correlations between mood and the residuals from the time series analyses. While more of the relationships are apparently ‘significant’ in the simple correlational analysis of raw heart rate than in the full time series analyses, the overall pattern is similar. In most cases subjects showing a significant relationship in the full autoregressive method also showed significant relationships with the simple correlations. When this did not occur it was often largely a matter of the simple correlation being similar in size to the partial correlations from the autoregressive analysis, but failing to achieve statistical significance. The significant positive correlations ranged between 0.38 and 0.68. The correlations of mood and the residuals of heart rate were broadly similar to the results obtained in the full time series analysis. Both these findings suggest that the

TABLE I.-THE NUMBER OF SIGNIFICANT RELATIONSHIPS BETWEEN MOODANU HEART KATF AFTER AUTOREGRESSIVE MODELLING, WITH AND WITHOUT ACTIVITY ENTERED INTO MODEL

Heart rate only

Mood

n

Positive

Arousal Stress Time txessure

31 31 30

3 6 4

Negative 2 2 1

Heari rate (activity in model)

Positive 2 4 2

Negative

I I 1

Heart

rate and mood

in real life

25

TABLEII.-SIGNIRCANTCORRELAT~ONSBETWEENHEARTRA~,HEARTRA~~RESIDUALSANDMOODRECORDED EVERY 30 MINIJlF.9

Mood

n

Arousal Stress Time pressure

31 31 30

Heart Positive 5 8 5

rate Negative 2 2

1

Heart rate residuals Positive Negative 0 4 2

2 0

I

Heart rate residuals (allowing for activity) Positive Negative 2 2 4

2 1 0

autoregressive procedures used to generate the mood estimates are not producing spurious or unlikely relationships. The 11 subjects who showed significant mood/heart rate correlations (in the univariate time series analysis) were compared with the remainder on the Spielberger Anxiety Inventory, the Profile of Mood States, blood pressure and the various mood measures taken during the day. They had significantly higher levels of Trait (46.5 vs 38.9; t = 2.88, p < 0.01) and State Anxiety (39.9 vs 34.8; t = 2.24, p < 0.05), experienced more anger on the Profile of Mood States (3.6 vs 1.5; t = 2.54, p -C 0.02) and had higher Systolic Blood Pressure (122.3 vs 110.4; t = 2.59, p < 0.05). They did not differ in their mood as assessed by their mood diaries, although there was a tendency for them to experience more Stress. It may be of interest that subjects reporting above the median on anger during the day had higher Systolic Blood Pressure (121.5 vs 109.2; t = 2.18, p < 0.05). DISCUSSION

Our data show that heart rate averaged over a 6-min period throughout the waking day relates in a minority of subjects to one of the three moods assessed every 30 min. In a few subjects these relationships still hold when rigorous statistical methods are used and physical activity allowed for. Our study is directly comparz!ible to that of Sokolow et af. [5]. We adapted our measure of Time Pressure from their work and our measures of Arousal and Stress are similar to their alertness and anxiety scales. They found, using simple correlations, that heart rate correlated positively with Time Pressure in 29% of subjects, while alertness and anxiety correlated with heart rate in approximately 15%. The comparable figures (in Table II) are 17%, 16% and 26%. While there are slight differences between the two sets of results they are less striking than the similarities. Using comparable methods, at most one third of the subjects show a positive relationship between a specific mood and heart rate. Inspection of the autocorrelograms of the heart rate and mood data confirmed that the data gathered every 30 min was positively autocorrelated, suggesting that the significance levels of the correlations might have been overestimated. It is therefore not surprising that the picture changes when more rigorous autoregressive modelling procedures were used to produce defensible estimates of the significance of the relationships. The number of significant relationships between mood and heart rate was reduced for all mood measures. When activity was allowed for the number of significant positive relationships dropped even further to a maximum of 4 from 31 (13%) for Stress. When the heart rate residuals were used in simple correlational analyses the number of

26

DEREK W. JOHNSTONand PAVLOS ANASTASIAI)ES

significant relationships was again reduced. It therefore appears likely that the earlier study may have overestimated the relationship between mood and heart rate because of inadequate statistical methods and the failure to take into account the effects of concurrent physical activity. These findings may appear to contrast with a recent report [13] claiming that the ambulant blood pressure of hypertensive patients relates very reliably to the patients’ emotional state at the time of measurement. The blood pressure of a substantial number of patients was recorded frequently and a comparison of blood pressure made, across subjects, of periods when patients were experiencing either happiness, anxiety or anger. No allowance was made for whether each patient experienced all the emotions or how many times the emotion was experienced by any one patient. In addition all the data was treated as independent although individual patients contributed from 1 to 42 readings. The confounding of patient and emotion makes the results very difficult to interpret. Furthermore treating repeated observations from the same individual as independent inflates the degrees of freedom, and hence the likelihood of obtaining spuriously ‘significant’ results. It appears therefore that the relationship between mood and heart rate in healthy young volunteers is modest. It is not clear what significance should be placed on the finding that most of the relationships were with Stress. The numbers of significant relationships are small and the differences between the findings for the various moods very slight. It may, however, be significant that Stress is the most unambiguously negative of the three moods. Southard et al. [14] has recently reported that subjects reporting the most negative emotions when blood pressure was measured had the highest pressures. Specific negative emotions such as stress may reflect more accurately the subjective effects of environmental stressors with more sympathetic effects on the heart. It may also be relevant that the subjects who showed a reliable heart rate/mood relationship had higher State and Trait anxiety and experienced more anger during the measurement period. They may therefore be more prone to experience distress or other negative emotions with sympathetic effects on the heart. Alternatively mood measures may reflect the effects of stress more accurately in such people. Additionally, the fact that they had higher systolic blood pressure may suggest that they also had more responsive cardiovascular systems. The demonstration that the number of significant relationships between mood and heart rate tends to decrease when activity is allowed for does not necessarily mean that the results without activity taken into account are artefactual. It is quite possible that mood and activity are related and that there is therefore a metabolically mediated relationship with heart rate. Such a relationship would be genuine and of interest for some purposes. However most models of psychophysiological disorders that place some importance on psychological factors, emphasise the non-metabolic effects of stress [4]. Are our findings likely to be a realistic estimate of the percentage of normal subjects showing a relationship between stress and heart rate in daily life? Laboratory studies suggest that most subjects respond to psychological challenge with elevations in heart rate which can be marked [15]. However, it is not known if such responses relate reliably to concurrent mood. In addition laboratory stressors may overestimate the importance of psychological stress in producing cardiovascular change and pathology. Most laboratory stressors are rarely encountered in real life and

Heart

rate and mood

in real life

27

furthermore the response to laboratory stressors often habituates rapidly with repeated presentation of the stressors. Since in real life one is usually dealing with the familiar then most real life responses may be well habituated, particularly in the relatively routine days our subjects experienced. On the other hand we found, in another part of the present study [9], that aspects of the heart rate response to challenging laboratory tasks related to heart rate variability in real life. The latter presumably reflects, in part, the effects of every day stressors on the heart rate of susceptible individuals and suggests that the present findings may be an underestimate based on the limitations of the measurement of real life stress in this study. It is very likely that assessing three moods every 30 min does not adequately characterise the psychological stressors to which our subjects were exposed, or even their response to them, because of the low frequency of measurement and the fact that some stressors will affect other subjective states or may have no measurable subjective effects. This reflects the practical problems inherent in relating stress to physiological activity in freely responding humans. There is clearly a need for better methods of measuring the variations in stress people are subjected to during the day. However, it is also necessary to measure subjective states since it is a truism that stress cannot be defined purely in terms of environmental stimuli. Acknowledgement-This

research

was supported

by the Medical

Research

Council.

REFERENCES 1.

TAGGART P, GIBBONSD, SOMERVILLE W. Some

effects of motor-car driving on the normal and Br med J 1969; 4: 130-134. PAYNERL, RICK JT. Psychobiological markers of stress in surgeons and anaesthetists. In Biological and Psychological Factors in Cardiovascular Disease (Edited by SCHMIDTTH, DEMBROSKITM, BLUM&EN G), pp. 321-342.Berlin: Springer-Verlag, 1986. _ abnormal

2.

3. 4. 5.

6. 7. 8. 9. 10.

heart.

MICHEL~CIN L, MAVE~AKIAN M. Psychophysiological outcome of behavioral and pharmacological treatments of agoraphobia. J Con Clin Psych 1985; 53: 229-236. OBRISTPA. Ca;diov&dar Psychophysiology: A perspective. New York: Plenum, 198 I. SOKOLOWM. WERDEGARD. PERLOFF DB. COWAN RM, BRENENSTUHL H. Preliminary studies relating

portably recorded blood pressure to daily life events in patients with essential hypertension. 1; Psychosomarics in Essential Hypertension (Edited by KOSTERM, MUSAPHH, VISSERP.), pp. 161-189. Basle: Karger, 1970. GXK TD, CAMPBELLDT. Quasi-experimental experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand McNally, 1979. ANASTASIADES P, JOHNSTONDW. A simple measure of activity for use with ambulatory subjects, Psychophysiology (in press). MACKAY C, Cox T, BURROWSG, LAZZERINIT. An inventory for the measurement of self-reported stress and arousal. Br J Sot Clin Psych 1979; 17: 283-284. JOHNSTON DW, ANASTASIADES P, WCI~DC. The relationship between cardiovascular responses in the laboratory and in the field. Psychophysiology (in press). SPIELBERGER C, GORSUCHR, LUSHENER. Manual for the Stale-Trait Anxiety Inventory. Palo Alto:

Consulting

Psychological

Press,

1970.

11. 12.

MCNAIR DM, LQRR M. An analysis of mood in neurotics. J Abn Sot Psych 1964; 69: 620-627. GOTTMANJM. Time-Series Analysis: A Comprehensive Introduction for Social Scientists. Cambridge:

13.

JAMEXGD, YEE LS, HARSHFIELD GA, BLANK SG, PICKERINGTG. The influence

Cambridge

15.

Press,

1981.

of happiness, anger Psychosom Med 1986; 48: 502-508. SOUTHARDDR, COATESTJ, KOLODNERK, PARKERFC, PADGE~TNE, KENNEDYHL. Relationship between mood and blood pressure in the natural environment: An Adolescent population, Health Psych 1986; 5: 469480. and anxiety

14.

University

on the blood pressure

of borderline

hypertensives.

KRANTZ DS, MANUCK SB. Acute psychophysiologic reactivity and risk of cardiovascular disease: A

review and methodological

critique.

Psych BUN 1984; 96: 434464.