OUP user menu

Personality traits and heart rate variability predict long-term cardiac mortality after myocardial infarction

Clara Carpeggiani, Michele Emdin, Franco Bonaguidi, Patrizia Landi, Claudio Michelassi, Maria Giovanna Trivella, Alberto Macerata, Antonio L'Abbate
DOI: http://dx.doi.org/10.1093/eurheartj/ehi252 1612-1617 First published online: 12 April 2005


Aims To investigate personality traits and sympatho-vagal modulation of heart rate variability (HRV) during acute myocardial infarction (AMI), assessing their relationships and their long-term prognostic value.

Methods and results Psychological traits and 24 h HRV were prospectively investigated in 246 patients at discharge of an AMI. Patients were followed-up to 8 years for the occurrence of cardiac death and non-fatal reinfarction. Low coping and anxiety traits associated with reduced HRV characterized the study population. At univariate analysis, low emotional sensitivity and insecurity, relative tachycardia, reduced high frequency (HF), and low frequency power and pNN50 were predictive of cardiac death at 8-year follow-up. At multivariable analysis, low emotional sensitivity and low HF power remained predictive, with a relative risk of 4.18 (P=0.003) and 2.76 (P=0.007), respectively; also the type of infarction (Q vs. non-Q) and hospital length of stay were independent predictive variables.

Conclusion Anxiety and emotional sensitivity were significant predictors of 8-year cardiac mortality after AMI. Reduced HF power, a recognized marker of vagal withdrawal, increased the risk.

  • Risk factors
  • Myocardial infarction
  • Prognosis
  • Heart rate variability


Negative emotions are generally associated with coronary heart disease.1,2 In particular, patients with acute myocardial infarction (AMI) are characterized by specific personality traits3,4 related to poor prognostic outcome.57 Personality may favour emotional stress, which in turn may alter autonomic drive, reduce coronary blood flow, and induce ischaemia.810 Moreover, clinical depression is associated with cardiac mortality in patients after AMI6,7,11 likely because of an autonomic dysfunction.12

Autonomic imbalance may trigger mechanical and electrical complications, leading to sudden death,13,14 both in the acute and in the chronic phase of MI. Heart rate variability (HRV) has been conventionally accepted as a tool for describing cardiac autonomic function,15 reduced variability is thought to reflect increased sympathetic and/or depressed vagal modulation of heart rate and to have independent prognostic value for late cardiac events after an AMI.16,17

The purpose of this study was to investigate a possible link between psychological traits and autonomic imbalance in AMI, with respect to hard cardiac events during long-term follow-up, extending prediction to a maximum of 8 years.

The paper reports the results of a multi-centre Italian study aimed at assessing patients' psychological traits at discharge of AMI and at investigating their correlation with HRV and their relative prognostic value on a long-term basis.


Study population

The original study was a prospective multi-centre study including 13 centres from northern and central Italy, that enrolled 455 consecutive patients with AMI over a 5-year period since 1990. The study was approved by the CNR Institutional Review Board and the research was conducted according to the principles of the Declaration of Helsinki. The inclusion criteria were age <65, sinus rhythm, admission within 6 h from the onset of symptoms, and diagnosis of AMI on the basis of the presence of two or more of the following criteria: prolonged chest pain of ≥20 min, prolonged electrocardiographic ST segment changes, new Q wave development, and increase of creatine kinase to at least two-fold normal reference value with myocardial band isoenzyme assay >10%. Exclusion criteria were unrelated disease limiting survival such as cancer, renal failure, psychological or psychiatric illness; need for psychotropic therapy; and alcohol abuse. Patients over 65 years were excluded to avoid bias in psychological characterization, for which controls are usually young volunteers. Participating centres were asked to perform 24 h ECG monitoring upon entry, and at discharge were asked to evaluate patients' psychological traits and behavioural patterns, a second 24 h ECG recording and 2D-echocardiography with computation of wall motion score index according to the 16-segment model. Decision about medical therapy was left to each centre.

Of the initial 455 patients, 246 had 24 h ECG recording, and psychological evaluation at discharge, and follow-up data. This group was used to assess the correlation between psychological traits and HRV and their relative long-term prognostic impact; data on prognostic impact of early assessment of HRV have been already published.18 Psychological traits of normal volunteers were previously published.4 Controls were selected within family for the initial group of 455 patients. They should be free of history of coronary heart disease, hypertension, metabolic and endocrine disease, cancer and psychiatric illness, continued consumption of hypnotic or anxiolytic drugs. Any disease-free family members were initially selected and then, in a second stage, controls similar to the study sample in age, gender, and educational level were chosen and only 158 of such patients were found.

Twenty-four hour ECG monitoring and HRV analysis

All patients underwent 24 h Holter monitoring by a frequency modulated two-channel system. HRV, ST segment, and analysis of arrhythmias was done by a Core Centre. Tapes were digitized at a sampling rate of 250 Hz and QRS-labeled. Ectopic beats and artifacts were filtered and outliers (such as ectopic beats) were substituted by linear-spline interpolated points.19 For each recording, we computed 24 h average values of normal RR intervals (RR), standard deviation of all RR (SDNN), standard deviation of 5 min mean values of RR (SDANN), square root of the mean of the sum of the squares of differences between adjacent RR (RMSSD), and the number of adjacent RR differing by more than 50 ms, as per cent of the total number of RR (pNN50). Autoregressive spectral analysis was computed20 for consecutive series of 256 beats and was averaged over the 24 h period to obtain the power in the low frequency (LF) (0.03–0.15 Hz) and high frequency (HF) (0.15–0.40 Hz) band; LF/HF ratio was also computed.15

Psychological traits

Personality was investigated one day before discharge using the Sixteen Personality Factors Questionnaire (16 PF) of Cattell et al.21 (D Form), Italian version.22 It consists of 105 items, each with three response categories; the answers are combined to yield a score of personality expressed by 16 traits, relatively stable over time. These 16 traits are measured on a scale from 1 to 10: a low score (≤5) defines a meaning opposite to that obtained from a high score (>5) as shown in Table 1. Finally, second-order factors were extracted applying factor analyses to the 16 traits:23,24 the values of such factors were normalized and measured on a scale having mean 0 and variance 1. Primary traits and second-order factors are reported here. Low emotional sensitivity is an inability to express emotion and adapting to society or coping. Social inhibition reflects tendency to inhibit the expression of emotions and behaviours in social interactions and can lead to a breakdown in adaptation and, over time, create a chronic distress which facilitates autonomic imbalance.

View this table:
Table 1

Cattell's 16 personality factors

FactorLow score (−)High score (+)
AReserved nessWarmth
BLow mental capacityHigh mental capacity
CEmotional instabilityEmotional stability
FSoberness, introspectivenessImpulsivity
GLack of conformityConformity
IRealistic, rejecting illusionSensitivity
LRelaxed securitySuspiciousness
OConfident adequacyInsecurity
Q1Conservatism of temperamentRadicalism
Q2Group dependentSelf-sufficiency
Q3Low self-sentiment integrationSelf-discipline
Q4Low tensionHigh tension
Second order factors
QIIILow imaginationSensitivity

Follow-up data

Follow-up was planned for a minimum of 5 years. Follow-up data were obtained from one of the following four sources: review of the patient's record, communication with patient's physician, telephone interview, or examination of the patient at the outpatient clinic.25 The outcome events considered were cardiac death and non-fatal MI. The cause of death was derived from medical records or death certificates. The definition of cardiac death required the documentation of significant arrhythmias, or cardiac arrest, or death attributable to congestive heart failure, or MI in the absence of any other precipitating factor. In patients undergoing coronary revascularization after discharge, follow-up was censored at the time of revascularization and patients were considered alive at that time. Only one event was considered for each patient; all events following the first were excluded.

Statistical analysis

The frequency distribution of each variable was plotted and assessed for skewness using the value of the standardized third moment around the mean. If the distribution was markedly skewed, i.e. had a skewness coefficient exceeding 1.00, a natural logarithmic transformation was applied; this was done for LF and HF power but for simplicity, data are presented in a non-logarithmic form. The Pearson's product–moment correlation coefficient was used to evaluate the association among HRV and psychological variables (continuous data). Two-tailed tests were used, unpaired Student's t-test (continuous data) was used to compare groups; Bonferroni correction was used for subgroup analysis (patients and controls) when many comparison statements were made. P-value <0.05, and when necessary, P<0.001 was considered significant.

The factor analysis23 was performed using the principle components methods.26 To test the goodness of the matrix we applied Kaiser–Meyer–Olkin measure of sampling adequacy, which tests whether the partial correlations among variables were small; Bartlett's test of sphericity, which tests whether the correlation matrix is an identity matrix, which would indicate that the factor model is inappropriate. The variance explained by each factor is the Eigen value for that factor. The values of each factor were normalized and measured on a scale having mean 0 and variance 1.

To identify significant prognostic variables, their individual association with follow-up events was assessed by univariate and multivariable Cox regression analyses.26 Variables have initially been assessed as continuous variables. For ease of communication and for clinical use, we dichotomized the variables when estimating their association with mortality. The best cut-off point for each variable was obtained by means of parametric receiver operating characteristic (ROC) analysis.27 According to a forward stepwise selection process, variables were entered in the model (enter=0.05; remove=0.10) on the basis of the computed significance probability (maximized partial likelihood ratio). Confidence interval (CI) for hazard risk (HR) was calculated at 95%. The HRV measures selected for examination were as follows: 24 h average value of RR, SDNN, SDANN, RMSSD, pNN50, LF, HF, and LF/HF. The psychological measures were the 16 primary traits obtained by 16 PF. Family history of ischaemic heart disease, smoking, arterial hypertension (systolic blood pressure >140 mmHg and/or diastolic pressure >90 mmHg), hypercholesterolaemia (plasma cholesterol >220 mg/dL), hypertrygliceridaemia (plasma triglycerides >160 mg/dL), obesity (body mass index >30 kg/m2), diabetes (fasting plasma glucose >120 mg/dL) were the risk factors analysed. We ran the Cox model separately for HRV measures, for psychological measures, and for risk factors, and then for those variables that resulted significant, adjusted for previously reported post-infarction risk predictors as age, sex, wall motion score index and type of infarction (non-Q vs. Q),1618,25 creatine kinase, and hospital length of stay were also included as covariates. Results were expressed as HR of cardiac death and AMI.

Finally, Kaplan–Meier survival curves were compared with the log-rank test (Mantel–Cox).


The baseline characteristics of the 246 patients and 158 controls are reported in Table 2. Median patient age was 56.5 (51–62 CI 95%). Compared with normal values reported in the literature,15 all HRV variables were markedly lower. Seventy patients were on beta-blockers, 177 on nitrates, and 71 on calcium antagonists.

View this table:
Table 2

Baseline characteristics and test results of studied populations

VariablesPatients (%)Controls (%)
Age years (mean±SD)55±846±10
Sex (males)220 (89)138 (87)
Family history of ischaemic heart disease92 (37)50 (32)
Smokers177 (72)95 (60)
Hypercholesterolaemia83 (34)25 (16)
Hypertrygliceridaemia42 (17)15 (9)
Obesity48 (20)35 (22)
Diabetes mellitus28 (11)
Hypertension90 (37)
Previous MI26 (11)
Type of AMI
 Q204 (83)
 Non-Q42 (17)
Site of AMI
 Anterior109 (44)
 Inferior/lateral137 (56)
Echo wall motion score index ≥1.673 (30)
HRV (patients)Mean±SDMedianCI 95%
RR interval (ms)915±130899816–998
SDNN (ms)111±3210889–129
pNN50 (%)5±631–8
RMSSD (ms)19±81813–24
LF Power (ms2)500±438391224–668
HF Power (ms2)267±29017691–325

Cattell's 16 PF questionnaire

Cattell's 16 PF questionnaire was obtained 13±7 days following admission. When compared with our own control values, the analysis of 16 PF primary traits showed a low score on emotional stability (P=0.0000), sensitivity (P=0.0369), and progress of ideation (P=0.0008); high score on tension (P=0.0000), and insecurity (P=0.0024) (Table 3). By factor analysis of the 16 factor scores, four second-order personality factors were extracted: extraversion, anxiety, sensitivity, and self-control. High score on anxiety (P=0.0000) and low score on extraversion (P=0.0340) and sensitivity (P=0.0233) were obtained. No difference was observed related to type or site of infarction.

View this table:
Table 3

Cattell's 16 personality factors in patients and control subjects

FactorPatients (mean±SD)Controls (mean±SD)P-value
Second order factors

Factors acronyms as in Table 1.

Correlations among HRV and 16 PF

No correlation was found between 16 PF and variability indices, except pNN50 which showed significantly negative correlation with both insecurity and tension (P=0.003).

Prognostic value of HRV and 16 PF for cardiac events

Patients were followed up for 8 years, median 98 (CI 23–110) months. During the follow-up there were 30 cardiac deaths and 19 non-fatal AMI; 72 patients underwent coronary revascularization (coronary bypass or angioplasty). Seventeen patients were lost at follow-up.

Considering survival alone, the most important predictors by univariate analysis (Table 4) were sensitivity <4 and HF power <221 ms2; death was also significantly associated with insecurity <5, pNN50 <5%, mean RR <870 ms, and LF power <148 ms2. In a multivariable analysis, independent and additive prognostic value was found for sensitivity (HR=4.18), HF power (HR=2.76), together with hospital length of stay >15 days and non-Q AMI (Table 4). Interestingly, non-Q had higher mortality rate than Q-AMI (20 vs. 11%, P=0.049). In non-Q, psychological factors only had a predictive power (suspiciousness <5=HR 6.14; imagination <6.6=HR 4.9), while in Q-AMI also HRV indices had prognostic value. The Cox analysis for both death and non-fatal AMI gave similar results: the HRs for low value of HF power and sensitivity factor were independent of one another and events were more likely to occur in non-Q patients (Table 4).

View this table:
Table 4

Univariate and multivariable Cox proportional regression analyses

HR (95% CI)P-value
Predictors of death
Univariate model
 I (<4)3.41 (1.59–7.30)0.037
 HF power (<221 ms2)3.08 (1.44–6.50)0.003
 pNN50 (<5%)2.77 (1.06–7.23)0.021
 O (<5)2.32 (1.13–4.77)0.022
 RR interval (<870 ms)2.25 (1.03–4.92)0.033
 LF power (<148 ms2)2.21 (1.06–4.59)0.031
Multivariable model
 HF power (<221 ms2)2.76 (1.28–5.92)0.003
 I (<4)4.18 (1.88–9.26)0.007
 Hospital length of stay (>15 days)2.60 (1.23–5.52)0.040
 Non-Q vs. Q3.18 (1.35–7.46)0.014
Predictors of death+MI
Univariate model
 Non-Q vs. Q2.07 (1.09–3.90)0.035
 HF power (<221 ms2)2.41 (1.36–4.29)0.002
 O (<5)1.94 (1.10–3.40)0.022
 pNN50 (<5%)1.90 (0.97–3.73)0.046
 I (<4)1.80 (1.02–3.16)0.011
Multivariable model
 HF power (<221 ms2)3.07 (1.65–5.73)0.002
 I (<4)2.69 (1.40–5.18)0.002
 Non-Q vs. Q2.53 (1.31–4.85)0.015

Psychological factors acronyms as in Table 1; pNN50 as in Table 2.

No predictive value could be assigned to LF/HF ratio or major statistical HRV indices other than pNN50. Traditional risk factors as smoking and diabetes failed to show significant predictive power.

Because of the concern of the possible confounding effect of treatment, data were adjusted for beta-blockers, calcium antagonists, and nitrates but no significant differences were found between treated and non-treated groups.

Kaplan–Meier survival curves were constructed identifying patients on the basis of reduced or preserved HF power and/or sensitivity and the 8-year mortality rate is showed in Figure 1. The combination of preserved HF power (>221 ms2) and sensitivity factor >4 defined a subset of patients (35% of the entire population) with a low 8-year mortality rate of 6% which significantly differed from that of patients with a combination of both risk predictors (62%).

Figure 1 Kaplan–Meier survival curves for cardiac death: patients with HF power below 221 ms2 and sensitivity, factor I<4 had higher number of events than those with one or both preserved variables.P-value refers to differences in event rates between the groups. HF, high frequency; I, sensitivity, factor I.


This study clearly indicated that specific personality traits, such as low emotional sensitivity and insecurity, greatly impact long-term prognosis after AMI, and are associated with cardiac death and re-infarction. Moreover, the study confirmed the long-term prognostic value of frequency domain indices of HRV. Other factors such as type of infarction and hospital length of stay were also associated with adverse outcome.

Personality traits in AMI

In accordance with the previous studies, patients resulted to be anxious, introvert, socially inhibited, with a high control over emotions.1,3,28 This inhibition is expressed by low emotional sensitivity, an inability to express emotion, and adapting to society or coping. Low emotional sensitivity has been associated with serious diseases including hypertension and coronary heart disease.29,4 Social inhibition reflects tendency to inhibit the expression of emotions and behaviours in social interactions and can lead to a breakdown in adaptation 30 and, across time, create a chronic distress which facilitates autonomic imbalance, primary source of sudden death.12,31 This study suggests, in quite a large population with a great number of events, that personality may matter in predicting prognosis even independently of left ventricular dysfunction, as it was the case in other papers.7

Our results add to previous knowledge in the same field by showing that personality traits can emerge in the sub-acute phase of MI, sufficiently far from the confounding effect of the acute phase. Moreover, it is interesting to underline that this cardiovascular population was clinically free of history of psychiatric disease and was not on psychotropic medications. It might be expected that personality would be investigated by specific personnel. The choice of a simple bedside questionnaire routinely collected by cardiologists played a role in patient satisfaction.

Risk factors and prognosis

Few studies, the majority on depressive patients, have tried to explain the connection between psychological distress and prognosis in coronary heart disease; different mechanisms have been suggested, such as chronic immune stimulation,32 abnormalities in platelet and endothelial function,33 impairment of serotonin re-uptake,34 and impairment of autonomic modulation as detected by HRV.35

At univariate analysis more than one HRV index, low LF and HF power, relative tachycardia, and low pNN50, were predictors of cardiac death at 8-year follow-up. However, at multivariable analysis, only low HF power remained predictive, together with personality trait and two clinical variables, type of AMI, and duration of hospital stay. When both death and new non-fatal MI were considered, the independent prognostic power of the same variables, except for duration of hospital stay, was confirmed. The combination of both risk predictors identified a subset of patients (10% of this population) with a very high 8-year mortality rate (38%).

The prognostic value of HRV has already been reported;16,17 the novelty of our results is the very long-term prognostic impact of spectral indices. High frequency oscillations reflect the parasympathetic outflow;15 thus, the apparent decrease in a marker of tonic parasympathetic activity might indicate a reduction of vagal protective effect on the emergence of life-threatening arrhythmias persisting over time in patients with poor outcome.

In the literature, the 70 ms cut-off for SDNN identified patients at high risk for cardiac events during long-term follow-up.36 In the present study SDNN was ≤70 ms in 19 patients (8%), and lower than 50 ms in only five patients. This might explain the lack of statistical significance of this index in our population.

The lack of effect of therapy on HRV indexes in our population of post-infarction patients is in agreement with the literature15 and may be explained by the limited period of drug treatment. We recognize that some, or perhaps many, of HRV and psychological factors could have changed during this long follow-up interval and more measurements could have been done. Although it is extremely burdensome to recall all patients in for physical examination, we cannot neglect the long-term power of the parameters collected during the acute event on the prediction of mortality.

Clinical variables

Non-Q wave AMI and prolonged hospital length of stay were the only other predictors which could be the expression of perdurable clinical instability.

One of the characteristics of the study was the relatively young age of the population in accordance to the design of the study, which aimed to assess the prognostic role of personality and behavioural traits. This cut-off certainly limited the number of hard events during the follow-up, as supported by previous reports.36,37 However, such a cut-off also minimized the well-known confounding effect of age on both HRV15 and personality,4 and restricted the study to a homogeneous young population for which finding strong predictors is particularly important.


Study limitation

The reduction from the initial 455 patients enrolled in the original multi-centre study to the final group of 246 patients used for this paper could erase the issue of selection bias. The necessity of having patients with a second Holter recording, psychological questionnaire and follow-up was the only reason to justify this reduction.

One other characteristic of the study was that it was conducted in a predominantly Italian male population. This limitation is related to the predominance of male patients with MI, which is valid in all countries. The per cent of smokers was far higher than that would be found in other post-MI groups. Certain personality traits go along with smoking which could be a surrogate of them. The level of physical activity which could have an impact on HRV was not determined.

In conclusion, this research underlines that low HF power is predictive of cardiac death at 8-year follow-up and that clinicians should be alert to the potential prognostic power of personality and should consider expanding their assessment of post-AMI patients to include simple measure of personality to detect patients at risk, in whom an ad hoc treatment could be of some benefit.


We thank Antonella Niccolini and Elena Barberini for their expert assistance in follow-up telephone interviews. We also thank Manuella Walker for the revision of English style. Study was supported in part by Grant 104299/41/93/04986 from the CNR, Targeted Project FATMA.SP8.

Appendix: Enrolling Centres

  • Istituto di Fisiologia Clinica CNR (C Carpeggiani, A L'Abbate) Pisa

  • Divisione di Malattie Cardiovascolari (G Bigalli) Pisa

  • UTIC (A Pesola, R Gistri) Viareggio

  • Divisione di Cardiologia (M Lupetti) Pescia, Pistoia

  • Divisione di Cardiologia (G Maggini) Livorno

  • UTIC - USL-16 Valdera (PL Topi, D Levantesi) Pontedera

  • Divisione di Cardiologia (G Micheli) Piombino

  • Divisione di Cardiologia (E Nannini) Castelnuovo Garfagnana

  • Divisione di Cardiologia (AM Ballestra) Volterra

  • UTIC - Ponte a Niccheri (M Barchielli, R Vergassola) Firenze

  • Divisione di Cardiologia (M Lombardi, D Bernardi) Barga

  • Istituto Scientifico di Medicina Interna (P Spallarossa) Università di Genova

  • II Divisione Cardiologica (D Granata, C Belli) Niguarda, Milano

  • Divisione di Cardiologia (A Pozzolini) Urbino


View Abstract