The sum of time–voltage QRS areas in the 12-lead electrocardiogram (ECG) has outperformed other 12-lead ECG indices for detection of left ventricular hypertrophy (LVH). We assessed indices of time–voltage QRS and T-wave (QRST) areas from body surface potential mapping (BSPM) for detection of and quantitation of the degree of LVH. We studied 42 patients with echocardiographic LVH (LVH group) and 11 healthy controls (controls). QRST area sums were calculated from 123-lead BSPM and from the 12-lead ECG for comparison. Leadwise discriminant indices and correlation coefficients were used to identify optimal recording locations for QRST area-based LVH assessment. BSPM QRS area sum was greater in the LVH group than in controls (3752 ± 1259 vs 2278 ± 627 μV s, respectively; P<0.001) and at 91% specificity showed 74% sensitivity for LVH detection. The 12-lead QRS area sum performed similarly. Taking T-wave areas into account did not improve the results. QRS area sum from two most informative leads (located in the upper and lower right precordium) also separated the LVH group from controls (61.1 ± 23.5 vs 27.8 ± 6.5 μV s, respectively; P<0.00001). This 2-lead QRS area sum showed 90% sensitivity with 100% specificity for LVH detection and maintained high correlation to indexed left ventricular mass (r=0.732; P<0.001). In conclusion, the BSPM QRS area sum compared to 12-lead QRS area sum does not substantially improve LVH assessment. The 2-lead QRS area sum may improve ECG QRS area-based LVH assessment.
Both identification of patients with left ventricular (LV) hypertrophy (LVH) and assessment of the degree of LVH are clinically important, because detection of LVH independently predicts cardiovascular morbidity as well as mortality and the relative risk of these events increases with increasing LV mass.1,2,3 Echocardiography and magnetic resonance imaging, in particular, provide accurate estimates of anatomic LV mass, but neither of these methods is applicable to wide-scale screening of LVH.4,5,6
According to the solid-angle theory, LVH augments QRS voltages on the body surface electrocardiogram (ECG),7 but for clinical purposes QRS voltage-based LVH indices fail to identify LVH patients with sufficient accuracy.8 In the orthogonal-lead signal-averaged ECG, the QRS time–voltage area in the horizontal plane significantly improves ECG detection of LVH.9,10 Even better performance for eccentric LVH detection can be achieved by summing all the QRS time–voltage areas in the 12-lead ECG.11 However, the electrode positioning used in the 12-lead ECG may not be optimal for QRS area-based LVH detection.12 In addition, the T-wave may contain information of LV mass not revealed by restricting the measurements to the QRS complex.12,13,14,15
In the present study, our first aim was to assess the capability of body surface potential mapping (BSPM) QRS area analysis for detection and quantification of pressure overload-induced LVH compared to the 12-lead ECG QRS area sum as well as to the Sokolow–Lyon index,16 and to study if calculating QRST areas improves the BSPM methodology. Our second aim was to identify the optimal electrode positioning for QRS area-based LVH assessment and thereby derive a simple, but informative ECG LVH index.
Subjects and methods
The study population (n=53) consisted of two groups. The LVH group (n=42) included patients with haemodynamically significant aortic valve stenosis (n=27) and patients with essential arterial hypertension (n=15), all of whom showed echocardiographic LVH by gender-specific criteria (see below). The patients with aortic valve stenosis were being invasively evaluated in the Helsinki University Central Hospital. None of these patients showed ≥70% stenosis of luminal diameter of epicardial coronary arteries in routine coronary angiography. No patient with essential arterial hypertension had a history of effort angina. No patient in the LVH group showed pathological Q-waves in the ECG or regional wall-motion abnormalities in LV cineangiography or two-dimensional echocardiography suggesting a previous myocardial infarction. Patients with a right or left bundle branch block were not included.
The control subjects (n=11; controls) were healthy middle-aged volunteers who were recruited with a newspaper advertisement. Consecutive subjects who had no major cardiovascular risk factors and who had no history or signs of cardiovascular disease were further examined. If they showed a normal routine echocardiogram and a symptom-limited bicycle exercise test without angina or ST-segment changes, they were included as controls.
All patients in both study groups were in sinus rhythm and none was under medication known to influence QRS or T-wave morphology. The study complied with the Declaration of Helsinki. Informed consent was obtained from all participants and the study protocol was approved by the ethical review committee of the Helsinki University Central Hospital.
A skilled cardiologist (MaK) performed a standard M-mode and two-dimensional echocardiography to all study subjects. Two-dimensionally guided M-mode tracings were recorded from the parasternal long-axis view with particular care taken to ensure proper orientation of the ultrasonic beam. End-systolic and end-diastolic LV wall thicknesses and dimensions were then measured on a digitizing table by the cardiologist who was blinded to any clinical data according to the recommendations of the American Society of Echocardiography.17 An average of up to five measurements from consecutive cardiac cycles for each measure was used in the final analyses. LV mass was calculated with an anatomically validated formula (r=0.90 vs autopsy LV mass) from M-mode measurements made according to the Penn convention.4 Calculated LV mass and also LV mass indexed to body surface area (to account for the effect of physiological variation in LV mass) were both used in the analyses. The patients included in the LVH group showed indexed LV mass >116 g/m2 in men and >104 g/m2 in women.18 In patients with echocardiographic LVH, the LV geometric pattern was considered as concentric if the relative wall thickness19 exceeded 0.43; otherwise, it was considered to be eccentric.
Measurement protocol: The BSPM recording device and procedure have been described in detail elsewhere.20,21,22 In brief, the BSPM device records unipolar potentials with 120 electrodes placed on the subject's thorax (Figure 1), and three standard limb potentials. The recorded signals were digitized with a sampling frequency of 1 kHz, and selectively averaged off-line.
LVH measures: From the signal-averaged BSPM data, the time instants of QRS onset, QRS offset, and T-wave offset in each lead were determined as described previously.23,24 The following BSPM indices were calculated in each lead: (1) QRS area by integrating the absolute value of the BSPM signal over the QRS complex.11 (2) T-wave area by integrating the value of the BSPM signal from the QRS offset to the T-wave offset. To relate this index value to QRS-T discordance (opposite polarity of main QRS deflection and T-wave), areas of the T-wave in leads with discordant QRS and T waves were considered negative (Figure 2). (3) QRS-T area by subtracting the T-wave area parameter from the QRS area parameter. The rationale of calculating QRS and T-wave areas differently was based on a previous study showing that the ECG response to LVH is flattening of the T-wave or polarity reversal of the T-wave in relation to the main QRS deflection.13 QRS area sum, T-wave area sum, and QRS-T area sum were derived by calculating the sum of respective values from all leads.
Optimal recording locations for QRS area: To identify the recording locations with the best discriminative power for LVH detection, we used the discriminant index (DI).12 In each lead and for each index, DIs were calculated by subtracting the control group mean value from the LVH group mean value. The obtained difference was then divided by the pooled standard deviation of the study groups to derive leadwise DIs.12,25 The pooled standard deviation (s) was derived from pooled variance (s2) calculated by the equation s2=[(n1−1)s12+(n2−1)s22]/[n1+n2−2], where s1 and s2 are the standard deviations of the two groups of sizes n1 and n2.12,25,26 A large DI value indicates good performance of that lead in separating the LVH patients from the controls. We then constructed discriminant maps12 displaying DI value distributions over the thorax.
As the DI uses groups based on dichotomized values, it does not necessarily identify informative leads for quantification of a continuous parameter, such as LV mass. Therefore, we also constructed correlation coefficient maps displaying the correlation coefficient between QRS area and LV mass index in each lead. In an attempt to derive a simple index for LVH detection and quantification, we searched for particularly informative leads with both good discriminative power between LVH group and controls as well as with a good correlation between QRS area and LV mass index for further analysis.
A 12-lead digital ECG was recorded before the BSPM measurement. However, because in some of the ECGs there were a few missing leads, we used a derived 12-lead ECG from the BSPM measurement in all study subjects. Precordial leads V1–V2 and V4–V6, as well as bipolar limb leads are part of the BSPM layout (Figure 1). The signal in precordial lead V3 was interpolated from four neighbouring leads (distances ≈2 cm) and the signals in unipolar augmented limb leads were calculated according to standard equations.27
From the averaged 12-lead ECG signal, we measured the Sokolow–Lyon index (SV1+RV5/V6 (whichever is taller)).16 In each lead, the time–voltage QRS area was calculated similarly as in the BSPM leads, and the areas from the 12 leads were summed (12-lead QRS area sum).11 In patients with complete measurements, the QRS area sum from the BSPM-derived 12-lead ECG correlated extremely closely with that obtained from the original 12-lead ECG (r=0.956).
Statistical data analysis was performed with SPSS version 10.1 software (SPSS Inc, Chicago, IL, USA). All continuous data are presented as mean±S.D. The significance of difference in continuous variables between the study groups was determined with the Mann–Whitney U-test. Correlation between continuous variables was calculated using Pearson's correlation coefficients and using partial correlation coefficients when adjusting for covariates. All categorical variables were tested with the χ2 test or with Fisher's exact test when appropriate. Two-tailed P-values <0.05 were considered statistically significant.
Clinical and echocardiographic characteristics
The clinical and echocardiographic characteristics of the study groups are shown in Table 1. Of the 42 patients in the LVH group, 33 (78.6%) had concentric and nine (21.4%) eccentric LVH.
BSPM and 12-lead ECG indices of LVH in the study groups
In BSPM, the QRS area sum separated the LVH group from controls better than T-wave area sum or QRS-T area sum (Table 2). In the 12-lead ECG, the Sokolow–Lyon voltage as well as the QRS area sum performed equally well as the BSPM QRS area sum for identifying LVH patients (Table 2). Using the conventional cutoff value of 3.5 mV, the sensitivity of Sokolow–Lyon voltage for detection of LVH was 50% with 91% specificity. With matched specificity of 91% (one false positive in the controls) and optimizing cutoff points in the present study population, the sensitivity of the 12-lead QRS area sum (cutoff value 440 μV s) was 69% and that of BSPM QRS area sum (cutoff value 3000 μV s) was 74%.
Optimal BSPM leads for QRS area assessment in LVH detection and quantitation
By use of discriminant maps and correlation maps, we identified several locations with both a high DI value for QRS area (Figure 3) and a high correlation coefficient value between QRS area and LV mass index. These were leads 12 and 13 on the lower right precordium, leads 10 and 16 in the upper right precordium, and lead 82 on the left lower flank area (see Figure 1). The sum of QRS areas in leads 13 and 16 (2-lead QRS area sum) also separated the LVH group from controls (Table 2), and the sensitivity for LVH detection (cutoff value 39 μV s) was 90% with 100% specificity (Figure 4). Other additional sum combinations from these five leads did not improve the performance. Furthermore, summing up other additional combinations of QRS area, T-wave area and/or QRS-T area values from informative leads did not perform substantially better (data not shown).
Correlation of BSPM and 12-lead ECG QRS area measures to LV mass and structure
Both in the overall study population and within the LVH group, the BSPM QRS and QRS-T area sums showed a stronger correlation to LV mass and LV mass index than the Sokolow–Lyon voltage, whereas the BSPM index correlations were only slightly stronger than or equal to those of the 12-lead QRS area sum (Table 3). In fact, the BSPM and 12-lead QRS area sums were almost linearly correlated (r=0.967; P<0.001), suggesting that they provide practically identical information. Scatterplots showing the actual values for BSPM 2-lead QRS area sum and 12-lead QRS area sum vs LV mass index are shown Figure 4. In the overall study population, BSPM QRS area sum and 2-lead QRS area sum correlated strongly with echocardiographic interventricular septal wall thickness (r=0.694 and 0.654, respectively; P<0.001 in both) as did also Sokolow–Lyon voltage and 12-lead QRS area sum (r=0.542 and 0.685, respectively; P<0.001 in both). None of these indices correlated with LV internal diastolic diameter (r values betweeen −0.085 and 0.103; P>0.4 in all). All of the above correlations remained significant and substantially in the same order when using partial correlation coefficients adjusting for the possible effect of age and/or gender on these parameters.
Our results show that BSPM QRS area sum can separate the group of LVH patients without evidence of coronary artery disease from controls, but for identifying patients with LVH it does not perform substantially better than the 12-lead QRS area sum. The sum of QRS areas from the two most informative leads identified by leadwise BSPM analysis improved detection of LVH compared to 12-lead ECG analyses and retained the ability to quantitate the degree of LVH. Both these leads were located outside the conventional 12-lead ECG electrode positions.
BSPM and 12-lead ECG detection and quantitation of LVH
Detection of patients with LVH is clinically important, because the risk of cardiovascular morbidity and death is 2–4-fold higher in subjects with LV mass above normal limits than in those with normal LV mass.1,2,3,6,18 In addition, quantification of the degree of LVH is equally important, since the risk of these events increases as LV mass increases.2,3 Direct visualization of the LV wall thicknesses and dimensions by echocardiography enables calculation of LV mass based on certain geometrical assumptions,4,6 but the accuracy with this methodology is obviously expertise-dependent. Magnetic resonance imaging offers accurate assessment of anatomic LV mass, but the upper normal limits with this methodology have only recently been established28 and the method atpresent is unsuitable for screening purposes.5 Therefore, an attempt to develop a widely applicable screening method for LVH assessment is still warranted.
Pressure-overload of the LV induces myocardial cellular hypertrophy and, thus, an increase in electrically active LV mass. According to both computer simulations and the solid-angle theory, an increase in LV mass is expected to strengthen QRS voltage on the body surface ECG.7,13 However, clinically QRS voltage-based LVH indices in the 12-lead ECG fail to identify individual LVH patients with sufficient accuracy. This may in part be explained by the effect of several extracardiac factors, such as age, gender, chest diameter, and body build, on ECG QRS voltages.8 Attempts to take these factors into account alone does not suffice for critical improvement of the methodology. Measuring only maximum QRS deflection amplitudes overlooks the more powerful discriminating and quantitation capabilities of information in other segments within the QRS complex.12,29,30 Accordingly, the time–voltage integrated QRS area in the horizontal plane from the orthogonal signal-averaged ECG improves ECG detection of LVH,9,10 and the sum of QRS areas in the 12-lead ECG results in further improvement of performance for LVH detection.11 Importantly, the 12-lead QRS area sum may be relatively independent of extracardiac factors and gender.11 Even though the QRS area scans the temporal aspects of the depolarization wavefront, it is possible that the positioning of the electrodes in the 12-lead ECG is not spatially optimal for QRS area-based LVH detection. This is suggested by the distribution of most informative leads in BSPM outside the conventional 12-lead ECG electrode positioning12,29,30 and by the fact that orthogonal ECG does not capture all the available information for the assessment of the activation wavefronts.9,10,12,29,30 In addition, the T-wave may contain complementary information of LV mass,12,13,14,15,29,30 but QRST areas have not been assessed for this purpose. Methods developed for LVH detection utilizing the BSPM technique have improved the identification and even quantification of anatomic LVH,12,29,30,31,32,33,34 suggesting that accurate assessment of anatomic LVH from electrical signals is indeed possible.
Based on the above background, we hypothesized that BSPM mapping of QRS and QRST areas may improve the detection and quantitation of LVH. In addition, we determined whether the information redundancy inherent in BSPM might be compressed to yield a simple index that could be recorded by a simple modification of lead placement with the 12-lead ECG equipment. Our results show that BSPM QRS area sum can separate the group of patients with pressure overload-induced LVH with mostly concentric LVH from controls and can quantify the degree of LVH, but the 12-lead QRS area sum is, in both respects, a very accurate estimate of BSPM QRS area sum. Furthermore, with the approach we used taking T-wave areas into consideration does not improve performance. However, our results suggest that a simple sum of QRS areas from the upper and lower right precordial areas improves performance compared to the 12-lead QRS area sum. Although this 2-lead QRS area sum was not an a priori index, the previous BSPM mapping studies have shown that for QRS voltage measurements the right precordial area is particularly informative for LVH assessment.12,30 This suggests that our results may be more generalizable and prompts the prospective testing of our approach.
An obvious limitation to our study is the relatively small study population and the use of echocardiography as a reference method,5 since even with meticulous measurement technique, the assessment of LV mass with M-mode echocardiography is not very accurate.35 However, at present screening of patients for a study like ours with magnetic resonance imaging is impractical. The numerical values, including the values of upper limit of normal, in the article by Okin et al11 should be treated with caution and their approach was not tested in an independent test set. Thus, because of the lack of values for upper normal limits, we compared our results with matched specificities. The upper limits of normal should be derived from a larger sample of unselected subjects and include investigation of potential gender differences, although Okin et al11 found their criteria not gender dependent. Deriving the 2-lead QRS area sum index from a training set and not testing it subsequently in an independent test set is likely to optimize the results. Therefore, to test if the 2-lead QRS area index truly improves clinical LVH assessment, the sensitivity and specificity of our approach should be assessed in a separate population and also including more patients with borderline LVH. Although with our approach we found no additional information in the T-wave area sums, other methods incorporating separately QRS and T-wave area information or using, for example, the concept of QRST gradients may offer additional information from repolarization for LVH assessment.36 Indexing LV mass to body surface area is an attempt to account for the variation in physiological determinants of LV mass and to enable assessment of correlation to pathologically increased LV mass, but has certain limitations.37
Our present results show that the 12-lead QRS area sum captures the same information for LVH detection and quantitation as the BSPM QRS area sum. The 12-lead QRS area sum performs relatively well in a study population of patients with mainly concentric LVH, which has not been previously shown.11 A simple 2-lead QRS area sum from the right precordium may be a way to improve ECG detection of LVH with 12-lead ECG equipment. If prospectively validated, this index may be useful in screening patients with LVH and in preliminary assessment of the degree of LVH as well as in interventional studies for assessment of regression of LVH.
Casale PN et al. Value of echocardiographic measurement of left ventricular mass in predicting cardiovascular morbid events in hypertensive men. Ann Intern Med 1986; 105: 173–178.
Levy D et al. Prognostic implications of echocardiographically determined left ventricular mass in the Framingham Heart Study. N Engl J Med 1990; 322: 1561–1566.
Schillaci G et al. Continuous relation between left ventricular mass and cardiovascular risk in essential hypertension. Hypertension 2000; 35: 580–586.
Devereux RB et al. Echocardiographic assessment of left ventricular hypertrophy: comparison to necropsy findings. Am J Cardiol 1986; 57: 450–458.
Myerson SG, Bellenger NG, Pennell DJ . Assessment of left ventricular mass by cardiovascular magnetic resonance. Hypertension 2002; 39: 750–755.
Devereux RB, Pini R, Aurigemma GP, Roman MJ . Measurement of left ventricular mass: methodology and expertise. J Hypertens 1997; 15: 801–809.
Holland RP, Arnsdorf MF . Solid angle theory and the electrocardiogram: physiologic and quantitative interpretations. Prog Cardiovasc Dis 1977; 19: 431–457.
Surawicz B . Electrocardiographic diagnosis of chamber enlargment. J Am Coll Cardiol 1986; 8: 711–724.
Okin PM et al. Electrocardiographic diagnosis of left ventricular hypertrophy by the time–voltage integral of the QRS complex. J Am Coll Cardiol 1994; 23: 133–140.
Okin PM, Roman MJ, Devereux RB, Kligfield P . Time–voltage area of the QRS for the identification of left ventricular hypertrophy. Hypertension 1996; 27: 251–258.
Okin PM et al. Time–voltage QRS area of the 12-lead electrocardiogram. Detection of left ventricular hypertrophy. Hypertension 1998; 31: 937–942.
Kornreich F et al. Identification of best electrocardiographic leads for diagnosing left ventricular hypertrophy by statistical analysis of body surface potential maps. Am J Cardiol 1988; 62: 1285–1291.
Thiry PS, Rosenberg RM, Abbott JA . A mechanism for the electrocardiogram response to left ventricular hypertrophy and acute ischemia. Circ Res 1975; 36: 92–104.
Devereux RB, Reichek N . Repolarization abnormalities of left ventricular hypertrophy. Clinical, echocardiographic and hemodynamic correlates. J Electrocardiol 1982; 15: 47–54.
Okin PM et al, for The LIFE Study Investigators. Relationship of the electrocardiographic strain pattern to left ventricular structure and function in hypertensive patients: The LIFE Study. J Am Coll Cardiol 2001; 38: 514–520.
Sokolow M, Lyon TP . The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J 1949; 37: 161–168.
Sahn DJ, DeMaria A, Kisslo J, Weyman A . Recommendations regarding quantitation in M-mode echocardiography: results of a survey of echocardiographic measurements. Circulation 1978; 58: 1072–1083.
Liu JE, Devereux RB . Clinical assessment of cardiac hypertrophy. In: Sheridan DJ (ed). Left Ventricular Hypertrophy. Churchill-Livingstone: London, 1998, pp 11–16.
Reichek N, Devereux RB . Reliable estimation of peak left ventricular systolic pressure by M-mode echographic-determined end-diastolic relative wall thickness: identification of severe valvular aortic stenosis in adult patients. Am Heart J 1982; 103: 202–209.
Simelius K et al. A body surface potential mapping system in clinical use. Med Biol Eng Comput 1996; 34: 107–108.
Takala P et al. Magnetocardiographic and electrocardiographic exercise mapping in healthy subjects. Ann Biomed Eng 2001; 29: 501–509.
Hänninen H et al. ST-segment level and slope in exercise-induced myocardial ischemia evaluated with body surface potential mapping. Am J Cardiol 2001; 88: 1152–1156.
Simson MB . Use of signals in the terminal QRS complex to identify patients with ventricular tachycardia after myocardial infarction. Circulation 1981; 64: 235–242.
Oikarinen L et al. Magnetocardiographic QT interval dispersion in postmyocardial infarction patients with sustained ventricular tachycardia: validation of automated QT measurements. Pacing Clin Electrophysiol 1998; 21: 1934–1942.
Kornreich et al. Identification of best electrocardiographic leads for diagnosing anterior and inferior myocardial infarction by statistical analysis of body surface potential maps. Am J Cardiol 1986; 58: 863–871.
Altman DG . Comparing groups — continuous data. In: Altman DG (ed). Practical Statistics For Medical Research. Chapman & Hall: London, 1995, pp 179–228.
Macfarlane PW . Lead systems. In: Macfarlane PW, Lawrie TDV (eds). Comprehensive Electrocardiology. Pergamon Press: Oxford, 1989, pp 315–352.
Salton CJ et al. Gender differences and normal left ventricular anatomy in an adult population free of hypertension. A cardiovascular magnetic resonance study of the Framingham Heart Study offspring cohort. J Am Coll Cardiol 2002; 39: 1055–1060.
Kornreich F et al. Diagnostic body surface potential map patterns in left ventricular hypertrophy during PQRST. Am J Cardiol 1989; 63: 610–617.
Kornreich F et al. Improved prediction of left ventricular mass by regression analysis of body surface potential maps. Am J Cardiol 1990; 66: 485–492.
Holt JH, Barnard ACL, Lynn MS . A study of the human heart as a multiple dipole electrical source. II. Diagnosis and quantitation of left ventricular hypertrophy. Circulation 1969; 40: 697–710.
Holt JH, Barnard ACL, Kramer JO . Multiple dipole electrocardiography. A comparison of electrically and angiographically determined left ventricular masses. Circulation 1978; 57: 1129–1133.
Dunn RA et al. Performance of conventional orthogonal and multiple-dipole electrocardiograms in estimating left ventricular muscle mass. Circulation 1979; 60: 1350–1353.
Yamaki M et al. Improved diagnostic performance on the severity of left ventricular hypertrophy with body surface potential mapping. Circulation 1989; 79: 312–323.
Palmieri V et al. Reliability of echocardiographic assessment of left ventricular structure and function. The PRESERVE Study. J Am Coll Cardiol 1999; 34: 1625–1632.
Montague TJ et al. Isointegral analysis of body surface maps: surface distribution and temporal variability in normal subjects. Circulation 1981; 63: 1166–1172.
Rautaharju PM, Zhou SH, Park LP . Improved ECG models for left ventricular mass adjusted for body size, with specific algorithms for normal conduction, bundle branch blocks, and old myocardial infarction. J Electrocardiol 1996; 29(Suppl): 261–269.
This work was supported by grants from the Finnish Foundation for Cardiovascular Research and the Aarne Koskelo Foundation.
About this article
Cite this article
Oikarinen, L., Karvonen, M., Viitasalo, M. et al. Electrocardiographic assessment of left ventricular hypertrophy with time–voltage QRS and QRST-wave areas. J Hum Hypertens 18, 33–40 (2004). https://doi.org/10.1038/sj.jhh.1001631
- body surface potential mapping
- left ventricular hypertrophy
Relationship between QRS measurements and left ventricular morphology and function in asymptomatic individuals
American Heart Journal (2017)
γ‐Glutamyltransferase and Risk of Sudden Cardiac Death in Middle‐Aged Finnish Men: A New Prospective Cohort Study
Journal of the American Heart Association (2016)
Baseline and long-term fibrinogen levels and risk of sudden cardiac death: A new prospective study and meta-analysis
Heart Rhythm (2015)