Available online at www.sciencedirect.com
Journal of Electrocardiology 41 (2008) 190 – 196 www.jecgonline.com
Review article
Intraindividual variability in electrocardiograms Bob J.A. Schijvenaars, PhD, Gerard van Herpen, MD, PhD, Jan A. Kors, PhD ⁎ Department of Medical Informatics, Erasmus University Medical Center, Rotterdam, The Netherlands Received 21 December 2007; accepted 31 January 2008
Abstract
The electrocardiogram (ECG) can be affected by intraindividual variations from various sources that may confuse the diagnosis of the underlying cardiac condition and impair the accuracy of ECG interpretation. Intraindividual variability is a hindrance in serial ECG analysis, where ECGs of the same individual, but taken at different times, are compared. Two sources of intraindividual variability can be distinguished as follows: variability related to the technical circumstances during ECG recording (technical sources) and nonpathologic biologic variability (biological sources). Among the technical sources, variation in electrode positioning between recordings is the most confusing. Of the biological sources, respiratory variations are effective at any time scale, but the most important are age and weight that work on prolonged time scales. Technical problems are best prevented by rigorously sticking to a standard acquisition protocol. Criteria can be adapted to changing circumstances (age, weight), and by computer modeling, it may be possible to correct the ECG diagnosis for some sources of intraindividual variability. © 2008 Elsevier Inc. All rights reserved.
Keywords:
Electrocardiogram; Intraindividual variability; Interindividual variability; ECG parameters
Introduction To interpret electrocardiograms (ECGs), one must have knowledge about the distribution of ECG measurements in a normal population. A number of studies have provided data on normal limits (eg, Simonson1 and Macfarlane and Lawrie2). The variability implied in the term “normal limits” becomes evident when normal ECGs are taken from different individuals. This interindividual variability should be distinguished from intraindividual variability and from the variability within one ECG, also called beat-to-beat variability. Intraindividual variability is an important confounder in serial ECG analysis, where ECGs of the same individual, but taken at different times, are compared. In the following review, restricted to the resting ECG, 2 sources of intraindividual variability are distinguished, viz, the technical circumstances during ECG recording (technical sources) and nonpathologic biologic variability (biological sources). Interindividual ECG variability will be briefly considered to put the 2 sources of intraindividual variability
⁎ Corresponding author. Department of Medical Informatics, Erasmus University Medical Center, PO Box 2040, 3000 CA Rotterdam, The Netherlands. Tel.: +31 10 7043045; fax: +31 10 7044722. E-mail address:
[email protected] 0022-0736/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jelectrocard.2008.01.012
into context. Possible ways to deal with intraindividual variability are discussed. The reader is referred to the literature for intraindividual ECG variability in exercise ECGs,3 ambulatory monitoring,4 patient monitoring, and signal-averaged ECGs.5 Technical sources of ECG variability In the past, considerable variation was caused by differences in the quality and type of recording equipment and procedures and in the lead system used. This type of variation still exists but has become of less importance. In the following, the technical sources that still can play a role are discussed. Many of the pertinent studies are older than 10 years, but this does not detract from their validity today. Electrode placement Correct, standardized placement of electrodes is crucial for accurate ECG recording. The 12-lead ECG system is nowadays widely accepted as the standard.6 Electrodes for monitoring are often located at specific places according to local preferences. Comparison of such ECGs with the standard ECG should be done with caution.7,8 The precordial electrodes require careful positioning guided by palpation of the bony structures of the chest. Already in 1958, August et al9 presented a few normal ECGs
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196
that mimicked acute anterior myocardial infarction after slight electrode-positional changes. Several studies have since determined the occurrence rate and severity of these errors.7,10,11 Common errors are placing an electrode on a rib or on the sternum rather than on an intercostal space or placing it in the wrong intercostal space. Accurate placement was performed by only 13%7 to 25%11 of critical care nurses during placement tests, regardless of their experience. In nontest situations, this percentage even decreased to 9%.11 Wenger et al10 found that the average distance from the actual electrode positions to the prescribed locations was 2.9 cm. Shifts of more than 1.6 cm upward were encountered in 50% of V1 to V2 placements and to the left and downward in 30% to 50% of V4 to V6 placements. Devices have been constructed to facilitate and guide precordial lead placement.12-14 Willems et al15 found a reduction of 25% in day-to-day variability when marked electrode positions were used. Intracutaneous dye injections have been recommended for the purpose.15,16 The men who, long ago, proposed the standard chest electrode positions had little eye for the female breast. When placed under the breast, electrodes tend to be pushed downwards. Rautaharju et al,17 from a study involving 6814 women, recommend placing the electrodes on the breast. Adequate instruction and proper supervision of the technicians and nurses are the best guarantee for proper electrode placement. The limbs are in effect not much more than extensions of the electrode cables so that variation in placement up or down arm or leg is of no consequence. Shifting the extremity electrodes to the trunk, however, may not be without undesired effects. Chest electrode malpositioning In how far will chest electrode malpositioning result in changes in ECG measurements and diagnosis? Two studies on a small number of subjects (n = 13 and n = 23) indicate that most precordial ECG measurements undergo considerable changes by electrode displacement.12,18 The effects on diagnosis, however, seem smaller; clinically significant changes in interpretation by cardiologists were found for 2 of 13 subjects and in computerized interpretation for five.18 In another study, only clinically inconsequential changes were found.12 Herman et al14 compared ECGs recorded using their electrode-positioning device with ECGs obtained after deliberate electrode misplacement more than 2 cm upward and downward. Considerable measurement variations were observed in all 15 cases, whereas the interpretation changed in 9 and 10 cases for computer and human reader, respectively. In another group of 80 patients, where device-aided placement was compared to routine placement, 60% of the ECGs had a considerable change in at least one of 25 measurements.19 The measurement changes had clinical significance in 16% of the subjects when analyzed by an experienced reader and in 10% when interpreted by computer. Schijvenaars et al20 simulated 4 types of electrode position changes in a set of 746 ECGs from healthy individuals and from patients with myocardial infarction and left ventricular hypertrophy (LVH). At most, 6% of the cases
191
showed important changes in diagnostic classification by computer. Expert cardiologists made about half that number of changes. Lead reversal A different type of electrode misplacement is the reversal of electrodes, which may mimic a pathologic condition and have a considerable effect on ECG interpretation.21-25 For the 4 extremity electrodes, the possible number of interchanges is 4! − 1 = 23; for the chest electrodes, it is 719; for all electrodes together, more than 3.5 million. Obviously, only the most common reversals have to be considered. Interchanges between extremity and precordial electrodes are rare because the cables are often ordered group-wise and made of different length in modern ECG equipment. A common and easily detected error is left/right arm electrode reversal.24 Leg-arm reversals are hard to detect. Haisty et al25 reported that none of 25 cardiologists recognized a right arm/right leg electrode switch in a normal ECG although lead II is reduced to zero in this case and present a detection algorithm for it. Several others have also developed electrode reversal detection algorithms.26-32 Reversal of precordial electrodes results in an incongruity in the wave sequences over the precordium. Detection algorithms have been designed for some of these errors.28,32 The best algorithms have a sensitivity for detection of electrode reversal of about 80% with a specificity of almost 100%. Variant limb electrode placement For certain applications, it is necessary to move the limb electrodes to the torso to reduce motion artifacts; it may also shorten recording time in emergencies. In 1966, Mason et al33 placed the arm electrodes in the infraclavicular fossae and the left leg electrode on the anterior axillary line, midway between the rib margin and the iliac spine. This Mason-Likar lead system became widely used for exercise stress testing. However, often a baseline (resting) ECG is taken before the exercise test using the same lead positions. Mason et al,33 and later on also Diamond et al,34 concluded that ECGs recorded with this system can be compared with standard ECGs. Others, however, reported profound amplitude and waveform changes.35-40 Therefore, the Mason-Likar ECG appears not to be “essentially identical” with the standard 12-lead ECG, as was originally claimed by Mason et al. Several transformation matrices have been proposed to reconstruct the standard limb leads from the Mason-Likar leads.41,42 In our opinion, like should be compared with like, that is, standard ECG with standard ECG and not with a simulacrum. The same holds for ECGs derived from other exercise adaptations such as those of Edenbrandt et al43 and Takuma et al.44 Recording procedure and equipment Electrode size Ideally, an ECG lead records the signal between 2 points on the body surface. In practice, electrodes cover a fair-sized area and record an averaged potential over that area. Berson et al45 studied the effect of 2 different electrode sizes (1 cm2
192
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196 2
and 7.5 cm ) in 20 subjects. Differences in R or S amplitudes ranged from −350 μV to +220 μV, with 8 subjects having a difference larger than 100 μV. Larger electrode size resulted in systematically smaller amplitudes in all leads. It appears that only a relatively small area close to the heart has potential gradients high enough for electrode size to be of importance. When conductive electrode paste is richly applied, potentials may be averaged over the entire area of application (gel short-circuiting).1 A disadvantage of smaller electrodes is their higher electrode impedance, which might pose a problem for equipment with low input impedance. Electrode type Zywietz46 gives a review of various electrode types. Self-adhesive silver-silverchloride electrodes are widely used. Suction electrodes are another frequently used type.47 The electrical characteristics (impedance and offset potential) of various types of electrodes differ19,48 but as long as they meet AHA standards,49 the effect on the ECG will be negligible. Skin preparation and electrode gel The ratio of skin-electrode contact impedance to recorder input impedance is important for the quality of the recording. Zywietz46 covers this aspect as well in his review. When the skin is unprepared, the skin-electrode impedance may vary between 50 and 200 kΩ at 60 Hz.50 A high-input impedance of the recording device can minimize the effect of these impedance variations.51 The 1975 recommendations of the AHA52 require 5 MΩ for frequencies up to at least 60 Hz; modern ECG recorders have an input impedance in the order of 10 MΩ. Skin-electrode impedance can be reduced by skin abrasion53,54 and the use of conductive electrode gel or spray. Abrasion also reduces potential variations when the electrode is pressed on the skin55 or moved relative to it.56 Because electrode impedance decreases with time,57,58 skin abrasion is not very useful for long-term recordings. For this application, at most light abrasion is recommended to diminish the chance of skin irritation and infections.59,60 Salted water or household substances such as hand cream, toothpaste, mayonnaise, mustard, and others seem to work as well as gel.61,62 Probably any water-soluble lubricant containing some free electrolyte and of a consistency, able to maintain a thin aqueous film between skin and electrode, would be satisfactory. Bandwidth and sampling rate A too low frequency response of the recording system can affect the ECG by reduction of amplitudes, as was already observed by Einthoven.63 Wave durations are less vulnerable. Berson et al64 demonstrated amplitude errors greater than 50 μV in more than 10% of adult ECGs recorded at 100 Hz bandwidth. For the pediatric ECG, larger bandwidths are necessary.65-67 To keep amplitude errors smaller than 25 μV in 95% of the ECGs, Rijnbeek et al67 showed that a bandwidth up to 250 Hz is necessary. The 2007 AHA
recommendations, therefore, specify a high-frequency cutoff of 150 Hz for adults and of 250 Hz for children.6 The theoretical minimum digital sampling rate should be twice the high-frequency cutoff. On the safe side of this, the AHA recommends sampling rates 2 to 3 times the minimum cutoff frequency.68 A too low sampling frequency can jeopardize computer-generated measurements and interpretations.69 Nevertheless, long-term storage is often done after lowering the sampling frequency to 250 Hz or by “lossy” compression (see Zywietz et al70 for a review on ECG compression). The loss of information incurred in this way hampers the use of these ECGs in serial comparison. Furthermore, ECG technicians may be in the bad habit of selecting low-pass filter settings to produce smooth signals and may be encouraged in this habit by doctors who are insufficiently aware of the loss of information involved. The computer analysis of the signals, however, usually precedes this cosmetic filtering and is unaffected by it. Noise Different noise sources may disturb the ECG.46 Myographic noise stems from electrical muscular activity and is caused by muscular tension and tremor. Electromagnetic fields generated by alternating current wires cause interference at 50 or 60 Hz or harmonics of these frequencies. Patient movement may result in gradual baseline wander at frequencies not exceeding 1 Hz or abrupt baseline jumps because of loss of electrode contact. Capacitative variations because of swinging cables compound these effects. Because of the discrete nature of diagnostic criteria used in computer programs, these disturbances can swing diagnostic decisions.71,72 For noise suppression, dedicated filters have been developed.73-75 Creating an average or a median complex out of the individual beats is also a powerful means to reduce noise. Notwithstanding this, noise may still impair ECG measurement and contribute to intraindividual variability.76,77 Biological sources of ECG variability The ECG variability of biological origin may be interindividual and be caused by factors such as race,26,78-80 sex,81-84 and chest configuration,85,86 or it is intraindividual and attributable to constitutional sources, such as age and weight, and to functional and physiologic sources, such as pregnancy, posture, and respiration. In many of these circumstances, the position of the heart plays a role. This touches on the physical basis of electrocardiography. Implicit in the Einthoven-Wilson model of electrocardiography is the assumption that the electric source is a dipole. The dipole can change in direction and strength but is “stationary,” that is, fixed in position. Furthermore, the body is supposed to be a homogeneous conductor. If the ECG shows variations because of translational displacements or changes in the conducting medium, this is in violation of the dipole model. Numerous studies were devoted in the third quarter of the previous century to the problem of dipolarity.87,88 Recently, a model study by
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196 89
Huiskamp et al showed that translational heart displacements as small as 0.5 cm can produce changes in QRS amplitude in the order of the SD for normal people. Likewise, according to Hoekema et al,90 variations in heart position, as measured by magnetic resonance imaging, can have considerable effects on QRS amplitudes. Dougherty91 reports that anatomical rotation in the frontal plane seen on the x-ray is responsible for a shift in QRS axis 3 times larger than expected from the same rotation of a single dipole. Positional changes of the heart can be brought about by many of the biological sources. A further complicating factor is the Brody effect, the short-circuiting of the electrical signals by the intracavitary blood mass.92 It may come into play under conditions of blood or fluid loss or accumulation. The direction of the effect, however, is entirely unpredictable because of counteracting and complicating factors.93-99 Constitutional sources Simonson1 in his classic book of 1961 already pointed out that the constitutional variables age, weight, and heart position are interdependent. Aging is mostly accompanied by weight change that in turn affects heart position. The following discussion will, nevertheless, try to differentiate somehow between these sources. It may be added here that at higher age, there is a natural and sometimes considerable loss of body height, caused by loss of elastic substance in the intervertebral discs. This causes compression of the abdominal cavity with upward displacement of the diaphragm. No systematic study of this factor has been made to our knowledge. Additional complications are pathologic alterations setting in with age, from which nonpathologic age-related variability has to be distinguished.100 Age In the age group of 0 to 18 years normal ECG limits are heavily dependent on age.101-103 In older age groups, the ECG changes are smaller and have a larger time scale. Simonson104 reviewed the general age trends among adults. There is a general decrease of amplitudes (eg, QRS spatial magnitude diminishes with approximately 8% per decade), a leftward shift of the frontal plane axis (approximately 10° per decade), and an anterior rotation of the horizontal plane axis. Interval durations increase for PR and QT. Most trends flatten out after the age of 50. These data were collected mainly in cross-sectional studies but were confirmed in the scarcely available longitudinal studies.105,106 In the latter, a decrease in QRS duration was also found. Other factors such as race, sex, and obesity interact with the age effect in a sometimes erratic fashion.79,104,107 For example, the R amplitude in V5 decreases with age for normal-weight men but increases for normal-weight women. 79 This leftprecordial R-wave amplitude increase in women was also found in 2 studies on normal Chinese.80,108 Furthermore, the decrease of the R amplitude in V5 in Hispanic men is about 40% of that in white and black men.79 Obesity tends to reinforce the age trends as pointed out by Simonson,1 but the interactions are complex. For example, the QRS axis in obese subjects aged 20 to 29 years is more inferior than those
193
in nonobese subjects aged 50 to 59 years, although the general trend in obesity is a superior shift. If these effects are compared with the interindividual variability (in the form of mean and SD of the measurements in white normal males aged 40-49 years), it appears that for 3 decades, the intraindividual change in QRS spatial amplitude (−25%) has become comparable to the interindividual SD.2 This does not hold for all measurements— PR duration, for instance, increases with about 3 milliseconds per decade, but the SD amounts to 22 milliseconds.2 With respect to cardiac rhythm, there is a marked rise in the incidence of supraventricular premature complexes (SVPC) and ventricular premature complexes (VPC), ranging from less than 1% in the age group 20 to 29 years to 5% (SVPC) and 10% (VPC) in those older than 70 years.109 Weight Obesity is a long-term factor in variability, associated with a superior and anterior deviation of the QRS axis and a decrease in precordial voltages.1,79,110-112 Not unreasonably, left axis deviation has been ascribed to horizontal heart position and voltage decrease to increased distance from heart to skin. Selzer et al113 already remarked that most false positive LVH diagnoses were in emaciated people. In other publications, a negative correlation was found between precordial amplitudes and weight or a weight-related index (eg, ponderal index).114,115 Furthermore, decreasing the distance between heart and precordial electrodes, for example, brought about by an α1-adrenergic substance or by turning the body on the left side, resulted in larger R waves. 116 Contrarily, Brohet and Tuna 117 observed a decrease in Q and S, as well as in maximum P and QRS vector amplitudes 1 year after jejunoileal bypass surgery in 37 markedly obese patients. Although statistically significant, the effect was too small to have clinical meaning; −70 μV for the maximum spatial QRS vector, for which the SD in normal people is 440 μV.2 Proof of the effect of tissue thickness between heart and electrode might have come from a study by LaMonte et al118 on the consequences of mastectomy on ECG amplitudes in 39 subjects. Paradoxically, peak-to-peak QRS amplitude showed significant increases after right mastectomy in V1 but also in the leftprecordial leads V5 and V6 (124-303 μV), whereas after left mastectomy, QRS amplitudes in V3R and V1 to V4 increased by 66 to 284 μV. There were no statistical differences between the changes in P or T amplitudes. Clearly, reduced distance between underlying myocardium and electrode fails as an explanation for these findings. The obvious consequence of decreased precordial voltages is a loss of sensitivity for LVH.119-122 On the other hand, obese people may very well have LVH because of various causes.123-126 This has led to suggestions for weight-adjusted LVH criteria.121,127-129 Functional and physiologic sources Pregnancy The cardiovascular changes during pregnancy include sodium and water retention, increased blood volume, and a
194
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196 130
higher resting cardiac output, concomitant with a rise in heart rate.131,132 Carruth et al132 studied the ECG in 102 normotensive women in the first, second, and third trimester of pregnancy, 1 to 3 days after delivery and 6 to 8 weeks postpartum. The mean resting heart rate increased from 77 in the first trimester to 87 in the third, dropping off to 66 postpartum. One might expect that the growing uterus brings the heart in a more horizontal position causing a leftward shift of the frontal QRS axis. However, the behavior of QRS and T axis proved to be unpredictable, as also reported by others.133,134 An early study even signaled opposite trends—a leftward QRS axis in early pregnancy and a rightward before delivery.135 The capricious behavior of the axis in this situation also comes to light when its directions before and after delivery are compared—there is no difference.134 Changes in PR, QT, and QTc durations were not clinically important, as was reported before.136
prominent are bradycardia, prolonged intraventricular conduction, and higher voltages in QRS, T and U, and ST elevation.149-154 Some of these differences are less pronounced in women155 or even reversed in the form of sinus tachycardia. 156 In addition to these findings, Bjørnstad et al151,157-161 mention PQ prolongation associated with low heart rates and incomplete right bundlebranch block. Some measurements (heart rate, conduction times, and less significantly, indices of right and left ventricular hypertrophy) seem to reflect the level of fitness.157,159,161 The type of training also plays a role. Endurance athletes have significantly lower heart rates than, for example, gymnasts or sprinters.159 It is likely that this interindividual variability is transferable to intraindividual variability because a number of ECG findings appeared reversible in athletes who were followed during detraining.150
Respiration
A meal can cause considerable ECG changes.1,162,163 In normal people, there may be an increase of heart rate, a decrease of T-wave amplitude and QT interval, and small leftward shifts of the QRS and T axes.162 In cardiac patients, ST depressions and/or T-wave inversions can be observed.163 Autonomic stimulation is a possible explanation.162 Dota et al164 studied interindividual and intraindividual variability between ECGs recorded right after breakfast and just before lunch. QT and QTc intervals were shortened and T amplitudes reduced between morning and noon. Harris et al165 performed a similar study and also found QT shortening but no significant change in QTc. The temperature of the ingested mass might play a role. Fluids can reach the stomach much faster and in larger portions than food and can warm up or cool down the overlying inferior cardiac wall before temperature equalization will take place. Repolarization changes have been provoked by the drinking of cold fluids.166,167
The observation that respiration induces changes in the ECG goes far back in the history of electrocardiography.137 The changes concern amplitudes, axis directions, rhythm, and durations of intervals and waves.138-141 Because the respiratory cycle is shorter than the usual ECG record length, the variation is mostly seen as “beat-to-beat” variation. Intraindividual variation may occur if recordings at different moments are made under different respiratory conditions (frequency and depth of respiration, breath holding at inspiration or expiration). In a study in 194 patients,139 changes in diagnostic interpretation because of deep inspiration occurred in about 17% of the cases, as compared to 12% because of an electrode position shift of one intercostal space and 12% because of posture. The obvious method to reduce respiration-related ECG variability is breath holding in a specified phase of respiration.142 The explanation of these changes of amplitude and axis is sought in the positional changes of the heart,143,144 which, together with the inflation or deflation of the lung, also change the distance between precordial electrodes and heart,138 in addition to altering the electrical conductivity of the lungs.140,144 “Respiratory arrhythmia” is typically prevalent in the younger age groups and is characterized by periodic acceleration (during inspiration) and deceleration (during expiration). It is the result of various autonomic reflexes (see, eg, Hirsch et al145 ) and will be suppressed by breath holding and also mostly disappear when heart rate is increased. If the slow phase is pronounced, interference dissociation by the atrioventricular node is not unusual. Intraindividual variation may then arise. Absence of this type of arrhythmia where it might typically be expected is considered abnormal and has been observed in alcoholic individuals.146 The phenomenon has been used to estimate the respiration frequency147 or the tidal lung volume.148 Physical training The changes because of physical training have mostly been studied as differences between normal individuals and athletes, that is, as interindividual variability. Most
Food and fluid intake
Posture According to Simonson,1 amplitudes are not significantly different between the supine and the sitting positions. Later studies on the vectorcardiogram (VCG), however, did reveal that going from the supine to a sitting or standing position went along with an increase of QRS spatial amplitude and of R amplitude in lead Z and a decrease in Q amplitude in lead Y, whereas azimuth and elevation angles were not significantly altered.37,139,168 Changes from supine to 60° headup resulted in similar ECG changes.169 Posture-induced changes also occur without a discernable trend. According to Riekkinen et al,139 changes in diagnostic interpretation because of postural change occurred in 12% of 144 cases, and pathologic patterns may appear and disappear.37 Wave durations are not affected by posture.168 Differences by lying on the back or one side or the other have also been studied.116,170-174 Such changes may not be negligible and, when not taken into account, may create confusion in a coronary care unit. For the normal practice of standard ECG recording they are not of interest.
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196
Neurophysiologic mechanisms Heart rate and electrophysiologic parameters are under autonomic control.175 Frederiks et al176 applied 2 stressors to 13 healthy young individuals in a sitting position, viz handgrip (isometric stress) and leg-lowering (gravitational stress). At matching heart rates, the ECGs of the same subject showed significant differences in depolarization and repolarization parameters. For example, the QT interval was 435 ± 21 milliseconds for handgrip vs 418 ± 15 milliseconds for leg-lowering (P b .01). Emotional stress has a partly different effect on the ECG than physical stress.177,178 The ECGs of subjects under emotional stress show increased heart rates and minor ST changes.179 Doi et al180 found a significant decrease of R-wave amplitude in lead V5 during mental arithmetic stress. This they attributed to enhanced myocardial contractility and increased cardiac output, with ensuing reduction of cardiac volume and increase of the distance between heart and chest wall. Methods to induce mental stress are, however, difficult to standardize and quantify.181 Evaluation of the specific emotional response is important, however, as comes forward from a study involving QT duration from Huang et al.182 Temperature Having a sauna leads to shorter ST-T segments and decreased T amplitudes.183 These repolarization changes were attributed to the temperature dependence of the action potential duration. In addition, the average heart rate was raised considerably from 66 to 109 beats per minute after 25 minutes of heat exposure, dropping after a cool shower. This rate response cannot be explained by temperature alone, but an autonomic effect has to come into play as well.183 The effects of swallowing cold fluids, mainly affecting the repolarization phase, have been mentioned above. Effect of time interval between ECGs Intraindividual variability between ECGs recorded minutes apart, or a few hours at most, is called minute-to-minute variability. Sometimes the electrodes remain in place, leaving respiration and noise as the main sources of variability; sometimes they are reapplied on marked locations. Other sources of variability may come into play such as changes of posture and of breathing depth, meals, and mental disposition. De Bruyne et al184 recorded 2 ECGs 30 minutes apart using marked electrode positions in 97 elderly subjects. They report a coefficient of variation of 4.4% and 6.5% for PR and QRS duration, respectively, and of 12.2% for the Sokolow-Lyon voltage. Others report smaller mean minute-to-minute variability of the SokolowLyon voltage, with coefficients of variation of 3.1%185 and 2.6%,186 but in these studies, the electrodes remained attached between the recordings. When the time interval between recordings spans 1 or more days, we speak of day-to-day variability. Electrodes seldom remain in place, but locations may be marked. Even then, however, larger changes than those caused by minuteto-minute variability are found. Reported coefficients of
195
variability for day-to-day changes of the Sokolow-Lyon voltage were 18.5%,186 14.9%,184 and 10.0%.185 If electrodes remain in place, minute-to-minute variability is about 2.5 times smaller than day-to-day variability185,186; if electrodes are reapplied to marked positions, it is only 1.2 times smaller.184 No position marking at all enlarges variability even more, as was shown by Willems et al.15 This demonstrates the importance of correct electrode placement. The results confirm findings by Larkin et al.187 The ECG recordings taken 1 or more years apart presumably show the largest intraindividual variability. Sources such as age, weight, or physical fitness then come into play, in addition to sources already mentioned.105,188 The magnitude of year-to-year variability is considerable. According to Michaelis et al,188 the 95% range of serial ECG changes for a period of 1 to 3 years for adult males of age 36 to 45 years is 25 milliseconds for QRS duration and 700 and 1550 μV for R amplitudes in aVF and V5, respectively. Corresponding 95% ranges for the interindividual variability2 are 39.6 milliseconds for QRS duration and 1468 and 1972 μV for the R amplitudes. The magnitude of year-to-year variability then comes down to 50% to 75% of the interindividual variability. Contrarily, McManus et al189 following up 243 elderly men found that year-to-year variation was virtually identical to day-to-day variation in almost all VCG measurements. Discussion Intraindividual variability in the resting ECG can have many sources, but only a few play an important role in routine diagnosis provided that adequate provisions are taken. The most bothersome are the ones that vary in a random fashion such as chest electrode malpositioning or respiratory modulation. Noise is a perpetual evil but has the advantage that it is clearly visible on the ECG. Thorough instruction and rigorously sticking to a standard protocol can prevent most problems. Variability in the ECG manifests itself on 3 different levels, viz the signals, the measurements derived from those signals, and the diagnostic interpretation. Human electrocardiographers have less trouble discarding erratic signal fluctuations than computer programs.71 Computerized ECG analysis eliminates human measuring bias190,191 but is sensitive to small changes in the ECG signal that may cause the crossing of a diagnostic threshold.69,76,77 One method to solve this problem is by smoothing the discrete diagnostic criteria.192 An alternative method was followed by Kors et al193 : the combination of the separate interpretations of individual beats into a single diagnosis performed better than the interpretation of one representative beat. Others have tried different ways of correcting for beat-tobeat variability.194,195 There might even be diagnostic information in beat-to-beat variability. Beat-to-beat morphological changes may precede ventricular arrhythmias196 but do not appear to have discriminative power for coronary artery disease.197 Intraindividual ECG variability is in principle smaller than interindividual variability 16 but may be important in
196
B.J.A. Schijvenaars et al. / Journal of Electrocardiology 41 (2008) 190–196
individual cases. The ECG diagnosis could be improved by taking causes of variability into consideration such as age and obesity (sex may be added as a source of interindividual variability). The interpretation process can be made less sensitive to variability by using less discrete thresholds for diagnostic criteria (eg, “fuzzy logic”) or by analyzing beats separately, possibly resulting in alternative diagnostic statements to consider. This last method could be taken one step further by modeling sources of intraindividual variability and simulating their effect on the ECG recorded.198,199 Interpretation of these simulated ECGs may provide insight into whether the diagnostic interpretation of the ECG at hand is sensitive to intraindividual variability. Appendix A. Supplementary data Because of space limitations, only part of the references are listed at the end of this article; the full reference list is available at doi: 10.1016/j.jelectrocard.2008.01.012. References 1. Simonson E. Differentiation between normal and abnormal in electrocardiography. St Louis: Mosby; 1961. 2. Macfarlane PW, Lawrie TDV. Normal limits. In: Macfarlane PW, Lawrie TDV, editors. Comprehensive electrocardiology, vol. 3. New York: Pergamon Press; 1989. p. 1442. 6. Kligfield P, Gettes LS, Bailey JJ, et al. Recommendations for the standardization and interpretation of the electrocardiogram: part I: the electrocardiogram and its technology. J Am Coll Cardiol 2007;49:1109. 11. Bupp JE, Dinger M, Lawrence C, Wingate S. Placement of cardiac electrodes: written, simulated, and actual accuracy. Am J Crit Care 1997;6:457. 15. Willems JL, Poblete PF, Pipberger HV. Day-to-day variation of the normal orthogonal electrocardiogram and vectorcardiogram. Circulation 1972;45:1057. 18. Hill NE, Goodman JS. Importance of accurate placement of precordial leads in the 12-lead electrocardiogram. Heart Lung 1987;16:561. 20. Schijvenaars RJA, Kors JA, van Herpen G, et al. Effect of electrode positioning on ECG interpretation by computer. J Electrocardiol 1997; 30:247. 26. Macfarlane PW, Lawrie TDV. The normal electrocardiogram and vectorcardiogram. In: Macfarlane PW, Lawrie TDV, editors. Comprehensive electrocardiology, vol. 1. New York: Pergamon Press; 1989. p. 407. 27. Hedén B, Ohlsson M, Edenbrandt L, Rittner R, Pahlm O, Peterson C. Artificial neural networks for recognition of electrocardiographic lead reversal. Am J Cardiol 1995;75:929. 32. Kors JA, van Herpen G. Accurate automatic detection of electrode interchange in the electrocardiogram. Am J Cardiol 2001;88:396. 33. Mason RE, Likar I. A new system of multiple-lead exercise electrocardiography. Am Heart J 1966;71:196.
45. Berson AS, Pipberger HV. Effects of electrode area on electrocardiographic voltages. J Electrocardiol 1978;11:147. 46. Zywietz C. Technical aspects of electrocardiogram recording. In: Macfarlane PW, Lawrie TDV, editors. Comprehensive electrocardiography, vol 2. New York: Pergamon Press; 1989. p. 353. 67. Rijnbeek PR, Kors JA, Witsenburg M. Minimum bandwidth requirements for recording of pediatric electrocardiograms. Circulation 2001;104:3087. 71. Hurst JW, Treasure CB, Sathavorn CS. Computer errors in electrocardiography. Clin Cardiol 1996;19:580. 77. Zywietz C, Willems JL, Arnaud P, van Bemmel JH, Degani R, Macfarlane PW. Stability of computer ECG amplitude measurements in the presence of noise. The CSE Working Party. Comput Biomed Res 1990;23:10. 89. Huiskamp GJ, van Oosterom A. Heart position and orientation in forward and inverse electrocardiography. Med Biol Eng Comput 1992; 30:613. 104. Simonson E. The effect of age on the electrocardiogram. Am J Cardiol 1972;29:64. 105. Harlan WRJ, Graybiel A, Mitchell RE, Oberman A, Osborne RK. Serial electrocardiograms: their reliability and prognostic validity during a 24-yr. period. J Chronic Dis 1967;20:853. 106. Bachman S, Sparrow D, Smith LK. Effect of aging on the electrocardiogram. Am J Cardiol 1981;48:513. 132. Carruth JE, Mivis SB, Brogan DR, Wenger NK. The electrocardiogram in normal pregnancy. Am Heart J 1981;102:1075. 138. Dougherty JD. The relation of respiratory changes in the horizontal QRS and T-wave axes to movement of the thoracic electrodes. J Electrocardiol 1970;3:77. 150. Oakley DG, Oakley CM. Significance of abnormal electrocardiograms in highly trained athletes. Am J Cardiol 1982;50:985. 162. Simonson E, Alexander H, Henschel A, Keys A. The effect of meals on the electrocardiogram in normal subjects. Am Heart J 1946;32:202. 165. Harris RI, Steare SE. A meta-analysis of ECG data from healthy male volunteers: diurnal and intra-subject variability, and implications for planning ECG assessments and statistical analysis in clinical pharmacology studies. Eur J Clin Pharmacol 2006;62:893. 169. Bergman KS, Stevenson WG, Tillisch JH, Stevenson LW. Effect of body position on the diagnostic accuracy of the electrocardiogram. Am Heart J 1989;117:204. 175. Swenne CA. Neurocardiological basis for intraindividual ECG variability. J Electrocardiol 2002;35(Suppl):239. 184. De Bruyne MC, Kors JA, Visentin S, et al. Reproducibility of computerized ECG measurements and coding in a nonhospitalized elderly population. J Electrocardiol 1998;31:189. 185. Van Den Hoogen JP, Mol WH, Kowsoleea A, Van Ree JW, Thien T, Van Weel C. Reproducibility of electrocardiographic criteria for left ventricular hypertrophy in hypertensive patients in general practice. Eur Heart J 1992;13:1606. 186. Farb A, Devereux RB, Kligfield P. Day-to-day variability of voltage measurements used in electrocardiographic criteria for left ventricular hypertrophy. J Am Coll Cardiol 1990;15:618. 188. Michaelis J, Lippold R, Gluck E, Schindler W, Scheidt E. Automated serial ECG-analysis within an epidemiological study. J Electrocardiol 1987;20(Suppl):34.
