Journal of Electrocardiology 40 (2007) 228 – 234.

e8
[Link]

Electrocardiographic reference ranges derived from

79,743 ambulatory subjects
Jay W. Mason, MD,a,b,4 Douglas J. Ramseth, PhD,a Dennis O. Chanter, DPhil,a
Thomas E. Moon, PhD,a Daniel B. Goodman, MD,a Boaz Mendzelevski, MDa
a
Covance Cardiac Safety Services, Inc, Reno, NV, USA
b
Cardiology Division, University of Utah, Salt Lake City, UT, USA
Received 11 April 2006; accepted 27 September 2006

Abstract Background: Reference ranges for electrocardiogram (ECG) intervals, heart rate, and QRS axis
in general use by medical personnel and ECG readers are unrepresentative of true age- and sex-
related values in large populations and are not based on modern electrocardiographic and ECG
reading technology.
Methods and Results: The results of ECG interpretation by cardiologists using digital technology
for viewing and interpreting ECGs were compiled from single, baseline ECGs of 79,743 individuals
included in pharmaceutical company–sponsored clinical trials. Women comprised 48% of the total
population. Ages ranged from 3 months to 99 years, and the bulk of the population (56%) was aged
40 to 70 years. Striking differences in numerical ECG values based on age and sex were observed.
A subgroup of 46,129 individuals with a very low probability of cardiovascular disease was
identified. The following were the reference ranges for this subgroup, determined using the 2nd and
98th percentiles: heart rate, 48 to 98 beats/min; PR interval, 113 to 212 milliseconds; QRS interval,
69 to 109 milliseconds; frontal plane QRS axis, 408 to 918; QT interval, 325 to 452 milliseconds;
QTc-Bazett, 361 to 457 milliseconds; and QTc-Fridericia, 359 to 445 milliseconds. There were
marked age- and sex-related variations in the reference ranges of this subgroup, and they differ
substantially from previously reported norms. Small differences were observed in ECG values
obtained using our digital methods as compared with readings done using paper tracings and values
computed by 2 commercial computer algorithms.
Conclusions: We observed large differences in electrocardiographic heart rate, interval, and axis
reference ranges in this study compared with those reported previously and with reference ranges in
general use. We also observed a large influence of age and sex upon normal values. Very large
cohorts are required to fully assess age- and sex-related variation of reference ranges.
Electrocardiographic reference ranges should be modernized.
D 2007 Elsevier Inc. All rights reserved.
Keywords: Electrocardiogram; Normal value; Reference range

Introduction defined decades ago. Populations throughout the world

have different age and ethnic compositions from before,
Commonly accepted reference ranges for the electrocar-
and the nature, incidence, extent, and medical treatment of
diogram (ECG) have been in use, with little change, for
cardiac and noncardiac diseases that influence these
many years. A reexamination of established reference
intervals has changed.
values is needed. Electrocardiograms are now recorded
The purposes of this study are to examine the range of
and read differently from when the time-honored upper and
numerical ECG values encountered in an ambulatory
lower limits of normal for heart rate, PR interval, QRS
population, to define new reference ranges for ECG
interval, QT interval, and frontal plane axis were being
intervals based upon digitally acquired ECGs interpreted
by experienced cardiologists using high-resolution viewing
4 Corresponding author. Covance Cardiac Safety Services, Inc, Reno,
techniques, and to examine a sufficiently large and diverse
NV 89521, USA. Tel.: +1 775 858 1000x1014; fax: +1 775 858 1038. sample to allow reliable distribution estimates for sex, age,
E-mail address: [Link]@[Link] and the two combined.
0022-0736/$ – see front matter D 2007 Elsevier Inc. All rights reserved.
doi:10.1016/[Link].2006.09.003
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 229

Methods that the end of the T wave be determined, if possible, using an

end of voltage change method (99.7% of ECGs). In this
Selection of subjects
method, the end of T is the time at which the increase or
The 79,743 subjects included in this analysis are normal decrease in voltage of the final limb of the T wave approaches
human volunteers and patients who were screened for zero. In the tangent method (0.3% of ECGs), a tangent is
enrollment in pharmaceutical company–sponsored clinical drawn at the point of the largest positive or negative dV/dt on
studies for which Covance Cardiac Safety Services, Inc the final limb of the T wave, and its crossing point on the
(CCSSI), served as the core ECG laboratory. Only one isoelectric line is taken as the end of T. The tangent method
ECG, done at screening for study enrollment or before was also used in place of the voltage change method if merger
institution of therapy, is included for each subject. All with a U wave obscured the end of the T wave.
screened subjects in our database whose ECGs were read in Annotated ECGs were then presented to a cardiologist in
the years 2003 (n = 14,677), 2004 (n = 29,944), and Digitography, a CCSSI-proprietary viewing station that
through November 3, 2005 (n = 35,122) were included displays the ECG on a high-resolution, 20-in, flat-screen
except for those whose ECGs were not reviewed on-screen. monitor. The waveforms and annotations were viewed at a
The subjects were drawn from 156 clinical studies variety of resolutions. At the highest resolution, 1 millisec-
sponsored by 53 pharmaceutical companies. ond is represented by 1 pixel. Cardiologists examined and
adjusted or confirmed the location of the annotation points
Electrocardiography on the ECG waveforms at high resolution to assure the
Electrocardiograms were acquired using 2 types of accuracy of the measurements. Intervals were measured
electrocardiographs programmed and maintained by CCSSI. using the standardized methods noted above. After measur-
The first, the MTX-2, is a proprietary device developed by ing intervals, rate, and axis, the cardiologists then used a
CCSSI, which holds a 510(k) approval by the US Food and uniform code set and uniform criteria to assign diagnostic
Drug Administration (FDA) and a Conformité Européenne statements to the tracing. Because measurement of intervals
b(CE)Q for use in the European Union. This device records in Digitography might produce different results compared to
data at 500 samples per second with a resolution of 5 lV. reading of paper ECGs and to computer algorithms, we
The second device, the MAC 1200, is a commercially performed analyses to quantify these potential differences
available electrocardiograph manufactured by GE Medical and included them with a methodological description in the
Systems. The data acquisition rate of this device is also 500 electronic supplement to this article.
samples per second, at 5 lV resolution. There were no Reference range subset
statistically or clinically significant differences in interval,
We identified a subset of subjects expected to be free of
rate, and axis readings between these 2 acquisition devices.
disease-associated ECG effects. We used this subset to
Interval, heart rate, and axis measurement and diagnostic determine reference ranges for heart rate, ECG intervals,
interpretation and axis. To derive this subset, we excluded subjects from
the full cohort of 79,743 for the following reasons: evidence
Electrocardiograms received from the MTX-2 and the
of a cardiac pacemaker on ECG (n = 256), screen failure for
MAC1200 are stored electronically at CCSSI. The wave-
any reason (n = 756 of the remaining), enrollment in a trial
forms are then submitted to a computerized algorithm or to a of cardiovascular (n = 7066 of the remaining), diabetes
human technician for initial placement of markers on (n = 9190 of the remaining), or hyperlipidemia therapy
multiple beats at the onset of the P wave, the onset of the Q (n = 6571 of the remaining). The disease target of the
wave (or R wave in absence of a Q wave), the J point, the therapy under study identified 41 disease states in the
offset of the T wave, and the peak of the dominant R or remaining patients. From this group, an additional 1320 sub-
S wave. Eighty-five percent of the ECGs were annotated jects were excluded because of noncardiac disorders that
using CCSSI’s standard method in which the lead with the could affect electrocardiographic intervals: end-stage renal
longest QT was used for interval measurements. During a disease (n = 392), hypertension (n = 451), and acute stroke
period of rate stability and good signal quality, 3 consecutive (n = 477). From the remaining group of 54,584 subjects, an
beats in that lead were annotated, including the R or S peak of additional 8455 were removed because of one or more
the beat preceding the first measured beat. An average value specific diagnoses on ECG: nonsinus rhythm (4286); acute,
for RR, PR, QRS, and QT was calculated from these recent, or remote myocardial infarction (1050); ventricular
annotations. Methods using an alternate choice of leads were preexcitation (20); a diagnosis of right or left ventricular
used when specified by the pharmaceutical sponsor. In 14% hypertrophy with both voltage and associated abnormalities
of the ECGs, a single lead was selected from a priority list, (229); complete right bundle branch block (RBBB) (833),
most commonly lead II followed by lead V2 and then V5. In complete left bundle branch block (LBBB) (261), IVCD
1% of the tracings, intervals were measured using median (1761), and technical issues (35). This process left 46,129 to
beats calculated from each of the 12 leads, time-registered, form the reference range subset.
and displayed superimposed on one another in a single
Statistical analysis
viewing pane. Less than 1% of the ECGs were measured
using variations of the above methods, such as measurement Descriptive statistics, as provided in JMP release
of 5 beats rather than 3. The methodology for both the CCSSI 5.1.5 (SAS Institute, Inc, Cary, NC, 2004), were used for
standard and most of the sponsor-specified methods require this analysis.
230 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

All authors have read and agreed to the article as written. Table 1
This study qualifies for exempt status according to Section Demographic information
46.101(b)(4) of 45CFR56, bThe Common Rule,Q because N Mean SD Range
the analysis involves existing data recorded in such a way Total population 79,743
that the authors cannot identify the subjects. Subjects could Men 41,392
Women 38,351
not be asked to give informed consent, and this study did
Total with interval measurements 79,487
not undergo IRB review. Total with diagnostic coding 70,728
Previous studies used for comparison Age (y) 79,743 51.7 16.6 0-99
Height (in) 68,843 66.4 4.4 32-84
The normal limits listed in the textbook of Macfarlane Weight (lb) 68,995 179 45 24-500
Reference range subset 46,129
and Lawrie1 were derived from 1338 apparently normal
Men 21,567
government employees in Western Scotland before 1985. Women 24,562
The age distribution of this group was quite different from Total with interval measurements 46,129
that of our study, but data were provided in age and sex Total with diagnostic coding 46,129
groups similar to ours, allowing some direct comparisons. Age (y) 46,129 47.5 17.4 0-99
Height (in) 38,767 66.1 4.6 32-84
Lamb2 reported another study of moderate size in his
Weight (lb) 38,846 173 45 24-500
textbook. This analysis was based upon the study of Hiss
et al3 performed in 6014 male Air Force recruits, and because of the presence of atrial fibrillation and flutter and
Lamb’s unpublished data in 782 women. Lamb reported the other arrhythmias in some subjects, and frontal plane QRS
percent of subjects in ordinal data categories. axis was only measured in 79,122 subjects because of
Dmitrienko et al4 recently reported another large data set indeterminacy of axis in some. The average age of this
from Eli Lilly and Company. They analyzed baseline ECGs population, which ranges from infants to nonagenarians,
obtained with GE Medical Systems electrocardiographs in was 51.7 F 16.6 years, and average height and weight were
13,039 patients enrolled in Lilly-sponsored trials in 2000 66.4 F 4.4 in and 179 F 45 lb, respectively. Probability
and 2001. These ECGs were first computer-interpreted by distributions of weight, height, and age in this population
the GE-Marquette 12SL program. Paper copies of the were roughly Gaussian as shown in Appendix Fig. 1, panels
ECGs, with the computer-generated results, were then A-C, in the online edition of this article. Although the
overread by cardiologists. The population included only subjects of this study resided in 57 countries, most (70%)
819 subjects more than 65 years of age. Most subjects were were from the United States and Canada, and the next
enrolled in neuroscience trials, and those enrolled in largest group was from Europe (21%). Regions and
cardiovascular drug trials were excluded. The authors countries of origin of the study population are shown in
proposed normative values based upon different age ranges Appendix Table 1 (online).
(decades beginning at middecade) and different lower and Because this population includes subjects enrolled in all
upper percentiles (1st and 99th) from ours, and they reported clinical trial phases (phase 1, 8%; phase 2, 18 %; phase 3,
on subjects with normal and abnormal ECGs separately. 72%; and phase 4, 1%), most of the subjects had a known
The study of reference ECG values by Simonson5 was the clinical disorder. The most common diseases were neuro-
most statistically advanced treatment of the topic when it was psychiatric (27%), diabetes mellitus (12%), and cardiovas-
published in 1961. He used the percentile distribution to cular (9%).
differentiate between normal and abnormal, and set the upper Characteristics of the subjects included in the reference
and lower thresholds at 97.5% and 2.5%, respectively. His range subset are summarized on the bottom of Table 1. The
study involved 960 adults between 20 and 60 years of age, of mean age is a few years younger, and there is a greater
which 649 were men, and 311 were women. His subjects proportion of women than men as compared with the entire
were drawn from several groups of hospital, railroad, and group. Their weight, height, and age distributions were
insurance company employees. The subjects were screened approximately normal and similar to those of the whole
by history and physical examination, and Simonson indi- population (Appendix Fig. 1, panels D-F [online]).
cates that he excluded individuals with bmajor electrocar- The study population includes 990 subjects aged 0 to
diographic abnormalities, such as patterns of old myocardial 9 years. Within this group, there were 6 children less than
infarction and right or left bundle branch block.Q5 1 year (aged 3, 6, 8, 8, 8, and 11 months), 3 aged 1, 15 aged 2,
25 aged 3, 57 aged 4, 66 aged 5, 179 aged 6, 174 aged 7, 258
aged 8, and 207 aged 9 years. These numbers are insufficient
Results for estimation of age-specific norms in children, and the
summary values for ages 0 to 9 must be recognized as averages
Characteristics of the study population
derived from age groups with large group differences.
Table 1 displays the basic characteristics of our study
Electrocardiographic findings
population. The study includes a single screening or
baseline ECG from 79,743 subjects. Fifty-two percent were All subjects
men. Heart rate, QRS interval, and QT interval were Appendix Tables 3-6 (online) list ECG data for the entire
measured in the 79,487 individuals without cardiac pace- population with measurable values. Separate data for the
makers. PR was measured in fewer subjects (78,846) sexes, 10-year age cohorts, and sex/age cohorts are
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 231

provided. Mean, standard deviation of the mean, median, Table 3

98th and 2nd percentiles, 99th and 1st percentiles, and the QTcB, QTcF, and QT by age and sex, reference range subset
number of subjects in each group are displayed. There were Group 2% Median 98% N
many striking differences between ages within each sex and QTcB interval (ms)
between age/sex cohorts. For example, in women aged 20 to All 361 409 457 46,129
Male 356 401 449 21,567
29 years, the lower limit of QRS axis was 38, whereas it
Female 369 414 460 24,562
was 638 in women aged 90 to 99 years. The 96% range of M, 0-9 368 408 452 579
QRS in women aged 30 to 39 years was from 69 to 110 M, 10-19 352 403 448 776
milliseconds, whereas it was 65 to 170 milliseconds for men M, 20-29 347 390 436 2528
aged 90 to 99. In general, the observed ranges in older age M, 30-39 353 396 443 3411
M, 40-49 357 401 446 4316
groups were surprisingly broad.
M, 50-59 361 405 451 4460
M, 60-69 362 407 456 3275
Reference range subgroup M, 70-79 362 406 456 1718
Appendix Tables 7-10 (online) provide similar informa- M, 80-89 361 409 455 483
tion for the reference range subgroup, from which we M, 90-99 363 421 442 21
F, 0-9 365 411 461 384
propose derivation of reference ranges for the ECG. In
F, 10-19 371 408 457 569
Tables 2 and 3, only the age and sex groups with clinically F, 20-29 364 409 454 2469
F, 30-39 367 412 455 3954
Table 2 F, 40-49 371 414 460 6047
Heart rate, PR, QRS, and QRS axis by age and sex, reference range subset F, 50-59 370 416 463 4899
Group 2% Median 98% N F, 60-69 370 416 462 3323
F, 70-79 370 417 467 2072
Heart rate (beats/min) F, 80-89 370 417 467 801
All 48 68 98 46,129 F, 90-99 373 423 476 44
M 47 66 98 21,567
QTcF interval (ms)
F 49 68 98 24,562 All 359 400 445 46,129
0-9 60 84 120 963 Male 355 394 438 21,567
10-19 49 70 101 1345 Female 365 405 450 24,562
M, 20-29 44 63 91 2528
M, 0-9 346 388 428 579
F, 20-29 49 69 96 2469 M, 10-19 354 391 430 776
30-69 48 68 97 33,685 M, 20-29 351 387 426 2528
70-99 46 65 93 5139 M, 30-39 353 389 430 3411
PR interval (ms) M, 40-49 356 393 435 4316
All 113 154 212 46,119 M, 50-59 359 397 438 4460
M 115 157 218 21,561 M, 60-69 361 399 444 3275
F 112 151 205 24,558
M, 70-79 363 401 446 1718
0-9 92 127 167 963 M, 80-89 366 407 452 483
10-19 104 141 186 1345 M, 90-99 369 414 448 21
20-59 114 153 203 32,083 F, 0-9 347 387 428 384
M, 60-69 120 163 226 3273
F, 10-19 362 396 437 569
F, 60-69 116 156 213 3323 F, 20-29 362 400 440 2469
M, 70-79 123 168 250 1716 F, 30-39 364 403 441 3954
F, 70-79 116 160 221 2070 F, 40-49 367 405 447 6047
M, 80-99 121 177 282 502 F, 50-59 367 407 450 4899
F, 80-99 118 163 237 844 F, 60-69 367 407 452 3323
QRS interval (ms) F, 70-79 369 410 459 2072
All 69 91 109 46,129
F, 80-89 366 411 464 801
M 71 94 109 21,567 F, 90-99 370 416 454 44
F 68 88 107 24,562 QT interval (ms)
M, 0-9 61 79 101 579 All 325 384 452 46,129
F, 0-9 57 77 96 384
Male 322 380 449 21,567
M, 10-19 65 89 108 776 Female 327 387 455 24,562
F, 10-99 69 89 107 24,178 M, 0-9 286 347 408 579
M, 20-99 73 95 109 20,212 F, 0-9 277 344 399 384
QRS axis (8)
10-19 321 373 435 1345
All 40 37 91 45,944 20-29 328 381 447 4997
M 45 36 91 21,466 M, 30-59 324 378 443 12,187
F 35 38 90 24,478
F, 30-49 329 386 448 10,001
0-19 0 60 102 2293 F, 50-69 330 390 458 8222
20-29 10 60 95 4978 M, 60-69 325 384 455 3275
30-39 25 47 91 7337 M, 70-79 329 394 466 1718
40-49 31 40 90 10,327
F, N 70 335 397 471 2917
50-59 41 30 87 9316 M, N 80 342 406 467 504
60-69 46 20 87 6567
70-79 51 9 87 3779
80-99 60 1 76 1347
M indicates male; F, female.
232 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

significant differences from the other groups are shown. In Comparison to previous studies
these tables, median values, 98th and 2nd percentiles, and
Electrocardiogram interval reference values reported in
the number of subjects in each group are displayed. Those
this study are compared to those of Macfarlane and Lawrie1 in
groups without significant dissimilarities in the 2nd or 98th
Appendix Table 11 (online). A few differences between our
percentiles ( N3 beats/min for rate, N5 milliseconds for
interval values and theirs stand out. Our reference range
intervals, or N 58 for axis) are combined into single groups. subset (as well as the entire group of 79,743 subjects)
The data demonstrate a strong influence of age and sex upon demonstrated a small decrease in heart rate with increasing
heart rate, ECG intervals, and axis. Fig. 1 displays the age, whereas Macfarlane and Lawrie1 observed the oppo-
distributions of heart rate, PR, QRS, and QT intervals for all site. Like us, they did observe a consistently lower heart rate
subjects (panels A-D) and the reference range subgroup in men than in women. Both we and Macfarlane and
(panels E-H). The distributions are nearly normal, although Lawrie1 recorded a progressive increase in PR interval with
high kurtosis is present in the QRS distribution of the aging, and higher values in men than in women. Both
full population (panel C), and there are subtle variations observed a longer mean QRS duration in men than in
from normal in the other distributions. Thus, the use of women, but they reported a progressive decrease in mean
percentiles rather than a multiple of the standard QRS duration, whereas we observed stability of the mean
deviation was preferred for estimating the reference range with increasing age. The most striking difference was in the
for this population. Bazett-corrected QT interval. The group in the study of

A E
0.10 0.10
Probability

Probability
0.08 0.08
0.05 0.05
0.03 0.03

36 48 60 72 84 96 108 132 156 36 48 60 72 84 96 108 132 156

Heart Rate, beats per minute Heart Rate, beats per minute

B F
Probability

Probability
0.15 0.15

0.10 0.10

0.05 0.05

100 200 300 400 500 100 200 300 400

PR interval, msec PR interval, msec

C G

0.10
Probability

0.15
Probability

0.08
0.10
0.05
0.05
0.03

50 70 90 110 130 150 170 190 210 230 48 60 72 84 96 108 120 132 144 156
QRS interval, msec QRS interval, msec

D H
0.13
0.13
Probability

Probability

0.10 0.10
0.08 0.08
0.05 0.05
0.03 0.03

300 400 500 600 300 400 500 600

QT interval, msec QT interval, msec

Fig. 1. Distributions of HR, PR, QRS, and QT in all subjects (A-D) and the reference range subset (E-H). A normal curve (red) is fitted to each distribution. The
distributions show subtle variations from normal. The distributions in the 2 groups are very similar in most cases. However, in the case of the QRS interval, the
distribution is narrower in the reference range subset, and the upper tail is abruptly cut off because of the exclusion of ventricular conduction delay diagnoses in
this subset.
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 233
1
Macfarlane and Lawrie generally had considerably higher in current use for adults are 60 to 100 for resting heart
values. An increase of QTcB with aging was observed in rate, 140 to 210 milliseconds for PR interval, 70 to
both studies. Their group and we reported the 98th and 2nd 110 milliseconds for QRS duration, 308 to 908 for QRS
percentile values of our study populations. Using these as axis, and 460 milliseconds for the upper limit of QTc, as
the upper and lower limits of normal, considerably different stated in Marriott’s Practical Electrocardiography.6 Elec-
ranges of normal values are observed in the 2 studies. For trocardiographers typically apply these ranges across most
illustration, 4 age/sex groups are compared in Appendix or all adult age groups, usually without regard to sex. These
Table 11 (online). Macfarlane and Lawrie’s1 upper and ranges are not consistent with published age- and sex-
lower heart rate limits were generally 5 to 10 beats/min specific norms; rather, they have been handed down from
higher than ours, and their upper and lower QTcB limits teachers to students of electrocardiography over many
were strikingly higher than ours, ranging from 21 to decades as easily remembered, although imprecise, esti-
28 milliseconds and 18 to 23 milliseconds greater for the mates of normal. Unfortunately, their use results in
lower and upper limits, respectively. incorrect categorization as normal or abnormal of a
Appendix Table 12 (online) displays the age- and sex- substantial proportion of subjects.
related percentage distribution of QRS durations in the Review of the medians and 96% ranges of the entire
study of Lamb2 and ours. The differences are quite striking, group of 79,743 ambulatory subjects provides important
as durations in both male and female subjects in Lamb’s insight to the extent of the influences of age and sex upon
study show higher frequencies in the lower duration numerical ECG values. Variations among age and sex
categories than we observed in our subjects. cohorts are large enough to strongly impugn the common
As shown in Appendix Table 13 (online), subjects in the use of a single range of values to distinguish between
Lilly study4 that were grouped in the normal ECG category normal and abnormal among adults. A similar pattern of
had narrower ranges than those in the abnormal ECG variation, although not as large, between age and sex
category. The ranges of our reference range subset were cohorts exists in the reference range subgroup.
intermediate between these 2 Lilly groups, although we We propose that the 96% ranges of our subset of
used 2% and 98% cutoffs, whereas they used 1% and 99%. 46,219 subjects who had a very low probability of a
Unexpectedly, the upper PR range in men in the Lilly disease-related ECG abnormality be used to establish new
normal ECG group exceeded that of the Lilly abnormal reference ranges to replace existing reference ranges for
ECG group for both age cohorts, and considerably exceeded heart rate, ECG intervals, and axis values. Those ranges
our upper limit for men aged 36 to 45 years by 23 mill- appear in Appendix Tables 7-10 (online). The proposed
seconds. There were consistent differences between Lilly’s ranges are condensed in Tables 2 and 3. The greatest effect
QTcL limits and ours (recalculated using Lilly’s log linear of use of these new ranges will be more accurate diagnosis
formula). The lower limit values in all of their age/sex of bradycardia, which is overdiagnosed with current
categories were 3 to 14 millseconds higher than ours, and reference ranges; reduction in overdiagnosis of first-degree
their upper limit values were 10 to 11 millseconds higher. atrioventricular block; a more accurate recognition of the
The data of Simonson5 are compared with ours in effect of age upon QRS axis; more accurate diagnosis of QT
Appendix Table 14 (online). He stated that, although some prolongation according to age and sex; and a generally
differences in mean values for his age groups were better understanding of differences between men and
significant, the differences in upper and lower limits were women and among age and age/sex cohorts.
so minor that he combined all ages. In contrast, we observed These new reference ranges are applicable to ECGs
progressive changes with aging. The most striking differ- interpreted with the high-resolution digital display and
ence between our data and Simonson’s data is the absence of analysis systems that are in wide use today. All previously
longer QT values in women compared with men.5 reported ranges were derived either from human interpre-
tation of ECGs printed on paper or from computer
Comparison of Digitography to other reading methods
algorithms without on-screen, computer-assisted human
As expected, Digitography does result in a systematic review. Measurements done on paper by humans and those
difference in interval values from measurements performed done by computer algorithms do vary.7-9 Readers who make
directly on paper ECGs and measurements performed by direct measurement on paper ECGs recorded at standard
commercial electrocardiographic software. For example, gain and a paper speed of 25 mm/s or those who rely upon
QRS duration was shorter in Digitography by 5.64 milli- the software-derived values of Mortara or GE electrocardio-
seconds as compared with paper, 1.82 milliseconds as graphs can calculate the expected differences between those
compared with the GE 12 SL algorithm and 0.05 milli- methods and our Digitography results using the data in
seconds as compared with the Mortara Veritas algorithm. Appendix Tables 15-17 (online). Systematic differences
These differences are described in the online edition of this between manual, digital, and automated measurement
article in Appendix Tables 15-17. methods are well recognized10 and should be accounted
for with use of the reference ranges proposed in this study.
Our reference range subset includes both healthy
Discussion
volunteers and subjects with known noncardiac disease.
Derivation of the commonly used reference values for These 46,219 individuals were screened by trained medical
the ECG is surprisingly obscure. Typical reference ranges personnel at the investigative sites for evidence of cardiac
234 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

and other disease by history, physical examination, and and cultural differences in the populations, differences in
ECG before their enrollment. All subjects that failed recording methodology and equipment, differences in read-
screening for any reason were excluded from this subset. ing methodology, differences in health status among the
We consider use of this population to establish normative populations, and, especially, the smaller size and homogene-
ranges preferable to the use of small, highly screened, ity of the populations used in the previous studies.
homogeneous, apparently disease-free groups, as has been A single definitive set of normative values is needed so
common in the past.1-3,5,11 The latter approach eliminates that all ECGs can be interpreted consistently. For that
genetic diversity and results in samples that are too small to purpose, we believe our database has several advantages
permit reliable calculation of age- and sex-specific ranges. over older reported data sets1-3,5: (1) much larger sample
More importantly, highly culled, homogeneous groups are size; (2) greater genetic and geographic diversity of the
not reflective of the populations seen in either an ambula- population; (3) modern, well-maintained recording equip-
tory or in-patient setting in which physicians use the ECG to ment; (4) uniform recording methodology, including simul-
screen for cardiac disease. The restricted groups used in the taneous lead acquisition; (5) digital interval annotation using
past to establish normal values have generally resulted in a standardized measurement methodology at high resolu-
excessively narrow ranges, leading to a falsely high tion; (6) use of a single diagnostic code set and uniform
incidence of abnormal ECGs. Choice of the population diagnostic criteria; (7) review of all ECGs by cardiologists;
from which normative ranges are derived should be based (8) a limited subject recruitment period (3 years); (9) a more
upon the nature of the group the norms will be applied to contemporaneous subject population (years 2003-2005);
and upon the purpose for comparing findings to the norms. and (10) inclusion of disease states with low potential to
Practicing physicians use electrocardiographic norms pri- directly affect the ECG, yielding a study population that
marily to evaluate symptomatic patients in the office or more closely represents the patient population in which
hospital with the purpose of screening for clinically clinicians use the ECG to judge cardiac status.
unapparent cardiac disease. We believe that norms derived The major objective of this work was to establish well-
from populations that are not seen by the medical profession founded electrocardiographic heart rate, interval, and axis
will not serve that purpose well. reference ranges for clinical use. We recommend that
By selecting 96% ranges, we arbitrarily set the false- existing reference ranges be abandoned and replaced by
positive error rate at 2% for high and 2% for low values. age and sex-specific values derived from large, diverse
Other ranges have been used commonly, such as 95% and samples such as ours, and that these more robust norms be
98%. Although there is no established preference or incorporated into ECG computer analysis algorithms and
scientific rationale for one or the other, it is important to teaching curricula.
understand the implicit error rate associated with the
selected range. The choice of a lower false-positive error References
rate (eg, a 98% range) is inevitably associated with a
1. Macfarlane PW, Lawrie TDV. Comprehensive electrocardiology:
higher false-negative error rate (ie, the chance that an theory and practice in health and disease. New York7 Pergamon Press;
abnormal subject would incorrectly be categorized as 1989.
normal), but the absolute magnitude of this error rate 2. Lamb LE. Electrocardiography and vectorcardiography: instrumenta-
depends on the distribution that the parameter takes in the tion, fundamentals, and clinical applications. Philadelphia7 W.B.
Saunders; 1965.
abnormal population.
3. Hiss RG, Lamb LE, Allen MF. Electrocardiographic findings in 67,375
What is the origin of the differences between our asymptomatic subjects X Normal values. Am J Cardiol 1960;6:200.
reference ranges and those of previous studies? In the case 4. Dmitrienko AA, Sides GD, Winters KJ, et al. Electrocardiogram
of the Lilly study, the fact that all of the ECGs were reference ranges derived from a standardized clinical trial population.
overread on paper copies might have systematically affected Drug J Inf 2005;39:395.
5. Simonson E. Differentiation between normal and abnormal in
interval measurements because of the low resolution
electrocardiography. St. Louis7 Mosby; 1961.
provided by paper. Lamb2 and Simonson5 also used paper. 6. Wagner GS. Marriott’s practical electrocardiography. 10th ed. Phila-
In addition, Lilly’s strategy of excluding a priori ECGs that delphia7 Lippincott, Williams and Wilkins; 2001.
were deemed by the reader to have abnormal numerical 7. Willems JL, Arnaud P, van Bemmel, et al. A reference data base for
values may have partially defeated the objective of multilead electrocardiographic computer measurement programs. J Am
Coll Cardiol 1987;10:1313.
determining the true normative range. Electrocardiograms
8. Murray A, McLaughlin NB, Bourke JP, et al. Errors in manual
are frequently assigned an abnormal overall assessment in measurement of QT intervals. Br Heart J 1994;71:386.
normal, healthy individuals because the meaning of 9. Kligfield P, Hancock EW, Helfenbein ED, et al. Relation of QT interval
bnormalQ to most electrocardiographers is not based upon measurements to evolving automated algorithms from different
valid reference ranges but, rather, upon easily remembered, manufacturers of electrocardiographs. Am J Cardiol 2006;98:88.
10. Darpo B, Agin M, Kazierad DJ, et al. Man versus machine: is there an
generally accepted pseudoreference ranges, such as heart
optimal method for QT measurements in thorough QT studies? J Clin
rate between 60 and 100 beats/min. Pharm 2006;46:598.
Additional plausible explanations for differences between 11. Hiss RG, Lamb LE. Electrocardiographic findings in 122,043
our findings and the others reviewed previously are genetic individuals. Circulation 1962;25:947.
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 234.e1

Appendix A Table 2: Primary disease categories in all subjects

Table 1: Geographic location of all subjects Disease category n %
Regions Countries No. of % Total Cancer 1915 2
represented subjects Cardiovascular 7124 9
Diabetes 9315 12
Africa 1 2104 3
HIV 609 1
Asia 12 1451 2
Hyperlipidemia 6594 8
Europe 30 16 471 21
Neuropsychiatric 21 725 27
Latin America 10 2491 3
No systemic disease 23 204 29
Northern America 2 55 670 70
Others 6553 8
Oceania 2 1556 2
Respiratory 2704 3
Total 57 79 743 100

Table 3: Heart rate and PR interval (all subjects)

Group Heart rate (beats per min) PR interval (milliseconds)
Mean SD Median 98% 2% 99% 1% N Mean SD Median 98% 2% 99% 1% N
All 68 12 67 97 47 102 45 79 487 159 26 157 221 115 235 110 78 846
Male 67 13 65 97 46 102 43 41 222 163 27 160 229 117 246 112 40 773
Female 69 12 68 97 49 103 47 38 265 155 24 153 211 113 223 108 38 073
0-9 85 15 84 120 60 129 58 990 128 18 127 167 92 183 86 989
10-11 71 13 70 100 48 105 46 1477 142 20 141 186 104 194 100 1475
20-29 67 12 66 95 46 98 43 6086 150 21 149 198 112 207 106 6086
30-39 69 12 68 95 48 100 46 9569 153 21 152 200 114 208 110 9567
40-49 69 12 68 96 48 101 46 15 392 155 22 154 205 115 215 111 15 379
50-59 68 12 67 97 48 102 46 18 578 160 23 158 213 118 224 113 18 516
60-69 67 12 65 96 46 102 44 16 585 165 27 163 228 119 244 113 16 414
70-79 65 12 64 95 45 101 43 8432 171 31 167 246 120 268 115 8188
80-89 65 12 64 95 46 101 44 2259 176 36 171 276 120 308 114 2124
90-99 70 15 66 125 45 146 43 119 180 37 177 283 116 337 73 108
M, 0-9 84 14 83 117 58 127 57 598 129 18 128 169 94 189 89 597
M, 10-19 70 13 69 99 47 106 45 847 142 21 141 188 104 197 100 846
M, 20-29 64 12 63 91 45 97 42 3127 153 21 151 200 114 210 110 3127
M, 30-39 67 12 66 95 46 99 44 4605 156 21 154 203 118 212 114 4604
M, 40-49 68 12 67 96 47 101 45 7104 159 22 157 208 118 219 114 7096
M, 50-59 68 12 66 97 47 102 45 9936 163 24 161 219 121 232 115 9885
M, 60-69 65 13 64 96 44 102 42 9457 169 28 166 235 122 253 116 9316
M, 70-79 64 13 62 96 44 102 42 4509 177 34 172 263 123 284 117 4333
M, 80-89 63 12 61 94 44 98 41 1001 186 40 180 306 125 326 115 933
M, 90-99 64 11 62 95 43 95 43 38 197 45 191 342 140 342 140 36
F, 0-9 88 15 86 125 62 133 60 392 127 18 126 165 91 169 86 392
F, 10-19 73 12 71 101 53 105 50 630 141 19 140 181 103 191 99 629
F, 20-29 69 11 69 96 49 99 46 2959 148 21 147 195 110 202 104 2959
F, 30-39 70 11 69 96 50 100 48 4964 150 21 149 198 111 205 107 4963
F, 40-49 70 11 69 97 50 102 48 8288 152 22 151 202 113 212 109 8283
F, 50-59 69 11 68 96 49 102 47 8642 156 22 155 207 115 215 111 8631
F, 60-69 68 12 67 95 48 101 46 7128 160 24 158 217 116 227 111 7098
F, 70-79 67 12 66 95 46 101 44 3923 164 26 162 227 118 238 112 3855
F, 80-89 67 12 66 97 47 102 46 1258 168 31 164 246 118 265 112 1191
F, 90-99 72 16 69 144 46 147 44 81 172 30 171 234 91 242 69 72
234.e2 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

Table 4: QRS interval and QRS axis (all subjects)

Group QRS interval (milliseconds) QRS axis (8)
Mean SD Median 98% 2% 99% 1% N Mean SD Median 98% 2% 99% 1% N
All 94 15 93 136 70 149 67 79 487 28 38 30 91 50 98 60 79 122
Male 98 15 96 144 73 155 69 41 222 26 40 28 91 56 100 68 40 985
Female 91 13 90 123 68 138 65 38 265 31 35 33 91 42 95 50 38 137
0-9 79 10 78 101 60 103 57 990 44 36 54 100 0 111 1 981
10-19 89 11 88 113 66 118 63 1477 56 31 63 104 3 110 15 1468
20-29 93 11 93 116 72 121 69 6086 55 29 60 99 14 106 27 6060
30-39 93 11 92 116 71 121 68 9569 43 32 46 91 26 105 36 9528
40-49 92 12 92 118 70 124 67 15 392 34 34 37 91 35 95 45 15 330
50-59 94 14 93 130 71 144 68 18 578 24 36 25 90 45 91 56 18 495
60-69 96 16 94 148 71 158 68 16 585 17 38 15 87 56 91 67 16 510
70-79 98 19 95 155 71 163 68 8432 11 40 1 87 66 91 75 8382
80-89 100 21 96 158 71 166 68 2259 5 41 1 87 73 91 80 2250
90-99 101 25 94 165 66 182 60 119 3 45 14 89 74 124 87 118
M, 0-9 80 10 79 101 61 107 59 598 45 36 54 99 0 120 0 593
M, 10-19 90 12 90 117 65 122 61 847 56 32 64 103 12 110 30 839
M, 20-29 97 11 97 120 75 124 71 3127 56 31 62 101 22 107 34 3112
M, 30-39 96 11 96 120 74 125 70 4605 42 34 45 95 31 105 44 4582
M, 40-49 96 12 96 122 74 130 71 7104 32 36 35 91 42 100 50 7064
M, 50-59 97 14 96 138 74 150 71 9936 22 38 21 90 51 91 63 9879
M, 60-69 100 17 97 152 74 162 71 9457 15 39 10 87 61 91 74 9407
M, 70-79 102 20 98 159 74 170 71 4509 9 44 1 90 73 98 85 4478
M, 80-89 106 22 100 164 75 171 70 1001 3 44 1 87 80 91 87 994
M, 90-99 106 28 95 170 65 170 65 38 7 49 15 76 89 76 89 37
F, 0-9 77 10 77 96 57 101 55 392 43 37 53 102 0 107 15 388
F, 10-19 87 10 87 109 67 112 63 630 56 29 60 105 1 110 4 629
F, 20-29 90 10 89 110 70 112 67 2959 55 27 59 95 3 105 15 2948
F, 30-39 89 10 89 110 69 114 67 4964 45 31 46 91 20 104 30 4946
F, 40-49 89 11 89 111 69 117 66 8288 36 32 39 90 30 91 39 8266
F, 50-59 90 12 90 115 69 129 66 8642 28 34 29 88 40 91 45 8616
F, 60-69 92 15 90 136 69 150 65 7128 20 35 18 87 46 91 56 7103
F, 70-79 94 16 91 143 69 153 66 3923 12 36 1 87 55 90 65 3904
F, 80-89 96 19 92 153 69 160 66 1258 7 38 1 87 66 91 73 1256
F, 90-99 99 23 93 167 64 185 59 81 1 43 13 106 63 132 69 81

Table 5: QT and QTcB intervals in milliseconds (all subjects)

Group QT interval (milliseconds) QTcB Interval (milliseconds)
Mean SD Median 98% 2% 99% 1% QT-RR N Mean SD Median 98% 2% 99% 1% N
slope
All 391 33 389 466 327 478 319 0.1518 79 487 412 26 411 468 362 479 356 79 487
Male 389 34 387 466 325 479 317 0.1490 41 222 406 26 405 465 358 476 352 41 222
Female 393 32 391 465 330 478 321 0.1637 38 265 418 24 417 471 370 482 364 38 265
0-9 347 29 346 405 284 415 276 0.1898 990 409 21 409 455 367 463 363 990
10-19 375 28 373 437 322 448 315 0.1412 1477 405 23 405 453 359 459 350 1477
20-29 383 29 382 447 328 458 321 0.1295 6086 400 24 400 450 352 457 345 6086
30-39 382 29 381 447 326 456 318 0.1397 9569 406 24 405 455 358 462 352 9569
40-49 386 30 385 450 327 460 319 0.1463 15 392 411 24 410 461 364 471 357 15 392
50-59 391 31 390 462 330 473 321 0.1496 18 578 414 25 413 468 365 479 359 18 578
60-69 398 34 396 476 333 490 324 0.1523 16 585 415 26 414 473 366 485 359 16 585
70-79 404 36 402 484 334 498 324 0.1513 8432 418 28 416 481 365 495 359 8432
80-89 408 37 407 493 336 503 327 0.1537 2259 422 29 420 486 366 497 357 2259
90-99 405 43 405 519 328 541 307 0.1765 119 430 34 429 522 364 533 342 119
M, 0-9 349 29 347 408 287 429 277 0.1906 598 408 21 408 453 367 455 363 598
M, 10-19 375 29 372 437 321 449 310 0.1366 847 401 23 402 448 351 454 348 847
M, 20-29 382 29 381 447 326 457 319 0.1316 3127 391 22 391 439 347 446 342 3127
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 234.e3

Group QT interval (milliseconds) QTcB Interval (milliseconds)

Mean SD Median 98% 2% 99% 1% QT-RR N Mean SD Median 98% 2% 99% 1% N
slope
M, 30-39 379 29 377 443 324 455 316 0.1382 4605 397 22 397 446 353 453 349 4605
M, 40-49 382 30 381 448 324 457 316 0.1432 7104 404 23 403 454 358 463 352 7104
M, 50-59 388 32 386 459 327 471 319 0.1470 9936 409 24 408 464 362 474 356 9936
M, 60-69 398 35 396 477 332 491 323 0.1493 9457 411 26 410 471 362 483 356 9457
M, 70-79 404 37 402 487 332 501 323 0.1462 4509 414 29 411 480 362 497 357 4509
M, 80-89 413 37 411 496 340 506 327 0.1447 1001 418 29 417 486 364 493 355 1001
M, 90-99 416 47 423 545 328 545 328 0.2073 38 424 31 427 498 363 498 363 38
F, 0-9 343 28 344 399 277 409 273 0.1897 392 411 22 411 460 365 466 360 392
F, 10-19 375 28 375 438 321 448 315 0.1564 630 410 21 409 458 371 466 363 630
F, 20-29 384 28 382 448 329 462 324 0.1495 2959 410 22 410 455 364 459 358 2959
F, 30-39 386 29 385 448 329 458 320 0.1545 4964 413 22 413 459 368 467 364 4964
F, 40-49 389 29 388 453 332 463 323 0.1563 8288 417 23 416 465 372 474 366 8288
F, 50-59 394 31 394 464 333 476 324 0.1588 8642 420 24 419 472 372 482 365 8642
F, 60-69 399 34 397 475 333 489 325 0.1649 7128 421 25 420 475 372 487 366 7128
F, 70-79 404 35 402 482 335 496 327 0.1641 3923 422 26 420 482 371 492 364 3923
F, 80-89 405 37 404 488 335 502 326 0.1652 1258 424 28 422 490 370 500 361 1258
F, 90-99 400 41 399 499 319 511 302 0.1659 81 434 35 430 530 356 535 337 81

Table 6: QTcF in milliseconds (all subjects)

Group Mean SD Median 98% 2% 99% 1% N
All 404 23 403 458 361 469 355 79 487
Male 400 23 398 455 358 467 353 41 222
Female 409 22 408 460 367 471 361 38 265
0-9 387 19 387 428 347 433 341 990
10-19 394 19 393 434 357 440 353 1477
20-29 394 20 394 437 354 443 350 6086
30-39 397 20 397 441 357 448 352 9569
40-49 402 21 401 447 361 456 355 15 392
50-59 406 22 405 456 364 466 358 18 578
60-69 409 23 407 464 367 476 361 16 585
70-79 413 25 410 473 366 485 360 8432
80-89 417 26 415 478 366 486 360 2259
90-99 421 32 422 503 356 514 351 119
M, 0-9 387 19 388 429 346 438 342 598
M, 10-19 392 19 391 431 354 434 348 847
M, 20-29 388 19 387 429 351 437 347 3127
M, 30-39 391 19 390 434 354 441 350 4605
M, 40-49 396 20 395 441 357 449 352 7104
M, 50-59 401 22 400 451 361 463 355 9936
M, 60-69 406 23 404 462 364 474 359 9457
M, 70-79 410 26 407 472 364 486 359 4509
M, 80-89 416 26 413 476 366 485 359 1001
M, 90-99 421 34 427 507 364 507 364 38
F, 0-9 387 20 387 428 347 433 336 392
F, 10-19 398 18 396 439 362 451 359 630
F, 20-29 401 19 400 440 362 449 357 2959
F, 30-39 404 20 403 446 364 453 359 4964
F, 40-49 407 20 406 452 368 460 363 8288
F, 50-59 411 22 410 459 368 470 362 8642
F, 60-69 413 23 411 468 370 477 363 7128
F, 70-79 415 25 414 474 370 484 361 3923
F, 80-89 417 27 416 480 366 488 360 1258
F, 90-99 422 31 419 504 354 516 351 81
*Patients with pacemakers excluded.
234.e4 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

Table 7: Heart rate and PR interval (reference range subset)

Group Heart rate (beats per min) PR interval (milliseconds)
Mean SD Median 98% 2% 99% 1% N Mean SD Median 98% 2% 99% 1% N
All 69 12 68 98 48 103 46 46 129 156 24 154 212 113 224 109 46 119
Male 68 12 66 98 47 103 45 21 567 159 25 157 218 115 230 110 21 561
Female 70 12 68 98 49 103 47 24 562 153 23 151 205 112 217 107 24 558
0-9 85 15 84 120 60 129 58 963 128 18 127 167 92 179 86 963
10-11 72 13 70 101 49 105 47 1345 142 20 141 186 104 194 100 1345
20-29 67 12 66 94 46 98 43 4997 150 21 149 198 112 206 108 4997
30-39 69 11 68 95 48 99 46 7365 153 21 151 201 114 208 110 7364
40-49 69 12 68 97 49 102 47 10 363 154 22 153 203 114 213 111 10 363
50-59 69 12 68 98 49 102 47 9359 157 23 156 208 116 218 112 9359
60-69 69 12 67 98 48 103 45 6598 161 25 159 221 118 232 112 6596
70-79 66 12 65 93 46 99 44 3790 166 28 164 232 118 251 113 3786
80-89 65 11 64 89 47 96 45 1284 172 34 168 260 120 284 114 1281
90-99 67 12 66 95 47 95 47 65 181 32 179 272 116 282 116 65
M, 0-9 84 14 83 116 59 127 57 579 129 18 128 169 94 185 88 579
M, 10-19 71 13 70 99 48 106 47 776 142 20 141 188 105 195 100 776
M, 20-29 64 12 63 91 44 97 42 2528 153 21 151 200 114 211 111 2528
M, 30-39 67 12 66 95 47 100 44 3411 156 21 154 203 118 210 114 3411
M, 40-49 68 12 67 97 47 101 45 4316 158 21 156 206 117 215 114 4316
M, 50-59 69 12 68 98 48 103 46 4460 160 23 159 212 120 222 115 4460
M, 60-69 68 13 67 99 47 104 45 3275 165 26 163 226 120 239 115 3273
M, 70-79 65 12 63 94 45 100 43 1718 172 29 168 250 123 264 115 1716
M, 80-89 62 10 61 87 46 95 44 483 182 37 177 282 121 308 114 481
M, 90-99 62 12 59 95 48 95 48 21 191 39 188 282 140 282 140 21
F, 0-9 88 15 86 126 62 133 60 384 127 18 126 165 91 171 86 384
F, 10-19 73 12 71 101 52 104 50 569 141 19 140 181 102 190 99 569
F, 20-29 69 11 69 96 49 99 46 2469 148 21 147 195 110 200 105 2469
F, 30-39 70 11 69 95 50 99 48 3954 150 21 148 198 111 204 108 3953
F, 40-49 70 11 69 97 50 103 48 6047 151 21 150 201 112 211 109 6047
F, 50-59 69 11 68 97 49 102 47 4899 155 22 153 203 115 211 111 4899
F, 60-69 69 12 68 97 49 102 47 3323 158 23 156 213 116 224 111 3323
F, 70-79 67 11 66 93 47 98 45 2072 161 25 160 221 116 231 110 2070
F, 80-89 66 11 65 91 48 96 47 801 166 29 162 237 119 264 114 800
F, 90-99 70 11 68 94 47 94 47 44 176 26 177 225 116 225 116 44

Table 8: QRS interval and QRS axis (reference range subset)

Group QRS interval (milliseconds) QRS axis (8)
Mean SD Median 98% 2% 99% 1% N Mean SD Median 98% 2% 99% 1% N
All 91 10 91 109 69 109 66 46 129 34 35 37 91 40 93 49 45 944
Male 94 10 94 109 71 112 68 21 567 33 37 36 91 45 97 54 21 466
Female 88 10 88 107 68 108 65 24 562 35 34 38 90 35 91 45 24 478
0-9 79 10 78 101 59 103 57 963 44 36 53 101 0 112 1 955
10-19 88 10 88 108 65 109 62 1345 56 30 63 102 3 109 15 1338
20-29 92 10 93 109 71 110 69 4997 55 29 60 95 10 103 25 4978
30-39 91 10 92 109 70 109 68 7365 44 31 47 91 25 98 35 7337
40-49 91 10 91 109 70 109 67 10 363 36 33 40 90 31 91 42 10 327
50-59 91 10 92 109 70 110 67 9359 27 34 30 87 41 91 49 9316
60-69 91 10 92 109 70 109 67 6598 20 34 20 87 46 87 56 6567
70-79 91 10 92 109 69 109 66 3790 14 35 9 87 51 87 62 3779
80-89 91 10 91 109 70 109 67 1284 8 36 1 77 60 87 70 1283
90-99 89 10 89 112 63 114 59 65 3 36 8 76 60 76 60 64
M, 0-9 80 10 79 101 61 107 59 579 44 36 53 100 0 122 0 574
M, 10-19 89 11 89 108 65 110 61 776 56 31 65 102 6 110 30 770
M, 20-29 95 9 96 109 75 112 71 2528 56 30 62 100 22 104 30 2519
M, 30-39 94 9 95 109 74 112 70 3411 43 33 47 95 30 101 40 3397
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 234.e5

Group QRS interval (milliseconds) QRS axis (8)

Mean SD Median 98% 2% 99% 1% N Mean SD Median 98% 2% 99% 1% N
M, 40-49 94 10 94 109 73 112 70 4316 34 35 39 91 38 95 46 4296
M, 50-59 94 9 94 109 73 112 71 4460 24 36 25 87 45 91 56 4435
M, 60-69 94 9 95 109 74 112 71 3275 18 36 17 87 51 90 63 3259
M, 70-79 94 10 95 109 72 109 69 1718 12 37 1 87 59 88 65 1714
M, 80-89 94 9 95 109 74 111 71 483 8 38 1 87 65 87 76 482
M, 90-99 88 7 89 100 74 100 74 21 1 43 1 76 60 76 60 20
F, 0-9 77 10 77 96 57 101 55 384 42 37 52 103 0 107 15 381
F, 10-19 86 9 86 107 65 108 63 569 55 29 61 104 3 108 7 568
F, 20-29 89 9 89 107 70 108 67 2469 55 26 59 91 3 100 12 2459
F, 30-39 89 9 89 107 69 109 66 3954 45 29 47 91 16 91 28 3940
F, 40-49 88 10 88 107 68 108 66 6047 38 31 41 90 25 91 33 6031
F, 50-59 89 10 89 108 68 108 65 4899 30 32 32 87 35 91 45 4881
F, 60-69 89 10 89 108 68 108 64 3323 22 33 23 87 45 87 49 3308
F, 70-79 89 10 89 108 68 109 65 2072 15 34 11 87 46 87 56 2065
F, 80-89 89 10 89 108 69 109 66 801 8 35 1 70 56 85 66 801
F, 90-99 90 11 89 114 59 114 59 44 5 34 13 75 60 75 60 44
Table 9: QT and QTcB intervals in milliseconds (reference range subset)
Group QT interval (milliseconds) QTcB Interval (milliseconds)
Mean SD Median 98% 2% 99% 1% QT-RR N Mean SD Median 98% 2% 99% 1% N
slope
All 385 31 384 452 325 463 315 0.1518 46 129 409 23 409 457 361 463 355 46 129
Male 381 31 380 449 322 459 313 0.1441 21 567 402 23 401 449 356 456 350 21 567
Female 388 31 387 455 327 466 318 0.1585 24 562 415 22 414 460 369 467 363 24 562
0-9 347 29 346 406 283 416 276 0.1902 963 410 21 410 455 367 461 363 963
10-19 374 28 373 435 321 448 315 0.1450 1345 405 22 405 453 359 458 351 1345
20-29 382 29 381 447 328 457 322 0.1404 4997 400 23 400 448 353 454 345 4997
30-39 381 28 380 444 326 453 318 0.1470 7365 405 23 404 452 358 457 352 7365
40-49 384 29 383 448 326 457 318 0.1508 10 363 409 23 409 457 363 464 357 10 363
50-59 386 30 385 451 326 463 320 0.1532 9359 412 23 411 459 365 466 359 9359
60-69 389 31 388 457 328 470 319 0.1572 6598 412 23 412 459 366 465 360 6598
70-79 397 33 395 469 333 478 322 0.1560 3790 413 24 412 463 366 471 359 3790
80-89 402 33 402 469 338 482 333 0.1575 1284 415 24 414 462 365 470 357 1284
90-99 400 33 401 466 333 469 332 0.1793 65 420 23 422 471 364 476 363 65
M, 0-9 349 29 347 408 286 429 277 0.1906 579 408 20 408 452 368 454 364 579
M, 10-19 374 28 371 434 321 448 310 0.1366 776 402 23 403 448 352 454 348 776
M, 20-29 381 29 380 447 326 457 317 0.1316 2528 391 22 390 436 347 442 343 2528
M, 30-39 377 28 375 439 324 451 316 0.1382 3411 396 22 396 443 353 449 347 3411
M, 40-49 379 29 377 444 323 453 314 0.1432 4316 401 21 401 446 357 454 351 4316
M, 50-59 381 29 381 444 325 453 317 0.1470 4460 406 22 405 451 361 457 354 4460
M, 60-69 386 31 384 455 325 466 318 0.1493 3275 407 22 407 456 362 460 357 3275
M, 70-79 395 33 394 466 329 478 320 0.1462 1718 407 23 406 456 362 462 357 1718
M, 80-89 405 31 406 466 342 478 331 0.1447 483 409 23 409 455 361 459 352 483
M, 90-99 407 36 410 469 335 469 335 0.2073 21 411 24 421 442 363 442 363 21
F, 0-9 343 28 344 399 277 409 273 0.1897 384 411 22 411 461 365 466 360 384
F, 10-19 375 28 375 436 321 448 316 0.1564 569 409 21 408 457 371 463 363 569
F, 20-29 384 28 382 449 330 462 325 0.1495 2469 409 21 409 454 364 458 357 2469
F, 30-39 385 28 384 447 329 456 320 0.1545 3954 412 21 412 455 367 460 363 3954
F, 40-49 388 29 387 449 331 460 322 0.1563 6047 415 22 414 460 371 469 365 6047
F, 50-59 391 30 390 457 331 468 322 0.1588 4899 417 23 416 463 370 469 364 4899
F, 60-69 391 31 391 458 330 474 322 0.1649 3323 417 22 416 462 370 468 364 3323
F, 70-79 398 33 396 471 334 478 327 0.1641 2072 418 23 417 467 370 475 364 2072
F, 80-89 400 33 399 474 336 482 332 0.1652 801 418 24 417 467 370 480 360 801
F, 90-99 396 31 398 455 332 455 332 0.1659 44 424 22 423 476 373 476 373 44
234.e6 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

Table 10: QTcF in milliseconds (reference range subset)

Group Mean SD Median 98% 2% 99% 1% N
All 399 21 398 446 359 454 353 26 498
Male 394 20 393 440 355 448 351 14 297
Female 405 21 404 450 365 459 360 12 201
0-9 387 21 387 432 343 442 335 469
10-11 394 18 393 434 357 439 353 550
20-29 393 20 393 434 354 440 350 3623
30-39 396 20 395 439 357 447 352 4813
40-49 399 20 398 442 360 451 355 6855
50-59 402 21 401 448 363 456 356 5707
60-69 406 22 404 454 364 463 359 3385
70-79 411 23 409 461 367 472 361 947
80-89 416 25 413 470 365 480 354 143
90-99 411 31 406 467 377 467 377 6
M, 0-9 387 22 388 436 341 449 334 263
M, 10-19 391 19 391 433 353 437 348 322
M, 20-29 388 18 388 428 352 437 347 2149
M, 30-39 390 19 389 433 353 441 350 2639
M, 40-49 394 19 393 435 356 442 352 3445
M, 50-59 398 20 398 444 360 452 354 3250
M, 60-69 401 21 400 448 362 454 357 1785
M, 70-79 406 23 405 461 365 483 356 385
M, 80-89 419 25 417 479 351 482 349 57
M, 90-99 386 13 386 396 377 396 377 2
F, 0-9 386 20 387 431 347 435 334 206
F, 10-19 397 17 396 437 362 449 360 228
F, 20-29 400 19 400 439 362 447 357 1474
F, 30-39 402 19 402 443 365 449 361 2174
F, 40-49 405 20 404 448 368 458 363 3410
F, 50-59 407 21 407 450 366 461 361 2457
F, 60-69 411 22 410 461 370 471 362 1600
F, 70-79 414 22 413 462 372 471 361 562
F, 80-89 414 25 410 472 366 478 365 86
F, 90-99 424 30 417 467 395 467 395 4

Table 11: Comparison of our 96% interval reference ranges Table 12: Comparison of our distribution of QRS interval
to Macfarlane’s and Lawrie’s1 durations to Lamb’s2
Macfarlane Current QRS Lamb study2 Current study
and Lawrie1 study study (reference (milliseconds) (reference range subset)
range subset) Male Female Male Female
Male Female Male Female N 45-55 0.3 0.8 0.0 0.0
Age 30-39 y N 55-65 5.0 3.8 0.4 0.9
N 218 115 3411 3954 N 65-75 19.9 23.0 3.0 7.0
HR (beats per min) 52-99 57-105 47-95 49-96 N 75-85 47.6 62.9 14.7 26.7
PR (milliseconds) 116-206 114-184 114-201 118-203 N 85-95 19.6 6.8 33.0 37.4
QRS (milliseconds) 78-114 76-106 74-109 69-107 N 95-105 6.7 2.3 34.9 22.7
QTcB (milliseconds) 375-468 395-473 353-443 367-455 N 105-115 0.8 0.4 13.5 5.0
Age 40-49 y N 115 * * 0.5 0.1
N 119 72 4316 6047 *Excluded in Hiss3 study.
HR (beats per min) 49-96 59-106 47-97 50-97
PR (milliseconds) 116-210 108-200 117-206 112-201
QRS (milliseconds) 78-114 74-108 73-109 68-107
QTcB (milliseconds) 377-464 350-483 357-446 371-460
HR indicates heart rate.
 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8 234.e7
4
Table 13: Comparison of our interval reference ranges to Lilly’s
Lilly study4 Lilly study4 Current study
Normal ECG Abnormal ECG Reference range subset
(98% Range) (98% Range) (96% Range)
Male Female Male Female Male Female
Age 36-45 y
N 619* 902* 506* 741* 4118 5509
HR (beats per min) 60 -99 60-98 45-117 45-110 47-96 50-97
PR (milliseconds) 112-236 104-208 112-200 112-192 117-205 112-200
QRS (milliseconds) 69-112 68-104 72-144 68-116 73-109 69-107
QTcL** (milliseconds) 371-446 373-459 358-462 380-479 357-437 370-449
Age 46-55 y
N 686* 917* 563* 753* 4445 5534
HR (beats per min) 60-96 60-96 44-108 46-107 49-98 49-97
PR (milliseconds) 116-228 108-224 112-200 112-196 120-208 114-201
QRS (milliseconds) 72-112 68-100 72-152 64-132 74-109 69-107
QTcL** (milliseconds) 370-452 383-464 381-474 383-492 360-441 370-455
*Approximation estimated from known gender distribution and known normal/abnormal distribution.
**QTcL = Lilly log-linear correction, where QTc = QT/RR0.413.
Table 14: Comparison of our interval reference ranges to automatically by the Veritas computer algorithm (version
Simonson’s5 1.20.23) and once by a cardiologist using digitography, who
did not have access to the Veritas algorithm results. From the
Simonson study5* Current study
total of 28 372 ECGs, heart rate, QRS and QT values were
2.5%-97.5% 2%-98% available from both Veritas and digitography in all cases, and
Age 20-59 y Male Female Male Female PR values were available in 28 234 ECGs.
n 649 311 14 715 17 369 GE 12SL
RR 630-1140 600 -1090 624-1284 623-1213 This analysis is summarized in Table 17. It includes
(milli- 43 906 ECGs of 185 subjects who participated in two
seconds) definitive QT studies. One study involved 4707 ECGs
PR 130-200 120-200 118-207 112-200 obtained in 49 subjects with solid tumors, and the other
(milli- included 39 199 ECGs recorded in 136 normal volunteers.
seconds) Each ECG was read once automatically by the 12 SL
QRS 70-120 60-110 73-109 69-107 computer algorithm (version 5.21) and once by a cardiol-
(milli- ogist using digitography, who did not have access to the 12
seconds) SL algorithm results. From the total of 43 906 ECGs, heart
QT 330-440 320-420 324-444 330-451 rate, QRS and QT values were available from both 12 SL
(milli- and digitography in 43 826 cases, and PR values were
seconds) available in 43 566 ECGs.
*Data expressed to the nearest centisecond.

Table 15: Interval values read on paper minus values read in

Methodology: Comparison of digitography
Digitography to Other Reading Methods
Interval Difference SD % Difference
Reading on Paper HR (beats per min) 0.05 0.94 0.07
The analysis summarized in Table 15 was derived from PR (milliseconds) 2.03 4.02 1.42
180 readings of 10 ECGs. Three cardiologists performed the QRS (milliseconds) 5.64 7.31 6.49
readings. Each read each of the 10 ECGs three times on QT (milliseconds) 9.32 6.15 2.33
paper and three times on screen in digitography. The repeat
readings were separated by at least a week. Only the Table 16: Mortara Veritas interval values minus values read
interreading variability is assessed in Table 15. in digitography
Mortara Veritas
Interval (milliseconds) Difference SD % Difference
This analysis is summarized in Table 16. It includes 28 372
ECGs of 242 subjects who participated in two definitive QT RR 1.90 29.8 0.21
studies. One study involved 21 439 ECGs obtained in 198 PR 4.72 9.08 2.99
normal volunteers, and the other included 6933 ECGs QRS 0.05 9.01 0.05
recorded in 44 normal volunteers. Each ECG was read once QT 5.22 10.21 1.35
234.e8 J.W. Mason et al. / Journal of Electrocardiology 40 (2007) 228–234.e8

Fig. 1. Weight, height, and age distributions of all 79 743 subjects (panels A-C) and the reference range subset (panels D-F).

determinations for QRS axis performed using standard

Table 17: GE 12SL interval values minus values read in methods, we did the following:
digitography
! Rotated the whole distribution by 608 so that the
Interval (milliseconds) Difference SD % Difference median value (30) became 908 (exactly 1808 from the
RR 6.71 24.37 0.70 break between +2708 and 898)
PR 3.75 7.86 2.39 ! Subtracted 3608 from any derived values that
QRS 1.82 6.87 1.99 exceeded 2708
QT 7.84 16.96 1.99 ! Recalculated all the summary statistics.
! Rotated the derived statistics back to the original
position of the distribution by subtracting 608 from all
Methodology: QRS Axis Distribution variables (except for the SD of the mean).
The QRS axis is distributed on a circle, such that the Because this did not change the resultant distribution
largest and smallest values are adjacent to one another. To statistics significantly, we believe that use of the usual
determine if this significantly distorted our distribution method of calculating distribution of QRS axis is acceptable.