Electrocardiograms in Healthy North American Children in the Digital Age

Abstract

Background

Interpretation of pediatric electrocardiograms (ECGs) is limited by lack of accurate sex- and race- specific normal reference values obtained with modern technology for all ages. We sought to obtain contemporary digital ECG measurements in healthy children from North America, to evaluate the effects of sex and race, and to compare our results to commonly used published datasets.

Methods

Digital ECGs (12-lead) were retrospectively collected for children ≤18 years old with normal echocardiograms at 19 centers in the Pediatric Heart Network. Patients were classified into 36 groups: 6 age, 2 sex, and 3 race (white, African-American, and other/mixed) categories. Standard intervals and amplitudes were measured; mean ± SD and 2^nd/98^th percentiles were determined by age group, sex, and race. For each parameter, multivariable analysis, stratified by age, was conducted using sex and race as predictors. Parameters were compared to two large pediatric ECG data sets.

Results

Among ECGs from 2400 children, significant differences were found by sex and race categories. The QTc in lead II was greater for females compared to males for age groups ≥3 years (p ≤ 0.03) and for whites compared to African-Americans for age groups ≥12 years (p < 0.05). The R wave amplitude in V6 was greater for males compared to females for age groups ≥12 years (p < 0.001), for African-Americans compared to white or other race categories for age groups ≥3 years (p ≤ 0.006), and greater compared to a commonly used public data set for age groups ≥12 years (p < 0.0001).

Conclusions

In this large, diverse cohort of healthy children, most ECG intervals and amplitudes varied by sex and race. These differences have important implications for interpreting pediatric ECGs in the modern era when used for diagnosis or screening, including thresholds for left ventricular hypertrophy.

Introduction

The electrocardiogram (ECG) is a cornerstone in the evaluation of children with acquired and congenital heart disease and may provide a basis for cardiac screening in the future.^1,2 As a diagnostic tool, the modern ECG has multiple advantages including the potential to detect cardiac disease, accurate automated interpretation, digital data storage, low cost, and essentially no risk. Currently, the utility of ECGs in the identification of heart disease in children is limited by a lack of reliable reference values leading to poor accuracy.³ The amplitude and duration of surface ECG waveforms are affected by age, sex, cardiac rhythm, heart position, and the size of cardiovascular structures.^4–12 Electrocardiographic data may also be affected by race and ethnicity,^13–15 but there is little supportive evidence in the pediatric population. Recent data including children indicate refinement of population and patient specific reference values improves diagnostic accuracy of the ECG,^16–17 and the aim of this study is to further define the effects of sex and race on patient specific ECG reference values for healthy children in North America from birth to age 18 accounting for age.

Prior studies defining normal ECG data in children are limited by wide variation in methodology, inclusion criteria, number of subjects, and population.^4–8 Indeed, the two most referenced normal data sets, from Davignon⁴ and Rinjbeek⁵, were published before current AHA recommendations were issued in 2009 in response to a shift from analog to modern, computerized, digital ECG recording and analysis.² Available pediatric reports are largely constructed from single clinical sites and failed to account for analog versus digital ECG recording, geographic variations, or the influences of sex and race. Due to these differences in methodology and populations, there have been wide variations in reported normative values including the QT interval and precordial R wave amplitudes.^4–8 Focusing specifically on the ability to identify increased LV mass, the sensitivity and specificity of current normal ECG parameters are estimated at 80–90% for children in the United States.³

We aimed to describe common contemporary ECG measurements adjusted for age, sex, and race for children with normal hearts who enrolled in the Pediatric Heart Network (PHN) Echocardiogram Z-Score and ECG Database Project.^18,19 We explored the relationship between age group, sex and race for these measurements within the data set. Finally, we compared the median, lower (2^nd) and upper (98^th) percentiles of these measurements to existing public data sets from the Davignon⁴ and Rijnbeek⁵ studies including statistical assessment of differences where possible.

Patients and Methods

Study Design

Demographic and clinical data, echocardiogram images and ECGs were collected from the records of healthy children at 19 centers in the PHN under the Echocardiogram Z-Score and ECG Database protocol. Because all data and images were collected retrospectively and were de-identified before submission, most children were enrolled under a waiver of consent after Institutional Review Board (IRB) or Research Ethics Board (REB) approval was obtained at each participating center. Race and ethnicity information were not routinely obtained at one center, so these data were collected after obtaining informed consent and IRB/REB approval. The subset of children enrolled in the PHN Echocardiogram Z-Score and ECG Database Project who had a digital 12-lead ECG uploaded to the database was included in this analysis. The first and second authors had full access to the data in the study and take responsibility for its integrity and the data analysis.

Study Population

Healthy children ≤18.0 years old with echocardiograms performed after January 2008 and with documentation of age, height, weight, sex, and self-reported race were eligible for this study. Healthy children were identified by having a normal echocardiogram and no evidence of cardiac, inherited or medical disease on review of the available medical records (Supplemental Table 1). Age groups were pre-specified in the PHN Echocardiogram Z-score study. A pre-appointed committee adjudicated potentially normal anatomic variants for the echocardiograms. Available ECGs with minimum standards of 12 leads, 10-second recording, sampling rate >500 Hz, and 150 Hz bandwidth in exportable digital format (MUSE, GE Healthcare, Waukesha, WI or Philips Healthcare, Andover, MA) were collected closest to the date of the echocardiogram and included in the analysis. The study population (Table 1) represents a convenience sample subset of the PHN Echocardiogram Z-Score study. Self-reported race and ethnicity were divided into three categories for the study: white, African-American, and other (Hispanic, Asian, Pacific Islander, Native American, and multiracial).

Table 1A.

Male Subjects with Measurable ECG by Age and Race^*.

Race	<1 Month	[1 Month - <3 Years)	[3–6) Years	[6–12) Years	[12–16) Years	[16–18] Years	Total
White	54	80	68	106	86	78	472
African-American	34	68	71	75	70	63	381
Other or Mixed†	57	81	73	73	75	63	422
Total	145	229	212	254	231	204	1275

^*Sample sizes smaller than the target N=60 are bolded

^†“Other or Mixed” race includes data for race categories “Asian”, “American Indian or Alaska Native” and “Native Hawaiian or other Pacific Islander”, any non-Hispanic subject where more than one racial category was marked as “Yes” indicating mixed race, and those whose ethnicity were marked as “Hispanic or Latino/Latina”.

Children were excluded (Supplemental Table 1) for evidence of acquired or congenital heart disease, corrected gestational age <37 weeks, obesity, acute or systemic disorder typically associated with cardiovascular manifestations, first-degree relative with a non-ischemic cardiomyopathy, first-degree relative with a left-sided obstructive congenital heart lesion, or ECG waveforms that did not meet minimum digital standards or were inadequate for analysis.

Data Collection

Demographic, clinical, echocardiogram and ECG data were obtained by participating centers and uploaded to a cloud database housed within the PHN Data Coordinating Center Bioinformatics Grid. All patient-specific information on the ECGs from the clinical centers was removed through PHN software tools with concurrent assignment of blind identifiers. All data transfers were completed through secure protocols using Hypertext Transfer Protocol Secure (HTTPS). Real time validations, including both inter- and intra-instrument data checks, were integrated into the data entry system.

ECG Waveform Analysis

Deidentified eXtensible Markup Language (XML) format ECGs were transferred to the PHN ECG Core Lab server and were analyzed using Cal-ECG (A.M.P.S. LLC, New York, NY) software. The AMPS software suite is FDA-certified and widely used in clinical research trials and throughout the pharmaceutical industry for assessment of drug-induced ECG changes. The AMPS algorithms for interval measurements have been validated and published in peer-reviewed journals.^20,21 Quality assurance metrics consisted of 3 distinct measures: analysis of the waveforms to detect artifact, inter-, and intra-reviewer variability. The Cal-ECG software automated artifact detection program identified a disproportionate number of ECGs for infants with a heart rate above 150 bpm, so two board certified pediatric cardiologists reviewed all ECGs automatically excluded for artifact and manually determined if the waveforms were interpretable. Each of 64 parameters was measured by the Cal-ECG software and all were subsequently reviewed, including visual inspection of the digital ECG waveform, and validated by at least one board-certified pediatric cardiologist (Supplemental Table 2). The QT interval was measured from the beginning of the QRS complex to the end of the T wave defined as the intercept between the isoelectric line with the tangent drawn through the maximum down slope of the T wave; the second phase for biphasic T waves were included but U waves were excluded from QT interval. Additional customary parameters were calculated from these 64 measured values by the PHN Data Coordinating Center using standard methods (Supplemental Table 2).

Statistical Analysis

Thirty-six study groups were evaluated based on 6 age categories (<1 month, 1 month to <3 years, 3 to <6 years, 6 to <12 years, 12 to <16 years, 16 to 18 years), sex (male, female), and 3 race categories (white, African American, other). The sample size was greater than the target of 60 in 27/36 groups. A sample size of 60 per group was chosen so that the margin of error for the mean would be 25% of the observed standard deviation.

For every ECG parameter, descriptive statistics were determined for each age, sex, and race subgroup. All analyses were conducted using SAS v9.4 (SAS Institute Inc., Cary, NC). For each parameter, a two-way analysis of variance (ANOVA), stratified by age group, was conducted using sex, race category and the race category*sex interaction term as predictors. The interaction term tested whether the effect of sex depended on race or vice versa. In addition, for each parameter, a multivariable regression analysis, stratified by age group, was conducted using only sex and race category as predictors (excluding the interaction term). This model allowed us to test if there was a difference between sex categories after accounting for race and whether there was a difference among race categories after accounting for sex. Furthermore, for each parameter, a one-way analysis of variance was conducted comparing the 12–16 year old age group to the 16–18 year old age group. No formal adjustment for multiple testing was performed, but consistency of results across multiple outcomes was emphasized, as well as comparison of the number of significant effects to that which would be expected due to chance alone.

In addition, children in our study cohort were re-categorized into the Davignon age groups to make a direct comparison with their published data.² Because only published means and standard deviations (SDs) were available from Davignon, we used a one-sample t-test to compare the data. Comparisons were limited to amplitudes where mean values were published including males and females age 12 to 16 years. No mean values were reported in the normal ECG data set published by Rijnbeek et al. in 2001, so no statistical comparisons could be made with the PHN Normal ECG data (Supplemental Tables 3–9). We reported raw rather than adjusted means to allow some comparison with the Rijnbeek dataset.³

Results

Among 2619 ECGs uploaded to the PHN data grid, the core lab excluded 219 due to inadequate quality of waveforms, leaving 2400 for this analysis (Table 1A and 1B; Supplemental Tables 10–82). Intra-reader variability was <2 ms for all intervals and <0.10 mV for all amplitudes. Inter-reader variability was a maximum of +/−5 ms for intervals including the QT and a maximum of 0.15 mV for the amplitudes including the R waves. Small but statistically and clinically significant differences were found in many intervals and amplitudes between sex and among race categories (Supplemental Tables 83–84).

Table 1B.

Female Subjects with Measurable ECG by Age and Race^*.

Race	<1 Month	[1 Month -<3 Years)	[3–6) Years	[6–12) Years	[12–16) Years	[16–18] Years	Total
White	55	77	75	79	87	75	448
African-American	28	62	46	66	66	42	310
Other or Mixed†	29	68	72	78	69	51	367
Total	112	207	193	223	222	168	1125

^*Sample sizes smaller than the target N=60 are bolded

^†“Other or Mixed” race includes data for race categories “Asian”, “American Indian or Alaska Native” and “Native Hawaiian or other Pacific Islander”, any non-Hispanic subject where more than one racial category was marked as “Yes” indicating mixed race, and those whose ethnicity were marked as “Hispanic or Latino/Latina”.

QT Intervals

Differences in QTc measurements up to 10 milliseconds (ms) in lead II were found using both Bazett’s (Table 2) and Fridericia (Supplemental Table 85) correction for heart rate between sex and among race categories. Compared to males, females had a longer mean QTc by 5 to 10 ms at ≥3 years of age (p = 0.008, 0.03, 0.03, 0.002 for male vs. female in age groups 3–<6 years, 6–<12 years, 12–<16 years and 16–18 years, respectively) using Bazett’s correction (Table 2). After accounting for sex, above age 12 years the mean QTc measured 9–10 ms longer for white children than for African-American children in the same age categories (p = 0.001 and 0.02 for white vs. African-American in age groups 12–<16 years and 16–18 years, respectively) using Bazett’s correction (Table 2). No interactions between sex and race were detected for the QTc using Bazett’s correction for heart rate (p > 0.05 for all).

Table 2.

Bazett’s Corrected QT Interval - II (ms) for the Six Primary Age Categories by Race and Sex^*.

	Male		Female		White		African-American		Other/Mixed		P-value Sex	P-value Race
	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)
<1 Month*	138	392±29 393 (304,455)	111	398±25 395 (349, 459)	104	397±28 398 (347, 461)	61	387±27 388 (347, 450)	84	397±25 399 (343, 455)	0.11	0.03
1 Month - <3 Years*	229	388±25 387 (342, 446)	205	393±26 390 (345, 455)	156	392±24 389 (352, 451)	130	391±26 388 (338, 454)	148	390±27 388 (339, 464)	0.11	0.74
3 - <6 Years*	211	384±21 383 (343, 431)	191	390±24 389 (356, 466)	143	390±22 387 (345, 449)	115	386±23 386 (343, 443)	144	386±23 385 (343, 427)	0.008	0.29
6 - <12 Years*	254	390±21 388 (356, 439)	223	394±24 392 (353, 455)	185	395±25 392 (350, 459)	141	389±20 384 (355, 439)	151	392±22 391 (356, 452)	0.03	0.10
12 - <16 Years*	231	390±25 387 (342, 445)	222	395±23 392 (357, 445)	173	397±25 394 (356, 450)	136	387±24 385 (342, 444)	144	391±23 388 (351, 435)	0.03	0.002
16 - 18 Years*	202	384±27 382 (335,453)	168	394±22 394 (344, 442)	152	393±25 393 (344, 442)	105	384±26 383 (334, 441)	113	388±26 387 (337, 452)	0.002	0.047

^*interaction between sex and race was not statistically significant

R Waves

Differences in precordial R wave measurements were noted between sex and among race categories in older children (Table 3). Overall, African-Americans had the tallest precordial R waves in all age categories ≥3 years (Table 3) with a maximum difference in the mean of 0.305 mV above age 16 compared to subjects with other/mixed race (p < 0.001) and a maximum difference in the 98^th percentile of 0.677 mV compared to subjects with white race (p < 0.001). Males had taller precordial R waves than females in V6 for ages ≥12 years (Table 3) with a mean difference of about 0.350–0.400 mV (p <0.001 for male vs. female in age groups 12-<16 years and 16–18 years, respectively). There was one statistically significant race*sex interaction calculated for the R wave in V6 in the age group 3–<6 years (Table 3, p = 0.04), but in the context of multiple testing this is likely due to chance.

Table 3.

R Wave Amplitude in V6 (millivolts) for each of Six Primary Age Categories by Race and Sex.

	Male		Female		White		African-American		Other/Mixed		P-value Sex	P-value Race
	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)	N	Mean ± SD Median (2nd, 98th)
<1 Month	143	0.756±0.425 0.665 (0.157, 2.027)	111	0.868±0.445 0.773 (0.189, 1.846)	108	0.825±0.444 0.752 (0.127,1.942)	61	0.875±0.509 0.834 (0.225, 2.184)	85	0.729±0.358 0.663 (0.159, 1.606)	0.07	0.17
1 Month - <3 Years	229	1.446±0.558 1.370 (0.407, 2.911)	207	1.455±0.588 1.386 (0.413, 2.953)	157	1.490±0.566 1.408 (0.447, 2.914)	130	1.406±0.622 1.299 (0.358, 3.170)	149	1.447±0.530 1.402 (0.439, 2.650)	0.89	0.46
3 - <6 Years*	212	1.629±0.648 1.561 (0.663, 3.457)	193	1.626±0.661 1.521 (0.651, 3.241)	143	1.531±0.563 1.488 (0.676, 3.011)	117	1.801±0.628 1.705 (0.820, 3.547)	145	1.582±0.729 1.476 (0.503, 3.841)	0.74	0.002
6 - <12 Years	254	1.757±0.698 1.677 (0.656, 3.571)	222	1.701±0.572 1.587 (0.717, 3.203)	185	1.704±0.656 1.598 (0.670, 3.530)	140	1.870±0.653 1.778 (0.627, 3.525)	151	1.634±0.595 1.549 (0.656, 2.843)	0.38	0.006
12 - <16 Years	231	1.798±0.631 1.700 (0.744, 3.357)	222	1.383±0.482 1.334 (0.619,2.552)	173	1.590±0.587 1.534 (0.348, 3.169)	136	1.714±0.632 1.592 (0.741, 3.197)	144	1.486±0.564 1.398 (0.619, 2.903)	<.001	0.003
16 - 18 Years	204	1.695±0.543 1.677 (0.705, 2.974)	168	1.346±0.393 1.267 (0.631, 2.344)	153	1.512±0.410 1.479 (0.680, 2.298)	105	1.714±0.573 1.711 (0.620, 2.975)	114	1.409±0.531 1.288 (0.747, 2.845)	<.001	<.001

^*interaction between sex and race statistically significant p=0.

Comparison to prior large pediatric ECG data sets

Supplemental Table 86 lists the breakdown in number of patients in our study limited to age groups used by Davignon.⁴ The R wave amplitudes in the PHN data set were statistically significantly different (all p ≤ 0.01) for all precordial leads except V5 in both sex categories and V1 and V4 in males (Table 4).

Table 4.

Comparison Sex Differences in Precordial R Wave Amplitudes (millivolts) in Age Group 12–16 Years Between Means from PHN to Davignon et al, 1979⁴.

Lead	Dataset	N	Sex	Mean	Std.	P-value
V1	Davignon	142	Female	0.37	0.23	<0.001
	PHN	216	Female	0.29	0.21
	Davignon	105	Male	0.44	0.25	0.49
	PHN	227	Male	0.42	0.24
V2	Davignon	142	Female	0.92	0.36	<0.001
	PHN	220	Female	0.62	0.36
	Davignon	105	Male	1.11	0.40	<0.001
	PHN	231	Male	0.92	0.46
V4	Davignon	142	Female	1.76	0.57	0.003
	PHN	222	Female	1.56	0.67
	Davignon	105	Male	2.58	0.72	0.19
	PHN	231	Male	2.46	0.88
V5	Davignon	142	Female	1.66	0.50	0.12
	PHN	222	Female	1.57	0.56
	Davignon	105	Male	2.40	0.60	0.30
	PHN	231	Male	2.32	0.75
V6	Davignon	142	Female	1.23	0.30	<0.001
	PHN	222	Female	1.38	0.48
	Davignon	105	Male	1.58	0.40	0.001
	PHN	231	Male	1.80	0.63

^*p-values are obtained by T-test comparing means between the two datasets for each sex within each lead.

Mean values were not available in the Rijnbeek⁵ report, so statistical comparisons were not possible, but we were able to make observational assessments of discrepancies in the parameters reported in our study from those reported by Rijnbeek. For example, the 98^th percentile for R waves in V6 in our study were higher for both males and females with a 0.3 millivolt difference in many categories (Supplemental Tables 3–8). Additionally the 98^th percentile for the global QTc using Bazett’s correction in our study was 23 ms longer for males in the <1 month age category (Supplemental Table 4).

Discussion

This is the first large pediatric data set from North America of ECG parameters obtained from digital acquisition, stratified by age, sex and race, with echocardiogram confirmation of normal cardiac anatomy and exclusion of significant cardiac or systemic disease by medical record review. Our data, including interval and voltage parameters from digital 12 lead ECGs, indicate differences among age groups, between sex, and among race categories, as previously described in children and adults.^4–15 These data have important implications for interpretation of the ECG in children in the current era and they contrast existing pediatric public data sets.

Importantly, the QTc interval measurements from this PHN data set diverge from previous reports by expanding age stratification by sex and race.¹⁹ Our data indicate longer normal values for the QTc in girls age 3 years and older as well as clinically important racial differences in the QTc interval, where white children have longer values compared to African-American children. The PHN ECG data also suggests a broad range of QTc measurements in children less than 3 years of age, with the 98^th percentile extending out to 464 ms in lead II for children with race other than white or African-American (Table 2).

Further, the PHN Normal ECG data set diverges from prior reports for the R wave in V6. The most commonly referenced public ECG data set ⁴ is based on waveforms acquired using analog technology in 2141 children living within a narrow geographic region of Canada without stratification by sex or race. Therefore, it is not surprising that the parameters documented by Davignon contrast with subsequent reports, most notably the Dutch pediatric study by Rijnbeek of 1912 digital ECGs.^4,5 Comparison of Davignon⁴ values to the Dutch study reported by Rijnbeek⁵, indicates that the median R wave in V6 shifts from 1.5 mV to 2.0 mV and the 98^th percentile from 2.3 mV to 3.1 mV for an adolescent, a 30% difference in the maximal “normal” R wave. Our data suggest an even further shift in the 98^th percentile to 3.357 mV in males and 3.197 mV in African Americans (Table 3). Similar differences can be identified when modern proprietary reading algorithms or other geographic samples are included.^6–8

Our data support both a broader range for defining normal and the need to stratify norms by sex and race for ECGs in children. It is possible that body size could negate the effects of sex and race on the ECG in healthy children, but this hypothesis has yet to be tested. These findings have important clinical implications since ECG parameters are currently used by clinicians and other groups who screen youth for the risk of sudden cardiac death from disorders like long QT syndrome and hypertrophic cardiomyopathy and may guide diagnostic decisions (e.g. referral for echocardiogram to rule out left ventricular hypertrophy in a patient with an elevated R wave in V6 above the 98% for age). The PHN ECG data set suggests that the frequent high rate of false positive ECG screening results may be skewed toward specific sex and race categories.

Study Limitations

The main limitation of this study relates to the study cohort consisting of a convenience sample drawn from healthy children with normal echocardiograms enrolled in the PHN Echocardiogram Z-score study rather than a random sample of the healthy population. Since the indication for the echocardiogram was not collected as part of the PHN Echocardiogram Z-Score and Database Project, the cohort could over represent children with “abnormal” ECGs as the indication for the echocardiogram. However, our data did not indicate extreme outliers and all measured intervals were normally distributed. In addition, it is important to note that obese children were excluded from the PHN study; since it is well described that obesity may affect the ECG,^13,22,23 our findings are not generalizable to the large percentage of North American children who are obese but still considered healthy – 17% overall and up to 25% in some age group, sex and race categories.²⁴

This study was also inherently limited by its retrospective design. The study protocol required rigorous review of medical records and strict elimination of subjects with abnormal findings on any diagnostic study, but no records were reviewed after the study period to exclude subsequent abnormal findings. The ECGs included in this study were performed for clinical purposes and investigators did not confirm correct lead placement. In addition, the study did not collect information about concomitant medication use in enrolled healthy children including non-cardiac medications that may change the ECG including prolong the QT interval. As clinical researchers, we believe that the “gold standard” for setting normal values for any pediatric medical test should include selection of subjects at random prospectively from a healthy population and performance of the given test under supervised conditions with quality control. Therefore, because our study population was a convenience sample and because the study is retrospective, the aim of this study is descriptive.

Additionally, the National Institutes of Health definitions for race and ethnicity frequently differed from local definitions, leading to a widely diverse “other” race category. Indeed several enrolling centers recorded “Hispanic” as a race category, and these subjects could not be included in our study because this local practice did not align with NIH definitions of race and ethnicity; these potential subjects were not enrolled because they did not meet inclusion criteria. There is also a higher percentage of children with African-American race or Hispanic ethnicity than white race who are obese in the United States,²⁴ and this may have contributed to under enrollment in the African-American and other race categories. This study therefore focused primarily on whites and African-Americans, so our findings may be less applicable to children of other races.

Another potential limitation of this study is the PNH ECG Core Lab process whereby the pediatric cardiologists who reviewed the ECG waveforms were not blinded to the results of the automatic waveform interval and amplitude measurements performed by the AMPs ECG software. This lack of blinded, independent analysis may have biased the results.

Finally, this retrospective study included ECGs recorded by GE or Philips systems that collect waveforms using a bandwidth of 150 Hz and fail to comply with current recommendations of using a bandwidth of up to 250 Hz in children.² This shift would make it more likely to miss rapid, high frequency elements of the ECG and reduce the amplitudes of Q, R, and S waves, particularly in younger children with faster heart rates and rapid cardiac depolarization. However, given the current clinical practice in North America where most centers use ECG recording systems that do not comply with current AHA recommendations,² the PHN data are representative of contemporary real-world clinical practice.

Conclusion

In this large, diverse cohort of healthy children with normal echocardiograms from North America, numerous ECG intervals and amplitudes varied by sex and race.

In particular, the QTc interval in lead II and the R wave amplitude in V6 varied by sex and race and differed from previously published pediatric data sets and norms commonly used in practice. Relatively small differences in mean values, correspond with substantial and potentially clinically important differences in the 98^th percentile that represent the practical criteria for ECG diagnoses of prolonged QT or left ventricular enlargement. These differences have important implications for the interpretation of the pediatric ECG during diagnostic evaluations and screening for cardiac disease, and this dataset can be used to refine the criteria based on ECG intervals, maximum and/or minimum voltages, and aggregate voltages for triggering further cardiac testing in seemingly healthy children with benign histories and normal physical examinations. Future prospective studies evaluating digital ECGs in healthy children should include stratification not only by age but also sex and race to validate our findings and further refine measurements in the normal pediatric population.

What is Known

• Interpretation of pediatric electrocardiograms is limited by lack of sex- and race- specific normal reference values obtained with digital, modern technology for all ages.

What the Study Adds

• This dataset can be used to refine the criteria based on ECG intervals, maximum and/or minimum voltages, and aggregate voltages for triggering further cardiac testing in seemingly healthy children with benign histories and normal physical examinations.

• The PHN ECG dataset suggests that the frequent high rate of false positive ECG screening results may be skewed toward specific sex and race categories.

• QTc interval in lead II and the R wave amplitude in V6 varied by sex and race and differed from previously published pediatric data sets and norms commonly used in practice. Relatively small differences in mean values, correspond with substantial and potentially clinically important differences in the 98th percentile that represent the practical criteria for ECG diagnoses of prolonged QT or left ventricular enlargement.