Anda di halaman 1dari 9

A Review and Critique of the Statistical Methods Used to Generate Reference Values in Pediatric Echocardiography

Wadi Mawad, MD, Christian Drolet, MD, BSc, Nagib Dahdah, MD, and Frederic Dallaire, MD, PhD, Quebec, Montreal, and Sherbrooke, Quebec, Canada

Several articles have proposed echocardiographic reference values in normal pediatric subjects, but adequate validation is often lacking and has not been reviewed. The aim of this study was to review published reference values in pediatric two-dimensional and M-mode echocardiography with a specic focus on the adequacy of the statistical and mathematical methods used to normalize echocardiographic measurements. All articles proposing reference values for transthoracic pediatric echocardiography were reviewed. The types of measurements, the methods of normalization, the regression models used, and the methods used to detect potential bias in proposed reference values were abstracted. The detection of residual associations, residual heteroscedasticity, and departures from the normal distribution theory predictions were specically analyzed. Fifty-two studies met the inclusion criteria. Most authors (87%) used parametric normalization to account for body size, but their approaches were very heterogeneous. Linear regression and indexing were the most common models. Heteroscedasticity was often present but was mentioned in only 27% of studies. The absence of residual heteroscedasticity and residual associations between the normalized measurements and the independent variables were mentioned in only 9% and 22% of the studies, respectively. Only 14% of studies documented that the distribution of the residual values was appropriate for Z score calculation or that the proportion of subjects falling outside the reference range was appropriate. Statistical suitability of the proposed reference ranges was often incompletely documented. This review underlines the great need for better standardization in echocardiographic measurement normalization. (J Am Soc Echocardiogr 2013;26:29-37.) Keywords: Echocardiography, Reference values, Pediatric, Normalization, Z scores

Echocardiography is a reliable, noninvasive tool to evaluate heart structure and function in children and adults. Many important clinical decisions are routinely based on the absolute sizes of cardiac structures.1 Evaluation is highly dependent on the quality of the measurements but also on the quality of the reference values with which these measurements are compared. The American Society of Echocardiography Pediatric and Congenital Heart Disease Council recently published recommendations for quantication methods during the performance of pediatric echocardiography.2 However, reference values for the proposed methods often lack adequate validation. Unbiased reference values require appropriate normal subjects, standardized reproducible measurements, and appropriate sample sizes.3 In children, reference values are also highly dependent on
From the Division of Pediatric and Congenital Cardiology, Department of Pediatrics, Laval University Hospital, Faculty of Medicine, Laval University, Quebec City, Quebec, Canada (W.M., C.D., F.D.); the Division of Pediatric Cardiology, Sainte-Justine University Hospital, University of Montreal, Montreal, Quebec, Canada (N.D.); and the Division of Pediatric Cardiology, University Hospital of Sherbrooke, University of Sherbrooke, Sherbrooke, Quebec, Canada (F.D.). Reprint requests: Frederic Dallaire, MD, PhD, Division of Pediatric Cardiology, Department of Pediatrics, Faculty of Medicine, University of Sherbrooke, 3001, 12e Avenue Nord, Sherbrooke, QC J1H 5N4, Canada (E-mail: frederic.a.dallaire@ usherbrooke.ca). 0894-7317/$36.00 Copyright 2013 by the American Society of Echocardiography. http://dx.doi.org/10.1016/j.echo.2012.09.021

accurate adjustment for body size.1 Although nonparametric approaches have sometimes been used, parametric methods, such as Z scores, are now becoming the standard for body size adjustment in pediatric echocardiography.2,4,5 However, parametric methods rely on an appropriate distribution of the data, on the absence of residual associations, and on constant variance of the normalized measurements throughout the entire sample. These important requirements have not always received the attention they deserve. A recent review by Cantinotti et al.6 underlined several limitations of the available reference values in pediatric echocardiography, including a lack of standardization in data acquisition, a limited number of healthy subjects, and heterogeneous methods of normalizing and reporting reference values. However, their review did not specically address the statistical methods used or the potential pitfalls of parametric normalization. In this article, we present a systematic review of available reference values in two-dimensional (2D) and M-mode echocardiography in infants, children, and adolescents with a focus on the statistical validity of the methods used to generate the proposed reference ranges. For each reviewed article, we analyzed how the reference values were estimated, what type of normalization was used, and how the authors documented the detection of potential bias.

METHODS Literature Search Strategy A search of the National Library of Medicines PubMed database was performed using the Medical Subject Headings controlled 29

30 Mawad et al

Journal of the American Society of Echocardiography January 2013

vocabulary from the National Library of Medicine. The BSA = Body surface area search strategy was built to retrieve all articles containing 2D = Two-dimensional the Medical Subject Headings terms echocardiography and reference values or their equivalents: (reference values OR biometry OR anthropometry OR regression analysis) AND {echocardiography OR [ultrasonography AND (heart OR cardiovascular system)]}. We limited the search results to articles whose subjects were <18 years of age and that were available in English or French. References of selected articles were also reviewed for other, potentially missed relevant articles.
Abbreviations

tion on the distribution of the residual values and on the adequacy of the proportion of subjects falling outside the normal ranges across the entire range of the independent variable.

RESULTS Search Results and Selected Articles The search strategy returned 1,016 articles. Initial screening by title and abstract identied 117 potentially relevant articles. Sixty-ve were further excluded (27 did not propose reference values, ve were related to other echocardiographic modalities, 20 did not include pediatric data, ve did not include normal subjects, and eight were conducted before 1980), leaving 52 articles for analysis.7-58 The main characteristics of selected recent articles are summarized in Table 1 (2D studies) and Table 2 (M-mode studies). Reference values for almost all cardiac structures have been published, but the dimensions of the left ventricular outow tract, ascending aorta, and coronary arteries; M-mode measurements of left ventricular size; and estimations of left ventricular volume and mass were the most commonly measured structures. Three studies included reference ranges for $10 structures.7,16,41 Population Studied The populations studied ranged in age from infancy to early adulthood. Twelve studies included young adults up to 27 years of age.10,11,13,17-19,24,37,42,43,47 One study including adults of all ages was also included in our analysis because almost 75% of the subjects were children.39 Seven studies included only newborns,35,44,48,51,52,55,56 and three focused only on preterm infants.44,48,52 Most studies included strictly healthy subjects, but some did not exclude current or past history of a minor congenital heart defect such as a small atrial septal defect, a patent foramen ovale, or a small patent ductus arteriosus.28,41,48,52,53 One study included patients with histories of Kawasaki disease without documented coronary artery abnormalities.32 Twenty-one studies (39.6%) recruited subjects for research purposes only, and 18 studies (34.0%) relied on echocardiographic studies performed on clinical grounds but subsequently read as normal. The authors of the remaining 15 studies (28.3%) failed to specify why echocardiography was performed. Adjustment for Growth Parameters Seven studies (13.5%) did not attempt parametric normalization and presented reference ranges as percentile limits or means and standard deviations stratied by age or weight.7,13,23-25,48,50 The majority of authors (45 studies [86.8%]) used various methods of parametric normalization to account for growth. Sixteen studies (31.4%) considered only one independent variable (body surface area [BSA] in nine studies). The authors of the remaining studies tested two or more independent variables, the most common being BSA, weight, and height. BSA was used as the independent variable in 22 studies (42.3%). Weight was used in seven studies (mostly for infants), and height was used in eight studies (mostly for older children). Normalization using multivariate regression was performed in nine studies. The approach to parametric normalization was heterogeneous among the reviewed articles. Twenty of the studies (44.2%) using

Article Selection The search was performed in July 2011. Titles and abstracts were rst reviewed for identication of relevant studies. Full-text articles were reviewed if they matched the inclusion criteria or if they could not be condently excluded from the abstracts alone. We included studies proposing reference values for cardiac dimensions measured using 2D or M-mode imaging by transthoracic echocardiography in a normal healthy pediatric population. Articles were excluded when one or more of the following was applicable: inclusion of subjects with potential cardiopathies or other conditions that could alter cardiac dimensions, measures from imaging methods other than transthoracic 2D or M-mode echocardiography (fetal echocardiography, transesophageal echocardiography, etc), functional studies (ventricular function or strain, speckle tracking, tissue Doppler, etc), use of reference values from other studies, and publication before 1980. Thirteen studies focusing on reference values in pediatric echocardiography also included young adult subjects, and these were included in our analysis to better represent the upper range of growth. Data Abstraction and Analysis Each selected article was reviewed separately by two authors (W.M. and F.D.) using a standard data collection form. Discrepancies were resolved by consensus. Information on the study subjects (age, reason for echocardiography, inclusion and exclusion criteria), the cardiac structures measured, and the echocardiographic techniques used (views, mode, cardiac cycle, etc) was extracted. We also noted if interobserver or intraobserver variability was considered by the authors. We then thoroughly examined the method of adjustment for body size and reviewed the type and number of independent variables used. For parametric normalization, we extracted information on the type of regression, the mathematical transformation of the independent or the dependent variables, and how the authors justied their choice of regression strategy. Because most cardiac structures display inconstant variance across most growth variables (heteroscedasticity), we noted if heteroscedasticity was assessed, by what method (visual assessment of plots or statistical tests), and, if present, how its effect was taken into account in the regression models. Finally, we extracted information on how the authors assessed their proposed reference values for the absence of bias. For any parametric normalization, we noted if the authors conrmed the absence of residual associations with the selected independent variable and the absence of residual heteroscedasticity. We also extracted informa-

Journal of the American Society of Echocardiography Volume 26 Number 1

Table 1 Summary of 2D studies published after 1990


Was the detection of potential biases documented? Testing for the distribution of residual or normalized values

Study

Age

Structures

Model and type of normalization

Testing for heteroscedasticity

Testing for residual association

Testing for residual heteroscedasticity

Lytrivi et al. (2011)28 Dallaire and Dahdah (2011)8 Gautier et al. (2010)14

100 1,033 353

0 to 3 y 2 mo to 18 y

LVEDV (bullet shape formulas) Coronary arteries

Indexed to BSA1.38 (allometric model) Linear with square root of BSA Log-log with BSA

No Yes No

Yes Yes No

No Yes No

No Yes Percentage of subjects with Z > 2 No Yes No

Mean, AoV, sinus, STJ, 12 6 4.5 y AscAo 0 to 20 y 1 to 17 y 0 to 18 y Coronary arteries LA 14 structures by 2D imaging, 7 structures by M-mode imaging AoV, aortic sinus, STJ RV volume

Olivieri et al. (2009)37 Neilan et al. (2009)33 Pettersen et al. (2008)41

432 4,109 782

Kaldararova et al. (2007)24 Suleymanoglu et al. (2007)50

702 213

0 to 20 y 0 to 15 y

Poutanen et al. (2006)43 Zilberman et al. (2005)58 Makan et al. (2005)29 Poutanen et al. (2003)42 Tan et al. (2003)53 Joyce et al. (2001)22 Tan et al. (2001)52 Tacy et al. (1995)51 Sheil et al. (1995)47 Nidorf et al. (1992)34 Domanski et al. (1991)12 Pearlman et al. (1990)39

168

2 to 27 y

Mitral and aortic area AoV, PV, TV, MV annuli LV dimensions and mass, LA, aortic root AoV, sinus of Valsalva, STJ, aortic arch Coronary arteries RV free wall mass AoV, PV, PAs AoV, PV, MV, TV LVOT, AoV, sinus of Valsalva, STJ, AscAo AoV, LA, LVEDD, LV length LVEDD, LA LA

748 0 to 18 y 250 controls, 14 to 18 y 900 elite athletes 168 2 to 27 y 390 44 62 70 48 196 10 196 children, 72 adults 2 mo to 8 y 0 to 17 y Preterm infants 0 to 10 d 0 to 23 y 6 d to 18 y 4 to 12 y 6 d to 18 y (children)

Log-log model with BSA Log-log model with BSA or weight Third-order polynomial with log-transformed dependent variable Nonparametric Nonparametric normalization; authors report 5th and 95th percentiles for several independent variables Linear with BSA (range presented as 5th and 95th percentiles) Log-log model with BSA No normalization, mean 6 2SDs according to age Linear with BSA (range presented as 5th and 95th percentiles) Linear (results presented with 5 different independent variables Indexed to BSA Linear with weight Linear with weight Linear with height and indexed to AoV Indexed to height Indexed to LVOT dimension Power model with BSA

Yes Yes Unclear

No Yes No

No Yes No

NA NA

NA NA

NA NA

NA NA

NA No No

No No NA

NA No NA

No No NA

NA No No No No No No No Yes

No No No No No No Yes Yes No

NA No No No No No No No No

No No No No No No No

Mawad et al 31

No No

AoV, Aortic valve; AscAo, ascending aorta; LA, left atrium; LV, left ventricular; LVEDD, left ventricular end-diastolic diameter; LVEDV, left ventricular end-diastolic volume; LVOT, left ventricular outow tract; MV, mitral valve; NA, not available; PA, pulmonary artery; PV, pulmonary valve; RV, right ventricular; STJ, sinotubular junction; TV, tricuspid valve.

32 Mawad et al

Table 2 Summary of M-mode studies published after 1990


Was the detection of potential biases documented? Testing for the distribution of residual Testing Testing or normalized for residual for residual Testing for values heteroscedasticity association heteroscedasticity

Study

Age

Structures

Type of normalization

Nagasawa (2010)31 Foster et al. (2008)13 Bonatto et al. (2006)7 Overbeek et al. (2006)38 Kervancioglu et al. (2006)25 Huicho et al (2005)18

243 440 595 747 229

Linear with height (with 2 inexion points) 0 to 21 y LVM Nonparametric. lambda-mu-sigma method 1 mo to 12 y 8 structures by M-mode Nonparametric (centile curves with imaging and LVM BSA) 0 to 18 y Left ventricular Log-log model with weight dimensions 0 to 15 y Aortic sinus Nonparametric Linear with BSA

0 to 12 mo

LVEDD

No NA NA Yes NA No

No NA NA Yes NA No

No NA NA Yes NA No

No NA NA No NA No

Joyce et al. (2004)23 Skelton et al. (1998)48

Nagasawa et al. (1996)32 Daniels et al. (1995)9 Huwez et al. (1994)19 Malcolm et al. (1993)30 de Simone et al. (1992)11 Gupta and Jain (1991)15

321 (subjects 2 mo to 19 y 10 structures living at high altitude) 45 0 to 4 mo Left and right ventricular volumes, masses, wall thicknesses 79 Newborns LV, Ao, LA including preterm infants 437 1 mo to 17 y LVEDD 192 127 904 444 183 6 to 17 y LV 7 mo to 19 y LV, aortic root, LA, RV 6 to 16 y LVM 4 mo to 23 y LVM 3 to 12 y LV

No normalization, mean 6 SD according to age Nonparametric, mean and SD for strata of GA and BW

NA

NA

NA

NA

NA

NA

NA

NA

Journal of the American Society of Echocardiography January 2013

Linear with height Indexed to height3 (allometric model) Linear with age and linear with BSA Multiple regression with height, age, and height age interaction Indexed to height2.7 (allometric model) Indexed to height

No No No No No No

No Yes No No Yes No

No No No No No No

No No No No Yes No

Ao, Aorta; AoV, aortic valve; BW, birth weight; GA, gestational age; LA, left atrium; LV, left ventricle; LVEDD, left ventricular end-diastolic diameter; LVM, left ventricular mass; NA, not available; RV, right ventricle.

Journal of the American Society of Echocardiography Volume 26 Number 1

Mawad et al 33

parametric normalization tested one regression model only. Of them, eight articles reported linear regression only, six reported indexing (including two using allometric indexing), two reported power or logarithmic models only, and three used multivariate regression but tested only one model. In the remaining reports, two or more models were assessed, usually linear, polynomial, power, and log-log models. Pearsons correlation coefcient (R2) was often used to assess the goodness of t of the regressions. Various mathematical transformations of both the dependent and the independent variables were commonly used to deal with the nonlinearity of most echocardiographic measurements with growth parameters. Logarithmic transformation of either the dependent or the independent variable, or both, was the most common strategy to adjust for nonlinearity, followed by the square and cube roots of the independent variable. Unequal Variance and Heteroscedasticity Heteroscedasticity is dened as the inconstant variance of the dependent variable across the entire range of the independent variable. In parametric normalization, one uses the magnitude of dispersion around the mean to establish reference ranges. Such ranges are thus dependent on any variation of that dispersion. Heteroscedasticity was mentioned in the text of 12 of the 45 studies in which some form of parametric normalization was done. However, there were four studies in which heterogeneous variance seemed to have been taken into account, although we could not nd any mention of it in the text. The most common methods to account for increasing variance were logarithmic transformation of the dependent variables (seven studies) and weighted regression models (seven studies). Types of Reference Values The reference values proposed by the selected studies were expressed in a variety of ways. Z score equations were suggested in eight studies. The remaining studies proposed graphs (14 studies), tables (two studies), equations to derive the upper and lower ranges of normal (12 studies), indexes (10 studies), means and standard deviations (four studies), or other methods (two studies). Evaluation of Bias Only 14 studies using parametric normalization mentioned any type of further evaluation of the proposed reference values. The authors of 10 studies assessed if residual associations between the normalized measurements and the chosen independent variables were still present. The authors of six studies stated that the distribution of the dependent variable was appropriate for Z score calculation or that the proportion of subjects falling outside the reference range was appropriate. Testing for absence of residual heteroscedasticity was mentioned in only four studies. Finally, intraobserver and interobserver variability was evaluated in 13 and 19 studies, respectively.

DISCUSSION In pediatric and congenital cardiology, many clinical, interventional, and surgical decisions are based on the sizes of cardiac structures. Moreover, follow-up of children with repaired and unrepaired cardiac defects often depends on the identication of structural growth that deviates from expected. The development of reliable and validated

reference values in transthoracic echocardiography is thus of great importance, because a biased or inaccurate cardiac growth curve could lead to inappropriate clinical and surgical management. In this review, we identied multiple articles recommending reference ranges for most cardiac structures. However, information on the detection of potential biases was absent from more than half of them, and few authors seemed to have gone beyond the goodness of t of their regressions to assess the quality and validity of the reference ranges they proposed.8,11,13,33,38,40,59 Furthermore, although most cardiac structures displayed clear nonlinearity with weight, height, BSA, or age, many authors put forward simple linear models or indexes. Nonparametric methods do not assume that the response variable adopts a given distribution and are therefore less prone to bias. However, because echocardiographic reference ranges change as the body grows, if one wants to dene precise reference ranges across growth using nonparametric methods, one must compute percentiles for several growth strata of the studied population. Because each stratum must include a sufcient number of subjects to estimate reliable percentiles,60 and because of the large number of strata needed to generate precise percentile curves from birth to early adulthood, nonparametric methods are rarely used in pediatric echocardiography. Bonatto et al.7 used this approach, and although their study included almost 600 subjects, some strata had only 20 subjects to compute reference ranges. Consequently, most of the reviewed studies relied on parametric normalization to dene reference ranges. In parametric normalization, one uses the known distribution parameters of a population to estimate reference values according to one or more independent growth variables. The response variable is modeled mathematically on the independent variable, and the results of that regression yield the predicted mean of the response variable according to the independent variable. Several regression approaches exist. It has been advocated that the choice of a regression model should be in accordance with what is expected physiologically.61 For example, Sluysmans and Colan59 estimated the optimal vessel dimensions that would minimize ow-mediated energy loss and then showed that cardiac structure size predictions were in agreement with their theoretical model. Whereas a regression approach that is in accordance to what is predicted by physiologic principles is likely to be superior to an empirical model, several steps are needed to ensure that no important biases were introduced by the modeling of the response variable, regardless of the regression approach used. First, the regression model should be chosen so that the t is adequate across the entire population. A signicant lack of t could result in a predicted mean value higher or lower than the true mean for certain strata of the population. Figure 1 provides an example of two different models from a sample of sinus of Valsalva diameters in children (see the legend for details on the method). In Figure 1A, a poorly tted linear model for sinus of Valsalva diameter against BSA is shown. Figure 1B shows a visually more adequate t using a gamma function model (Y = aBSAbexp[lBSA]) proposed by Nevill et al.61 The quality of the t should be evaluated statistically but should also be carefully inspected visually using plots of the dependent on the independent variables and plots of the residual values (or Z scores) on the independent variable. Adequate t should result in no signicant residual associations between the residual values and the independent variable. Figure 2 shows the Z scores according to BSA for the two models from Figure 1. Note the strong residual association in Figure 2A (red curve) and the absence of a residual association in Figure 2B. The current review identied only 12 studies that

34 Mawad et al

Journal of the American Society of Echocardiography January 2013

Figure 1 Sinus of Valsalva diameter according to BSA. The dashed curves represent the predicted mean and the solid curves the Z = +2 and Z = 2 limits. (A) Linear model. The predicted mean is poorly tted, especially for smaller patients. The parallel Z score boundaries do not capture the clear heteroscedasticity. (B) Gamma function weighted model. The predicted mean displays a more adequate t, and the weighted model allows Z score limits to follow the increasing variance with body surface area. The sinus of Valsalva diameters were extracted from the Sainte-Justine University Hospital database (Montreal, QC, Canada). Studies were performed on children ranging from 1 day to 17 years of age who were referred for murmurs, syncope, or chest pain from May 2001 to May 2003. Patients above or below 2 standard deviations from the mean body mass index for age were excluded.

documented assessment of the residual association. It should be noted that having a good R2 value does not rule out residual association. In our example, the linear model had an R2 value of 0.80, which could be considered adequate, although visually it was obviously poorly tted. Second, the distribution of the residual values should be inspected. When Z scores are computed, or when the standard deviation is used to estimate percentiles, residual values must be normally distributed. Importantly, they must be so across the entire range of the independent variable. To better detect departure from the normal distribution, it was previously suggested that the residual values should be divided into at least three equal groups according to the independent variable.62 The distribution of each of these groups should then be assessed. Any departure from the normal distribution in any stratum of the population studied could lead to biased reference values. Moreover, when Z scores are computed, each stratum must adopt a normal distribution with a mean of 0 and a standard deviation of 1. In Figure 2A, small children in the rst tertile of BSA had Z scores with a distribution close to normal, but the mean was signicantly smaller than zero, which indicates a high likelihood of Z score underestimation. This review identied only four studies documenting assessment of response variable distribution. Third, heteroscedasticity should be assessed, and when it is present, a model taking it into account should be used. In the example from Figure 1, clear heteroscedasticity was present. A weighted model was used only in Figure 1B, and the reader can appreciate that the distance between the Z = 2 and Z = 2 limits increases with BSA (represented by the blue lines). Adequate heteroscedasticity management should result in no signicant residual heteroscedasticity, which should also be veried thoroughly. Although no consensus exists on how to detect residual heteroscedasticity, the presence of a statistically signicant slope between the absolute residual values and the dependent variable and a statistical test aimed at the detection of heteroscedasticity (the White test or the Breusch-Pagan test) yielding a low P value both indicate that resid-

ual heteroscedasticity is likely present. Residual heteroscedasticity could lead to underestimated or overestimated variance for some strata of the population, which in turn could bias the reference values or Z scores. Clear residual heteroscedasticity was present in Figure 2A. Although the authors of nearly half of the studies recognized and corrected for unequal variance, only four studies documented assessment of residual heteroscedasticity. Logarithmic transformation of the dependent variable was used by many authors to adjust for nonlinearity and heteroscedasticity. When logarithmic transformation of the dependent variable is used, the assumption is usually that the dependent variable has a log-normal distribution so that the regression of the logtransformed value will produce residual values with a normal distribution.63 It should also be noted that logarithmic transformation may mask potentially strong outliers.63 Although logarithmic transformation was used by the authors of 13 studies, assessment of adequate distribution of the residual values was noted in only two studies. Our previous observations led to the conclusion that many echocardiographic measurements were normally distributed at any given stratum of growth, and when logarithmic transformation was used, parametric normalization failed to produce normally distributed residual values, which could introduce bias (unpublished results and Dallaire and Dahdah8). Finally, whatever the method used to estimate reference values with parametric methods on the basis of the normal distribution, one should always ensure that the proportion of individuals falling outside the reference range does not deviate from what is predicted by the normal distribution theory. For example, in a normal population, 2.28% of the population will have Z scores $ 2.0. A signicant difference between the predicted and observed proportions (i.e., Z score > 2 in >2.28% of the subjects) in any stratum of the population studied is a strong indicator that bias is present. Such verication was described in very few of the reviewed studies. In our example, the proportion of small patients falling below Z = 2 in the linear model (Figure 2A) was 9.7%, well above the predicted 2.28%.

Journal of the American Society of Echocardiography Volume 26 Number 1

Mawad et al 35

Figure 2 Relation between Z scores and BSA computed with the models from Figure 1. (A) Z scores computed with the unweighted linear model. (B) Z scores computed with the gamma function weighted model. Dashed curves represent residual association with the Z scores and BSA. Adequate Z scores should be evenly distributed around 0, with 95.4% of the population within the red boundaries for all BSA strata. In its recent recommendations, the American Society of Echocardiography advocated that when parametric normalization is done, reference values should be expressed as Z scores.2 Z scores are superior to dichotomous normal values because they allow clinicians to appreciate the magnitude of abnormality. Z score estimates are now part of the daily decision making in clinical and surgical management in pediatric cardiology.64,65 When adequately validated Z score equations are available, their inclusion in simple computer software renders the interpretation of cardiac structure size and growth simple and intuitive. An ascending aorta with a Z score diameter increasing from 1.8 to +1.8, although within the normal range, is not following its expected normal growth curve. This can be easily appreciated by a clinician without having to refer to cumbersome normal value tables. Furthermore, although the Z score is used to estimate percentiles, extreme Z scores are easier to interpret than percentiles; for example, a Z score of 3 corresponds to the 99.865th percentile, while a Z score of 4 corresponds to the 99.997th percentile. Nine studies proposed indexes as a way to normalize measurements. Likely because of their simplicity, indexes have often been used to normalize echocardiographic and hemodynamic measurements. This simplicity comes at a signicant cost. Indexes are prone to the same biases as any other parametric normalization. To be valid, they also need to meet very stringent criteria: perfect linear correlation, a zero intercept, and absence of heteroscedasticity. These criteria are almost never present in pediatric echocardiographic measurements, and previous studies have repeatedly shown that simple linear indexing introduces bias.8,13,33,66 In our opinion, such a method for normalization in pediatric echocardiography should be abandoned. In structurally normal hearts, the sizes of cardiac structures are a function of the cardiac output. The linear relationship between BSA and cardiac output has been widely recognized and has led many authors to use BSA to normalize echocardiographic measurements.67 Surprisingly, the vast majority used the formula of Du Bois and Du Bois68 to estimate BSA, although it has been shown to underestimate BSA in young children.69,70 None of the available equations for calculating BSA are perfect, and it is likely that the sizes of cardiac structures are a function of both weight and height and that each of them affects cardiac output in different proportions as children develop and grow. Indeed, it has previously been shown that the relationship between cardiac output and growth parameters changes as children grow.71 Several articles reviewed in this study showed relatively linear relationships between cardiac structure sizes and various transformations of BSA. However, because BSA does not capture differences in body composition (fat/muscle proportion), it remains an imperfect surrogate of cardiac output. Normalization across a wide range of body sizes has many practical benets, but more subtle effects in the extremes of the pediatric ages could be masked by the search of a single model to describe the complex mechanisms of heart growth. The extremes of the pediatric age range, particularly newborns and infants, should be studied separately to ensure adequately validated reference ranges, especially because many crucial interventional and surgical decisions are made very early in life. Other factors also complicate the anthropometric equations such as gender, obesity, and physical tness. In 2001, Lipshultz et al.72 showed that systematic biases among laboratories existed for some measurements of left ventricular dimension. Such biases are likely present for other structures as well, and their magnitudes probably relate to the technical difculty of the measurement. Systematic error, whether among laboratories or among observers, will greatly affect reference values. This should be kept in mind when a laboratory uses reference ranges derived at another institution. The change in Z score over time for a specic measurement in a given individual will, however, be less affected by systematic bias. There is a great need for adequately validated multicenter reference values derived from large populations of healthy children. However, we recommend that validation be performed on local controls to ensure that a systemic bias does not lead to an underestimation or overestimation of the normalized measurements.

CONCLUSIONS Choice of population, technical standardization of echocardiographic measurements, and detailed strategies for model selection were outside the scope of this review, as they were recently addressed elsewhere.6,59,61 The recent recommendations of the American Society of Echocardiography on quantication methods in pediatric echocardiography concluded that standardizing quantication

36 Mawad et al

Journal of the American Society of Echocardiography January 2013

methods is the rst step in the task of generating a normative database that encompasses the range of body sizes and ages encountered in the pediatric population.2 This review underlines that there is also a great need for a more thorough approach in the detection of bias in parametric normalization. Of course, incomplete validation does not necessarily mean that a given set of reference values is biased. However, if any set of reference values is expected to be used routinely with condence, its authors must provide to readers and clinicians adequate proof that signicant bias is not present.

REFERENCES
1. Lipshultz SE, Miller TL. Establishing norms for echocardiographic measurements of cardiovascular structures and function in children. J Appl Physiol 2005;99:386-8. 2. Lopez L, Colan SD, Frommelt PC, Ensing GJ, Kendall K, Younoszai AK, et al. Recommendations for quantication methods during the performance of a pediatric echocardiogram: a report from the Pediatric Measurements Writing Group of the American Society of Echocardiography Pediatric and Congenital Heart Disease Council. J Am Soc Echocardiogr 2010;23:465-95. 3. Vasan RS, Levy D, Larson MG, Benjamin EJ. Interpretation of echocardiographic measurements: a call for standardization. Am Heart J 2000;139: 412-22. 4. Sluysmans T, Colan SD. Structural Measurements and adjustment for growth. In: Lai WW, Mertens LL, Cohen SC, et al., editors. Echocardiography in pediatric and congenital heart disease. West Sussex, United Kingdom: Wiley-Blackwell; 2009. pp. 52-62. 5. Kaski JP, Daubeney PE. Normalization of echocardiographically derived paediatric cardiac dimensions to body surface area: time for a standardized approach. Eur J Echocardiogr 2009;10:44-5. 6. Cantinotti M, Scalese M, Molinaro S, Murzi B, Passino C. Limitations of current echocardiographic nomograms for left ventricular, valvular and arterial dimensions in children: a critical review. J Am Soc Echocardiogr 2012;25:142-52. 7. Bonatto RC, Fioretto JR, Okoshi K, Matsubara BB, Padovani CR, Manfrin TC, et al. Percentile curves of normal values of echocardiographic measurements in normal children from the central-southern region of the State of Sao Paulo, Brazil [article in English and Portugese]. Arq Bras Cardiol 2006;87:711-21. 8. Dallaire F, Dahdah N. New equations and a critical appraisal of coronary artery Z scores in healthy children. J Am Soc Echocardiogr 2011;24:60-74. 9. Daniels SR, Kimball TR, Morrison JA, Khoury P, Meyer RA. Indexing left ventricular mass to account for differences in body size in children and adolescents without cardiovascular disease. Am J Cardiol 1995;76:699-701. 10. Daniels SR, Meyer RA, Liang YC, Bove KE. Echocardiographically determined left ventricular mass index in normal children, adolescents and young adults. J Am Coll Cardiol 1988;12:703-8. 11. de Simone G, Daniels SR, Devereux RB, Meyer RA, Roman MJ, de Divitiis O, et al. Left ventricular mass and body size in normotensive children and adults: assessment of allometric relations and impact of overweight. J Am Coll Cardiol 1992;20:1251-60. 12. Domanski MJ, Cunnion RE, Roberts WC. Usefulness of the subaortic diameter for normalizing left ventricular and left atrial dimensions. Am J Cardiol 1991;67:785-6. 13. Foster BJ, Mackie AS, Mitsnefes M, Ali H, Mamber S, Colan SD. A novel method of expressing left ventricular mass relative to body size in children. Circulation 2008;117:2769-75. 14. Gautier M, Detaint D, Fermanian C, Aegerter P, Delorme G, Arnoult F, et al. Nomograms for aortic root diameters in children using twodimensional echocardiography. Am J Cardiol 2010;105:888-94. 15. Gupta R, Jain BK. Norms and clinical correlates of echocardiographic left ventricular mass in 3-12 year old children. Indian Heart J 1991; 43:445-7.

16. Hanseus K, Bjorkhem G, Lundstrom NR. Dimensions of cardiac chambers and great vessels by cross-sectional echocardiography in infants and children. Pediatr Cardiol 1988;9:7-15. 17. Henry WL, Gardin JM, Ware JH. Echocardiographic measurements in normal subjects from infancy to old age. Circulation 1980;62:1054-61. 18. Huicho L, Muro M, Pacheco A, Silva J, Gloria E, Marticorena E, et al. Crosssectional study of echocardiographic characteristics in healthy children living at high altitude. Am J Hum Biol 2005;17:704-17. 19. Huwez FU, Houston AB, Watson J, McLaughlin S, Macfarlane PW. Age and body surface area related normal upper and lower limits of M mode echocardiographic measurements and left ventricular volume and mass from infancy to early adulthood. Br Heart J 1994;72:276-80. 20. Ichida F, Aubert A, Denef B, Dumoulin M, Van der Hauwaert LG. Cross sectional echocardiographic assessment of great artery diameters in infants and children. Br Heart J 1987;58:627-34. 21. Ichida F, Denef B, Dumoulin M, Van der Hauwaert LG. Cardiac chamber growth pattern determined by two-dimensional echocardiography. Heart Vessels 1988;4:26-33. 22. Joyce JJ, Denslow S, Kline CH, Baylen BG, Wiles HB. Estimation of right ventricular free-wall mass using two-dimensional echocardiography. Pediatr Cardiol 2001;22:306-14. 23. Joyce JJ, Dickson PI, Qi N, Noble JE, Raj JU, Baylen BG. Normal right and left ventricular mass development during early infancy. Am J Cardiol 2004;93:797-801. 24. Kaldararova M, Balazova E, Tittel P, Stankovicova I, Brucknerova I, Masura J. Echocardiographic measurements of the aorta in normal children and young adults. Bratisl Lek Listy 2007;108:437-41. 25. Kervancioglu P, Kervancioglu M, Tuncer CM. Echocardiographic study of aortic root diameter in healthy children. Saudi Med J 2006;27: 27-30. 26. King DH, Smith EO, Huhta JC, Gutgesell HP. Mitral and tricuspid valve anular diameter in normal children determined by two-dimensional echocardiography. Am J Cardiol 1985;55:787-9. 27. Lappen RS, Riggs TW, Lapin GD, Paul MH, Muster AJ. Two-dimensional echocardiographic measurement of right pulmonary artery diameter in infants and children. J Am Coll Cardiol 1983;2:121-6. 28. Lytrivi ID, Bhatla P, Ko HH, Yau J, Geiger MK, Walsh R, et al. Normal values for left ventricular volume in infants and young children by the echocardiographic subxiphoid ve-sixth area by length (bullet) method. J Am Soc Echocardiogr 2011;24:214-8. 29. Makan J, Sharma S, Firoozi S, Whyte G, Jackson PG, McKenna WJ. Physiological upper limits of ventricular cavity size in highly trained adolescent athletes. Heart 2005;91:495-9. 30. Malcolm DD, Burns TL, Mahoney LT, Lauer RM. Factors affecting left ventricular mass in childhood: the Muscatine Study. Pediatrics 1993;92: 703-9. 31. Nagasawa H. Novel regression equations of left ventricular dimensions in infants less than 1 year of age and premature neonates obtained from echocardiographic examination. Cardiol Young 2010;20:526-31. 32. Nagasawa H, Arakaki Y, Yamada O, Nakajima T, Kamiya T. Longitudinal observations of left ventricular end-diastolic dimension in children using echocardiography. Pediatr Cardiol 1996;17:169-74. 33. Neilan TG, Pradhan AD, King ME, Weyman AE. Derivation of a sizeindependent variable for scaling of cardiac dimensions in a normal paediatric population. Eur J Echocardiogr 2009;10:50-5. 34. Nidorf SM, Picard MH, Triulzi MO, Thomas JD, Newell J, King ME, et al. New perspectives in the assessment of cardiac chamber dimensions during development and adulthood. J Am Coll Cardiol 1992; 19:983-8. 35. Oberhansli I, Brandon G, Friedli B. Echocardiographic growth patterns of intracardiac dimensions and determination of function indices during the rst year of life. Helv Paediatr Acta 1981;36:325-40. 36. Oberhoffer R, Lang D, Feilen K. The diameter of coronary arteries in infants and children without heart disease. Eur J Pediatr 1989;148: 389-92. 37. Olivieri L, Arling B, Friberg M, Sable C. Coronary artery Z score regression equations and calculators derived from a large heterogeneous population

Journal of the American Society of Echocardiography Volume 26 Number 1

Mawad et al 37

38.

39.

40.

41.

42.

43.

44. 45.

46.

47.

48. 49.

50.

51. 52. 53.

of children undergoing echocardiography. J Am Soc Echocardiogr 2009; 22:159-64. Overbeek LI, Kapusta L, Peer PG, de Korte CL, Thijssen JM, Daniels O. New reference values for echocardiographic dimensions of healthy Dutch children. Eur J Echocardiogr 2006;7:113-21. Pearlman JD, Triulzi MO, King ME, Abascal VM, Newell J, Weyman AE. Left atrial dimensions in growth and development: normal limits for two-dimensional echocardiography. J Am Coll Cardiol 1990;16: 1168-74. Pearlman JD, Triulzi MO, King ME, Newell J, Weyman AE. Limits of normal left ventricular dimensions in growth and development: analysis of dimensions and variance in the two-dimensional echocardiograms of 268 normal healthy subjects. J Am Coll Cardiol 1988;12:1432-41. Pettersen MD, Du W, Skeens ME, Humes RA. Regression equations for calculation of Z scores of cardiac structures in a large cohort of healthy infants, children, and adolescents: an echocardiographic study. J Am Soc Echocardiogr 2008;21:922-34. Poutanen T, Tikanoja T, Sairanen H, Jokinen E. Normal aortic dimensions and ow in 168 children and young adults. Clin Physiol Funct Imaging 2003;23:224-9. Poutanen T, Tikanoja T, Sairanen H, Jokinen E. Normal mitral and aortic valve areas assessed by three- and two-dimensional echocardiography in 168 children and young adults. Pediatr Cardiol 2006;27:217-25. Reller MD, Meyer RA, Kaplan S. Normal aortic root dimensions in premature infants. J Clin Ultrasound 1983;11:203-5. Riggs TW, Lapin GD, Paul MH, Muster AJ, Berry TE. Measurement of mitral valve orice area in infants and children by two-dimensional echocardiography. J Am Coll Cardiol 1983;1:873-8. Roman MJ, Devereux RB, Kramer-Fox R, OLoughlin J. Two-dimensional echocardiographic aortic root dimensions in normal children and adults. Am J Cardiol 1989;64:507-12. Sheil ML, Jenkins O, Sholler GF. Echocardiographic assessment of aortic root dimensions in normal children based on measurement of a new ratio of aortic size independent of growth. Am J Cardiol 1995;75:711-5. Skelton R, Gill AB, Parsons JM. Reference ranges for cardiac dimensions and blood ow velocity in preterm infants. Heart 1998;80:281-5. Snider AR, Enderlein MA, Teitel DF, Juster RP. Two-dimensional echocardiographic determination of aortic and pulmonary artery sizes from infancy to adulthood in normal subjects. Am J Cardiol 1984;53:218-24. Suleymanoglu S, Okutan V, Yozgat Y, Lenk MK. Determination of normal echocardiographic values for right ventricular volume in children with two-dimensional transthoracic echocardiography. Turk J Pediatr 2007; 49:141-7. Tacy TA, Vermilion RP, Ludomirsky A. Range of normal valve annulus size in neonates. Am J Cardiol 1995;75:541-3. Tan TH, Heng JT, Wong KY. Pulmonary artery diameters in premature infants: normal ranges. Singapore Med J 2001;42:102-6. Tan TH, Wong KY, Cheng TK, Heng JT. Coronary normograms and the coronary-aorta index: objective determinants of coronary artery dilatation. Pediatr Cardiol 2003;24:328-35.

54. Voogd PJ, Rijsterborgh H, Lubsen J, Arntzenius AC, Monsjou LK, Godijn EH. Reference ranges of echocardiographic measurements in the Dutch population. Eur Heart J 1984;5:762-70. 55. Walther FJ, Siassi B, King J, Wu PY. Normal values of aortic root measurements in neonates. Pediatr Cardiol 1985;6:61-3. 56. Walther FJ, Siassi B, King J, Wu PY. Echocardiographic measurements in normal preterm and term neonates. Acta Paediatr Scand 1986;75:563-8. 57. Wessel A. Normal values of two-dimensional echocardiographic evaluation of left and right ventricular geometry in children. Herz 1985;10:248-54. 58. Zilberman MV, Khoury PR, Kimball RT. Two-dimensional echocardiographic valve measurements in healthy children: gender-specic differences. Pediatr Cardiol 2005;26:356-60. 59. Sluysmans T, Colan SD. Theoretical and empirical derivation of cardiovascular allometric relationships in children. J Appl Physiol 2005;99:445-57. 60. Solberg HE. The IFCC recommendation on estimation of reference intervals. The RefVal program. Clin Chem Lab Med 2004;42:710-4. 61. Nevill AM, Bate S, Holder RL. Modeling physiological and anthropometric variables known to vary with body size and other confounding variables. Am J Phys Anthropol 2005;(Suppl 41):141-53. 62. Royston P. Constructing time-specic reference ranges. Stat Med 1991;10: 675-90. 63. Packard GC, Boardman TJ. Model selection and logarithmic transformation in allometric analysis. Physiol Biochem Zool 2008;81:496-507. 64. Loand GK, McCrindle BW, Williams WG, Blackstone EH, Tchervenkov CI, Sittiwangkul R, et al. Congenital Heart Surgeons Society. Critical aortic stenosis in the neonate: a multi-institutional study of management, outcomes, and risk factors. J Thorac Cardiovasc Surg 2001;121:10-27. 65. Manlhiot C, Millar K, Golding F, McCrindle BW. Improved classication of coronary artery abnormalities based only on coronary artery Z-scores after Kawasaki disease. Pediatr Cardiol 2010;31:242-9. 66. Gutgesell HP, Rembold CM. Growth of the human heart relative to body surface area. Am J Cardiol 1990;65:662-8. 67. Grollman A. Physiologic variations in the cardiac output in man. Am J Physiol 1929;90:210-7. 68. Du Bois D, Du Bois EF. Clinical calorimetry: tenth papera formula to estimate the approximate surface area if height and weight be known. Arch Intern Med 1916;17:863-71. 69. Ahn Y, Garruto RM. Estimations of body surface area in newborns. Acta Paediatr 2008;97:366-70. 70. Haycock GB, Schwartz GJ, Wisotsky DH. Geometric method for measuring body surface area: a height-weight formula validated in infants, children, and adults. J Pediatr 1978;93:62-6. 71. de Simone G, Devereux RB, Daniels SR, Mureddu G, Roman MJ, Kimball TR, et al. Stroke volume and cardiac output in normotensive children and adults. Assessment of relations with body size and impact of overweight. Circulation 1997;95:1837-43. 72. Lipshultz SE, Easley KA, Orav EJ, Kaplan S, Starc TJ, Bricker JT, et al. Reliability of multicenter pediatric echocardiographic measurements of left ventricular structure and function: the prospective P(2)C(2) HIV study. Circulation 2001;104:310-6.