Search Strategy/Search Protocol

L.H.Y., A.H., and S.Y., the medical librarians at Washington University School of Medicine (St. Louis, Missouri), trained in systematic reviews, created search hedges to cover the concepts of pediatrics and children, speckle-tracking echocardiography, and the left-heart ventricle using terms harvested from standard term indices and on-topic articles ( Appendix 1 ). To exclude animals, L.H.Y. used the human filter for PubMed recommended in the Cochrane Handbook for Systematic Reviews of Interventions and then used that as a model created by S.Y. to create similar filters for the other searched databases. The search strategy was launched in MEDLINE, Embase, Scopus, the Cumulative Index of Nursing and Allied Health Literature, the Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov . Searches were completed by November 2015. The reference lists of all selected reports were screened to identify additional studies.

Study Selection/Eligibility Criteria

Studies were included if they reported using strain derived by 2DSTE to measure LV function in healthy pediatric normal or control subjects. Studies that exclusively included children <21 years of age were considered eligible for the meta-analysis. The systematic review incorporated observational studies that used pediatric control groups with normal results on echocardiography (who were recruited for specific studies) or if healthy children were the primary objective. Studies were excluded if they were review articles or abstracts only, without full text.

LV GLS and global longitudinal strain rate (GLSr) from a 17- or 18-segment model (calculated from segmental averaging of the three apical views [apical four-, three-, and two-chamber]) were included in this meta-analysis. GCS and GRS, calculated from segmental averaging of the short-axis views at the apical, midventricular, and basal levels, were also included in the meta-analysis. In addition, we also evaluated the LV free wall LS measures and included segmental LS (SLS) at the apical, midventricular, and basal levels of the LV free wall from segmental averaging of the three apical views. Clinically, LS is also reported from the weighted average of the six segments from the apical four-chamber view, and CS and RS are reported from weighted averages of the six segments from the midventricular level at the papillary muscle. We therefore stratified our meta-analysis by the different methods, “global” strain (GLS, GCS, and GRS) and “six-segment method” (LS, CS, and RS), of reporting LV strain and incorporated the publications that reported these different methods in the meta-analysis to account for the different techniques used among studies.

Data Collection

Each eligible article meeting the inclusion criteria was reviewed by two independent reviewers (P.T.L and A.M.), and the following data were extracted and entered into an electronic database: (1) study: first and last authors and year of publication; (2) demographics: number of control subjects, age, and gender; (3) clinical: HR, BSA, or body metabolic index; and (4) echocardiographic parameters: vendor-customized ultrasound and model, vendor-customized software and version, probe frequency, FR, FR/HR ratio, tissue-tracking methodology, (endomyocardial, epicardium to endocardium), and reported location of the strain value along the stain curve (systolic strain, end-systolic strain, or postsystolic strain). All the authors of the eligible studies were contacted by e-mail to notify them of the meta-analysis and obtain any missing information not reported in their individual studies.

Quality Assessment

To assess the quality and reporting of studies, two reviewers (P.T.L. and A.M.) independently assessed 12 items that were considered relevant to this meta-analysis topic, on the basis of the quality assessment methodology of Downs and Black and previously validated by our group ( Appendix 3 ).

Statistical Analysis/Data Synthesis

The meta-analysis was performed using Stata/IC version 13 (StataCorp LP, College Station, TX). The means and 95% CIs of strain measures were computed using random-effects models weighted by inverse variance. Between-study statistical heterogeneity was assessed using the Cochran Q statistic and was quantified using the I ² method by measuring inconsistency ( I ² , the percentage of total variance across studies attributable to heterogeneity rather than chance). These results are presented as a forest plot, according to our previously described methodology. The forest plot is used as a graphical display of the relative strength of the effect estimates and CIs for each of the individual studies and the entire meta-analysis. Publication bias was assessed using funnel plots and the Egger test. The funnel plots are also presented according to our previously described methodology and combined with the results of the Egger test to provide a quantitative evaluation of publication bias. The sources of variation among the studies were sought using metaregression to estimate the percentage of heterogeneity on the influence of the variation in normal strain measurements.

Eligibility Criteria

The search identified 603 articles. After excluding duplicate and triplicate articles ( n = 180), 423 studies were screened for relevance. Studies not conducted exclusively in children ( n = 73), articles unrelated to the topic ( n = 92), abstracts without text or reviews ( n = 178), and articles that did not have data on control or normal children ( n = 15) were then excluded. Searching the reference lists did not reveal any additional results. No ongoing studies were found in the clinical trial registries. Sixty-five published observational or case-control studies met the inclusion criteria ( Figure 1 ). All studies that met the search criteria were in English, although the search criteria imposed no language limitations.

Process of inclusion of studies in the meta-analysis.

Electronic Communication

The first or last author of each of the eligible 65 studies was contacted by e-mail. Twenty reports used the same control data set in multiple published studies, and two studies did not have the raw data available. In total, 43 data sets of strain measures from 63 articles of strain measures with 2,325 children were considered eligible for assessment in the meta-analysis ( Figure 2 ). Forty of 43 authors (93%) responded and provided either the raw strain data and/or filled in the missing information on study quality and potential sources of variability. The remaining three data sets (7%) from the authors who did not respond via e-mail were included with the available information provided in their respective reports. Authors who used the same control data sets in multiple published studies were consulted, and one control data set was either provided or chosen from one article on the basis of author recommendation, ( Tables 1 and 2 ). Specific correspondence regarding the handling of the use of the same control data sets in multiple published studies from the same authorship groups is described in Appendix 2 .

Normal value of LV GLS stratified by age distribution and view. (A) LV “global” LS derived from the segmental averaging of the three apical views and (B) LV LS from the apical four-chamber view only. The forest plot lists the names of the included studies by age distribution and in chronological order and the means and CIs with the upper (95%) and lower (5%) limits. Each study is represented by a square that reflects the mean at the point estimate of effect and is proportional to the study’s weight in the meta-analysis. A horizontal line extending from either side of the square reflects the 95% CI. The overall meta-analysis measure of effect is plotted as a diamond with the lateral points of the diamond indicating CIs for this mean estimate. ^∗ Klitsie et al ., Labombarda et al ., Lorch et al ., Marcus et al ., Takayasu et al ., and Takigiku et al . all performed cross-sectional studies and reported strain values for multiple mean age groups from birth to 21 years of age.

Study Selection on the Basis of Strain Measures

Forty-one data sets with 2,084 patients were eligible for the meta-analysis of GLS or LS, 26 data sets with 1,506 patients were eligible for the meta-analysis of GCS or CS, and 15 data sets with 1,092 patients were eligible for the meta-analysis of GRS or RS. The patient characteristics of the included studies are listed in Table 1 . The echocardiographic variables included from the studies are listed in Table 2 .

Study Quality Assessment

Critical appraisal of the studies demonstrated moderate quality in all the studies included. The eligible data sets met >75% of the quality checklist items ( Appendix 3 ). Specifically, all studies clearly defined the objectives, the primary outcomes that were measured, and the main findings. All studies used a deformation imaging acquisition and postprocessing protocol. Reproducibility analysis was performed in 32 of 43 data sets and referenced in four of 43.

Normal Ranges of LS

Normal Ranges of CS and RS

CS

GCS from the combined short-axis views at the base (level of the mitral valve), midventricle (level of the papillary muscle), and apex was reported in 10 of the 43 data sets ( n = 474 children). The normal mean values of GCS varied from −12.9% to −31.4% (mean, −22.3%; 95% CI, −19.9% to −24.6%) ( Figure 3 A). Between-study heterogeneity was evidenced by a Cochran Q statistic of 569 ( P < .0001) and inconsistency by an I ² value of 98.1%. CS from the midventricular level (papillary muscle) was reported in 23 of the 43 data sets ( n = 1,380 children). (Seven studies presented both GCS and CS. ) The normal mean values of CS varied from −14.2% to −26.2% (mean, −20.3%; 95% CI, −19.4% to −21.2%) ( Figure 3 B). Between-study heterogeneity was evidenced by a Cochran Q statistic of 777 ( P < .0001) and inconsistency by an I ² value of 96.8%.

Normal value of LV GCS stratified by age distribution and view. (A) LV “global” CS derived from the segmental averaging of the three short-axis views at the base (mitral valve), midventricular (papillary muscle), and apical levels views and (B) LV CS at the level of the papillary muscle only. The forest plot lists the names of the included studies by age distribution and in chronological order and the means and CIs with the upper (95%) and lower (5%) limits. Each study is represented by a square that reflects the mean at the point estimate of effect and is proportional to the study’s weight in the meta-analysis. A horizontal line extending from either side of the square reflects the 95% CI. The overall meta-analysis measure of effect is plotted as a diamond with the lateral points of the diamond indicating CIs for this mean estimate. ^∗ Klitsie et al ., Labombarda et al ., Lorch et al ., Marcus et al ., Takayasu et al ., and Takigiku et al . all performed cross-sectional studies and reported strain values for multiple mean age groups from birth to 21 years of age.

RS

GRS from the combined short-axis views at the base (level of the mitral valve), midventricle (level of the papillary muscle), and apex was reported in six of the 43 data sets ( n = 377 children). The normal mean values of GRS varied from 33.9% to 54.5% (mean, 45.2%; 95% CI, 38.8%–51.7%) ( Figure 4 A). Between-study heterogeneity was evidenced by a Cochran Q statistic of 283 ( P < .0001) and inconsistency by an I ² value of 97.5%. RS from the midventricular level (papillary muscle) was reported in 12 of the 43 data sets ( n = 946 children). (Three studies presented both GRS and RS. ) The normal mean values of RS varied from 28.8% to 58.1% (mean, 46.7%; 95% CI, 42.3%–51.1%) ( Figure 4 B). Between-study heterogeneity was evidenced by a Cochran’s Q statistic of 1,811 ( P < .0001) and inconsistency by an I ² value of 99.0%.

Normal value of LV GRS stratified by age distribution and view. (A) LV “global” RS derived from the segmental averaging of the three short-axis views at the base (mitral valve), midventricular (papillary muscle), and apical levels views and (B) LV RS at the level of the papillary muscle only. The forest plot lists the names of the included studies by age distribution and in chronological order and the means and CIs with the upper (95%) and lower (5%) limits. Each study is represented by a square that reflects the mean at the point estimate of effect and is proportional to the study’s weight in the meta-analysis. A horizontal line extending from either side of the square reflects the 95% CI. The overall meta-analysis measure of effect is plotted as a diamond with the lateral points of the diamond indicating CIs for this mean estimate. ^∗ Klitsie et al ., Labombarda et al ., Lorch et al ., Marcus et al ., Takayasu et al ., and Takigiku et al . all performed cross-sectional studies and reported strain values for multiple mean age groups from birth to 21 years of age.

The heterogeneity for GCS, CS, GRS, and RS was not explained by the different methods or the age, gender, BSA, HR, tissue-tracking methodology, reporting of strain value along the strain curve, FR, FR/HR ratio, or probe size.

Age and Global Strain

Age did not explain the heterogeneity of the reported normal ranges of values for GLS, GCS, or GRS. Lorch et al ., Marcus et al ., Klitsie et al ., and Labombarda et al . performed cross-sectional studies with patients from birth to 21 years of age. Takayasu et al . ¹⁶ generated strain measures in two separate cohorts of children and adolescents. The breakdown of the age distribution for the remaining 38 data sets was as follows: four data sets recruited patients 0 to 1 years of age, 18 data sets had patients with age ranges of 2 to 9 years, 10 data sets had patients with age ranges of 10 to 13 years, and six data sets examined patients with age ranges of 14 to 21 years. We performed a separate meta-analysis for LV strain measures stratified by age distribution using the mean age from each study as a continuous variable and also by categorizing each study into one of the four age distribution categories (0–1, 2–9, 10–13, and 14–21 years) ( Figures 2–4 ). The Cochran Q statistic ranged from 39 to 370 ( P < .0001) and the I ² value remained the same for both methods and ranged from 82.19% to 98.0%. The means and 95% CIs for GLS, GCS, and GRS and for LS from the apical four-chamber view and CS and RS at the midventricular level are listed in Table 3 . A similar analysis stratified by age was done for SLS from the segmental averaging of the three apical views and from the apical four-chamber view only. An apex-to-base gradient existed for all ages and the results are listed in the Appendix 4 .

Publication Bias

Both visual inspection of the funnel plot and the nonsignificant results of the Egger test for the GLS, GCS, and GRS measures ( P = .40) suggested the absence of publication bias ( Figure 5 ).

Publication bias. Funnel plot for studies of LV GLS. The standard error of the effect estimate is plotted on the vertical axis. The mean of LV GLS is plotted on the horizontal axis. Visual inspection shows symmetry in the distribution of the studies that suggests the absence of publication bias ( P = .40 from the Egger test for statistical funnel plot symmetry).

Sources of Variability

In this meta-analysis age, gender, BSA, HR, FR, FR/HR ratio, tissue-tracking method, location of reported strain value along the strain curve, ultrasound vendor (model), and software (version) were tested to determine if any of these parameters influenced the variability in reporting of normal strain and strain rate measures in children ( Table 2 ). We stratified the meta-analysis by the method of generating the strain measurements: the six-segment method versus the global (17- or 18-segment average) method. To account for maturational changes in hemodynamic parameters from infancy to adolescence, we stratified the meta-analysis by age distribution to determine its contribution to the reported ranges of normal values.

Individual meta-regression analysis on each dependent strain measure and each independent variable was performed to examine which parameter might statistically influence the variation in strain measures in this meta-analysis. None of the demographic, clinical, or echocardiographic variables were significantly associated with the mean values for any of the strain measures ( Table 4 ). Intervendor equipment and software was independently assessed; nine of the 43 data sets (21%) used non-GE equipment, and all but one of the studies acquired and then analyzed the strain images with the same vendor and software package. Six data sets used Philips equipment and software, two used Siemens products, and two used Esaote (Mylab 50/XStrain). One study used both Philips and GE, and one study used Philips, GE, and Toshiba ultrasound machines. We created vendor-specific normal ranges of values for GE, Philips, Siemens, and Esaote ( Appendix 5 ).

Table 4

Metaregression for strain measures

Variable	P Value (GLS)	P value (GCS)	P value (GRS)
Age	.56	.38	.19
Gender	.67	.56	.56
BSA	.67	NA	.34
HR	.34	.78	.22
FR	.14	.47	.48
FR/HR ratio	.23	.56	.14
Ultrasound scanner ∗	.19	.62	.47
Model †	.43	.12	.17
Vendor software ∗	.22	.35	.36
Version ‡	.23	.37	.69
Probe size	.26	.43	.32
Tissue-tracking methodology	.54	.34	.19
Location of strain value	.14	.47	.48

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Patterns of Mechanical Inefficiency in Pediatric Dilated Cardiomyopathy and Their Relation to Left Ventricular Function and Clinical Outcomes
2. Right Ventricular Systolic Function Responses to Acute and Chronic Pulmonary Hypertension: Assessment with Myocardial Deformation
3. Correction
4. Continuing Education and Meeting Calendar

Stay updated, free articles. Join our Telegram channel

Join