Occult RV systolic dysfunction detected by CMR derived RV circumferential strain in patients with pectus excavatum

Abstract

Aims

To investigate the right ventricular (RV) strain in pectus excavatum (PE) patients using cardiac magnetic resonance tissue tracking (CMR TT).

Materials and methods

Fifty consecutive pectus excavatum patients, 10 to 32 years of age (mean age 15 ± 4 years), underwent routine cardiac magnetic resonance imaging (CMR) including standard measures of chest geometry and cardiac size and function. The control group consisted of 20 healthy patients with a mean age of 17 ± 5 years. RV longitudinal and circumferential strain magnitude was assessed by a dedicated RV tissue tracking software.

Results

Fifty patients with images of sufficient quality were included in the analysis. The mean right and left ventricular ejection fractions were 55 ± 5% and 59 ± 4%. The RV global longitudinal strain was -21.88 ± 4.63%. The RV circumferential strain at base, mid-cavity and apex were -13.66 ± 3.09%, -11.31 ± 2.79%, -20.73 ± 3.45%, respectively. There was no statistically significant decrease in right ventricular or left ventricular ejection fraction between patients and controls (p > 0.05 for each). There was no significant difference in RV global longitudinal strain between two groups (-21.88 ± 4.63 versus -21.99 ± 3.58; p = 0.93). However, there was significant decrease in mid-cavity circumferential strain magnitude in pectus patients compared with controls (-11.31 ± 2.79 versus -16.19 ± 2.86; p < 0.001). PE patients had a significantly higher basal circumferential strain (-13.66 ± 3.09% versus -9.76 ± 1.79; p < 0.001) as well as apical circumferential strain (-20.73 ± 3.45% versus -12.07 ± 3.38) than control group.

Conclusion

Mid-cavity circumferential strain but not longitudinal strain is reduced in pectus excavatum patients. Basal circumferential strain as well as apical circumferential strain were increased as compensatory mechanism for reduced mid-cavity circumferential strain. Further studies are needed to establish clinical significance of this finding.

About the Authors:

Vien T. Truong

Roles Data curation, Formal analysis, Methodology, Writing – original draft

Affiliations Department of Cardiology, The Christ Hospital; Cincinnati, Ohio, United States of America, Department of Cardiology, Pham Ngoc Thach University of Medicine, Ho Chi Minh City, Viet Nam

Candice Y. Li

Roles Data curation, Methodology, Resources, Writing – review & editing

Affiliation: Department of Cardiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Rebeccah L. Brown

Roles Conceptualization, Investigation, Writing – review & editing

Affiliation: Department of Surgery, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Ryan A. Moore

Roles Investigation, Methodology, Supervision, Writing – review & editing

Affiliation: Department of Cardiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Victor F. Garcia

Roles Data curation, Supervision, Validation, Writing – review & editing

Affiliation: Department of Surgery, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Eric J. Crotty

Roles Conceptualization, Writing – review & editing

Affiliation: Department of Radiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Michael D. Taylor

Roles Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Validation, Writing – review & editing

Affiliation: Department of Cardiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, United States of America

Tam M. N. Ngo

Roles Data curation, Formal analysis, Writing – review & editing

Affiliation: Department of Cardiology, Pham Ngoc Thach University of Medicine, Ho Chi Minh City, Viet Nam

Wojciech Mazur

Roles Conceptualization, Formal analysis, Investigation, Methodology, Supervision, Writing – review & editing

* E-mail: [email protected]

Affiliation: Department of Cardiology, The Christ Hospital; Cincinnati, Ohio, United States of America

ORCID http://orcid.org/0000-0002-0286-9127

Introduction

Pectus excavatum (PE) is the most common congenital deformity of the anterior chest wall that affects both appearance and function [1]. Pectus excavatum occurs in 1 in 300–400 births with a male predominance [1, 2]. The degree of abnormal chest wall deformity determines its functional effect, particularly its cardiopulmonary impairment and physiologic limitations [2]. Patient may be completely asymptomatic or presents with exertional dyspnea, decreased endurance, tachycardia, and palpitations [2, 3]. The cause of PE remains unknown. Impairment in the growth of the sternum and biomechanical properties of costal cartilage, are proposed in the pathogenesis [4]. Many patients with PE have associated alterations in right ventricular (RV) morphology and function and cardiopulmonary disturbance may contribute to symptoms in patients with PE [5–7]. Guidelines for the assessment of cardiovascular function remain undefined.

A non-contrast computerized tomographic (CT) scan is helpful to examine the deformity of the bony and cartilaginous skeleton, which clearly finds any cardiac compression or displacement. Assessing RV function using echocardiography is limited by the complex RV geometry, retrosternal position, and complex motion [8], especially in pectus excavatum due to deformity of the anterior chest wall. In contrast, MRI provides highly reproducible information on RV myocardial motion and function and can be used instead of CT scan to reduce radiation exposure in pectus excavatum. Cardiac magnetic resonance (CMR) imaging has emerged as validated tool in patients with pectus excavatum [6, 9, 10]. The purpose of our study is to investigate the RV strain in pectus excavatum patients using cardiac magnetic resonance tissue tracking (CMR TT).

Materials and methods

Patient population

Fifty pectus excavatum patients without known RV pathology, 10 to 32 years of age (mean age 15 ± 4 years), underwent cardiac magnetic resonance imaging (CMR) as part of standard clinical evaluation. The control group consisted of 20 healthy patients with a mean age of 17 ± 5 years. The control group was referred for the following suspected conditions: atypical chest pain, suspected coronary artery anomaly, cardiac mass by echocardiography (typically prominent moderator band). We excluded not only patients with reduced LVEF, segmental wall motion abnormalities but also abnormal longitudinal and circumferential strain. Pectus excavatum patients or the control group with known conditions affecting RV function such as congenital heart disease, cardiomyopathy, valvular heart disease, sleep apnea, morbid obesity, known or suspected pulmonary hypertension, were excluded from the study. This study was approved by the institutional review board (IRB) of Cincinnati Children’ Hospital Medical Center.

CMR acquisition

Standard CMR acquisition was performed as reported previously [11]. Briefly, imaging was performed on a 1.5 Tesla scanner (Ingenia, Philips Healthcare; Best, Netherlands), using a phased-array coil. A horizontal long-axis image and a short-axis RV stack from the atrioventricular ring to the RV apex were acquired using an SSFP pulse sequence (repetition time of 3.2 ms; echo time of 1.7 ms; flip angle of 60°; sequential 7 mm slices with no interslice gap). There were 30 phases per cardiac cycle resulting in a mean temporal resolution of 30–40 ms.

RV longitudinal and circumferential strain magnitude was assessed by a dedicated RV tissue tracking software [12]. The following clinical variables were collected for each patients: age, sex, body mass index (BMI- kg/m2), body surface area (BSA-m2), and heart rate (bpm). The cardiac and chest MR variables were: LV and RV end—diastolic volume (EDV-ml), end—systolic volume (ESV-ml), stroke volume (ml) as well as indexed (BSA) values were calculated, ejection fraction (EF-%), left ventricular mass index (g/m2), cardiac output (L/min), cardiac index (L/min/m2), Haller index, correction index, and depression index. The Haller index is a ratio of the transverse diameter of the chest to the anterior-posterior diameter, measured from the inner aspect of the sternum to the anterior aspect of the vertebral body at the level of greatest sternal depression (Fig 1)[13]. Correction index and depression index were evaluated according to methods previously described [14, 15].

[Figure omitted. See PDF.]

Fig 1. Pectus index and CMR tissue tracking.

(A) The Haller Index (HI) is a ratio of the transverse diameter of the chest (line a) to the distance between the posterior aspect of the sternum and the anterior portion of the vertebra (line b): HI = a/b. The correction index (CI) measures the depression of the sternum relative to the anterior chest: CI = [(c-b)/c] x 100. (B) Right ventricular longitudinal strain. (C) and (D) Mid-cavity circumferential strain and peak value was recorded. The yellow colored contours show the tracking of the ventricle.

https://doi.org/10.1371/journal.pone.0189128.g001

Tissue tracking

The RV myocardial deformation was quantified using a prototype of RV specific CVI42 Tissue Tracking software (Circle Cardiovascular Imaging, Calgary, Canada). First, an experienced operator traced the RV endocardial and epicardial borders at the end diastolic (ED) phase in both short-axis and long-axis cine images. The software then constructed a 3D deformable myocardial model based on the tracing, assuming the myocardium is nearly incompressible [16]. In each of the subsequent frames the displacements of the myocardial tissues, including the borders were determined using a gradient-based optical flow method with an incompressible model constraint. The propagated myocardial tissue across the cardiac cycle was verified by the operator to ensure the accuracy of the propagation. Strain values (along the longitudinal, circumferential, and radial directions) for each tissue point as well as the global strain values for the short-axis and long-axis views were automatically derived by the software [12]. The right ventricle was divided into basal, mid-cavity, and apical segments to derive regional deformation parameters.

Strain analysis

The horizontal long axis was used for calculation of longitudinal strain, while the short axis at the level of greatest sternal depression was used to calculate mid-cavity circumferential strain of the right ventricle. The apical and basal circumferential strain were calculated from the apical and basal short-axis views respectively. The interventricular septum was not included in the strain calculation. Endocardial and epicardial contours were drawn in the cardiac phase with the most distinct myocardium boundaries; RV trabeculations were carefully excluded. The software automatically propagated contours throughout all phases. Longitudinal and mid-cavity circumferential strain were computed as shown in Fig 1 [12].

Statistical analysis

Continuous variables are expressed as mean ± standard deviation (SD) for normal distributions and median + interquartile range for non-normal distributions. Normality was tested using the Shapiro–Wilk test. For the evaluation of qualitative variables, we used the Chi-Square test. To test for significant differences between continuous variables in two groups, independent sample t-tests were performed for normally distributed variables and Mann-Whitney U test was performed for variables with non-normal distribution. Spearman's Rho correlation test was performed to evaluate the relationships among continuous variables. Statistical analysis was performed using the SPSS 22 software program (SPSS Inc., Chicago, IL, USA). A P value of < 0.05 was considered statistically significant.

Results

Demographic, and CMR-derived parameters for the whole sample are shown in Table 1 and S1 File. The mean age of patients was the 15 ± 4 (years). There were 39 men with a mean age of 16 ± 4 years and 11 women with a mean age of 14 ± 4 years (P = 0.12). The mean value of LVEF and RVEF were 58.97 ± 4.07; 55.06 ± 4.94, respectively (Table 1). The mean Haller index was 5.71 ± 2.93 with a median index of 4.68 (4.05 to 6.78) (Table 2). There were 46 patients with severe chest wall phenotype, defined as a Haller index over 3.25. The LV global longitudinal strain and was -21.88 ± 4.63%. The RV circumferential strain at base, mid-cavity and apex were -13.66 ± 3.09%, -11.31 ± 2.79%, -20.73 ± 3.45%, respectively. Differences were not statistically significant for age, right ventricular or left ventricular ejection fraction (p > 0.05 for each). There was no significant difference in RV global longitudinal between patients and controls (-21.88 ± 4.63% versus -21.99 ± 3.58%; p = 0.93). However, there was significant decrease in mid-cavity circumferential strain magnitude in pectus patients compared with controls (-11.31 ± 2.79 versus -16.19 ± 2.86; p < 0.001) (Table 3). PE patients had a significantly higher basal circumferential strain (-13.66 ± 3.09% versus -9.76 ± 1.79; p < 0.001) as well as apical circumferential strain (-20.73 ± 3.45% versus -12.07 ± 3.38) than control group.

[Figure omitted. See PDF.]

Table 1. Baseline characteristics of fifty pectus excavatum patients.

https://doi.org/10.1371/journal.pone.0189128.t001

[Figure omitted. See PDF.]

Table 2. Chest index.

https://doi.org/10.1371/journal.pone.0189128.t002

[Figure omitted. See PDF.]

Table 3. Basic demographics and patients characteristics between two groups.

https://doi.org/10.1371/journal.pone.0189128.t003

There were no correlations between RVEF and Haller index (Spearman’s = 0.039, p = 0.79), RV longitudinal strain and Haller index (Spearman’s = -0.029, p = 0.84), RV mid-cavity circumferential strain and Haller index (Spearman’s = -0.148, P = 0.31), RVEF and correction index (Spearman’s = 0.013, p = 0.93), RV longitudinal strain and correction index (Spearman’s = 0.036, p = 0.81), RV mid-cavity circumferential strain and correction index (Spearman’s = -0.202, p = 0.16) (Fig 2).

[Figure omitted. See PDF.]

Fig 2. Relationship between pectus severity index and right ventricular ejection fraction, myocardial strain.

https://doi.org/10.1371/journal.pone.0189128.g002

Discussion

CMR TT has emerged as a useful technique for assessing RV myocardial function [12, 17]. This study assessed RV myocardial strain using CMR TT in a cohort of 50 patients with pectus excavatum. Our study showed no significant difference in RV longitudinal strain between the patients and normal controls. Patients with pectus excavatum had trend towards lower RVEF compared to the healthy control group, but it did not reach statistical significance. However, mid-cavity circumferential strain magnitude is significantly decreased in patient group. This may be explained by geometric distortion of the RV due to sternal compression in pectus excavatum. Saleh et al studied 30 patients with pectus excavatum with age ranges 14–67 years who underwent CMR. They found decreased RV ejection fraction (53.9 ± 9.6 versus 60.5 ± 9.5; p = 0.013), reduced right ventricular short axis dimensions (p < 0.05) both at end diastole and systole, and increased right ventricular long axis dimensions at end diastole (p < 0.05) compared to controls [10]. Töpper et al performed CMR in 38 patients (mean age 21 ± 8.3; 31 men) before and after surgical correction and found RVEF was decreased before surgery and improved significantly after successful correction (45.7% vs 48.3%, P = 0.0004) [9]. RVEF is significantly decreased in study of Saleh et al and Töpper et al but not in our study. This can be explained by severity of pectus excavatum. In our study, the Haller index was 5.71 ± 2.93 lower than those of Saleh et al (9.3 ± 5.0) and Töpper et al (9.64 ± 0.75). Furthermore, our study found that patients with PE had the higher basal and apical circumferential strain which may be explained as compensatory mechanism for reduced mid-cavity circumferential strain. Severeal recent studies demonstrated,in conditions leading to RV pressure or volume overload such as tetralogy of Fallot, pulmonary hypertension, or Systemic RV, the contractile pattern changes from longitudinal to circumferential shortening [18–20]. This might be the reason why longituidinal strain was not increased in our study. In addition, the RV with deep longitudinal myocardial layers that are aligned base to apex are likely to affect RV strain [21], which is also supported by the fact that a significant base-to-apex segmental strain gradient with much higher apical circumferential strain value in PE.

Recent studies demonstrate that for evaluating ventricle dysfunction, myocardial strain value can be used as a highly sensitive marker as opposed to the ejection fraction, since a reduction in the strain often precedes a decline in EF [22, 23]. De Siqueira et al retrospectively enrolled 116 patients (age 52.2 ± 12 years, 73.6% women) referred to CMR for pulmonary hypertension evaluation who underwent right heart catheterization. The author found the all strain values (longitudinal strain and circumferential strain) were significantly decreased in patients with pulmonary hypertension with normal RVEF compared to control group [23]. We did not find correlations between RVEF, RV longitudinal strain and RV circumferential strain with Haller index and correction index. Saleh et al also found no significant correlation between the RVEF and pectus severity index (r = -0.2061, p = 0.2746) [10]. The author also explain the possible reasons that included chest reconfiguration and cardiac left lateral shift [10]. These factors could be compensatory mechanisms to avoid the impairment in RV systolic function. Sigalet found the similar result that pectus index has not been shown to have any correlation with stroke volume [24].

Limitations

Limitations of our study include its retrospective nature leading to possible selection bias. We were unable to report segmental longitudinal strain as well as radial strain due to poor reproducibility which is known as current software limitation. Moreover, the measured strain will be slightly different than the actual strain that would be measured by tracking the same myocardium without through plane motion. There is no easy way to fix this limitation using the 2-d system that we employ for strain measurement. The only true way to correct for this is using a 3-d cine acquisition to get an isotropic 3-d data set combined with a 3-d strain analysis. While that has been shown to work in limited research studies, the technique suffers from significantly reduced spatial and temporal resolution. This limitation affects equally PE and control patients, as such it is unlikely that it would affect findings of our study.

Conclusion

Mid-cavity circumferential strain but not longitudinal strain is reduced in pectus excavatum patients and is more sensitive marked for occult RV dysfunction than RVEF. There was a compensatory increase in basal as well as apical circumferential strain representing RV adaptation to maintain systolic RV function without simultaneous increase in RV longitudinal strain. Further studies are needed to establish clinical significance of this finding.

Supporting information

[Figure omitted. See PDF.]

S1 File. Datasheet for pectus excavatum study.

https://doi.org/10.1371/journal.pone.0189128.s001

(SAV)

Citation: Truong VT, Li CY, Brown RL, Moore RA, Garcia VF, Crotty EJ, et al. (2017) Occult RV systolic dysfunction detected by CMR derived RV circumferential strain in patients with pectus excavatum. PLoS ONE 12(12): e0189128. https://doi.org/10.1371/journal.pone.0189128

References

1. Fokin AA, Steuerwald NM, Ahrens WA, Allen KE. Anatomical, Histologic, and Genetic Characteristics of Congenital Chest Wall Deformities. Seminars in Thoracic and Cardiovascular Surgery. 2009;21(1):44–57. http://dx.doi.org/10.1053/j.semtcvs.2009.03.001. pmid:19632563

2. Jaroszewski D, Notrica D, McMahon L, Steidley DE, Deschamps C. Current Management of Pectus Excavatum: A Review and Update of Therapy and Treatment Recommendations. The Journal of the American Board of Family Medicine. 2010;23(2):230–9. pmid:20207934

3. Jaroszewski DE, Fonkalsrud EW. Repair of pectus chest deformities in 320 adult patients: 21 year experience. The Annals of thoracic surgery. 2007;84(2):429–33. Epub 2007/07/24. pmid:17643611.

4. Feng J, Hu T, Liu W, Zhang S, Tang Y, Chen R, et al. The biomechanical, morphologic, and histochemical properties of the costal cartilages in children with pectus excavatum. Journal of pediatric surgery. 2001;36(12):1770–6. Epub 2001/12/06. pmid:11733904.

5. Oezcan S, Attenhofer Jost CH, Pfyffer M, Kellenberger C, Jenni R, Binggeli C, et al. Pectus excavatum: echocardiography and cardiac MRI reveal frequent pericardial effusion and right-sided heart anomalies. European Heart Journal—Cardiovascular Imaging. 2012;13(8):673–9. pmid:22298154

6. Tandon A, Wallihan D, Lubert AM, Taylor MD. The effect of right ventricular compression on cardiac function in pediatric pectus excavatum. Journal of Cardiovascular Magnetic Resonance. 2014;16(Suppl 1):P250–P.

7. Kowalewski J, Brocki M, Dryjanski T, Zolynski K, Koktysz R. Pectus excavatum: increase of right ventricular systolic, diastolic, and stroke volumes after surgical repair. The Journal of thoracic and cardiovascular surgery. 1999;118(1):87–92; discussion -3. Epub 1999/06/29. pmid:10384189.

8. Haddad F, Hunt SA, Rosenthal DN, Murphy DJ. Right Ventricular Function in Cardiovascular Disease, Part I. Anatomy, Physiology, Aging, and Functional Assessment of the Right Ventricle. 2008;117(11):1436–48. pmid:18347220

9. Töpper A, Polleichtner S, Zagrosek A, Prothmann M, Traber J, Schwenke C, et al. Impact of surgical correction of pectus excavatum on cardiac function: insights on the right ventricle. A cardiovascular magnetic resonance study†. Interactive CardioVascular and Thoracic Surgery. 2016;22(1):38–46. pmid:26487434

10. Saleh RS, Finn JP, Fenchel M, Moghadam AN, Krishnam M, Abrazado M, et al. Cardiovascular magnetic resonance in patients with pectus excavatum compared with normal controls. Journal of Cardiovascular Magnetic Resonance. 2010;12(1):73. pmid:21144053

11. Hor KN, Gottliebson WM, Carson C, Wash E, Cnota J, Fleck R, et al. Comparison of magnetic resonance feature tracking for strain calculation with harmonic phase imaging analysis. JACC Cardiovascular imaging. 2010;3(2):144–51. Epub 2010/02/18. pmid:20159640.

12. Truong VT, Safdar KS, Kalra DK, Gao X, Ambach S, Taylor MD, et al. Cardiac magnetic resonance tissue tracking in right ventricle: Feasibility and normal values. Magnetic Resonance Imaging. 2017;38:189–95. http://dx.doi.org/10.1016/j.mri.2017.01.007. pmid:28093270

13. Haller JA Jr., Kramer SS, Lietman SA. Use of CT scans in selection of patients for pectus excavatum surgery: a preliminary report. Journal of pediatric surgery. 1987;22(10):904–6. Epub 1987/10/01. pmid:3681619.

14. Fagelman KM, Methratta S, Cilley RE, Wilson MZ, Hollenbeak CS. The Depression Index: an objective measure of the severity of pectus excavatum based on vertebral diameter, a morphometric correlate to patient size. Journal of pediatric surgery. 2015;50(7):1130–3. Epub 2015/03/19. pmid:25783321.

15. St. Peter SD, Juang D, Garey CL, Laituri CA, Ostlie DJ, Sharp RJ, et al. A novel measure for pectus excavatum: the correction index. Journal of pediatric surgery. 2011;46(12):2270–3. pmid:22152863

16. Bistoquet A, Oshinski J, Skrinjar O. Myocardial deformation recovery from cine MRI using a nearly incompressible biventricular model. Medical image analysis. 2008;12(1):69–85. Epub 2008/02/01. pmid:18234539.

17. Harrild DM, Han Y, Geva T, Zhou J, Marcus E, Powell AJ. Comparison of cardiac MRI tissue tracking and myocardial tagging for assessment of regional ventricular strain. The International Journal of Cardiovascular Imaging. 2012;28(8):2009–18. pmid:22392105

18. Raina A, Vaidya A, Gertz ZM, Susan C, Forfia PR. Marked changes in right ventricular contractile pattern after cardiothoracic surgery: implications for post-surgical assessment of right ventricular function. The Journal of heart and lung transplantation: the official publication of the International Society for Heart Transplantation. 2013;32(8):777–83. Epub 2013/06/19. pmid:23773508.

19. Pettersen E, Helle-Valle T, Edvardsen T, Lindberg H, Smith HJ, Smevik B, et al. Contraction pattern of the systemic right ventricle shift from longitudinal to circumferential shortening and absent global ventricular torsion. Journal of the American College of Cardiology. 2007;49(25):2450–6. Epub 2007/06/30. pmid:17599609.

20. Kind T, Mauritz GJ, Marcus JT, van de Veerdonk M, Westerhof N, Vonk-Noordegraaf A. Right ventricular ejection fraction is better reflected by transverse rather than longitudinal wall motion in pulmonary hypertension. Journal of cardiovascular magnetic resonance: official journal of the Society for Cardiovascular Magnetic Resonance. 2010;12:35. Epub 2010/06/08. pmid:20525337

21. Levy PT, Sanchez Mejia AA, Machefsky A, Fowler S, Holland MR, Singh GK. Normal ranges of right ventricular systolic and diastolic strain measures in children: a systematic review and meta-analysis. Journal of the American Society of Echocardiography: official publication of the American Society of Echocardiography. 2014;27(5):549–60, e3. Epub 2014/03/04. pmid:24582163

22. Smiseth OA, Torp H, Opdahl A, Haugaa KH, Urheim S. Myocardial strain imaging: how useful is it in clinical decision making? European Heart Journal. 2016;37(15):1196–207. pmid:26508168

23. de Siqueira ME, Pozo E, Fernandes VR, Sengupta PP, Modesto K, Gupta SS, et al. Characterization and clinical significance of right ventricular mechanics in pulmonary hypertension evaluated with cardiovascular magnetic resonance feature tracking. Journal of cardiovascular magnetic resonance: official journal of the Society for Cardiovascular Magnetic Resonance. 2016;18(1):39. pmid:27306901.

24. Sigalet DL, Montgomery M, Harder J. Cardiopulmonary effects of closed repair of pectus excavatum. Journal of pediatric surgery. 2003;38(3):380–5; discussion -5. Epub 2003/03/13. pmid:12632353.