A novel, data-driven conceptualization of Critical LVOTO based on baseline echocardiograms CHSS Fall Work Weekend November 18, 2016 James M. Meza, MD and Brian W. McCrindle, MD, MPH (Study PI) for the LVOTO Baseline Echo Working Group
Overview Background, Rationale, Hypothesis Cluster Analysis – Methods and Results Defining the groups – Methods and Results Differentiating characteristics Group characteristics Outliers Discussion, Interpretation, and Clinical Relevance
Traditional Classification Derived from morphologic analysis Qualitative in nature Current IPCCC classification of HLHS is complex Association with outcomes and prognostic potential is unclear
Echocardiographic diagnosis is sophisticated and ubiquitous Detailed quantitative and qualitative evaluation of function and morphology possible CHSS Imaging Core Lab
Rationale and Hypothesis Research questions How do these measures relate to one another? Can infants be grouped based solely on their baseline echocardiographic characteristics, both quantitative & qualitative? Hypothesis Novel groups of similar infants with greater clinical relevance can be identified using baseline echocardiographic quantitative and qualitative morphologic and functional characteristics
Patient Source CHSS Critical LVOTO Registry Inclusion criteria Neonate ≤ age 30 days at admission to a CHSS institution 2005 onward Concordant AV & VA connections Precludes an adequate systemic cardiac output through the aortic valve Exclusion criteria First intervention at a non-CHSS institution Discordant AV or VA connections Atrioventricular Septal Defect
Echocardiogram available for review Study Population Eligible N=716 Echocardiogram available for review N=651
Echocardiographic Data Single reviewer Protocol of 194 variables Variables for analysis 136 total
Classification – Cluster Analysis Do the variables in the data set tell us anything about the structure of the observations (patients)? Identifies groups with similar characteristics NOT which variables differentiate the groups Hierarchical, unsupervised, non-overlapping, agglomerative Ward’s method
Tree Diagram N=651 2 N=338 1 N=215 3 N=98
Characterizing Groups Multinomial and logistic regression Descriptive statistics Survival analysis
Differentiating Factors Odds Ratio P value Group 1 Aortic atresia 0.2 0.01 LV end diastolic area (cm2) 0.03 < 0.0001 Group 2 3.9 83% concordant if perform classification analysis Group 3 = reference
Distribution of AVA 87% 11% 8%
Distribution of LVEDA cm2 Median LVEDA 1.05 0.57 1.96
1 vs 2 Probability modeled = membership in Group 2 Classification = 89% concordant Parameter Estimate P value Odds Ratio 95% CL Aortic atresia 1.72 < 0.0001 31.33 15.5-63.2 Tunnel type subvalvar LVOTO -1.89 0.02 0.005-0.11 LV end diastolic endocardial length (cm) 3.5 33.58 13.07-86.31 Group 1 Group 2 P value Aortic atresia 11% (24/215) 87% (294/338) < 0.0001 Tunnel type subvalvar LVOTO 15% (33/215) 1% (4/338) LV end diastolic endocardial length (cm) 2.41 1.60
1 vs 3 Probability modeled= membership in Group 3 Classification = 88% concordant Parameter Estimate P value Odds Ratio 95% CL LV hypoplasia -1.23 < 0.0001 0.09 0.03-0.21 LV end diastolic epicardial area (cm2) -0.93 0.39 0.28-0.54 Group 1 Group 3 P value LV hypoplasia 86% (185/215) 9% (9/98) < 0.0001 LV end diastolic epicardial area (cm2) 3.42 6.90
2 vs 3 Probability modeled= membership in Group 3 Classification = 97% concordant Parameter Estimate P value Odds Ratio 95% CL Mitral valve annular abnormality -2.28 < 0.0001 0.01 0.001-0.07 Apex forming LV 2.58 172.68 32.90-906.22 Group 2 Group 3 P value Mitral valve annular abnormality 99% (336/338) 8% (8/98) < 0.0001 Apex forming LV 1% (3/338) 95% (93/98)
Operative Strategy Differs by Group Group 1 Group 2 Group 3 P value First procedure type < 0.0001 Operation 91% 98% 39% Catheter 9% 2% 61% First operation Single Ventricle Palliation 90% 58% Biventricular repair 1% 42%
Survival Differs by Group %
How to explain potential “outliers?”
Outlier analysis, aortic atresia in Group 1 Group 1, no AA (N=191) Group 1, AA (N=24) P value LV end diastolic area 2.04 1.35 0.0009 LV end diastolic epicardial area 3.53 3.64 0.66 Mitral valve annular abnormality 82% (157/191) 92% (22/24) 0.38 Apex forming LV 23% (43/191) 4% (1/24) 0.03 LV hypoplasia 84% (161/191) 100% (24/24) Tunnel type subvalvar LVOTO 14% (27/191) 25% (6/24) 0.22 Retrograde flow in PDA 5% (9/191) 29% (7/24) < 0.0001 Death 28% (45/191) 26% (14/24) 0.86
Outlier analysis, mitral atresia in Group 1 Group 1, no MA (N=194) Group 1, MA (N=21) P value LV end diastolic area 2.07 0.97 < 0.0001 Mitral valve annular abnormality 81% (158/194) 100% 0.03 Apex forming LV 22% (43/194) 5% (1/21) 0.08 LV hypoplasia 85% (164/194) 0.05 Tunnel type subvalvar LVOTO 16% (31/194) 10% (2/21) 0.75 First intervention 0.23 Catheter-based 10% 0% Operation 90% First operation type 0.17 AVR 1% Norwood 69% 95% Hybrid 19% 5% Other Arch procedure Death 25% (49/194) 17% (10/21) 0.04 Is the mitral atresia actually just very severe mitral stenosis?
Outlier analysis, aortic atresia in Group 3 Group 3, no AA (N=90) Group 3, AA (N=8) P value LV end diastolic area 4.6 3.7 0.15 Mitral valve annular abnormality 8% (7/90) 13% (1/8) 0.51 Apex forming LV 96% (86/90) 88% (7/8) 0.35 LV hypoplasia 25% (2/8) 0.16 Tunnel type subvalvar LVOTO 7% (6/90) 0% 1.0 First intervention < 0.0001 Catheter-based 67% (60/90) Operation 33% (30/90) 100% First operation type Norwood 33% (10/30) 50% (4/8) Hybrid 20% (6/30) Arch procedure 37%(11/30 Yasui 10% (3/30) Death 13% (12/90) 0.59 Is aortic atresia in this group just really severe aortic stenosis?
Outlier Analysis, Aortic stenosis in Group 2 Group 2, no AS (N=300) Group 2, AS (N=38) P value LV end diastolic area 0.59 1.09 < 0.0001 Mitral valve annular abnormality 99% (298/300) 100% 1.0 Apex forming LV 0.3% (1/300) 5% (2/38) 0.03 LV hypoplasia 99% (299/300) 97% (37/38) 0.21 Tunnel type subvalvar LVOTO 0.67% (2/300) 0.06 Retrograde flow in PDA 0% First intervention 0.51 Catheter-based 2% 3% Operation 98% 97% First operation type 0.36 Norwood 81% 73% Heart transplant 1% Hybrid 17% 24% Arch procedure Death 42% (126/300) 32% (12/38) 0.29
Do 4 clusters exist? No Group 2 (“HLHS”) and Group 3 (“AS”) remain Group 1 breaks into two different groups Group 1 (N=137) Group 4 (N=78) P value Aortic valve atresia 17% (23/137) 1% (1/78) <0.0001 LV end diastolic area 1.87 2.14 0.05 Mitral valve annular abnormality 83% (114/137) 83% (65/78) 1.0 Apex forming LV 8% (11/137) 42% (33/78) First intervention 0.80 Catheter-based 9% (13/137) 8% (9/78) Operation 91% (124/137) 92% (72/78) First operation type AVR 0% 3% (2/72) Single Ventricle Palliation 98% (122/124) 75% (54/72) Heart transplant Other 1% (1/72) Arch procedures 2% (2/124) 21% (15/72) Death 30% (41/137) 23% (18/78) 0.34 Six groups also evaluated
In summary Data driven classification based on: LV size (end diastolic area) Presence or absence of aortic atresia Groups align with surgical strategy and with mortality Three groups best differentiate patients
Interpreting Group 1 “Multiple levels of left-sided obstruction” AS/MS end of spectrum Some LV development Arch hypoplasia and coarctation common Most underwent an operation (SVP) Mortality – middle (27%) “Multiple levels of left-sided obstruction”
Interpreting Group 2 AA/MA or AA/MS Hypoplastic aorta Smallest LVs Apex-forming RV Almost all had an operation initially – SVP Highest mortality (41%) “True HLHS”
Interpreting Group 3 “Isolated AS” Most -no MV abnormality No AA ~80% moderate or severe AS Apex forming LV Antegrade flow in ascending aorta and arch Most had a catheter-based first procedure Lowest mortality (12%) “Isolated AS”
Clinical Relevance Conceptualization vs. Classification Categories not new Concepts now supported by data Brings management ideas into alignment with classification May relate to underlying etiology
Next steps Develop method to classify infants into these groups, incorporating clinical data
THANK YOU james.meza@sickkids.ca